Datasets:

CShorten

/

CORD-19-prototype

like 0

SEED Servers: High-Performance Access to the SEED Genomes, Annotations, and Metabolic Models

The remarkable advance in sequencing technology and the rising interest in medical and environmental microbiology, biotechnology, and synthetic biology resulted in a deluge of published microbial genomes. Yet, genome annotation, comparison, and modeling remain a major bottleneck to the translation of sequence information into biological knowledge, hence computational analysis tools are continuously being developed for rapid genome annotation and interpretation. Among the earliest, most comprehensive resources for prokaryotic genome analysis, the SEED project, initiated in 2003 as an integration of genomic data and analysis tools, now contains >5,000 complete genomes, a constantly updated set of curated annotations embodied in a large and growing collection of encoded subsystems, a derived set of protein families, and hundreds of genome-scale metabolic models. Until recently, however, maintaining current copies of the SEED code and data at remote locations has been a pressing issue. To allow high-performance remote access to the SEED database, we developed the SEED Servers (http://www.theseed.org/servers): four network-based servers intended to expose the data in the underlying relational database, support basic annotation services, offer programmatic access to the capabilities of the RAST annotation server, and provide access to a growing collection of metabolic models that support flux balance analysis. The SEED servers offer open access to regularly updated data, the ability to annotate prokaryotic genomes, the ability to create metabolic reconstructions and detailed models of metabolism, and access to hundreds of existing metabolic models. This work offers and supports a framework upon which other groups can build independent research efforts. Large integrations of genomic data represent one of the major intellectual resources driving research in biology, and programmatic access to the SEED data will provide significant utility to a broad collection of potential users.

The growth in the number of sequenced genomes has exceeded the most ambitious past estimations [1, 2] . Today, tens of thousands of genomes are being sequenced by public and private institutions around the globe [3, 4] , and whole-genome sequencing/re-sequencing has become a routine step of virtually any analysis of microbial behavior. In some fields, researchers are sequencing hundreds of genomes from distant taxa [5] to explore microbial diversity, while elsewhere, hundreds of closely related genomes are being sequenced to understand environmental adaptation, to support pangenome reconstruction [6, 7, 8] , and to build maps of microbial variomes [9] . Sequencing is also being applied to aid in solving urgent biological problems, including the development of strategies to combat emerging or reemerging biothreats, such as the severe acute respiratory syndrome (SARS) virus [10] , the 2009 H1N1 influenza virus [11, 12] , and the 2011 German Escherichia coli outbreak [13, 14] . With the vast speed and minimal costs of modern sequencing technology, new demands are being placed on the bioinformatics pipelines used to analyze sequence data [1] . Modern bioinformatics must be high throughput and high volume with a strong emphasis on comparative genomics, and the SEED family of resources for genome interpretation and analysis were specifically constructed to embody these characteristics [15] . The SEED resources are built upon the organization of biology into a modular set of subsystems, each describing a specific biological functionality (e.g., flagellar motility or histidine biosynthesis) [16] . Subsystems-based technologies were developed in the SEED with the view that the interpretation of one genome can be made more efficient and consistent if hundreds of genomes are simultaneously annotated in one subsystem at a time [16, 17] . The SEED Project, a multi-institutional effort coordinated by the Fellowship for Interpretation of Genomes (FIG) and Argonne National Laboratory, began with the ambitious goal of consistently annotating 1,000 genomes (http://theseed.org/wiki/ Annotating_1000_genomes); today, over 50,000 viral and prokaryotic genomes have been annotated by the subsystems technology. More recently, the SEED Project has been expanded to include the capacity to automatically generate genome-scale metabolic models based on SEED annotations of genome sequences, and the Model SEED system developed from this pipeline has now been applied to construct metabolic models for thousands of genomes [18, 19, 20] . The SEED family of resources now collectively serves as a repository for almost 5,000 distinct complete prokaryotic genomes, associated with approximately 30,000 annotations, 11,000 metabolic models, 178,000 protein families, 10,250 functional roles, and 1,060 subsystems. While a web interface is available for visualizing the data within the SEED (http://pubseed.theseed.org), this interface is inadequate to make full effective use of the massive quantities of data now embodied within the SEED. To truly expose the SEED data for efficient large-scale use by the global scientific community, we initially developed a Simple Object Access Protocol (SOAP) server to allow the needed programmatic access to the SEED data [21] ; however, the SOAP server could not scale up with the increased volume of data. To cope with the increasing demands placed on the SEED web services, as well as the increasing volumes of data included within the SEED resources, we developed a secondgeneration of web services to improve throughput, response time, and access to SEED functionality. Here, we describe in detail the four primary servers now available for remotely accessing SEED data and we describe the client-side distribution of server scripts that we provide as examples of how the servers may be applied to perform complex analyses of genome data. The Servers and the Services The four servers described here, collectively referred to as the SEED servers, currently support 181 methods that can be invoked to extract data and services (Fig. 1) . The server code resides at the location of the SEED data. Users download a distribution with their choice of runtime environment that they may use to write programs to access SEED data or perform a number of common bioinformatics tasks using a supplied set of preprogrammed scripts. Currently, Perl and Java integrations are supported. The following subsections describe the four servers in more detail. The Sapling Server offers access to the underlying integration of genomic data-including genomes, genes, proteins, annotations, subsystems, FIGfams, and co-occurrence data. On the server side, the Sapling Server accesses a database implemented by an entity-relationship data model (ERDB). The ERDB model is defined by a set of XML metadata describing the entities, relationships, and attributes in a form that can be used to generate queries as well as the documentation and an up-to-date database diagram may be viewed at http://servers.nmpdr.org/ figdisk/FIG/ErdbDocWidget.cgi?database = Sapling. Thus, the public description of the database always remains synchronized with the internal data structures-an important benefit in a database designed for public use. The Sapling Server is architected such that new features can be added quickly. New data tables may be added as updates to the XML metadata, which is processed by a special load program to build the initial database tables. The list of services offered is maintained on the server, so that client software does not need to be updated in order for users to access new features. A web application that converts general database queries to Perl code helps speed implementation of new functions. A database query is specified by naming the entities and relationships along a path through the ERDB diagram ( Fig. 2) along with a list of the data items to be returned and a filter clause that limits the results to the desired data objects (e.g., a particular genome or identifier). The Sapling Server allows direct queries against the database; however, a set of common data requests is implemented as direct server functions. Sapling Server functions typically accept multiple input values within a single call, allowing a client to minimize the number of requests that must be made to the server. Additional input parameters allow a client to modify the query, for example, to request that the output be in FASTA format or to ask for protein rather than DNA sequences. In a sample ids_to_sequences request (Fig. 3A) , the user specifies four identifiers, and the server returns them as a table (actually a Perl hash) with the associated DNA sequences attached. The Sapling Server currently supports 127 access functions. These functions are listed on a web page generated automatically from the latest version of the code, ensuring that the documentation remains up to date. A sample showing the web page description of ids_to_sequences is shown (Fig. 3B ). The Annotation Support Server supports two distinct capabilities relating to the annotation of genomes: de novo annotation of protein or DNA sequences, and aggregation of annotations into subsystems. The Annotation Support Server accepts either DNA or protein sequence as input and, depending on the user options, can either use existing gene calls or invoke standard gene callers (e.g., GLIMMER-3 for protein-encoding genes). The server also houses newly developed high-performance methods to assign function to protein sequences or regions of genomic DNA sequences, based on FIGfams [22] . Below is an example application using these methods that produces a relatively accurate annotation of most microbial genomes within a few minutes. To evaluate the technology, users are encouraged to simply submit a known prokaryotic genome to the server for annotation. Sequences can be submitted to the server in three ways: 1. Programs can directly access the services needed to call genes and assign functions to the proteins encoded within the genome. 2. If the protein-encoding genes have already been identified, the program can assign functions to these sequences. An example program is provided in the download library and is described at http://servers.theseed.org/sapling/server.cgi?pod=svr_assign_ using_figfams.pl. A program can take as input fragments of DNA (e.g., from a metagenomic sample) and use the services to detect pieces of protein-encoding genes. Again, an example program is provided in the download library and is described at http://servers.theseed. org/sapling/server.cgi?pod=svr_assign_to_dna_using_figfams.pl). The Annotation Support Server also provides the ability to take as input a set of functional roles (in the controlled vocabulary established by the subsystem collection) and to produce a detailed estimate of which subsystems are represented by those functional roles. That is, one can also use the server to develop a metabolic reconstruction based on the functional roles that have been assigned to the protein-encoding genes. The RAST (Rapid Annotations using Subsystems Technology) [23] service provides access to high-throughput, high-quality annotation of prokaryotic genomes, and routinely processes ,25-75 genomes a day, with peak throughputs that have exceeded 200 genomes per day. This service is accessible for individual genome submission via a graphical web interface, and the next generation SEED servers provide new access for programmatic batch submission of genomes via the RAST submission and retrieval server. This server supports programmatic submission of genomes to the RAST Server, retrieval of job status, and retrieval of the final set of annotations. These scripts and the underlying application programming interface (API) enable users to submit genomes to the RAST Server, test the status of submitted jobs, and retrieve the output (i.e., annotated genomes). Three types of input are supported: 1. A FASTA file of contigs that make up the genome to be annotated 2. A file of GenBank-formatted entries (with the option to retain the gene calls as given in the uploaded files) 3. An Entrez ID or a Genome Project ID. In the case of the first type of input (FASTA file), the RAST Server will perform de novo annotation, i.e., will start by calling RNA-and protein-coding genes, and then will proceed to functional annotation, subsystems assignment, and subsequent metabolic reconstruction. In the second case (GenBank-formatted files), gene calling is optional, since GenBank-formatted files already include the coordinates for open-reading frames and RNA genes. In the last case (Entrez ID or Genome Project ID), the tools provided within the RAST Submission/Retrieval Server will query NCBI using the query ID(s) and will retrieve the sequence (whether that sequence consists of one or a set of contigs that make up the genome project); that sequence then becomes the input to the RAST Server. The Metabolic Modeling and Flux Balance Analysis (FBA) Server provides programmatic remote access to the Model SEED biochemistry and genome-scale metabolic model database, as well as some model analysis algorithms. The Model SEED biochemistry database integrates into a single, nonredundant set all the reactions and compounds found in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database [24, 25, 26] , together with additional curated reactions and compounds [18] , and a continuously growing number of published genome-scale metabolic models. Currently this database consists of 16,279 compounds and 13,272 reactions. For compounds, the database also includes database IDs from KEGG compounds and SEED models, names/ synonyms, molecular masses, molecular formulas, molecular charge, and estimated Gibbs free energy of formation [27] . For reactions, the database includes database IDs from KEGG reactions and SEED models, names/synonyms, stoichiometry, EC numbers, pathways, and estimated Gibbs free energy change of reactions [27] . Compound charge, formula, formation energies and reaction stoichiometry are all calculated for aqueous conditions at neutral pH. All API functions used to access the Metabolic Modeling and FBA Server capabilities are documented in detail at http://servers.theseed.org/. The SEED database also contains a large number of genomescale metabolic models, including 14 published models [28] and approximately 11,000 models generated from the annotated genomes stored in the SEED [29] . The Metabolic Modeling and FBA Server also provides the user with an API to remotely obtain a list of the models in the SEED and to download data on the compounds and reactions in each SEED model. The server returns the following data for each reaction in a specified model: (i) all data from the SEED biochemistry database, (ii) a list of the genes associated with each reaction in the model in a format that captures how the protein products encoded by the genes function to catalyze the reaction (as either independent enzymes or multienzyme complexes), and (iii) a list of compartments in the model where the reaction takes place and the directionality/ reversibility of the reaction in each compartment. For the model compounds, the server returns the data from the SEED Figure 2 . Entities and relationships in the SEED. The entities (boxes) are connected to each other by a series of relationships (diamonds) that describe how the two entities relate. To move from one entity (e.g., ''Identifier'') to another (e.g., ''DNA Sequence''), the series of connections shown by the gray arrow is made. This way, any entity can be connected, either directly or indirectly, to any other entity. doi:10.1371/journal.pone.0048053.g002 biochemistry database. As with the biochemistry data, all the model data in the server are accessible either via Perl programs or the API. The Metabolic Modeling and FBA Server also enables users to run various FBA studies on any of the genome-scale metabolic models stored in the SEED database. These studies can be performed while simulating any of 525 distinct media conditions currently encoded in the SEED database (which includes all BiologH [30] media conditions and a variety of complex media formulations). Both the Perl program and the API enable users to obtain a list of the media conditions currently stored in the SEED and details on the compounds included in each formulation. Once a model and media condition have been selected for simulation, the server provides an interface for running three types of FBA simulation: (i) simple growth simulation to predict maximum growth rate of the organism in the selected media, (ii) flux variability analysis (FVA) [31] to classify the reactions and compounds in the model according to their behavior during growth in the selected media, and (iii) single gene knockout analysis to predict the genes essential for growth in the selected media. The simple growth simulation returns the maximum predicted growth rate of the model given the input parameters, the predicted flux through the model reactions during maximum growth, and the predicted uptake and production of nutrients from and to the environment during maximum growth. The FVA simulation returns the predicted class of every reaction and compound in the model during growth given the input parameters. Reactions in the model are classified as forward essential or reverse essential if they are required for growth to occur, with forward and reverse referring to the direction in which the reactions must proceed. Reactions that are not essential for growth, but still active, are classified as forward variable, reverse variable, and variable, with forward and reverse indicating when reactions proceed only in a single direction. Reactions are classified as blocked if they cannot carry flux under the conditions specified by the user. Metabolites in the model are classified as essential nutrients or essential products if their uptake or secretion is required for growth in the input conditions, and they are classified as transported if they can be taken up or secreted but are not essential for growth. In addition to classifying the reactions and compounds in the model, the FVA simulation returns the maximum and minimum values for the flux through each reactions and the uptake/secretion of each metabolite. The single gene knockout analysis rapidly simulates the individual knockout of every gene represented in the model during growth in the input conditions. Based on these simulations, the analysis produces a list of the predicted essential genes and the predicted nonessential genes in the model. Both the Perl program and API allow the user to run any of the three simulation types from the command line. All three simulation types accept the same user input: the name of the model to be run, the name of the media formulation that growth should be simulated in, a list of genes in the model that should be knocked out during the simulation, and a list of the reactions in the model that should be knocked out during the simulation. See http://servers.theseed.org/for detailed documentation on all Metabolic Modeling and FBA Server functions. The SEED servers can have limitless applications. Below, we demonstrate various applications of SEED servers through a small set of potential applications and coding examples. Dealing with IDs of genes or the proteins they encode is often nontrivial. In the SEED database, we use IDs that specify proteinencoding genes in a rapidly growing set of genomes, and we support correspondences between these IDs and those used by other annotation efforts. The SEED has two notions of equivalence: (i) two IDs that represent either protein-encoding genes or protein sequences are said to be sequence equivalent if the protein sequences are identical and (ii) two IDs that represent either exactly the same protein-encoding gene or the precise protein encoded by the gene (that is, ''the protein sequence of gene X in genome Y'') are said to be precisely equivalent. Unfortunately, in the presence of multiple versions of thousands of genomes, perfect maintenance of the ''precisely equivalent'' correspondence is virtually impossible. Our first example script takes a command-line argument containing a single ID and produces a table for all assertions of functions for sequence equivalent IDs. Each ID in the input is associated with the name of the genome containing it, the function for that ID, the source of the functional assignment assertion, and an indication of whether the source of the assertion provided a confidence for their estimate. The code is available at http://servers.theseed.org/sapling/server. cgi?code = server_paper_example1.pl (also provided as Supporting Information S1). Given a set of functional roles, one often wishes to understand which subsystems can be inferred from the set. The following example script reads as input a set of functional roles and constructs a table of subsystems that can be identified, along with their variation codes. The data displayed in this simple example could form the start of a research project to gather the functional roles not connected to subsystems to determine whether they were not connected because a small set of functional roles were not present in the input, and to seek candidates for such "missing functional roles." The ability to easily map functional roles into subsystems will improve as the SEED annotation effort improves its collection of encoded subsystems [32] . The code for this example is shown at http://servers.theseed.org/sapling/server. cgi?code=server_paper_example2.pl. (also provided as Supporting Information S1). The SEED provides the ability to graphically display the chromosomal regions around a set of genes (normally from distinct genomes); for example, see http://seed-viewer.theseed.org/ seedviewer.cgi?page=Annotation&feature=fig|83333.1.peg.4. The SEED also offers an alternative for creating custom interfaces, moreover, one that does not require the user to know appropriate SEED IDs. This approach exploits the conversion capabilities of the SEED for creating a program to accept arbitrary protein IDs. It also exploits the ability of SEED to map functional roles into subsystems as described in the preceding example. The result is a tool that enables the user to take a SEED ID and a region size, and extract the genes that are found within a region centered on the designated gene. The code for this example is shown at. http://servers.theseed.org/ sapling/server.cgi?code=server_paper_example3.pl (also provided as Supporting Information S1). A great deal has been learned from studying genes that tend to occur close to one another in diverse genomes [33, 34, 35, 36, 37, 38] . In particular, the co-occurrence of hypothetical and non-hypothetical proteins can be exploited to suggest the function of the former based on the function of the latter. The following program at http://servers.theseed.org/sapling/ server.cgi?code=server_paper_example4.pl illustrates the potential for constructing custom tools by going through all of the protein-encoding genes in all of the complete prokaryotic genomes maintained within the SEED looking for ''hypothetical proteins'' that tend to co-occur with genes encoding functions that can be connected to subsystems. The program constructs a table showing the following: This table can therefore be used to suggest functions for hypothetical proteins that could be tested experimentally. A copy of the code is provided (Supporting Information S1). The SEED can be used to assign functions to a file of protein sequences. The code for this example is at http://servers.theseed. org/sapling/server.cgi?code=server_paper_example6.pl (also provided as Supporting Information S1). This program reads a FASTA file of protein sequences and attempts to assign function to those sequences using a K-merbased algorithm. When a function is proposed, the program will produce a ''score'' (the number of distinct K-mers that were matched) and an estimate of phylogenetic neighborhood-a representative genome that is ''phylogenetically close'' to the genome containing the protein, if an estimate can reasonably be given. A similar approach has been adopted for rapid, real-time analysis of metagenomics samples, which might elsewhere take days or months for BLAST-based analysis. This real-time metagenomics analysis (URL: http://edwards.sdsu.edu/rtmg) can be performed on computers or cellular phones [39] . Here, we demonstrate how to run a variety of FBA algorithms on the SEED model of E. coli and how to print all data from the E. coli model and the results of the FBA into an output table. For the code, see http://servers.theseed.org/sapling/server. cgi?code=server_paper_example7.pl (also provided as Supporting Information S1). The program starts by obtaining a list of all compounds and reactions in the SEED E. coli model (Seed83333.1) using the ''get_compound_id_list'' and ''get_reaction_id_list'' functions, respectively. The program then uses these lists to obtain detailed data on all the E. coli compounds and reactions (using the ''get_compound_-data'' and ''get_reaction_data'' functions, respectively). These data are stored in two tables: one for compounds and one for reactions. Next the ''classify_model_entities'' function is used to run a FVA on the SEED E. coli model. In this particular FVA, the reactions and compounds in the E. coli model are classified while simulating growth in LB media (called ArgonneLBMedia in the SEED model). At this point, the data returned by the ''classify_model_entities'' function is added onto the compound and reaction tables prepared previously. In the next step, the code uses the ''simulate_model_growth'' function to run a standard FBA on the SEED E. coli model, maximizing the model growth rate in simulated glucose minimal media (called Carbon-D-Glucose in the Model SEED). The data returned by this function are also added to the reaction and compound tables. In the final call to the server, the program uses the ''simulate_all_single_gene_knockout'' function to simulate the single knockout of all E. coli genes, and the results of this study are stored in a gene table. The remainder of the program handles the printing of the compound, reaction, and gene tables to the files CompoundTbl.txt, ReactionTbl.txt, and GeneTbl.txt, respectively. The Gene Ontology (GO) project aims to unify biology by providing a controlled vocabulary of terms for all genes and gene products [40] , but has long had a focus on eukaryotes with less emphasis on prokaryotes and their viruses. On the other hand, the SEED database contains high quality annotations for hundreds of microbial and viral genomes, using subsystems-based controlled vocabulary. Mapping SEED functional roles to GO annotations for a given set of genes or gene products can be achieved via SEED servers. A workflow, detailed elsewhere (Short URL: http://bit.ly/server_paper_example8 ), uses two SEED serversbased programs to update SEED to GO comparisons through the use of UniProt [41, 42] protein identifiers. The workflow consists of the following steps: The SEED project [15] focuses on developing technology to support rapid, high-volume, accurate annotation of genomes, and has so far achieved four advances of central importance: 1. The subsystems strategy, adopted as the guiding principle of the effort [16] , centers on leveraging expert annotations to define a small set of functional roles in all genomes rather than all the functional roles in a small number of genomes. 2. The subsystem effort provided a convenient framework for the curation of a set of protein families that became known as FIGfams [22, 23] , intended to contain only isofunctional homologsthat is, each family was intended to contain only homologous proteins playing the same functional role. When errors in FiGfams are detected, the underlying subsystems are updated and then the FIGfams are regenerated to correct those errors. The rapid evolution of the FIGfam collection has made possible a number of the services described in this article. 3. Using subsystems and FIGfams as the underlying technology, the RAST server was developed and made available in 2007 [23] . Thousands of viral and prokaryotic genomes have been annotated with the RAST system, and hundreds more are being annotated each week. 4. The high-quality annotations generated by RAST together with the ability to manually modify those annotations by human experts allowed for the automatic generation of genome-scale metabolic models in the Model SEED [18, 19, 20] . As indicated above, the improvement in sequencing speed and efficiency led to a rapid accumulation of genome sequences and necessitated more efficient methods for large-scale access to genomes, annotations, subsystems, and genome-scale models within the SEED database. A SOAP server was first developed to allow programmatic access to the SEED data [21] , but several performance issues prevented the service from scaling with the volume of data contained within the SEED database. The server abstraction layer required the loading of numerous large modules on each invocation of server functionality, resulting in a noticeable delay in response to each server request. The encapsulation of the results in SOAP XML conferred significant overhead on the data being transferred. Finally, each operation of the SOAP server was atomic, accepting a single argument and returning a single datum. Trivial requests such as retrieving all the functions for all of the proteins in a genome took unacceptably long to complete, requiring a separate call for each protein and instantiating many threads on the server. The four SEED servers described here provide programmatic access to the SEED data and methods. They expose the current data in a form that is conveniently accessed computationally. The installation and maintenance of the client-side software require minimal effort. We have constructed the underlying methods to support relatively large-grained data transfers, allowing the construction of relatively efficient programs. In comparison to the SOAP server [21] , the new web services provide access to larger amounts of data in less time, and they have been engineered to respond to server requests with little or no server-side delay. Furthermore, the new web services provide a more efficient and flexible computing approach because they are designed to process batches of requests at a time, streaming the responses as they complete. These services provide access to the integrated genomic data, subsystems, FIGfams, co-occurrence data, annotation services, RAST annotation submission and job retrieval (thereby offering access to our continuing improvements in microbial annotation), and metabolic modeling. All client modules, code examples and documentation are available online at http:// servers.theseed.org, and we are continually expanding these services and improving the underlying documentation. In conclusion, the SEED integration of genomic data now contains over 5,000 complete or nearly complete genomes, a constantly updated set of curated annotations embodied in a large and growing collection of encoded subsystems, and a derived set of protein families. The client-server code discussed in this article gives users easy programmatic access to the data in the SEED. It has already been successfully been used for building other applications, such as finding prophages in microbial genomes [43] , desktop-based RAST genome annotation and comparison (manuscript in preparation), and real-time annotation of metagenomes (RTMG, URL: http://edwards.sdsu.edu/rtmg [39] ). We encourage researchers to use this software to retrieve and analyze data from the SEED database and we believe that the underlying implementation of these new servers is efficient enough to address the needs of most users and we will continue providing occasional stand-alone versions of the SEED to users who need more performance or privacy. The latest documentation and downloads are available at the following web addresses: http://servers.theseed.org and http:// blog.theseed.org/servers/ N Project name: SEED Servers N Project home page: http://www.theseed.org/ N Operating system(s): Mac OS, Linux N Programming language: Perl, JAVA N License: SEED Toolkit Public License N Any restrictions to use by non-academics: no limitations The SEED servers project is documented and can be downloaded from the servers' web site, http://servers.theseed.org. The Perl distribution contains the following: Client Packages. 1. The Sapling server -SAPserver.pm 2. The MODEL server -MODELserver.pm 3. The Annotation Support Server -ANNOserver.pm 4. The RAST server -RASTserver.pm Programming using the servers. The SEED servers provide all necessary network operations in a client package that can be used to access the server functions. One uses these like any other Perl package. For instance, to find all genomes in the SEED, one does the following: #!/usr/bin/perl w use strict; use SAPserver; my $sapObject = SAPserver-.new(); my $genomes = $sapObject-.all_genomes(); foreach my $g (sort { $genomes-.{$a} cmp $genomes-.{$b} } keys(%$genomes)) { print "$g\t$genomes-.{$g}\n"; } The function call $sapObject-.all_genomes() marshals the correct server-side function call and arguments into a network package, transmits that package to the server, waits for and retrieves the answer, processes any returned error codes, decodes the return package into a Perl data structure, and returns the result. All function calls in all the client packages perform these basic services. The Java Distribution Contains the Following Client Packages. The org.theseed.servers.serverConnections package handles connecting to the server, transmitting and receiving the data, and converting data structures from the server into Java data structures. The classes in org.theseed.servers.servers packages handle connecting to each of the servers and making the appropriate calls. Programming using the servers. We recommend that the code be accessed in eclipse (http://www.eclipse.org/), netbeans (http://www.netbeans.org/), or a similar graphical integrated development environment (IDE). These are used like any other class. For instance, to find all genomes in the SEED, one does the following import java.util.HashMap; import org.theseed.servers.SAPserver; public class AllGenomes { public static void main(String [35] args) { SAPserver sapling = new SAPserver(); HashMap,String, String.genomes = sapling. allGenomes(); for (String id : genomes.keySet()) System.out. println(id + "\t"+ genomes.get(id)); } } Future directions include expanding the applications and releasing packages for use by other programming languages such as Python. Supporting Information S1 Examples of programming using the SEED servers (coded in Perl). Eight application examples are provided, seven of which (1-4 and 6-8) are described in detail in the text. (PDF)

Discovery and Targeted LC-MS/MS of Purified Polerovirus Reveals Differences in the Virus-Host Interactome Associated with Altered Aphid Transmission

Circulative transmission of viruses in the Luteoviridae, such as cereal yellow dwarf virus (CYDV), requires a series of precisely orchestrated interactions between virus, plant, and aphid proteins. Natural selection has favored these viruses to be retained in the phloem to facilitate acquisition and transmission by aphids. We show that treatment of infected oat tissue homogenate with sodium sulfite reduces transmission of the purified virus by aphids. Transmission electron microscopy data indicated no gross change in virion morphology due to treatments. However, treated virions were not acquired by aphids through the hindgut epithelial cells and were not transmitted when injected directly into the hemocoel. Analysis of virus preparations using nanoflow liquid chromatography coupled to tandem mass spectrometry revealed a number of host plant proteins co-purifying with viruses, some of which were lost following sodium sulfite treatment. Using targeted mass spectrometry, we show data suggesting that several of the virus-associated host plant proteins accumulated to higher levels in aphids that were fed on CYDV-infected plants compared to healthy plants. We propose two hypotheses to explain these observations, and these are not mutually exclusive: (a) that sodium sulfite treatment disrupts critical virion-host protein interactions required for aphid transmission, or (b) that host infection with CYDV modulates phloem protein expression in a way that is favorable for virus uptake by aphids. Importantly, the genes coding for the plant proteins associated with virus may be examined as targets in breeding cereal crops for new modes of virus resistance that disrupt phloem-virus or aphid-virus interactions.

Virus species in the Luteoviridae, including Cereal yellow dwarf virus (CYDV)-RPV, referred to herein as luteovirids, are each exclusively transmitted by one or a few species of aphids in a persistent, circulative, non-propagative manner. Aphids can transmit acquired virus for days or weeks and even after molting [1, 2, 3] . Virus is acquired when the virus moves from the aphid hindgut (HG) lumen to the hemolymph. Once in the hemolymph, the virus passes into the accessory salivary gland (ASG) and is released into the salivary duct. Aphids transmit the virus into phloem cells during salivation and feeding. Through specific interactions with putative cellular receptors [4, 5, 6] , luteovirids actively cross the membranes of the aphid HG and ASG. The HG and the ASG are the major barriers to the successful aphid acquisition and transmission of luteovirids. The movement of luteovirids across these barriers is under genetic control by an unlinked set of aphid genes [7] that are additive in effect [8] . These genes encode for proteins that are differentially expressed in aphids with varying virus transmission efficiency [9, 10] . The virus encodes two structural proteins that make up the capsid. These proteins orchestrate systemic movement within host plants and transcytosis within aphids [11, 12, 13, 14] . The 22-24 kDa coat protein (CP) expressed from the viral open reading frame (ORF) 3 is the major capsid protein. Occasionally, readthrough translation of the CP stop codon results in the translational fusion of ORF3 with ORF5, resulting in the minor protein species component of the capsid called the readthrough protein (RTP). Virions can be assembled from the CP alone, but the RTP is required for luteovirid movement across the ASG as well as efficient systemic infection [13, 14, 15, 16, 17] . These activities in aphids and plants are regulated by virion topological features, specifically interactions between CP monomers and welldefined interfacial regions within the RTP [18] . Although a number of aphid proteins that bind luteovirids have been identified, the molecular details of virus movement in aphids are not well understood. Due to the paucity of molecular tools for the study of aphids, most virus-binding proteins have been identified using proteomics approaches. In Sitobion avenae, four 2-D gel spots were differentially expressed between parthenogenetically-reproducing F1 clones that differed in their vectoring capacity for two isolates of the species Barley yellow dwarf virus (BYDV)-PAV [19] . Two proteins from head tissues of S. avenae were identified as potential receptors for BYDV-MAV based on virus overlay and 2-D immunoblot assays [5] . Other studies progressed beyond 2-D gel spot patterns to use mass spectrometry (MS) to investigate the identity of aphid proteins associated with virus transmission [9, 10, 20] in effort to link the protein function to vector biology. Membrane-associated actin, rack-1, and GAPDH-3 were identified to interact with a wild-type isolate of Beet western yellows virus (BWYV). Actin and GAPDH-3, but not rack-1, also interacted with two BWYV RTP mutants displaying a reduction in aphid transmission ability [20] indicating either (a) their assay for virus binding does not reflect the in vivo functions of these interactions in aphids, or (b), that these RTP mutants still retain the interaction interface for binding with these aphid proteins. Cyclophilin and a luciferase homologue were identified from the greenbug aphid Schizaphis graminum as binding to CYDV-RPV. These proteins were also differentially expressed among different S. graminum F2 genotypes with different CYDV-RPV vectoring capacity [9, 10] . Intriguingly, all studies investigating proteomic differences of aphid F1 [19] or F2 progeny genotypes [9, 10, 19] reported the differences between vectors and non-vectors were not related to gross changes in protein expression levels. Instead, vectoring capacity was associated with a small shift in the isoelectric point of only a small number of proteins, indicating that the control of luteovirid transmission in multiple aphid species was via the expression of allelic variants of proteins that either differentially bind to the virion or participate in the various steps of the virus transmission pathways. In plants, luteovirids move from cell-to-cell and long distance as virions. However, in plants luteovirids must also replicate, assemble into virions, thwart the plant immune response and exhibit phloem-tissue tropism for host-to-host transmission by aphids, processes which likely involve the recruitment of a wide range of host factors to accomplish [21] . Interestingly, Bencharki and colleagues showed that the addition of soluble proteins to the aphid diet, including phloem-specific proteins that can interact with virions, can enhance transmission efficiency of the polerovirus, cucurbit aphid borne yellows virus [22] . Thus, we hypothesize that plant host proteins are also involved in aphid transmission. Serendipitously, we discovered that common additives to the plant tissue homogenization buffer during virus purification do not appear to alter the virus morphology or protein structure, but do render the virus non-transmissible by aphids. By comparing the host plant proteomes associated with transmissible and non-transmissible virions, proteins with potential involvement in transmission were identified. Moreover, our data indicate an increased accumulation of several virus-associated host proteins within aphids that have fed on infected plants, supporting a role for these host proteins in the circulative transmission process. Sodium sulfite (Na 2 SO 3 ) and ethylenediaminetetraacetic acid (EDTA) are commonly added to infected plant homogenate during virus purification to minimize the effects of polyphenolic compounds and virion aggregation, respectively [23, 24] . When sodium sulfite and EDTA were added to the tissue homogenization buffer, and the subsequent purified CYDV-RPV virions were fed to aphids through ParafilmH membranes, the virus was not transmitted in two independent experiments (Table 1 ). In contrast, CYDV-RPV purified without these additives was highly transmissible by aphids (.95%) in six independent experiments (Table 1) . To determine whether treated virions were unstable and being degraded during membrane feeding by aphids, treated and non-treated virions were collected from the diet following the acquisition access period (AAP) and analyzed by transmission electron microscopy (TEM). Importantly, treated virions remained intact in the diet following the AAP (Fig. 1 ). Treated and untreated virions were morphologically indistinguishable at this point ( Fig. 1) . Broken capsids and capsid swelling were not observed, indicating virions were structurally stable for transmission (Fig. 1) . To determine whether sodium sulfite or EDTA was responsible for the reduction in transmission efficiency, virions were purified with only EDTA or sodium sulfite as additives and transmissibility was determined. When only EDTA was added, the purified virus was transmissible at levels similar to virus prepared with neither of the additives, 92% compared to 95%, respectively (Table 1 ). In contrast, the sodium sulfite treatment reduced transmissibility to ,5%. The negative impact of sodium sulfite on virus transmission by aphids was independent of virus concentration since the results were consistent across four independent experiments where the Transmission of CYDV-RPV is dependent upon the virus circulating through the aphid. The interaction between the virions from each purification treatment and the aphid was evaluated, particularly to identify possible barriers to transmission. To determine if aphids ingested and acquired virus, aphids were collected following a 24-h AAP on membranes containing each type of purified virus: buffer only, EDTA-treated, or sodium-sulfite treated. Total RNA was extracted from half of the aphids from each treatment and the remaining aphids were transferred to healthy oat seedlings for a 3-day inoculation access period (IAP). Total RNA was also extracted from those aphids immediately following the 3-day IAP. Viral RNA (vRNA) was detected within aphids after a 24-h AAP for each virus treatment preparation (Fig. 2) using RT-PCR and CP-specific primers. After 24-h, vRNA can be detected in all three groups of aphids. At 24-h, virus could be in the gut lumen, and/or in the HG cells and possibly in the hemolymph. This indicates aphids were ingesting, and possibly acquiring, the virus under all conditions. For aphids collected after a 3-day IAP, vRNA was detected for the buffer only treatment (Fig. 2) indicating uptake of the virus into the hemolymph, which was supported by high level of transmission (92% transmission, Table 2 ). No vRNA was detected after 3 day IAP (Fig. 2) for EDTA-or sodium sulfite-treated virions in the aphids. EDTAtreated virions were still transmitted by aphids to oat seedlings (92% transmission efficiency, Table 2 ), suggesting there was virus uptake and persistence occurred within the aphid, albeit at levels not sensitive enough to detect using RT-PCR. However, the sodium sulfite treatment reduced the rate of virus transmission into oat seedlings (Tables 1 and 2) . Taken together with the RT-PCR results, these data suggest that sodium sulfite-treated virions may not be acquired. If the virus is not acquired, any virus remaining in the lumen after 24-h may flow out of the aphid in the honeydew [25] . To determine if either additive affected the ability of virus to move into gut cells, we used TEM to visualize virion attachment and virus penetration across the gut membrane (Fig. 3) . Aphids were fed on sodium sulfite-or EDTA-treated as well as nontreated virions for a 48-hr AAP. Non-treated (Fig. 3A, B) and EDTA-treated virions (Fig. 3C , D) were observed to attach to apical plasmalemma lining the HG lumen, in coated pits invaginating and budding into the cytoplasm, as well as retained Figure 1 . Negative stained grids, coated with CYDV-RPV coat protein antibody, of purified virus from each virus preparation after purification and virus recovered from membranes fed on by R. padi for a 24 h AAP. Virion morphology was similar within each group and a representative picture for each is shown. Transmissible virions after purification (A) and after membrane feeding (B) look morphologically similar. Non-transmissible virions after purification (C) and after membrane feeding (D) look morphologically similar to each other and are indistinguishable from transmissible virions in shape and size. Scale bars = 100 nm. doi:10.1371/journal.pone.0048177.g001 in vesicles dispersed throughout the cytoplasm. In contrast, sodium sulfite-treated virions were rarely observed to interact with the apical plasmalemma nor were they observed within the cytoplasm of the HG. However, virions were observed in the HG lumen ( Fig. 3E , F). The patterns of virus transmission and virus distribution within the aphid for each purified virus (non-treated, EDTA-or sodium sulfite-treated) were consistent in five aphids fed on the different virus preparations ( Table 2) . After virus egress from the HG, virions must remain stable in the hemolymph and pass through the ASG for transmission to occur. To determine whether the treatments had an effect on these latter steps in the circulative transmission pathway, treated and non-treated virions were microinjected into aphid hemolymph to bypass the HG barrier. Transmission efficiency was reduced when aphids were injected with sodium sulfite-treated or sodium sulfite plus EDTA-treated virions, 8% and 0%, respectively, relative to aphids microinjected with EDTA-treated or untreated virions, 78% and 91%, respectively (Table 3) . These data were consistent over multiple independent experiments, indicating the effect was reproducible and not dependent on the concentration of virus. Initial results from the membrane-feeding experiments indicated that the EDTA and sodium sulfite could be acting additively or synergistically to block CYDV-RPV transmission since no transmission was observed. An identical trend was observed in the microinjection experiments (Table 3) . We performed a loglinear analysis of a 3-way contingency table using sodium sulfite, EDTA, and transmissibility as the three variables. We were unable to conclusively show any interaction effects of EDTA and sodium sulfite on aphid transmission of CYDV-RPV with the number of replicates performed for either the feeding or microinjection assays. To test the hypothesis that sodium sulfite treatment changes the host-virus interactome, LC-MS/MS was used to analyze tryptic digests of non-transmissible and transmissible virion preparations purified from infected oat plants. Although the genome of oat is not available, we were able to use predicted proteins from available genomes of related cereals for homology-based protein identification of the plant proteins found in the virus preparations. Twenty proteins were identified in the transmissible, but not the non-transmissible virus preparations (Table 4 ). Sixteen proteins ( Table 4 ) are candidate proteins that may interact with transmissible virus directly or in complex. These include three chloroplast proteins (326533372, 20302473, and 2565305), seven mitochondrial proteins (115472339, 115474041, 115471693, 326500100, 115477529, 115448577, and 357139868), two isoforms of remorin (357164942 and 115456099), one predicted cytoplasmic protein (115449199) and three proteins with predicted functions in the nucleus (357144283, 357121487, and 226531758). Six host proteins were found in both transmissible and non-transmissible virus preparations. These include a 33-kD secretory protein, thaumatin-like protein 5 (TLP-5), ATP synthase CF1 beta subunit, cellulose synthase, triosephosphate isomerase (TIM), and a cysteine-rich repeat secretory protein (Table 4 ). Two proteins, prophobilinogen synthase and adenosylhomocysteinase, were identified exclusively from preparations of sodium-sulfite treated, non-transmissible virions. Proteins homologous to most of these proteins are involved in plant defense and have been previously identified in plant phloem sap proteomes [26, 27, 28, 29] . Among the potential virus genome-derived protein products, peptides from only the two virus structural proteins (CP and RTP) were identified (not shown). To rule out co-sedimentation of plant proteins with purified virus based on density alone, we used LC-MS/MS to thoroughly characterize sucrose gradient fractions of healthy plant homogenates (Table S1 ). Healthy plants were grown in the greenhouse alongside infected plants and the tissue was used in the virus purification protocol. Fractions from the gradient that would normally contain virus particles when infected plant tissue is used in the purification were subjected to trypsin digestion and LC-MS/MS to identify any plant proteins that could co-sediment with virus particles based on density and not a physical association. Four proteins ( Table 4 ) that are likely not associated with virions, Figure 2 . RT-PCR of total RNA extracted from R. padi aphids that fed on different virus preparations was performed to detect RPV using RPV coat protein primers, amplifying a 614 bp product. Aphids were collected after initially feeding for a 24 h AAP (A) and after 3 d IAP (B). Three independent replicates using RNA collected from small pools of aphids are shown for each treatment. doi:10.1371/journal.pone.0048177.g002 Table 2 . Summary of R. padi transmission and virion detection following membrane feeding of different cereal yellow dwarf virus-RPV preparations used for RT-PCR and TEM evaluations. including homologues of three ribosomal proteins and fructokinase, were also identified in the analysis of the gradient fractions from healthy plants. Other abundant proteins detected in the sucrose gradients from healthy tissue included Rubisco, pyrophosphate-dependent phosphofructo-1-kinase, 6-phosphofructokinase, ribosomal proteins, and histone proteins (Table S1 ). Although many of these proteins have been detected in phloem sap in other plant species [26] and in the gradients of purified virus, the presence of these proteins is probably unrelated to the transmissibility of virions since they were detected in the fractions from healthy tissue. Importantly, no peptides were identified as belonging to CYDV-RPV indicating that the plants used for the healthy controls were not infected. Analysis of purified sodium sulfite and EDTA-treated and nontreated CYDV-RPV using 1-D gel electrophoresis often showed distinct protein profiles (Fig. S1A ). These differences were initially thought to be due to degradation products of the RTP; since the C-terminal half of the RTP is truncated during purification [30, 31] . For both transmissible and non-transmissible virus, we analyzed gel bands corresponding to viral and host proteins. MS analysis of peptides recovered from an in-gel trypsin digestion of band containing the truncated RTP from purified virions produced nearly identical tryptic fragments representing both the CP and RTP from both treated (Fig. S1B ) and untreated virions (Fig. S1C ). These data indicate minimal or no differences in the virus protein composition of the virion particles. Consistent with previous reports of C-terminal truncation of the RTP during purification, no peptides were discovered in the C-terminal portion of the RTP from either type of purified virus. The highly abundant proteins (Table S1 ) were also found in analysis of the gel bands from Fig. S1 (data not shown) and enabled us to attribute some of the variability in the gel bands to differences in plant protein complexes that co-sediment based on their density alone. The LC-MS/MS data suggest that treatment of plant tissue homogenate with sodium sulfite disrupts the host-virus interactome that is required for virus transmission by aphids. Here, the host-virus interactome is defined as the complement of host plant proteins binding directly to or in complex with the purified virus. The host proteins may play a direct role in virus uptake, maintaining virion stability, or help to provide enzymatic or cofactor activity during initial steps of entry into aphid cells. We hypothesized that if host plant proteins were involved in transmission, either via direct interactions with virions or by another indirect mechanism, we should be able to detect evidence of these proteins accumulating to higher levels in aphids that have fed on infected plants compared to aphids that have fed on healthy plants. A targeted proteomics approach called selective reaction monitoring (SRM) mass spectrometry [32, 33, 34, 35] was used to detect peptides derived from virus-interacting plant proteins in aphid protein homogenates. SRM detects and quantifies selected tryptic peptides within a total protein extract that are unique to the proteins of interest by monitoring specific intact tryptic peptides ions and their collisionally-induced dissociation (CID) fragments (ions derived from fragments of the tryptic peptides) based on in silico predictions of their mass:charge ratios. In contrast to the discovery approach that enabled us to ask ''What host proteins are in the different virus preparations?'' (Table 4) , SRM enables us to ask the hypothesis-driven question, ''Are virus-interacting plant proteins A, B, and C, present in our aphid protein sample?''. The candidate host proteins considered for SRM studies included all of the proteins listed in Table 4 that were found to be exclusively associated with transmissible virus. Proteins found in both transmissible and non-transmissible virus were also considered since they may be found in different amounts in the aphid. All of these plant proteins were checked for similarity against all available aphid and aphid bacterial endosymbiont protein sequences. Peptides from proteins that were identical to aphid or endosymbiont proteins at the amino acid level were not considered for further analysis as only peptides unique to the plant proteins could be informative for the specific measurements of the plant proteins in aphids. Those that remained included cellulose synthase, thaumatin-like protein 5, 33-kD secretory protein, pyruvate dehydrogenase E1, pyruvate dehydrogenase E2, remorin, cysteine-rich repeat secretory protein 55-like, structural maintenance of chromosomes protein 3, and predicted ribosomal protein. We also included b-D-glucosidase which was abundant in healthy controls as well as the purified virus preparations. To perform relative quantification of these proteins, a method was created in Skyline [32] that identified peptides specific to each of the candidate virus-interacting host proteins and exported to a TSQ Vantage mass spectrometer operating in SRM mode. Total protein extracted from the efficient RPV-vector aphid species Rhopalosiphum padi (all developmental stages) fed on healthy or infected plants for 21 days was digested with trypsin and analyzed by SRM. Peptides from five of the plant proteins listed in Table 4 Figure 3. Effect of sodium sulfite and EDTA on CYDV-RPV virion (arrowhead) attachment to apical plasmalemma and endocytosis into the HG cells of R. padi following membrane acquisition. (A and B) no sulfite or EDTA (buffer only) virus is internalized into cells of R. padi HG and can be found in tubular vesicles; (C and D) EDTA only treatment shows no effect of EDTA on acquisition of virions into the aphid HG which is consistent with the transmission data presented in Tables 1 & 2 ; (E-F) sodium sulfite treatment prevents attachment and acquisition of virus into cells of the HG. APL, apical plasmalemma; HG, HG; L, lumen; TV, tubular vesicle; T, tubule; R, ribosome. doi:10.1371/journal.pone.0048177.g003 Table 3 . Effects of EDTA and sodium sulfite on R. padi transmission of purified cereal yellow dwarf virus-RPV using microinjection into the aphid hemolymph. [90] . Each unique peptide match is based on at least 2 distinct spectra, but in some cases, many more. In the case of cysteine-rich repeat secretory protein 55-like, more than 300 total spectra were matched, highlighting the limitations of a homology-based search strategy for protein identification [79] . doi:10.1371/journal.pone.0048177.t004 were detected in the R. padi homogenate (Table 5) . A new method was created containing only these peptides, and three biological replicate samples of digested homogenates from aphids reared on infected or healthy plants were analyzed. The chromatographic retention time for each peptide was highly reproducible among all six samples ( Table 5 ). The transitions are generally free from interference when they were detected (Fig. S2 , Table S2 ) in the digested homogenates that were derived from aphids fed on infected plant tissue. Only one peptide from thaumatin-like protein 5, 33-kD secretory protein and pyruvate dehydrogenase E1 and E2 was detected (Table 5) , whereas two peptides from cellulase were detected (Table 5 , Figure S2 ). Until the oat genome is sequenced, we do not know exactly what protein we are monitoring, i.e., a single plant protein isoform or a mixture of proteins sharing the same peptide. Total peak areas for the selected peptide ions and their CID fragments were used to calculate a fold-change, and the log [2] of the fold-change is reported so that a positive value indicates an enrichment of the peptide in the extract from aphids fed on CYDV-RPV infected plants, and a negative value indicates a lower level of the peptide in the extract from aphids fed on CYDV-RPV infected plants. Pyruvate dehydrogenase E1 and E2 were found to associate with only transmissible virions (Table 4 ). Peptides derived from both of these proteins were found to accumulate to higher levels in aphids fed on CYDV-RPV infected plant tissue (Table 5 ). In fact, the peptide from pyruvate dehydrogenase E2, GLGMIAEEVK, was only detected in aphids fed on infected plant tissue (Table 5) and not in aphids fed on healthy plants. Cellulose synthase, thaumatin-like protein, and the 33-kD secretory protein co-purified with both transmissible and non-transmissible virions (Table 4 ). Cellulose synthase was detected at similar levels in aphids reared on healthy and infected plants, with fold-changes consistent across two different peptides from this protein (20.3 and 0). In contrast, peptides specific to thaumatin-like protein 5 and 33-kD secretory protein were found to accumulate to higher levels in aphids that had fed on infected plants (Table 5 ). These data suggest that the amount of these proteins was increased in aphids fed on CYDV-RPV infected tissue, with fold-changes of 2.0 and 1.6, respectively. Peptides from remorin, cysteine-rich repeat secretory protein 55-like, structural maintenance of chromosomes protein 3, predicted ribosomal protein, and b-D-glucosidase were not detected. SRM is excellent at reproducibly monitoring peptide signals; however, upon finding interesting differences such as the ones described above, the next step is validation and absolute quantification via the use of stable isotope labeled peptides. Successful virus purification depends on obtaining high titers of virus in host plants and having methods to efficiently extract biologically active virus from infected tissue. The latter is most critical since virus particles must retain their infectivity while facing the harsh oxidative environment during plant tissue homogenization. Ideally, purification methods should be optimized to obtain a high titer while maintaining biological activity for each virus species under investigation. Purified BYDV (later to be recognized as BYDV-MAV) from Coast black oats was first reported by Rochow and Brakke [36] , using 0.1 M pH 7 phosphate buffer with no additives included in the homogenization buffer. Pierpoint [37] identified a number of substances (ascorbate, ethyl xanthate diethyldithiocarbamate, cysteine, 2mercaptobenzothiazole) that prevented oxidation of polyphenolic compounds in homogenized, infected plant tissue (which can be observed as browning of the plant homogenate). As a result, reducing agents became commonplace additives to virus extraction buffers and, in many cases, with positive outcomes. In contrast to our current study, the addition of sodium sulfite and EDTA were necessary for successful purification of the potyvirus, peanut mottle virus [38] . When purifying BYDV-PAV, Hammond (1983) incorporated sodium sulfite in the extraction buffer to prevent browning, and its addition did not affect infectivity nor did reducing agents affect other luteovirids such as the potato leafroll virus (PLRV) [14] or soybean dwarf virus (B. Tian, personal communication). However these studies did not examine the virus transmissibility in the absence of reducing agents so it is unknown whether sodium sulfite imparts any negative (or positive) effects on the transmissibility of other purified luteovirids. Those reviewing plant virus purification protocols [24] cautiously warn against adoption of a one-size-fits-all protocol for purifying plant viruses. Together with the previous works, our data show such caution should be duly noted by those purifying luteovirids. The vectorvirus specificity that defines the luteovirids undoubtedly reflects the complex chemical nature of protein interactions mediated by the CP and RTP. Care should be taken in understanding how the chemistry of the purification protocol can impart biochemical changes in the virion that will alter virus-host and virus-vector protein interactions. The action of sulfite on disulfides (e.g. cystines) can be represented by the equilibrium equation: R.S.S.R + SO 3 22 « RS 2 + RS.SO 3 2 . Above pH 9, the equilibrium constant does not favor the total cleavage of the disulfide bond, and large excesses of sulfite are required to drive the reaction to completion. However, at lower pH values (e. g. pH 7 of the homogenization buffer used in these virus preparations) the thiol predominates over the thiolate ion resulting in more complex reaction kinetics and more favorable equilibriums [39] . In general, the equilibrium constant can be shifted to the right by any process that removes the thiol and in the presence of divalent metal ions, known to stabilize several icosahedral viruses [40] , and all thiol and disulfide groups can be readily converted to S-sulfonates by a process that reduces the metal ion [41] . The reaction of sulfite with protein disulfides is further complicated by an acute sensitivity to the ionic atmosphere in the neighborhood of the disulfide bond. For example, anionic disulfides have been shown to react much more slowly than neutral or cationic disulfides [42] . For cystines (disulfides) having two positively charged flanking amino acids, the rate constant for reaction of the negative disulfide is 132,000 s 21 M 21 as compared to one having two neutral neighbors, 367 s 21 M 21 , a 10 6 -fold range in rate constants [43] . Furthermore, the rate of reaction can be greatly affected by steric factors, and the disulfide bonds of many proteins show great variability in their susceptibility to cleavage by sulfite in the absence of denaturants such as guanidine hydrochloride or urea [44, 45, 46, 47] . Thus, it can be expected that in the absence of denaturants, the thiols and disulfides most susceptible to modification by sulfite would be those that are most exposed and that have positively charged amino acids as their nearest neighbors. With these thoughts in mind, the three cysteines in the CP or RTP of CYDV-RPV may provide some clues (Fig. S2 ) as to why sodium sulfite treatment would affect the virus-plant interactome. Two of the cysteines are flanked by at least one basic amino acid. These are likely the most reactive cysteines in the CP or RTP with respect to the formation of S-sulfonates and would readily participate in thiol-disulfide exchange reactions initiated by the sulfolysis of other disulfide bonds. Blocking the critical thiol by converting it to a thiosulfonate or otherwise modifying critical interactions via the formation of a non-native disulfide could have negative consequences for the internalization of RPV into aphid cells, as is observed for a wide range of other viruses (enveloped and non-enveloped) that rely on disulfide bond formation for internalization into and transport through host cells [48, 49, 50] . In contrast to the sodium sulfite treatment, EDTA treatment did not prevent entry of virus into aphid cells or transmissibility, but it did appear to reduce the long term stability of the virus inside the aphid. Hence, in vivo interactions between divalent cations and the virus capsid may be required for long-term virion stability. Indeed, other icosahedral plant viruses such as rice yellow mottle virus, tomato bushy stunt virus, southern bean mosaic virus and cowpea chlorotic mottle virus are stabilized by divalent cations such as Ca 2+ and Mg 2+ , [40, 51, 52, 53, 54] . These data support the hypothesis that transmission of CYDV-RPV requires the formation of a critical disulfide bond pairing either intramolecular, within the CP or RTP, or intermolecular with a specific host protein, and that treatment with sodium sulfite promotes a random process of thiol-disulfide exchange that creates structures that interfere with normal virus transmission. Sodium sulfite-treated virions did not enter into the HG or into the ASG and thus, a partial overlap in the biochemical mechanisms for virus entry may exist in these two aphid tissues. Previous work using infectious mutants of other luteovirids supports the role of cysteine residues within the capsid as contributing to virus-host specificity and even aphid transmission. PLRV mutants with cysteine deletions or modification of residues flanking cysteines residues show phenotypes in a host-dependent manner Two mutants D-P-K (which alters two cysteine-flanking residues) and H-C-K (which deletes the cysteine residue) showed defects in systemic infection in the host Physalis floridana, but not in Solanum tuberosom, Nicotiana benthamiana, or Nicotiana clevelandii [11] . Aphid transmission of the D-P-K mutant was severely impaired when acquired from or inoculated into P. floridana [11] . LC-MS/MS enabled us to detect numerous host proteins copurifying with virion particles in the sucrose gradients. The SRM data indicate a complex picture on the roles of these plant proteins in luteovirid transmission by aphids. How might host proteins also participate in aphid transmission? As a result of natural selection, aphid acquisition of CYDV-RPV by aphids may have been favored from phloem cells with higher protein expression, a phenotype that could have evolved as a response from virus infection, the act of aphid feeding, or both in combination. Coingestion of soluble plant proteins that associate with virions may help stimulate endocytosis into the epithelial cells of the aphid HG in a host-dependent manner. These data are consistent with previous observations of host proteins as associated with insecttransmitted viruses [21, 22] and that addition of any soluble protein into the diet acquired together with virus can enhance transmission efficiency [22] . In tobacco BY-2 cells, sugar levels tightly regulate pyruvate dehydrogenase E1 and E2 promoters. Promoter activity is markedly increased by sugar depravation [55] . We detected pyruvate dehydrogenase E1 and E2 at increased levels in aphids fed on CYDV-RPV infected plants. In contrast, cellulose synthase was detected at similar levels in aphids fed on healthy or CYDV-RPV infected plants. This is consistent with a more general role for the latter protein in aphid-plant interactions that CYDV-RPV may exploit. In varieties of wheat that are susceptible to the phytotoxic Russian wheat aphid, Diuraphis noxia, the mRNA encoding for cellulose synthase is up-regulated 4 to7fold during aphid infestation [56] . Intriguingly though, evidence suggests that the Russian wheat aphid is a poor vector for yellow dwarf viruses [57, 58] . Cellulose synthase is a member of the glycosyltransferase family A protein family. These proteins synthesize glycoconjugates by transferring a sugar moiety to a donor molecule, such as a protein or lipid. Little is understood about the role of protein glycosylation in non-enveloped virusesspecifically plant viruses [59] and a role for glycosylation in luteovirid transmission is not well understood [59, 60] . Independent of a direct role in aphid transmission, cellulose synthase and the other virus-associated host proteins may help to orchestrate virion functions in planta. Most plant viruses, including CYDV-RPV, move from cell-to-cell in host plants via plasmodesmata (PD, reviewed in [61] . Luteovirids are targeted to PD early during infection [62] . PD permeability is regulated by carbohydrate metabolism via callose deposition [63] . Callose turnover regulates PD size exclusion limit, as ectopic expression of m-type thioredoxin that is expressed in non-green plastids, which controls callose deposition, causes an increase in plasmodesmal permeability [63] . For cell-to-cell movement, plant viruses have evolved specialized mechanisms to tap into the plants endogenous system for controlling PD permeability for their own cell-to-cell transport. We propose cellulose synthase may be involved in modification of the cell-wall encasing the specialized PD to assist in cell-to-cell virion translocation. A number of other plant viruses have been known to directly interact with and use host enzymes involved in carbohydrate breakdown for cell-to-cell movement through plasmodesmata [21, 64, 65] . Susceptibility to virus infection is decreased in a class I beta-1,3-glucanase-deficient mutant of tobacco generated by stable transformation of tobacco with an antisense construct. The mutant exhibited delayed intercellular trafficking via PD of a tobamovirus (tobacco mosaic virus), of a potexvirus (potato virus X), and of the movement protein 3a of a cucumovirus (cucumber mosaic virus), as well displayed alterations in callose deposition. Through interactions with host beta-1,3 glucanase, the triple gene block protein of PVX, TGB 12 modulates plasmodesmal permeability, probably to mediate cellto-cell spread [66] . Viruses that move from cell-to-cell as ribonucleoproteins (RNPs) also recruit cell-wall modifying enzymes for cell-to-cell movement [64] , suggesting that enzymatic modulation of cell wall carbohydrates to alter PD permeability is a strategy widely used by plant viruses for intercellular trafficking. Several of the proteins found associated with CYDV-RPV have well-described functions in nuclei, for example SMC1, SMC3 [67] , and NAP [68] . Luteovirids are positive-sense, single stranded RNA viruses. Although the coat protein and the RTP of the related luteovirid PLRV can localize to the nucleolar compartment of plant cell nuclei, this localization is lost in the presence of replicating vRNA [69] . Furthermore, in the early stages of infection of oats with BYDV, infected cell nuclei become morphologically distorted and filaments associated with cytoplasmic virions appear in the nucleoplasm and in nuclear pores [62] . Thus, a physical interaction between coat protein and these proteins, in vivo, is an intriguing possibility. However, the latter three proteins have been reported to have additional, non-nuclear functions and/or localization [70, 71, 72] . Furthermore, aphid SMC proteins are commonly found associating with purified PLRV in co-immunoprecipitation studies (Cilia, unpublished) . It is also possible that maturation of phloem sieve elements may release these proteins into the sap and provide a mechanism for functional protein interactions with luteovirids, or that these proteins are transported into the phloem sap to carry out functions yet to be defined. A detailed LC-MS/MS analysis of the pumpkin phloem sap proteome revealed numerous ribosomal proteins and homologues to NAP and many other proteins with nuclear functions [26] . Excitingly, we found two remorin proteins to co-purify with transmissible virus (Table 3) . Consistent with remorin localization in plant cell membranes [73] , peptides derived from remorin were never detected from aphid homogenates (data not shown). Remorins are plant-specific proteins with unknown functions [74] but have been receiving wide attention because of their involvement in plant defense against viral, bacterial, and rhizobial infections [reviewed in [75] ]. In vivo, remorins cluster in the plasma membrane within PD and lipid rafts [73] . Remorin proteins accumulate in mature and aging tissues and in source tissue [76] where mature, branched PD are in the majority. Remorin can physically interact with the PVX movement protein TRIPLE GENE BLOCK PROTEIN 1. Remorin association with PVX is inversely proportional to the ability of PVX to move from cell-to-cell [73] . These data show that remorin proteins may function in vivo to retain virus within individual cells. Remorin association with luteovirids is particularly interesting because the luteovirid RTP retains virus in the phloem. Phloem-retention of luteovirids is critical for virus dispersal by aphids [13] . How this occurs is not known, but one possibility could be via protein-protein interactions between the C-terminal domain of the RTP and remorin. Luteovirid mutants that are no longer restricted to the phloem [13] will be particularly useful to probe direct interactions with remorin proteins. There is a paucity of tools to study vector biology and vectorvirus-host interactions at the molecular level; however, mass spectrometry technologies are emerging as one of the most powerful tools to develop a comprehensive understanding of virusvector-host interactions [9, 18, 22, 77, 78, 79, 80, 81, 82, 83] . We used mass spectrometry to describe several host proteins that associate with virions and to show they may even be ingested by aphids during feeding. The mounting body of evidence is that luteovirids commandeer their hosts and vectors to ensure their own survival and transmission. For instance, luteovirids move from companion cells into sieve elements but natural selection has favored their retention in the phloem of host plants and hence, dispersal by aphids. Luteovirid infection may also exert changes in the phloem proteome, changes that may also facilitate virus dispersal by aphids. A critical next experiment would have to distinguish between the following two hypotheses a) that virion-host protein complexes are internalized into aphids or b) that the virus manipulates the phloem to have a higher concentration of these proteins during infection. Due to the dynamic nature of protein trafficking in plants [61, 63, 84, 85, 86, 87] and cell-type specific transcriptional regulation [88] , the latter experiment is not a trivial undertaking. The current work has laid the groundwork for these future experiments. The virus-host interactome we describe in this study provides critical insights into the biochemical mechanisms that luteovirids use for movement in plants and aphids. These plant proteins (and associated biochemical pathways) are novel targets for developing host-resistance to luteovirid infection in cereals and other crops. Furthermore, the exciting, serendipitous discovery that sodium sulfite reduces transmissibility of virions provides biochemical evidence that intra or intermolecular disulfide bonding may be required for luteovirid entry into aphid cells and may also be exploited as part of a strategy to disrupt aphid-virus interactions and ultimately to mitigate virus transmission. CYDV-RPV was purified from oat plants (Avena byzantina K. Koch cv. Coast black) inoculated 7 to 8 weeks previously with viruliferous R. padi as described [10] . Infection was determined by yellowing and dwarfing symptoms. Tissue harvested was divided into 200-300 g batches, chopped into 2.5 cm pieces and frozen at 280uC until used. Virus was purified using a modified version of the protocol of Hammond et al. [89] . Tissue was homogenized using 0.1 M phosphate (K 2 HPO 4 ) buffer (pH adjusted to 7 using 0.1 M KH 2 PO 4 ) at 2.5 ml g 21 tissue. Tissue was homogenized with phosphate buffer containing 1% cellulase with and without 0.01 M EDTA and 0.5% sodium sulfite together or individually. Sucrose gradients were fractionated using a density-gradient fractionator (Teledyne-ISCO) at sensitivity 0.5 and chart speed set at 60 cm/h. Two milliliter fractions were collected along the entire gradient for each gradient. The virus fractions were collected as 4 ml gradient fractions and were concentrated by centrifugation for 4 h at 113,6136g in a type 70Ti rotor (Beckman Coulter). The remaining fractions were stored at 280uC until needed. Supernatant was discarded and the pellet was resuspended overnight in 0.01 M phosphate buffer, pH 7. Purified virus was equally divided among several tubes and stored at 280uC. As a control, 500 g of healthy tissue was homogenized in phosphate buffer containing 1% cellulase and purified following the above protocol. Purified virus from each sample preparation after purification was evaluated by negative staining. In addition, virus was recovered from membranes after aphids fed for 24 h to assess the stability of the virus. 300 Mesh copper carbon-coated formvar grids were incubated for 30 min on a 10 ml drop of RPV coating antibody diluted 1:500 in phosphate buffered saline (PBS). Excess antibody was wicked off with a wedge of filter paper and grids were rinsed in 3 drops of PBS, wicking off excess with filter paper after each rinse. Antibody-coated grids were incubated for 1 h on a 10 ml drop of virus. Virus was wicked off using filter paper and grids were rinsed in 2 drops of PBS, followed by 3 drops of sterile distilled water, wicking excess after each rinse. Grids were stained by incubating for 3 min on a 20 ml drop of 2% aqueous uranyl acetate, excess stain was wicked off and grids were stored dry in a grid box. Grids were viewed on a Jeol 1200 TEM at the Electron Microscope Facility for The Huck Institute of the Life Sciences at The Pennsylvania State University. For each purified virus preparation (buffer only; buffer including EDTA and sodium sulfite; buffer including EDTA only; buffer including sodium sulfite only), healthy R. padi aphids were allowed a 24-48 h AAP on stretched ParafilmH membranes made by standard protocol by stretching in two directions until very thin. Approximately 75 ml of the virus preparation, at 20-60 mg/ml concentration containing 15% sucrose was used in each membrane. After 24 h, aphids were allowed a 3-4 day IAP on 7 day old oat seedlings (Coast black), 5 aphids per plant, 12-16 plants per treatment. Plants were fumigated with Orthene to kill the aphids. The fumigated plants were placed in the greenhouse and observed for symptom expression 3-4 weeks later. Infected plants were evaluated by obvious yellowing and reddening of the leaves, and dwarfing of the plant. In addition, a randomly selected subset of the leaves (symptomatic and asymptomatic) was tested by double-antibody sandwich enzyme linked immunosorbent assay (DAS-ELISA) using anti-CYDV-RPV antibodies. Each type of virus preparation was tested 2-4 times. In addition to membrane feeding, virus was directly injected into the hemocoel of the aphid to bypass the HG. For each purified virus preparation, 10 nl of a 20-60 mg/ml virus preparation was injected into an aphid using a Microinjector IM300 (Narishige). Three aphids were placed onto a 7 day old oat seedling (Coast black), 12-16 plants per virus treatment, for a 4 day IAP, after which plants were fumigated to terminate feeding. Plants were evaluated for infection 3-4 weeks later as described above. To test for synergistic effects of EDTA and sodium sulfite treatment on CYDV-RPV transmission by R. padi, we performed a log-linear analysis for a three-way contingency table on the transmission results following microinjection using +/2 EDTA, +/2 sodium sulfite and +/2 transmission as the three variables. Simulations and the experiment were run using the statistical software package JMP (SAS). Healthy R. padi aphids were allowed a 24 h AAP on membranes, as described previously, which contained 40 mg/ml of purified virus in 15% sucrose buffered with 0.01 M phosphate buffer, pH 7. The following virus treatments used were preparations containing 1) phosphate buffer only, 2) phosphate buffer including EDTA only, and 3) phosphate buffer including Na 2 SO 3 only. To examine an aphid's ability to acquire virus, after 24 h, six aphids were collected, divided randomly into three groups of two for each treatment and stored at 280uC in 25 ml nuclease-free water in an RNase-free microcentrifuge tube. The remaining aphids were transferred to healthy oat seedlings (Coast black), placing five aphids per seedling. Aphids were allowed a 3 day IAP to test for transmissibility of the virus, and aphids were collected to examine the treatment effects on virus uptake into the hemocoel. For each virus treatment, six aphids were collected, divided randomly into three groups of two for each treatment and stored at 280uC in 25 ml nuclease-free water in an RNase-free microcentrifuge tube. Total RNA was isolated, five ml of RNA was used in an RT-PCR as described [14] , with the exception that primers specific for the coat protein of CYDV-RPV (RPV-CP For: 59-ATGAGTACGGTCGTCCTTAGATCC-39; RPV-CP Rev: 59-CTATTTTGGGTTTTGTAGCTGGAC-39) were used to amplify a 614 bp fragment. TEM was also used to detect virus within the aphid. R. padi were allowed a 48 h AAP on membranes containing 40 mg/ml per virus preparation. For each virus preparation, ten aphids were collected after feeding, the head and cauda were removed using a razor blade, and heads and abdomens were fixed for TEM in 1% formaldehyde-2% gluteraldehyde in 0.02 M sodium cacodylate (pH 7.2) containing 10 mM calcium chloride and 0.05% sodium azide for 24 h. Aphids were subsequently prepared for TEM accordingly [13] . The HGs of five aphids were examined per virus treatment. Grids were viewed on a Jeol 1200 transmission electron microscope (TEM) at the Electron Microscope Facility for The Huck Institute of the Life Sciences at The Pennsylvania State University. Electrophoresis 25 mg of virus preparations were separated on 10-20% Novex tricine gels (Invitrogen) according to the manufacturer's instructions, at 125 V for 2 hr at room temperature in the SureLock XCell minicell (Invitrogen). Gels were fixed for 30 min in 100 ml of a solution containing 50% methanol and 7% acetic acid. After 30 min, the fixing solution was replaced with fresh fixing solution. The gels were stained overnight with a 1:1 dilution of Sypro Ruby stain (Invitrogen) in nanopure water at room temperature in the dark. The gels were then washed for 30 min in a solution containing 10% methanol and 7% acetic acid. Gels were scanned on the Typhoon Variable Mode Imager (GE Healthcare) at 100 dpi and visualized with the 532 nm laser using the 610BP30 filter. The bands were excised from the gel using a bench top UV transilluminator at a wavelength of 302 nm. Gel bands. Each gel band was subjected to an in gel tryptic digestion and extraction as described [77] . Dried samples were reconstituted with 12 mL of 3% acetonitrile (ACN) with 0.1% trifluoracetic acid. Nano-LC (nLC) separation of tryptic peptides was performed with a nanoACQUITY system (Waters), equipped with a Symmetry C 18 5 mm, 20 mm6180 mm trapping column and a UPLC BEH C 18 1.7 mm, 15 cm675 mm analytical column (Waters). The samples were transferred to the trapping column using a 5 mL partial loop injection with a 0.1% solution of formic acid (FA) in water at a flow rate of 7 mL/min for 3 min. Mobile phase A consisted of 0.1% FA in water and mobile phase B consisted of 0.1% FA in ACN. Following desalting and concentration, the trapping column was subjected to a reverse flush to the analytical column and separated with a gradient of 2-40% mobile phase B over 60 min at a flow rate of 300 nL/min, followed by a 5 min rinse with 95% of mobile phase B. The column was re-equilibrated at initial conditions for 20 min. Column temperature was maintained at 35uC. 100 fmol/mL [Glu 1 ]-fibrinopeptide B in 25% ACN with 0.1% FA was used as the lock mass compound and was delivered via the auxiliary pump of the LC system at a flow rate of 300 nL/min to the reference sprayer of the NanoLockSpray source of the mass spectrometer. The eluent from the analytical column was delivered to the analytical sprayer of the same source through a PicoTip emitter (New Objective, Woburn, MA) with 10 mm tip diameter. Mass spectrometric analysis of tryptic peptides was performed using a Synapt HDMS mass spectrometer (Waters, Manchester, UK) or a 4700 Proteomics Analyzer (Applied Biosystems) as described in [78] and [77] , respectively. In solution digests of purified virions. For analysis of purified CYDV-RPV virions in solution, disulfide bonds were reduced with 5 mM dithiothreitol (DTT) for 30 min, followed by alkylation with 10 mM iodoacetamide (IAA) for 30 min, in the dark. Reduction and alkylation were performed at 25uC. CYDV-RPV virions were then digested with trypsin (1:200 ratio, Promega) at 35uC for 16 h. The peptide mixture was desalted using a C18 sep-pak (Waters) and stored at 280uC for 1 week prior to MS analysis. For nLC-MS/MS analysis on the LTQ-Orbitrap Velos (Thermo-Fisher Scientific, San Jose, CA), the tryptic digest was reconstituted in 10 mL of 2% ACN with 0.5% formic acid (FA). The mass spectrometer was equipped with a ''Plug and Play'' nano ion source device (CorSolutions LLC, Ithaca, NY). The nanoLC was performed using a Dionex UltiMate3000 MDLC system (Dionex, Sunnyvale, CA). The gel extracted peptides (5-10 mL) were injected using a ''User Defined Program'' onto a PepMap C18 trap column (5 mm, 300 mm65 mm, Dionex) at a 20 mL/min flow rate for on-line desalting and then separated on a PepMap C18 reverse phase (RP) nano column (3 mm, 75 mm615 cm, Dionex) which was installed in the ''Plug and Play'' device with a 10-mm spray emitter (NewObjective, Woburn, MA) mounted in front of the Orbitrap ion transfer tube. The peptides were then eluted in a 60 min gradient of 10% to 40% ACN in 0.1% FA at 300 nL/min. The Orbitrap Velos was operated as described previously [9] . For repeat injections of the same samples, an exclusion list containing m/z values identified in the previous DDA run were generated using Proteome Discovery 1.1 software and applied to prevent resampling of the same ions. All data were acquired using Xcalibur 2.1 software (Thermo-Fisher Scientific). Tandem mass spectra from purified CYDV-RPV were converted mascot generic format (MGF) peak list files using Proteome Discovery 1.1. An in-house FASTA protein database was created from all NCBI entries, including all the translations of all the available cereal genomes, common contaminants, and viruses, was downloaded on January 31, 2012. This strategy was used as opposed to restricting the search to green plants to minimize false matches due to the presence of virus gene products or to the presence of other unanticipated sources of proteins (such as from plant-infecting bacteria of fungi) in the samples. All data were searched against this database using Mascot v2.3.02 (Matrix Science, Boston, MA) as follows. Fixed carbamidomethyl and variable methionine oxidation were used as modifications. Precursor ion tolerances were set to 30 parts per million (ppm), and fragment tolerance was 0.8 Dalton (Da). ESI-Trap was selected as the instrument type. The enzyme selected was trypsin with 1 missed tryptic cleavage permitted. Mascot *.dat files were created in Mascot and loaded into Scaffold (version 3_00_05). Peptide and protein probabilities were calculated using PeptideProphet and ProteinProphet algorithms [90] . We reported protein accession numbers that could be identified on the basis of at least one peptide with a Mascot score exceeding the identity threshold and E-value ,0.05. The FDR was less than 1.0%. Spectral counts were normalized to the total and compared between treated and non-treated virions, as well as a healthy control. The healthy control consisted of the fraction in the sucrose gradient corresponding to the virus sedimentation position. Proteins were not reported if they were also detected in the healthy control (Table S1 ). To minimize redundancy due to effects of homology-based searching, only one protein per protein family were reported. Rubisco was abundant in all virus and healthy plant samples. Parthenogenetically reproducing aphid colonies of R. padi were maintained on CYDV-RPV infected or healthy caged oats at 20uC with an 18-h photo-period for 21 days. Aphids were harvested from plants for protein extraction as described [77] . Proteins were extracted from aphids using the TCA-Acetone method as described [77] . The pellets were dried and stored at 280uC until used for mass spectrometry, approximately 2 weeks. Protein pellets were prepared for mass spectrometry as previously described [80] . Pellets were solubilized by adding a volume of 50 mM ammonium bicarbonate (Sigma; St. Louis, MO)/0.1% RapiGest SF (Waters Corp.; Milford, MA) solution. Samples were left to stir overnight at 4uC and then centrifuged at 16,0006g for 5 min to pellet insoluble debris. Protein concentration of the supernatant was determined using a Quickstart Bradford assay (Biorad) and verified using 1-D SDS PAGE as described [77] . For each sample replicate, 100 mg of protein was diluted in 50 ml 50 mM ammonium bicarbonate/0.1% RapiGest SF (Waters Corp., Milford, MA) and used as the starting material for the digestion procedure. Samples were reduced with DTT at a 5 mM final concentration for 30 min at 50uC and then alkylated with IAA at a 15 mM final concentration for 30 min at room temperature, in the dark. For digestion, a 200 ng/ml trypsin (Promega; Madison, WI) solution was prepared using 0.01% acetic acid. Two mg of trypsin was added to each sample at a trypsin:protein ratio of 1:50 and incubated at 37uC for 3.5 h, with gentle vortexing every 15 min. To hydrolyze the RapiGest surfactant, samples were acidified with HCl to a pH #2, final HCl concentration of 200 mM, incubated at 37uC for 45 min, and centrifuged at 16,0006g for 10 min. The supernatant was transferred to new tubes and frozen at 280uC until mass spectrometry, approximately 48 h. Impurities were removed using mixed mode RP SCX SPE cation exchange cartridges (Waters Oasis 1cc MCX cartridge). Nano-flow liquid chromatography was performed using an Eksigent 1D nanoLC system (Dublin, CA) with direct column injection. Tips were pulled from silica capillary (75 mm I.D. 6 360 mm O.D.) in-house using a commercial CO 2 laser puller (Sutter Instruments Co., Novato, CA), and then packed to a length of 15 cm with 4 mm C12 reverse phase particles (Phenomenex, Torrance, CA). Two mL of the 1 mg/mL digested aphid protein extracts were injected directly on the column and eluted with a flow-rate of 300 nL/mn. The gradient ramped from 2% B (80:20 ACN/H 2 0) to 37% B across 50 min, and then increased to 80% B and held constant for 5 min. Initial conditions were restored for the final 15 min of the run. Electrospray ionization (ESI) was initiated by applying 2.2 kV via a liquid junction distally from the ESI tip. The capillary voltage and temperature were 42 V and 275uC, respectively. MS analyses were performed using a TSQ Vantage (ThermoFisher, San Jose, CA) operating in SRM mode. For SRM-mass spectrometry, the doubly charged precursor ions were monitored in Q1 with a resolution of 0.7 full width at half-maximum (FWHM) and three to four singly charged y-ions for each peptide were monitored in Q3 at 0.7 FWHM. Each transition was monitored for 25 ms (dwell time) enabling a maximum duty cycle of 2.5 s. Targeted protein sequences were imported into Skyline [32] and converted into trypsin fragments. Refinement was performed as described [91] . Briefly, to optimize collection of SRM data, we focused initial analysis on peptides from host proteins that could be detected in the matrix of total protein homogenates extracted from R. padi fed on CYDV-RPV infected tissues. From these samples, MS/MS data were collected for Skyline-predicted tryptic peptide ions from host proteins of interest. These data were imported back into Skyline for refinement of the SRM method. During refinement, we selected proteotypic peptides that ionized well (3-4 abundant y-ions for each peptide) and showed reproducible chromatographic retention properties and made a new SRM method. The new, data-driven, refined SRM method was exported to the mass spectrometer. Three biological replicates were analyzed using the refined SRM method, and a Student's Ttest was used to compare total peak areas. A normalization factor of 0.92 was calculated by monitoring for peptides derived from two different proteins that were not differentially expressed between treatments and applied to the peptides derived from healthy samples. Both raw and normalized peak areas for each transition are reported (Table S2 ). Figure S1 (A) 1-D PAGE of proteins from purified virus showing bands that were excised and digested with trypsin. Lane 1, Broad range molecular weight standards (Biorad) in kDa; Lane 2, transmissible virus; Lane 3; non-transmissible virus. The band containing the RTP is encircled in red. Other bands subjected to LC-MS/MS analysis are indicated with a red *. Multiple proteins were found in each lane. These proteins were abundant contaminants (reported in the text and Table S1) also found in sucrose gradients separating healthy tissue. Tryptic peptides matched to the full-length RTP in (B) non-transmissible virus purified with sodium sulfite and EDTA and (C) transmissible virus are highlighted in red. Virions were purified from the same infected source tissue (oats), the only difference was addition of sodium sulfite and EDTA in homogenization buffer. Peptides were identified from the same region of the protein in both treated and untreated virions indicating no cleavage of the RTP in the nontransmissible virion preparations. (TIF) Figure S2 Clustal W alignment showing CYDV-PRV cysteine residues in the RTP in the context of a multiple alignment of twelve luteovirid species. C136 and C373 are highly conserved among luteovirids whereas C112 is unique to CYDV-RPV. In CYDV-RPV, all three cysteine residues within the RTP are flanked by at least one basic amino acid, making them especially reactive and likely to be involved in disulfide bonding. (TIF) Figure S3 SRM transitions of the peptides from host plant proteins that were detected in tryptic digests of pooled, whole-body R. padi protein samples (from data shown in Table 5 ). One replicate per peptide is shown. Two peptides from cellulose synthase show no statistically significant fold-change in aphids reared on CYDV-RPV infected or healthy plants (A) SQTGDFDHNR detected aphids fed on CYDV-RPV infected plants or (B) healthy plants, (C) IPMFAYVSR detected in aphids fed on CYDV-RPV infected plants or (D) healthy plants. The peptide FGGDTYCCR from thaumatin-like protein 5 detected in aphids fed on CYDV-RPV infected plants in (E) or healthy plants in (F). The peptide VLYSSCYVR from 33-kD secretory protein detected in aphids fed on CYDV-RPV infected plants (G) or healthy plants (H). The peptide VLYSSCYVR was at the lower limit of detection in the samples derived from aphids reared on healthy plants. The peptide SDSIITAYR from pyruvate dehydrogenase E1 derived from samples of aphids collected from CYDV-RPV infected plants (I) or healthy plants (J). One peptide was detected from pyruvate dehydrogenase E2: GLGMIAEEVK was only detected in samples of aphids reared on CYDV-RPV infected tissues (K), and not in aphids reared on healthy tissue (L). The next step to confirm these differences in aphids fed on healthy or infected plants is validation and absolute quantification of these peptides via the use of stable isotope labeled peptides. (TIF) Table S1 Plant proteins that were identified in the sucrose gradient fractions from healthy oat plants. Table S2 Raw and normalized peak areas, T-test results, and retention time coefficient of variation for plant peptides detected using SRM in aphids fed on CRDV-RPV infected or healthy plants. (PDF)

Mortality, Severe Acute Respiratory Infection, and Influenza-Like Illness Associated with Influenza A(H1N1)pdm09 in Argentina, 2009

INTRODUCTION: While there is much information about the burden of influenza A(H1N1)pdm09 in North America, little data exist on its burden in South America. METHODS: During April to December 2009, we actively searched for persons with severe acute respiratory infection and influenza-like illness (ILI) in three sentinel cities. A proportion of case-patients provided swabs for influenza testing. We estimated the number of case-patients that would have tested positive for influenza by multiplying the number of untested case-patients by the proportion who tested positive. We estimated rates by dividing the estimated number of case-patients by the census population after adjusting for the proportion of case-patients with missing illness onset information and ILI case-patients who visited physicians multiple times for one illness event. RESULTS: We estimated that the influenza A(H1N1)pdm09 mortality rate per 100,000 person-years (py) ranged from 1.5 among persons aged 5–44 years to 5.6 among persons aged ≥65 years. A(H1N1)pdm09 hospitalization rates per 100,000 py ranged between 26.9 among children aged <5 years to 41.8 among persons aged ≥65 years. Influenza A(H1N1)pdm09 ILI rates per 100 py ranged between 1.6 among children aged <5 to 17.1 among persons aged 45–64 years. While 9 (53%) of 17 influenza A(H1N1)pdm09 decedents with available data had obesity and 7 (17%) of 40 had diabetes, less than 4% of surviving influenza A(H1N1)pdm09 case-patients had these pre-existing conditions (p≤0.001). CONCLUSION: Influenza A(H1N1)pdm09 caused a similar burden of disease in Argentina as in other countries. Such disease burden suggests the potential value of timely influenza vaccinations.

Early during the 2009 pandemic, the number of deaths attributed to influenza A(H1N1)pdm09 in Argentina (population 40 million [1] ) was only surpassed by the number of deaths in United States (population 300 million [2] ). A case-series conducted during the first months of the pandemic suggested that influenza A(H1N1)pdm09 mortality rates were at least 0.5-1.1/100.000 and concentrated among middle-aged adults with comorbidities [3] . The case-fatality proportion among hospitalized case-patients was 18% during June, 2009 [3] . Such findings suggested that the burden of influenza A(H1N1)pdm09 in Argentina was greater than that elsewhere. A subsequent review of surveillance data suggested that the apparent influenza-associated disease burden in Argentina was likely an artifact of the way that the initial severe case patients were preferentially sampled and reported to authorities. Indeed subsequent surveillance data from Argentina suggested that the proportion of persons infected with influenza A(H1N1)pdm09 who died as a result of their illness may have been similar to that elsewhere. In order to explore whether Argentina's influenza A(H1N1)pdm09 burden was higher or similar to the burden documented elsewhere, we use active facility-based influenza surveillance and health utilization surveys from three cities in Argentina to estimate rates of influenza A(H1N1)pdm09-associated mortality, hospitalization, and influenza-like illnesses. This study was conducted in three cities in Argentina, Mar de Plata (central Argentina, population 701,096), Tucumán (the largest city in northern Argentina, population 1,493,488) , and Santa Fe (east-central Argentina, population 396,243) that comprised 6% of the total population of 40 million persons in Argentina. During April-December 2009 (epidemiological weeks , surveillance staff actively searched city and hospital ledgers for all decedents with a history, in the previous two-weeks, of influenza-like illness (ILI), defined as persons with sudden onset fever [$38uC], with cough or sore throat [4] . In addition, surveillance staff actively searched for all persons hospitalized with severe acute respiratory infection (SARI), defined as persons with sudden onset fever [$38uC] with cough or sore throat requiring hospitalization as a result of complications from ILI [4] . Last, because of limited personnel, surveillance staff actively searched for a convenience sample of ILI case-patients who sought care among all sentinel city providers. Clinicians in these sentinel cities were also mandated to report all ILI case-patients whom they tested for influenza to the national surveillance system. For each SARI and ILI case patient, staff recorded the person's age; sex; date of illness onset; history of asthma or chronic obstructive pulmonary disease, diabetes, obesity, pregnancy; and survival status. As part of clinical care, a proportion of these casepatients provided both nasal and throat swabs for influenza testing during the course of their acute illness [4] . These were then tested by indirect immunofluorescence (sensitivity of ,95%) [5] and/or reverse transcription-polymerase chain reaction to identify influenza type A and B, universal swine (primers and probes designed to identify swine influenzas), and 2009 H1N1 at one of the National Influenza Centers in Argentina using methods previously described (sensitivity of ,98%) [4, 6, 7] . We conducted door to door cross-sectional surveys in each of the sentinel city populations during May-November 2010, the seasonal influenza epidemic months in Argentina, to determine if household members had developed ILI, if they had sought care, and the proportion who visited a physician multiple times for a single illness event (and therefore generated multiple case-reports in the national surveillance system). We assumed that the proportion of case-patients visiting physicians multiple times were similar during 2009 and 2010 and used bootstrapping to determine the 95% confidence interval for this proportion. We also estimated the proportion of persons in the population who were pregnant, obese, diabetic, asthmatic, or had chronic obstructive pulmonary disease and developed ILI. For each case-patient (decedent with a history of ILI, SARI, and ILI-case-patient) of a particular age group, we estimated the number of case-patients associated with influenza A(H1N1)pdm09 illness each week by adding case-patients which tested positive for influenza A(H1N1)pdm09 to the number of untested case-patients who may have tested positive for influenza A(H1N1)pdm09 if a respiratory sample had been obtained (Appendix S1). We obtained this number by multiplying the number of untested case-patients identified each week by the proportion of case-patients of the same age group which tested positive for influenza A(H1N1)pdm09 and this proportion's 95% confidence interval (Appendix S1). We calculated the number of ILI associated with influenza A(H1N1)pdm09 by adding the number ILI-case patients which tested positive for influenza A(H1N1)pdm09 to the number of untested ILI cases reported to the obligatory surveillance system multiplied by the proportion of ILI cases testing positive for influenza A(H1N1)pdm09 identified through active surveillance each week and this proportion's 95% confidence interval (Appendix S1). We adjusted each numerator by the proportion of persons without information about their date of illness onset (i.e. epidemiologic week of illness) and, in the case of ILI case-patients, for the proportion of persons who sought care multiple times per illness event as estimated by the health utilization surveys (Appendix S1). Next, we estimated the rate of influenza A(H1N1)pdm09 -associated case-patients by dividing the sum of influenza-associated decedents, SARI-case patients, and ILI-casepatients by the age-specific census population in the sentinel sites catchment during 2009 (Appendix S1) [1] . Last, we compared the probability that decedents with a history of ILI, SARI, and ILI case-patients were more likely to be of a certain age or have a preexiting medical condition using rank-sum tests and Fisher's Exact tests. The research protocol was reviewed and approved by the Argentina Ministry of Health and the Argentina office of the Pan American Health Organization. Health authorities strived to maintain case-patient confidentiality by protecting data elements that help third parties identify them. Health utilization survey participants provided verbal informed consent prior to participation. Investigators identified 108 decedents with a history of ILI during the study period (Table 1 ) ( Figure 1 ). Sixty-five (61%) of 108 were males. The median age of the 74 decedents tested for influenza was 50 years compared to 61 years among the 34 untested decedents (p = 0.01). Laboratorians identified influenza A(H1N1)pdm09 among 49 (66%) of 74 decedents tested for influenza. Accounting for age and epidemiologic week, we estimated that 66 persons died with influenza A(H1N1)pdm09 illness within our study population. We divided this estimate by the age-appropriate census population, adjusted for the proportion of decedents without known symptom onset (4 [4%] of 108), and estimated that the influenza A(H1N1)pdm09-associated mortality rate per 100,000 ranged from 1.5 (95%CI 1.5-1.7) among persons aged 5-44 years to 5.6 (95%CI 5.6-5.6) among persons aged $65 years. Investigators identified 1,622 SARI case-patients (Table 2 ) ( Figure 1 ). Eight-hundred and twenty-two (51%) were males. Among those with available age information, the median age of the 687 SARI case-patients tested for influenza A(H1N1)pdm09 was 27 years compared to 36 years among the 887 untested SARI case-patients (p,0.0001). Laboratorians identified influenza A(H1N1)pdm09 among 279 (45%) of 621 SARI case-patients with adequate laboratory samples. After accounting for age and epidemiologic week, we estimated that 600 persons of all ages developed SARI associated with influenza A(H1N1)pdm09. We divided these estimates by the age-appropriate census population and estimated that the influenza A(H1N1)pdm09-associated SARI rate per 100,000 ranged from 26. 9 Clinicians at the three sentinel cities reported 101,179 ILI cases to the national obligatory surveillance system when physicians submitted samples for respiratory virus testing (Table 3) . Investigators at the sentinel cities also identified through active surveillance 22,474 ILI case-patients who sought medical care Staff interviewed 14,535 households with 22,066 household members (mean 1.5 persons per household). Seven hundred fiftytwo (3.4%) of the 22,066 household members reported a history of ILI during the month before the interview. The ILI case patients had a median age of 20 years (IQR 7-40 years) and 339 (45%) of 752 were male (Table 4 ). Of ILI case-patients with available data, 50 had asthma, 47 had chronic obstructive pulmonary disease, 22 had diabetes, and 13 had obesity. Nine of 164 women aged 15-50 We estimated that the influenza A(H1N1)pdm09-associated ILI rates per 100 py were 1.6 (95% CI 3.8-0.32) among children aged ,5 years, 15.0 (95% CI 26.6-5.7) among persons aged 5-44 years, 17.1 (95% CI 42.8-1.7) among persons aged 45-64, and 7.1 (95% CI 9.5-5.2) among persons aged $65 years ( Our findings suggest that influenza A(H1N1)pdm09 caused a significant burden of disease in Argentina during 2009. If we assume that influenza activity was similar throughout country, we could multiply the age specific influenza-associated rates from Santa Fe, Tucumán, and Mar del Plata and their 95% confidence intervals by the census population of Argentina (3,240,001 persons aged ,5 years; 25,055,187 aged 5-44 years; 7,717,549 aged 45-64; and 4,121,684 aged $65 years) to estimate that approximately 1,300,000 (95% CI 10,000,000-1,800,000) persons visited clini-cians for ILI, 10,000 (95% CI 13,000-7,200) were hospitalized, and 990 died (95% CI 1,100-970) throughout the country as a result of the 2009 pandemic. Our mortality pandemic rates were more conservative than those estimated using linear models of 2009 Argentina pneumonia and influenza mortality data (8.4/100,000py (95% CI 6.5-10.3/ 100,000py) a method that is not well suited to differentiate the impact of influenza from that or RSV [9] . Our estimates are similar, however, to age-adjusted influenza A(H1N1)pdm09associated respiratory and cardiovascular rates for the southern cone countries (2.1-4.0/100,000py) [8] . Indeed, our estimates were similar to those of diverse countries such as Bangladesh (4/ 100,000py) [9] and subpopulations within the United States (0.9-3.7/100,000py) [9, 10] . Influenza A(H1N1)pdm09 mortality and hospitalization rates were also similar to those of seasonal influenza [11, 12, 13] . Influenza A(H1N1)pdm09 mortality was similar to Argentina's seasonal influenza mortality estimated using pneumonia and influenza diagnostic codes and Serfling models (2.3-10.6/100,000 person-years during 2002-2009) [9] and auto-regressive integrated moving averages models (0-4.6/100,000py during 1992-2002) [11, 12] . Our findings, therefore, suggest that early case-fatality proportions overestimated the actual burden of influenza A(H1N1)pdm09. Plausible explanations for this early overestimation include clinicians' preferentially identifying, sampling, and reporting severely ill case-patients. Nevertheless, it is important to note that while the overall influenza rates may be have been similar during 2009 and previous years, the years to life lost were likely greater during the pandemic because a greater proportion of ill persons were aged ,65 years than during a typical influenza season. Influenza A(H1N1)pdm09-associated hospitalizations were similar to those reported in Australia during 2009 (23/100,000py) [14] but higher to those of low income countries such as Bangladesh (13/100,000py hospitalizations) [15] . The similarities between Argentina's medically attended ILI rates and those of Bangladesh (6.6/100py) [15] , however, suggests that differences in hospitalization rates may be a factor of health utilization. [16] . Such findings suggest that ministries of health may be justified in exploring the burden of seasonal influenza in these groups and whether measures used to control and prevent influenza during the pandemic [17] would be applicable to prevent and mitigate disease among subpopulations at high risk of complications from seasonal influenza illness [18] . Our study found that the proportion of decedents with positive 2009 H1N1 samples (66%) was significantly higher than that of hospitalized SARI case-patients (42%) and ILI case-patients (25%). Such a finding suggests that while influenza was identified in a fraction of ILI case-patients [15] and community acquired pneumonias [19, 20] , influenza can be identified among a significant proportion (1/3-2/3) of severe hospitalized illness case-patients and decedents in during epidemic periods. If consistent among other surveillance platforms, such findings could have implications for countries exploring whether to empirically treat SARI case-patients with oseltamivir or other antivirals during influenza epidemic periods [21] . This study had several important limitations. We assumed that after accounting for case-definition, age-group, and epidemiologic week, the proportion of tested and untested case-patients with influenza A(H1N1)pdm09 was likely similar. This may be incorrect if physicians were more likely to test severely ill younger case-patients without pre-existing medical conditions, if the laboratory used one assay preferentially to test severely ill casepatients, and if the probability of testing positive for influenza A(H1N1)pdm09 was greater among severely ill case-patients. Although improbable, it is mathematically feasible that all untested case-patients had influenza A(H1N1)pdm09 infection, (a theoretical scenario where our influenza A(H1N1)pdm09-associated ILI rates would have been 50/100py, the hospitalization rates 48/100,000py, and the mortality rates 3.3/100,000py). Conversely, all un-tested case-patients may have been infected with other pathogens and not influenza (a theoretical scenario where our influenza A(H1N1)pdm09-associated ILI rates would have been 1/100py, the hospitalization rates 11/100,000py, and the mortality rates 2.0/100,000py). Last, we assumed that proportion of persons seeking care multiple times for ILI was similar during the 2010 and 2009 epidemic periods. Our study suggests that influenza A(H1N1)pdm09 burden in Argentina was similar to that elsewhere and caused a large number of deaths, hospitalizations, and cases of ILI. Indeed, the majority of decedents with a history of ILI identified during the pandemic tested positive for influenza A(H1N1)pdm09. Influenza-associated mortality and hospitalization rates were similar to those elsewhere and to those of Argentina during seasonal influenza epidemics. Such findings suggest that it may be prudent to examine interventions used during the pandemic to determine their potential value to prevent and mitigate Argentina's annual seasonal influenza burden. Appendix S1 Equation for calculating the rates of influenza A(H1N1)pdm09-associated mortality among decedents with a history of influenza-like illness (ILI) at three sentinel cities (DOC)

C. difficile 630Δerm Spo0A Regulates Sporulation, but Does Not Contribute to Toxin Production, by Direct High-Affinity Binding to Target DNA

Clostridium difficile is a Gram positive, anaerobic bacterium that can form highly resistant endospores. The bacterium is the causative agent of C. difficile infection (CDI), for which the symptoms can range from a mild diarrhea to potentially fatal pseudomembranous colitis and toxic megacolon. Endospore formation in Firmicutes, including C. difficile, is governed by the key regulator for sporulation, Spo0A. In Bacillus subtilis, this transcription factor is also directly or indirectly involved in various other cellular processes. Here, we report that C. difficile Spo0A shows a high degree of similarity to the well characterized B. subtilis protein and recognizes a similar binding sequence. We find that the laboratory strain C. difficile 630Δerm contains an 18bp-duplication near the DNA-binding domain compared to its ancestral strain 630. In vitro binding assays using purified C-terminal DNA binding domain of the C. difficile Spo0A protein demonstrate direct binding to DNA upstream of spo0A and sigH, early sporulation genes and several other putative targets. In vitro binding assays suggest that the gene encoding the major clostridial toxin TcdB may be a direct target of Spo0A, but supernatant derived from a spo0A negative strain was no less toxic towards Vero cells than that obtained from a wild type strain, in contrast to previous reports. These results identify for the first time direct (putative) targets of the Spo0A protein in C. difficile and make a positive effect of Spo0A on production of the large clostridial toxins unlikely.

Sporulation is an adaptive strategy that enables bacteria to survive harsh environmental conditions for prolonged periods of time, and is an integral part of the transmission of sporulating pathogens and their tolerance and resistance towards antimicrobial compounds. Spo0A is the key regulator for sporulation [1, 2] . Most of our knowledge about the protein is based on work in Bacilli. Spo0A is a response regulator that demonstrates phosphorylation dependent binding to DNA [3] [4] [5] . Phosphorylation occurs through the concerted action of several proteins that together form a so called phosphorelay [6] . The signaling cascade allows for the integration of environmental signals into the regulation of Spo0A dependent processes, including sporulation. The two functional domains, the N-terminal phosphorylation and dimerization domain (receiver domain), and the C-terminal DNA binding (effector) domain are separated by a hinge region that is relatively poorly conserved [7] . Phosphorylation is believed to result in a structural rearrangement that facilitates dimerization [8, 9] , resulting in the disruption of transcription-inhibitory contacts between the receiver and effector domains. The isolated DNA binding domain can bind legitimate targets of the Spo0A protein due to the absence of the transcription inhibitory contacts, thereby bypassing the need for phosphorylation [10] . Extensive characterization of Spo0A targets has revealed a motif that represents a high affinity Spo0A binding site, the 0A box [10, 11] . The crystal structure of the DNA binding domain confirms specific and non-specific contacts between the protein and the consensus sequence [12, 13] . It is noteworthy that Spo0A regulates many other processes than sporulation, such as competence for genetic transformation, DNA replication, and biofilm formation in B. subtilis [14] [15] [16] , virulence factors and stress responses in for instance B. anthracis and B. thuringiensis [17] [18] [19] [20] [21] , and solvent production in Clostridium acetobutylicum [22, 23] . C. difficile is a Gram positive, anaerobic bacterium that is the causative agent of C. difficile infection (CDI) (for recent reviews see [24, 25] ). Though many people are asymptomatically colonized by C. difficile, the bacterium can cause serious health problems, such as pseudomembranous colitis and toxic megacolon, under the influence of risk factors such as age and antibiotic use. As a result, CDI was long regarded a nosocomial infection. Recently, however, an increase in the cases of community acquired CDI can be observed [26] . Outbreaks of CDI have been linked to so called hypervirulent strains, such as PCR ribotypes 027 (BI/ NAP1) and 078 [27, 28] . Its main virulence factors are the major clostridial toxins A and B [29, 30] . In addition, certain strains of C. difficile, including ribotypes 027 and 078, additionally encode a binary toxin [31, 32] . C. difficile is transmitted via the fecal-oral route. It is believed that spores are crucial to successfully infect new hosts, as they are able to withstand the harsh environment of the stomach, and survive antibiotic treatments that alter the endogenous flora, after which C. difficile can overgrow [24, 25] . There is limited knowledge about the regulation of sporulation in C. difficile. It has been reported that spo0A, as expected, is required for the formation of spores [33] and the gene is required for persistence and transmission in mice [34] . Though the pathways downstream of Spo0A seem to a large extent conserved between B. subtilis and Clostridia, this is less so for the pathways leading to activation of Spo0A [2] . It has been suggested that the orphan histidine kinase CD2492 is involved in the activation of Spo0A [35] . Similarly, it was reported that multiple orphan histidine kinases can phosphorylate Spo0A in C. acetobutylicum [36] . Recently, it was reported that spo0A can be transcribed from a SigH-dependent promoter [37] . It is unknown which genes are regulated by direct binding of Spo0A to their upstream regions. Here, we establish an in vitro binding assay for C. difficile Spo0A and demonstrate for the first time direct binding of this transcription factor to DNA upstream of several putative target genes. Escherichia coli strains were routinely grown in Luria-Bertani broth or plates, supplemented with appropriate antibiotics. Chloramphenicol was used at a final concentration of 20 mg/mL for agar plates and 10 mg/mL for liquid cultures. Ampicillin was used at a final concentration of 100 mg/mL. Kanamycin was used at a final concentration of 20 mg/mL. Cloning was carried out using E. coli DH5a, overexpression was performed in E. coli Rosetta(DE3) pLysS (Novagen). C. difficile strains were grown in a glucose-free trypton-yeast based medium (TTY; 3% w/v bactotrypton (BD), 2% yeast extract (Fluka), 0.1% w/v thioglycollate (Sigma) pH 7.4), supplemented with 20 mg/mL of lincomycin when appropriate, or on CLO or TSS plates (Biomerieux). All plasmids are listed in Table 1 . Primers (obtained from Sigma Aldrich) are listed in Text S1 and specific cycling conditions are available on request. Unless noted otherwise, PCR reactions were carried out using Pfu polymerase (Fermentas) according to the instructions of the manufacturer. Plasmid pWKS1251, for the overproduction of Spo0A-DBD carrying a C-terminal 66His-tag, was constructed as follows. A sequence corresponding to the DNA binding domain of Spo0A was amplified using primers oWKS-1123a and oWKS-1124 using chromosomal DNA from C. difficile strain 630Derm as a template. The resulting fragment was cloned into pCR2.1-TOPO (Invitrogen), yielding pWKS1247. This plasmid was digested with NdeI and XhoI, separated on a 1% agarose/0.56 TAE (20 mM Tris Acetate, 0.5 mM EDTA) gel, the fragment corresponding to the DNA binding domain was recovered by gel-isolation (using a GeneJET Gel Extraction kit, Fermentas) and cloned into similarly digested pMF14 [10] that had been gel-isolated in the same manner. The construct was verified by PCR, restriction analyses and DNA sequencing using primers oWKS-135 and oWKS-136 (see below). Plasmid pWKS1245, for the production of full length Spo0A carrying a C-terminal 6xHis-tag, was constructed in a similar manner using chromosomal DNA from C. difficile 630Derm as a template, but using the PCR product of primers oWKS-1122 and oWKS-1123a. Plasmids used as PCR templates for generating EMSA probes were constructed by cloning the PCR products into pCR2.1-TOPO. The inserts, and in the case of the mutated PabrB promoters the presence of the desired point mutations in the consensus 0A box, were verified by DNA sequencing using primers oWKS-24 and oWKS-25 (see below). Sequence grade plasmids were isolated using a Nucleospin Plasmid QuickPure kit (Macherey Nagel) according to the manufacturer's instructions, except that two lysis reactions were combined onto a single filter and eluted with 65uC prewarmed AE buffer. All constructs were sequenced using BigDye Terminator chemistry (Invitrogen) on an ABI3130 sequencer (Perkin Elmer), according to the instructions of the manufacturers. In short, ,200 ng of plasmid was mixed with 3.2 pmol of primer, 1 mL Terminator Ready Reaction Mix (Invitrogen) in a final volume of 20 mL. After thermocycling, DNA was precipitated and washed with 65% isopropanol, and dissolved in 12 mL HiDi formamid (Invitrogen) at 96uC for 2 mins and stored in the dark at 4uC until the sequencing run. Sequence analyses were performed in CloneManager Professional Suite 7 (SciEd) and Geneious version 5.6.2 (Biomatters Ltd). Plasmids pWKS1245 and pWKS1251 were transformed into E. coli Rosetta(DE3) pLysS (Novagen). Transformants were used to inoculate 25 mL of LB with appropriate antibiotics. After overnight incubation, the cells were 1:100 diluted in 500 mL fresh medium containing appropriate antibiotics. Protein production was induced with 1 mM IPTG at an OD600 of 0.7 and growth was continued for another three hours before harvesting. Cells were washed with ice cold PBS and stored at 280uC for later use. Purification of the proteins was essentially done as described [10] . In short, cells were disrupted in 4 mL lysis buffer (2 mM PMSF, 10 mM imidazole, 5 mM beta-mercaptoethanol, 300 mM NaCl, 50 mM NaH 2 PO 4 , pH 7.9). Cleared cell lysates we incubated with 2 mL pre-equilibrated 50% TALON slurry (Clontech) in a final volume of 15 mL lysis buffer for 1 hr. The resin was allowed to settle on a Poly-Prep column (BioRad) and washed with 2 mL wash buffer (20 mM imidazole, 300 mM NaCl, 50 mM NaH 2 PO 4 , pH 7.9). The protein was stepwise eluted in 1 mL fractions after applying 2 mL elution buffer to the column (identical to wash buffer but with 50, 100, 250 or 500 mM imidazole). The whole procedure was carried out at 4uC. Fractions were assayed for purity and yield and suitable fractions were dialysed against 26 1L dialysis buffer (50 mM Tris-HCl pH 8, 1 mM EDTA, 0.5 mM DTT) using Slide-A-Lyzer cassettes with a molecular weight cut-off of 3.5 kDa (Pierce). Proteins were stored at 280uC in storage buffer (identical to dialysis buffer but containing 20% glycerol). Protein concentrations were determined using Bradford reagent (BioRad), according to the manufacturer's instructions. DNA fragments for use in EMSA experiment were generated by PCR using GoTaq polymerase (Promega) and chromosomal DNA from B. subtilis JH642 (Bacillus Genetic Stock Center 1A96; http://www.bgsc.org), plasmids listed in Table 1 , or chromosomal DNA from C. difficile 630Derm [38] as a template. Primers and specific cycling conditions for generation of the EMSA probes are listed in Text S1. DNA fragments of the expected size were isolated from a 16TAE/8% native polyacrylamide gel using diffusion buffer (0.5 M ammonium acetate, 10 mM magnesium acetate, 1 mM EDTA pH 8, 0.1% SDS) and a QIAExII kit (Qiagen), according to the manufacturer's instructions. Recovered DNA was end-labeled with 32P-c-ATP using FR buffer and T4 kinase (Invitrogen) according to the instructions of the manufacturer. Specific activity was determined on a LS6000 scintillation counter (Beckman). EMSA conditions were based on previous studies [10] . In short, binding reactions were carried out in binding buffer (10 mM Tris-HCl pH 7.6, 1 mM EDTA, 50 mM NaCl, 1 mM DTT, 5% glycerol) in the presence of 200 mg/mL bovine serum albumin (NEB) and 200 cpm/mL radiolabeled DNA fragment. Reactions were incubated for 20 minutes at 30uC prior to loading on a 16TAE/8% non-denaturing polyacrylamide gel that was prerun for 20 minutes at 50 V in 16 TAE buffer. Electrophoresis was carried out for 120 min at 85 V. After vacuum drying the gels onto filter paper, they were imaged after overnight exposure on Phosphorimager screens on a Typhoon instrument (GE Healthcare). The toxic effects of C. difficile culture supernatants on Vero cells (a kind gift of Eric Snijder [39] ) were determined as follows. Supernatant from a bacterial culture was harvested by centrifuging cells for 3 minutes at 140006g and filtered on a 0.45 mM cellulose acetate filter using a syringe. Supernatants were 2-fold serially diluted in cell culture medium (Dulbecco modified Eagle medium (Lonza) supplemented with 100 mg/mL penicillin, 100 U/mL streptomycin, 10% fetal calf serum), before applying them to a monolayer of Vero cells, and incubation was continued for another hour. As a positive control, 50 mL 1:10 diluted purified toxin (Techlab) was added to the cells. To determine if observed cytotoxic effects were specific for the large clostridial toxins, commercially available anti-toxin against TcdA and TcdB (Techlab) was added to 10-fold diluted bacterial supernatant for 60 min prior to incubation on the Vero cells. Toxin end-point titres were defined as the lowest dilution at which no cytopathological effects (cell rounding) were observed. Statistical significance was evaluated with an independent sample t-test. Immunization of mice with full length C. difficile Spo0A-6xHis was kindly performed at the Welcome Trust Sanger Institute (Hinxton, UK). Cells from 1 mL of C. difficile culture were collected by centrifugation for 1 min at 14000 rpm in a table top centrifuge and resuspended in 200 mL resuspension buffer (10 mM Tris HCl pH 8, 10 mM EDTA, 0.5 mg/mL lysozyme, 1 mM Pefabloc SC (Roche)). After incubation for 30 mins at 37uC, 50 mL of 56 SDS sample buffer (0.1 M DTT, 2% SDS, 50 mM Tris HCl pH 6.8, 10% glycerol, 0.0025% BPB) was added, and samples were heated to 96uC for 5 mins. Total cell lysates (amounts corrected for OD 600 ) were separated on a 12% SDS-PAGE gel prior to semi-dry blotting for 1 h at 10 V to a polyvinylidene fluoride (PVDF) membrane. Membranes were blocked in PBST buffer (phosphate buffered saline with 0.1% v/v Tween-20) containing 5% membrane blocking reagent (Amersham Biosciences). To visualize Spo0A protein cleared polyclonal serum from a single mouse at a 1:3000 dilution was used, followed by either a goat-anti-mouse HRP-conjugated secondary antibody followed by ECL+ detection (Amersham Bioscience), or a goatanti-mouse-biotin-conjugated secondary antibody (Dako) followed by a tertiary mouse-anti-biotin Cy3-conjugated antibody (Jackson). Detection was done using on a Typhoon instrument (GE Healthcare). Background corrected peak volumes were quantified using ImageQuant TL (Amersham Biosciences). Alignments of B. subtilis and C. difficile spo0A were made using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/) on the basis of the published genome sequences, Genbank accession numbers AL009126 and AM180355, respectively, and the 630Derm spo0A sequence as determined in this study. The sequence for spo0A of C. difficile strain 630Derm was deposited in Genbank (accession no JX050222). Consensus Spo0A boxes were identified using a Single string Search command in Genome2D [40] , allowing 0 mismatches. The box positions were linked to upand downstream genes using the ''Add nearest gene to List of DNA Motifs'' feature and Microsoft Excel. The results were manually inspected for those boxes within 500 bp upstream of a gene on the same strand. Figures for publication were prepared using ImageQuant TL (Amersham Biosciences), Adobe Photoshop CS3 (Adobe Systems Inc) and Corel Graphics Suite X5 (Corel Corporation). In order to characterize C. difficile Spo0A, the full length protein and its DNA binding domain (DBD) were expressed as a Cterminally 66His-tagged protein in the heterologous host Escherichia coli (Fig. 1A) and purified to near homogeneity using metal affinity chromatography ( Fig. 1A ; lanes P). Full length protein was used to raise antibodies to detect Spo0A in total lysates of C. difficile strains, and the purified DNA binding domain was used in subsequent in vitro binding assays (see below). We determined the expression of C. difficile Spo0A throughout growth. We found that the protein is present in lysates from exponential to stationary growth phase cells. We performed immunoblotting using polyclonal antibodies against C. difficile Spo0A on total lysates of wild type and spo0A mutant cells grown in a trypton-yeast based medium (TTY). We found a clear signal of the size expected for full length Spo0A (,31 kDa) as early as 3 hours post inoculation (exponential growth phase), through transition phase (8 h) as well as 24 and 48 hours post inoculation (stationary growth phase) ( Figure 1B; 630Derm) . The signals were specific for C. difficile Spo0A as they were absent from lysates from the C. difficile spo0A mutant (Fig. 1B , CT::spo0A). We obtained similar results in other media, such as the commonly used supplemented brain heart infusion broth (BHIS; data not shown). To determine relative levels of Spo0A throughout growth, we performed an immunoblot experiment using fluorescent antibodies, which gives more quantitative information compared to the use of horseradish peroxidase conjugated antibodies in our hands. We found that the levels of Spo0A increases approximately 20-fold from 6 hours post inoculation and remains at similar levels from 8 to 48 hours post inoculation ( Figure 1C ). Though it should be noted that the Western blots do not provide information on the phosphorylation state of the protein, we conclude that the protein in active or inactive form is present throughout growth and is more abundant in stationary growth phase. Spo0A of C. difficile Strain 630Derm Contains a 6aminoacid Duplication BLAST homology searches readily identify a homolog of the well-characterized B. subtilis Spo0A protein in C. difficile 630 (CD1214) and previous work demonstrated that a spo0A mutant (an insertional inactivation of cd1214) -as expected -no longer forms spores [41] . In silico analyses suggest a similar secondary structure for both proteins ( Fig. 2A) , with a conserved dimerization and DNA binding domain, separated by a poorly conserved hinge region [7, 12] . We compared the sequence of CD1214 obtained from our lab strain 630Derm [38] to that of the published C. difficile 630 genome [42] . Strain 630Derm is a spontaneous erythromycin sensitive strain, which is commonly used in mutagenesis studies and was obtained by serial passaging of strain 630 [33, 38] . The 630Derm spo0A sequence (Genbank accession no JX050222) was derived from the expression plasmids constructed for this study, and confirmed in a whole genome sequence of strain 630Derm generated in our lab (data not shown). We found that 630Derm spo0A contains an 18 base pair direct repeat, resulting in a 6 amino acid (NVGNIE) duplication compared to the published reference sequence. The duplication maps to a region of the protein with relatively low sequence conservation (hinge), flanking the highly conserved DNA binding domain ( Fig. 2A and B) . We verified the absence of this duplication in strain 630 by PCR (Fig. 2C ) as well as sequencing from the chromosomal DNA of C. difficile 630 (data not shown), to rule out an error in the original genome sequence and to demonstrate that the difference in size of the PCR product was specific to the 18 bp insertion. In addition, we checked several other strains of PCR ribotypes 12 (to which 630 and 630Derm belong) by PCR, but the duplication was found to be unique to 630Derm among the isolates tested (data not shown). C. difficile Spo0A-DBD Shows Similar Specificity as B. subtilis Spo0A-DBD Next, we examined the conservation of the DNA binding domain of Spo0A (Spo0A-DBD) between B. subtilis and C. difficile. In B. subtilis amino acid residues contacting the backbone of the DNA and interacting with specific residues of the Spo0A binding sequence have been defined [13] . We found that all these residues were conserved in the C. difficile protein sequence (Fig. 2B) , indicating that the protein likely recognizes a similar motif. DNA binding by full length Spo0A in B. subtilis requires phosphorylation dependent dimerization [8, 9] . However, it was shown that the isolated DBD is capable of binding to legitimate targets of the full length protein [10] . Analogously, we purified the C. difficile Spo0A-DBD for use in in vitro binding assays. As no direct targets for the C. difficile protein have been reported so far, we used the upstream region of the abrB gene (PabrB) of B. subtilis. PabrB is commonly used as a high-affinity control in binding assays with the B. subtilis Spo0A or Spo0A-DBD protein [43, 44] . It is noteworthy that we failed to identify a homolog of abrB in C. difficile using BLAST, indicating that potential indirect regulation by Spo0A cannot occur through abrB in C. difficile as it does in B. subtilis. We found that C. difficile Spo0A-DBD bound with high affinity to PabrB (Fig. 2D and E) . We performed electrophoretic mobility shift assays (EMSAs) using radiolabeled PabrB and increasing amounts of purified C. difficile Spo0A-DBD that was purified using a C-terminal 66His-tag. The addition of protein leads to a dose-dependent retardation of the DNA fragment with an apparent K D of ,50 nM. In the same range of protein concentrations, no binding was observed for a negative control (a DNA fragment of B. subtilis citG [45] ) (Fig. 2E) , suggesting that binding was specific for the abrB promoter region. B. subtilis Spo0A recognizes a distinct sequence (0A box), that is characterized by a 7 bp core motif (TGTCGAA) [10, 11] . Structural studies have revealed that the protein makes specific contacts with the G at position 2 (G2), and the C at position 4 (C4) and 5 (G5) of this motif [13] . We introduced G2A, C4A, G5A, G2A/C4A and C4A/G5A mutations in the perfect consensus core 0A-box present in PabrB. We found that the affinity of C. difficile Spo0A for these mutated PabrB fragments was highly reduced (Fig. 2E) . We performed EMSAs using radiolabeled PabrB containing the mutated core sequence. For the single point mutations in the DNA, the affinity decreased ,10-fold. There did not seem to be an additive effect of a second point mutation for the two combinations tested. None of the mutations abolished binding of C. difficile Spo0A completely, most likely as the result of binding of Spo0A to other (non-consensus) 0A boxes in the abrB promoter [44] . Taken together, we conclude that the guanine and cytosine residues in the core TGTCGAA motif of PabrB are important for specific binding of this fragment by C. difficile Spo0A-DBD. Value for Binding by C. difficile Spo0A-DBD Above, we have established that the Spo0A-DBD of C. difficile is highly homologous to that of the B. subtilis Spo0A protein, and that the proteins recognize a similar consensus sequence (Fig. 2 ). Based on this information, we identified the several genes as putative direct targets of C. difficile Spo0A. We queried the C. difficile 630 genome sequence for perfect matches to the core 0A box using Genome2D [40] . Such an analysis revealed the presence of 102 matching motifs, of which 45 were located within 500 bp of the initiating ATG of an open reading frame on the same strand (see Table S1 ). Our attention was drawn to spo0A and sigH, as these two genes were previously found to be regulated by Spo0A in B. subtilis and/or play important roles in sporulation [3, [46] [47] [48] . We found that C. difficile Spo0A bound to DNA sequences upstream of spo0A and sigH. We performed EMSAs with DNA encompassing 220-281 bp upstream of the initiating ATG codon of the spo0A, sigH and spoVG open reading frames. We found that the addition of Spo0A-DBD to the reactions caused retardation of the spo0A and sigH DNA fragments (Fig. 3A) , but not of a spoVG fragment which did not contain a consensus 0A box (Fig. 3B) . It should be noted that the affinity of Spo0A-DBD for the region upstream of spo0A was the highest we have observed so far for any C. difficile DNA. Moreover, the presence of multiple shifted species could indicate the presence of more than one strong binding site. These results establish that spo0A and sigH are likely legitimate targets of Spo0A in C. difficile, and confirm that spoVG is not, in line with results obtained in B. subtilis [10] . We were interested to see if Spo0A in C. difficile could potentially regulate genes that have no documented function in sporulation. Our in silico analysis identified several genes with no obvious link to sporulation that had a consensus 0A box within 100 bp upstream of their start codon. This positioning is similar to that observed for spo0A (275) and sigH (278). We confirmed in vitro binding of the C. difficile Spo0A-DBD to the promoter regions of lplA and ssuA. We carried out EMSA experiments using probes that included the perfect consensus site and purified Spo0A-DBD protein. We observed binding of the protein to fragments upstream of the lplA gene (CD1654; box at 267) and the ssuA gene (CD1484; box at 282) (Fig. 3A) . The lplA gene encodes a predicted lipoate-protein ligase, and ssuA is annotated as an aliphatic sulfonates ABC transporter; to our knowledge, neither of these have been directly implicated in sporulation or have found to be targets for Spo0A in other organisms. Together our results establish the potential for binding of Spo0A to DNA upstream of spo0A and sigH, two genes that are important for sporulation, and indicate that Spo0A may have functions that go beyond the regulation of sporulation in C. difficile. It has been established that a spo0A mutant of C. difficile does not produce any spores, consistent with a crucial role in the sporulation pathway [33] . However, the in silico identification of upstream regions with a consensus Spo0A binding site did not point to any of the early sporulation genes (downstream of spo0A itself) as direct targets of Spo0A. This is likely the result of variations in the 0A-box in these promoters that were disregarded in the box search. In support of this, many well-characterized legitimate direct targets of B. subtilis Spo0A (such as spoIIAA and spoIIE) do not contain a 100% match to the core motif, but rather one or more near-consensus boxes [5, 49] . We found that Spo0A- We performed EMSA experiments using increasing amounts of purified Spo0A-DBD from C. difficile 630Derm and the DNA fragments indicated above (Fig. 3B ). For spoIIAA (encoding an antianti sigma-factor) and spoIIE (encoding a serine phosphatase), we observed a low intensity shifted species at concentrations as low as 150 nM. For spoIIGA (encoding a sporulation specific protease) we observed the shifted species only at higher concentrations of protein (.200 nM). The negative control (spoVG) did not demonstrate binding of Spo0A-DBD at these concentrations. Moreover, the shift we observed was reversible using unlabeled DNA containing a high affinity binding site, but not using unlabeled DNA that lacked such a site ( Figure S1B-D) . Therefore, we consider the binding to spoIIAA, spoIIE and spoIIGA genes to be specific, despite the fact that increasing the amount of protein did not seem to cause a significant increase in the amount of DNA in the complex. Together, these results suggest that Spo0A in C. difficile might regulate the transcription of at least a subset of early sporulation genes by direct binding to their promoter regions. C. difficile Spo0A-DBD Binds to DNA Upstream of tcdB It has previously been reported that the deletion of Spo0A in C. difficile results in a significantly lower toxin production and a ,1000-fold reduction in the toxicity of culture supernatant derived from spo0A negative cells towards Vero cells [35] . Considering the absence of a homolog of the abrB repressor, direct binding of Spo0A and concomitant activation of toxin gene transcription is a likely mechanism through which this could occur. We found evidence for direct binding of Spo0A-DBD to the region upstream of tcdB, encoding one of the major clostridial toxin genes, and possibly tcdC, but this did not seem to result in lower toxin levels in our hands. We performed EMSAs using DNA upstream of tcdR (encoding a sigma factor responsible for the activation of toxin gene transcription), tcdB (encoding toxin B), tcdA (encoding toxin A). In order to test regions upstream of all open reading frames in the PaLoc, we also tested binding of Spo0A to DNA upstream of tcdE (encoding a holin-like protein [50, 51] ) and tcdC (encoding a putative negative regulator of toxin production [52] [53] [54] ), even though this regulator does not have a significant effect on toxin levels under the conditions we used [55, 56] . Of the regions tested, we only observed a clear shifted species, indicative of Spo0A binding, for tcdB ( Figure 4A ); the shifted species in our EMSA assay was reversed by the addition of unlabeled DNA containing a high affinity binding site, but not by DNA lacking such a site ( Figure S1E ). For tcdC, some smearing was observed at all concentrations of proteins tested ( Figure 4A ), and there did not seem to be a clear effect of the addition of unlabeled DNA fragments ( Figure S1F ). The probes for tcdA, tcdE and tcdR were indistinguishable from those obtained with our negative control, spoVG. We wanted to determine if toxin levels in culture supernatants were directly or indirectly affected by Spo0A, as was previously suggested. We found no lower toxicity towards Vero cells of culture supernatants derived from spo0A mutant cells compared to wild type. We grew three independent biological replicates of a wild type (630Derm) or Clostron-generated spo0A mutant (CT::spo0A -a kind gift of the Minton lab) in glucose-free TTY medium. We harvested culture supernatant at late-exponential phase (approximately 7 hours post inoculation), the transition phase between exponential and stationary growth phase (approximately 9 hours post inoculation), as well as two time points in stationary phase (24 and 48 hours post inoculation) and determined the toxin endpoint titres (see Materials and Methods). In contrast to previous findings, we observed a small (#4-fold) increase in the toxicity of supernatants derived from spo0A mutant cells compared to wild type, but in all cases this difference was not statistically significant (p.0.05, independent sample t-test). In other medium (BHIS), we observed no differences at all (data not shown). We conclude that Spo0A does not positively affect toxin production in C. difficile 630Derm and the in vivo relevance of the binding to regions upstream of tcdB and/or tcdC is therefore limited under our experimental conditions. The Spo0A-box of C. difficile In B. subtilis, the binding site of Spo0A on target DNA has been well-characterized, through a combination of in vitro binding assays, determination of in vivo binding profiles and mutagenesis of regulated promoter sequences. This work has led to the identification of a conserved core motif, TGTCGAA, or Spo0A box [5, 10, 11, 45] . Depending on the analysis, this motif is flanked by one or more adenine or thymine residues [10, 11] . Interestingly, many target genes do not harbor a perfect match to this consensus sequence, but rather contain one or more degenerate motifs. The differences in these motifs may reflect different promoter architectures (e.g. AT content), modes of action (e.g. activation or repression) or levels of regulation. Spo0A genes in B. subtilis can be divided in different classes that respond to different levels of phosphorylated Spo0A [43, 57] . For C. difficile, we conclude that the Spo0A protein likely recognizes a motif that is similar to the B. subtilis Spo0A box on the basis of four lines of evidence; 1. All DNA binding/contacting residues are conserved (Fig. 2B) , 2. C. difficile Spo0A can bind with high affinity to a target of B. subtilis Spo0A (Fig. 2D) , 3. Mutagenesis of key residues in the B. subtilis Spo0A box reduces affinity of C. difficile Spo0A for DNA (Fig. 2E ) and 4. A B. subtilis Spo0A box has predictive value for DNA binding by C. difficile Spo0A (Fig. 3A) . It is conceivable that our model system, using the purified DNA binding domain, does not accurately reflect binding to all target sites, if target site selectivity is determined in part by other parts by of the full length protein. It is likely that differences do exist between the preferred binding sites for both proteins that will be evident when a comprehensive analysis is performed of in vivo DNA binding of C. difficile Spo0A; based on the limited data set of this study, a MEME analysis [58] already suggests possible differences in the extended Spo0A motif (W.K. Smits, unpublished observations). These differences may relate to the much higher AT content of C. difficile compared to B. subtilis (71 vs. 56.5%, respectively), or phosphorylation dependent dimerization, for instance. The initiation of sporulation in B. subtilis is subject to complex regulation (for review see ref [1, 59] ). The activation of Spo0A is controlled by a multi-component phosphorelay that can integrate environmental cues [60] and ensures a gradual increase in the level of phosphorylated Spo0A in the cell [57] . In addition, the transcription of the spo0A gene is controlled by multiple feedback loops. For instance, Spo0A regulates its own transcription by binding to the spo0A promoter [46] , as well as by indirectly stimulating the transcription of sigH, encoding a sigma factor that recognizes the spo0A promoter [48] . In C. difficile, there are some interesting differences and similarities in the regulatory pathways. Most notably, there seems to be no phosphorelay [2] and the phosphorylation state of Spo0A is supposedly controlled by orphan histidine kinases [35] . The transcription of spo0A in C. difficile is under control of the transition state sigma factor Sigma H [37] , as it is in B. subtilis [61] . Our data indicate that both spo0A and sigH could be targets for direct regulation by Spo0A in C. difficile (Fig. 3A) , raising the possibility of auto-regulation of spo0A. The putative direct regulation of sigH by Spo0A may reflect that the C. difficile genome does not harbor a homolog of the pleiotropic regulator AbrB, which is responsible for the Spo0A-dependent regulation of sigH in B. subtilis [48] . Consistent with a model in which spo0A is positively autoregulated, we noted a sharp increase in the levels of Spo0A as cells approach the stationary growth phase ( Figure 1C) . Downstream of Spo0A, we found binding of Spo0A to DNA upstream of several early sporulation genes, such as spoIIAA, spoIIE, and spoIIGA (Fig. 3B ). All these observations are consistent with direct regulation of these genes by Spo0A in other organisms [5, 45, 49, 62] , and the conservation of the sporulation pathway [2] . Though Spo0A is the key regulator for sporulation in Firmicutes, it regulates numerous other processes in various bacteria. In the non-pathogenic B. subtilis, for instance, the protein also affects competence development, biofilm formation, the production of and resistance to antimicrobial compounds, chromosome dynamics and aspects of phage biology [10, [14] [15] [16] . Importantly, several of these processes are indirectly regulated, through the Spo0A-dependent repression of abrB. Additionally, transcription of abrB responds already to low levels of Spo0A,P [43] . As a result these effects are detectable in late-exponential and early stationary phase, as some Spo0A is present throughout growth in B. subtilis cells. Though abrB is absent from C. difficile, this does not exclude the possibility of indirect transcriptional regulation through Spo0Adependent effects on other regulators. Alternatively, Spo0A may exert a direct effect. In Clostridium acetobutylicum and C. beijerinckii, Spo0A is a direct regulator of solvent formation, as well as sporulation [22, 23] . It seems therefore conceivable that Spo0A in C. difficile also affects aspects of metabolism. In this respect, it is important to note that also in C. difficile Spo0A is detectable from early exponential growth phase on ( Figure 1B) . We observed direct binding of C. difficile Spo0A to the promoter region of sigH (Fig. 3A) . This gene encodes the key sigma factor for the transition phase, and regulates processes outside sporulation as well [37] . Moreover, we found significant levels of Spo0A from early stationary phase on ( Fig. 1B and unpublished observations) , indicating the regulatory actions of Spo0A need not be limited to stationary phase in C. difficile. In line with this idea, we found a potential regulatory link between Spo0A and two genes that to our knowledge are not related to the sporulation process, the lipoate ligase lplA and the aliphatic sulfonates transporter ssuA (Fig. 3A) . The presence of a putative Spo0A binding site upstream of these genes, as well as the spacing compared to the start codon, is conserved in the problematic Stoke-Mandeville strain (R20291), a member of PCR ribotype 27. This could indicate that these aspects of regulation by Spo0A are conserved in multiple strains of C. difficile. It should be noted that our work so far has been limited to an in vitro analysis of Spo0A binding, and therefore does not indicate whether activation or repression of the putative target genes occurs in vivo. To answer this question, detailed transcriptome and/or proteome studies have to be performed. In order to distinguish direct from indirect effects, in vivo binding profiles of Spo0A should be performed. The antibodies generated for this study should prove to be useful for this type of experiments. Amongst the pathogenic Firmicutes, Spo0A has been reported to affect toxin production in multiple species. In B. anthracis a spo0A mutation results in elevated levels of AbrB, and concomitantly lower levels of the toxin genes pagA, cya and lef that are under AbrB control [17] . Similarly, the production of the emetic toxin cereulide in B. cereus is greatly repressed in a spo0A mutant, in an AbrB-dependent manner [63] . In contrast, Spo0A directly represses the expression of the cry toxin genes in B. thuringiensis and a spo0A mutant is therefore a hyper-producer of the insecticidal crystal protein [18, 21] . In Clostridium perfringens TpeL, a member of the large clostridial toxins just like TcdA and TcdB, is directly dependent on Spo0A [64] and also the production of enterotoxin in this organism seems to be (indirectly) dependent on sporulation [65, 66] . In C. difficile an insertional spo0A mutant generated using Clostron technology was reported to have ,10-fold reduced levels of toxin A (TcdA), both intracellularly and extracellularly as well as ,1000-fold reduced toxicity towards Vero cells, which are primarily sensitive towards toxin B (TcdB) [35] . Our in vitro binding data indicate a potential binding site for Spo0A upstream of tcdB and possibly tcdC (Fig. 4A) . However, the in vivo relevance of this binding seems limited as in our hands an independently derived but otherwise identical mutant (a kind gift of the Minton lab; [33] ) did not demonstrate a reduced toxicity towards Vero cells. In contrast, we found that in TTY medium toxin levels were slightly elevated in spo0A mutant cells compared to wild type (#2fold in exponential phase cells up to 4-fold in late-stationary phase cells). The small, and not significant, differences in toxin levels in our experiments might be attributed to differences in the susceptibility of cells for lysis rather than the production of toxin, but could also indicate a negative regulatory effect of Spo0A on toxin production. In support of the latter hypothesis, it was recently reported that a spo0A mutant of C. difficile strain R20291 (a PCR ribotypes 027/BI/NAP1 epidemic strain) demonstrates ,10fold higher toxin levels than its isogenic wild type 30 h post inoculation, and is significantly more virulent in a mouse model of disease [34] . The differences between Underwood et al [35] on the one hand and our study as well as the study of Deakin and coworkers [34] on the other hand may be explained by differences in experimental conditions, such as the medium used. However, we observed no difference in cytotoxicity between supernatant derived from wild type or spo0A mutant cells when they were grown in BHIS, a medium nearly identical to that used previously (data not shown). Alternatively, the differences could indicate integration of the group II intron at more than one location in the chromosome in the strain used in Underwood et al [35] . In the absence of a complementation experiment and/or Southern blot data, this remains to be established. In summary, our data are consistent with a model in which the regulation of the major clostridial toxins in C. difficile is not positively affected by Spo0A, in contrast to previous findings and other pathogenic Clostridia. Whether Spo0A is truly a negative regulator of toxin production remains to be confirmed using in vitro and in vivo transcription assays. In the present study we have for the first time demonstrated direct binding of the DNA binding domain of C. difficile Spo0A to putative target DNA. This work has revealed that aspects of Spo0A binding are conserved between Bacillus and C. difficile (0A box, possible auto-regulation and binding to early sporulation promoters), whereas others are not (the absence of abrB as a direct target in C. difficile, binding to DNA upstream of lplA, ssuA). The effects of Spo0A on toxin production may be similar to those observed for B. thuringiensis [18, 21] . Future work will be aimed at determining the effect of Spo0A on the transcription of the putative target genes, and carry out a comprehensive analysis of Spo0A binding in vivo. The identification of genes affected by Spo0A in C. difficile may shed light on the role of the protein in virulence and pathogenesis of this organism. Figure S1 Specificity controls for binding by Spo0A-DBD-his6. Arrows indicate the position of shifted species (DNA:protein complexes). Titrations with PCR fragments of PabrB (containing a high affinity binding site) and PtcdA (lacking such a site) correspond to approximately 0.1 nM/mL -0.03 nM/ mL. A. Comparison of binding of Spo0A-DBD-his6, Spo0A-his6 and CD2195-his6 binding to the upstream region of spoIIAA. B. Binding of Spo0A-DBD-his6 to the upstream region of spoIIAA is reversed by the addition of PabrB, but not by the addition of PtcdA). C. Binding of Spo0A-DBD-his6 to the upstream region of spoIIE is reversed by the addition of PabrB, but not by the addition of PtcdA. D. Binding of Spo0A-DBD-his6 to the upstream region of spoIIGA is reversed by the addition of PabrB, but not by the addition of PtcdA. E. Binding of Spo0A-DBD-his6 to the upstream region of tcdB is reversed by the addition of PabrB, but not by the addition of PtcdA. F. Binding of Spo0A-DBD-his6 to the upstream region of tcdC is not or moderately affected by the addition of PabrB and/or PtcdA. (TIF) Text S1 Oligonucleotides used in this study and PCR cycling conditions for the EMSA probes. (PDF)

Sodium taurocholate cotransporting polypeptide is a functional receptor for human hepatitis B and D virus

Human hepatitis B virus (HBV) infection and HBV-related diseases remain a major public health problem. Individuals coinfected with its satellite hepatitis D virus (HDV) have more severe disease. Cellular entry of both viruses is mediated by HBV envelope proteins. The pre-S1 domain of the large envelope protein is a key determinant for receptor(s) binding. However, the identity of the receptor(s) is unknown. Here, by using near zero distance photo-cross-linking and tandem affinity purification, we revealed that the receptor-binding region of pre-S1 specifically interacts with sodium taurocholate cotransporting polypeptide (NTCP), a multiple transmembrane transporter predominantly expressed in the liver. Silencing NTCP inhibited HBV and HDV infection, while exogenous NTCP expression rendered nonsusceptible hepatocarcinoma cells susceptible to these viral infections. Moreover, replacing amino acids 157–165 of nonfunctional monkey NTCP with the human counterpart conferred its ability in supporting both viral infections. Our results demonstrate that NTCP is a functional receptor for HBV and HDV. DOI: http://dx.doi.org/10.7554/eLife.00049.001

Approximately 2 billion people have been infected with human hepatitis B virus (HBV) worldwide. Over 350 million people currently are chronically infected and are at high risk for progression to cirrhosis, liver failure, or cancer. More than 50% of liver cancers worldwide are attributed to HBV infection. HBV-related liver diseases remain a major public health problem, causing approximately 1 million deaths per year. Individuals coinfected with HBV and HDV are at greater risk for rapid progression and severe disease (Lavanchy, 2004; Hughes et al., 2011) . Despite its enormous medical and social relevance, progress in HBV research has been impeded by the lack of understanding of HBV entry by which the virus specifically infects human liver cells. HBV is an enveloped virus containing a small genome of 3.2 kb of partially double-stranded DNA encoding four overlapping reading frames. The HBV envelope consists of the small (S), middle (M), and large (L) envelope proteins, which are multiple transmembrane spanners sharing the same C-terminal domain corresponding to the S protein but differing at their N-terminal domains ( Figure 1A ) (Heermann et al., 1984; Seeger et al., 2007) . HDV is a small satellite RNA virus of HBV carrying all three HBV envelope proteins and can only propagate when coexisting with HBV. The mechanism of viral entry of HDV is believed to be similar to that of HBV, and HDV has been used as a surrogate for studying HBV infection at the entry level (Barrera et al., 2004; Sureau, 2006; Hughes et al., 2011) . The L protein and integrity of S protein are critical for HBV Photoreactive ligand peptides for identification of interacting protein(s) of pre-S1 domain of L envelope protein To identify the pre-S1 interacting molecule(s), we employed a photo-cross-linking approach using a synthetic peptide derived from the native pre-S1 peptide with particular residues replaced by eLife digest Liver diseases related to the human hepatitis B virus (HBV) kill about 1 million people every year, and more than 350 million people around the world are infected with the virus. Some 15 million of these people are also infected with the hepatitis D virus (HDV), which is a satellite virus of HBV, and this places them at an even higher risk of liver diseases, including cancer. The viruses are known to enter liver cells by binding to receptors on their surface before being engulfed. Both HBV and HDV have outer coats that consist of three kinds of envelope proteins, and a region called the pre-S1 domain in one of them is known to have a central role in the interaction between the viruses and the receptors and, therefore, in infecting the cells. However, the identity of the HBV receptor has remained a mystery. Now Yan et al. have identified this receptor to be sodium taurocholate cotransporting polypeptide. This protein, known as NTCP for short, is normally involved in the circulation of bile acids in the body. In addition to humans, only two species are known to be susceptible to infection by human HBV and HDV-chimpanzees and a small mammal known as the treeshrew. Yan et al. started by isolating primary liver cells from treeshrews, and then used a combination of advanced purification and mass spectrometry analysis to show that the NTCP on the surface of the cells interacts with the pre-S1 domain in HBV. The authors then performed a series of gene knockdown experiments on liver cells of both human and treeshrew origin: when the gene that codes for NTCP was silenced, HBV infection was greatly reduced. Moreover, they were able to transfect HepG2 cells-which are widely used in research into liver disease, but are not susceptible to HBV and HDV infection-with NTCP from humans and treeshrews to make them susceptible. Similarly, although monkeys are not susceptible to HBV, replacing just five amino acids in monkey NTCP with their human counterparts was enough to make the monkey NTCP a functional receptor for the viruses. In the past, basic research into HBV and the development of antiviral therapeutics have both been hindered by the lack of suitable in vitro infection systems and animal models. Now, the work of Yan et al. means that it will be possible to use NTCP-complemented HepG2 cells for challenges as diverse as fundamental studies of basic viral entry/replication mechanisms and large-scale drug screening. It is also possible that HBV and HDV infection might interfere with some of the important physiological functions carried out by NTCP, so the latest work could also be of interest to medical scientists working on other diseases related to these infections. Research article Figure 1 . Developing photoreactive peptide ligands and an antibody for identifying pre-S1 binding partner(s) by zero distance cross-linking. (A) Schematic diagram of HBV envelope proteins and N-terminal peptides of pre-S1 domain. Pre-S1 (2-47): 2-47 th residues of the pre-S1 domain of the L Figure 1 . Continued on next page nonnatural amino acids (L-photo-leucine, L-2-amino-4,4-azi-pentanoic acid) ( Figure 1A) . L-photoleucine contains a photoactivatable diazirine ring. Irradiation of ultraviolet (UV) light at 365 nm induces a loss of nitrogen of the diazirine ring and yields a reactive carbene group with short half-life for covalent cross-linking at nearly zero distance (Suchanek et al., 2005) . Primary hepatocytes isolated from treeshrews (Tupaia belangeri), the only species susceptible to human HBV infection other than humans and chimpanzees (Su et al., 1987; Walter et al., 1996; Glebe et al., 2003) , were used as target cells. To maximize the efficiency of photo-cross-linking, two residues (leu 11 and phe 14 ) in a region (aa 9-15) known to be critical for viral infection (Schulze et al., 2010) were chosen for substitution with L-photoleucine. Leu 11 is 100% conserved among HBV genotypes, and the 14th residue is a phenylalanine in most genotypes but a leucine in some HBV strains of genotypes F and G. Changing phe 14 to leucine (F14L) did not significantly affect the binding of HDV virion to primary Tupaia hepatocytes (PTHs) ( Figure 1B) . The activity of the synthesized peptide ligand Myr-47/WT b (or WT b hereafter) containing photo-leucines at positions 11 and 14 was also confirmed ( Figure 1C,D) . WT b inhibited HDV binding to PTHs with efficiency comparable to Myr-47/WT that is comprised of all natural amino acids ( Figure 1A,C) . A peptide Myr-47/N9K b (or N9K b hereafter) similar to WT b but with an additional mutation at the ninth residue (N9K) did not block HDV binding to PTHs ( Figure 1C ). WT b but not N9K b inhibited viral infection of HBV and HDV on PTHs ( Figure 1D ). Both WT b and N9K b peptides were myristoylated at the N-terminus and conjugated with a biotin tag on a C-terminal lysine residue ( Figure 1A ). N9K b differs from WT b by only one amino acid but completely lost these blocking activities. Thus, N9K b was used as a negative control for WT b . In addition, a monoclonal antibody (mAb) 2D3, which specifically recognizes an epitope adjacent to the critical receptor-binding region of the peptides and shared by both WT b and N9K b , was developed ( Figure 1E ). Identification of NTCP as a specific binding protein of pre-S1 The WT b or control N9K b peptide at 200 nM was then applied to PTHs in culture and near zero distance cross-linking was induced by UV irradiation. The cross-linked peptide and associated partners were precipitated by streptavidin T1 beads and separated by SDS-PAGE. Western blotting using 2D3 as a probe revealed several bands including a major smeared band with apparent molecular weight of ∼65 kDa in the WT b but not N9K b cross-linked sample. The 65-kDa band shifted to ∼43 kDa upon treatment with the deglycosylation enzyme PNGase F (Figure 2A, left) , indicating that it is highly N-glycosylated. The WT b cross-linked protein apparently contained no intermolecular disulfide bonds as it migrated similarly under both nonreducing and reducing conditions (Figure 2A, right) . The non-photoreactive Myr-47/WT peptide but not its N9K mutant peptide effectively competed with WT b for cross-linking to the 65-kDa band ( Figure 2B) . The cross-linked protein from PTHs decreased in abundance rapidly over time during culture ( Figure 2C) . We also examined primary human hepatocytes (PHHs) in the crosslinking experiments. Bands with slightly smaller molecular weights than those seen in the PTH cells were also observed in PHHs ( Figure 2D) . We then proceeded to identify the target protein(s) using affinity purification followed by mass spectrometry (MS) analysis. The purification procedure included three tandem steps after photo-cross-linking: protein of HBV (S472 strain, genotype C). Residue numbering is based on genotype D. Asterisk indicates highly conserved residues among genotypes. Epitope of mAb 2D3 was shaded in gray. (B) Effect of alterations of the critical N-terminal residues within pre-S1 region of L protein on HDV binding to PTHs. Both wild-type (WT) and mutant HDV virions carry HBV envelope proteins. Mutant HDV carries point mutation as indicated in the pre-S1 region of L protein. PTHs were incubated with HDV at 16°C for 4 hr and followed by extensive wash; bound virions were quantified by qRT-PCR for virus genome RNA copy, and the data are presented as percentage of virus binding, the binding of WT virus was set as 100%. (C) Myr-47/WT b bait peptide dosedependently inhibited HDV virion binding. The binding assay was performed similarly as panel B except that PTHs were pre-incubated with indicated peptides. (D) Inhibition of viral infection by the photoreactive peptides. Left: PTHs were pre-incubated with peptides at indicated concentrations at 37°C for 1 hr and then inoculated with HDV virus. Viral infection was examined by measuring viral RNA in infected cells with qRT-PCR 6 days post-infection (dpi). Data are presented as percentage HDV infection. Right: peptides at indicated concentrations were added to PTHs before HBV inoculation. The cell culture medium was replenished every 2 days. Secreted viral antigen HBeAg was measured by ELISA on 6 dpi, and the data are presented as percentage of that in the absence of peptides. (E) Antibody 2D3 recognizes residues 19-33 of pre-S1. Peptide NC36 (aa 4-36 of pre-S1, NLSVPNPLGFFPDHQLDPAFGANSNNPDWDFNP) conjugated with keyhole limpet hemocyanin (KLH) was the immunogen peptide for generating mouse mAb 2D3. Binding activity of 2D3 with full-length pre-S1 protein was measured by ELISA in the presence of competition peptides at indicated concentrations. LD15 peptide compassing residues 19-33 of pre-S1 inhibited 2D3 binding in a dose-dependent manner, indicating that 2D3 recognizes an epitope within this region. HBV: hepatitis B virus; mAb: monoclonal antibody; HDV: hepatitis D virus; PTH: primary Tupaia hepatocytes; HBeAg: HBV e antigen. DOI: 10.7554/eLife.00049.003 ) or Myr-47/N9K b (N9K b ), followed by Streptavidin Dynal T1 beads precipitation and Western blot analysis using mAb 2D3. The protein cross-linked by WT b is sensitive to PNGase F treatment and shifted from ∼65 to ∼43 kDa. Right: WT b cross-linked samples were treated with 100 mM DTT and/or PNGase F as indicated and detected similarly as in the left panel. (B) Non-photoreactive Myr-47/WT peptide (WT) but not its N9K mutant competed with 200 nM of WT b peptide for cross-linking with PTHs in a dose-dependent manner. (C) The abundance of the target protein(s) in PTH cells decreased over time. PTHs on different days of in vitro culturing were photo-cross-linked with 200 nM WT b . The cross-linked samples were analyzed by Western blot. The two bands at ∼65 and ∼43 kDa were due to incomplete deglycosylation by PNGase F. (D) WT b cross-linking with primary human hepatocytes (PHH). Frozen PHH cells were thawed and plated 1 day before cross-linking. With same procedure as in panel A, 200 nM WT b but not N9K b cross-linked with a glycoprotein of molecular weight at ∼60 kDa, which shifted to ∼39 kDa upon PNGase F treatment. (E) Purification of target protein(s) for MS analysis. PTHs photo-cross-linked with 200 nM of WT b or N9K b peptide were lysed, then the peptides and their cross-linked proteins were purified in tandem with Streptavidin Dynal T1 beads, mAb 2D3 conjugated beads, and Streptavidin Dynal T1 beads in 1× RIPA buffer. Extensive wash was applied for each purification step. The samples were treated with or without PNGase F as indicated prior to the last step of Streptavidin beads precipitation. The final purified samples were subjected to SDS-PAGE followed by silver staining (left). Bracketed areas indicate the bands cut for MS analysis. Western blot analysis (right) of the same cross-linked samples were performed similarly as in panel A. The top 10 nonredundant proteins identified in the 3 samples by MS analysis are listed in Figure 2 -Source data 1. The common protein hit identified by MS analysis of the ∼65and ∼43-kDa bands cut from the WT b cross-linked sample was Tupaia NTCP (tsNTCP), and the representative MS/MS spectra and parameters of the peptide hits are shown in Figure 2 -figure supplement 5. The control band cut from N9K b cross-linked sample did not generate any hits on any of these peptides. (F) Predicted tsNTCP protein sequence. A 30-amino acid insertion unique to tsNTCP is underlined. Two peptides identified by LC-MS/MS were highlighted in green. All lysine and arginine are highlighted in red to Figure 2 . Continued on next page capturing all biotin-labeled proteins with streptavidin T1 beads, sorting out the target protein(s) with 2D3 antibody affinity beads, and then purifying with streptavidin T1 beads again to remove residual molecules that were not covalently cross-linked with the bait peptide. The purified samples were subsequently subjected to SDS-PAGE followed by silver staining. Similar to the Western blotting results with the 2D3 antibody, a ∼65-kDa protein band was visible by silver staining. The band was also shifted to ∼43 kDa upon PNGase F treatment ( Figure 2E ). Both the original 65-kDa and the shifted 43-kDa bands were subsequently excised from the gel and subjected to LTQ-Orbitrap Velos (Thermo Fisher Scientific, MA. USA) MS analysis after trypsin digestion. The tandem mass spectra were searched against a Tupaia hepatocyte protein database, which we had established by deep sequencing of the transcriptome (Figure 2-figure supplements 1-4) . Two different tryptic peptide fragments, which were identified from both the ∼65-kDa and ∼43-kDa bands ( Figure 2-figure supplement 5 ), matched to a protein homolog of human NTCP. Tupaia NTCP (tsNTCP) shares 83.9% protein sequence identity with its human counterpart and has an insertion of 30 aa near its C-terminus ( Figure 2F ). The peptide (TEETIPGTLGNSTH) containing 4 aa of this insertion (underlined) was one of the two peptides identified by the MS analysis at a high confidence level ( Figure 2-figure supplement 5) . These data suggest that NTCP is the protein specifically interacting with the WT b bait peptide. Confirmation of NTCP as a specific binding protein of pre-S1 We next cloned human and Tupaia NTCPs and validated the binding of the exogenously expressed NTCPs with the WT b peptide and an N-terminal myristoylated pre-S1 peptide with native residues. Both human NTCP (hNTCP) and tsNTCP could be efficiently cross-linked by WT b but not N9K b when expressed in 293T cells as shown by Western blotting with the anti-WT b antibody 2D3 as well as an anti-C9 antibody recognizing the C-terminal C9 tag of the recombinant hNTCP and tsNTCP proteins ( Figure 3A ). WT b but not the control N9K b peptide bound to 293T cells expressing a green fluorescent protein (GFP)-tagged tsNTCP (tsNTCP-EGFP) and co-localized with tsNTCP-EGFP on the cell surface. This binding was readily competed off by the free Myr-47/WT peptide ( Figure 3B ). Moreover, a native pre-S1 peptide specifically recognized the human hepatocellular carcinoma Huh-7 cell line transfected with hNTCP ( Figure 3C ). Consistently, Huh-7 cells transfected with either tsNTCP or hNTCPs had markedly increased HDV binding to the cells. The Myr-47/WT peptide readily competed with binding of the wild-type HDV, whereas a noninfectious mutant HDV virus bearing a single N9K mutation in the pre-S1 domain of its L envelope protein failed to bind either hNTCP-or tsNTCPexpressing Huh-7 cells ( Figure 3D) . Collectively, these data demonstrated a specific interaction between NTCP and the pre-S1 domain of the L protein, which directly mediates the binding of HDV virions to target cells. To test the requirement of endogenous expression of NTCP for HBV and HDV infection, we first examined the effect of NTCP gene silencing on viral infection of PTHs. PTHs were transfected with tsNTCP-specific or a control small interfering RNA (siRNA) prior to viral inoculation. When tsNTCP Figure 3 . Binding of NTCP with N-terminal peptide of pre-S1 and HDV virions. (A) 293T cells transfected with an expression vector or plasmid containing cDNA of hNTCP or tsNTCP fused with a C9 tag at its C-terminus were cross-linked with 200 nM Myr-47/WT b or Myr-N9K b similarly as in Figure 2A at 24 hr post-transfection. Cross-linked protein samples were precipitated by Streptavidin Dynal beads followed by treatment with PNGase F as indicated, and then analyzed by Western blotting using mAb 2D3 or anti-C9 tag antibody. (B) 293T cells transfected with tsNTCP-EGFP or a control hSDC2-EGFP (encoding human heparan sulfate proteoglycan core protein fused with EGFP at C-terminus) expression plasmid were incubated with WT b or N9K b in the presence or absence of 200 nM non-photoreactive Myr-47/WT as indicated. Bound peptides were probed with PE-streptavidin and the colocalization of peptide and NTCP on cell surface was shown in the merged images. (C) FACS analysis of pre-S1 peptide binding with hNTCP transiently transfected Huh-7 cells. 24 hr post-transfection with hNTCP or a control plasmid, the cells were stained with 200 nM FITC-pre-S1 (FITC-labeled lipopeptide corresponding to the N-terminal 59-amino acid of pre-S1). The binding was analyzed by flow cytometry. (D) Huh-7 cells, after 24 hr of transfection of indicated plasmids, were incubated with wild-type HDV or HDV with a N9K mutation on its L protein. Bound virions were quantified by qRT-PCR. The result is presented as fold changes of binding over the background virus binding to pcDNA6-transfected cells. mAb: monoclonal antibody; tsNTCP: Tupaia NTCP; NTCP: sodium taurocholate cotransporting polypeptide. DOI: 10.7554/eLife.00049.011 mRNA level was reduced to ∼30% in tsNTCP siRNA-transfected cells ( Figure 4A , upper-left), total HDV RNA copies were markedly reduced in these cells comparing to those transfected with control siRNA. We further quantified the HDV genome and antigenome RNA copies using strand-specific reverse transcription followed by quantitative real-time polymerase chain reaction (qPCR). The HDV antigenome is a circular replication intermediate that is complementary to the genome. It is not present in the inoculum and only appears in infected cells (Chen et al., 1986) . As shown in Figure 4A (upper-middle panel), both HDV genomic and antigenomic RNA copies were greatly reduced in cells transfected with tsNTCP-specific siRNA but not the control siRNA, indicating that tsNTCP is required for de novo HDV infection. By contrast, lenti-VSV-G virus infection, for which viral entry is mediated by glycoprotein protein G of VSV, was not affected in the tsNTCP-and siRNA-transfected cells ( Figure 4A , upper-right). These data demonstrate that HDV viral entry requires NTCP. As HDV is enveloped by HBV envelope proteins and can only infect target cells in a single round in the absence of HBV, these data support that tsNTCP functions at entry level for viral infection mediated by HBV envelope proteins. We then tested HBV infection on tsNTCP knockdown PTHs. Infection with HBV can be assessed by measuring secreted viral antigens HBV S antigen (HBsAg) and HBV e antigen (HBeAg). HBV inocula may contain residual HBsAg that can release and interfere with the detection of newly synthesized HBsAg during the first few days of infection. To differentiate de novo HBsAg synthesis from the contaminating inoculum, we assayed HBsAg secretion over time from days 6 to 12 after infection with the culture medium changed every 2-3 days. In addition, the kinetics of production of HBeAg with minimal or no residuals in the inoculum was also examined in the same time course experiment. As shown in Figure 4A (lower-left), both HBsAg and HBeAg levels were markedly reduced by transfection of tsNTCP-specific but not a control siRNA at all three time points tested, demonstrating that tsNTCP expression is required for bona fide HBV infection. To confirm that tsNTCP functions at the viral entry level for HBV as it does for HDV, we tested AAV8-HBV virus infection on tsNTCP knockdown PTHs. AAV8-HBV is a recombinant adenovirus-associated virus containing a 1.05× overlength HBV genome, for which viral entry is mediated by AAV8 capsid instead of HBV envelope proteins. AAV8-HBV infection of PTHs can nevertheless transduce the HBV genome into cells and lead to subsequent HBV viral antigen expression. NTCP knockdown did not affect AAV8-HBV infection in PTHs, as shown by the kinetics of HBeAg ( Figure 4A , lower-right). This result shows that NTCP has no effect on post-entry steps of HBV infection. We next examined the effect of silencing human NTCP on HBV and HDV infections in human hepatocytes. Human hepatoma cell line HepaRG is the only cell line known to date to be susceptible to HBV and HDV infections upon differentiation into a mixture of hepatocyte-like and biliary-like cells (Gripon et al., 2002) . HepaRG differentiation requires a lengthy cell culture procedure, including maintaining undifferentiated cells for 2 weeks before induction, followed by induction with corticoids and DMSO for another 2-4 weeks (Gripon et al., 2002) . The NTCP mRNA level was low in HepaRG cells before induction when examined on days 5 and 10 after initial plating, but increased dramatically when the cells differentiated after induction (Figure 4B, . To examine if the acquired hNTCP expression on differentiated HepaRG cells is required for HDV and HBV infections, the cells were transfected with siRNAs targeting hNTCP. About 70% HDV infection was reduced by hNTCP knockdown as indicated by decreased levels of HDV viral RNAs ( Figure 4B , upper-right). Similarly, HBV infection was also inhibited as indicated by significantly reduced HBeAg at multiple time points ( Figure 4B , lower-left), as well as viral RNAs including the 3.5 kb RNA for HBV pre-C and pregenome RNA (pgRNA) and HBV total RNA ( Figure 4B , lower-right) quantified at the end of the experiment. We further validated the critical role of hNTCP on HBV infection in PHHs, the natural host of the virus. Consistently, knockdown of hNTCP significantly reduced HBV infection, which was correlated with the NTCP mRNA knockdown efficiency. Both viral antigens and viral RNAs were decreased in cells transfected with hNTCP-specific siRNAs but not with the control siRNA ( Figure 4C ). Taken together, these data demonstrate NTCP as a common key cellular receptor component necessary for HBV and HDV infections of hepatocytes. We then investigated the ability of NTCP to render nonsusceptible cells susceptible to viral infection. NTCP mRNA expression is low in human hepatocarcinoma cell lines that are not susceptible to HBV or HDV infection. The levels of NTCP mRNA in Huh-7 and HepG2 cells were about 10,000 times lower than that in primary human and Tupaia hepatocytes ( Figure 5A ). We first examined if NTCP expression renders Huh-7 susceptible to HDV infection. Human NTCP-transfected Huh-7 cells supported HDV infection with an efficiency comparable to that of PTHs; nearly 10% of cells were infected as shown by staining of the HDV delta antigen that mainly locates in cell nuclei, whereas Huh-7 cells transfected with a vector plasmid allowed no HDV infection ( Figure 5B ). Moreover, the infection could be blocked by known HBV entry inhibitors, such as pre-S1 lipopeptide and hepatitis B immune globulin (HBIG), demonstrating a genuine infection of HDV mediated by HBV envelope proteins on these cells ( Figure 5C ). HDV RNAs, including antigenomic RNA that is only produced during HDV replication, rapidly increased over time in the infected cells ( Figure 5D ). Moreover, the infection efficiency correlated with both the inoculation dose of HDV ( Figure 5E ) and the expression level of hNTCP ( Figure 5F ). HDV also infected HepG2 cells transiently transfected with hNTCP ( Figure 5 -figure supplement 1) as well as a cell line established by G418 selection of HepG2 cells after hNTCP transfection, which expresses hNTCP stably and could be readily stained by the FITC-pre-S1 peptide ( Figure More efficient HBV infection was achieved on stable HepG2-hNTCP cells with about 5-10% of the cells being infected as revealed by intracellular staining of HBsAg, whereas there was no HBV infection in the parental HepG2 cells ( Figure 6A ). HBV infection in the HepG2-hNTCP stable cells was further evidenced by the continuously increased production of HBeAg during the testing period. HBV entry inhibitors, in particular the Myr-59 peptide and 17B9, efficiently inhibited the infection ( Figure 6B ). The infection efficiency as evidenced by HBV total and 3.5 kb RNA levels correlated with the inoculation dose. Moreover, the formation of HBV covalently closed circular DNA (cccDNA), which is a replicative intermediate and transcriptional template for production of viral RNAs, was confirmed by Southern blot analysis ( Figure 6D ). To further demonstrate that the replicative intermediates of HBV were synthesized de novo in HepG2-hNTCP cells after infection, we performed additional time course experiments. HBV viral replicative intermediates, including cccDNA, the 3.5 kb HBV RNA, as well as the total HBV RNA in the HBV-infected HepG2-hNTCP cells were quantified at different time points. The cccDNA became detectable at 24 hr post-infection. It markedly increased at day 3 post-infection and maintained a relatively stable level for the rest of the time points examined, whereas the formation of HBV cccDNA was completely abolished if entry inhibitor Myr-59 was included with the initial virus inoculation ( Figure 6E) . Consistently, HBV RNA levels of the 3.5 kb transcript and the total HBV viruses AAV8-HBV and Lenti-VSV-G. For HDV and HBV, PTHs were infected at 500 and 100 genome equivalent copies per cell, respectively. The level of HDV viral RNAs in infected cells was quantified by qRT-PCR on 6 dpi. Strand-specific primers were used to differentiate the HDV genomic and antigenomic RNAs (see 'Materials and methods'). For VSV-G control virus infection, recombinant lentivirus pseudotyped by VSV-G carrying a luciferase reporter was inoculated to PTHs 3 days after siRNA transfection. The luciferase activity was assessed on 6 dpi. For HBV infection, the kinetics of secreted viral antigens HBsAg and HBeAg were measured by ELISA. The medium was changed every 3 days. For AAV8-HBV infection, PTHs were infected with a recombinant AAV8 carrying 1.05× overlength HBV genome. Secreted HBeAg was assessed on indicated days post-infection. The effect of tsNTCP silencing in all viral infections was independently evaluated with a total of four siRNAs against tsNTCP (see 'Materials and methods'). The data shown are the result of a representative siRNA out of the four tested. (B) Differentiated HepaRG cells express high level of NTCP mRNA and knockdown NTCP in these cells inhibited HDV and HBV infections. HDV and HBV infection of siRNA-transfected HepaRG cells was conducted similarly as in panel A. HDV RNA levels in the infected cells were measured on 9 dpi. For HBV infection, secreted HBeAg was collected every 2 days as indicated and analyzed by ELISA. The copy numbers of HBV total RNA and 3.5 kb RNA in the infected cells were measured at the end of the experiment, 10 dpi. (C) Knockdown hNTCP in PHHs hampered HBV infection. Frozen PHHs were thawed and plated 1 day before transfecting with siRNAs against hNTCP or a control siRNA. Similar to panels A and B, 3 days after transfection, PHHs were inoculated with 100 genome equivalent copies of HBV per cell, and the levels of secreted HBeAg were determined at indicated dpi. HBV RNAs were quantified at the end of the experiment, 9 dpi. The knockdown efficiency of siRNA targeting tsNTCP or hNTCP shown in panels A-C was determined by real time RT-PCR on day 4 after transfection. NTCP: sodium taurocholate cotransporting polypeptide; HBV: hepatitis B virus; HDV: hepatitis D virus; PTH: primary Tupaia hepatocytes; tsNTCP: Tupaia NTCP; siRNA: small interfering RNA; dpi: days post-infection; hNTCP: human NTCP. DOI: 10.7554/eLife.00049.012 The Huh-7 was used to normalize the relative expression levels in other cells. (B) 1 × 10 5 Huh-7 cells were transfected with 100 ng hNTCP/pcDNA6 or a vector control in 24-well plate and maintained in PMM, 24 hr after transfection, transfected cells were infected with HDV at 500 genome equivalent copies per cell. On 8 dpi, HDV delta antigen, which typically locates in nuclei, was stained with 4G5 antibody in green, nuclei were stained with DAPI in blue. (C) Huh-7 cells transfected with hNTCP were infected with HDV similarly as in panel B in the presence or absence of HBV entry inhibitors: HBIG (hepatitis B immune globulin), Myr-59, and anti-HBsAg mAb, 17B9. 4G5 was used as an antibody control. HDV RNA copies of infected cells were quantified by real-time RT-PCR on 6 dpi. (D) Huh-7 cells transfected with hNTCP were infected with HDV similarly as in panel B. The HDV viral RNAs in infected cells at indicated time points were quantified by real-time RT-PCR. (E) HDV infection with increasing multiplicities of genome equivalents (mge). With 100 ng hNTCP/pcDNA6, 1 × 10 5 Figure 5 . Continued on next page transcripts in the infected HepG2-hNTCP cells gradually increased during first several days of infection and reached a steady level after day 5 ( Figure 6F ). Together these data show that NTCP contributes substantially to HBV infection. We next compared the efficiency of HBV infection in HepG2-hNTCP cells with that in PHHs. As shown by intracellular staining of HBV core antigen (HBcAg) on day 8 post-infection, about 10% HepG2-NTCP cells were infected at multiplicities of genome equivalents (mge) of 100, which is comparable to the efficiency of HBV infection of PHHs ( Figure 6-figure supplement 2) . In contrast to PHHs, HepG2-NTCP cells propagate in cultures, thus the actual infection efficiency of HepG2-NTCP cells may be more likely than not underestimated by the observed end-point HBcAg staining. We also compared the levels of secreted HBeAg and intracellular viral RNAs in these two types of cells infected with three inoculation doses. The level of secreted HBeAg from HepG2-NTCP appeared to be higher than that in PHHs from two donors, whereas the levels of viral RNAs per nanogram of total cell RNA in both infected cell types are comparable ( Figure 6-figure supplement 3) . This may be partially explained by their different abilities in propagation and supporting viral replication and protein expression that would require more detailed studies. Efficient HBV infection of PHHs or HepaRG cells in vitro normally requires high dose of virus inoculums, and only limited progeny viruses are produced after infection (Gripon et al., 1988; Gripon et al., 2002; Boehm et al., 2005) . To assess viral particles released from HBV-infected HepG2-NTCP cells, we first quantified viral DNA in the medium collected at different time points after the infection. As indicated by drastic decline of viral DNA level on day 4 post-infection, the majority of residual viruses from inocula were removed by changing the medium and washing during the first few days of infection. The levels of viral DNA in the media resulted from the ongoing infection during days 4-13 post-infection were low (equivalent to ∼1% of input viral DNA copies) despite significant amount of HBeAg secretion during this period. Similarly, only low levels of viral DNA were detected in the medium from HBV-infected PHHs ( Figure 6-figure supplement 4, right) . It is reasonable to speculate that some host factors that are lacking in cell cultures might be needed for efficient viral particles formation or releasing; or some cellular factors in cultures may hinder these processes during infection. The culture medium collected from infected HepG2-NTCP cells was subsequently tested for infection of PHHs. In line with the low HBV viral DNA level in the medium inoculum, very low number of intracellular HBV total RNA copies were detected in PHHs on day 13 post-infection ( Figure 6-figure supplement 5) , indicating that only very limited HBV infection might have occurred, which may be attributed to the low multiplicity of infection. Residues 157 to 165 of hNTCP are critical for pre-S1 binding and viral infections We finally investigated the molecular determinants of NTCP for HBV and HDV infections. Crab-eating monkey (Macaca fascicularis) NTCP (mkNTCP) shares high protein sequence identity with hNTCP (96.3%) (Figure 7-figure supplement 1) . However, mkNTCP neither supports HDV infection nor pre-S1 peptide (Myr-59) binding ( Figure 7A) , consistent with the known narrow species specificity of the We then made a series of human NTCP variants to cover all the different amino acids between hNTCP and mkNTCP. In each variant, two or a few residues were mutated to their mkNTCP counterparts (Figure 7-figure supplement 1) . Whereas most mutations did not significantly interfere with Myr-59 binding or HDV infection, alteration of five residues of hNTCP between aa 157-165 and its monkey counterpart (from KGIVISLVL to GRIILSLVP, distinct residues are underlined) completely abolished Myr-59 binding and the ability to support HDV infection. Remarkably, replacing the motif of aa 167-156 in mkNTCP with the corresponding human residues converted mkNTCP to an efficient receptor for HDV infection ( Figure 7A) . All the NTCP variants tested were examined for NTCP expression, and comparable levels of cell surface expression were confirmed ( Figure 7B ). Similar to HDV, HBV infection was also abolished on HepG2 cells expressing hNTCP carrying monkey-like mutations, GRIILSLVP, while mkNTCP-bearing human residues KGIVISLVL within the motif of aa 157-165 largely restored HBV infection ( Figure 7C) . These data show that residues between 157 and 165 of NTCP are crucial for binding to the receptor-binding region of the pre-S1 domain of the L protein of HBV, and critically contribute to NTCP-mediated HBV and HDV infections. In this study, by employing a unique approach of tandem affinity purification combined with MS analysis against a Tupaia hepatocyte proteome database established by deep sequencing, we revealed that the liver bile acid transporter, NTCP, specifically interacts with a key region in the pre-S1 domain of the HBV envelope L protein. By performing a series of virological analyses, we showed that silencing NTCP expression markedly inhibited viral infection of HBV and HDV in Tupaia as well as human hepatocytes. Exogenous expression of NTCP rendered nonsusceptible human hepatoma cells susceptible to the viral infections. The authentic viral infections in cells complemented with NTCP were shown by the kinetic analyses of several markers of viral infections, in particular the quantification of newly synthesized viral replicative intermediates. Moreover, the NTCP-rendered infections were blocked by known entry inhibitors. NTCP residues 157 to 165 were identified to be critical for pre-S1 binding and viral infections. These data clearly demonstrate that NTCP is a functional receptor for both HBV and HDV. Identification of cellular receptor(s) of HBV and HDV has been challenging. In our study, we utilized a short peptide ligand, WT b , which was originated from the known receptor-binding domain of the L protein (Barrera et al., 2005; Glebe et al., 2005; Gripon et al., 2005; Engelke et al., 2006; Schulze et al., 2010) , but with specially designed properties suitable for photo-cross-linking and tandem purification. form) at the mutated positions of NTCP are shown for hNTCP, crab-eating monkey NTCP (mkNTCP), and tsNTCP. Huh-7 cells were transfected with plasmids encoding tsNTCP, hNTCP, mkNTCP, or NTCP mutants as indicated. The mutant NTCPs include hNTCP-bearing mutations of mkNTCP residues and mkNTCP-bearing mutations of human residues at indicated positions. The transfected cells were maintained in PMM for 24 hr and then either stained with 200 nM FITC-pre-S1 or infected with 500 mge HDV. HDV delta antigen in infected cells was detected with mAb 4G5 on 7 dpi. Replacing aa 157-165 of mkNTCP with human counterpart rendered mkNTCP an efficient receptor for pre-S1 binding and HDV infection. (B) All NTCP variants expressed comparable levels of NTCP. Huh-7 cells transfected as in panel A were biotinylated 24 hr after the transfection, then lysed and analyzed for cell surface NTCP expression (top), total NTCP expression (middle), and GAPDH (bottom), respectively. For cell surface expression, cell lysates were pulled down with streptavidin T1 Dynabeads and subsequently examined by western blot with mAb 1D4 recognizing a C9 tag at the C-terminus of each NTCP variant. For total NTCP expression, cell lysates were directly subjected to SDS-PAGE, followed by Western blot analysis with 1D4. (C) Effects of NTCP mutations on HBV infection. HepG2 cells were transfected with plasmids encoding hNTCP, mkNTCP, or hNTCP variants bearing the indicated monkey residues, or mkNTCP variants with the indicated human residues. Transfected cells were maintained in PMM for 24 hr, and subsequently infected with HBV at 100 mge. HBeAg and HBV 3.5 kb RNA were assayed on 6 dpi. Similar to panel B, comparable NTCP surface expression levels in the transfected HepG2 cells were confirmed for all the NTCP variants tested (Figure 7-figure supplement 2 Figure 7C were analyzed for total or cell surface NTCP expression at 24 hr post-transfection as described in Figure 7B . DOI: 10.7554/eLife.00049.025 Two photo-leucines were incorporated into the critical receptor-binding region of WT b without interfering with its receptor-binding activity, which allowed highly specific zero distance cross-linking of its direct binding partner(s) but not other neighboring molecules. A biotin moiety of WT b facilitated purification of the complex of WT b and its binding partner(s) by streptavidin beads. An mAb, 2D3, was developed to recognize WT b on an epitope outside the receptor-binding site, serving as a highly specific tool for detection as well as additional affinity purification of the complex. Thus, the binding partner(s) was first cross-linked by WT b , and then purified by using streptavidin and 2D3 beads in tandem. The covalent interaction between the WT b ligand and its partner(s) enabled a purification process under high-stringency conditions, and efficient isolation was achieved irrespective of the nature of the binding partner(s) even if it is a membrane protein(s) with multiple transmembrane domains, like NTCP identified here. NTCP (Slc10a1) is the founding member of the SLC10 family of solute carrier proteins. It is a hepatic Na + bile acid symporter and is responsible for cotransportation of sodium and bile acids across cellular membranes to maintain the enterohepatic circulation of bile acids Stieger, 2011) . NTCP is a multiple transmembrane glycoprotein presumed to span the cellular membrane up to 10 times with small extracellular loops (Mareninova et al., 2005; Hu et al., 2011) . It is mainly expressed in the liver (Stieger, 2011) , consistent with the liver tropism of HBV and HDV. NTCP localizes to the sinusoidal (basolateral) plasma membrane of hepatocytes (Stieger et al., 1994) , a location that fits well with its receptor role for blood-borne HBV and HDV. Whereas HBV first attaches to hepatocytes mainly through heparan sulfate (Schulze et al., 2007; Leistner et al., 2008) , our data demonstrate that the interaction between NTCP and L protein of HBV is highly specific, and NTCP is crucial for productive viral entry of hepatocytes. Consistent with previous reports on primary cultures of rat hepatocytes (Liang et al., 1993; Rippin et al., 2001) , NTCP expression rapidly decreased over time in cultured PTHs after isolation. This may at least partially explain the observations that primary hepatocytes typically remain susceptible to HBV infections in vitro for only a few days after isolation from liver tissues (Gripon et al., 1988; Seeger et al., 2007) . NTCP is functionally conserved in mammalians, but protein sequences of NTCP vary among species, which is likely to contribute to the narrow species tropism of viral infection. Strikingly, despite the high level of protein sequence homology between human and monkey NTCP, the later did not support HBV and HDV infection. Replacing a small motif of aa 157-165 of mkNTCP with the corresponding hNTCP residues converted mkNTCP to a receptor for pre-S1 binding as well as HDV and HBV infection. Further studies are warranted to determine if and how NTCP contributes to the species specificity of HDV and HBV infection in other species. It also remains to be determined if other molecule(s) additional to NTCP contributes to the cellular entry of HBV and/or HDV as a coreceptor(s) or receptor component(s), and if other host factors such as the microenvironment or architecture of hepatocytes in liver, or soluble blood components like those that have been shown to involve in infections of other viruses (Shayakhmetov et al., 2005; Morizono et al., 2011) , contribute to HBV and/or HDV infection. Expression and subcellular distribution of NTCP are precisely regulated under physiological conditions. NTCP accounts for most, if not all, hepatic Na + -dependent bile acid transport (Stieger, 2011) . NTCP expression is low and inversely correlated with the degree of dedifferentiation of cancer cells in human hepatocellular carcinoma (Kullak-Ublick et al., 1997; Zollner et al., 2005) and the severity of HBV-related liver cirrhosis (Lee and Kim, 2007) . The newly discovered role of NTCP as an entry receptor for HBV and HDV raises interesting questions regarding its involvement in viral pathogenesis. Identification of NTCP as a functional receptor for HBV and HDV advances our understanding of their entry into host cells and may lead to new prevention and treatment strategies against these viruses and related diseases. Adult tree shrews (Tupaia belangeri chinensis) were housed in a Tupaia animal facility at the National Institute of Biological Science, Beijing. All studies were performed in accordance with institutionally approved protocols and adherent to guidelines of the National Institute of Biological Sciences Guide for the care and use of laboratory animals. PTH cells were obtained from anesthetized Tupaia (100-150 g) with a two-step perfusion method as previously described (Walter et al., 1996) . Cell suspensions after perfusion were filtered through a 70-μm cell strainer and centrifuged at 50 g for 3 min. The cell pellet containing PTHs was resuspended in plating medium of Williams E medium supplemented with 10% FBS, 5 μg/ml transferrin, 5 ng/ml sodium selenite, 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. The cells were then plated on collagen-coated cell culture dishes or plates. 4 hr after plating, medium were changed to primary hepatocytes maintenance medium (PMM), that is, Williams E medium supplemented with 5 μg/ml transferrin, 10 ng/ml EGF, 3 μg/ml insulin, 2 mM L-glutamine, 18 μg/ml hydrocortisone, 40 ng/ml dexamethasone, 5 ng/ml sodium selenite, 2% DMSO, 100 U/ml penicillin, and 100 μg/ml streptomycin. Cells were maintained in 5% CO 2 humidified incubator at 37°C with regular medium change every 2-3 days. PHHs were purchased from Becton Dickinson (United States) or Shanghai RILD Inc. (Shanghai, China). The cells were cultured similarly as PTHs using the same plating medium and maintaining PMM medium as described above. Human embryonic kidney cell lines 293 and 293T, human cervix carcinoma cell line Hela, and human hepatocellular carcinoma cell line HepG2 were from American Type Culture Collection (ATCC); human hepatocellular carcinoma cell lines Huh-7, SMMC-7721 (SMMC), and Bel-7404 (BEL) were from the Cell Bank of Type Culture Collection, Chinese Academy of Sciences. The cells were cultured with Dulbecco's Modification of Eagle's Medium (DMEM; Invitrogen, United States) supplemented with 10% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin at 37°C in 5% CO 2 humidified incubator except otherwise indicated. HepaRG cells were purchased from Biopredic International (Rennes, France) and were cultured following the product manual. Differentiated HepaRG cells were obtained following a two-step procedure as described by Gripon et al. (2002) . A plasmid containing a head to tail trimer of 1.0× HDV cDNA of a genotype I virus (Genebank accession number: AF425644.1) under the control of a CMV promoter was constructed with de novo synthesized HDV cDNA for the production of HDV RNPs. A pUC18 plasmid containing nucleotide 2431-1990 of HBV (Genotype D, Genebank accession number: U95551.1), or the same plasmid bearing mutation generated by site-directed mutagenesis, was used for expressing HBV envelope proteins under the control of endogenous HBV promoter. HDV virions were produced by transfection of the plasmids in Huh-7 as previously described by Sureau et al. (1992) . HBV HBV genotype B virus was obtained by ultracentrifugation of plasma from an HBV chronic carrier with written consent. HBV genotype D virus was produced by transfection of Huh-7 cells with a plasmid containing 1.05 copies of HBV genome under the control of a CMV promoter similarly as previously described by Blanchet and Sureau (2006) . The Genebank accession numbers for the viruses are JX978431 and U95501.1, respectively. Recombinant adeno-associated virus 8 (AAV8) carrying 1.05 copies of HBV genome was produced similarly as previously described (Xiao et al., 1998) by cotransfection of 293 cells with plasmids for AAV8 packaging, 1.05× overlength HBV genome (genotype D) and adenovirus helper. An HIV-1 genome-based lentivirus pseudotyped by glycoprotein of vesicular stomatitis virus and carrying a firefly luciferase reporter gene was produced by cotransfection of 293T cells with plasmids for VSV-G expression, HIV genome packaging, and luciferase reporter, respectively, as described (Sui et al., 2005) . Virus-related experiments were conducted in a BSL-2 facility at the National Institute of Biological Sciences, Beijing. Peptides with nonnatural amino acid L-2-amino-4,4-azi-pentanoic acid (L-photo-leucine) were synthesized by American Peptide Company Inc. (United States). Other peptides corresponding to the N-terminal of pre-S1 domain of HBV L protein (genotype C, strain S472, GeneBank EU554535.1) were synthesized by SunLight Peptides (Beijing, China). Mouse monoclonal antibodies (mAb) 2D3, 1C10, and 4G5 were generated in the laboratory; all are of IgG1 isotype. 2D3 specifically recognizes the 19-33 amino acids of the pre-S1 domain of HBV L protein; 1C10 recognizes HBcAg; 4G5 targets HDV delta antigen. 17B9, a mouse mAb-specific to HBV S protein, was provided by Dr. Lin Jiang, China National Biotec Group. Hepatitis B immune globulin (HBIG) was from the National Institutes for Food and Drug control, Beijing, China. 2D3 magnetic beads were prepared by covalently cross-linking 2D3 to Dynabeads M-270 Epoxy following manufacturer's instructions. Secondary antibodies for immunofluorescence staining and Western blot were purchased from Invitrogen or Sigma-Aldrich (United States). ELISA kits for HBsAg and HBeAg measurement were purchased from Wantai Pharm Inc. (Beijing, China). SYBR Premix Ex Taq quantitative real-time PCR kit and Reverse Transcriptase (RT) kit were from Takara Inc. (Beijing, China). Streptavidin-coupled magnetic beads (Dynabeads MyOne Streptavidin T1) and magnetic beads coated in glycidyl ether (Epoxy) groups (Dynabeads M-270 Epoxy) were purchased from Invitrogen. Other reagents were purchased from New England Biolabs (United States), Life Technologies (United States), or Sigma-Aldrich. HBV viral antigens HBsAg and HBeAg were examined using 50 μl supernatants with commercial ELISA Kits (Wantai Pharmacy, Beijing, China) following manufacturer's instructions. In most cases, HBsAg level was normalized with WHO HBsAg reference serum (kindly provided by Dr. Zhenglun Liang from the National Institutes for Food and Drug control, Beijing, China) and presented as international units per milliliter. Quantification of HDV total RNA (genome equivalent) copies and HBV genome equivalent copies HDV Viral RNA was isolated with Trizol reagent following manufacturer's instructions. Total RNA was reverse transcribed into cDNA with random primers (PrimeScript RT kit; Takara) and 2 μl of the cDNA was used for real-time PCR assay. Primers for quantifying HDV total RNA or genome equivalent copies are complemented with the delta antigen coding region of HDV RNA genome: forward primer HDV-1184F, 5′-TCTTCCTCGGTCAACCTCTT-3′, and backward primer HDV-1307R, 5′-ACAAGGAGAGGCAGGATCAC-3′. HBV Viral DNA was isolated by standard genomic DNA isolation method. The DNA was quantified using specific primers: 5′-GAGTGTGGATTCGCACTCC-3′ (forward) and 5′-GAGGCGAGGGAGTTCTTCT-3′ (backward) by real-time PCR. The viral genome equivalent copies were calculated based on a standard curve generated with known copy numbers. Real-time PCRs were performed using SYBR Premix Ex Taq kit on an ABI Fast 7500 real-time system instrument (Applied Biosystems, United States). With 5 × 10 7 copies of genome equivalent HDV, 1 × 10 5 of target cells were incubated at 16°C for 4 hr in the presence of 4% PEG8000, followed by extensive wash with cold PBS for four times. The cells were then lysed directly with Trizol reagent and followed by reverse transcription. RNA copy numbers of viral genome and internal control glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA were determined by real-time PCR. For HDV binding inhibition assay, peptides were pre-incubated with target cells at 16°C for 1 hr before incubating with the virus; antibodies against viral envelope protein were pre-incubated with viruses before adding to target cells. Viral infections of HDV and HBV were conducted in 48-well plates at multiplicities of genome equivalents of 500 and 100, respectively. Normally, 5 × 10 7 copies of genome equivalent HDV or 1 × 10 7 copies of genome equivalent HBV were inoculated in the presence or absence of entry inhibitors with 1 × 10 5 cells and incubated for 16 hr except otherwise indicated. Cells were then washed with medium for three times and maintained in PMM medium with medium change every 2-3 days. For HDV infection, 4% PEG 8000 was present during the 16 hours viral inoculation period similarly as described by Barrera et al. (2004) . Viral infection at different time points was analyzed by measuring viral DNA/ RNAs and viral antigen expression. Quantitative real time RT-PCR was used to quantify HDV total RNAs, strand-specific real time RT-PCR to determine copies of HDV genome and antigenome RNA (see below). For HBV viral infection on HepaRG cells and PHH, ∼ 4% PEG 8000 was present during the inoculation period as previously described by Gripon et al. (Gripon et al., 1993; Gripon et al., 2002; Schulze et al., 2007) . Viral infection of PTH was conducted in the absence of PEG8000. Culture medium was changed every 2-3 days. Secreted HBsAg and/or HBeAg were determined with commercial ELISA kits. Real-time PCR, with or without a prior reverse transcription step, was used for quantification of HBV-specific 3.5 kb pre-C and pregenomic RNA, total HBV sub-genomic RNA, and HBV cccDNA copies. The strand-specific qRT-PCR was performed as previously described by Freitas et al. (2012) . Briefly, the genomic and antigenomic RNAs were reverse transcribed separately with strand-specific primers into cDNAs: primer HDV398R (5′-CGCTTCGGTCTCCTCTAACT-3′) for genomic RNA; primer HDV288F (5′-GCAGACAAATCACCTCCAGA-3′) for antigenomic RNA. The reverse transcribed cDNAs of genmomic or antigenmoic HDV RNAs were used as templates for real-time PCR using the HDV398R and HDV288F primer pair. TaqMan probe was 5′ FAM-AGAGCTCTGACGCGCGAGGAGTAAGC-TAMRA 3′. Real-time PCR assays were conducted with an ABI Fast 7500 real-time PCR instrument. Total RNA from HBV-infected cells was isolated with Trizol reagent (Invitrogen). About 400 ng total RNA was reverse transcribed into cDNA with PrimeScript RT kit (Takara) in a 10 μl reaction. cDNA derived from 20 ng total RNA was used as template for real-time PCR amplification. In a separate realtime PCR reaction, 20 ng of total RNA was directly used as template to assess the possible HBV viral DNA contamination in the RNA preparation. Primers (HBV2270F: 5′-GAGTGTGGATTCGCACTCC-3′) and (HBV2392R: 5′-GAGGCGAGGGAGTTCTTCT-3′) were used for HBV 3.5 kb transcripts; (HBV1805F: 5′-TCACCAGCACCATGCAAC-3′) and (HBV1896R: 5′-AAGCCACCCAAGGCACAG-3′) were for total HBV-specific transcripts. Amplification of 123-bp fragment for 3.5 kb transcripts and 92-bp product for total HBV-specific transcripts were both conducted by denaturation at 95°C for 30 s, followed by 40 cycles of 95°C denaturation for 3 s, and 60°C annealing/elongation for 30 s. Real-time PCR was performed using SYBR Premix Ex Taq kit on an ABI Fast 7500 real-time system instrument. Real-time PCR using either set of the primers generated highly specific amplification product. HBV RNA copy numbers were deduced from a standard curve generated from known nucleic acid quantities. Then the HBV RNA copy number per nanogram RNA in the infected cell cultures was calculated by subtracting the background amplification noise derived from the viral DNA contamination in the RNA preparation from that cDNA amplification. The signal-to-noise ratio for HBV total transcripts is usually ≥50, and for 3.5 kb transcripts ≥20. The real-time PCR detection limits for total HBV-specific transcripts and 3.5 kb transcripts are ∼0.5 and ∼3.5 copies per nanogram cellular total RNA, respectively. HBV cccDNA Southern blot was conducted following a similar procedure as described by Summers et al. (1990) with modifications. Briefly, to selectively extract HBV cccDNA, infected hepG2-NTCP cells in 6-cm dishes were lysed with 1 ml lysis buffer at 37°C for 60 min, followed by addition of 0.25 ml of 2.5 M KCl and incubation at 4°C overnight. The lysis buffer was not supplemented with proteinase K, containing 50 mM Tris-HCl, pH 7.4, 10 mM EDTA, 150 mM NaCl, 1% SDS. The lysate was then clarified by centrifugation at 12,000 g for 30 min at 4°C and extracted with phenol and phenol:chloroform. DNA was precipitated with equal volume of isopropanol in the presence of 20 µg glycogen (Roche) and finally dissolved in TE buffer. The prepared DNA sample was then treated with plasmid-safe adenosine triphosphate (ATP)-dependent deoxyribonuclease DNase (Epicentre Technologies) following manufacturer's instructions. For Southern blotting, the plasmid-safe DNase-treated DNA was separated on a 1.3% agarose gel and then transferred to a nylon membrane (Hybond-N + ; Amersham) using a standard neutral transfer procedure. A 3280-bp plasmid constructed by inserting a 588-bp HBV DNA fragment (from 1805 to 2392, genotype D, Southern blot probe) into a 2692 bp pMD18T vector (Takara) was also run on the same agarose gel to serve as the molecular marker for cccDNA in Southern blot analysis. The plasmid is of similar size of HBV genome and was mainly in supercoiled form; therefore it runs at similar size as HBV cccDNA in agarose gel. The nylon membrane was hybridized with a [α-32 P] dCTP-labeled HBV probe (genotype D HBV DNA fragment from 1805 to 2392) prepared by random primer DNA labeling kit (Ver.2.0; Takara). Hybridization was carried out in 7 ml of Perfect Hyb Plus Hybridization Buffer (Sigma) with 1 hr pre-hybridization, followed by overnight hybridization at 67°C. The membrane was then washed once with 2× SSC/0.1% SDS, 1× SSC/0.1% SDS, and 0.5× SSC/0.1% SDS at 67°C for 20 min, respectively. Finally, the membrane was subjected to autoradiographic exposure. HBV cccDNA (double-stranded DNA without nick and gap) was quantified by real-time PCR using a protocol as previously described Werle-Lapostolle et al., 2004) with modifications. In particular, specific primers for cccDNA detection reported by Glebe et al. (2003) (ccc-1582F: 5′-TGCACTTCGCTTCACCT-3′; ccc-2316R: 5′-AGGGGCATTTGGTGGTC-3′) were validated and used for quantifying copy numbers of cccDNA using real-time PCR. Viral DNAs other than cccDNA, including single-stranded and relaxed circular DNAs, were degraded prior to amplification by treatment of the DNA templates with plasmid-safe adenosine triphosphate (ATP)dependent deoxyribonuclease DNase (Epicentre Technologies). In brief, HBV-infected cells were lysed for 4 hr at 65°C in lysis buffer (50 mM Tris-HCl, pH 8.0, 50 mM EDTA, 100 mM NaCl, 1% SDS) supplemented with proteinase K (200 μg/ml) and followed by phenol-chloroform extraction. A total of 250 ng of the extracted DNA was digested with 5-10 units plasmid-safe DNase in a 50 μl volume for 8 hr at 37°C followed by DNase inactivation at 70°C for 30 min. 2 μl of the 50 μl reaction was then added to 20 μl of a real-time PCR reaction. Amplification of 735 bp cccDNA product was conducted by denaturation at 95°C for 5 min, followed by 45 cycles of denaturation at 95°C for 30 s, 62°C annealing for 25 s, and 72°C elongation for 45 s. HBV cccDNA copy numbers were calculated with a standard curve from plasmid with known nucleic acid quantities. The detection limit for cccDNA is ∼10 copies cccDNA per reaction (equivalent to 10 ng of total cell lysate DNA). Realtime PCR was performed with SYBR Premix Ex Taq kit on an ABI Fast 7500 real-time system instrument. PTH cDNA library construction, deep sequencing of Tupaia transcriptome, and bioinformatics analysis of Illumina deep sequencing-determined transcriptome Primary Tupaia hepatocytes were isolated as described above. PTH mRNA was purified from 10 μg of total RNA using Oligo-dT magnetic beads. The mRNA was fragmented into small pieces by incubation with divalent cations at 94°C for exactly 5 min. The first strand cDNA was synthesized using random primers and SuperScript II reverse transcriptase (Invitrogen) with fragmented mRNAs. RNA template was then removed by RNase H, and double-stranded cDNA was prepared with DNA polymerase I. cDNA with blunt ends was created by T4 DNA polymerase and Klenow DNA polymerase and an 'A' base was subsequently added to the 3′ end of the blunt phosphorylated DNA fragments by Klenow fragment (3′ to 5′ exo minus). The cDNA was then ligated with adapters and then ran on a 2% agarose gel. The fragments with a size range from 200 ± 25 bp were purified, followed by amplification using the manufacturer's primers. The PCR products were then purified using QIAquick PCR purification Kit (Qiagen), quantified and diluted for cluster generation and deep sequencing. The 72-cycle pair-end sequencing was performed with Sequencing Kits (Version 5) on an Illumina Genome Analyzer IIx (Illumina, San Diego, United States). Illumina CASAVA pipeline v1.8.1 was used for sequence extraction and filtering. Bioinformatics analysis of Illumina deep sequencing-determined transcriptome of primary Tupaia hepatocytes De novo reconstruction of transcriptome from cDNA library deep sequencing data Total 253,919,616-pair 72 nt sequences with 36.6G base from the sequencing results of the hepatocyte cDNA library described above were fed to Trinity (Grabherr et al., 2011) r20110519 using pair end RNA-seq protocol, with which 209,063 transcripts of average length of 1421 nt (minimum 300 nt, maximum 21,043 nt, and scaffold N50 of 3674 nt) were generated. Following assembly, GENSCAN (Burge and Karlin, 1997) was used with default parameters to identify coding sequences and the encoded protein sequences of these transcripts. A total of 91,479 protein sequences were identified. Each chosen protein sequence was first annotated with its corresponding blastp (Camacho et al., 2009 ) matches from NCBI human protein sequences. Those not annotated in the first step were then submitted for similar annotation process with UniprotKB human proteome and NCBI nonredundant protein sequence database. Protein sequences that were not annotated by previous steps were submitted for annotation with their corresponding transcripts. Protein sequences with their corresponding transcripts that can be annotated by the blastx (Camacho et al., 2009) hits of NCBI human protein sequences, UniProtKB human proteome or NCBI nonredundant protein sequence database were annotated with these hits from transcripts. All identified protein sequences were included in the hepatocyte protein sequences database. The protein sequences were labeled with corresponding functional annotation results. Any identified protein sequences that were not successfully annotated are labeled with 'Uncharacterized protein'. Total 50,951 annotated and 40,528 uncharacterized protein sequences generated from the Tupaia hepatocytes transcriptome were combined into the database. A numeric ID (gi) was generated for each protein sequence in the database. The corresponding cDNA sequences were deposited to NCBI Transcriptome Shotgun Assembly (TSA) database of GenBank with accession number from JU120276 to JU170736 after removal of cDNAs shorter than 200 bp and a vector sequence. PANTHER database (Mi et al., 2005) was used to determine the protein class distribution of annotated transcripts and protein sequences in primary Tupaia hepatocytes (PTHs) generated in this study and transcriptome from primary human hepatocytes (PHHs) reported by Hart et al. (2010). Photo-cross-linking with peptide ligand and tandem purification of the target molecule(s) L-photo-leucine-bearing wild-type bait peptide (WT b ) or control bait peptide (N9K b ) was dissolved in DMSO in dark. L-photo-leucine contains a photoactivatable diazirine ring, irradiation of UV light at 365 nm induces a loss of nitrogen of the diazirine ring, and yields a reactive carbene group with short half-life for covalent cross-linking at nearly zero distance. For tandem purification, WT b or N9K b at indicated concentrations was applied to ∼1 × 10 7 hepatocytes plated on collagen-coated dishes. Cells were cross-linked by UV irradiation and then washed to remove residual free peptides and subsequently lysed with 1 ml radioimmunoprecipitation assay (RIPA, pH 7.4) buffer containing 20 mM Tris, 150 mM NaCl, 0.1% SDS; 0.5% sodium deoxycholate, 1% NP40, and 1× protease inhibitor cocktail (Roche). The cell lysates were precipitated with 100 μl streptavidin T1 magnetic beads, eluted with 50 μl nonreducing SDS-PAGE loading buffer and then diluted with cold RIPA buffer to a final volume of 1 ml and was precipitated with 100 μl (1 × 10 8 ) 2D3-conjugated M-270 dynabeads, then eluted with 100 μl nonreducing loading buffer. The elute, with or without PNGase F treatment, was diluted to 1 ml with RIPA buffer and was precipitated again with 100 μl streptavidin T1 magnetic beads, followed by extensive washing, and finally eluted by boiling 5 min with 20 μl SDS-PAGE loading buffer. The samples were then analyzed with 12% SDS-PAGE and silver staining. For photocross-linking analysis of primary cells, cell lines, or NTCP-or control plasmid-transfected Huh-7 or 293T cells, WT b or N9K b bait peptide in the presence or absence of competing peptide was applied to 2 × 10 6 cells, photo-cross-linking was conducted similarly as described above. The cross-linked samples were precipitated with streptavidin T1 magnetic beads and separated by SDS-PAGE, and followed by Western blotting with mAb 2D3 (recognizing bait peptides) or mAb 1D4 against C-terminal tag C9. Silver stained gel bands were cut, followed by in-gel reduction, alkylation, and trypsin digestion as previously described (Shevchenko et al., 2006) . Digested peptide mixtures containing 0.1% formic acid were loaded onto a 4 cm, 75-μm inner diameter fused silica capillary column packed with 10-μm YMC C18 material (YMC, Kyoto, Japan). After desalting, samples were separated with a Waters nano ACQUITY UltraPerformance LC (Waters, United States) and eluted to a LTQ-Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, United States). The UPLC separation gradient included a 30-min gradient from 0% to 30% acetonitrile, followed by a 10-min gradient to 80% acetonitrile, then 10 min of 80% acetonitrile and back to 0% acetonitrile within 5 min. The mass spectrometer was operated in the data-dependent mode. Survey MS scans were acquired in the orbitrap with the resolution set to a value of 60,000. Each survey scan (300-2000 m/z) was followed by four data-dependent CID tandem mass (MS/MS) scans at 35% normalized collision energy and four data-dependent HCD tandem mass (MS/MS) scans at 40% normalized collision energy with 15,000 resolution in orbitrap. AGC target values were 500,000 for the survey scan, 10,000 for the ion trap MS/MS scan, and 50,000 for the orbitrap MS/MS scan. Target ions already selected for MS/MS were dynamically excluded for 30 s. Tandem mass spectra were searched against the Illumina deep sequencing-determined Tupaia hepatocyte protein database that was concatenated with reversed sequences to estimate false positives and was supplemented with the sequence of bait peptide under the Linux operating system using the ProLuCID (Xu et al., 2006) protein database search algorithm with peptide mass tolerance of ±100 ppm, fragment ion mass tolerance of ±400 ppm, half tryptic specificity, and a static modification of 57.0215 on Cys due to carboxyamidomethylation. ProLuCID search results were then filtered with DTA Select 2.0 (Tabb et al., 2002) using a cutoff of 1% for peptide false identification rate (-fp 0.01). Peptides with DeltaMass > 10 ppm (-DM 10) were rejected; the minimum number of peptides to identify a protein was set to 1 (-p 1). Immunofluorescence microscopy and FACS analysis of NTCP binding with pre-S1 peptides Plasmid encoding human, or Tupaia NTCP, with or without a tag, or a control plasmid was transfected into cells. The cells were washed and blocked with 3% BSA 24-36 hr after transfection and then stained with biotin-labeled peptides at 4°C about 1-2 hr, followed by fixation with 4% paraformaldehyde (PFA) for 10 min. Cells were then stained with PE-labeled Streptavidin (eBioscience). In some cases, Myr-59 peptide containing the first 59 residues of pre-S1 domain with an N-terminal myristoylation modification and labeled with FITC (FITC-pre-S1) was applied to cells directly. The cell images were captured with a Nikon Eclipse Ti Fluorescence Microscope or a Zeiss LSM 510 Meta Confocal Microscope. For FACS analysis, the FITC-pre-S1 stained cells without fixation were detached with 0.5 mM EDTA/PBS, washed and resuspended in PBS, and analyzed with a FACS LSRII instrument (BD). Four siRNAs (tsNTCP-si1: 5′-CUAUGUAGGCAUUGUGAUAdTdT-3′, tsNTCP-si2: 5′-GUGUUAUCCU GGUGGUUAUdTdT-3′, tsNTCP-si3: 5′-GGACAUGAAUCUCAGCAUUdTdT-3′, tsNTCP-si4: 5′-GGGCAAGAGCAUCAUGUUUdTdT-3′) specifically targeting Tupaia slc10a1 were designed through siDESIGN Center (www.thermo.com/sidesign). The specificities of these siRNAs were examined by searching against NCBI cDNA databases and the in-house Tupaia hepatocyte transcriptome to ensure they are free of off-target. siRNA with sequence Ctrl-si: 5′-UUCUCCGAACGUGUCACGUdTdT-3′ that is a scramble sequence with no known mammalian target sequence was used as a negative control. 20 nM siRNA were transfected into PTHs 24 h after initial seeding with lipofectamine 2000 (Invitrogen). tsNTCP mRNA level in the siRNA-transfected cells was quantified by real-time RT-PCR 3 days after siRNA transfection. Transfected PTH cells were infected on day 4 after siRNA transfection with HBV, HDV, AAV8-HBV (carrying 1.05 copies of HBV genome), or Lenti-VSV-G-Luc viruses. The secreted viral antigen HBsAg or HBeAg from HBV-and AAV8-HBV-infected cells were examined as indicated. For Lenti-VSV-G-Luc virus-infected cells, luciferase activity was determined on 6 days post-infection. For gene knockdown experiment on HepaRG cells SLC10A1 from HepaRG cells was cloned, and the sequence was deposited to Genebank (accession number: JQ814895). Differentiated HepaRG cells in 48-well plate were transfected using RNAiMax (Invitrogen) with 20 nM of a siRNA pool (Qiagen) containing four specific siRNAs targeting different regions of human SLC10A1 (5′-GGAUCGUCCUCAAAUCCAAdTdT-3′, 5′-GGAGUCAGCCGGAGAACAAdTd T-3′, 5′-GGACAAGGUGCCCUAUAAAdTdT-3′, 5′-GGUGCUAUGAGAAAUUCAAdTdT-3′) or a negative control siRNA (Ctrl-si: 5′-UUCUCCGAACGUGUCACGUdTdT-3′) as indicated. Gene knockdown efficiency was examined 4 days after transfection. Cells were then inoculated for 16 hr with HBV in the presence of 3.6% PEG8000. The secreted HBeAg and the 3.5 kb HBV RNA were determined on indicated days after infection. Frozen PHH cells were thawed and subsequently transfected as that of PTH with human SLC10A1specific siRNA 5′-CACAAGUGCUGUAGAAUUAdTdT-3′ (siR405) or a siRNA pool containing four SLC10A1 specific siRNAs from QIAGEN (5′-GGAUCGUCCUCAAAUCCAAdTdT-3′, 5′-GGAGUCAGCCGGAGAACAAdTdT-3′, 5′-GGACAAGGUGCCCUAUAAAdTdT-3′, 5′-GGUGCUA UGAGAAAUUCAAdTdT-3′). Control siRNA 5′-UUCUCCGAACGUGUCACGUdTdT-3′ (Ctrl-si) has a scramble sequence with no known mammalian target sequence. Total 20 nM siRNA was transfected into ∼1.4 × 10 5 PHH cells per well in 48-well plate 24 h after initial cell seeding; the cells were then inoculated with HBV 72 hr after transfection. hNTCP mRNA level in the Research article siRNA-transfected cells was quantified by qRT-PCR 3 days after siRNA transfection. The secreted antigens HBsAg and HBeAg, and the intracellular HBV RNAs were determined on indicated days after infection. HepG2 or Huh-7 cells were transfected with a plasmid expressing human, treeshrew, monkey NTCP, or an NTCP variant, or a vector control. Stable cell line expressing hNTCP was established by transfection of HepG2 cells with a plasmid encoding hNTCP (hNTCP/pcDNA3.1) followed by selection with 500 μg/ml G418 and maintained in DMEM supplemented with 10% FBS, 500 μg/ml G418, 100 U/ml penicillin, and 100 μg/ml streptomycin. The transfected cells, or HepG2-hNTCP stable cells, were cultured in PMM 24 hr before infection. With 5 × 10 7 genome equivalent copies of HDV, 1 × 10 5 cells were incubated, or otherwise indicated, for 24 hr in the presence or absence of entry inhibitors. HDV inoculation was conducted in the presence of 4% PEG8000. PMM was replenished every 2 days. On indicated days postinfection, cells were treated with 100% methanol for 10 min, followed by incubating with 5 μg/ml FITC-labeled mAb 4G5 for 1 hr at RT to stain HDV delta antigen. The nuclei were stained with 4′-6-diamidino-2-phenylindole (DAPI) before analyzing. HDV viral RNA copies in cell lysates were quantified by qPCR. With ∼1 × 10 7 genome equivalent copies of HBV, 1 × 10 5 cells were inoculated, or indicated otherwise, in the presence of ∼4% PEG8000 as reported for primary human hepatocyte and HepaRG cell (Gripon et al., 1993; Gripon et al., 2002; Schulze et al., 2007) . The cells were maintained subsequently in PMM and the medium was changed every 2-3 days. For immunofluorescence microscopy analysis, HBV-infected cells, with or without replating on glass coverslips for imaging, were fixed with 4% Paraformaldehyde (PFA) and permeabilized with 0.5% TritionX-100, and then stained either with 10 μg/ml 17B9 against HBsAg followed by FITC-labeled goat anti-mouse IgG, or with 5 μg/ml 1C10 against HBcAg followed by Qdot 655 VIVID donkey anti-mouse IgG. 1 μg/ml of DAPI was added to stain the nucleus before analyzing. The cell images were captured with a Nikon Eclipse Ti Fluorescence Microscope or a Zeiss LSM 510 Meta Confocal Microscope. Secreted viral antigens and intracellular viral replication intermediates cccDNA and/or RNAs were examined on indicated days after infection. For cell surface expression, the transfected cells expressing NTCPs or mutants were surfacebiotinylated with sulfo-NHS-LC-biotin (Pierce) following the manufacturer's instruction manual. The biotinylated cells were then lysed in 600 μl of 1× RIPA buffer supplemented with 1× protease inhibitor cocktail (Roche) and then treated with PNGaes F. Total cellular protein in the lysate was determined by using Bio-Rad DC Protein assay. About 300 μg Streptavidin T1 Dynabeads (Invitrogen) were then used to pull down surface-biotinylated proteins in the supernatants containing ∼160 μg of total cellular protein each sample. After extensive washing with 1× RIPA buffer, bound proteins were eluted, and separated by SDS-PAGE and subsequently examined with anti-C9 mAb 1D4 by Western blotting. For total NTCP expression, the same transfected cells expressing wild-type or mutant NTCPs were lysed with 1× RIPA buffer and treated with PNGase F. Each sample containing same amount of total cellular protein (∼8 μg) were loaded for SDS-PAGE followed by Western blotting analysis with mAb 1D4 that recognizes the C9 tag fused at the C-terminus of NTCPs. All experiments were repeated 2-6 times with duplicate or triplicate samples for each condition unless indicated otherwise. A representative result of multiple independent experiments is present (n = 2-6) in each figure. Error bars shown in all figures represent standard deviation of the mean (n = 2-4). Dotted lines show detection limit except otherwise specified. Reporting standards: We followed the reporting standards for the Transcriptome Shotgun Assembly Sequence Database of NCBI, which are: Submitted sequences must be assembled from data experimentally determined by the submitter. Screened for vector contamination and any vector/linker sequence removed. This includes the removal of NextGen sequencing primers. Sequences cannot be less than 200 bp. Sequences should have no more than 10% n's or greater than 14 n's in a row. If the submission is a single-step, unannotated assembly and the output is a BAM file(s) these should be submitted as a TSA project to SRA.

The Evolutionary Pattern of Glycosylation Sites in Influenza Virus (H5N1) Hemagglutinin and Neuraminidase

Two glycoproteins, hemagglutinin (HA) and neuraminidase (NA), on the surface of influenza viruses play crucial roles in transfaunation, membrane fusion and the release of progeny virions. To explore the distribution of N-glycosylation sites (glycosites) in these two glycoproteins, we collected and aligned the amino acid sequences of all the HA and NA subtypes. Two glycosites were located at HA0 cleavage sites and fusion peptides and were strikingly conserved in all HA subtypes, while the remaining glycosites were unique to their subtypes. Two to four conserved glycosites were found in the stalk domain of NA, but these are affected by the deletion of specific stalk domain sequences. Another highly conserved glycosite appeared at the top center of tetrameric global domain, while the others glycosites were distributed around the global domain. Here we present a detailed investigation of the distribution and the evolutionary pattern of the glycosites in the envelope glycoproteins of IVs, and further focus on the H5N1 virus and conclude that the glycosites in H5N1 have become more complicated in HA and less influential in NA in the last five years.

Influenza A viruses (IVs), which belong to the orthomyxoviridae family, consist of eight negative RNA strands. Hemagglutinin (HA) and neuraminidase (NA) are two glycoproteins that are encoded by the IV genome, expressed from segments 4 and 6, respectively. The selection due to various host immune systems and anti-flu drugs accelerate the mutation rates of viral proteins, especially for these two membrane proteins [1, 2] . There are 17 HA subtypes and 10 NA subtypes, designated H1-H17 and N1-N10, respectively. Over 118 combinations of IVs can be isolated from wild birds, which are also the natural reservoir of these viruses (except the H17N10 virus, which, until recently, was isolated only from bat) [3] [4] [5] . The species jumping ability of IVs can result in the infections of poultry and mammals, such as chicken, swine, equine or whale species, with different virulence levels [6] [7] [8] [9] . The H1N1, H2N2 and H3N2 viruses have been responsible for tens of millions deaths during the deadly history of human influenza epidemics. Furthermore, the H5N1, H7N7, H7N2, H7N3 and H9N2 viruses have been isolated from sporadic human infections and deaths [10] [11] [12] [13] . It is worth noting that the H5N1 virus is the most severe for human and avian species, with sudden onset and high mortality. The mortality rate in hundreds of patients who were hospitalized for H5N1 infections was roughly 59.05%, much higher than the mortality rates of the Spanish Flu or the 2009 influenza pandemic (H1N1) [11, 14, 15] . As a requirement for infection, the homotrimeric HAs play a key role in binding to the host sialic acid (SA) receptors and membrane fusion. The nascent HA of all subtypes consists of conserved structures, including the signal peptide, the cytoplasm domain, the transmembrane domain and the extracellular domain [16] . The mature HA monomer can be cleaved by proteases into the global HA1 and stalk HA2 subunits [17] . When IVs are located in the host digestive tract or respiratory tract cell, the receptor binding domains (RBDs) at the tip of HA1 bind to the SAa2-3Gal or SAa2-6Gal receptors, which are essential for endocytosis [18, 19] . HA unfold and expose the interior HA2 subunits in the acid environment, then the fusion peptides in HA2 insert themselves into the host membrane (viral membrane fusion) [20, 21] . Homotetrameric NA is a type II membrane protein, whereas HA is a type I membrane protein. The nascent NA consists of four parts: the cytoplasm tail (in amino-terminus), the transmembrane domain, the stalk domain and the global domain [22] . Different subtypes of NA are composed of 450,480 amino acids, displaying low sequence similarity. Although there is variable homology among the various NA subtype sequences, especially in the N1 and N2 subtypes with the deletion of 4,30 amino acids in the stalk domain [23] , NA subtypes display stable topologies: a six-bladed b-propeller fold makes an enzymatic activity domain that functions in the release of progeny virions [24, 25] . HA and NA have a distribution ratio of 4:1 on the influenza viral envelop and maintain the basic functions of host recognition, infection and viral diffusion [26] . Various studies have reported that some of the factors that influence HA include the number of the basic residues in the HA0 cleavage site, the mutation of key residues in the RBD, the changing of antigenic sites or Nglycosylation sites (glycosites) and the variation of the topology of N-glycan structures [15, 27] . Meanwhile, the factors that influence NA include deletion of the stalk domain, the mutations drugresistance, as well as the changing of antigen sites or glycosites and the variation of the topology of N-glycan structures [28] . As a kind of glycan-binding protein (GBP), HA works with NA, which functions as an exoglycosidase, cooperatively. Only by achieving a dynamic equilibrium between attaching to the host and releasing progeny virions can the IVs gain a long-term mechanism for infection and diffusion. The existence of N-glycosylation is necessary for viral membrane glycoproteins. The biosynthesis and modification of nascent secretory or membrane proteins occurs in the endoplasmic reticulum and Golgi, N-linked glycans encode crucial information for the folding, maturation, transport or degradation of proteins [29] . To escape both the host's humoral and cellular immune systems, the potential glycosites in viral envelope proteins can provide the identical glycans as those of the host's cells to mask the antigenic sites [28, 30] . Additionally, glycosylation also impacts the sensitivity of HA to temperature, the protection of cleavage sites and the stalk domain, and even the receptor-binding preferences [31] [32] [33] . As the ideal model for the influence of N-glycosylation in pathogen-host interaction, the present studies show that the envelope glycoproteins of IVs appear to only have N-glycosylation, with no O-glycosylation and GPI-anchors [34] ; hence, the glycosites discussed in this paper only pertain to N-glycosylation site. Figure 1 . The N-J trees of two glycoproteins in IVs with the corresponding distribution chart of glycosites. The phylogenetic trees of HAs and NAs were constructed using three to ten representative amino acid sequences in each subtype (File S1). The distribution charts of glycosites, colored according to the statistics of conservation in each HA or NA subtype (File S2), are shown in various strips. The red, green and blue color represent the levels of conservation of ''.95%'', ''5%,95%'' and '',5%'', respectively. The conserved cysteines are shown in yellow strips. (A) The N-J tree of HA subtypes with the corresponding distribution chart of glycosites. (B) The N-J tree of NA subtypes with the corresponding distribution chart of glycosites. doi:10.1371/journal.pone.0049224.g001 The Glycosylation Sites in Influenza Virus PLOS ONE | www.plosone.org Current studies have analyzed the evolutionary dynamics of Nglycosylation sites of select subtypes or HA/NA as a whole [35] [36] [37] . Although most influenza evolution can be accounted for by genetic drift, there is also evidence of adaptive evolution of mutations which are under positive selection [38, 39] . We have extended previous studies of basic similarity alignment scoring of similarity to analyze the position-specific glycosites that are under selective pressure in IVs [40] . Here we present a detailed investigation of the distribution and the evolutionary pattern of the glycosites in the envelope glycoproteins of IVs, especially in the H5N1 virus. Protein sequence data from all subtypes of HA and NA The amino acid sequences of HA and NA were obtained from the NCBI (National Center for Biotechnology Information) Influenza Resource (http://www.ncbi.nlm.nih.gov/genomes/ FLU, accessed 15th March 2012) [5] . To fully understand the distributed regularity of the glycosites in each subtype, we downloaded the 29 sets of HA and NA with the customized definition ''.{serotype} {strain} {segname}'' using the following combinations: H1Nx,H17Nx and HyN1,HyN10 (where x and y represent ''any'' by default), in addition to the HA and NA from the H5N1 virus. All the sequence alignments of the various HAs and NAs were performed using Clustal 2.0. Repetitive, incomplete and mixed sequences were removed. For the purpose of convenience, one representative sequence with the generally longest length from each subtype was chosen for further description (Table S1 and S2). To assess the relationship between the distributed regularity of the glycosites in the envelope glycoproteins and the evolutionary position of various HAs and NAs, firstly, the preliminary trees for each HA and NA subtype were constructed using all available sequences; secondly ten representative sequences from discrepant clades, with consideration of host source and areas, were picked up (excluding H14, H17 and N10 due to limited records). Finally, the phylogenetic trees of HA and NA were constructed by these hundreds of sequences (File S1). The phylogenetic trees of the HA and NA were constructed by MEGA5.05 with N-J methods and a p-distance model with a bootstrap value of 1000 [41, 42] . As is known, N-X-T/S (X cannot be a proline) is the glycosite motif. Although some researchers have concluded that not all the potential glycosites would be glycosylated in mature glycoproteins, Kelley WM et al. argued that the 14-sugar glycans would be transferred to all the glycosites in nascent proteins and be truncated and modified in the subsequent process [42] . Due to the fixed pattern of glycosite, many programs and software provide the prediction of glycosites [43] , such as the ''NetNGlyc 1.0 Server'' (www.cbs.dtu.dk/services/NetNGlyc). By submitting the alignment files to a prediction server, we obtained a series scores for the potential glycosites. In consideration of experimental error, those occasional glycosites, which may result from the genomic sequencing or translated by different genetic codes, were excluded. For example, the 429NLS in the HA of A/ duck/Hunan/3315/2006(H5N1) only appears in H5 set once, and it is also embedded inside of HA trimer [44] . Statistics of the position, pattern and the levels of conservation in various subtypes were used for further discussion (Table S1 , File S2). The available structures of HA and NA would help us to explore the distinctive function of various glycosites. Therefore, we collected the structures of various HA or NA subtypes from the PDB (Protein Data Bank, http://www.rcsb.org/pdb/home/ home.do). To obtain the whole multimer, these monomer structures were processed with a VMD 1.9 transformation matrix using a Tcl script [45] . The code for each representative structure of the various subtypes of HA and NA and the corresponding source of IVs are shown in Table S2 . Most coordinate files of envelope glycoproteins are obtained from X-ray crystallography or NMR. Furthermore, complete larger glycans are too exible to yield sufficient electron density [46] . Isolation and purification of membrane glycoproteins by particular enzyme treatments lead to the lack of partial domains The H5N1 virus can be divided into 10 clades, according to the evolutionary position of the HAs defined by the WHO (World Health Organization) in 2008. These clades are numbered 0 to 9 [48, 49] . Clades 0,2 are responsible for all the human infections, resulting in 359 deaths since 2003 [11] . Many clades have not been reported for the last four years, while the dominant H5N1 viruses are concentrated in clade 0,2. Clade 2.1, clade 2.2 and clade 2.3 have further evolved into third-or fourth-order clades, and these newly formed viruses have become geographic strains. For the purpose of investigating the co-evolution relationship between the glycosite patterns of HA and NA, we reconstructed a phylogeny tree of HAs and determined ten clades. When discrete monophyletic groups with a common node meet a bootstrap value of $60 at the clade-defining node and average percentage pairwise nucleotide distances between and within clades of .1.5% and ,1.5%, respectively, they could split into the second-, third-, or even fourth-order clades [50] . The statistics of the glycosites of each HA, as well as the corresponding NA, were recorded according to their clades (Files S3, S4 and Figure S1 ). As shown in Figure 1A , the current 17 HA subtypes were concentrated in two evolutionary groups. One large group, represented by H1 and H5, contained H2, H6, H8, H9, H11, H12, H13 and H16; Another large group was represented by H3, and contained H4, H7, H10, H14 and H15. Although each subtype of HA had a lower similarity to each other and the distribution of the glycosites differed in sequential numbering, we found that the sequence alignment of all HA subtypes indicated that partial glycosites appeared in similar domains. Furthermore, the structure alignment of available HAs also showed that various HA subtypes had a highly conserved structure, and the distribution of glycosites was also regular ( Figure S2 ). By analyzing the position of glycosites and their conservative rates, we conclude that two types of glycosites appear in HAs: one with a high level of conservation in all HA subtypes and another with various conservative rates in different HA subtypes. Two highly conserved glycosites are located near the HA0 cleavage site (e.g., the 27NNST in H1 or the 30NGT in H8) and the fusion peptide of the HA2 (e.g., the 498NGT in H1 or 500NGS in H5) respectively in all subtypes, and these two glycosites play necessary role in viral life cycle for protecting the HA0 cleavage sites and fusion peptide [34, 51] . In addition, another highly conserved glycosite appears at C-terminal part of the HA1 sequence, which is near the connection of the global and stalk domains, except in H7 and H15 sequences [52] . Three to ten characteristic glycosites were distinctive in each subtype. Their conservative rates were influenced by different internal evolution branches, ranked from 0.5% (e.g., 292NGS with 2.85% in H3) to 100% (e.g., 38NGT with 99.94% in H3), distributed mainly in the global domain. It is worth mentioning that some highly conserved glycosites were near to cysteines (Figure 1 ). Kozlov et al. hypothesized that ERp53, which is involved in the formation of disulfide bonds during the folding of nascent proteins, would form a complex with calreticulin/calnexin, which depends on precursor N-glycans [53] . One most obvious characteristics of glycosites was that the increasing samples or cross-species reports would lead to more glycosites. There are many factors contributing to the existence of numerous glycosites in H1, H3 and H5 subtypes. Because the RNA virus is rapidly mutating, long-term cross-species infections have resulted in the accumulation of adaptive mutations and glycosites. Due to the selection pressure of different host immune systems, a disadvantageous mutation would be eliminated while a virus with the gains and losses of glycosites would be conserved. For example, the HAs from avian H1N1 or H3N2 virus had less glycosites than those of the seasonal human H1N1 or H3N2 virus (e.g., 286NAS in A/Memphis/28/1983 (H1N1). Figure 2A and File 5S). The H3 set had the most glycosites (16, as shown in File S6) among all HA subtypes, which were mainly contributed by H3N2 and H3N8 virus. The earliest H3 record had only one moderate conservative glycosite (79NCT, e.g. A/equine/Miami/1/1963 (H3N8)) except for six highly conserved glycosites, which is also conserved in most seasonal human H3N2 virus and mammalian H3N8 virus (e.g. A/Denmark/22/2011 (H3N2) or A/equine/ Yokohama/aq79/2011 (H3N8)), but not in subsequent avian H3N1, H3N9 virus (e.g. A/duck/Zhejiang/5/2011 (H3N3) or A/duck/Saitama/2/2009 (H3N8). File S5). Over the last four decades, seasonal human H3N2 viruses have gradually acquired additional glycosites within the globular HA1 [37, 40] . The occurrence of 294NSS has increased the glycosite numbers to 13 in parts of current HA of H3N2 virus from 2010 (A/Singapore/ GP5/2011(H3N2). File S6). Not all the glycosites shared the same numbering in one subtype. For this reason, we chose a representative sequence for each subtype (Table S1 ). Most glycosites were influenced by the deletion or insertion of the mutants in the upstream sequence, which is also known as ''antigenic drift'' [54] . In the HAs of the H1N1 or H5N1 viruses, some nearby glycosites were determined by ''antigenic shift'' [55] . In the H1 subtypes, the 144NVT glycosite from the seasonal H1N1 flu, which were isolated from 1940s to 1980s (e.g., A/Memphis/1/1984 (H1N1)), was replaced by the 142NHT glycosite after the 1990s (e.g., A/California/04/ 2007 (H1N1)). Additionally, no such glycosylation appeared in those H1N1 viruses isolated from swine H1N1 flu (e.g., A/swine/ Guangdong/1604/2010 (H1N1)), Spanish flu (e.g., A/South Carolina/1/1918 (H1N1)) and the 2009 influenza pandemic (H1N1) (e.g., A/Mexico city/CIA1/2009 (H1N1)). It has been partially shown that vaccination of mice with the 1918 influenza strain protected against subsequent lethal infection by the 2009 virus; however, the 1918 strain did not protect against the seasonal H1N1 flu [56] . It is interesting to note that the migration of the 179NKS to 177NLS tracks with these effects. Almost all of the 179NKS glycosites identified were centralized in swine-origin influenza viruses (S-OIVs, e.g., A/Swine/Guandong/1604/2010 (H1N1)) and parts of the 2009 influeza pandemic H1N1 virus (e.g., A/Mexico City/014/2009 (H1N1)), while the 177NLS glycosite was mainly found in the swine or human H1N2 viruses (e.g., A/ New York/481/2003 (H1N2)) and most seasonal flu viruses (e.g., A/California/04/2007 (H1N1)). These date further confirm that the HA of the 2009 influenza pandemic (H1N1) originated from S-OIV, not previous seasonal virus [57] . Similar positional conversion of the glycosites in H5N1 caused by ''antigenic drift'' was the ''179NYT/181NNT''. The conserved 181NNT glycosites have been replaced by the 179NYT in parts of clade 7 viruses since 2005. It reminds us that although these glycosites were adjacent, their N-glycans would shield different antigenic sites, which would provide some suggestions for the development of influenza vaccine. Moreover, in the smaller cluster of H7, H10 and H15, one highly conserved glycosite appeared in the long a-helix of HA2 (e.g., 431NWT in A/turkey/Chile/4418/02 (H7N3)). This uncommon glycosite requires further investigation (File S1). It can be observed from Figure 1B that N10 is highly divergent from other subtypes [4] . The remainder of the current 9 NA subtypes are concentrated into two evolutionary groups: one group was represented by N2 and contains N3, N6, N7, and N9; and another group contains N1, N4, N5, and N8. The glycosites of NA can be divided into two types according to the distributed region: two to four highly conserved glycosites are located in the stalk domain in each subtype; two conserved glycosites and most middle-low conserved glycosites are mainly located in the global domain, which are near the tip of NA, the connection of the global and stalk domains, or the antigenic sites. The glycosite of 146N is conservative in all NA subtypes (e.g., 146NDT in A/Boston/20/2008(H3N2), 144NGT in A/duck/ Taiwan/4201/99(H7N7) or 146NGT in /Viet Nam/1203/ 2004(H5N1)). It has been shown that the N-glycan at this glycosite affects NA enzymatic activity, causing a 20-fold decrease in activity [58] . Similar to the description of HAs, a large number of H1N1, H3N2, H5N1, H7N2 and H9N2 viruses have accumulated numerous glycosites in N1 and N2, especially in the global domain, mainly participating in immune evasion. Moreover, one conserved glycosite, 12NTT (Conservative rate: 93%, e.g., A/ turkey/Italy/3807/2004 (H7N3)), located in the transmembrane domain in N3 and N10, requires further investigation [59] . The significance of the conserved glycosites in the stalk domain was providing the N-glycans to avoid the cleavage by host enzyme (e.g., Trypsin) [60, 61] . The variance of the glycosites was closely related to the deletion of the stalk domain. Although the threedimensional structure of stalk domain has not been determined yet, it is speculated that the presence of an a-helix motif in the uncrystallized structure has also been provided by cryoelectron microscopy [62] . Wagner et al. believed that a longer stem domain would enhance the replication capacity of the virus, while the deletion of the stem domain would decrease the enzymatic activity of NA [63] . Various subtypes had stalk domain deletions of 3 to 24 residues, except for N4, N8, N9 and N10. The numbers of deletions were also distinctive across different combinations of IVs and even within one subtype, such as the N2 subtype. Generally, no deletions were found in the NA of the H3N2 virus; deletion of 3 residues and one corresponding glycosite with the pattern ''E-R-61N-3-64T-V-H'' (meaning 3 residues missing between 61N and 64T, e.g., A/chicken/Zhejiang/611/ 2011 (H9N2)), appeared in most of the NA subtypes of the H9N2 virus. A similar deletion of 20 residues and two glycosites (I-E-60R-20-80N-I-I) appeared in most NA subtypes of the H6N2 virus (e.g., A/duck/Fujian/3193/2005 (H6N2)). The deletion of 16 residues and two glycosites as ''C-E-55P-16-72T-T-E'' were a distinctive part of the H7N2 virus (e.g., A/unknown/New York/19501-5/2006 (H7N2)). Parts of the H5N2 virus were characterized by the deletion of 20 residues and two glycosites (''R-N-62I-20-83G-Y-R'', e.g., A/chicken/Ibaraki/3/2005 (H5N2)). Moreover, in some avian H2N2 viruses, the deletion of 22 residues resulted in the loss of two glycosites; however, a new glycosite appeared in the newly created sequences (''P-A-47N-22-70N-T-V'', e.g., A/chicken/New York/Sg-00300/1997 (H2N2)). In all, the diversity of deletions in the stalk domain indicates that the glycosylation pattern of HA and NA has a complex relationship. Globally, the researchers have been paying attention to the highly pathogenic avian influenza (HPAI) since the first human death caused directly by avian H5N1 virus in 1997. It is generally considered that the HPAI viruses were characterized by polybasic residues in the HA0 cleavage site in HA and the deletion of the stalk domain in NA [64] . However, these characteristics had been reported even before 1990s. The earliest H5N1 virus, ''A/ chicken/Scotland/1959 (H5N1)'', had four continuous basic residues in the HA0 cleavage site (compared to 5,6 continuous basic residues in most common H5N1 HAs); the stalked deletion of NA also existed in A/turkey/Ontario/84/1983 (H5N1). The records of H5N1 virus have increased rapidly since 2003. Since then, a number of new clades and subclades have emerged and resulted in various new glycosites ( Figure 3A and 3B) . The percentage of unconserved glycosites in HA also increased and diversified rapidly after 2003, meanwhile the percentage of unconserved glycosites in NA reduced gradually, regardless of those highly-conserved glycosites ( Figure 3C and 3D) . Both twelve glycosites were found in the HA and NA of the H5N1 virus, shown in the Tables 1 and 2 and Files S3 and S4. Most H5N1 viruses were grouped into clade 0 before it appeared in Hong Kong again in 2003. Various patterns of the glycosites in HA and NA had co-existed in these original viruses. These original NAs contained the known glycosites, including four highly conserved glycosites in the stalk domain and seven in the global domain (except the occasional 341NGT which only appeared in clade 1 and Thailand records during 2004,2010, e.g., A/chicken/Thailand/CU-354/2008 (H5N1)). In contrast, six highly conserved glycosites together with 170NST exist widely in clade 0. Since 2003, the WHO has recorded a three-wave epidemic of H5N1, which resulted in hundreds of deaths and huge economic losses [11] . Until recently, the glycosite patterns were highly conserved in all the avian clades except for clade 7 (e.g., clade 3, 4, 5, 6, 8 and 9), as shown in Figure 4 and Figure S2 . Most currently recorded H5N1 viruses were concentrated in the fourth-order clades. During this decade, there have been increasing human infections and new glycosite patterns of HA and NA. It has been reported that the H5N1 virus of clade 2.2 was involved in the outbreak that occurred among the migratory bird population near Qinghai Lake in 2005. Since then clade 2.2 spread westward. This resulted in a number of deaths of wild birds in Europe [65] . In 2006, the H5N1 virus appeared in Africa for the first time, followed by hundreds of mortally infected humans in Egypt, Nigeria and Djibouti [66] . Figure 5D , File S3). The 181NVT glycosite, which is located at the apical b-folding of HA, is conserved in all human IVs (e.g., A/Anhui/1/2007 (H5N1)). Previous statistics have indicated that this glycosite has a lower level of conservation in avian IVs (e.g., A/chicken/ Vietnam/NCVD-093/2008 (H5N1)). The results of the molecular dynamics simulation indicated that the a2-3-sialoglycans adopted a straight-like and outward topology structure while the a2-6sialoglycans were fishhook-like and inward; therefore, we inferred that the that the deficiency of the glycosite would benefit the binding of SAa2-3Gal sialoglycans [68] . Actually, all the viruses that had deficiencies of the 181N glycosite were isolated from the avian host, which were concentrated in clades 7.1, 7.2 and 2.2.1.1 ( Figure 6A, File S3) . Interestingly, the deletion of the stalk domain in H5N1 NA is variable. As the first human death reported in 1997, most H5N1 viruses that belonged to clade 0 and isolated in Hong Kong were characterized by ''N-Q-S-I-54I-18-73N-F-Y'', which remained a glycosite: 50NQS. Although previous studies conjectured that the H6N1 virus (e.g., A/Teal/Hong Kong/W312/97 (H6N1)) was the donor of the NA gene in 1997 HPAI virus [69, 70] ; however, similar motif could be found even in 1983. Since 2000, the most common pattern of ''A-E-48P-20-69I-S-N'' with four glycosites missing has dominated in the H5N1 NAs ( Figure 6B, File S4 ). There is one kind of glycoprotein that participates in the recognition and membrane fusion in most virus envelopes, such as the spike (S) protein in the SARS virus, the gp160 in HIV and HA in IVs [71, 72] . These glycoproteins can bind to one specific glycan structure which is known as the lectin or GBP. Other viral glycoproteins, such as HN in the Newcastle Disease Virus or NA in IVs, function as the exoglycosidase in the release of virus particles [73] . The IVs have an innate capacity for high mutation rates because they are RNA viruses. Humans are under continuous attacks by newly emerging IVs which constantly undergo ''antigenic drift'' and ''antigenic shift''. Accumulation of substantial sequences and 3D coordinates of IV proteins have provided the ideal tools for the investigation of how mutations affect transfaunation, vaccine design and drug-resistance [74, 75] . N-glycosylation not only influences the folding and secretion of glycoproteins such as HA and NA but also provides the same glycans (similar to the host's own glycans) to escape the host's immune system. As the key modification of biological significance in viral glycoproteins, we found the glycosites in 17 known HA subtypes and 10 known NA subtypes that have complex characteristics. In general, two highly conserved glycosites near the HA0 cleavage site and fusion peptide site may maintain the basic function of HA; these sites were seldom absent in the HA subtypes. More glycosites were identified in the global domain or in the connection of global and stalk domains, along with longterm and large-scale epidemics in several of the subtypes. The distributional regularity of the glycosites in the NA subtypes is also complex; two to four glycosites located in the stalk domain are highly conserved in various subtypes, and are affected by the deletion of the stalk domain. Another highly conserved glycosite was found at the tip of tetramer NA in all subtypes. Other glycosites were found to be mainly concentrated in the global domain, which surrounds the antigenic sites. The HA or NA subtypes exhibited low similarities of amino acid sequences in all subtypes while maintaining identical structures, which revealed that the functions of various HA or NA subtypes remain conserved. Notably, some of the glycosites near the cysteines were found to also be conserved. The cysteines take the main role in stabilizing the tertiary structure; the conservation verifies that cysteines and N-glycans played an important role in the protein folding and quality control. We have further investigated the H5N1 virus to elaborate the collaborative relationship of glycosites in HA and NA. Five highly conserved glycosites in HA had existed before the H5N1 virus first crossed species barriers and infected humans, as well as two additional glycosites, 181NVT and 170NST. Since 2003, the H5N1 virus has exhibited a rapid evolutionary dynamic. Under the selection pressure of different hosts and antigenic drift, the glycosite pattern of current H5N1 viruses in different geographical locations has been distinctive: the HA in clade 2.2.1.1, isolated from Egypt, lacked the 181N glycosite but had added the 88NVS glycosite. In addition, the HA in clade 2.3.2.1, isolated from China or Vietnam, lacked the 170N glycosite but had added the 152N glycosite. In contrast, as the glycosites in HA became more diverse in the H5N1 virus, the glycosylation in NA was impaired by a decreasing number of glycosites. The NAs of current H5N1 viruses lack a stalk domain and the four corresponding glycosites; except the four high conserved glycosites in the global domain, others glycosites have rarely been identified. The envelope glycoproteins, which play a crucial role in virus recognition, invasion and spread. The analysis of the glycosites in HAs and NAs has provided basic information for vaccine design, host selection and changing virulence. However, infection is a complex process; the alternation of glycosites and glycan shapes may affect the functions of glycoproteins. In addition, there are other mutations also worthy of further consideration, such as the E627K in the PB2 protein that enhances the avian viral replication capacity in mammalian cells [76, 77] . Drug resistance is also related to viral proteins, such as the M2 protein [78] . Thus, to prevent the next influenza pandemics, more research needs to be done. Figure S1 The N-J trees of H5N1 HA. The phylogenetic tree was inferred from protein sequences by the Neighbor-Joining method and rooted using A/chicken/Scotland/1959. Estimates of the statistical significance of the phylogenies were calculated by performing 1,000 bootstrap replicates. The clades classified by the WHO are shown as colored bars. (TIF) Figure S2 The superposition of crystal structures from various HA and NA subtypes. The red, green, blue and yellow color represent the H1, H3, H5 and H7 in HAs or N1, N2, N8 and N9 in NA respectively. File S5 The analytic alignment file containing the 5973 protein sequences of H3 HA. As is shown in the output file, the red, green and blue color represent the levels of conservation of ''.95%'', ''5%,95%'' and '',5%'', respectively. (ZIP) File S6 The N-J trees of H3 HA. The phylogenetic tree was inferred from H3 HA sequences by the N-J method and rooted using A/equine/Miami/1/1963(H3N8). Estimates of the statistical significance of the phylogenies were calculated by performing 1,000 bootstrap replicates. 633 representative H3 HA sequences were chosed in consideration of the age, hosts, areas, as well as the known HA sequences from H3N1 and H3N3,H3N7 viruses. Those unconserved glycosites are labeled (conservative rate ,95%).

Niclosamide Is a Proton Carrier and Targets Acidic Endosomes with Broad Antiviral Effects

Viruses use a limited set of host pathways for infection. These pathways represent bona fide antiviral targets with low likelihood of viral resistance. We identified the salicylanilide niclosamide as a broad range antiviral agent targeting acidified endosomes. Niclosamide is approved for human use against helminthic infections, and has anti-neoplastic and antiviral effects. Its mode of action is unknown. Here, we show that niclosamide, which is a weak lipophilic acid inhibited infection with pH-dependent human rhinoviruses (HRV) and influenza virus. Structure-activity studies showed that antiviral efficacy and endolysosomal pH neutralization co-tracked, and acidification of the extracellular medium bypassed the virus entry block. Niclosamide did not affect the vacuolar H(+)-ATPase, but neutralized coated vesicles or synthetic liposomes, indicating a proton carrier mode-of-action independent of any protein target. This report demonstrates that physico-chemical interference with host pathways has broad range antiviral effects, and provides a proof of concept for the development of host-directed antivirals.

1 Supporting Text S1 infection (relative to control)

A Single Native Ganglioside GM(1)-Binding Site Is Sufficient for Cholera Toxin To Bind to Cells and Complete the Intoxication Pathway

Cholera toxin (CT) from Vibrio cholerae is responsible for the majority of the symptoms of the diarrheal disease cholera. CT is a heterohexameric protein complex with a 240-residue A subunit and a pentameric B subunit of identical 103-residue B polypeptides. The A subunit is proteolytically cleaved within a disulfide-linked loop to generate the A1 and A2 fragments. The B subunit of wild-type (wt) CT binds 5 cell surface ganglioside GM(1) (GM(1)) molecules, and the toxin-GM(1) complex traffics from the plasma membrane (PM) retrograde through endosomes and the Golgi apparatus to the endoplasmic reticulum (ER). From the ER, the enzymatic A1 fragment retrotranslocates to the cytosol to cause disease. Clustering of GM(1) by multivalent toxin binding can structurally remodel cell membranes in ways that may assist toxin uptake and retrograde trafficking. We have recently found, however, that CT may traffic from the PM to the ER by exploiting an endogenous glycosphingolipid pathway (A. A. Wolf et al., Infect. Immun. 76:1476–1484, 2008, and D. J. F. Chinnapen et al., Dev. Cell 23:573–586, 2012), suggesting that multivalent binding to GM(1) is dispensable. Here we formally tested this idea by creating homogenous chimeric holotoxins with defined numbers of native GM(1) binding sites from zero (nonbinding) to five (wild type). We found that a single GM(1) binding site is sufficient for activity of the holotoxin. Therefore, remodeling of cell membranes by mechanisms that involve multivalent binding of toxin to GM(1) receptors is not essential for toxicity of CT.

dation by the proteasome due to the paucity of lysine residues in its structure. It then forms an allosterically activated complex by binding to an ADP ribosylation factor (ARF), and it ADP ribosylates the alpha subunit of the stimulatory G protein, leading to constitutive activation of adenylate cyclase. In the human intestinal cell line T84, an increased concentration of cyclic AMP elicits a Clsecretory response which can be measured electrophysiologically in real time as a change in short circuit current, I sc (11) . A key step in the intoxication process is the transport of the toxin from the cell surface to the ER. How GM 1 confers the specificity for lipid trafficking has not yet been determined. Pentameric binding of GM 1 by the B subunit of CT may itself induce membrane curvature and induce invagination to begin the entry process, as is seen with simian virus 40 (SV40) (12) and Shiga toxin (13) (14) (15) . Interestingly, invasion by SV40 occurs only with GM 1 that has native long-chain acyl groups. Clustering of GM 1 may enable the toxin to associate with lipid rafts that serve as platforms for trafficking of CT through the retrograde pathway or parts of it. Lipid rafts are viewed as highly dynamic microdomains that may self-assemble in membranes from sphingomyelin, cholesterol, glycolipids, and proteins that favor a lipid-ordered microenvironment (reviewed in references 16 to 18). Some functions attributed to lipid rafts may require interactions with protein components or scaffolds to stabilize them, extend their lifetimes, or facilitate their coalescence into larger physiologically significant structures. By binding to and cross-linking five GM 1 molecules, CTB might serve as such a protein scaffold and promote the function of lipid rafts in toxin trafficking. Conversion of PM sphingomyelin to ceramide or acute depletion of membrane cholesterol both prevent endocytosis of CT (19, 20) , consistent with a role for lipid rafts in CT trafficking. A requirement for the lipid raftassociated proteins flotillin 1 and flotillin 2 in a zebra fish model of intoxication by cholera toxin (21) also supports the key role of lipid rafts in trafficking of CT. Our previous study (22) showed that a mixture of CT holotoxins produced in vivo and having chimeric B pentamers with from zero to two wild-type (wt) GM 1 binding sites (BS) was still capable of intoxicating host cells. Because genetic methods, not chemical modifications, were used to prepare these toxin variants, the structures of their GM 1 BS were fully defined and were known to be either the same as wt sites or inactive. Thus, having at most 2 wt GM 1 BS was sufficient for CT to intoxicate host cells, albeit at reduced efficiency. Left unanswered, however, was the question of whether a holotoxin with only one binding site for the GM 1 receptor can function. A toxin molecule able to bind only a single GM 1 molecule would be completely unable to cluster GM 1 molecules or scaffold them into microdomains and would therefore be unable to induce membrane curvature (2, 23) . We designed the experiments reported here to produce defined holotoxin variants that have from zero to five native GM 1 BS in their B pentamers, and we compared their abilities to bind to GM 1 and intoxicate T84 cells. We found that holotoxins able to interact with only a single GM 1 molecule can nevertheless still complete the intoxication process, demonstrating directly that binding a single GM 1 molecule permits the toxin to enter the host cell, complete the trafficking process, and deliver the toxic CT-A1 fragment to the cytoplasm. However, we also found that eliminating even a single binding site from wt holotoxin produced a detectable attenuation in toxicity. Production and purification of cholera toxin variants with 0, 1, 2, 3, 4, or 5 native GM 1 binding sites. To produce CT holotoxins with defined combinations of wt and mutant CTB subunits, we made three compatible plasmid constructs encoding either tagged wt CTB, CTB-G33D, or wt CTA and expressed them in the same Escherichia coli strain to produce a holotoxin pool with B pentamers consisting of tagged wt CTB and/or CTB-G33D, from which to purify all six possible holotoxin variants. The tag is a 34-aminoacid peptide (denoted GSH6) which is genetically appended after the Met103 codon of native ctxB and encodes glycosylation (bold) and sulfation sites (underlined) (24) , SSSGGGGSSH-PNNTSNNTSSAEDYEYPS, followed by six His residues. Each plasmid has a different replication origin, antibiotic selection, and combination of promoters ( Fig. 1A ; Table 1 ). The ctxA gene, encoding CTA, is expressed from dual pLac and pBAD promoters. Holotoxins and free B pentamers were purified from whole-cell lysates by metal-ion affinity chromatography on Talon resin, and free B pentamers were removed by passing the eluate pool over cation exchange resin, resulting in binding of the free CTB pentamers to the resin and recovery of holotoxin in the flowthrough (Fig. 1B) . Holotoxins with different numbers of BS were then separated by anion-exchange chromatography (Fig. 1C) . By varying the amount of each inducer, we sought to alter the ratio of wt and mutant B subunits in the holotoxin pool. However, for practical purposes, we found that a single condition of relatively low levels of arabinose (0.0005% to express GSH6-tagged wt-CTB) and high levels of isopropyl-␤-d-thiogalactopyranoside (IPTG) (400 M to express native-size CTB-G33D) gave acceptable yields of holotoxin with homopentameric and singly or doubly tagged heteropentameric B subunits (with zero, one, and two wt BS, respectively) and detectable but lower yields of the triple, quadruple, and homopentameric tagged B subunits as described in detail in the next paragraph. This production strain we designated AMBT (for A, mutant CTB, wt CTB-tagged). Simply by swapping the wt and G33D alleles in the respective CTB-encoding vectors and keeping the same inducer ratios, we were able to express the single or double tagged heteropentameric and homopentameric wt CTB subunits with three, four, and five wt BS, respectively, from the production strain designated ABMT (A, wt CTB, mutant CTBtagged). Using these strategies, from two production strains, we produced all six variant holotoxins, which had from zero to five wt BS and a maximum of two tagged subunits. A third production strain, designated ABBT (A, wt CTB, wt CTB-tagged), in which both the GSH6-tagged and untagged B polypeptides were wt CTB, was used under the same expression conditions to produce control holotoxins with five native BS and zero, one, or two tagged CTB subunits. Figure 1 shows the purification process for the AMBT strain (mutant CTB-G33D, wt CTB tagged). Since CTB pentamers naturally bind to metal affinity resins (25) , cholera toxin variants can be purified to near-homogeneity by a single passage over Talon resin. The imidazole eluate pool from the Talon resin contained a mixture of holotoxin and some free CTB pentamers. These free pentamers were removed by cation exchange chromatography, where at pH 8.0 in 50 mM Tris buffer, holotoxin passed through the column while free pentamers bound and could be eluted subsequently with a salt gradient (Fig. 1B) . The heterogenous mixture of holotoxins was further separated into its components by anion-exchange chromatography. The GSH6 tag not only changed the size of the B monomer but also changed its predicted pI from 8.24 for native CTB to 6.21 for CTB-GSH6. The respective predicted pIs for the CTB-G33D variant changed from 7.31 for the native length to 5.91 for the GSH6-tagged variant. All holotoxins bound to the anion exchange matrix, and at least five peaks could be discerned following elution with a salt gradient (Fig. 1C) . Samples from individual fractions were analyzed by SDS-PAGE either without (Fig. 1D , upper) or after (Fig. 1D , lower) boiling in sample loading buffer. The unboiled samples in lanes 4 to 13 showed a single band corresponding to assembled holotoxin (AB 5 ), with a small amount of contaminating protein of slower mobility in lanes 2 and 3. The major fractions from each peak are homogenous and are consistent with each peak containing a single species CTB-G33D-Linker-His 6 , pAra BAD , Cm r , Ori p15a This study of variant holotoxin, The assembled holotoxins in the unboiled samples bind less SDS than if they were fully denatured and therefore do not migrate in direct proportion to their molecular weights. The relative molecular weight (M r ) for each holotoxin variant is increased by the presence of one or more tagged CTB subunits. Each of these holotoxins completely disassembles into its component polypeptides after boiling (Fig. 1D , lower panel; CTA, 29 kDa; CTB-GSH6-tagged monomer, 14.4 kDa; and native CTB-G33D monomer, 11.5 kDa). The holotoxin in peak 0 ( Fig. 1C ) has homopentameric CTB-G33D subunits and no functional GM 1 BS, while holotoxins in peaks 1 and 2 have a mixture of wt CTB-GSH6 and CTB-G33D pentamers with 1 and 2 GM 1 BS, respectively. The fractions loaded in lanes 2 to 5 contain CTA plus 5 CTB-G33D (peak 0), and those in lanes 6 to 9 contain CTA plus 1 wt CTB-GSH6 and 4 CTB-G33D (peak 1), and those in lanes 10 to 13 contain CTA plus 2 wt CTB-GSH6 and 3 CTB-G33D (peak 2). Only these first three peaks were well enough separated to be purified efficiently from the extract of production strain AMBT. The fractions corresponding to each peak were pooled, dialyzed, and rerun over the same anion-exchange column to achieve a very high degree of purity. To make holotoxins with 3 and 4 native GM 1 BS, an expression and purification run was done using strain ABMT, which produced three peaks similar to those shown in Fig. 1C , containing holotoxins with zero, one, or two GSH6tagged CTB-G33D subunits and five, four, or three native GM 1 BS, respectively. Finally, as noted above, strain ABBT was used for expression and purification of wt control holotoxins, all with 5 GM 1 BS and zero, one, or two GSH6-tagged CTB subunits. Analysis of the eight final highly purified holotoxin preparations by SDS-PAGE and Coomassie blue staining is shown in Fig. 2 . The upper panel shows a gel run with unboiled samples, and the lower panel shows a gel run with samples boiled in sample buffer to dissociate the components. Without boiling, holotoxins with untagged or tagged wt CTB pentamers (Fig. 2 , upper panel, lanes 2 to 4) dissociated into free CTA and stable pentameric B subunits that ran near the 45-kDa marker (lane 1). Interestingly, the presence of one or more CTB-G33D polypeptides increased stability of the heterohexameric AB 5 holotoxins. Consequently, unboiled holotoxins with three or more CTB-G33D polypeptides (Fig. 2 , upper panel, lanes 5 to 7) remained fully associated and ran near the 66-kDa marker (lane 1), whereas unboiled holotoxins with one or two CTB-G33D subunits exhibited partial dissociation into B pentamer and free CTA (Fig. 2 , upper panel, lanes 8 and 9). The high stability of the heterohexameric AB 5 holotoxins containing five, four, or three CTB-G33D polypeptides is also evident in the upper panel of Fig. 1D . As expected, increasing the number of tagged CTB polypeptides resulted in moderate decreases in observed mobilities of the pentamer bands (Fig. 2 , upper panel, compare lanes 2, 3, and 4) and the holotoxin bands (Fig. 2 , upper panel, compare lanes 5, 6, and 7), and increasing the number of CTB-G33D polypeptides resulted in slight increases in mobilities of the holotoxin bands (Fig. 2 , upper panel, compare lanes 6 to 8 and lanes 7 to 9). Upon boiling, all holotoxins dissociated and resolved into their individual polypeptide components (Fig. 2, lower panel) . All holotoxins have one CTA subunit. Holotoxins in lanes 2 and 5 are predicted to have five native CTB monomers; holotoxins in lanes 3, 6, and 8 have are predicted to have one tagged CTB and 4 untagged CTB monomers; and holotoxins in lanes 4, 7, and 9 are predicted to have two tagged and three untagged CTB monomers. Stoichiometry of the individual polypeptides in each purified holotoxin variant was confirmed experimentally by densitometric scanning of the gel in the lower panel of Fig. 2 . To adjust for differences in loading of the purified holotoxins (ranging from 0.6 g in lane 6 to 2.9 g in lane 2), the observed density for each band was expressed as a fraction of the total density for all bands in the same lane. The observed fractional densities were then compared with the expected values based on the predicted molecular mass of each polypeptide (CTA, 27.2 kDa; wt CTB, 11.6 kDa; and tagged CTB, 15.3 kDa) and the assumption that binding of Coomassie blue is proportional to the mass of each peptide. The results (see Table S1 in the supplemental material) were generally within 15% of expected values, but all observed values for tagged B subunits were higher than expected, suggesting that the tagged subunits bound proportionately more Coomassie blue stain than native CTA or CTB. Corrected for loading differences, the expected stoichiometric ratios for the tagged B subunit bands in lanes 3 versus 4, 6 versus 7, and 8 versus 9 were 1:2, and the corresponding observed ratios were 1:2.1, 1:1.7, and 1:1.9, respectively. Ganglioside GM 1 -binding activities of cholera toxin variants. Relative binding activity of each holotoxin preparation with from zero to five binding sites (and with zero, one, or two tagged B subunits) to ganglioside GM 1 receptor was measured by ELISA with a fixed amount of toxin (5 ng, 60 fmol) added to individual wells coated previously with serial dilutions of GM 1 , and bound toxin was detected with polyclonal rabbit anti-CTB and horserad- ish peroxidase (HRP)-conjugated secondary antibody. Figure 3 shows that at high GM 1 density, toxins with more than one native GM 1 BS bound almost as well as native cholera toxin. The single-BS holotoxin showed a lower plateau signal at high GM 1 density than holotoxins with more than one BS, which we interpret to be due to a less favorable equilibrium between binding and release for the holotoxin with one BS for GM 1 . With more than one BS, a toxin molecule is expected to exhibit faster initial binding to immobilized GM 1 , multivalent binding as the density of immobilized GM 1 increases, and slower dissociation from GM 1 , resulting in increased avidity of binding. In wells coated with 75 nM GM 1 (50 l, 3.75 pmol), toxins with 2 or more native GM 1 BS bound at more than 90% of the wt level, toxin with a single BS bound significantly less at 60% of the wt level, and toxin with no BS gave a minimal signal (4% or less of the wt level). At lower GM 1 densities, there were significant differences between all variants, and in wells coated with 1.2 nM GM 1 (50 l, 60 fmol), toxins with four or three BS bound at 54 and 49% of wt levels, respectively, and toxins with 2, 1, or no BS bound at 25, 12, and 1% of wt levels, respectively. The amount of GM 1 required to coat each well and give 50% of the maximal signal was calculated to be 65, 100, 150, 240, and 850 fmol per well for wt holotoxin and for holotoxin variants with 4, 3, 2, or 1 wt BS, respectively. Biological activities of the eight holotoxin preparations (the untagged and singly and doubly tagged native holotoxin controls and the variant holotoxins with zero to four wt BS and one or two tags) were initially tested in an overnight assay on mouse Y1 adrenal cell monolayers, on which cholera toxin causes an easily scorable morphological change (rounding of intoxicated cells). In this assay, approximately 2 ng of unnicked native cholera toxin caused rounding of 75 to 100% of the cells (48 U/100 ng). The G33D holotoxin showed some rounding at only the highest concentration tested (33 ng, extrapolating to less than 1 U/100 ng). All holotoxin variants with one or more BS showed rounding of Y1 cells at between 24 and 48 U/100 ng, showing that a single GM 1 binding site is sufficient to intoxicate mouse Y1 cells at near-wt levels. Failure to detect significantly less toxicity after overnight exposure of mouse Y1 adrenal cells to holotoxin variants with from 1 to 4 GM 1 binding sites suggests that any differences in delivery of the wt CT-A1 fragment from the cell surface to the cytosol by these holotoxin variants were not rate limiting for development of the morphological manifestations of intoxication. To assess the biological consequences of decreased numbers of BS in a more quantitative manner, we determined the real-time electrophysiological effects of the holotoxin variants on polarized human intestinal cells (T84 cell line) by measuring the short circuit current required to eliminate the potential difference induced by the cAMP-dependent Clsecretory response resulting from CT-A1-mediated ADP ribosylation of Gs␣ and constitutive activation of adenylate cyclase. Figure 4A to 4F shows the results of these experiments. Panel A shows that loss of even a single GM 1 BS attenuated the activity of cholera toxin to some degree, an effect which increased as the number of native GM 1 BS was lost. The time of onset of intoxication was also delayed as the number of native GM 1 BS decreased. Nevertheless, toxin with a single native GM 1 BS had clearly detectable activity. As expected, the G33D holotoxin with no native GM 1 BS had almost no activity over the time frame of the experiment: 1.5% of wt toxin signal over baseline at 90 min and less than 9% of wt signal at 120 min. Since these variants have differing numbers of C-terminal GSH6 tags (zero, one, or two), we also examined the effect that the number of tags had on the activity of wt holotoxin with five BS (Fig. 4B) . The presence of the tags modestly affected both the time of onset of intoxication, a measure of toxin-GM 1 trafficking from the PM to the ER of host cells, and the rate of increase of I sc . The effect was greater for doubly tagged holotoxin, although it eventually showed a similar maximal I sc . To control for these effects of the B-subunit tags, toxins with different numbers of native GM 1 BS but the same number of tags were compared. For these analyses, data were normalized by setting the maximal signal for the five native GM 1 BS toxins in each comparison to 1.00 (Fig. 4C to F) . For the toxin variants with one tag on the B subunit, the holotoxin with four native GM 1 BS exhibited a slight (25%) attenuation of toxicity relative to holotoxin with five native GM 1 BS, seen as a decrease in maximal I sc but with no delay in onset of intoxication (Fig. 4C) . A similar result (30% decrease of maximal I sc ) was seen for comparisons of the holotoxin variants with three or five native GM 1 BS and two CTB-GSH6-tagged subunits (Fig. 4D) . When the number of native GM 1 BS was reduced to two or one in doubly or singly wt B-subunit-tagged holotoxin variants, the peak I sc decreased by more than 50% and also the onset of intoxication was delayed ( Fig. 4E and 4F) , suggesting defects in entry of CT into the cell or transport to the ER (or both). As an initial step toward investigating whether CT variants with five and one native GM 1 BS trafficked from the cell surface to the ER by the same pathway, we examined the effects of brefeldin A (BFA) on their toxicity for T84 cell monolayers. We found that BFA completely inhibited the I sc (Fig. 5) induced by both single and five native GM 1 BS toxin variants. BFA-treated cells, however, still responded to addition of the cAMP agonist vasoactive intestinal polypeptide (VIP) at 90 min, showing that the toxin-treated cells retained viability and were competent for cAMP-dependent Clsecretion (I sc ). Thus, a toxin variant with a single native GM 1 BS, like native CT holotoxin, traffics through a BFA-sensitive pathway to exert its toxic effects. Previous studies have shown that holotoxins with reduced numbers of GM 1 BS are still able to intoxicate host cells, albeit with attenuated activity. The initial studies (26, 27) used chemical modification and denaturation-renaturation to generate mixed populations of holotoxins predicted to have one or two (nonnative but active) GM 1 BS. In our previous study (22) , we used genetic methods to create populations of chimeric holotoxins with only one or two completely native GM 1 BS and showed that these chimeras were still capable of intoxicating host cells but with attenuated activity. More recently, we studied how a membrane lipid might specify trafficking of toxin in the retrograde pathway by using fluorophore-labeled GM 1 and imaging toxin trafficking in live cells (28) . We found that a subset of GM 1 species, those with unsaturated ceramide domains, sorted efficiently from the PM to the trans-Golgi network and the ER. Cross-linking by toxin binding was dispensable for such GM 1 trafficking, but membrane cholesterol and the lipid raft-associated proteins actin and flotillin were required. Our results implicated an endogenous proteindependent mechanism of lipid sorting that is dependent on cer- Maximum signal for the tagged wt holotoxin in panels C through F was adjusted to 1.00, and all data points in the panel were normalized to that value. Data points show the mean response Ϯ SE; n ϭ 2 to 4 for A; n ϭ 3 to 4 for B; n ϭ 2 for C to F; each study was reproduced at least once. amide structure and could explain how the toxin gains access to the ER of host cells to induce disease (28) . This work shows that multivalent binding to GM 1 is indeed dispensable for CT toxicity. To do this, we prepared and purified homogenous preparations of holotoxins with defined numbers of GM 1 BS. Monomeric binding of CT to single molecules of GM 1 permits the holotoxin to intoxicate the host cell. Our results confirm and expand our studies on trafficking of the single GM 1 lipids. Conversely, loss of even a single GM 1 BS resulted in a measurable diminishment of toxicity in the T84 line of polarized human intestinal cells. This result is also consistent with findings of our studies of trafficking of the non-cross-linked GM 1 molecules, where we found that GM 1 cross-linking by toxin binding enhanced entry of CT into the ER. Thus, although multivalent binding to GM 1 by CT is fundamentally dispensable, it does have a significant effect on retrograde trafficking from the cell surface to the ER and on toxicity. One way that multivalent binding could affect CT function would be by enhancing the binding avidity for cell membranes containing GM 1 . In this and in our previous studies of the CT binding site mutants, we observe a loss of avidity for binding GM 1 when the GM 1 binding pockets are mutated. In the current studies, the loss of apparent avidity for the one-binding-site species is 10-fold, but the toxin still binds to GM 1 applied to the well at nanomolar concentrations, suggesting this may not fully explain the strong loss of toxin function. It is possible that the attenuated toxicity seen with holotoxins with reduced numbers of native GM 1 BS could be enhanced by the presence of one or more CTB-G33D polypeptides that confer increased stability to the holotoxins (and thus might decrease the delivery of the CT-A1 subunit to the cytosol). We raise this idea because we observed that insertion of even a single CTB-G33D monomer into the pentameric B subunit rendered the holotoxin partially resistant to dissociation by SDS, and increasing the number of CTB-G33D monomers resulted in holotoxins that were completely resistant to dissociation by SDS. Conversely, there is no a priori reason to assume that the presence of one or more CTB-G33D polypeptides would affect chaperone-mediated release of CT-A1 from the nicked and reduced holotoxin in the ER, which appears to facilitate CT-A1 retrotranslocation from the ER to the cytosol (29) . Further experiments will be required to test these possibilities. Another way that multivalent binding to GM 1 might affect CT function would be through reorganization of membrane structure and function induced by scaffolding GM 1 into ceramide-based nanodomains, as suggested by studies in vitro and in vivo (14, 15, (30) (31) (32) (33) . Significantly, studies of the closely related AB 5 subunit Shiga toxin show that multivalent binding to the glycosphingolipid GB3 spontaneously induces high-curvature membrane tubules by coupling the toxin-lipid complex to membrane shape (31) . Multivalent binding of CT to GM 1 in model membranes also induces spontaneous membrane curvature, implying a similar coupling of the toxin-glycosphingolipid complex formation to membrane shape (32, 34, 35) . This could allow partitioning of the CT-GM 1 complex into highly curved sorting tubules of the sorting endosome, which is required for retrograde transport, and explain why the toxins with more than 1 GM 1 binding site are more potent in intoxication. It is also possible that multivalent binding to GM 1 may be needed to activate intracellular signaling pathways that enhance uptake and trafficking, as for Shiga toxin (36) . Effects either on binding avidity or on membrane structure and trafficking dynamics could underlie how the CTB subunit, and the other AB 5 toxins, evolved as pentameric structures. This and our other recent studies, however, show that the most fundamental function exploited by CT is to coopt an endogenous glycosphingolipid sorting pathway from the PM to the ER that is essential for toxin entry into the cytosol and the induction of disease. Bacterial strains and construction of expression clones. All chimeric toxins in this study were produced by expression from recombinant plasmids in Escherichia coli BW27784 (araFGH pCP18-araE [37] ). This strain does not metabolize arabinose and constitutively expresses the arabinose transporter, and therefore a uniform degree of induction occurs in all cells of a culture at any arabinose concentration. To produce the expression strains, genes encoding native-length CTB or carboxy-terminally extended CTB tagged with a glycosylation-sulfation signal (24) followed by a hexahistidine peptide were cloned on compatible plasmids with different selectable markers and inducible promoters. These encoded resistance to ampicillin (Ap) with the IPTG-inducible lacUV5 promoter or resistance to chloramphenicol (Cm) and the arabinose-inducible araBAD promoter, along with the regulator araC. Variants of these CTB clones were also made with G33D substitutions that eliminate GM 1 binding. A separate plasmid (pSlacbadCTA) selectable with spectinomycin (Sp) and compatible (with a pSC101 origin) with both ctxB plasmids was used to independently express the ctxA gene under control of both the lac and araBAD promoters. Each expression strain thus contained three plasmids, encoding a native CTB subunit (wt or G33D variant, Ap, and lac promoter), a GSH6-tagged CTB subunit (G33D or wt, Cm, and ara promoter), and the wt CTA subunit (Sp and ara and lac promoters). Details of construction, including full DNA sequences, are available on request. Plasmids created or used in this study are listed in Table 1 . Plasmids pAlacCTB and pAlacCTB G33D encode Ap resistance and the wt and G33D native-length variants of CTB, respectively, under control of the lac promoter and have previously been published as pLMP1 and pLMP148, respectively (38) . Carboxy-terminally linker-hexahistidine-tagged variants of CTB and CTB-G33D were made in a compatible Cm r arabinoseinducible vector, pAR3 (39), using the hexa-His tag from pT7sh6 (40) and a glycine-rich repeat from the M13 phage PIII protein, encoding XX (EX) 4 XDPRVPSS (where X is GGGS) inserted between residues 102 and 103 of CTB. Initial experiments to make wt and CTB-G33D mixed pentamers were done using these plasmids, but we saw significant proteolysis of the polyglycine linker-tagged variants, and subsequent experiments were done with variants that had the linker replaced with the GSH6 tag-pCbadCTBGSH6 and pCbadCTB G33D GSH6. Toxin expression and purification. A 400-ml LB culture at 30°C was inoculated with a 1/8 volume of an overnight culture with appropriate antibiotic selection (Sp and Ap at 100 g/ml and Cm at 25 g/ml) and grown to an A 600 of 1.2, when it was induced with 400 M IPTG and 0.0005% l-arabinose with incubation continued overnight. Cells were collected by centrifugation, resuspended in 20 ml phosphate-buffered saline (PBS), and lysed by mixing for 20 min at room temperature with 0.5 mg/ml lysozyme and Elugent detergent (EMD Biosciences, Inc., La Jolla, CA) to 2%. Viscosity was reduced by sonication four times (10 s each) on ice, and the lysate was cleared by centrifugation at 15,000 rpm for 20 min in an SS34 rotor. Toxin was purified from the supernatants by Talon chromatography as detailed by the manufacturer (Clontech Laboratories, Inc., Mountain View, CA). Imidazole eluates were dialyzed against 50 mM Tris-HCl (pH 8.0) (buffer A). All chromatographic separations were conducted on an Akta purifier (GE Healthcare Biosciences, Pittsburgh, PA) in 4.6-mm by 100-mm Poros perfusion chromatography columns packed with Poros HS 20 cation exchange medium or Poros HQ 20 anion exchange medium. For cation exchange, the column was equilibrated with 5 column volumes (CV) buffer A and washed with 5 CV buffer A after sample loading. Bound material was eluted with 10 CV of a linear gradient of 0 to 100% buffer B (50 mM Tris-HCl, pH 8.0, 1M NaCl). For anion exchange, the column was equilibrated with 5 CV buffer A, washed with 5 CV of buffer A after sample loading, and was eluted with a 40-CV (initial separation) or 10-CV (second purification) linear gradient of 0 to 100% buffer B. Anion exchange fractions were pooled and concentrated using Microcon Ultracel YM-10 filter devices (EMD Millipore, Billerica, MA). Ganglioside GM 1 ELISA. Ninety-six-well microtiter plates were coated overnight with 50 l of 2-fold serial dilutions of ganglioside GM 1 (Supelco, Sigma-Aldrich, St. Louis, MO) in PBS, starting at 150 nM, and then blocked with 10% horse serum in PBS. All steps were performed at 37°C for 1 h, after which samples were aspirated off and wells were washed three times with PBS plus 0.05% Tween 20. Each holotoxin was assayed in triplicate by incubating 100 l of 50-ng/ml holotoxin per well (approximately 60 fmol), followed by 1/5,000 rabbit anti-cholera toxin B subunit (B10) and then 1/2,000 HRP-conjugated goat anti-rabbit serum (Thermo Fisher Scientific Inc., Rockford, IL), followed by o-phenylenediamine (OPD) substrate (Sigma-Aldrich), and the reaction was stopped with 3 M HCl. Absorbance was read at 450 nm. Y1 assay. Mouse Y1 adrenal cells (ATCC CCL-79) were grown at 37°C in a humidified 5% CO 2 atmosphere in RPMI medium with 5% fetal bovine serum with 1ϫ penicillin-streptomycin (Life Technologies), and 96-well plates were seeded with 10 4 cells per well. One-hundredmicroliter volumes of 2-fold serial dilutions of toxins were added to semiconfluent monolayers and incubated overnight, followed by scoring for toxin-induced rounding. One unit of toxin was defined as the amount contained in the last dilution giving 75 to 100% rounding of cells. T84 cells and electrophysiology. Measurements of short-circuit current (I sc ) on monolayers (0.33-cm 2 inserts) of the polarized human intestinal cell line T84 were performed as previously described (41) .

A Single Residue Substitution in the Receptor-Binding Domain of H5N1 Hemagglutinin Is Critical for Packaging into Pseudotyped Lentiviral Particles

BACKGROUND: Serological studies for influenza infection and vaccine response often involve microneutralization and hemagglutination inhibition assays to evaluate neutralizing antibodies against human and avian influenza viruses, including H5N1. We have previously characterized lentiviral particles pseudotyped with H5-HA (H5pp) and validated an H5pp-based assay as a safe alternative for high-throughput serological studies in BSL-2 facilities. Here we show that H5-HAs from different clades do not always give rise to efficient production of H5pp and the underlying mechanisms are addressed. METHODOLOGY/FINDINGS: We have carried out mutational analysis to delineate the molecular determinants responsible for efficient packaging of HA from A/Cambodia/40808/2005 (H5Cam) and A/Anhui/1/2005 (H5Anh) into H5pp. Our results demonstrate that a single A134V mutation in the 130-loop of the receptor binding domain is sufficient to render H5Anh the ability to generate H5Anh-pp efficiently, whereas the reverse V134A mutation greatly hampers production of H5Cam-pp. Although protein expression in total cell lysates is similar for H5Anh and H5Cam, cell surface expression of H5Cam is detected at a significantly higher level than that of H5Anh. We further demonstrate by several independent lines of evidence that the behaviour of H5Anh can be explained by a stronger binding to sialic acid receptors implicating residue 134. CONCLUSIONS: We have identified a single A134V mutation as the molecular determinant in H5-HA for efficient incorporation into H5pp envelope and delineated the underlying mechanism. The reduced binding to sialic acid receptors as a result of the A134V mutation not only exerts a critical influence in pseudotyping efficiency of H5-HA, but has also an impact at the whole virus level. Because A134V substitution has been reported as a naturally occurring mutation in human host, our results may have implications for the understanding of human host adaptation of avian influenza H5N1 viruses.

H5N1 influenza virus is highly pathogenic in poultry, certain bird populations, and has occasionally infected human causing severe clinical outcomes [1] [2] [3] . Since the first human outbreak in 1997, there have been more than 600 confirmed human cases of H5N1 infection with a mortality rate of approximately 60% [4] . To initiate an infection, like all other subtypes of influenza viruses, H5N1 virus first binds to cell surface glycan receptors via its surface glycoprotein hemagglutinin (HA) and is subsequently internalized via endocytic pathways [5] [6] [7] . HA is a homotrimeric type I transmembrane glycoprotein, which can be cleaved into HA1 and HA2 subunits [8] . A furin-dependent polybasic cleavage site has been shown to be characteristic of highly pathogenic avian influenza viruses [9, 10] , although not all H5-HAs contain the polybasic cleavage site. In cells infected by influenza virus, HA protein is first synthesized as a precursor (HA0), which is then oligomerized, glycosylated and ultimately transported to the plasma membrane where assembly and budding of progeny virions takes place [11] . Then, during the final stage of the virus life cycle, neuraminidase (NA), the second major envelope glycoprotein of influenza, cleaves the terminal sialic acids from the cell surface glycans to allow release of the virus from the host cell [12] . Following the first report of H5N1 outbreak in 1996, the virus has evolved into different clades as shown by the phylogenetic analysis of H5-HA protein sequences [13] [14] [15] . Currently the spread of H5N1 virus in human population is limited. However, through mutation and reassortment, the virus may become more easily transmissible from bird to human or from human to human, posing a potential pandemic threat to public health worldwide [2, 3] . It is therefore important to fully understand the biology of H5N1 viruses and to develop sensitive and rapid diagnostic methods. However, an obstacle to the study of H5N1 viruses is the stringent safety requirement to work with them. Recently, we and other research groups have developed retroviral particles pseudotyped with H5-HA (H5pp) as an alternative strategy for large scale serological studies [16] [17] [18] [19] [20] [21] . Similar to the replication-competent virus, H5pp entry requires alpha-2,3 sialic acids, is pH-dependent, and can be neutralized by sera containing anti-H5N1 antibodies [18] , thus validating H5pp as very useful and safe tool for a wide range of applications, including entry mechanism studies, serodiagnosis and drug discovery [16, 18] . In our previous work, we have produced H5pp using the H5-HA of A/Cambodia/40808/2005 (H5Cam), which was isolated from a patient with a lethal infection of H5N1 virus [18] . In the current study, we have analyzed the ability of H5-HAs from different clades of avian influenza virus to pseudotype lentiviral particles and have found that they do not give rise to the same level of efficient H5pp production when compared with H5Cam. In particular, we have carried out a detailed comparison of the expression and cleavage of two H5-HAs, i.e., H5-HA of A/Anhui/ 1/2005 (H5Anh) and H5Cam, and of their ability to pseudotype lentiviral vector in HEK293T cells. Through several independent lines of evidence we have identified the molecular determinants in H5-HA for efficient incorporation into H5pp envelope and have delineated the underlying mechanism. Our results are discussed in the context of the understanding of human host adaptation of avian influenza H5N1 viruses. The ability of H5-HA to pseudotype lentiviral particles does not correlate with HA protein expression level in producer cells Similar to HA of other subtypes of influenza viruses, H5-HA is highly mutable as a result of antibody-selection pressure, leading to the rise of divergent H5N1 viruses that are categorized into various strains and clades [13, 14, 22] . To ascertain the flexibility and adaptability of H5pp production as an alternative approach for serological studies in the event of novel emerging H5N1 viruses, we sought to develop cladespecific H5pp and compared the ability of three other H5N1 viruses belonging to different clades to pseudotype lentiviral particles. H5-HA from clade 1 (H5Cam), clade 2.1 (H5Ind), clade 2.2 (H5Qin) or clade 2.3 (H5Anh) (see Table 1 ) was expressed in 293T cells together with lentiviral backbone plasmid to allow the production of H5-pseudotyped lentiviral particles (H5pp). Expression levels of H5-HAs in transfected 293T cells was monitored by Western blot using anti-FLAG antibody directed against the C-terminal tag (Fig. 1A , upper panel). Supernatants containing H5pp were harvested 48 hr post-transfection, and used to transduce MDCK cells for luciferase reporter activity assay (Fig. 1A , lower panel), as described in Materials and Methods. Unexpectedly, we observed significant differences in the transduction of MDCK cells by H5pp, depending on the clades of H5-HAs. In particular, H5Anh from A/Anhui/2005/01 resulted in very low luminescence levels after particle transduction in MDCK cells; whereas H5Cam from A/Cambodia/40808/2005 was the most efficient, inducing a consistent 3-4 log increase in luciferase activity compared with H5Anh ( Fig. 1A -B, lower panels). Analysis of cell lysates by Western blots, however, demonstrated that all H5-HAs tested were well expressed in the producer cells and, consequently, that luciferase reporter activity in MDCK target cells did not correlate with the level of HA protein expression in the cells (Fig. 1A) . Two main protein bands were detected, consistent with the expected electrophoretic mobility of the uncleaved protein (HA0) and the C-terminal portion of the cleaved form (HA2 subunit), whereas the N-terminal fragment (HA1 subunit) could not be recognized by the anti-FLAG antibody due to C-terminal tagging (Fig. 1A) . We next decided to compare in detail the behaviour of H5Cam and H5Anh. To determine whether the difference in luciferase reporter activity was due to the level of H5pp production, culture supernatants containing H5Cam-pp and H5Anh-pp were concentrated by ultracentrifugation, and the resulting H5pp pellets were analyzed by Western blotting. Our results showed that the number of particles produced in the culture supernatant was significantly less for H5Anh than for H5Cam in presence of soluble bacterial neuraminidase, as indicated by lower levels of p24 in concentrated supernatants to detect the lentiviral core and lower luciferase reporter activities in MDCK cells (Fig. 1B) . More importantly, incorporation of H5Anh into the pseudotyped lentiviral particles was not observed using anti-FLAG antibody (Fig. 1B, upper panel) . Altogether, these data suggest that H5Anh cannot be efficiently incorporated into pseudotyped particles and released into the cell culture supernatant. Swapping of HA2 domain (including the polybasic cleavage site) does not increase production of H5Anh-pp Sequence analysis of the polybasic cleavage site reveals that H5Anh has a deletion of a lysine residue when compared to H5Cam and moreover, there is an additional amino acid difference in the HA2 region at position 533, which is located at the border between the ecto-domain and the transmembrane domain (TMD) ( Fig. 2A) . Thus, H5Cam has an isoleucine at position 533 (I533), while H5Anh has a threonine (T533). Cleavage of HA into HA1 and HA2 subunits by host protease is a critical step for influenza viruses to gain membrane fusion capability [23, 24] ; whereas the TMD of HA is important for its association with lipid rafts at the plasma membrane [25] . To test the potential influence of these differences in the cleavage site and at position 533, we generated several chimerical constructs in which either the entire HA2 region including the cleavage site was replaced with that of H5Cam (AnhCam1), or only the cleavage site (AnhCam2) or a single T533I amino acid change was introduced (AnhCam3). All constructs were FLAG-tagged at the C-terminal end of H5 sequences as described in the Materials and Methods section ( Fig. 2A) . When transfected into 293T cells, all three mutant H5Anh proteins were well expressed in the producer cell lysates (Fig. 2B) ; however, analysis of transduction levels of MDCK target cells by H5pp produced with these H5AnhCam chimerical proteins suggests that none of them was able to increase the production of pseudotyped particles (Fig. 2C) . These data indicate that differences in the HA2 domain cannot account for the reduced ability of H5Anh to form pseudotyped particles. Two amino acid substitutions in the 130-loop of the receptor binding domain (RBD) are sufficient to confer H5Anh pseudotyping ability Sequence alignment of H5-HA proteins revealed a striking amino acid divergence at 9 positions over a short stretch of only 33-amino-acid-long region around the 130-loop of RBD (Table 1) , which accounted for over 30% of the amino acid differences found in the entire HA molecule (576 amino acids in length). Therefore this region was chosen for site-directed mutagenesis to generate a series of H5Anh mutants that were subsequently tested for their ability to pseudotype lentiviral vectors. The level of protein expression for all H5Anh mutants in producer cells was comparable, albeit slightly lower for AnhM1 and AnhM6 (Fig. 3A) . Interestingly, all H5Anh mutants that harbored residues alanine-valine at positions 133-134 (AnhM1-5, Table 1 ) displayed a largely restored ability of H5Anh to produce pseudotyped particles, despite other sequence differences at the 130-loop flanking region (Fig. 3B ). By contrast, AnhM6, which contains H5Cam-like 130-loop flanking sequences but serine-alanine at positions 133-134, did not generate efficiently H5Anh-pp in culture supernatant (Fig. 3B) . These data clearly demonstrate that amino acid residues at positions 133-134 are crucial for efficient H5pp production. More specifically, substitution of the two amino acids S133-A134 of H5Anh with A133-V134, which are unique to H5Cam sequence, confers H5Anh the ability to be incorporated into the lipid envelope of lentiviral particles and is essential for efficient production of transduction-competent H5Anh-pp. A single A134V mutation in the 130-loop of the RBD of H5-HA is the critical determinant for H5pp production To further delineate the respective roles of A133 and V134 for efficient pseudotyping, two additional H5Anh mutants were generated with a single amino acid substitution either at position 133 (S133A, AnhM7) or 134 (A134V, AnhM8) ( Table 1) . These experiments revealed that the A134V mutation was sufficient to confer H5Anh the ability to be incorporated into transductioncompetent pseudo-particles; whereas the S133A mutation was not (Fig. 4A) . We also generated reciprocal mutants of H5Cam that contained either the two residues found at positions 133-134 of H5Anh (viz., S133-A134; CamM1), or only one single V134A change (CamM2), or A133S substitution (CamM3) ( Table 1) . Again, the presence of valine at position 134 was found to be crucial for efficient H5pp production, whereas the A133S substitution had only a marginal effect, consistent with the results obtained with H5Anh mutants (Fig. 4A ). To confirm that the effect of valine at position 134 was indeed on the production of H5pp, we analyzed by Western blot both cellular lysates and culture supernatants containing H5pp that were concentrated by ultracentrifugation (Fig. 4B ). This series of experiments showed that incorporation of the ''Anhui-like'' single mutant CamM2 into the pseudotyped lentiviral particles was below the antibody detection limit, as also seen for wild-type H5Anh; whereas the ''Cambodialike'' single mutant AnhM8 induced H5pp production with an efficiency similar to H5Cam ( Fig. 4A-B) . Altogether, these experiments demonstrate that valine at position 134 (V134) is a critical residue for efficient H5pp production. The A134V mutation affects cell surface expression level of HA Because influenza virus, as well as particles pseudotyped with HA, buds from the plasma membrane [11, 18] , we reasoned that changes in surface expression of H5-HA could have an impact on the production of H5pp. Thus, we have compared by flow cytometry plasma membrane expression levels of HA protein in cells transfected with H5Anh and H5Cam. Cells were labelled with an anti-H5N1 antibody, fixed and then stained with a PEconjugated secondary antibody. As assessed by measuring mean fluorescence intensity (MFI), cell surface expression of H5Anh was significantly less compared to H5Cam ( Fig. 4C ; p,0.01). Interestingly, introduction of the A134V mutation into H5Anh (AnhM8) increased its cell surface expression ( Fig. 4C ; p,0.02), and conversely, a V134A mutation in H5Cam (CamM2) reduced transport to the plasma membrane to a level that was not significantly different from that measured with H5Anh (Fig. 4C ). These data demonstrate that the Ala to Val substitution at position 134 enhances surface expression of H5-HA. As residue 134 is in the 130-loop of the receptor binding site, we next investigated the impact of A134V mutation on receptor binding properties. We employed a cell-based assay using soluble H5-HA proteins that were engineered by removing TMD and Ctail of HA (Fig. 5A ) as described in Material and Methods. Stable cell lines were generated to express sH5Anh, sH5Cam and their reciprocal single amino acid mutant forms (sH5AnhM8 with A134V and sH5CamM2 bearing V134A; see also Table 1) , and soluble HA proteins were affinity-purified as described under Material and Methods. When analyzed on native gels, purified soluble H5-HA proteins contained mostly the homotrimeric form ( Fig. 5A ) that can bind to the sialic acid-containing cellular receptors. We observed that sH5Anh bound strongly to MDCK cells, whereas the A134V mutation reduced the binding to a much lower level (Fig. 5B ). By contrast, sH5Cam bound weakly to MDCK cells and, as predicted, the single V134A change induced a major increase in the binding of sH5Cam to MDCK cells (Fig. 5B) . The binding assay was also performed in MDCK-SIAT-1 cells which express two-fold higher amounts of alpha-2,6-link sialic acids than parental MDCK cells [26] . The results obtained were similar to that in parental MDCK cells ( Fig. 5C-D) . When cells were treated with bacterial neuraminidase NAvb before fixation with PFA, the binding of sHA proteins was diminished to background level in both MDCK and MDCK-SIAT-1 cells, indicating that the binding of sHA proteins is sialic acid dependent ( Fig. 5C-D) . Because we had previously found that H5pp with HIVbackbone bud from the plasma membrane in 293T cells [18] , a reduced cell surface expression of viral envelope proteins (see Fig. 4C ) would be expected to influence the formation of pseudotyped particles and, hence, could account for the observed differences in pseudotyping. It has been reported that retroviruses including HIV and Murine Leukemia Virus (MLV) can also bud from intracellular compartments [27, 28] , depending on the cell type and Gag expression systems. Therefore, we also used MLVbased pseudotyping system to compare the efficiency of H5pp production between H5Anh and H5Cam. As demonstrated in Fig. 6A , the results obtained with the MLV-backbone were similar to those with HIV-backbone, thus, indicating that inefficiency of H5Anh-pp production is not a mere consequence of the lentiviral system used for pseudotyping. However, co-transfection of the viral NA from A/Cambodia/JP52a/2005, rescued the inefficiency of H5Anh-pp production (Fig. 6A ). To further test whether reduced binding to sialic acid receptors, as a result of A134V mutation, is a major contributing factor for pseudotyping efficiency of H5-HA, we examined the production of H5Cam-pp and H5Anh-pp in Lec2 cells which are sialylationdeficient mutants of CHO cells [29] . As H5Cam binds weakly to sialic acid receptors (Fig. 5) , NAvb added exogenously post transfection was sufficient to release H5Cam-pp into culture supernatant in CHO cells; and the level of H5Cam-pp in CHO cells was not significantly different from that in Lec2 cells (Fig. 6B ). Similar to the results obtained in 293T cells, production of H5Anh-pp was lower than H5Cam-pp in CHO cells. By contrast, H5Anh-pp production in Lec2 cells was significantly increased in comparison to that in parental CHO cells and the level of H5Anhpp obtained in Lec2 cells was similar to that of H5Cam-pp, as indicated by the values of luciferase activity detected in MDCK cells 72 hr post H5pp transduction (Fig. 6B) . Together, these findings further suggest that binding of H5-HA to cellular sialic acid containing glycans is a major determinant of H5-HA incorporation into pseudo-particles. H5N1 viruses carrying the A134V mutation exhibit reduced capability to agglutinate horse red blood cells The reduced binding to sialic acid receptors as a result of the A134V mutation not only leads to changes in pseudotyping efficiency, but is also found to have an impact at the whole virus level. Reverse genetics generated RG-A/Cambodia/408008/2005 with the A134V mutation has been shown to agglutinate to the same degree both human red blood cells (RBCs), which express alpha-2,6-sialic acid, and guinea pig RBCs, which exhibit both alpha-2,3 and alpha2,6-linked sialic acid, but failed to agglutinate horse RBCs, which carry only alpha-2,3-sialic acid [30] . These observations provide an experimental evidence to support the notion that the A134V mutation leads to a reduced alpha-2,3sialic acid binding of the virus. To confirm the effect of A134V mutation on H5N1 viruses, we performed similar hemagglutination assays using another H5N1 virus strain A/Cambodia/ V0401301/2011, which also contains the same A134V mutation. In previous studies, we have reported the generation of H5pp and have characterized it as a safe alternative to the use of replicative H5N1 virus for sero-surveillance [16, 18] . Because H5pp mimics the entry mechanism of the avian virus while carrying only the H5-HA as envelope protein, it offers the advantage to be specifically neutralized only by anti-hemagglutinin antibodies, avoiding the confounding effect of antibodies directed against N1 neuraminidase due to infection of influenza virus subtypes other than H5N1. We report here that the efficiency to generate HA-only H5pp varies with HAs derived from different H5N1 virus clades, regardless of the lentiviral backbone used. Through serial mutagenesis of two H5-HAs, we have uncovered that differences in receptor binding ability, due to mutations in the receptor-binding domain of HA, may be the underlying mechanism. It is widely believed that HA is targeted to lipid rafts at the plasma membrane and the transmembrane domain has been described to be important for lipid rafts association of HA [25] . Therefore, we first swapped the transmembrane regions between H5Anh and H5Cam. We also noticed that the cleavage of H5Cam appears to be more efficient (Fig. 1A, 2B) . Thus, mutants with or without sequence variations found at the poly-basic cleavage site (AnhCam1, AnhCam2 and AnhCam3) were generated and analysed. However, none of these H5Anh mutants showed appreciable improvement in their ability to generate H5pp, when compared with wild type H5Anh. In fact, the production of H1 and trypsin-dependent H5 pseudo-particles has been reported [31, 32] , hence indicating that HA cleavage is not a determining factor for pseudotyping efficiency. Then by multiple sequence alignment, we identified a small region around the 130-loop of the receptor binding site of HA which appeared to be a ''hot-spot'', harboring several sequence variations among different H5N1 clades. Through a series of mutagenesis studies, we have found that one single residue at position 134 is a critical switch to dictate the ability of H5 HA to pseudotype lentiviral vectors for the production of H5pp. Similar to influenza virus, H5pp generated with an HIVbackbone bud at the plasma membrane [18] ; therefore the simplest explanation is that the mutation at position 134 may result in a change in cell surface expression of HA. Indeed we have observed a small but consistent change in cell surface HA expression due to mutations at position 134 (Fig. 4C) . To exclude the possibility that this finding merely reflected a differential binding to the two HA of the rabbit anti-H5 polyclonal serum (described in Material and Methods), we used another polyclonal serum from a different source (a duck anti-H5 serum described in Ref. 18 ) and found that the results of cell surface HA staining were similar (data not shown). The fact that the A134V mutation increased cell surface expression of H5Anh, may partially explain the effect of this amino acid substitution on H5pp production. Considering that the variation between H5Cam-pp and H5Anhpp production resulted in a 3 to 4 log difference in luciferase activity, it is likely that A134V mutation may have an impact on other properties of H5-HA, including binding to sialic acid receptors, which contribute to the observed phenotype. It has been reported in the case of H3-HA pseudotyping that lentiviral particles which incorporate sialic acid binding-incompetent H3-HA (derived from A/Aichi/2/68) can be efficiently generated and released into culture supernatant in the absence of exogenous bacterial NA; whereas the wild-type Aichi-HA fails to do so [33] . Although the difference in pseudotyping observed with wild-type Aichi-HA and its receptor binding-incompetent mutant is diminished when bacterial NA is added, the study by Bosch et al. [33] implies that changes in receptor binding properties can affect pseudotyping efficiency of lentiviral vectors by influenza HA. Regarding the potential influence of mutations at position 134 of H5-HA on receptor binding properties, there have been reports with contradictory results. First, Yamada et al. [34] found that A134T mutation did not change alpha-2,3 sialic acid binding preference of H5-HA. Then, Auewarakul et al. [35] reported that L129V/A134V allowed for dual binding to both alpha-2,3 and alpha-2,6-sialic acid receptors, although in their study, the effect of A134V mutation alone was not assessed. More recently, using virus elution assay, Imai and colleagues [36] found that H5N1 viruses containing alanine at position 134 (A134) show stronger binding than those harbouring threonine (T134) to both chicken erythrocytes (expressing both alpha-2,3 and alpha-2,6-sialic acid) and horse erythrocytes (expressing only alpha 2,3-sialic acid). Similar to the observation by Imai et al., we found in the current study that H5Anh which contains A134 displayed a strong binding to both MDCK and MDCK-SIAT-1 cells (expressing an increased level of alpha-2,6 and a decreased level of alpha-2,3-sialic acid than parental MDCK) [26] . As predicted by these observations, the A134V mutation reduced H5Anh binding to a dramatically lower level in both cell lines. It is likely that strong binding of H5Anh to cell surface sialic acid receptors makes it difficult to release H5pp from the producer cells even in the presence of exogenous bacterial NA; and the A134V mutation reduces binding, thus allowing for the release of H5pp. In keeping with this hypothesis, co-transfection with the viral NA gene from H5N1 led to the production of similar amounts of mixed HA-NA pseudoparticles for both H5Anh and H5Cam. Moreover, we did not observe an increase in binding to MDCK-SIAT-1 cells, which contain more alpha-2,6-sialic acids on the cell surface. In fact both sH5Anh and sH5CamM2 bind with slightly lower efficiency to MDCK-SIAT-1, compared with parental MDCK cells, suggesting that A134V mutation probably leads to a decreased binding of H5-HA to alpha-2,3-sialic acid rather than a switch to alpha-2,6sialic acid binding. Consistent with this notion, we observed an increased level of H5Anh-pp production in Lec2 sialylationdeficient cells, when compared with parental CHO cells (Fig. 6B) . We have found that the A134V mutation not only exerts a critical influence in the determination of pseudotyping efficiency, but has also an impact on H5N1 viruses. Both A/Cambodia/ 408008/2005 and A/Cambodia/V0401301/2011, two different H5N1 isolates carrying the same A134V mutation could agglutinate human and guinea pig RBCs but failed to agglutinate horse RBCs [30] (also Figure 7 of this paper); whereas two other strains of H5N1 viruses without the A134V mutation could also agglutinate horse RBCs (Figure 7) . These observations indicate that A134V mutation in H5-HA reduces virus binding to alpha-2,3-sialic acid. Although co-transfection with viral NA enables efficient lentiviral pseudotyping by H5Anh (Fig. 6A) , the differential RBC binding properties observed at the whole virus level, when both HA and NA are present, support the idea that A134V mutation in H5-HA can be biologically relevant. Interestingly, alanine at position 134 (A134) is highly conserved in avian H5N1 viruses and so far A134V mutation has only been found in human isolates of H5N1 viruses, both clade 1 and clade 2 viruses isolated from 2004 to 2011. Almost all avian H5N1 isolates possess A134 in the HA. So far only one avian H5N1 virus in the NCBI database has serine instead at position 134 of the HA protein. Notably, more diversity is observed at this position for human isolates of H5N1 viruses: three H5N1 viruses isolated from human patients have a threonine and eleven a valine at position 134 [36] . At least in two cases (A/Cam/408008/2005 and A/ Thailand/676/2005), viruses found in the original patient specimens were mixtures of both wild type, containing A134 in the HA, and mutant virus, containing V134 [30, 35] . It is possible that other human isolates of H5N1 viruses may actually contain the A134V mutation but failed to be detected in the process of either virus isolation or traditional capillary sequencing of viral genomes. Thousands of H5-HA sequences are available in the NCBI Influenza Database (http://www.ncbi.nlm.nih.gov/ genomes/FLU/FLU.html) from non-human isolates of H5N1 viruses, none of which contains this particular mutation. Altogether, these observations suggest that a valine at residue 134 of the receptor-binding domain is unlikely to be a random sequence variation but may be selected as the avian H5N1 viruses adapt for replication in human hosts. We speculate in general terms that changes in cell surface receptor binding of H5-HA, as a result of A134V mutation, may lead to changes in virus entry and virus release and, therefore, be considered an important factor for determination of host range. It is not clear whether intracellular sialic acid content and distribution may also influence this feature. As our data focus on the pseudotyping system, further studies are required to understand more precisely the biological consequences of A134V mutation and its potential influence on the adaptation of H5N1 viruses in humans. Our findings have also implications for the applicability of H5pp assay in serological surveys. H5pp has several advantages over the microneutralization method, which is the current gold standard serological assay for the detection of antibodies against avian influenza viruses [37, 38] . Pseudotyped particles are produced from synthetic genes without the need to have access to the virus and can be safely used in BSL-2 containment, making them ideal for widespread use, especially in areas where BSL-3 facilities are not available. Moreover, it has been reported that the H5pp assay is more sensitive than micro-neutralization [16, 39] . It appears, however, that its use in sero-epidemiological studies and to monitor the efficacy of candidate vaccines may be limited by the strain under investigation. Although we and others have found that production of mixed HA-NA pseudo-particles is consistently successful using N1 from either H1N1 or H5N1 [17, 20, 21] (also Fig. 6A of this paper) , the production of HA-only pseudotypes would be necessary to eliminate potential cross-reactivity that may be displayed by circulating anti-NA antibodies against N1 from avian H5N1 or seasonal influenza H1N1 [40] . We are cognizant that a single A134V mutation may result in a change of antigenicity but this limitation is not different from that of the microneutralization assay, which is the gold-standard to detect anti-H5 neutralization antibodies and utilizes an available H5N1 virus strain that may not be a perfect match of the viruses associated with the serum samples being tested. In fact, serologic surveys often use a collection of serum samples from human or animals without necessarily knowing the exact virus strain(s) involved. Moreover, if positive, samples shall contain polyclonal antibodies against multiple epitopes to an H5N1 virus, further minimizing the likelihood that a single amino acid substitution would compromise the usefulness of a pseudotype-based serological assay as a safer alternative to the microneutralization test. Careful assessment of the H5pp-based assay should obviously be performed when new strains emerge. In conclusion, by comparing the ability of different H5-HA to produce pseudotyped particles, we have demonstrated that when a single A134V mutation is introduced in the receptor binding site, the ability of the usually inefficient H5Anh to generate H5pp is largely restored. It is likely that the A134V mutation leads to an increased level of cell surface HA expression and reduced binding to sialic acid receptors, both of which contribute to the production of H5pp. The A134V mutation has been reported as a naturally occurring mutation in human host; and importantly, this mutation is so far only found in human isolates of H5N1 viruses. Our data with hemagglutination assays further demonstrate that viral isolates from human cases with avian influenza carrying the A134V substitution exhibit a reduced binding to alpha-2,3 linked sialic acids. Therefore, our results may have implications for the understanding of human host adaptation of avian influenza H5N1 viruses. It is possible that other mutations leading to reduction in receptor binding may exist and cause a change in pseudotyping efficiency. Thus, H5pp production together with soluble HA protein cell binding analysis may serve as convenient functional assays to screen for mutations with potential consequences on receptor binding properties and host adaptations of H5N1 viruses. Although zoonotic transmission from poultry to humans remains inefficient for H5N1, it may be of importance to monitor closely mutations in regions of the receptor binding site of H5-HA. 293T, MDCK, CHO and Lec2 cell lines were obtained from ATCC (Manassas, VA, USA). MDCK-SIAT-1 cells were generated by stable transfection of human alpha-2,6-sialyltransferase in MDCK cells and was described elsewhere [26] . This cell line overexpresses alpha-2,6-linked sialic acid compared to parental MDCK [26] . 293T, MDCK and CHO cells were cultured at 37uC with 5% CO2 in Dulbecco's Modified Eagle's Medium (DMEM, Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS, Invitrogen) and 1% penicillinstreptomycin. MDCK-SIAT-1 cells were grown in DMEM containing 10% FBS and 1 mg/ml G418. Lec2 cells, which lack terminal sialic acid in their glycoproteins and gangliosides due to a defect in the CMP-sialic acid transporter [29] , were cultured in Minimum Essential Medium (MEM-alpha, Invitrogen) supplemented with 10% FBS and 40 ug/ml L-proline. H5Cam (HQ664938, see also ref. 30 ), H5Anh (ABD28180), H5Ind (ABP51969), H5Qin (ABE68923)), mutants AnhM1-M6, CamM1-M3 (Table 1 ) and N1 gene (ABO10176) from A/ Cambodia/JP52a/2005 were synthesized as human codon optimized genes (GENEARTH, Regansburg, Germany) and subcloned into the mammalian expression vector pcDNA3.1 (Invitrogen). Mutants AnhM7 and AnhM8 were generated by sitedirected mutagenesis using QuikChange site-directed Mutagenesis Kit (Stratagene, Santa Clara, CA, USA) according to the manufacturer's instructions. To generate soluble H5-HA constructs, the transmembrane domain (TMD) of the HA was removed, and the polybasic cleavage site was changed into a monobasic cleavage site RESR by site-directed mutagenesis to avoid the potential influence of H5-HA cleavage in cells on the purification of sHA proteins, which involve multiple steps. The truncated HAs were then subcloned into pcDNA3.1 (Invitrogen). All H5 plasmids were tagged with the FLAG-epitope at the C-terminal and sequenced to confirm that they contain only the expected mutations as indicated in Table 1 . The production of lentiviral particles pseudotyped with H5 hemagglutinin was performed as previously described [18] . Briefly, HEK293T cells were co-transfected with a plasmid containing the coding sequence of the indicated H5-HA and a lentiviral backbone plasmid pNL-Luc E2 R2 which carries a modified proviral genome of HIV with env deleted and is engineered to express the firefly luciferase reporter. Alternatively, MLV-backbone plasmids (a kind gift from Dr. Michael Farzan, Harvard Medical School)described in [41] were used where indicated. To release particles into the culture medium, either soluble bacterial NA from Vibrio cholerae (NAvb; Roche, Mannheim, Germany) was added to the producer cells at a concentration of 6.25 mU/ml or co-transfection of N1 gene was used where indicated. Supernatants containing H5pp were harvested 48 hr post-transfection, filtered and used to transduce MDCK cells for luciferase reporter activity assay or concentrated by ultracentrifugation as indicated. Luciferase reporter activity assay MDCK cells (4000 cells/well) were seeded in 96-well white assay plates one day before H5pp transduction. Luciferase reporter activity assay was performed 72 hr post transduction using Bright-Glow Luciferase substrate (Promega, Mandison, WI, USA) according to the manufacturer's instructions. Samples were measured using a Microbeta Luminometer (PerkinElmer, Waltham, MA, USA) and data were expressed as Relative Luminescence Units. Equal amounts of protein from total cell lysates or equal volumes of H5pp concentrated by ultracentrifugation were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Proteins were then transferred onto Hybond-P polyvinylidene difluoride (PVDF) membranes (Invitrogen) that were blocked with 5% milk for 30 min at room temperature. H5-HA was detected by incubation for 2 hr at room temperature with a mouse monoclonal anti-FLAG M2 antibody (Sigma, St.Louis, MO, USA; 1:1000 dilution) conjugated with horse radish peroxidase (HRP) (Sigma). The core protein in the pseudotyped particles was detected using an anti-p24 antibody (Abcam, Cambridge, UK) for 1 hr at room temperature at a 1:1000 dilution, followed by an additional 1 hr incubation with a goat-anti-mouse secondary antibody conjugated with HRP (ZymedH, Invitrogen) at a 1:5000 dilution. The levels of cyclophilin B (detected with a rabbit anti-cyclophilin B antibody from Abcam, 1:5000 dilution) or GAPDH (detected with a mouse anti-GAPDH antibody from Abcam, 1:10000 dilution) were measured on the same blots to verify that equal amount of samples had been transferred. Proteins were visualized by chemiluminescence using ECL Western blot detection reagents (Invitrogen). The relative electrophoretic mobility was estimated using NovexH Sharp Pre-stained Protein Standards (Invitrogen). 293T cells transfected with H5 HA were detached with and resuspended in PBS, blocked in 10% horse serum and then labelled with a polyclonal rabbit anti-H5N1 antibody (Sino Biologicals Inc., Beijing, China) at a 1:400 dilution for 1 hr at 4uC. Unbound antibodies were removed by washing three times with cold PBS, followed by staining with a phycoerythrin (PE)-conjugated, donkey-anti-goat secondary antibody (Jackson Immunoresearch Laboratories, Suffolk, UK) for 30 min at 4uC. Data were collected from at least 5000 cells on an LSRII flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) and post-acquisition analyses of cell surface expression of H5-HA was performed using FlowJo software (TreeStar, Ashland, OR, USA). Expression and purification of soluble H5-HA proteins HEK293T cells stably expressing soluble HA (sHA) proteins were generated by selection of transfected cells in culture medium containing 300 ug/ml hygromycin for at least 1 month. To purify sHA proteins, cells were grown in DMEM with 5% FBS until 90% confluence. Culture supernatant containing secreted sHA proteins was cleared by centrifugation at 4000 rpm for 15 minutes at 4uC, concentrated by Amicon Ultra-15 Centrifugal Filter Units with Ultracel-100 membrane (Millipore, Billerica, MA, USA; 100 kDa cut-off) and stored at 280uC until use. Soluble HA proteins were affinity purified from concentrated supernatant using anti-FLAG M2 affinity gel (Sigma). Because it has been shown that HA-bound sialic acid could interfere with the accessibility of the receptorbinding site to cellular receptors [42] , anti-FLAG M2 resins bound with sHA proteins were washed twice in cold PBS and subjected to treatment with NAvb (Roche, 62.5 mU/ml) at 37uC for 45 minutes to remove terminal sialic acid residues, followed by three washes in cold PBS. Bound sHA proteins were eluted with 150 ml FLAG peptide (International Laboratory, USA, 0.4 mg/ml in PBS) for four times. To remove the FLAG peptides, all eluates were pooled and concentrated using Amicon Ultra-0.5 mL with Ultracel-100 membrane (Millipore, 100 kDa cut-off). To examine the oligomeric state of sHAs, proteins were resolved on a discontinuous native PAGE (6% of acrylamide) followed by western blot detection using a HRP-conjugated anti-FLAG M2 antibody (Sigma). Cell-based HA binding assay MDCK cells were grown in 96-well plate until complete confluence, then fixed with 4% paraformaldehyde (PFA, Sigma), washed three times in PBS and blocked for at least 2 hrs in 5% BSA. The indicated amount of purified soluble HA proteins, measured by the Bradford assay, were added to the wells in duplicates or triplicates and incubated overnight at 4uC. Cells were washed three times in PBS and then incubated with anti-FLAG antibody (Origene, Rockville, MD, USA; 1:1000 dilution for 2 hrs at room temperature) to detect sHA proteins bound to cell surface. After washing for three more times in PBS to remove unbound sHA proteins, cells were incubated with goat-anti-mouse secondary antibody conjugated with HRP (ZymedH, Invitrogen) at a 1:5000 dilution for 1 hr at room temperature. Unbound secondary antibody was removed by washing three times in PBS, and ABTS substrate (Invitrogen) was added to the plate according to the manufacturer's instructions. Forty minutes after the addition of substrate, absorbance at 415 nm (OD415) was measured using a Sunrise TM plate reader (Tecan, Mä nnedorf, Switzerland). Cambodian H5N1 virus strains were isolated from human clinical specimens by inoculation in Madin-Darby canine kidney (MDCK) cells in the biosafety level 3 laboratory of the Institut Pasteur in Cambodia, according to conventional protocols [43] . Hemagglutination titres were measured using 0.75% suspensions of human (type O), horse and guinea pig red blood cells, as previously described [30, 43] . Results are presented as mean values 6 SD of the indicate number of observations. Statistical difference between groups was determined by the unpaired Students's t-test with a 0.05 significance level.

Particulate matter air pollution disrupts endothelial cell barrier via calpain-mediated tight junction protein degradation

BACKGROUND: Exposure to particulate matter (PM) is a significant risk factor for increased cardiopulmonary morbidity and mortality. The mechanism of PM-mediated pathophysiology remains unknown. However, PM is proinflammatory to the endothelium and increases vascular permeability in vitro and in vivo via ROS generation. OBJECTIVES: We explored the role of tight junction proteins as targets for PM-induced loss of lung endothelial cell (EC) barrier integrity and enhanced cardiopulmonary dysfunction. METHODS: Changes in human lung EC monolayer permeability were assessed by Transendothelial Electrical Resistance (TER) in response to PM challenge (collected from Ft. McHenry Tunnel, Baltimore, MD, particle size >0.1 μm). Biochemical assessment of ROS generation and Ca(2+) mobilization were also measured. RESULTS: PM exposure induced tight junction protein Zona occludens-1 (ZO-1) relocation from the cell periphery, which was accompanied by significant reductions in ZO-1 protein levels but not in adherens junction proteins (VE-cadherin and β-catenin). N-acetyl-cysteine (NAC, 5 mM) reduced PM-induced ROS generation in ECs, which further prevented TER decreases and atteneuated ZO-1 degradation. PM also mediated intracellular calcium mobilization via the transient receptor potential cation channel M2 (TRPM2), in a ROS-dependent manner with subsequent activation of the Ca(2+)-dependent protease calpain. PM-activated calpain is responsible for ZO-1 degradation and EC barrier disruption. Overexpression of ZO-1 attenuated PM-induced endothelial barrier disruption and vascular hyperpermeability in vivo and in vitro. CONCLUSIONS: These results demonstrate that PM induces marked increases in vascular permeability via ROS-mediated calcium leakage via activated TRPM2, and via ZO-1 degradation by activated calpain. These findings support a novel mechanism for PM-induced lung damage and adverse cardiovascular outcomes.

Ambient particulate matter (PM) poses a threat to national public health in urban environments and other polluted areas throughout the US and around the world. Epidemiological studies have shown associations of exposure to low levels of urban particulate matter with increased cardiopulmonary morbidity and mortality [1, 2] . The assessment of PM-induced health effects is challenging. Various mechanisms have been proposed to explain the cardiopulmonary health effects of PM including increased pulmonary and systemic oxidative stress and inflammation, enhanced coagulation, and altered cardiac autonomic function [3, 4] . Airway epithelium represents a well-investigated target for environmental pollutants such as PM. Exposure of airway epithelium to airborne PM causes altered cytokine/chemokine gene expression and increased production of IL-1β, IL-6, IL-8 and TNF-α [5, 6] . Now, the lung endothelium is also gaining attention as a viable PM target tissue. Exposure of the endothelium to PM or its active components in the systemic circulation induces significant systemic endothelial inflammation and dysfunction, even at low levels of exposure [7, 8] . The water soluble fraction of PM (up to 35-50%) can easily diffuse through the epithelium/endothelial barrier to the systemic circulation. Bioavailable transition metals present in urban PM catalyze redox reactions in human lung endothelium, which cause oxidative stress, increase the production of inflammatory cytokines, and increase the activation of NF-κB signaling pathways, all of which trigger further endothelial damage [9] . Increased endothelial monolayer permeability is also observed in inflammatory pulmonary conditions such as acute lung injury (ALI), acute respiratory distress syndrome (ARDS), and sepsis; devastating lung disorders with mortality exceeding 30%, as well as in more subacute and chronic inflammatory disorders such as asthma [10, 11] . We recently described a murine asthma model with strong evidence for PM-mediated vascular barrier dysfunction with increased protein leakage into bronchoalveolar lavage (BAL), a marker of acute inflammatory lung damage [12] . By assessment of direct effects on endothelial barrier integrity in vitro, we further demonstrated that the vascular hyperpermeability mediated by intratracheal PM exposure is mainly dependent on acute endothelial barrier disruption by PM [13] . Exposure of human lung EC to PM resulted in significant ROS generation, which mediates p38beta MAPK activation, leading to the phosphorylation of HSP27 [13] . Phosphorylated HSP27 facilitates the synthesis of stress fibers and formation of paracellular gaps, which causes protein-containing fluid to leak from the microvessel lumen to the lung alveoli, leading to further pulmonary inflammation [13] . However, these previous findings did not explain the persistent character of PM-mediated endothelial barrier disruption. To fully examine PM-mediated vascular hyperpermeability, we explored effector targets such as tight junction and adherens junction proteins, which are known to be critical in EC barrier maintenance. We demonstrated that PM exposure induced the relocation of tight junction protein Zona occludens-1 (ZO-1) from the cell periphery, which was accompanied by a significant reduction in the level of ZO-1 protein but not in the levels of adherens junction proteins (VE-cadherin and β-catenin). PM also mediated intracellular calcium mobilization via the transient receptor potential cation channel M2 (TRPM2), in a ROS-dependent manner with subsequent activation of the Ca 2+ -dependent protease calpain. PM-activated calpain is responsible for ZO-1 degradation and EC barrier disruption. These observations not only provide new information as to how PM disrupts endothelial tight junctions, but also represent the first evidence establishing the critical role of calpain signaling in modulating endothelial cell barrier function under oxidative stress. These results increase our understanding of PM-induced adverse cardiopulmonary outcomes. Moreover, as the newly characterized signaling cascade of ROS/TRPM2/Calpain/ ZO-1 likely has fundamental roles in regulating the cytoskeleton under oxidative stress, these novel observations may have broad applicability to vascular pathophysiology in a variety of cell types. Molecular mass standards, polyacrylamide gels, and protein assay reagents were purchased from Bio-Rad (Hercules, CA Human lung microvascular ECs obtained from Lonza (Basel, Switzerland) were cultured as previously described [16] in EGMMV-2 complete medium (Lonza). Endothelial cells were grown to confluence in polycarbonate plates containing evaporated gold microelectrodes, and TER measurements were continuously obtained using an electrical cell-substrate impedance sensing system (ECIS) (Applied Biophysics, Troy, NY) as previously described in detail [17] . Human microvascular lung ECs were transfected with siRNA using siPORT Amine (Ambion, Austin, TX) according to the manufacturer's protocol as we described previously [13] . ECs plated on glass cover slips were loaded with 5 μM fura-2 AM (Invitrogen, Carlsbad, CA) in 1 ml of basal medium as previously reported [18] . The cover slips with ECs were inserted diagonally into 1 cm acrylic cuvettes filled with 3 ml of basal medium at 37°C. Fura-2 fluorescence was measured with an Aminco-Bowman Series 2 luminescence spectrometer (SLM/Aminco, Urbana, IL) at excitation wavelengths of 340 and 380 nm and emission wavelength of 510 nm. The cover slips with ECs were then moved to 35 mm dishes, treated with PM suspension, and incubated for 15 minutes at 37°C in 95% O 2 and 5% CO 2 . After every 15-minute incubation, the coverslips were withdrawn from the dish and inserted back into the acrylic cuvettes for Fura-2 fluorescence measurement, then returned to the dish with the resuspended PM preparation for another incubation period (15-60 min). Male A/J mice (10-12 weeks of age; Jackson Laboratories, Bar Harbor, ME) were housed in an environmentally controlled animal facility at the University of Illinois at Chicago (UIC) for the duration of the experiments. All animal procedures follow the guideline of the UIC Animal Care and Use Committee. PM (10 mg/kg, in 50 μl of saline) was delivered via intratracheal aspiration 1 hr after NAC or calpeptin treatment, as previously described [12, 15] . Animals were sacrificed 24 hr after PM treatment, and bronchoalveolar lavage (BAL) and lung tissue were collected [12, 19] . Total BAL cells were counted with a hemocytometer. The BAL fluid was used for protein and cytokine measurement (Bio-Rad, Hercules, CA) according to the user's manual. Data are presented as group means ± SEM. We performed statistical comparisons among treatment groups by randomized-design two-way analysis of variance followed by the Newman-Keuls post hoc test for more than two groups, or by an unpaired Student's t-test for two groups. In all cases, we defined statistical significance as p < 0.05. We assessed human lung microvascular EC barrier function as measured by transendothelial electrical resistance (TER), a highly sensitive measurement of permeability. PM challenge (10-100 μg/ml) induced doseand time-dependent reduction in TER (Additional file 1: Figure S1 ) or increases in FITC-dextran leakage through EC monolayer (Additional file 1: Figure S2 ) [20] , indicating a loss of EC barrier integrity. At the same time, PM induced a time-dependent (1-6 hr) reduction of the levels of tight junction proteins ZO-1 and ZO-2, but did not affect levels of adhesion junction proteins VE-cadherin or β-catenin ( Figure 1A ). No obvious cytotoxicity in the ECs was found after PM challenge (100 μg/ml, 0-16 hr) with MTT assay [13] or LDH release assay (Additional file 1: Figure S3 ). PM (100 μg/ml) induced substantial time-dependent ROS production in microvascular ECs as measured by DCFDA oxidation, which peaked around 30-60 min (Additional file 1: Figure S4 ). EC pretreatment with N-acetyl-cysteine (NAC, 5 mM, 1 hr), an ROS scavenger, or PEG-catalase (PEG-CAT, 250 U/ml, 1 hr), which degrades H 2 O 2 , prevented PM-induced DCFDA oxidation by ROS (Additional file 1: Figure S4 ). PM produced a sustained timedependent decrease in TER, with a maximal effect observed at 100 μg/ml (80% decrease in TER), which is similar to its effects on other endothelial cell types, as we have reported previously [13] . NAC pretreatment (5 mM, 1 hr pretreatment) prevented PM-induced ZO-1 degradation ( Figure 1B ), while NAC (5 mM, 1-24 hr) does not change ZO-1 protein levels by itself (Additional file 1: Figure S5 ). After ECs were treated with PM (100 μg/ml, 6 hr), ZO-1 was relocated from the cell periphery and degraded, which was followed by gap formation between ECs. VE-cadherin, on the other hand, underwent no relocation or degradation ( Figure 1C ). NAC pretreatment (5 mM, 1 hr pretreatment) prevented PM-induced ZO-1 relocation and gap formation ( Figure 1C ). These data strongly indicate that PM causes EC barrier disruption selectively via oxidative tight junction protein degradation. NAC (5 mM) or PEG-CAT (250 U/ml) pretreatment significantly inhibited PM-induced EC barrier disruption as measured by TER ( Figure 1D ). We also examined the effects of another ROS scavenger EUK-134 on PMchallenged ECs. As NAC, EUK (5 μM, 1 hr pretreatment) attenuated PM-induced ZO-1 degradation and TER reduction (Additional file 1: Figure S6 ). These results further confirmed that PM induces ROS-dependent EC barrier disruption. We previously demonstrated that high levels of ROS in endothelial cells activate calpain, a calcium-dependent protease [21] . We therefore investigated the role of calpain in PM-mediated ZO-1 degradation. PM (100 μM, 1 hr) induced a significant increase in calpain activity in ECs, which was inhibited by selective calpain inhibitors ALLN (30 μM, 1 hr pretreatment) and calpeptin (10 μM, 1 hr pretreatment) (Figure 2A ). PM-induced calpain activation was also inhibited by BAPTA-AM (50 μM, 1 hr pretreatment), a calcium chelator, and by NAC (5 mM, 1 hr pretreatment) (Figure 2A ). We next determined the role of calpain in PM-induced EC barrier disruption. PM-induced reduction in TER was partially inhibited by calpeptin or ALLN ( Figure 2B ). In parallel, ALLN and calpeptin also significantly prevented PM-induced ZO-1 degradation, as did chelation of intracellular calcium via BAPTA-AM ( Figure 2C ). Addition of PM (100 μg/ml) to human lung microvascular ECs produced slow time-dependent Ca 2+ influx ( Figure 3A ; Ca 2+ influx is represented by an increase in the 340/380 ratio of Fura-2 AM.). Notably, this finding of increased intracellular Ca 2+ is in accordance with our previous finding of increased activated calpain. Under oxidative stress, a key member of the transient receptor potential (TRP) cation channel, member M2 (TRPM2), is activated and causes slow calcium influx [22] . We therefore examined the role of activated TRPM2 in PM-stimulated calcium influx. Anti-TRPM2 blocking antibody (5 μg/ml, 4 hr pretreatment) significantly prevented the PM-induced Ca 2+ transients compared to ECs treated with control IgG ( Figure 3B ). Reductions in TRPM2 protein expression (by siRNA) also inhibited the PM-induced Ca 2+ influx compared to ECs treated with control siRNA ( Figure 3C ). We next investigated whether TRPM2 activation was ROS-dependent. Depletion of PM-induced ROS by NAC (5 mM, 1 hr pretreatment) significantly prevented PMmediated calcium influx ( Figure 3D ). Under oxidative stress, Poly ADP ribose polymerase (PARP) generates ADPribose [23, 24] , which activates TRPM2 by binding to its carboxyl terminus. The PARP inhibitor 3-aminobenzamide (3-AB, 1 mmol/L, 1 hr pretreatment)significantly reduced PM-induced Ca 2+ influx ( Figure 3E ), which further confirmed that TRPM2 is activated by PM via ROS and PARP. We next investigated the role of TRPM2 activation in EC barrier function and ZO-1 degradation. TRPM2 neutralizing antibody (5 μg/ml, 4 hr pretreatment) significantly prevented ZO-1 degradation induced by PM ( Figure 4A ). TRPM2 siRNA (100 ng/ml), which downregulated TRPM2 protein level ( Figure 4B ), also inhibited PM-induced ZO-1 degradation ( Figure 4C ). In parallel, antagonizing TRPM2 by either TRPM2 antibody (5 μg/ml, 4 hr pretreatment) or TRPM2 siRNA (100 ng/ml) significantly prevented PM-induced EC barrier disruption ( Figure 4D-E) , as indicated by TER measurements. A PM-mediated murine model of pulmonary inflammation has been well established [12] . We investigated the role of ROS and calpain in PM-induced pulmonary inflammation by examining protein leakage, white blood cell infiltration (inflammatory leukocytes), and the release of proinflammatory cytokines into BAL fluids ( Figure 5 ). Pre-administration of NAC (150 mg/kg) or calpeptin (1 mg/kg) 1 hr before PM challenge (10 mg/kg, 24 hr) significantly attenuated PM-induced protein leakage into BAL fluids (~50% reduction, Figure 5A ). PM challenge (10 mg/kg, 24 hr) resulted in an increase in inflammatory leukocytes, e.g. neutrophils and eosinophils [12, 15] , in BAL fluids ( Figure 5B ). Pre-administration of NAC (150 mg/kg) or calpeptin (1 mg/kg) attenuated PMinduced inflammatory leukocyte infiltration. Furthermore, NAC (150 mg/kg) or calpeptin (1 mg/kg) attenuated the release of PM-induced proinflammatory cytokines IL-6 and THF-α into BAL (~50% inhibition, Figure 5C-D) . These results suggest that ROS scavenging by NAC or Figure 2 Activated calpain is required for PM-mediated ZO-1 degradation and EC barrier disruption. (A) Human lung microvascular ECs grown in 60-mm dishes to approximately 95% confluence were treated with ALLN (30 μM), calpeptin (CALP, 10 μM), NAC (5 mM), or BAPTA-AM (50 μM) for 1 hr, and then challenged with PM (100 μg/ml) for 1 hr. Cell lysates were subjected to calpain activity assay by using the calpain activity kit (Calbiochem). Calpain activity was normalized to protein concentration and expressed as fold changed compared to control. *p < 0.05 compared to control. **p < 0.05 compared to PM only group. (B) ECs grown on ECIS gold electrodes were treated with ALLN (30 μM), calpeptin (CALP, 10 μM) for 1 hr, and then challenged with PM (100 μg/ml). Changes in TER were measured with ECIS. *p < 0.05 compared to PM only group. (C) ECs grown on 6-well plates were treated with ALLN (30 μM), calpeptin (CALP, 10 μM), or BAPTA-AM (50 μM) for 1 hr, and challenged with PM (100 μg/ml) for 1-6 hr. Cell lysates were analyzed by Western blotting with antibody to ZO-1. Changes in levels of ZO-1 are expressed as fold changes and normalized to β-actin. Shown are representative blots from three independent experiments. *p < 0.05 compared to PM-1 hr group. **p < 0.05 compared to PM-6 hr group. calpain activity inhibition by calpeptin leads to multiple protective effects including enhancement of lung endothelial barrier function, reduction of inflammatory cell infiltration, and prevention of proinflammatory cytokine release in the lungs of PM-challenged mice. We next examined tight junction ZO-1 levels in the PM-exposed lung. PM challenge (10 mg/kg, 24 hr) induced a significant reduction of ZO-1 protein levels in the murine lung, while pre-administration of NAC (150 mg/kg) or calpeptin (1 mg/kg) 1 hr before PM challenge attenuated PMinduced ZO-1 loss from lung tissues. Taken together, these data suggest a crucial role for the ROS-calpain-ZO-1 signaling pathway in the regulation of EC barrier disregulation in PM-mediated pulmonary inflammation. To further confirm the critical role of ZO-1 degradation in EC barrier disruption in vitro and pulmonary inflammation in vivo, we next examined the beneficial effects of over-expressing ZO-1 protein in endothelial cells in vitro and in vivo. ZO-1 over-expression ( Figure 6A ) significantly (but not completely) attenuated PM-induced EC barrier disruption ( Figure 6B ). These facts demonstrate that endothelial ZO-1 loss contributes to PM-mediated EC barrier disruption. We next over-expressed ZO-1 in vivo by a liposome delivery system labeled with ACE antibody, which successfully over-expressed ZO-1 in Human lung microvascular ECs were plated on glass cover slips and loaded with 5 μM Fura-2 AM in 1 ml of basic medium. ECs were rinsed twice and treated with PM (100 μg/ml). Fura-2 fluorescence was measured at excitation wavelengths of 340 and 380 nm and an emission wavelength of 510 nm (0-60 min). Fura-2 has excitation wavelengths of 380 nm in its free form and 340 nm when it is bound to calcium; therefore, relative calcium concentrations are represented by the ratio of 340/ 380. (B) ECs plated on glass cover slips were treated with TRPM2 antibody or control IgG (5 μg/ml, 4 hr pretreatment). ECs were then rinsed twice and subjected to the same calcium measurement procedures with PM treatment (100 μg/ml, 0-60 min). *p < 0.05 compared to control IgG group at the same time point. (C) ECs were transfected with TRPM2 siRNA or control siRNA for 48 hrs. ECs were then rinsed twice and subjected to the same calcium measurement procedures with PM treatment (100 μg/ml, 0-60 min). EC cell lysates were analyzed by Western blot with antibodies to TRPM2 and β-actin to confirm silencing. *p < 0.05 compared to control siRNA group at the same time point. (D-E) ECs plated on glass cover slips (95% confluent) were treated with NAC (5 mM, 1 hr pretreatment) or 3-AB (1 mM, 1 hr pretreatment). ECs were then rinsed twice and subjected to the same calcium measurement procedures with PM treatment (100 μg/ml, 0-60 min). *p < 0.05 compared to PM-only group at the same time point. murine lung tissues ( Figure 6C ). ZO-1 over-expression significantly attenuated BAL protein leakage ( Figure 6D) , BAL white blood cell infiltration ( Figure 6E) , and the release of proinflammatory cytokine IL-6 into BAL ( Figure 6F ), indicating the crucial role of ZO-1 loss in mediating PM-induced pulmonary inflammation and lung vascular hyperpermeability. The most significant finding of the present study is the novel characterization of a ROS-dependent pathway that causes calpain-dependent endothelial ZO-1 degradation in response to PM. These data represent the first evidence that calpain signaling, via calcium leakage from activated TRPM2 by ROS, plays a critical role in modulating endothelial cell barrier function, resulting in tight junction protein ZO-1 degradation (Additional file 1 Figure S7 ). The consequence of ZO-1 degradation is sustained endothelial hyperpermeability and persistent lung inflammation, both of which contribute to variety of acute or chronic cardiovascular disorders [25, 26] . These effects were also observed with other types of PM samples (1648a from Human lung microvascular ECs grown in 6-well dishes to approximately 95% confluence were treated with TRPM2 antibody or control IgG (5 μg/ml) for 4 hr, and then challenged with PM (100 μg/ml) for 6 hr. Cell lysates were analyzed by Western blotting with ZO-1 antibody. Changes in levels of ZO-1 are expressed as fold changes and normalized to β-actin. Shown are representative blots from three independent experiments. *p < 0.05 compared to control. **p < 0.05 compared to PM challenge. (B) ECs grown in 6-well dishes to approximately 80% confluence were treated with TRPM2 siRNA or control siRNA (100 ng/ml) for 48 hr, and cell lysates were analyzed by Western blotting with ZO-1 antibody. (C) ECs grown in 6-well dishes to approximately 95% confluence were treated with TRPM2 siRNA or control siRNA (100 ng/ml) for 48 hr, and then challenged with PM (100 μg/ml) for 6 hr. Cell lysates were analyzed by Western blotting with ZO-1 antibody. Changes in levels of ZO-1 are expressed as fold changes and normalized to β-actin. Shown are representative blots from three independent experiments. *p < 0.05 compared to control. **p < 0.05 compared to PM challenge. (D) ECs grown on ECIS gold electrodes were treated with TRPM2 antibody or control IgG (5 μg/ml) for 4 hr, and then challenged with PM (100 μg/ml). Changes in TER were measured with ECIS. *p < 0.05 compared to PM-challenged group. (E) ECs grown on 100 mm dishes were treated with TRPM2 siRNA or control siRNA (100 ng/ml) for 48 hr, and then plated onto gold electrodes for ECIS measurement. 24 hours after replating, the ECs were challenged with PM (100 μg/ml) and changes in TER were measured with ECIS. *p < 0.05 compared to PM-challenged group. National Institute of Standards and Technology, fine PM collected from New York city or Baltimore, data not shown), indicating a selective pathogenesis pathway by PM pollution. Previous studies report that PM triggers the generation of reactive oxygen species or ROS mainly from dysfunctional mitochondria [27, 28] , and we also noticed the massive generation of ROS by this PM sample is also mainly from mitochondria (unpublished observation). The high iron level of this particular PM might also contribute to the ROS generated via Fenton reactions. ROS released endogenously, have been implicated in the pathophysiology of several lung diseases, including asthma and COPD, as the biochemical mechanisms underlying the urban PM-induced airway inflammation and toxicity [29] . ROS are highly reactive and cause deleterious gene, protein, and tissue effects. ROS are increased in BAL or exhaled breath condensate from patients with inflammatory lung injuries and from people with cardiopulmonary disease who have been exposed to PM [30, 31] . This response may reflect the high oxidative potential of fine and ultrafine particulates. Residual oil fly ash (ROFA) and PM1.7-3.5 cause pulmonary inflammation mediated by oxidative stress [32, 33] . In vivo, exposing rats to PM leads to the formation of free radicals in the lung [34] . Since cardiovascular disease is considered a risk factor of PMrelated mortality and morbidity, it is interesting to note that spontaneously hypertensive rats, when exposed to PM, were more susceptible to pulmonary (inflammatory injury) and cardiovascular complications (acute depression of ECG activity) in an oxidant-dependent manner [35] . Besides ROS, PM might trigger adverse outcomes via other potential mechanisms including nonselective phosphatase inhibition (by vanadium) or competitive ion channel inhibition (by nickel) due to the complex and variable chemical components. In this study, we first define a novel pathway that mediates ROS-dependent tight junction disruption upon particulate matter challenge. Tight junctions, or zonula occludens, are the most apical component of the intercellular junctional complex, which also includes adherens junctions, desmosomes, and gap junctions [36] . ZO-1 was the first tight junction protein to be identified, and ZO-2 and ZO-3 were later isolated as proteins that coimmunoprecipitated with ZO-1 [37, 38] . ZO-1 is a peripheral membrane-associated component of the cytoplasmic plaque of tight junctions and is found ubiquitously within tight junctions of epithelial and endothelial cells [39] . ZO-1 interacts with many cellular proteins via its multiple protein-binding domains. ZO-1 has been reported to interact with other ZO family members or claudins via the PDZ domains [40, 41] . ZO-1 interacts with the C-terminus of occludin with its GuK domain and the acidic domain [42] . The proline-rich C-terminus of ZO-1 mediates its binding to F-actin in vitro, and thus links it to the cytoskeleton [43] . Clearly, ZO-1 interacts with a wide variety of cell skeleton components and plays a central role in orchestrating tight junction complexes. Any dysregulation of ZO-1 in endothelial cells by extracellular stimuli, such as virus shell proteins or alcohol, leads to persistent tight junction disruption and vascular hyperpermeability. Calpain is a regulator of endothelial integrity which helps control fundamental cellular processes including cytoskeletal remodeling, membrane fusion, cell proliferation and differentiation, and activation of proteolytical cascades leading to apoptosis [44, 45] . Under oxidative stress, activated calpain cleaves eNOS and cytoskeletal proteins and induces apoptosis [21, [46] [47] [48] . Particulate matter induces endothelial cell intracellular oxidative stress, which leads to the activation of calpain, one of the major cytoskeletal regulators. Here we describe the cleavage of tight junction protein ZO-1 by activated calpain both in vitro and in vivo, indicating that calpain plays a central role in PM-induced endothelial barrier disruption and vascular hyperpermeability. In addition, as activated calpain cleaves other critical cytoskeletal proteins including ezrin and MARCKS protein, the contribution of the other cytoskeletal proteins to the EC hyperpermeability induced by PM needs to be further investigated. Oxidative calcium influx is mediated by plasma membrane cation-permeable ion channels. The transient receptor potential protein (TRP) and its homologs are cation channels with a tetramer secondary structure which senses diverse stimuli from the extracellular and intracellular Figure 6 Over-expression of endothelial ZO-1 attenuates PM-induced EC barrier disruption in vitro and pulmonary inflammation in vivo. (A) Human lung microvascular EC were grown to 60% confluence and treated with ZO-1 expression plasmid with X-fect reagent for 48 hr, and over-expression of ZO-1 protein was confirmed by Western blot. (B) The ECs were then challenged with PM (100 μg/ml), and changes in TER after 6 hr were measured with ECIS. *p < 0.05 compared to PM-challenged group. AJ mice were treated with ZO-1 expression plasmid with an ACE antibody-conjugated liposome delivery system (5 mg/kg) for 3 days, then challenged with PM (10 mg/kg). After 24 hr of PM exposure, (C) lung ZO-1 levels were analyzed with Western blot. Shown is one of the three repeated blots. BAL was collected and (D) protein content, (E) total white blood cell, and (F) IL-6 levels were measured. N = 4. *p < 0.05 compared to PM-challenged group. **p < 0.05 compared to control. environments [49] . Mammalian TRPs comprise six major subfamilies. TRPM2, a member of the TRP channel M2 subtype, is a calcium-permeable channel activated by intracellular messengers such as ADP-ribose [50] . Massive ROS burden induced by PM contributes to DNA oxidation and damage, which activates poly-ADP ribose polymerase (PARP) to initiate DNA repair mechanisms. PARP binds to single-stranded and double-stranded DNA breaks and catalyses the breakdown of NAD into nicotinamide and ADPribose, the intracellular agonist of TRPM2 [22, 51, 52] . Oxidative stress-mediated activation of the PARP pathway serves as the major source of free ADP-ribose production in endothelial cells [53] . Intracellular ADP-ribose activates TRPM2, allowing calcium ions to enter the cell, which in turn trigger numerous physiological and pathological processes. An important limitation of our study is the high dose of PM that we employed. With 10-30 μg/m 3 ambient PM level in the US or Europe, it is hardly to achieve a high level of acute PM exposure. While 100 μg/ml (in vitro) or 10 mg/kg (in vivo) are typical doses used in particulate matter toxicology studies [12, 13, 27, [54] [55] [56] . With an assumed ambient PM level of 20 μg/m 3 , one man with 70 kg body weight and 8 m 3 /minute respiration rate would receive a dose of 10 mg/kg corresponding to about 16 years of exposure with 50% deposition rate. As noted, a lot of cities in the developing countries still have high levels of ambient PM. A report by world bank [57] Extensive epidemiologic and experimental evidence has demonstrated that particulate air pollution directly causes cardiopulmonary damage. Our observations demonstrate a novel mechanism of PM-mediated disruption of endothelial barrier function which is attributable to ZO-1 degradation by calpain, which is activated by extracellular calcium leakage through oxidant-sensitive TRPM2 channels. Therefore, inhibition of ROS/TRPM2/calpain/ZO-1 degradation may provide useful therapeutic strategies for the treatment of endothelial barrier dysfunction and lung inflammation. Additional file 1: Figure S1 . (A-B) PM induces dose-dependent reduction in transendothelial resistance (TER). (C) PM induces dosedependent (6 hr) reduction of ZO-1 protein levels. Figure S2 . PM induced FITC-dextran leakage across EC monolayer. Figure S3 . PM (100 μg/ml, 1-16 hr) does not induce LDH release from human ECs. Figure S4 . NAC or PEG-CAT attenuates PM-induced ROS in ECs. Figure S5 . NAC (5 mM, 1-24 hr) does not change ZO-1 protein levels in human ECs. Figure S6 . EUK-134 (5 μM, 1 hr pre-treatment) attenuates PM (100 μg/ml, 6 hr)-induced ZO-1 degradation and TER reduction. Figure S7 . We hypothesize that PM induces EC barrier disruption in delayed phase (via ZO-1 degradation) and acute phase (via stress fiber formation).

The expression of nicotinic receptor alpha7 during cochlear development

Nicotinic acetylcholine receptor alpha7 expression was examined in the developing and adult auditory system using mice that were modified through homologous recombination to coexpress either GFP (alpha7GFP) or Cre (alpha7Cre), respectively. The expression of alpha7GFP is first detected at embryonic (E) day E13.5 in cells of the spiral prominence. By E14.5, sensory regions including the putative outer hair cells and Deiters' cells express alpha7GFP as do solitary efferent fibers. This pattern diminishes after E16.5 in a basal to apex progression, as Hensen's cells and cells of the spiral ligament acquire alpha7GFP expression. At birth and thereafter alpha7GFP also identifies a subset of spiral ganglion cells whose processes terminate on inner hair cells. Efferent fibers identified by peripherin or calcitonin gene-related protein do not coexpress alpha7GFP. In addition to cochlear structures, there is strong expression of alpha7GFP by cells of the central auditory pathways including the ventral posterior cochlear nucleus, lateral lemniscus, central inferior colliculus, and the medial geniculate nucleus. Our findings suggest that alpha7 expression by both neuronal and non-neuronal cells has the potential to impact multiple auditory functions through mechanisms that are not traditionally attributed to this receptor.

Numerous neurotransmitter systems contribute to the normal development and function of the auditory sensory (cochlear) apparatus and the circuitry of the central nervous system. This includes members of the excitatory ligand-activated nicotinic acetylcholine receptor family (nAChR; Albuquerque et al. 2009 ). The nAChR subunit family consists of 16 distinct subunits that in various pentameric combinations form ligand-activated ion channels that each exhibit uniquely specialized pharmacological and functional properties (Albuquerque et al. 2009 ). One of these is the homomeric alpha7 nAChR (a7) whose functional uniqueness is in part due to its expression by both neuronal and non-neuronal cells in many tissues throughout the body and because it is responsive to multiple agonists (including acetylcholine and choline as well as nicotine). This results in its ability to modulate a diverse range of cellular functions including cell growth, cell survival, neurotransmission, and inflammation (Gahring and Rogers 2005; Levin et al. 2006; Albuquerque et al. 2009 ). Members of the nAChR family contribute to essentially all aspects of the auditory sensory system function and development (Morley and Happe 2000; Morley 2005) . This includes widespread changes in expression during embryogenesis that optimizes their contribution to signal transduction, fine-tuning of sensory hair cells, and modulating central auditory circuit neurotransmission (Elgoyhen et al. 1994 (Elgoyhen et al. , 2001a Happe and Morley 1998; Vetter et al. 1999 Vetter et al. , 2007 Morley and Happe 2000; Katz et al. 2004; Morley 2005) . This functional diversity is in part accomplished through strict spatiotemporal control of different nAChR subunit expression, as has been extensively described for the nAChRs composed of either homomeric (a9) or heteromeric (a9 + a10) subunits (Elgoyhen et al. 1994; Vetter et al. 1999 Vetter et al. , 2007 Elgoyhen et al., 2001b; Murthy et al. 2009) . Less is known about the role of other nAChRs including a7, although this receptor is implicated in modifying longer lived stimulation by high-frequency sound and supporting survival of spiral ganglion cells during development (Morley and Happe 2000; Morley 2005) . Because the measurement of a7 expression and function can be compromised by low receptor expression levels or the absence of conditions that best reveal its modulatory role (Gahring and Rogers 2005; Albuquerque et al. 2009 ), the participation by this receptor as an important contributor to the development and normal auditory sensory function remains to be fully explored. In this study, we examine a7 expression during development of the auditory sensory system. This was done using mice that were modified though methods of homologous recombination to introduce, at the a7 gene 3′ end, a hemagglutinin (HA) protein tag to the a7 receptor protein and a bicistronic IRES-driven tau + enhanced-GFP fusion protein reporter (a7 GFP ). An advantage of the tauGFP reporter construct is that the tau component directs GFP into the axon of cells expressing a7 GFP . Also generated was a mouse in which Cre-recombinase replaces the tauGFP. The expression of a7 GFP in these mice reveals extensive spatial and temporal remodeling of receptor expression during embryonic and postnatal development of the cochlear sensory structures. Furthermore, a7 GFP expression continues in both neuronal and non-neuronal cells of the adult cochlear structure and the central ascending auditory pathway. This suggests that a7 has the potential to impact functionally on auditory processes through multiple pathways and mechanisms that could impact upon the adult function in ways not traditionally attributed to this receptor. All animals were used and housed in accordance with protocols approved in advance by the Institutional Animal Care and Use Committee at the University of Utah . This includes adherence to the Guide for the Care and use of Laboratory Animals of the National Institutes of Health. Generation of alpha7-HA-IRES-tauGFP and alpha7-HA-IRES-Cre mice The construction of the a7 protein and gene (Chrna7) reporter mouse lines; Chrna7-HA-IRES-tauGFP (a7 GFP ) and Chrna7-HA-IRES-Cre (a7 Cre ) have been described in detail . Briefly, as diagramed in Fig. 1A , the methods of homologous recombination were used to introduce an epitope hemagglutinin (HA) and stop codon extension to the a7 C-terminus and a bicistronic IRES-tauGFP reporter cassette ). This generated the a7 GFP mouse (Fig. 1A) , which expresses the tauGFP protein as a marker of Chrna7 transcription. The Speed Congenic Program of the Jackson Laboratory was used to achieve 98% C57BL/6 background congenicity . For conditional cell ablation of the cells expressing Cre as in the a7 Cre mouse, we crossed this mouse with the LoxP conditional diphtheria toxin (DTA) mouse lines as described previously . Embryo (E) timing was based upon identification of coital plugs (equal to E0.5). Immunohistochemical methods were as described . Embryos were fixed in PBS/2% paraformaldehyde/5% sucrose, cryoprotected with sucrose in PBS to a final of 30%, embedded and sectioned using a Microm EM550 microtome. The 12-lm sections were mounted on glass slides, blocked, and permeabilized with 1% deoxycholate and 0.2% Triton X-100 in PBS, and then incubated overnight at 4°C with the appropriate primary antibodies. After washing, sections were incubated with secondary antibodies conjugated to fluorescent markers (Jackson Immuno-Research, West Grove, Pennsylvania) for 1 h at room temperature. The sections were again washed, and mounted in prolog gold antifade reagent (Invitrogen, Grand Island, New York; P36930) and cover-slipped before being photographed using fluorescence microscopy . Images were collected using a Microfire 24-bit CCD camera (Optronics, Goleta, California) and imported into Photoshop C2 for preparation of figures. The antibodies used were commercially obtained. These were anti-calcitonin gene-related protein (CGRP; rabbit; 1:30; Chemicon/Millipore, Temecula, Californa AB5920), anti-GFP (chicken; 1:800, Aves Labs, Tigard, Oregon GFP-1020), anti-HA (rabbit; 1:200; HA.11 Covance, Princeton, New Jersey PRB-101P), anti-peripherin (rabbit; 1:100; Abcam, Cambridge, Massachusetts #1530), anti-S100beta (rabbit; 1:100; Abcam ab868), rabbit anti-beta-III tubulin (TUJ1; 1:3000; Covance MMS-435P). Detection of GFP offers superior sensitivity that is well over background fluorescence ( Fig. 1B and C) . For this study, some inconsistent signal detection or autoflourescence was occasionally observed and these sites identified in the individual figures. We find the expression of GFP and HA are similar, although anti-HA expression is detected predominantly on the surface of cells identified by anti-GFP expression (Fig. 1D ). The expression of a7 exhibits distinct spatiotemporal patterning in developing cochlear structures. Previously, we demonstrated the earliest expression of a7 in the developing embryo to be in rhombomeres 3 and 5 of the E9.0 embryo ). Thus, we initiated studies of a7 GFP staining at this time. From E9.5 through approximately E12.5, the otic and cochlear structures did not express detectable a7 GFP ( Fig. 2A and not shown, see . The earliest detected expression of a7 GFP in the cochlear structures was at E13.5 in cells of the spiral prominence (SP; Fig. 2B ). The SP retains a7 GFP expression throughout embryonic and post-natal development (see below). By E14.5 ( Fig. 2C and D) , a7 GFP expression extends to cells in the sensory domain of the lesser epithelial ridge near the site of the presumptive outer hair cells (OHC) and Deiters' cells (Morsli et al. 1998; Lanford et al. 1999; Kiernan et al. 2005a,b) . Light staining of the greater epithelium ridge was also present from E14.5 and thereafter, although this staining is inconsistently observed (Fig. 1B and C and not shown). Coincident with this expression was strong staining of pioneering efferents that become separated into individually distinguished processes as they progress through the spiral ganglion (SG) to reach the external face of this sensory domain ( Fig. 2C ; see below). The staining of the epithelial cells of the lesser epithelial ridge intensifies thereafter (e.g., E15.5 in Fig. 2E ). At this stage, expression of a7 GFP by cells of the SG was in general only weakly observed in scattered cells (Fig. 2E ). By E16.5, a7 GFP expression continues to increase in cells of the lesser epithelial ridge of the prosensory domain where OHC and Deiters' cells can now be distinguished (Fig 2F and G and insert) . Cells throughout the SG were also revealed by expression of a7 GFP by this developmental stage. Pillar cells do not express a7 GFP and there were no identifiable efferent processes labeled by the expression of this receptor at this stage or thereafter (see the following sections). the a7 gene (Chrna7) was modified using homologous recombination to add a C-terminus epitope tag (hemagglutinin [HA] ) and inserted into the 3′ terminus of Chrna7 a reporter bicistronic internal ribosome entry sequence (IRES)-tau fusion to enhanced green fluorescent protein (eGFP) fusion protein cassette (a7 GFP ; see Methods and ). This construct was subsequently altered by replacing the tau:GFP cassette with the Cre-recombinase gene (a7 Cre ). (B, C) The visualization of the Chrna7 transcription using immunological detection of GFP compared with background. Shown are sagittal sections of the cochlear sensory structures of an E16.5 a7 GFP embryo in (B) and at greater magnification in (C). The panels on the left are stained for GFP expression (see Methods), whereas the image on the right shows an adjacent serial section that received the same staining treatment, only primary antibody was omitted. Photographs were collected at the same gain and exposure. The asterisk identifies cochlear ducts and the arrow points to the spiral prominence and the arrow head points to cell giving rise to the outer hair cells and Deiters' cells. Abbreviations are SG, spiral ganglion; and tg, trigeminal ganglion. In (B), the bar = 100 lm and in (C), the bar = 400 lm. (D) Examples of colabeling for a7 GFP (green) and anti-HA (HA) in cells associated with the spiral ganglion at E16.5. Examples of double-labeled cells are identified by with arrows. Some processes are also colabeled (arrow head). Bar = 50 lm. The pattern of a7 GFP expression in the E18.5 cochlear structure undergoes significant remodeling as both sensory hair cells and the associated supporting cells complete their differentiation ( Fig. 2H and I). This includes a decrease of a7 GFP expression by OHCs and underlying Deiters' cells that progresses away from the inner hair cells and proceeds in a basal-to-apical direction (next section). This is coincident with the appearance of signal in Hensen's cells that are most proximal to the outer line of OHCs (returned to below). Ganglionic afferent fibers ending at the base of the inner hair cells are also detected (see subsequent sections). In the postnatal mouse, as shown in the P6 cochlear sensory structure ( Fig. 2J and K), the expression of a7 GFP becomes limited to Hensen's cells immediately adjacent to the most distal OHC. Cells of the spiral ligament also acquire a7 GFP expression, while the spiral prominence remains unchanged. In the SG, the expression of a7 GFP is well established and the projections from these labeled cells can be followed to the vicinity of the inner hair cells (IHC) where their terminals appear to surround the base of the inner hair cell (IHC; Fig. 2J and K). A summary diagram illustrating the expression of the a7 GFP during these major developmental stages is shown in Fig. 2L . Remodeling of a7 GFP in the cochlear structure after E16.5 is in a basal-to-apical direction The remodeling of the sensory cell region of the cochlear structure between E16.5 and E18.5 as suggested by the progression in changing a7 GFP expression was examined further. Through E16.5, all otic structures exhibit a similar a7 GFP expression pattern (Fig. 3A) . This was not the case in the E18.5 cochlear structure where the loss of a7 GFP expression by OHC and Deiters' cells and acquisition of staining by Hensen's cells was first observed in the most basal structures and it then appears in the more apical structures successive developmental stages ( Fig. 3B and C and not shown). This generates a striking contrast in a7 GFP expression between cochlear structures at the apex relative to the base with intermediary turns, exhibiting the progressive stages of this change in a7 GFP expression (Fig. 3B ). By P4, this gradient was not evident (not shown) and the mature a7 GFP expression pattern first observed in the E18.5 basal cochlear structures was present across the entire structure. In Fig. 3D , a diagram depicts the remodeling of a7 GFP expression seen in the E18.5 developing cochlear structure. Nonsensory cells of the cochlear structure express a7 GFP As suggested by the preceding discussion, there was expression of a7 GFP by both neuronal and non-neuronal cells (Fig. 4) . This is particularly clear in the postnatal mouse (e.g., P6-P12), where the predominant expression of a7 GFP in neuronal cells was by cells of the SG (Fig. 4A) . The strongest labeling of cochlear structures was restricted to Hensen's cells and the spiral prominence ( Fig. 4A-E) . Evident at the P6 stage was a7 GFP signal in individual cells of the spiral ligament ( Fig. 4C and D) . Also evident were the extended branching that is characteristic of the morphology of type II fibrocytes located in this region (Fig. 4D ; Spicer and Schulte 1991; Sun et al. 2012 ). In the P12 cochlear structure, the branches were more abundant and form a 'feathered' structure that emanates from cell bodies defined by a7 GFP expression (Fig. 4E) . Cells of the stria vascularis or other members of the cell family composing the structures of the lateral wall and surrounding cochlear duct were not observed to express a7 GFP in these later stages of development (Fig. 4) . The expression of a7 GFP during innervation of the developing cochlear structure Innervation of cochlear sensory cells follows a series of distinct steps that were in part revealed by a7 GFP visualization (Fig. 5 ). As noted, the first detection of a7 expression was in the prominently labeled efferent processes that appear to form bundles upon entering the SG and then disperse into small solitary fibers (E14.5; Figs. 5A and 2C,D). These solitary processes exhibit a beaded structure as they proceed to the base of the developing sensory cells (Fig. 5B) . The origin of these efferent fibers was examined in serial sections of the E14.5 hind brain. These fibers appear to originate from a cell grouping in the basal brain stem caudal to trigeminal nucleus V that could be distinguished by their transient a7 GFP expression (Fig. 5C ). These cells occur in clusters (Fig. 5C insert) and their prominently labeled processes can be followed using serial section sets to the cochlear structure where they give rise to the fiber bundles and the point of dissemination associated with the SG (Fig. 5C and insert) . The anatomical location of these cells suggest that these cells are within the forming olive complex, which is consistent with the reports of pioneering fibers that originate from the developing olive complex and extend to the developing cochlea (Zuo et al. 1999 ). These fibers were not detected after E15.5. During the E15.5-16.5 period, there was essentially no labeling of neuronal processes by a7 GFP (Fig. 5D-F) . However, ongoing innervation of cochlear sensory cells was identified using peripherin labeling ( Fig. 5E ; see Simmons et al. 1996; Hafidi 1998; Huang et al. 2007) or for olivocochlear efferents that were identified by labeling for calcitonin gene-related protein (CGRP; Fig. 5F , Fritzsch 2003). By E18.5, the SG a7 GFP signal was present in afferent processes that extend to the base or near vicinity of the IHCs (Fig. 5G) . At birth and thereafter (P0-P12 analyzed), the expression of a7 GFP was strongly detected in SG afferent fibers where they terminate near or at the base of IHC sensory cells ( Fig. 5H and I) . This basic pattern of a7 GFP expression was reinforced during the remaining postnatal period as fibers continue to form a dense plexus that appears to surround the base of the IHCs. The other efferent fibers not detected by a7 GFP continue to be trimmed and also associate with their final targets (Merchan-Perez and Liberman 1996; Simmons et al. 1996; Hafidi 1998; Huang et al. 2007 ). The outcome of this remodeling was evident by P12 when the SG1 afferent terminals surrounding the IHC were distinguished by strong a7 GFP staining of the terminal clusters ( Fig. 5I and inset) . This was approximately the same time hearing onset occurs in mice (~P10; Kros et al. 1998 ). Processes originating from SG cells identified by peripherin expression that were not colabeled with a7 GFP form distinct efferent terminals on or very near OHCs cells and on the terminals that end on the IHC afferent terminals identified by a7 GFP labeling ( Fig. 5I ; Huang et al. 2007 ). While not entirely evident from the images shown, not all SG cells at P12 expressed a7 GFP , suggesting this could identify a functionally distinct subpopulation ( Fig. 5I ; Happe and Morley 1998) . Again, no a7 GFP labeling of olivocochlear efferents was detected. A diagram summarizing these findings is shown in Fig. 5J . Ablation of the a7 Cre -expressing cell lineage confirms a7 GFP expression during cochlear development Although a7 GFP expression was not detected in the developing cochlear structures until E13.5 (Fig. 2B) , as reported previously the earliest a7 expression we have defined is at P9.0 in rhombomeres 3 and 5 . Because cochlear morphogenesis includes signaling from rhombomere 5 (Liang et al. 2010) , the possibility of a7 GFP contributing to the development of this structure was examined. This was done using embryos from a7 Cre mice crossed with mice harboring the conditional ROSA26-loxp (diphtheria-A toxin (DTA; termed a7 Cre: DTA ; . In these embryos, a7 Cre: DTA -expressing cells and their direct lineages were ablated, thus revealing expression that could have been be missed by a7 GFP measurements ). An example of the cochlear structure at E16.5 taken from a7 Cre:DTA crosses is shown in Fig. 6 . Because there is only occasional overlap with a7 GFP (see Fig. 5E ), we used peripherin expression to aid in examining the fate of non-a7-expressing cells (Fig. 6A and B) . The overall patterning of the cochlear structure and the formation of major boney structures of the cochlea inclusive of the otic capsule and modiolus were intact, albeit somewhat distorted. The cochlear ducts were collapsed (Fig. 6B) , probably due to the absence or severe thinning of the distal lateral wall. Also absent was the sensory cell domain Figure 5 . The a7 GFP expression during cochlear innervation. Innervation of the developing cochlear structure is revealed by a7 GFP labeling. (A) An E13.5 sagittal section shows a group of efferent processes (arrow) that distribute to solitary fibers that are strongly labeled for a7 GFP expression (arrow heads). Cells of the putative sensory region (sr) and the spiral prominence (SP) are identified. (B) At greater magnification, these fibers (arrow heads) have a beaded appearance and project towards the base of sr. (C) The possible origin of the pioneering efferent fibers is suggested by the intense expression of a7 GFP in the E13.5 cell groups (arrow) located caudal to the trigeminal sensory nucleus (V) consistent with the early olive in this horizontal section through the posterior brain stem. At increased magnification (Insert), the cell clustering (arrow) and their projections (arrowhead) are identified. Serial sections (not shown) reveal continuity between these cells and those entering the cochlear structures (arrow heads). (D) E15.5 a7 GFP expression and colabeling with other neuronal process markers (red). The processes that express peripherin (arrow) end mostly in the vicinity of the inner hair cells (IHC). Occasional solitary fibers (arrow) extend towards the base of the outer hair cells (OHC) at the dorsal boarder of the Deiters' cells (D). (E) The E16.5 cochlear innervation pattern looks much the same as E15.5, although the peripherinlabeled fibers (arrow) are more distinct. These processes lack detectable a7 GFP expression. (F) Olivocochlear efferents identified by calcitonin generelated proteins (CGRP; arrows). (G) The E18.5 embryo exhibits afferents detected by a7 GFP expression (arrows). These extend from SG cells that are not colabeled with peripherin (not shown). (H) At birth (P0), there are distinctly labeled a7 GFP afferents (arrowhead) and peripherin-labeled efferents that extend to the Deiters' cells (D) and then turn (arrows) to contact the base of the OHCs. Hensen's cells are noted (H). (I) The P12 innervation pattern is similar to the P0. In this merged image of a7 GFP expression (green) and peripherin (red), many spiral ganglion (SG) cells and processes are labeled, but the labels only rarely overlap in the same processes (see insert). The a7 GFP identify mostly processes reaching the IHCs (arrow). Peripherin-labeled processes mostly terminate at the base of the outer hair cells (OHC) or onto the a7 GFP -labeled afferent fiber near the base of the IHC. Hensen's cells expressing a7 GFP is identified (H). The inset shows the sensory cell region at increased magnification. The arrows identify the a7 GFP -expressing afferent ending at the base of nonlabeled IHC, whereas the double arrow heads point to the peripherin-labeled terminal. Other peripherin processes extend to the base of the OHCs (individual arrow heads). (J) Diagrams as in Fig. 2 depicting the basic innervation patterns observed in this study. Green is a7 GFP and red is peripherin. Afferent (af) and efferents (ef). Bars = 50 lm containing presumptive OHCs and Deiters' cells ( Fig. 6C and D), as expected from results of a7 GFP expression (Figs. 2, 5) . The SG of a7 Cre:DTA embryos is reduced in size and the majority of cells remaining give rise to mostly peripherinlabeled efferents (see Fig. 5E ). These fibers also appear to be more densely aggregated relative to the a7 GFP control mouse ( Fig. 6A and B) . While peripherin-identified processes still project to the presumptive sensory cells (both IHC and OHC), they were less branched and those that did project to the former OHC target fields often turn and proceed backwards towards the vicinity of IHCs ( Fig. 6C and D) . These results are consistent with the earliest expression of a7 being after major cochlear structures are determined, and there was the expected selective ablation of OHCs and Deiters's cells. The necessity of the presence of the target sensory cell to coordinate the innervation process is also suggested by these findings. The results of studies examining a7 expression using in situ hybridization and functional measurements using electrophysiology have shown that this receptor is an important contributor to various nuclei of the central auditory system (Happe and Morley 1998; Vetter et al. 1999 Vetter et al. , 2007 Morley and Happe 2000; Morley 2005 ). The a7 GFP mouse system offers an excellent opportunity to view these central systems and their connections as shown in Fig. 7 . The connections between the SG and the cochlear nuclei were strongly identified at E18.5, presumably due to the dense projections from SG cells expressing a7 GFP that extend processes both to the IHC (Fig. 2 ) and the developing cochlear nuclei of the brainstem (Fig. 7A) . The expression of a7 GFP appears to intensify after P10, and by P12 signal is consolidated almost exclusively in the ventral-posterior cochlear nucleus (Fig. 7B ). This is in agreement with reports from in situ hybridization studies reporting the strong expression of a7 in this nucleus, whereas other major cochlear nuclear divisions exhibited only weak or sporadic labeling (Yao and Godfrey 1999; Morley and Happe 2000) . Also consistent with those studies was that the cells identified by a7 GFP expression resemble octopus cells (Fig. 7B, insert) . Essentially, no expression of a7 GFP was detected in the dorsal cochlear nucleus, although some dispersed and weakly stained cells were present in the granular aspect. Also evident was the strong staining of neuropil, presumably in part due to terminals of SG cells associated with the eighth cranial nerve (Fig. 7B, inset) . This strong labeling of the P12 SG and OHC afferents is consistent with other reports (Morley and Happe 2000) . The expression of a7 GFP also persists into the adult animal. This is apparent in the ascending central auditory system nuclei and their fibers (Fig. 7C) . After the cochlear nucleus, a7 GFP is present in the ventral lateral lemniscus, on through the dorsal lateral lemniscus, and to the inferior colliculus where dense staining of a7 GFP is present ( Fig. 7C ; Morley and Happe 2000; Yao and Godfrey 1999) . The commissural fibers of the inferior colliculus are also identified by a7 GFP expression (Fig. 7D) . Thereafter, efferents follow the brachium of the inferior colliculus to the medial geniculate nucleus where scattered cells expressing a7 GFP were seen. Not shown is that the expression of a7 GFP in the adult auditory cortex appears restricted to cells of layer 1. Labeling of olivocochlear fibers was not detected. This study extends the reports of spatiotemporal regulation of a7 expression during mouse embryonic development to include the cochlear sensory structure, as well as confirms the extensive expression of this nAChR in the ascending central auditory system. The novel finding that in addition to expression of a7 GFP in developing sensory cells of the cochlear structure and neuronal cells of the spiral ganglion, there is also considerable expression by nonsensory cells. Cells of the spiral prominence and ligament, Deiters' cells, and some Hensen's cells. Despite overall agreement between our studies and those using in situ hybridization (e.g., Happe and Morley 1998; Morley and Happe 2000) , these nonsensory cells were not reported previously to express a7. However, these comparisons are incomplete because the earlier studies did not necessarily show the comparable structures or the developmental stages at times where we observed peak a7 GFP expression. Also, our method of detecting GFP as a marker of a7 expression offers improved sensitivity and resolution that has previously not been available for this nAChR. The nicotinic receptors a9 and a10 are particularly well characterized in the auditory system (Elgoyhen et al. 1994 (Elgoyhen et al. , 2001a Vetter et al. 1999 Vetter et al. , 2007 Katz et al. 2004; Morley 2005) . Comparing the expression of a7 GFP to the results from these studies of the sensory hair cells and the nonsensory cells of the cochlea indicate that there are significant spatiotemporal differences during development between the expression of a7 versus a9 and/or a10. The a9 KO mouse also exhibits auditory deficiencies that are not observed in the a7 KO mouse, which is largely devoid of a phenotype in this sensory system under normal physiological conditions (Liberman and Brown 1986; Simmons and Morley 1998; Morley 2005; Lustig 2006 ). The a7 GFP is not detected in IHCs, which is consistent with a9 nAChR being the principle target of alpha-bungarotoxin in this cell type (Uziel et al. 1981; Glowatzki and Fuchs 2000) . Collectively, this suggests that functional redundancy between these receptor subtypes is unlikely (see also . This is also supported by the extensive studies by the Morley group Morley 1998, 2004; Morley and Happe 2000; Simmons and Morley 2011) who showed that multiple receptor subtypes are expressed in the cochlear and central auditory systems, but each exhibits distinct spatiotemporal patterns that likely preclude substantial or sustained functional overlap. Noteworthy is that the functional contribution of a7 towards modulating physiological systems may not be revealed unless the system is imbalanced as by genetic . Central auditory systems express a7 GFP . Central auditory nuclei identified by a7 GFP expression. (A) At E18.5 in this sagittal image of the entire otic complex and the adjacent basal brainstem is included. The cochlear nucleus (C) and the eighth cranial nerve (8n) are visible as is the fifth cranial nerve (5n), the trigeminal nucleus (TGN), and trigeminal ganglion (TGG). Also noted are cochlear ducts (asterisk) and a spiral ganglion (SG). (B) At P12, a7 GFP expression of cochlear complex reveals the strongly labeled cells of the ventralposterior cochlear nucleus (VCP). The dorsal cochlear nucleus (DC) and ventral-anterior cochlear nucleus (VCA) are identified and is the eighth nerve (8n) and a cochlear duct (asterisk). The inset shows the VCP at increased magnification. Cells clusters expressing a7 GFP (arrowhead) and individual cells that resemble the morphology of octopus cells described previously (Morley and Happe 2000; Morley 2005) to express a7 (arrow) are noted. (C) Another P12 sagittal section reveals a7 GFP expression in the ascending central auditory pathways. (D) The expression of a7 GFP in the inferior colliculus (CIC) of this horizontal section reveals staining of the commissural fibers (arrow). Structures identified are the brachium of the inferior colliculus (BIC); dentate gyrus (DG), inferior colliculus, central nucleus (CIC); lateral lemniscus, dorsal nucleus (DLL); lateral lemniscus, ventral nucleus (VLL); medial geniculate nucleus (MGN), and the substantia nigra (SN). Bars = 100 lm (A, B0; 20 lm (B-insert), and 1 mm (C, D). deficiencies, sustained exposure to pharmacological compounds, or other events such as inflammation (e.g., Faustman et al. 1992; Gahring and Rogers 2005; Venables et al. 2007; Albuquerque et al. 2009; Brown 2011; Severance et al. 2011) . For example, the dysfunction of a7 is implicated in several psychiatric syndromes associated with certain forms of autism and schizophrenia (particularly in patients who hallucinate) whose spectrum of disorders include abnormal sensitivity to sensory stimuli including an abnormal auditory gating phenotype (Khalfa et al. 2001; Veuillet et al. 2001; Araki et al. 2002; McEvoy and Allen 2002; Freedman et al. 2003; Lippiello 2006; Martin and Freedman 2007; Wallace and Porter 2011 and references therein) . Also, the association of certain auditory deficits and nicotine abuse, mostly associated with cigarette smoking, has further focused speculation on the role of a7 in these pathologies and the possible advantages of therapeutically targeting this receptor for symptomatic relief in these cases (Araki et al. 2002; McEvoy and Allen 2002; Simosky et al. 2002; Freedman et al. 2003; Levin et al. 2006; Lippiello 2006; Martin and Freedman 2007; Wallace and Porter 2011) . In this context, our results suggest additional lines of investigation. For example, in a 7Cre:DTA cell lineage ablation there are collapsed cochlear ducts and abnormal innervation indicating that the cells express a7 and the cells that do so contribute an obligatory role in the successful development and long-term function of these structures. The a7 receptor could also participate in auditory performance after birth, including functions related to the central auditory pathways. This study also adds the possibility of an effect by a7 on the performance of the spiral ligament. These cells exhibit a cholinergic response that is most often described in terms of muscarinic acetylcholine receptors (Khan et al. 2002; Maison et al. 2010) , and their dysfunction is related to several pathogenic auditory deficiencies (Spicer and Schulte 1991; Slepecky et al. 1995; Kikuchi et al. 2000; Sun et al. 2012) . The role of a7 has, to our knowledge, not been examined in these cells. Collectively, the potential for a7 functional pleiotropy in the auditory system is similar to other tissues we have recently examined . Thus, multiple defects that impact upon adult function could be expected depending upon the timing, duration, and nature of the receptor dysfunction.

Identification of serum proteomic biomarkers for early porcine reproductive and respiratory syndrome (PRRS) infection

BACKGROUND: Porcine reproductive and respiratory syndrome (PRRS) is one of the most significant swine diseases worldwide. Despite its relevance, serum biomarkers associated with early-onset viral infection, when clinical signs are not detectable and the disease is characterized by a weak anti-viral response and persistent infection, have not yet been identified. Surface-enhanced laser desorption ionization time of flight mass spectrometry (SELDI-TOF MS) is a reproducible, accurate, and simple method for the identification of biomarker proteins related to disease in serum. This work describes the SELDI-TOF MS analyses of sera of 60 PRRSV-positive and 60 PRRSV-negative, as measured by PCR, asymptomatic Large White piglets at weaning. Sera with comparable and low content of hemoglobin (< 4.52 μg/mL) were fractionated in 6 different fractions by anion-exchange chromatography and protein profiles in the mass range 1–200 kDa were obtained with the CM10, IMAC30, and H50 surfaces. RESULTS: A total of 200 significant peaks (p < 0.05) were identified in the initial discovery phase of the study and 47 of them were confirmed in the validation phase. The majority of peaks (42) were up-regulated in PRRSV-positive piglets, while 5 were down-regulated. A panel of 14 discriminatory peaks identified in fraction 1 (pH = 9), on the surface CM10, and acquired at low focus mass provided a serum protein profile diagnostic pattern that enabled to discriminate between PRRSV-positive and -negative piglets with a sensitivity and specificity of 77% and 73%, respectively. CONCLUSIONS: SELDI-TOF MS profiling of sera from PRRSV-positive and PRRSV-negative asymptomatic piglets provided a proteomic signature with large scale diagnostic potential for early identification of PRRSV infection in weaning piglets. Furthermore, SELDI-TOF protein markers represent a refined phenotype of PRRSV infection that might be useful for whole genome association studies.

Porcine reproductive and respiratory syndrome (PRRS) is one of the most important infectious swine diseases throughout the world [1] [2] [3] and is still having, more than two decades after its emergence, major impacts on pig health and welfare (reviewed by [4] ). The responsible agent is an enveloped, ca. 15 kb long positive-stranded RNA virus (PRRSV) that belongs to the Arteriviridae family [5] and that can cause late-term abortions in sows and respiratory symptoms and mortality in young or growing pigs. Once this virus has entered a herd it tends to remain present and active indefinitely causing severe economic losses and marketing problems due to high direct medication costs and considerable animal health costs needed to control secondary pathogens [6, 7] . Pigs of all ages are susceptible to this highly infectious virus, which has been shown to be present in most pigs for the first 105 days post infection [8] . However clinical manifestations vary with physiological status and age [9] , as the virus uses several immune evasion ways to complicate the ability of the host to respond to the infection process [4, 10, 11] . Weaning piglets, in particular, are likely to be exposed to the infection. Although PRRSV viraemia is often asymptomatic in these piglets, their productive performance is significantly decreased. Indeed, despite being sero-negative, persistently infected piglets still harbor PRRSV and have been shown to be a source of virus for susceptible animals [12] . SELDI-TOF MS analysis allows the comparison of protein profiles obtained from a large number of diverse biological samples by combining two principles, chromatography by retention on chip surface on the basis of defined properties (e.g. charge, surface hydrophobicity, or biospecific interaction with ligands) and mass spectrometry. It is thus distinct from common non-selective techniques, such as two-dimensional polyacrilamide gel electrophoresis (2D-PAGE) and matrix-assisted laser desorption ionisation (MALDI) MS. SELDI-TOF MS has been widely used for diagnostic biomarker discovery and validation across studies in blood serum/plasma, particularly in cancer research (reviewed by [13] ), but also to characterize and identify biomarkers associated with viral and other infectious diseases [14] [15] [16] [17] [18] [19] . The protein signatures identified by SELDI-TOF MS analysis have thus many potential applications in animal health, including early diagnosis of diseases, prediction of disease states, as well as monitoring of disease progression, recovery, and response to vaccination. Few reports have been published for livestock applications [19] [20] [21] [22] . Current needs in veterinary medicine and animal husbandry include the identification of tools that allow the early warning of diseases, especially during the incubation periods and before the onset of clinical signs. Therefore, the objective of this study was to identify by SELDI-TOF MS a proteomic profile able to differentiate PPRSV-positive from -negative weaning piglets raised in commercial farms and without clinical symptoms of the disease. We optimized the experimental conditions previously described [20] and validated 47 statistically significant discriminatory biomarkers. Among these, a combination of 14 biomarkers identified in F1 on CM10 at low focus mass permitted to correctly assign the piglets to the PPRSV-positive or PRRSV-negative groups with sensitivity and specificity of 77% and 73%, respectively. To enable identification of medium-low abundant proteins, only samples with a total content of hemoglobin lower than 4.52 μg/mL were included in the study. Total hemoglobin absorbance and the resulting hemoglobin content were calculated for all the piglet sera in both discovery (n = 50) and validation (n = 70) phases of the study [Additional file 1: Table S1 and Additional file 2: Table S2 , respectively]. Fractioning of the sera resulted in six different pH fractions; F1 = pH9, F2 = pH7, F3 = pH5, F4 = pH4, F5 = pH3, and F6 = organic solvent. The fractions F1, F4, and F6 were analyzed on the three surfaces CM10, IMAC30, and H50 at both low and high focus masses. Fractions F2 and F3 were excluded from further analyses because preliminary data with 3 serum samples showed that they still contained elevated quantities of abundant proteins (such as albumin), as well as the quality of the spectra and the number of signals detected were very low. Fraction F5 was excluded because no signals were detected. The fractions F1, F4, and F6 on the surfaces CM10, IMAC30, and H50 showed generally good signal intensities and low coefficient of variation (CV) values (< 30%) in both the discovery and validation phases. Exceptions were fraction F1 on IMAC30 (analyzed at high focus mass) and H50 (both low and high focus masses), as well as fraction F4 on H50 (low focus mass), which were therefore excluded from further analyses. A total of 50 pig sera, 25 from PRRSV-positive and 25 from PRRSV-negative piglets were analyzed during the discovery phase of the study [Additional file 1: Table S1 ]. We found a total of 785 protein peaks in the sera of all samples (Table 1 ). The most represented pH fraction was F6 (n = 381), followed by F4 (n = 223), and F1 (n = 181). On surface CM10 we identified 317 peaks, on IMAC30 302 peaks, and on H50 166 peaks. Furthermore, a much higher number of peaks (n = 512) was found on low mass range (1-20 kDa) compared to the high (n = 273; 20-200 kDa). Of the total 785 peaks, 200 were statistically significant (p < 0.05) and permitted to discriminate between PRRSV-positive and PRRSV-negative piglets. Discriminatory peaks were found in F1 (n = 80), F4 (n = 49), and F6 (n = 71) on the surfaces CM10 (n = 107), IMAC50 (n = 58), and H50 (n = 35), as well with low (n = 110) and high (n = 90) focus masses ( Table 1) . The highest sensitivity (80%) and specificity (76%) were obtained with the 22 discriminatory peaks of F1 on CM10 at low focus mass. Higher sensitivities were found with the 18 peaks of F4 on CM10 at low focus mass (87%), the 7 peaks of F6 on CM10 at low focus mass (85%), and the 12 peaks of F6 on CM10 at high focus mass (87%), however the specificities of these peaks were lower (64%, 66%, and 66%, respectively). The validation phase was performed on 35 new PRRSVpositive and 35 new PRRSV-negative piglets using the same experimental conditions applied in the discovery phase [Additional file 2: Table S2 ]. Of the total 200 peaks that were significant in the discovery phase, 47 were confirmed in the validation phase (Table 2 ). In particular, 28 peaks were confirmed on CM10, 19 on IMAC30, whereas none of the peaks could be validated on the surface H50. In the 3 fractions with different pH tested, F1 contained 28 peaks, F4 3 peaks, and F6 16 peaks. A higher number of peaks (n = 36) corresponded to small peptides (acquired at low focus mass 1-20 kDa), compared to big peptides (n = 11) that were acquired at high focus mass (Table 2 ). In line with the results of the discovery phase, the combination of peaks with the highest sensitivities (77% and 64.5%) and specificities (73% and 69.7%) were found on CM10 at low focus mass with the 14 discriminatory peaks of F1 and the 6 discriminatory peaks of F6, respectively ( Table 2 ). The correctly and incorrectly assigned piglets using these peaks are graphically illustrated in the heat map of Figure 1 ; part 1A shows the 14 peaks of F1 and part 1B the 6 peaks identified in F6. Principal component analysis (PCA) was performed on the profiles of the 47 discriminatory peaks identified during the discovery and confirmed during the validation phase to identify and quantify independent sources of variation observed in the data. PCA analysis showed that 58.2% (PCA1), 17.9% (PCA2), and 12.9% (PCA3) of the total variability within the data was accounted for the X, Y, and Z axes, respectively. These axes were used to plot the data ( Figure 2 ) and they provide an overview of the variation between the individual samples and show how samples grouped. Figure 2A showed three-dimensionally that the PCA peak profiles of piglets positive to PRRSV differed from piglets negative to PRRSV and revealed a good separation among the profiles of the two different groups, especially considering the high heterogeneity of the samples included in the study, as reported in the MM section and in [Additional file 1: Table S1 and Additional file 2: Table S2 ]. Furthermore, with the exception of few The 785 total number of peaks detected and the 200 statistically significant (p < 0.05) discriminatory peaks associated with PRRS infection that were identified by the Ciphergen Express software are reported with the fraction, the array surface, and the acquisition focus mass (low: 1-20 kDa; high: 20-200 kDa). outliers, PCA1 combined with PCA2 also separated well the two piglet populations ( Figure 2B ). To provide an overview of the current literature and to try to correlate the discriminatory peaks identified in this study with relevant proteins, we summarized in Table 3 the molecular weights of several peaks that have been shown to be related to PRRSV infection. First of all, we summarized the available information on the PRRS viral proteins. The PRRSV genome is ca. 15 kb in size and consists of the 5' untranslated region (UTR), at least nine open reading frames (ORFs), and Table 3 , along with the MW of the closest discriminatory peak identified in the current study. Interestingly, the MW of the viral proteins ORF2b, ORF4, and ORF7 were very similar (difference of MW ≤0.3 kDa) to up-regulated discriminatory peaks identified here ( Table 3) . As next, we compared proteins related to PRRSV infection that were identified in additional studies (Table 3) ; interestingly, all the 9 peaks found by [28] , and in particular the only up-regulated in PRRSV infected (corresponding to the Alpha 1 S (a1S)-subunit of porcine Haptoglobin), showed minimal MW differences (≤0.3 kDa) with up-regulated peaks identified in this study (Table 3) . Additional discriminatory peaks found in the current study were very similar (MW differences ≤0.3 kDa) to those identified in other PRRS-related proteomic studies (Table 3) . They corresponded to the following proteins: Glyceraldehyde-3-phosphate dehydrogenase, Proteasome activator hPA28 subunit beta, S100 calcium binding protein A10, Galectin 1, and Gastric-associated differentially expressed protein YA61P [26] ; Heat shock 27 kDa protein 1, Superoxide dismutase 2, Myoglobin, and Vacuolar protein sorting 29 [29] ; Heat shock protein 27 kDa and Nucleoside diphosphate kinase A [30] ; Heat shock 27 kDa protein 1, Galectin 1, and Ubiquitin [31] . In the present work, we show that proteomic fingerprint profiling is useful in researches on PRRS immuno-pathogenesis and might also be a robust, large scale diagnostic tool for the assessment of the proportion of PRRSV-positive weaning piglets without clinical symptoms in a herd. Indeed, we confirmed that the high-throughput capacity of the SELDI-TOF MS technology allows the screening for disease biomarkers of hundred of samples in a relative short-time period and with minimal sample preparation (as previously also reported by [32] ). Our results indicate that from the 200 significant peaks found in the discovery phase, a total of 47 could be confirmed in the validation phase. These values are comparable with another study where similar experimental conditions were applied to ovine sera [19] . Our findings also show that the combination of 14 discriminatory peaks in F1 on CM10 at low focus mass provided the highest sensitivity of 77% and specificity of 73% to correctly assign the piglets to the PPRSVpositive or PRRSV-negative groups. These percentages are in line with recent studies in humans using the [33, 34] . Also the PCA results showed a good separation of the piglets in the two groups under examination. This was reached even though the tested piglets had large variability and heterogeneity, as they were collected from several farms located in different regions, and underwent high environmental pressures, typical of the field conditions. This is mainly due to the careful choice of the serum samples, where we tried to minimize the environmental differences by using same experimental parameters (e.g. sample collection procedures, storage, handling) and by including a similar number of pigs from the same breed (Large White) and with very similar sex ratios and ages (at weaning). In a preliminary work [20] we had successfully transferred the experimental conditions used in profiling experiments of human sera to pig sera. However, in that work, none of the potential biomarkers identified in the discovery phase could be validated in the subsequent validation phase, because of high samples heterogeneity and high content of serum (e.g. albumin) and contaminant proteins (e.g. hemoglobin), having a negative effects on the detection of significant biomarkers, particularly those corresponding to the medium-low abundant proteins. It has been reported that low abundant proteins constitute about 1% of the entire human serum proteome, with the remaining 99% being comprised of only 22 proteins [35] . As it was therefore necessary to reduce the level of abundant proteins, in this follow up study, particular relevance was given to the content of the contaminant protein hemoglobin. Only non-hemolytic samples with similar, low contents of hemoglobin were included in the study. Additionally, to further increase the likelihood to identify statistically significant discriminatory biomarkers, we introduced a fractioning step based on anion-exchange chromatography. In a similar study performed with MALDI-TOF [28] , where serum samples were analyzed in the first weeks (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) Table 3 ). Furthermore, two peaks identified in this study (23. 162 and 14.843 kDa) were similar to peaks identified elsewhere (corresponding to Heat shock 27 kDa protein 1 [29] [30] [31] and Galectin 1 [26, 31] , respectively). In accordance with [31] , the identified peak corresponding to Heat shock 27 kDa protein 1 was upregulated, while the peak corresponding to Galectin 1 was down-regulated. Thus, these proteins seem to be very interesting and suitable candidates for future investigations. The preponderance of the significant biomarkers had a molecular mass lower than 20 kDa, confirming that small peptides are a rich source of relevant biomarkers in SELDI-TOF MS analyses as previously reported in human [36] and ovine [19] sera. This may also partly be caused by the fact that the low molecular weight region (LMW) of the serum proteome, called peptidome, is an assortment of small intact proteins and proteolytic fragments of larger proteins, including several classes of physiologically important proteins like peptide hormones and components of both the innate and adaptive immune systems (i.e. cytokines and chemokines) [35, 37] . This is particularly interesting as the patho-physiological state of the body's tissue is predominantly reflected in the LMW and low abundance region of the serum proteome, and specific protein fragments of the serum peptidome have been shown to contain a rich source of disease-specific diagnostic information and they have been correlated with disease stages in several studies (reviewed by [37] ). In agreement with other studies [29, 31] , we found that the majority of the discriminatory biomarkers were up-regulated in PRRSV-positive piglets. This seems to suggest that the corresponding proteins might be of viral origin or related to the innate or adaptive immune responses (e.g. cytokines, chemokines, acute phase proteins, toll like receptors). In fact, several peaks showed high similarities (MW differences ≤0.3 kDa) with previous works, in particular regarding viral proteins ( Table 3 ). The assignment of the discriminatory peak to a specific protein will require additional work, because the SELDI-TOF technology can only detect masses/peaks of proteins that are differentially expressed between samples but can not directly identify the proteins. This represents one of the major drawbacks of this technology compared to other methods. However, an advantage of the SELDI-TOF MS in this regard is that the results of this technique might lead to the identification of new proteins that were previously not correlated to the disease, and this might hopefully lead to the identification of new biomarkers representing the field situation. The interpretation of these results and the continuation of this project will benefit from the very imminent termination and publication of the sequence of the swine genome [38] , which will definitely contribute to a more precise annotation and a better identification of genes and proteins and thus will greatly facilitate genome wide mapping association studies. Although a combination of peaks identified with different experimental conditions (e.g. using different fractions and different surfaces) might have provided higher discriminatory power, here we developed a PRRSV diagnostic test based on peaks identified with the same experimental conditions (e.g. fraction, surface, and focus mass), which can be reproduced at high-throughput at reasonable costs. These results provide a set of proteomic biomarkers and related, optimized experimental conditions for high-throughput profiling of pig populations by SELDI-TOF MS for whole genome association studies, where identification of proteins underlying the phenotype can be made a posteriori. SELDI-TOF MS might therefore represent a complementary test or a possible alternative to classical (PCR) and more recent diagnostic methods (e.g. antibody detection in saliva) for profiling large flocks of pigs at reasonable costs, using blood samples that are routinely collected for general veterinary inspections. As well, these SELDI-TOF MS based tests could complement and provide a broader reference for emerging diagnostic methods and have potential applications for the detection of relevant proteins having highly heritable traits (e.g. acute phase proteins). A total of 120 serum samples of Large White piglets were selected from a well defined and characterized repository database, presently containing more than 20,000 swine samples from 18 different farms of the Lombardy region, Italy. Selection of the piglets aimed to minimize environmental factors and experimental conditions that might influence the results [39] . Hence, all piglets were from the same breed (Large White), had similar ages (weaning: 45-50 days), and their sera showed a low and comparable amount of hemoglobin (calculated as shown below). In the discovery phase of the study, a total of 50 pig sera, 25 from PRRSV-positive and 25 from PRRSVnegative piglets, as determined by PCR (see below), were analyzed [Additional file 1: Table S1 ]. The validation phase was performed with the same experimental conditions as the discovery phase. A total of 35 new PRRSVpositive and 35 new PRRSV-negative piglets were examined [Additional file 2: Table S2 ]. The actual duration of infection for each individual PRRSV-positive piglet was unknown, as sera were collected and analyzed once for each piglet (at weaning: 45-50 days of age). None of the piglets was treated, as they did not show any symptom of the disease. To ensure large variability and heterogeneity of the samples and minimize environmental differences, we included in the PRRSV-positive and -negative groups similar numbers of piglets with the same sex that originated from several farms located in different regions. In fact, PRRSV-positive piglets originated from 6 farms of the Lodi region (n = 8) and 7 farms of the Mantua region (n = 52), while PRRSV-negative piglets were collected in 5 farms around Lodi (n = 19) and 9 farms around Mantua (n = 41). Sex ratios males/females (44/76) were very similar in PRRSV-positive (21 vs. 39) and -negative (23 vs. 37) piglets, respectively. Veterinary inspections of the overall clinical status of the piglets at the day of serum collection did not evidence any clinical symptoms of PRRS, including respiratory distress or sneezing. All the serum samples were collected, stored, and handled in the same way. They were obtained for each piglet by storing two mL of whole blood without anticoagulants at room temperature (RT) for 4 h followed by centrifugation at 3,500 rpm for 4 min. As suggested in a previous work [20] , an abundant quantity of hemoglobin in the serum can hide early diagnostic biomarkers of PRRSV by competing with the other serum components for the binding site of the chromatographic surfaces. To avoid the consequent signal suppression of the medium-low abundant proteins, only non-hemolytic samples were included in the present study. A total of 200 clear, transparent sera without red pigmentation (low hemoglobin content) were first selected by visual screening from the total sera available in the database. Hemoglobin content of each serum sample was then determined according to [40] with minor modifications. A calibration curve was generated using five standard solutions (concentrations: 1.8, 3.6, 5.4, 7.2, and 9 μg/ml) of porcine hemoglobin diluted in 400 μL commercially available porcine serum (Sigma Aldrich, St Louis, MO, USA). Triplicate samples were incubated for 5Ámin at RT, then absorbance (E) was measured at 380, 415, and 440 nm. Absorbance at 380 and 440 nm was used to discern background absorbance flanking the absorbance peak (415Ánm) of oxygenated hemoglobin. Absorbance due to hemoglobin was calculated as: E415-[(E380 + E440)/2]. Hemoglobin absorbance values of the samples were converted to μg/mL of hemoglobin by means of the calibration curve. Of the 200 initial samples, a total of 120 samples having an absorbance ≤ 0.085 (corresponding to a hemoglobin content below 4.52 μg/mL) were included in the study; 50 in the discovery and 70 in the validation phases, respectively. Viral RNA extraction from the sera was performed following standard Roche procedures (High Pure Viral RNA Kit, Roche Diagnostics GmbH, Germany). Presence or absence of PRRSV was determined by multiplex PCR of conserved regions of viral ORF7 using primers and conditions previously described [41, 42] . The test also enabled to discriminate European and American genotypes and could detect all the different viral strains present in the Lombardy region at the time of sample collection. All the detailed steps of the SELDI-TOF MS process performed here are schematically represented [see Additional file 3: Figure S1 ]. The protocol follows the manufacturer's instructions with minor modifications (Bio-Rad Laboratories, ProteinChip W Serum Fractionation Kit manual). Briefly, serum samples were pre-fractionated with U9 buffer (9 M urea, 2% 3-[(3-Cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS), 50 mM Tris-HCl, pH = 9) to favor dissociation of protein complexes [Additional file 3: Figure S1A ]. Sera were fractionated using a ProteinChip Q strong anion-exchange resin filtration plate (Bio-Rad Laboratories, Hercules, CA). The filtration plate was re-hydrated and equilibrated with rehydration buffer (50 mM Tris-HCl, pH = 9) and the resin washed with rehydration buffer and U1 solution (1 M urea, 0.2% CHAPS, 50 mM Tris-HCl, pH = 9) [Additional file 3: Figure S1B ]. Serum samples were then mixed with U1 solution and added to the equilibrated filtration plate. Successive elutions with different buffers with decreasing pH and a final organic solvent (= different fractions) were collected by centrifugation. The buffers used included pH = 9 (50 mM Tris-HCl, 0.1% n-octyl β-D-glucopyranoside (OGP)), pH = 7 (50 mM 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES), 0.1% OGP), pH = 5 (100 mM Na acetate, 0.1% OGP), pH = 4 (100 mM Na acetate, 0.1% OGP), pH = 3 (100 mM Na citrate, 0.1% OGP), and organic solvent (33.3% isopropanol, 16.7% acetonitrile, 0.1% trifluoroacetic acid) [Additional file 3: Figure S1C ]. The six pH fractions obtained (F1 = pH9, F2 = pH7, F3 = pH5, F4 = pH4, F5 = pH3, and F6 = organic solvent) were profiled on weak cation-exchange (CM10), immobilized metal affinity capture-copper (IMAC30-CU), and reverse-phase (H50) ProteinChip W arrays. The arrays were initially placed in a Bioprocessor (C50-30011, Bio-Rad Laboratories) and then treated according to their surface [Additional file 3: Figure S1D ]. Each sample fraction was then bound/spotted randomly to the different Pro-teinChip W arrays using array-specific binding buffers [Additional file 3: Figure S1E ]. A 50% saturated sinapinic acid (SPA) matrix solution was finally added to each spot on the ProteinChip array prior to the final analysis [Additional file 3: Figure S1F ]. ProteinChip arrays were read using a Ciphergen Protein-Chip Reader PCS4000 model and data were analyzed with Ciphergen Express Software (Ciphergen Biosystems). Profiles were collected in the range 1-200 kDa at the two different ion focus mass 10 kDa ("low focus mass") and 50 kDa ("high focus mass"). The instrument was calibrated for dataset collection using all-in-one peptide standard (Bio-Rad Laboratories) in the 1-20 kDa range for 10 kDa low ion focus mass and all-in-one protein standard in the 20-200 kDa range for 50 kDa high ion focus mass [Additional file 3: Figure S1G ]. Spectra were normalized by total ion current, starting and ending at the M/Z of the collection ranges (1-20 or 20-200 kDa) after baseline subtraction and noise calculation. Outlier spectra were removed. The spectra were aligned to a reference spectrum with the normalization factor nearest 1.0. The spectra were aligned only if the percentage coefficient of variation was reduced after the alignment. Peaks from the different spectra were aligned using the cluster wizard function of the Ciphergen Express 3.0.6 software. The peak detection was automated within the M/Z range of analysis. Peaks were detected on the first pass when the signal-to-noise (S/N) ratio was 7 and the peak was 5 times the valley depth. Peaks below threshold were deleted and all first-pass peaks were preserved. Clusters were created within 0.15% of M/Z for each peak detected in the first pass. The clusters were completed by adding peaks with S/N ratio of 2 and two times the valley depth. P-values and ROC/AUC (Receiver Operating Characteristic/ Area Under Curve) values were calculated by using the P-value wizard. A 2-tailed t-test was used for statistical analysis of differences in peak intensity between groups. P-values below 0.05 were considered statistically significant. Principal component analysis (PCA) and agglomerative hierarchical clustering algorithm were applied to investigate the pattern among the different statistically significant peaks. PCA is a multivariate data analysis that transforms without a loss of essential information a number of correlated variables into a smaller number of uncorrelated variables called principal components (PCs), which can explain sufficiently the data structure. PCA transformation allows studying many variables simultaneously, showing how similar samples are correlated and grouped together. The data structure is visualized directly in a graphical way by projection of objects onto the space defined by the selected PCAs (for details see [43] ). Finally, to evaluate the influence of external variables (e.g. sample processing and acquisition) on the system under study and to calculate the dispersion of the acquired data, the coefficient of variation (CV), which is the normalized measure of dispersion of a probability distribution and shows the% dispersion of the data in rapport to the media (intensity variation), was also calculated. Six serum samples commercially available were prepared and analyzed in parallel with the pig samples of both, discovery and validation phases. The CV was calculated for all fractions and surfaces by choosing 6 peaks evenly distributed along the entire range. Additional file 1: Table S1 . Pigs tested with SELDI-TOF MS during the discovery phase of the study. List of the 25 positive and 25 negative pigs to PRRS (PCR-tested) analyzed with SELDI-TOF MS during the discovery phase of the study. The pig ID is reported with the total absorbance and the total amount of hemoglobin present in the sample, the status regarding the PRRS virus, as well as the sex and the number and location of the farm (MA = Mantua region, LO = Lodi region). Additional file 2: Table S2 . Pigs tested with SELDI-TOF MS during the validation phase of the study. List of the 35 positive and 35 negative pigs to PRRS (PCR-tested) analyzed with SELDI-TOF MS during the validation phase of the study. The pig ID is reported with the total absorbance and the total amount of hemoglobin present in the sample, the status regarding the PRRS virus, as well as the sex and the number and location of the farm (MA = Mantua region, LO = Lodi region).

Production and characterization of human anti-V3 monoclonal antibodies from the cells of HIV-1 infected Indian donors

BACKGROUND: Analysis of human monoclonal antibodies (mAbs) developed from HIV-1 infected donors have enormously contributed to the identification of neutralization sensitive epitopes on the HIV-1 envelope glycoprotein. The third variable region (V3) is a crucial target on gp120, primarily due to its involvement in co-receptor (CXCR4 or CCR5) binding and presence of epitopes recognized by broadly neutralizing antibodies. METHODS: Thirty-three HIV-1 seropositive drug naive patients (18 males and 15 females) within the age range of 20–57 years (median = 33 years) were recruited in this study for mAb production. The mAbs were selected from EBV transformed cultures with conformationally constrained Cholera-toxin-B containing V3C (V3C-CTB) fusion protein. We tested the mAbs for their binding with HIV-1 derived proteins and peptides by ELISA and for neutralization against HIV-1 viruses by TZM-bl assays. RESULTS: We isolated three anti-V3 mAbs, 277, 903 and 904 from the cells of different individuals. The ELISA binding revealed a subtype-C and subtype-A specific binding of antibody 277 and 903 while mAb 904 exhibited cross reactivity also with subtype-B V3. Epitope mapping of mAbs with overlapping V3 peptides showed exclusive binding to V3 crown. The antibodies displayed high and low neutralizing activity against 2/5 tier 1 and 1/6 tier 2 viruses respectively. Overall, we observed a resistance of the tier 2 viruses to neutralization by the anti-V3 mAbs, despite the exposure of the epitopes recognized by these antibodies on two representative native viruses (Du156.12 and JRFL), suggesting that the affinity of mAb might equally be crucial for neutralization, as the epitope recognition. CONCLUSIONS: Our study suggests that the anti-V3 antibodies derived from subtype-C infected Indian patients display neutralization potential against tier 1 viruses while such activity may be limited against more resistant tier 2 viruses. Defining the fine epitope specificities of these mAbs and further experimental manipulations will be helpful in identification of epitopes, unique to clade C or shared with non-clade C viruses, in context of V3 region.

The non-covalently associated surface (gp120) and transmembrane (gp41) subunits of the envelope glycoprotein are decorated on the surface of Human Immunodeficiency Virus Type-1 (HIV-1) as a trimeric spike [1] , and serve as a target for broadly neutralizing monoclonal antibodies (bNAbs) [2] [3] [4] . Because of its involvement in the primary steps of receptor [5] and co-receptor binding [6] , the envelope gp120 has been identified as a major target for HIV-1 NAbs [2, [7] [8] [9] [10] . However, the antigenic variability of exposed regions and low immunogenicity of conserved domains on gp120 impose great challenges to identify the vulnerable targets on HIV-1 [2, 4, 11] . Nevertheless, the conserved epitopes on gp120 have been identified using antibodies from neutralizing sera [12] [13] [14] and bNAbs [9, 10, [15] [16] [17] , which include the antibodies directed to the CD4 receptor binding site (CD4bs) and co-receptor binding site mainly the third variable region (V3) [10, [18] [19] [20] [21] [22] . The crystal structure of V3 resolved recently shows that V3 protrudes~30 Å from the CD4-bound gp120 core, and this extended structure can be divided into three regions: the base (residues 1-8 and [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] , the stem (9-14 and 18-24) and the crown (residues 15-17) (residue numbering w.r.t. V3) [23, 24] . The V3 loop of gp120 is highly antigenic in humans [25] [26] [27] [28] , and was previously recognized as the principal neutralizing domain of HIV-1 [29] . However its role was shown to be restricted to type specific viruses [30, 31] and such an observation was supported by the extensive sequence variation in V3 from different viral isolates [32, 33] . Given the critical interaction with the co-receptors (CXCR4 or CCR5) on host cells, V3 conventionally must retain structurally conserved elements required for binding [34, 35] . More recently, studies have revealed that the V3 domain possesses conserved structural motifs despite the sequence variation, and is often accessible on the virus surface as a target for bNAbs [21, [36] [37] [38] . Although, the V3 loop displays high structural conservation, yet the degree of cross reactive anti-V3 antibody response in individuals infected with diverse HIV-1 subtypes varies substantially [39] . This difference in antibody response to V3 loop, has been shown to be primarily determined by the four amino residues in the V3 crown (GPGQ or GPGR), which mostly form a type II β-turn [19, 40] . Interestingly, the anti-V3 monoclonal antibodies (mAbs) isolated from non-clade B infected individuals, bearing GPGQ at the tip of V3 display better neutralization capacity than subtype-B (having GPGR) derived anti-V3 mAbs [19] . Such an observation was substantiated by a study showing a high neutralization potential of the anti-V3 mAbs derived from Cameroon subjects infected with viruses harboring GPGQ (subtype-AG) at the V3 crown [21] . Further, in an immunization study carried out in rabbits with a gp120 DNA prime followed by a boost with various cholera toxin B (CTB) fusion proteins containing the V3 sequences from different HIV-1 subtypes, the CTB fusion protein containing a consensus-C (con-C) V3 sequence (V3C-CTB) (having a GPGQ motif at V3 crown), elicited a highly potent HIV-1 neutralizing response compared to the other V3-CTB constructs [41] . A limited number of mAbs have been generated so far against the HIV-1 subtype-C viruses including mAbs against the V3 loop despite the fact that subtype-C accounts for more than 50% of the global HIV-1 infections [42] . The only known human anti-V3 bNAb is from a clade B infected patient [15] . Keeping in view the above facts, we generated here three anti-V3 human mAbs from HIV-1 infected Indian patients, using the EBV immortalization method of human hybridoma technology. The functional analysis of the HIV-1 antibodies generated in this study revealed cross reactive binding and neutralization of the viruses tested. A total of 3321 culture wells (96 well plate) of PBMCs derived from 33 HIV-1 infected patients were established. After 2-3 weeks, the culture supernatants of approximately 6% transformed wells tested positive for reactivity with V3C-CTB (Table 1) . Three heterohybridomas producing anti-V3 mAbs were generated from different individuals (IDs; 277, 903 and 904). The mAbs belong to IgG1 subclass with one (904) using a lambda and the other two (277 and 903) with kappa light chain genes. Interestingly, the kappa light chain Abs displayed the same immunoglobulin heavy chain variable (IGHV) gene usage (3-30*03) while the other Ab (904) showed 1-8*01 gene usage ( Table 2 ). The complementarity determining region 3 for the heavy chain (CDRH3) were different for each mAb indicating their uniqueness ( Table 2) . Cross reactive binding and epitope mapping of anti-V3 mAbs by ELISA All the plasma samples from patients recruited for this study, were previously screened for their relative binding to V3C and V3B peptides (Andrabi et al., submitted) and the data is here shown for only three samples from which anti-V3 mAbs were isolated ( Figure 1C ). The peptide binding assays revealed the cross reactive binding potential of the anti-V3 Abs in the plasma sample 277 and 904 while 903 showed subtype-C V3 specific binding. In order to determine the specificity of the anti-V3 mAbs generated from these three patients, we performed the ELISA binding titration of the anti-V3 mAbs with recombinant envelope gp120 proteins and HIV-1 derived peptides (Table 3 ). In addition to three Indian anti-V3 mAbs, which were generated in this study, we also tested the binding of an anti-V3 bNAb (447-52D), isolated from a subtype-B HIV-1 infected individual living in USA. The binding curves of mAbs to con-C and B V3 peptides and subtype-C (Du156.12) and B (JRFL) gp120 proteins are depicted in the representative Figure 1 . The mAbs were also tested with a series of peptides and proteins for quality control purpose and the 50% maximal binding (Max50) binding titers are summarized (Table 3) . Overall, the mAbs 277 and 903 showed binding to subtype-A or C V3 while the antibody 904 also reacted with subtype-B V3 region ( Figure 1A -B, Table 3 ). We further tested the mAbs by ELISA binding assays using Three anti-V3 monoclonal Abs listed were derived from HIV-1 infected Indian donors. 2 Immunoglobulin gene usage for heavy (IGHV) and amino acid sequences of 3 CDRH3 domains was determined using IMGT system; an asterisk indicates allele. Figure 1 Relative binding affinity of anti-V3 antibodies to HIV-1 derived peptides and proteins. The binding pattern of anti-V3 mAbs derived from Indian donors (red) and 447-52D (an anti-V3 Ab isolated from HIV-1 subtype-B infected American individual) (green) to consensus-C and B V3 peptides (1A), and subtype-C (Du156.12) and subtype-B (JRFL) derived envelope gp120 proteins (1B). The binding of anti-V3 mAbs was tested by ELISA using mAbs at a concentration ranging from 10 to 0.00003 μg/ml (12 dilutions). Human anti-parvovirus B19 mAb 1418 was used as negative control. Relative reactivity of anti-V3 plasma antibodies to consensus-B and C V3 peptides is shown in terms of 50% ELISA binding titers (Max50) for three patients from whom antibodies were generated (1C). Two plasma samples 277 and 904 showed cross clade reactive binding while 903 displayed subtype-C specific binding. linear overlapping peptides encompassing mainly V3 region flanked with second constant (C2) and third constant (C3) region of gp120, to identify the core epitope recognized and found that all these anti-V3 mAbs, including mAb 447-52D, bind to the crown of V3 loop ( Table 4) . The binding of Abs to HIV-1 derived peptides or proteins does not necessarily mean that these mAbs will be able to bind intact viruses, which essentially present a more native conformation. In order to address this possibility, we tested the anti-V3 mAbs for binding with Du156.12 (subtype-C) and JRFL (subtype-B) viruses in an intact virion binding assay. The two viruses (Du156.12 and JRFL) were chosen for this experiment on the basis of their V3 sequence similarity with the corresponding con-C and B V3 sequence ( Figure 2C ). Consistent with the binding of anti-V3 mAbs to Du156.12 and JRFL derived envelope gp120 proteins, the three mAbs showed differential binding affinity to intact virions, 904 displaying high affinity as compared with 277 and 903. In addition, the mAb 904 retained the crossreactive binding potential to the intact viruses, evident from the binding pattern ( Figure 2A ). The experiment was validated by testing the intact virion binding of anti-V3 mAb 447-52D to SF162 virus, known to have a well exposed V3 region and allows accessibility to Ab without any interference [43] [44] [45] . The binding of 447-52D to SF162 intact virus revealed a very high binding affinity (more than five folds) of mAb 447-52D as compared to binding of three anti-V3 mAbs to Du156.12 and JRFL at equivalent viral concentration (i,e 25 ng/ml) (Figure 2A -B). This differential binding of mAbs to intact virions could also be attributed to the number of potential N- Table 3 Binding of mAbs to subtype-A, B and C derived HIV-1 proteins and peptides 2 anti-V3 mAbs 1 Protein/peptide Subtype I 277 I 903 I 904 A 447-52D 1418 List of recombinant proteins and peptides derived from subtype-A, B and C HIV-1 viruses, NA: Not applicable. 2 Four anti-V3 antibodies, three from Indian I (277, 903 and 904) and one from American A (447-52D) HIV-1 infected donor were tested for their relative binding. The binding activity of anti-V3 mAbs against proteins and peptides was tested by ELISA using mAbs at a concentration ranging from 10 to 0.00003 μg/ml (12 dilutions). The 50% binding titers (Max50, conc. μg/ml) of each antibody against the corresponding protein or peptide is depicted as numerical values in Bold (high affinity), Italic (low affinity) and >10, where Max50 value was not reached. Human anti-parvovirus B19 mAb 1418 was used as negative control. Each experiment was performed at least two independent times. Amino acid sequences of linear overlapping peptides encompassing the third variable region (V3: Underlined (middle)), flanked by second (C2: (left)) and third (C3: (right)) constant regions and are aligned with the corresponding consensus-C gp120 sequence. 2 Four anti-V3 antibodies, three from Indian I (277, 903 and 904) and one from American A (447-52D) HIV-1 infected donor were tested for their binding to overlapping peptides by ELISA using mAbs at a concentration ranging from 10 to 0.00003 μg/ml (12 dilutions). The 50% binding titers (Max50, conc. μg/ml) of each antibody against the corresponding peptides is depicted as numerical values in Bold (high affinity), Italic (low affinity) and >10, where Max50 value was not reached. Human anti-parvovirus B19 mAb 1418 was used as negative control. Each experiment was performed at least two independent times. glycosylation sites (PNGS) within the V1/V2 region (3 for SF162, while 6 and 7 PNGS for Du156.12 and JRFL respectively) of these viruses [46] . Overall the results suggest that the V3 epitopes recognized by three anti-V3 mAbs were exposed on both Du156.12 and JRFL, however mAbs 277 and 903 were not able to bind JRFL due to their subtype-C specific binding activity. The anti-V3 mAbs were tested to assess their capacity to neutralize a panel of eleven tier 1 and 2 viruses from different HIV-1 subtypes using the standard TZM-bl cell assay. The anti-V3 mAbs showed effective neutralization against two, a subtype-A (DJ263) and a subtype-C (MW965), out of five tier 1 viruses while only one virus (a subtype-C, HIV-001428) out of six tier 2 viruses was neutralized by two of the mAbs (903 and 904) with relatively low efficiency ( Table 5) . None of the Indian anti-V3 mAbs were able to neutralize any of the subtype-B viruses, despite antibody 904 showing cross reactive binding, nevertheless, the number of viruses tested here was limited. In contrast, the mAb 447-52D reached IC50 neutralization titers with 5/11 viruses tested, consistent with the previous studies [15, 19] . Human monoclonal antibodies against the HIV-1 envelope glycoproteins are useful tools in the structural and functional analysis of the viral envelope and have crucial roles in guiding the design of prophylactic anti-HIV vaccines. Despite a huge expansion of HIV-1 subtype-C worldwide, the clade-C viruses remain to be one of the least studied subtypes especially in terms of HIV-1 neutralizing antibodies. Using the rationale from previous studies showing that viruses with GPGQ residues at the tip of the V3 crown of the HIV-1 envelope induce potent and cross reactive NAbs as compared to viruses with GPGR motif, we generated here three anti-V3 mAbs from Indian donors presumably infected with subtype-C viruses bearing GPGQ at the V3 crown [42] . The functional analysis of the Abs generated reveals a potent neutralization potential with tier 1 viruses while such activity was limited with the tier 2 viruses tested. The anti-V3 Abs were selected from EBV-transformed B-cell cultures of 33 HIV-1 infected antiretroviral drug naïve patients using V3C-CTB fusion protein [47] . The advantage of using a conformationally constrained instead of a linear V3 peptide for selection of mAbs from cultures or as animal immunogens has been previously demonstrated [36, 41, [48] [49] [50] . We found 1-25% (mean = 6%) of the transformed wells positive for binding with V3C-CTB in the first screening. The characteristic nature of the B-cells from HIV-1 infected subjects and the conditions used to immortalize them apparently affects the number and type of Ab-producing cell lines that grow out [51] . The overall positive percentage of Ab secreting culture wells was relatively good and could be attributed to the high titers of anti-V3 Ab reactivity of the corresponding plasma [28] . In contrast to a high percentage of positive secretors in the initial screening, we were able to stabilize only three (277, 903 and 904) anti-V3 Ab producing B-cell clones. This loss could be in part due to the outgrowth of the non-secretor B cells over the secretor B cells in subsequent steps of secondary screening, cell fusion and dilution cloning process. Moreover, the B cells from HIV-1 infected patients are mostly dysfunctional and polyclonally activated [52] , and such properties have been associated with a low persistence of EBV infectivity [51, 53, 54] . The amino acid sequence variation of V3 across the various HIV-1 subtypes is often related to a differential immune response focused to V3 which is expected to originate due to the subtype specific conformational differences in the V3 region [55, 56] . For instance, the HIV-1 V3 crown residues GPGQ in non-clade B and GPGR in clade B viruses respectively are the major determinants of Ab binding and neutralization [19, 40] . Epitope mapping of the anti-V3 Abs with overlapping V3 peptides reveal that their core epitopes lie in the crown region only. Indeed, the recent immunological and structural studies of anti-V3 mAbs have observed similar pattern of binding, wherein essentially all the anti-V3 mAbs bind to~14 residues in the V3 crown [57] [58] [59] . Two of our anti-V3 mAbs (277, 903) showed binding to subtype-A or C but not to subtype-B derived proteins and peptides while mAb 904 displayed cross reactive binding with subtype-B as well. The binding pattern (in context of clade specific or cross reactive V3 antibodies) of the two anti-V3 mAbs 903 and 904 was similar to binding of polyclonal anti-V3 plasma antibodies from the respective patients, however it was different for mAb 277 in the context of the plasma Abs of this patient ( Figure 1A-C) . The finding highlights the importance of pre-screening of plasma for binding to peptides from different viral subtypes prior to isolation of mAbs. One plausible reason for the non-binding of mAb 277 to the V3B peptide in contrast to its corresponding polyclonal plasma may be attributed to the higher number of subtype-C specific B-cell clones in the B cell repertoire of this patient, as indicated by its very high binding to V3C ( Figure 1C ). It may also partly be ascribed to biased selection with a CTB construct containing only V3C sequence, which might allow it to preferentially pick up the B cell clone with clade-C V3 specificity. Interestingly, the anti-V3 mAbs 277 and 903, which show clade-A or C (both having a common GPGQ crown motif ) specific binding use the same variable heavy chain gene (VH3-30) whereas the cross reactive mAb 904 uses a different VH gene ( Table 1) . The finding suggests a possible association of antibody gene usage with epitope specificity, however the number of the mAbs generated in this study is too small for comparison. Remarkably, a recent analysis of anti-HIV Abs has pointed out a preferential usage of VH5-51 gene usage of anti-V3 Abs [60] , and such a preference was later shown to define a conserved antigenic structure within the V3 [61] . The Ab accessibility on the HIV-1 native virus is often challenged by the glycosylation pattern and epitope masking [39, 62, 63] . This effect has been particularly recognized for the V3 region wherein the neighboring regions including V1/V2 shield the epitopes recognized by anti-V3 Abs [46] . Although studies have suggested that the HIV-1 V3 loop remains accessible on most of the viruses [36] , however this information is limited to subtype-B viruses and remains to be explored for other subtypes. Consistent with the binding to gp120 proteins (Du156.12 and JRFL), the three anti-V3 mAbs were able to bind intact native virions with a similar binding pattern (Figure 2A) . The results suggest that V3 epitopes are well exposed over the intact trimeric viruses (Du156.12 and JRFL), and these findings are highly supported by previous work in the literature [36, 43] . The rationale of using same proteins (gp120) and its corresponding viruses (intact virion) for the binding assays was to minimize the effect of both, the V3 loop sequence and the neighboring regions, on the local and global orientation of V3 and on the subsequent binding of Abs. The anti-V3 mAbs showed potent neutralizing activity against subtype-A and C tier 1 viruses, however this activity was restricted for tier 2, especially the subtype-B viruses. The finding was intriguing given the ability of the anti-V3 mAbs to bind to two representative intact virions of subtype-B (JRFL) and subtype-C (Du156. 12) , and yet failing to reach IC50 neutralization titers. However, it should be noted that mAbs 903 and 904 which display a better affinity than 277, showed neutralization of up to 19% at 30 μg/ml, though not reaching IC50 neutralization titers, with these viruses (data not shown). Overall, the data suggest that higher concentrations of these mAbs may be effective in neutralization, however that needs to be confirmed in detail. Together, these results suggest that effective concentrations for binding and neutralization may vary substantially, and the high affinity binding by mAb might be critical for neutralization. The data are highly supported by various studies conducted previously [15, 36] . We isolated here three anti-V3 mAbs from HIV-1 infected donors from India which can effectively neutralize tier 1 viruses but are less effective with tier 2 viruses, however this needs to be confirmed by testing them with a broader panel of viruses from different HIV-1 subtypes. This study demonstrates the importance of pre-screening of plasma Abs for cross reactive binding, for production of mAbs and the idea can likewise be employed for other antigenic regions. Also the study highlights the significance of the antibody affinity, which may probably be equally important as its epitope accessibility, for effective viral neutralization. Furthermore, the analysis of the mAbs generated in this study will allow us to identify epitopes that are unique to clade C viruses and also those that are shared with other subtypes in the context of V3 loop, and the data may provide useful information for HIV-1 vaccine design. The study was approved by the ethics committee of All India Institute of Medical Sciences (AIIMS) New Delhi, and the written informed consent for research and publication of the data was obtained from all the participants. Thirty-three HIV-1 seropositive drug naive patients (18 males and 15 females) within the age range of 20-57 years (median = 33 years) were recruited in this study from the Regional STD Teaching Training & Research centre, Safdarjang Hospital, New Delhi, India during the period 2008-2011. The patients had a median CD4 count of 449 (range = 203-966) cells/cubic millimeter (Additional file 1: Table S1 ). The whole blood samples of HIV-1 positive donors were collected in EDTA vacutainers, plasma was separated by centrifugation at 300 g and stored in aliquots at -80°C until use. The plasma samples were heat inactivated at 56°C for 1 h before using in the assays. The patients have been previously shown to have high titers of anti-V3 Abs in their plasma and a good proportion of these V3 directed Abs displayed cross reactivity [28] , (Andrabi et al., submitted) . Presumably, the patients were infected with subtype-C viruses, which is a major subtype in India [64, 65] . Indeed, the envelope sequences (partial C2-C5 of envelope gp120) of a few patients revealed that majority of the patients were infected with subtype-C viruses (Andrabi et al., submitted) . The mAbs were generated using cellular techniques as previously described [66] . Briefly, peripheral blood mononuclear cells (PBMCs) were EBV transformed in 96-well plates and cultured with a polyclonal B cell activator, CpG [67] , which enhanced EBV infection and B cell transformation. The wells containing Ab-producing cells were identified by testing the culture supernatants for binding activity to V3C-CTB [47] . Cells from wells that test positive were expanded and fused with the heteromyeloma cell line SHM-D33 (ATCC; catalog no. 1668). The fused cells that continued to make functional Abs were repeatedly cloned until monoclonality was achieved. The Abs were purified from culture supernatants using Protein G affinity columns (GE Healthcare) and concentration of the mAbs was determined by noncommercial ELISA. The anti-V3 mAb 447-52D generated previously from a clade B infected individual and a mAb 1418, specific to parvovirus B19 [68] , were used as control Abs in this study. Nucleotide sequence of Ig variable genes of human anti-V3 mAbs was determined as previously described [60, 61] . The messenger RNA was extracted from hybridoma cell lines producing anti-V3 mAbs and reverse transcribed into cDNA using oligo dT primer. Amplification of the variable fragment was performed by PCR using different gene family-specific sets of primers and cDNA as template. were located at 5 0 end of V genes. Reverse primer [5 0 -CTTGGTGGARGCTGARGAGACGGTGACC-3 0 ] was located at the 3 0 end of JH segment and up to 12 nucleotides at the 5 0 of the constant region of IgG [69] . PCR amplification was performed using cycling program of 2 min at 94°C, 35 cycles of 60s at 94°C, 60s at 56°C, and 90s at 72°C, followed 8 min at 72°C. Ethidium bromidestained 0.8% agarose gels were used to visualize the PCR products. The bands of the appropriate size were excised and cleaned with GeneElute Minus EtBr Spin Column (Sigma, USA). PCR products were sequenced (Macrogen, South Korea) in both directions using the primers applied for amplification. The sequence data were analyzed using the International ImMunoGeneTics (IMGT) information system (http://imgt.cines.fr). Five recombinant gp120s representing sequences of primary HIV-1 isolates from clade A, B and C (produced in 293 cells) and a p24 protein were purchased from Immune Technology Corp. (New York, NY). A set of 12 linear overlapping peptides (each 15mer with an 11 amino acid overlap or a 4 amino acid walk) corresponding to the sequence of consensus subtype-C V3 gp120 and CEF Control Peptide Pool (PP) (Cat. No. 9808) were obtained from the NIH AIDS Research and Reference Reagent Program (NIH, ARRRP). Two full length (35mer) peptides corresponding to the consensus-B (CTRPNNNTRKSIHIGPGRAFYTTGEIIG DIRQAHC) (V3B) and con-C (CTRPNNNTRKSIRI GPGQTFYATGDIIGDIRQAHC) (V3C) of V3 gp120, a 24mer con-C MPER (DLLALDSWKNLWNWFDITNW LWYIK) and a 19mer con-C IDR (LGIWGCSGKLICT TAVPWN) peptides of gp41 were selected from Los Alamos HIV-1 sequence database (http://hiv.lanl.gov), and were synthesized from Sigma Genosys, USA. The peptides were HPLC purified to >95% purity (based on information provided by company). The V3-cholera toxin B (CTB) fusion protein (V3C-CTB) and wild type CTB (WT-CTB) used for screening of antibody cultures were kindly provided by Prof. Susan Zolla Pazner from New York University School of Medicine. The binding activity of anti-V3 mAbs against gp120 proteins and peptides (including the V3 overlapping peptides) were tested by enzyme-linked immunosorbent assay (ELISA) as described [18] . Briefly, ELISA plates were coated overnight with protein or peptide at 1.0 μg/ml, blocked with 2% bovine serum albumin (BSA) in PBS, and then incubated with mAbs at a concentration ranging from 10 to 0.00003 μg/ml (12 dilutions). The bound mAbs were detected by incubation with alkaline phosphatase-conjugated goat anti-human IgG (γ specific) (SouthernBiotech, Birmingham, AL) followed by adding substrate to develop color and the plates were read at 405 nm. The relative affinities of mAbs were determined by measuring the concentration of mAbs required for 50% maximal binding (Max50), defined when the binding curve reached the saturation level as described [70] . The binding of mAbs to intact virions was determined with a capture assay as previously described [71] . Briefly, a 96-well plate was coated overnight at 4°C with goat anti-human immunoglobulin G (IgG) Fc Abs (Sigma) at 4.0 μg/ml, and then anti-V3 human mAbs at a saturating level of 10 μg/ml were added for 1.5 h incubation at 37°C. The plate was blocked with 0.5% BSA in phosphatebuffered saline (PBS) containing 10% goat serum and 10 μg of human IgG/ml. The culture supernatant, containing pseudotyped viruses (Du156.12 and JRFL) at a p24 concentration of 25 ng/ml, and the two fold diluted primary isolate (SF162) with a starting p24/ml concentration of 50 ng were added. The plate was incubated overnight at room temperature. Viruses captured by immobilized mAbs were lysed with 1% Triton-X in PBS. Between each step of the assay, the plate was washed with PBS containing 0.05% Tween 20, pH 7.4. The p24 in the virus lysate was quantified by using a commercial ELISA. The anti-V3 Ab 447-52D and SF162, a subtype-B virus known to be sensitive to most of anti-V3 Abs were used as assay control while the human mAb, 1418, was used as a negative control. Eleven HIV-1 subtype A, B and C viruses including five primary isolates (MW965, DJ263, SF162, JR-CSF and 92RW009) and six pseudotyped viruses (HIV-001428, ZM109F.PB4, ZM233M.PB6, Du156.12, JRFL and RHPA4259.7) were used for this study. All the primary viruses and envelope clones were obtained from the NIH, ARRRP. The HIV-1 isolates were expanded by only one cycle of growth on phytohemagglutinin (PHA) and interleukin-2 (IL-2)-stimulated PBMCs, as described previously [45] to avoid alterations in env sequences due to multiple rounds of expansion. Pseudotyped viruses were produced by co-transfection of the rev/env expression plasmid and an env-deficient HIV-1 backbone vector (pSG3ΔEnv) into exponentially dividing 293 T cells (ATCC; catalog no. 11268), in 6-well tissue culture plates (Corning Inc) using calcium phosphate (Promega Inc) method. Pseudovirus-containing culture supernatants were harvested 48 hours post transfection filtered (0.45 μm pore size) and stored at −80°C in 1 ml aliquots. The 50% tissue culture infectious dose (TCID50) was determined in TZM-bl cells. Neutralization of viruses by anti-V3 mAbs was measured as a reduction in luciferase gene expression after a single round of infection of JC53bl-13 cells, also known as TZM-bl cells (NIH, ARRRP; catalog no. 8129), with viruses [72, 73] . Briefly, 200 TCID50 of pseudovirus was pre-incubated for 1 hr at 37°C, 5% CO 2 in 96-well flatbottom culture plates, with serial dilutions of mAbs, starting from 30 μg/ml. Freshly trypsinized TZM-bl cells (10,000 cells in 100 μl of growth medium containing DEAE Dextran and protease inhibitor indianavir (in case of primary isolates only), were added to each well of the mAb/virus mixtures in duplicates. One set of control wells received cells plus pseudovirus (virus control) and another set received cells only (background control). After 48 hours of incubation at 37°C, 5% CO 2 , luciferase activity was measured by using the Bright-Glo Luciferase Assay System (Promega Inc.). The 50% inhibitory concentration of mAb (IC50) was determined at which relative luminescence units (RLU) were reduced 50% compared to virus control wells.

Toxicology, biodistribution and shedding profile of a recombinant measles vaccine vector expressing HIV-1 antigens, in cynomolgus macaques

As a new human immunodeficiency virus type 1 (HIV-1) vaccine approach, the live-attenuated measles virus (MV) Schwarz vaccine strain was genetically engineered to express the F4 antigen (MV1-F4). F4 is a fusion protein comprising HIV-1 antigens p17 and p24, reverse transcriptase and Nef. This study assessed the toxicity, biodistribution and shedding profiles of MV1-F4. Cynomolgus macaques were intramuscularly immunized one or three times with the highest dose of MV1-F4 intended for clinical use, the reference (Schwarz) measles vaccine or saline, and monitored clinically for 11 or 85 days. Toxicological parameters included local and systemic clinical signs, organ weights, haematology, clinical and gross pathology and histopathology. Both vaccines were well tolerated, with no morbidity, clinical signs or gross pathological findings observed. Mean spleen weights were increased after three doses of either vaccine, which corresponded with increased numbers and/or sizes of germinal centers. This was likely a result of the immune response to the vaccines. Either vaccine virus replicated preferentially in secondary lymphoid organs and to a lesser extent in epithelium-rich tissues (e.g., intestine, urinary bladder and trachea) and the liver. At the expected peak of viremia, viral RNA was detected in some biological fluid samples from few animals immunized with either vaccine, but none of these samples contained infectious virus. In conclusion, no shedding of infectious viral particles was identified in cynomolgus monkeys after injection of MV1-F4 or Schwarz measles vaccines. Furthermore, no toxic effect in relation to the MV vaccination was found with these vaccines in this study.

The development of a safe and effective prophylactic vaccine against human immunodeficiency virus type I (HIV-1) is a global health priority. During the past two decades, significant efforts have been made to develop such a vaccine. Among the few candidate HIV-1 vaccines tested in large Phase IIb or III clinical trials, only the RV144 trial, evaluating a recombinant canarypox vector prime and HIV-1 gp120 protein boost, showed a modest efficacy (31 %) against HIV-1 acquisition (reviewed by McElrath and Haynes 2010; McMichael et al. 2010) . Although ultimately, a preventive vaccine against HIV-1 inducing sterile immunity would be optimal, vaccines that would reduce viral load and disease progression by induction of strong and polyfunctional T cell responses should also prove beneficial (McMichael et al. 2010) . Live attenuated measles virus (MV) vaccine strains, such as the widely used and strongly immunogenic Schwarz strain (Griffin 2007) , are replicating RNA viruses (Paramyxoviridae family), capable of inducing long-lived antibody and memory T cell responses (Ovsyannikova et al. 2003; Vandermeulen et al. 2007 ). Besides being highly efficacious, these vaccines are also recognized as safe (WHO 2009) , as MV replicates in the cytoplasm and does not integrate into the host cell genome. Moreover, reversion of the vaccine genome into a pathogenic form has never been observed. The experience accumulated with these vaccines in the past 50 years, and their capacity to induce both CD4 + and CD8 + T cells, render recombinant MV vectors an attractive platform for vaccines aimed to induce T cell responses specific for the HIV-1 transgene. For several of these vaccines expressing HIV-1 antigens, the immunogenicity has been preclinically demonstrated (Combredet et al. 2003; Guerbois et al. 2009; Liniger et al. 2009; Lorin et al. 2004) . We constructed the HIV-1 vaccine candidate MV1-F4, using an in vivo replication-competent MV vector (Combredet et al. 2003 ) derived from the Schwarz vaccine strain, to generate recombinant MV expressing the F4 antigen. F4 is a fusion protein comprising the clade B viral antigens p17, p24, reverse transcriptase and the regulatory protein Nef. Combined with AS01 (a liposome-based Adjuvant System containing 3-O-desacyl-4′-monophosphoryl lipid A (MPL) and QS21; Garçon et al. 2007 ), this F4 antigen was shown to induce potent polyfunctional CD4 + T cell responses in HIVseronegative volunteers (Van Braeckel et al. 2011 ). We conducted a study of the biodistribution, shedding and single-and repeated-dose toxicity of one or three intramuscular (IM) immunizations with MV1-F4 in cynomolgus macaques. The potential intrinsic toxicity of MV1-F4 was studied, as well as the potential immune-mediated toxicity resulting from the host response to the vaccine. The resulting toxicity and biodistribution profiles, including any target organs identified, could be used to guide clinical safety monitoring, while the shedding profile is crucial for determining the potential of infectious viral dissemination by future vaccine recipients, and thus of person-to-person transmission of the virus. The biodistribution, shedding and toxicity profiles of the MV1-F4 vaccine were compared to those of either the reference vaccine (the live attenuated monovalent Schwarz MV vaccine Rouvax), or saline. Humans are the natural hosts of MV. Measles pathogenesis has traditionally been studied in non-human primates as no suitable alternative models exist. Live attenuated MV vaccines are generally non-infectious in rodents, except for cotton rats (in which MV replication is restricted to the lungs) and transgenic mice (which reproduce only limited aspects of MV pathogenesis) (de Swart 2008) . In the current study, we used (MV-seronegative) cynomolgus monkeys. This species is highly sensitive to MV infection and able to develop pathologic lesions and clinical symptoms comparable to those in human MV infections (Kobune et al. 1996) . Moreover, their size permits the administration of a full human vaccine dose. As none of the known Schwarz strain virus traits, including viral envelope proteins, were altered for the construction of the MV1-F4 vector, we hypothesized that the toxicological profile, tropism and shedding capacity of MV1-F4 is similar to those of the parental strain (as represented by the reference vaccine). In addition, we hypothesized that both vaccine viruses have the potential to spread into a large variety of organs, as attenuated MV strains can use the CD46 receptor (which is ubiquitously expressed in both humans and cynomolgus macaques; Sakurai et al. 2008) , in addition to the receptors used by wild-type MV (Dorig et al. 1993) . Wildtype MV strains use primarily the signaling lymphocyte activation molecule (SLAM/CD150) expressed on certain immune cells, and likely other low-affinity receptors to enter CD150-negative epithelial cells (Tatsuo et al. 2000b; Watanabe et al. 2010) . As a result, these strains replicate predominantly in lymphoid organs and epithelial tissues (Griffin 2007; Takeda 2008) . Consequently, MV1-F4 and Rouvax were expected to spread not only to the latter organs and tissues, but also to other parts of the body. The MV vector pTM-MV-Schw that was used to construct the MV1-F4 recombinant virus has been described previously (Combredet et al. 2003) . This vector contains an infectious MV cDNA corresponding to the anti-genome of the Schwarz MV vaccine strain. An additional transcription unit (ATU) was inserted into pTM-MV-Schw in order to sub-clone the F4 fusion protein sequence. F4 has been described previously (Van Braeckel et al. 2011 ) and comprises HIV-1 subtype B antigens p24 (BH10), RT (HXB2), Nef (Bru-Lai) and p17 (BH10). The ATU contains a cloning-site cassette inserted into a copy of the original N-P intergenic region of the MV genome. This region contains the cis-acting sequences required for transcription of the viral P gene, enabling the recombinant MV to express the F4 protein like a measles gene. The ATU was introduced into the plasmid backbone by site-directed mutagenesis between the MV P and M genes, resulting in the plasmid pTM-MVSchw-ATU2_F4co_mut. MV1-F4 virus was rescued from the pTM-MVSchw-ATU2_F4co_mut plasmid using a helper cell-based system developed at the Institut Pasteur. Briefly, helper HEK293 cells expressing both the T7-RNA polymerase and the Schwarz MV N and P proteins were co-transfected with the pTM-MVSchw-ATU2_F4co_mut cDNA and a plasmid expressing the Schwarz MV polymerase L. Subsequently, transfected HEK293-T7-MV helper cells were gently harvested and co-cultured with MRC-5 cells for the amplification of the MV1-F4 virus. Virus titers were determined by endpoint titration on Vero cells and were expressed as 50 % cell culture infectious dose (CCID 50 )/ml. The selected route of administration (IM) and dose were identical to those intended to be used in the first MV1-F4 clinical trial, with the dose-level representing the highest intended human dose. One-dose and three-dose schedules were used, with the three-dose schedule representing the number of doses to be used in the clinical study, plus one. One milliliter of MV1-F4 vaccine (viral titer 1.6 × 10 4 CCID 50 /ml) was injected intramuscularly. The reference vaccine, the commercial (Schwarz) MV vaccine Rouvax (Sanofi-Aventis, Paris, France) was used according to the manufacturer's procedure for IM administration, with each vaccine dose containing 3.9×10 3 CCID 50 of the Schwarz measles vaccine. This virus titer was determined with the same assay as was used for the MV1-F4 vaccine (endpoint titration on Vero cells). All injections were administered in the thigh muscle. The study included purpose-bred cynomolgus monkeys (Macaca fascicularis), aged either 2-3 years (males and females) or 7-12 years (sexually mature males), obtained from Noveprim Ltd (Port Louis, Mauritius) that were seronegative for anti-MV antibodies. During the full experimentation period, the animals were housed in a dedicated primate unit under controlled environmental conditions. Due to the biosafety level of the MV1-F4 candidate vaccine (class 2, group II), they were housed individually in stainless steel cages (level A2, L2 confinement) and appropriate precautions were established. The study was conducted in compliance with the European regulations regarding the protection of animals used for experimental and other scientific purposes, and an ethical committee reviewed the study plan before the initiation of the study. The study design followed the guidelines published by the European Medicines Agency (EMA) as well as other relevant guidelines (EMEA 1995 (EMEA , 2000 WHO 2005) . All experiments were performed under good laboratory practices (GLP) conditions. Animals were inoculated and monitored at the laboratories of the Centre International de Toxicologie (CIT; Evreux, France). Of the 42 monkeys included in the study, 18 young males and 18 young females were allocated to treatment groups 1-6 (Table 1) . Animals were allocated to groups (by sex) using a computerized randomization procedure (CITOX software, developed in-house at CIT, Evreux, France). In addition, six sexually mature males (three per group) were allocated to treatment groups 7 and 8 using the same randomization procedure, and were only used for collection of semen and peripheral blood mononuclear cells (PBMC), as well as for selected clinical observations (body weight, body temperature and food consumption). Treatments included 3.9 × 10 3 CCID 50 of Rouvax, 1.6 × 10 4 CCID 50 of MV1-F4 vaccine, or saline, and were administered according to either a one-dose schedule (injection at day 1) or a three-dose schedule (injections at days 1, 29 and 57). Monkeys were monitored clinically until sacrifice (groups 1-6) or until return to laboratory stock at day 85 (groups 7 and 8). Upon completion of the observation period, animals of groups 1-6 were sedated with ketamine hydrochloride (Imalgène, Mérial, Lyon, France), then anesthetized with thiopental and sacrificed by exsanguination. Necropsy was performed at two time-points, either within the expected peak of viremia (10 days after the first vaccine dose (day 11) for groups 1-3) or when clearance of infectious MV was expected to have occurred (represented by 28 days after the third dose [day 85] for groups 4-6) (Auwaerter et al. 1999; Pan et al. 2005; Permar et al. 2003) . Single-and repeated-dose toxicity was assessed at various time points (Table 2 ). Shedding and post-mortem analyses were conducted at Texcell (Evry, France). All other procedures were conducted at CIT. Animals were monitored at least twice daily for mortality, morbidity and clinical signs of toxicity. Toxicological parameters included dermal reactions at the injection sites, body weight, food consumption, rectal body temperature, electrocardiography and ophthalmology. Complete clinical examinations were performed pre-treatment and thereafter once weekly. Sedation, used for electrocardiography, ophthalmology and occasionally for weighing, was done by IM administered ketamine hydrochloride (Imalgène). At 3 and 24 h after each immunization, dermal reactions, including edema and erythema formation, were evaluated using the Draize scale, and any other lesions were noted. Reactions persisting for 48 h after immunization were evaluated daily until disappearance. Electrocardiographic examinations were performed using a Cardioline Delta 3 Plus and Cardiovit AT-6 (Schiller AG, Baar, Switzerland) with standard leads I-III, starting with determining the heart rate, PQ and QT intervals and the QRS-complex duration on lead II. Ophthalmology included assessment of pupillary light reflexes using tropicamide (Mydriaticum, Théa, Clermont-Ferrand, France), examination of appendages, optic media and fundus by indirect ophthalmoscopy (Oméga 180, Heine, Germany) and of anterior segments and lenses (portable slitlamp biomicroscope, model SL-15; Kowa, Japan). Peripheral blood samples were collected without sedation at different time points after the first and the third dose (Table 2), in tubes containing EDTA, sodium citrate or lithium heparin (for haematology, coagulation parameters or blood biochemistry, respectively). Haematology (i.e., erythrocyte count, haemoglobin (HB), mean and packed cell volumes, mean cell HB concentration, mean cell HB, thrombocytes, leucocytes (differential) and reticulocytes) was determined by ADVIA 120 haematology analyser (Siemens, Saint-Denis, France). Leukocyte differential analysis (with cell morphology) was assessed in blood smears stained with May-Grünwald-Giemsa. Coagulation parameters (i.e., prothrombin time, activated partial thromboplastin time and fibrinogen) were measured with an ACL300 coagulation analyzer (Beckman Coulter, Instrumentation Laboratory, France). Complete blood biochemistry was assessed by the ADVIA 1650 Chemistry System (Siemens) using whole blood. Urinalysis (including volume, pH, specific gravity, proteins, glucose, ketones, bilirubin, nitrites, blood (HB) and urobilinogen) was done using a Clinitrek 500 urine chemistry analyzer (Siemens). Blood samples for assessment of humoral responses were taken at prevaccination and days 11, 29, 56, 67 and 85. Anti-MV humoral responses in sera were measured using an anti-MV enzyme-linked immunosorbent assay (ELISA) adapted from the commercial immunoassay Enzygnost Anti-Measles Virus/Ig (Dade-Behring, Illinois, USA). A peroxidase-labelled anti-monkey immunoglobulin (Ig) secondary antibody (Rockland, Pennsylvania, USA) was used in the assay. Animals were considered to be responders if they were seropositive for anti-MV antibodies, i.e., if the absorbance (measured at 490/620 nm) exceeded the cut-off value, for a serum dilution of 1/500. The cut-off values (percentile 100) were determined per gender for sera collected prior to vaccination (serum dilution 1/500), and were found to be equal to 1.74 and 1.22 for sera from female and male monkeys, respectively. At necropsy, organs were weighed and subjected to macroscopic and microscopic examination. The selection of organs to be examined for gross pathology and histopathology analyses followed the applicable European and international guidelines (EMEA 1995 (EMEA , 2000 WHO 2005) . Of this selection, the following organs were weighed: adrenals, brain, epididymides, heart, kidneys, cervical and iliac lymph nodes, liver, lungs with bronchi, ovaries, pituitary gland, prostate, spleen, testes, thymus, thyroid with parathyroid and uterus. Ratios of organ weight to body weight (at necropsy) were calculated. For gross pathology, the external surface of the body, orifices, the neck with organs and tissues, the cranial cavity with surfaces of the brain and spinal cord and the thoracic, abdominal and pelvic cavities with their contents were examined. For histopathological examination, 4-μm-thick sections of tissues were stained with haematoxylin and eosin. Sample processing for biodistribution and shedding analyses The selection of organs to be examined for biodistribution analyses followed the applicable European and international guidelines (EMEA 1995 (EMEA , 2000 WHO 2005) . For biodistribution analyses, tissue samples (3×3×3 mm 3 ) were dissected out, rinsed with cold phosphate-buffered saline (PBS) and snap-frozen in cryotubes in liquid nitrogen. For shedding analyses, selected excretions and biological fluids (i.e., semen [only for groups 7 and 8], serum, urine, throat and nasal swabs, saliva, vaginal secretion and faeces) and PBMC were collected. Throat swabs, saliva, nasal secretions and vaginal secretions were conserved in M4RT tubes (Remel, Oxoid, Cedex, France) at −70°C. Semen and faeces were directly frozen in cryotubes and conserved at −70°C. Immediately after collection, urine was centrifuged (800×g, 10 min, 4°C) and the pellet was resuspended in PBS and stored at −70°C. PBMC were isolated from EDTA blood samples by gradient centrifugation and frozen in PBS (−70°C). To obtain serum, blood was collected in plain tubes, incubated (20 min, room temperature) and centrifuged (3,000×g, 10 min, 4°C). Samples of organs, tissues, excretions and biological fluids were used for RNA extraction and MV N-specific, quantitative reverse transcriptase (RT) PCR analysis (RT-qPCR). Briefly, cDNA was synthesized from the RNA templates, and then amplified by qPCR specific for the MV N gene. This gene is identical in both vaccine viruses used in this study. Frozen organ samples were homogenized using a Tissue Lyser (30 Hz for ≥2 min; Qiagen, Valencia, CA) and kept on ice. Samples were then lysed using 1 ml of Qiazol lysis reagent (Qiagen). Frozen PBMC, excretions and biological fluids were directly (for urine, serum, throat swabs, nasal swabs and vaginal secretion), or with a 1/45 dilution factor (for semen) submitted to RNA extraction. Faeces was resuspended into PBS to 1/10 (v/v), centrifuged, and the supernatant was filtered (0.22 μm). a Body weights: recorded at pre-treatment, days 1 and 4 and then once weekly until study end. Food consumption: estimated daily from 5 days before start throughout the study b Dermal observations at the injection sites were recorded at 3 and 24 h after each vaccination c Performed for groups 4-6 only d On days 1 and 57, tests were done just before and 1 h post immunization. A test at day 85 was done before sacrifice for only one animal e Blood sampling for clinical pathology performed for groups 1-6 only f At day 55 or day 56, blood samples for haematology, coagulation parameters and blood biochemistry were taken before immunisation. On day 85, samples were taken before necropsy g Day 11, day 85 0 time points of termination of the animals of groups 1-3, groups 4-6 respectively Total RNA was extracted from the samples using an RNeasy 96 kit with the Biorobot 3000 (both Qiagen) and reagents and protocols recommended by the manufacturer. Briefly, the lysed samples in 0.2 ml of chloroform were homogenized (15 s), left at room temperature for 2-3 min and centrifuged (12,000×g, 15 min, 5°C). The aqueous phase was transferred to 96-well plates and placed on the Biorobot for automatic extraction (by vacuum or centrifugation). Ethanol (600 μl, 70 % (v/v)) was added to each well. After mixing, samples were placed in the column for RNA adsorption on the membranes. After three wash steps (once with 800 μl RW1 buffer and twice with 800 μl RPE buffer), columns were centrifuged (5,600×g, 10 min). RNA was then eluted with 60 μl elution buffer (RNase free water) and stored at temperatures below −70°C. RT and qPCR were performed using the AgPath-ID Onestep RT-PCR kit (Applied Biosystems, Inc., Foster City, CA) and the LightCycler 480 and software version 1.5.0.39 (Roche Applied Sciences). The forward primer was MeN_TMF(5′-GCGAGAGCTGCCCATCTTC), the reverse primer was MeN_TMR (5′-ACTCCGTTGCAGTGTCAATGTC) and the forward MeN_TMP probe was (6-FAM)~AACCGGCA CACCC~(NFQ), with the fluorogenic probe located in the region bracketed by forward and reverse primers. Cycling conditions were 45°C for 10 min, 95°C for 10 min, and 45 cycles of 95°C for 15 s and 60°C for 45 s, followed by 40°C for 2 min. The standard curve and RT-qPCR parameter determination was performed using an in vitro transcript RNA template with the sequence of the N gene of the measles vector. MV1-F4 suspension (GSK; EF4MA001A, viral titer 4.2 log CCID 50 /ml, 7.2 log genome equivalent (geq)/ml, diluted to 10 4 copies geq), was used as positive extraction control. Negative (RNase/DNase free H 2 O) and positive extraction controls were run in quadruplicate and duplicate reactions, respectively. MV1-F4 RNA recovery was assessed by N-specific RT-qPCR after spiking each type of tissue with MV1-F4 at various concentrations, and ranged from 90 % to 100 %. The presence of MV1-F4 RNA was considered positive if the measurements exceeded the limit of detection (LOD) of 100 geq/reaction, and if the acceptance criteria (error <0.2, efficiency>90 % and crossing point <45) were fulfilled. The LOD was defined as the lowest amount of copies that can be detected (but not necessarily quantified as an exact value) with 95 % probability. All samples were tested once. In vitro infectivity and immunofluorescence assays All samples in which a RT-qPCR signal was detected (either confirmed [above LOD], or at a very low level [below LOD]) were submitted to an in vitro infectivity assay. The presence of infectious viral particles for all fluids except semen was considered positive for values exceeding the LOD of 10 CCID 50 /well (with 95 % probability). Nevertheless, the assay was designed to detect one infected cell that propagates the virus to its adjacent cells, resulting in a plaque forming unit (as aided by the presence of agarose in the culture medium). Therefore, it was assumed that the system was able to detect 1 CCID 50 /well (but with a probability of detection of less than 95 %). However, for this to occur at least 10 CCID 50 should be present in the inoculum, in order to allow the adsorbtion onto the cells. For semen, validation of the infectivity assay could not be conducted due to high bacterial contamination. Vero cells were cultured in Dulbecco's modified Eagle's medium (DMEM)-5 % fetal calf serum (FCS)-4 mM glutamine. Before seeding the cells, a sterilized (ethanol, UV) coverslip (18×18 mm 2 ) was placed in each well. One day before inoculation, 6-well plates were seeded with 4×10 5 Vero cells/well. Biological fluid samples were diluted (1: 15.625 for serum, 1:10 for all other fluids) in dilution medium (DMEM-1 % antibiotics) and 700 μl of each fluid sample was added to Vero cells for 2 h of inoculation (37°C, 5 % CO 2 ). Inoculum was then removed and 2 ml of survival medium (DMEM-2 % FCS-4 mM of glutamine-1 % antibiotics) supplemented with 1 % (v/v) low-melting agarose was added. The plates were incubated (37°C, 5 % CO 2 ) for 6 days. As a positive control, Vero cells were inoculated with100 CCID 50 of MV1-F4 viral suspension (as described above) and diluted in 700 μl dilution medium. As a negative control, Vero cells were inoculated with 700 μl dilution medium. The presence of infectious viral particles was revealed by immunofluorescence assay (IFA), using an antibody directed against the measles N protein. The acceptance criteria of the assay were the absence or presence of fluorescence, respectively, in the negative or the positive controls. After 6 days of incubation, the culture medium was removed and 2 ml of 4 % paraformaldehyde was added (15 min, 5°C). The slides were washed with PBS and incubated overnight with 2 % goat serum at 5°C. Next, 100 μl of the primary antibody (mouse anti-MV-N monoclonal antibody [# MAB8906; Chemicon, Temecula, CA, USA]), diluted 1:200 in PBS-2 % goat serum-0.1 % saponin, was added to each well (60 min, room temperature). After washing twice with PBS, 100 μl of the secondary antibody (goat anti-mouse F(ab′) 2 -Cy3 [# LU1513615; Jackson ImmunoResearch, West Grove, PA, USA]) diluted to 1:500 in PBS-2 % goat serum-0.1 % saponin, was added to each well. After incubation (45 min, room temperature), coverslips were washed twice with PBS, rinsed with water, mounted on a slide using Vectashield DAPI mounting medium (Vector Laboratories, Peterborough, UK) and inspected under fluorescence microscope. Positivity or negativity was checked by the presence or absence of fluorescence, respectively. The following sequence was used for statistical analyses of body weights, body temperatures and electrocardiography, haematology, blood biochemistry and urinalysis data. A test of normality using the Kolmogorov-Lilliefors test was performed, followed by log-transformation of the data if normality assumptions were not satisfied. Assessment of homogeneity of variance was performed using the Bartlett test. If the homogeneity of variance was not rejected, the treatment groups were compared to the control group using the Dunnett test, in other cases the Dunn test was applied. The same procedure, but without logtransformation of the data, was performed for the statistical analysis of organ weights using PathData software (version 6.2b5). Statistical significance was expressed as p<0.05 or p<0.01. No mortality or morbidity was observed in any of the groups (Table 3 ). The only clinical effect recorded for a vaccine group was marked dehydration (without diarrhoea) in a Rouvax-treated male on days 56-62, which resolved within the following 2 days. Scab formation was noted in one Rouvax-treated male each of the one-dose and three-dose groups. These effects were incidental, self-limiting and of a mild-to-moderate nature, and therefore considered toxicologically irrelevant. In the controls of group 4, incidental vomiting (day 1; one male) and very slight erythema (days 1-3, one female) were noted. For both sexes, mean body weights of vaccine and control groups were not significantly different and remained virtually Organ weights -Elevated absolute and relative spleen weights on days 11 (in males) and day 85 (in both genders, statistically significant in Rouvax-treated females). No changes -Elevated mean absolute and relative weights of cervical and iliac lymph nodes, mainly in treated males (both vaccines; days 11 and 85). -Lowered absolute and relative thymus weights in treated males (both vaccines; days 11 and 85). Histopathology -Increased size of white pulp for both sexes on day 11 (MV1-F4) and day 85 (both treatments, corresponding with higher spleen weights). Increased size of white pulp in females (day 11). -Increased sizes/numbers of germinal centers in mandibular lymph nodes in MV1-F4 treated males (day 11). Increased sizes/numbers of germinal centers in mandibular lymph nodes in females (day 85). M/F male/female, IS Injection site, aPTT activated partial thromboplastin time, CK creatine kinase constant, with minor increases between days 0 and 85. Apart from scheduled fasting days prior to procedures, the individual food consumption was generally 75-100 % of the daily ration across groups. On days when animals were sedated, the consumption was typically 25-50 % of the daily ration, and occasionally 0 % (on days 1 and 57 for nearly all females of groups 4-6, and on day 84 for two Rouvax-treated males). These fasting periods had no impact on body weights and were similar across control and vaccine groups, and thus considered to be unrelated to immunization. Mean rectal temperatures of groups of vaccinated males were not significantly different from those of the control groups at all time points. For females, the sporadically noted statistical differences between groups remained within the ranges commonly recorded in animals of this age housed in similar experimental conditions. Consequently, no treatment-related effects on body temperature were observed in the study. There were no treatment-related abnormalities in the qualitative and quantitative electrocardiography parameters. Although low-amplitude R and P waves and high-or lowamplitude, biphasic and negative T waves were noted, these findings were isolated, observed both prior to and post treatment, and occurred in both controls and immunized groups. They were also commonly observed in non-treated cynomolgus monkeys kept under similar laboratory conditions. No statistical differences in quantitative parameters (PQ, QRS and QT intervals and heart rates) were recorded between any of the groups. Sinusal arrhythmia was observed prior to treatment in one female in the control group. No relevant ophthalmological findings were observed in any treatment group. In two control females, a hypopigmented fundus was noted at all three observation time points. No treatment-related effects were noted for haematological, blood clinical chemistry and urinalysis parameters in any of the vaccine groups. At all time points, values for control and immunized animals of both genders were similar. On day 56, the mean activated partial thromboplastin time (aPTT) for MV1-F4treated males was slightly (15 %) shorter than for control males (i.e., 17.4 vs. 20.4 s; p<0.01), which was not observed for females of this group or for other treatment groups. Prothrombin times (PT) were unaffected by immunization. For blood biochemistry parameters, the only relevant finding was a higher mean creatine kinase (CK) activity in Rouvax-treated males compared to control males (i.e., 3,760 vs. 1,010 IU/l; p<0.05) on day 85, due to elevated values recorded for two animals. To evaluate the humoral responses induced by both vaccines, anti-MV (Ig) antibody responses were measured in sera. Ten days after a single dose (day 11), responses were undetectable in the majority (87 %) of immunized animals (Fig. 1) . However, a 100 % seroconversion rate was observed in the animals of the three-dose Rouvax or MV1-F4 groups (groups 5-8) from 28 days after the first dose (day 29) onwards. This suggests that the anti-MV humoral responses may not have been fully developed at day 11, as confirmed by the kinetics of these responses in the vaccinated animals. To assess the potential release of infectious MV1-F4 viral particles in PBMC, excretions and biological fluids (including serum, urine, throat-and nasal swabs, saliva, vaginal secretions, semen and faeces), samples were first tested for the presence of MV viral sequences by N-specific RT-qPCR assay. All samples in which viral RNA was detected, even if below the defined LOD, were then tested further using an in vitro infectivity assay. Of the eight time points at which shedding analysis was performed, MV viral RNA was only detected at day 11 in some of the vaccine groups. At this time point, all signals in animals of all vaccine groups were below the LOD (Table 4) , with the exception of one positive result in the three-dose Rouvax group. In the MV1-F4 groups, viral RNA was detected (below the LOD) in faeces from two males of the one-dose group and in urine and vaginal secretions from one female of the three-dose group. In the Rouvax groups, viral RNA was detected in throat secretions (above the LOD, three-dose group), in faeces, throat swabs, vaginal secretion and serum (below the LOD, three-dose group), in nasal secretions (below LOD, both groups) and in PBMC (below the LOD, three-dose group and group of mature males). No positive results were detected in any of the samples of the control groups or in saliva samples of any treatment group. In the in vitro infectivity assay and at the LOD of 10 CCID 50 /well, no infectious viral particles were recovered from any sample in which viral RNA had been detected, with acceptance criteria (i.e., negative and positive controls showing respectively negative and positive results) being fulfilled. The potential presence of viral RNA in organs and tissues from vaccinated monkeys was detected by RT-qPCR specific for the N gene. This gene is identical in both vaccine viruses used in this study. At day 11, MV viral RNA was detected in a large number of animals and tissues from both vaccine groups (Table 5) , confirming the expected time point of the peak of viral replication. In animals of both one-dose vaccine groups, viral RNA was mainly detected in secondary lymphoid organs (lymph nodes, spleen, Peyer's patches) and in the non-lymphoid tissues of the intestine (with the majority of signals detected above the LOD) and to a lesser extent in the liver, trachea, larynx and urinary bladder. Viral RNA was also detected in few animals in lungs, eyes, adrenals (for animals in both vaccine groups), in ureters and salivary glands (for the MV1-F4 group) and in several other organs for the Rouvax group. Importantly, no viral RNA was detected in the brain, cerebellum, cerebrum, spinal cord respectively. a Young animals (N06/group) were immunized at day 1 and sacrificed at day 11 (groups 1, 2 and 3). Blood samples were taken at days 0 and 11. b Young animals (N06/group) were immunized at days 1, 29 and 57, and sacrificed at day 85 (groups 4, 5 and 6). Blood samples were taken at days 0, 11, 29, 56, 67 and 85. c Sexually mature animals (N03/group) were immunized at days 1, 29 and 57 (groups 7 and 8). Blood samples were taken at days 0, 11, 29, 56, 67 and 85 regions, thalamus and hypothalamus of any animal. No gender-determined patterns were detected. At day 85 (28 days after the third dose), viral RNA was still detectable in the spleen and lymph nodes of some animals (both vaccine groups), and also in Peyer's patches and the intestine (Rouvax group). Low signals were also detected in the larynx and the trachea (Rouvax group). No viral RNA was detected in the controls at both time points of necropsy. Gross pathology, organ weights and histopathology Necropsy after either one or three injections revealed no macroscopic findings attributable to either treatment. Recorded findings were sporadic or common for cynomolgus monkeys and therefore considered toxicologically irrelevant. At day 11, the mean absolute and relative spleen weights of males of both vaccine groups tended to be higher than for control males (without reaching statistical significance), Pia maters (cervical/lumbar/thoracic) -1/6<LOD --Spleen 4/6+; 1/6<LOD 5/6+ 1/6+; 2/6<LOD 3/6+; 1/6<LOD Sternum with bone marrow -2/6<LOD -- Urinary bladder 1/6+; 2/6<LOD 1/6<LOD -- No detections were made in the remaining organs (aorta, brain, cerebellum, cerebrum, epididymides, femoral bone with articulation, heart, hypothalamus, injection sites, knee joint, optic nerve, oesophagus, ovaries, oviducts, pituitary gland, prostate, sciatic nerve, seminal vesicles, skeletal muscle, spinal cord regions, stomach (fundus), testes, thalamus, thyroids and parathyroids) for these groups and time points N no. of animals with positive signal in tissue sample/no. of tested animals in that group; +/−: + 0 positive signal (signal > defined limit of detection [LOD] of the RT-qPCR assay [with 95 % probability]). -: no RNA signal detected; <LOD: RNA signal detected but the number of copies was less than the defined LOD of 100 geq/reaction (with 95 % chance to detect a positive signal) which was only observed to a minor extent for the females of these groups (Table 6 ). Similar effects were observed for the cervical lymph nodes. For the iliac lymph nodes, which were the draining lymph nodes for the injection sites (the thighs), there was a slight tendency for higher weights in all Rouvax-treated animals. Mean thymus weights tended to be lower in both treated male groups than in control groups, possibly due to high absolute and relative weights of one control male (129 % and 118 % of group mean values, respectively). No clear effects of immunization were observed on the thymus weights of females of any group. At day 85, mean absolute and relative spleen weights of both vaccine groups were higher than those of controls for both sexes, which reached statistical significance (p<0.05) in the Rouvax-treated females. Mean absolute and relative weights of cervical and iliac lymph nodes tended to be higher in both groups of vaccinated males as compared to the control groups, while a high variability in values were noted for the female groups. Mean thymus weights tended to be lower in both groups of vaccinated males and in Rouvax-treated females, as compared to control groups. None of these differences between groups or sexes, observed on either day of necropsy, were supported by macroscopic observations. For the majority of immunized animals that were sacrificed at days 11 or 85, the increase in the absolute and/or relative spleen weights could not be directly correlated to the detection of viral RNA in these spleens in the biodistribution analyses. An increased spleen weight was noted in some animals for which no viral RNA was detected in the spleen, and conversely, no increased spleen weight was observed in some animals for which viral RNA was detected in the spleen. Microscopically, enlargements of the white pulp in the spleen were occasionally noted in the animals sacrificed at day 11, in one of the three males and one of the three females treated with MV1-F4, and in one of the three control females. This enlargement was primarily due to generation of germinal centers, and was also noted in the animals sacrificed on day 85 (in one of the three MV1-F4treated males and two of the three Rouvax-treated males). At both time points, the increased sizes of the white pulp and/or germinal centers were not necessarily linked to the detection of viral RNA in the spleens. In addition, no enlargement was observed in some animals for which viral RNA was detected in the spleen. Furthermore, minimally increased germinal center sizes and/or numbers were occasionally noted in mandibular lymph nodes in two of the three MV1-F4treated males sacrificed on day 11. Although this was also seen in one of three control females sacrificed on day 85, a relationship to treatment could not be excluded. This effect was rarely seen in iliac, cervical or popliteal lymph nodes. The higher absolute and relative weights of cervical and iliac lymph nodes recorded in males of both vaccine groups on day 85 did not correlate with any consistent microscopic changes. No relevant microscopic changes were observed at the injection sites on both days of necropsy. All other microscopic findings occurred at equal frequency across groups and/or were within the normal laboratory ranges for this species, and were therefore considered as toxicologically irrelevant. Moreover, no effects suggestive of MV infection (follicular necrosis within the hair follicles, proliferative and necrotizing bronchointerstitial pneumonia, thymus atrophy or Warthin-Finkeldey cells within the lymph nodes or spleen) were found. In addition, there was no evidence of subacute sclerosing panencephalitis (a rare syndrome in humans occurring after infection with wild-type MV stains, characterized by progressive gliosis, demyelination and neuronal loss). In order to support the clinical development of the MV1-F4 vaccine, we have compared the toxicology, biodistribution and shedding profiles of the HIV-1 candidate vaccine MV1-F4 to those of the parental Schwarz vaccine strain, using a cynomolgus macaque model. In addition, the results may serve to increase our knowledge of the toxicity or in vivo tropism of MV vectors or vaccine strains, which has been the focus of few preclinical studies to date (e.g., de Vries et al. 2010; Lemon et al. 2011; Myers et al. 2007; Peng et al. 2003) . Our data show that no shedding of infectious virus was observed for either of the two vaccines. Furthermore, no toxic effect in relation to the MV vaccination was found with these vaccines. One or three IM injections of a full human dose of MV1-F4 or of the reference vaccine were well tolerated and induced no clinical symptoms of local or systemic reactogenicity. For both vaccines, virus replication was predominantly observed in secondary lymphoid organs (including spleens, Peyer's patches and all major lymph nodes) and to a lesser extent in epithelium-rich tissues (including intestine, larynx, trachea and urinary bladder) and liver, which were thus designated as potential target organs for intrinsic toxicity of the vaccines. However, no gross or histopathological effects indicative of toxicity were observed in these target organs, or in any other organs. Mean spleen weights were increased after three doses of either vaccine, which corresponded in some animals with enlargements of the white pulp, due to generation or growth of germinal centers. This effect likely resulted from immune activation of the secondary lymphoid organs by the vaccines, which has also been observed in other preclinical studies (Sheets et al. 2006; Speijers et al. 1988) . It is less likely that local virus replication in these organs has caused the increased spleen weights, as no clear relationship between increased spleen weights, enlargements of the white pulp and the presence of viral RNA in the spleens could be established. All other observations were considered to be unrelated to treatment, as they occurred incidentally, without correlation between genders, or at similar frequencies in control and immunized animals. The occasional reductions in food consumption had no impact on body weights and were not considered to be indicative of systemic toxicity. For DNA plasmid vaccines, it has been observed that food consumption data and body weights did not correlate with toxicity or other findings (Sheets et al. 2006) . A similar conclusion could potentially be drawn for recombinant live-attenuated vaccines such as MV1-F4. The only clinical pathological finding for MV1-F4 treatment groups was a slight but statistically significant decrease of the aPTT (but not of the PT) for males on day 56. However, shortening of the aPTT has generally little clinical relevance and may be associated with suboptimal sample collection or processing (Adcock et al. 1998; Awad et al. 2006) . Therefore this is not considered indicative of coagulation abnormalities of toxicological relevance. We also observed that on day 85, the mean CK activity for Rouvax-treated males was significantly higher than for control males. Elevated CK values are used as a marker of injury, heart attack, severe muscle breakdown, muscular dystrophy or acute renal failure (Lott and Abbott 1986) . None of these conditions were observed for the two males involved, and urinalysis showed no treatment-related effect for either vaccine. In addition, no elevated CK activity was found in these animals on previous time points. Possibly these elevations reflect the effects of IM-delivered sedation for ophthalmology on day 84. This effect is commonly seen in preclinical studies (Stewart et al. 2008 ) and is therefore not considered to be an adverse reaction to treatment. Cell tropism for vaccine and wild-type MV strains is largely determined by virus entry (Tatsuo et al. 2000a ). Wild-type MV, which use predominantly SLAM/CD150 expressed on activated immune cells (Tatsuo et al. 2000b) , is known to mainly replicate in lymphoid organs and epithelial tissues such as the skin, respiratory tract, intestine and urinary bladder (Griffin 2007; Takeda 2008) . Recent studies suggest that after aerosol infection, wild-type MV initially targets macrophages and dendritic cells in alveolar tissues, which is followed by replication in regional lymph nodes and systemic spreading to lymphoid organs (Lemon et al. 2011) . How MV infects CD150-negative epithelial cells remains largely unclear, although putative epithelial receptors have recently been described (Watanabe et al. 2010) . As MV vaccine strains use the ubiquitous CD46 receptor in addition to the receptors used by wild-type MV (Dorig et al. 1993) , the vaccine strains were expected to exhibit a wider tropism than that documented previously for wild-type MV (Griffin 2007; Lemon et al. 2011; Takeda 2008) . However, in macaques infected intratracheally or by aerosol with wild-type or attenuated (Edmonston tag) recombinant MV, only the pathogenic MV caused significant viremia and widespread distribution. The attenuated MV had a restricted systemic spread and was rarely detected in lymphoid tissues (de Vries et al. 2010) , in contrast to observations in the current study. Our data seem to be more aligned with studies in mice immunized intraperitoneally or intravenously with an attenuated oncolytic (Edmonston) recombinant strain, in which infected cells were mainly detected in secondary lymphoid organs, liver and lungs (Myers et al. 2007; Peng et al. 2003) . These disparities between studies may be explained by differences in administration routes, MV strains or methods of MV detection. The RT-qPCR results for tissues and biological fluids confirm a peak of replication around 10 days after the first dose, supporting other studies in monkeys with wild-type or attenuated MV strains (Auwaerter et al. 1999; Pan et al. 2005; Permar et al. 2003) . This resembles the clinical situation, as vaccine-induced peaks are known to occur a few days before those induced by natural infection taking place between 11 and 14 days after exposure (Strebel et al. 2004) . We observed that viral clearance for both vaccines was not fully completed at 28 days after the third dose, since viral RNA was still detectable in lymphoid and epithelium-rich tissues. This is consistent with an earlier report showing that MV RNA remained detectable in monkeys for 4-5 months (in this case, in PBMCs), even though clearance of viremia occurred within 14 or 29 days (Pan et al. 2005 ). An additional study objective was to assess the potential of virus shedding of the MV1-F4 vaccine (in parallel to that of the reference vaccine). In natural infection, MV can be isolated from urine up to 10 days after rash onset (Gresser and Katz 1960) , and viral RNA can be detected in PBMC and nasopharingeal specimens up to 100 days after rash onset (Riddell et al. 2007 ). The few known studies of shedding by attenuated MV vectors report detection of (Edmonston) viral RNA in buccal swabs from monkeys (after intravenous administration; Myers et al. 2007 ), but no detection in human urine or saliva (after intraperitoneal administration; Galanis et al. 2010) . Consistent with reports of detection of viral RNA from attenuated MV vaccine strains in human urine up to 14 days after vaccination (Rota et al. 1995) , we detected MV1-F4-derived RNA in urine at day 11. Infectious viral particles from MV vaccines have rarely been found in human throat or nasopharyngeal secretions, and there is only one reported case of isolation of (Schwarz) MV vaccine virus from throat swabs (Morfin et al. 2002) . While we detected viral RNA from both vaccines at the peak of viremia in a few samples such as throat and nasopharingal swabs, none of these contained infectious virus at a LOD of 10 CCID 50 /well (with 95 % probability). Therefore, transmission of MV1-F4 virus between persons is considered unlikely to occur, and has also never been documented previously for any of the current measles vaccines (WHO 2009) . The immunogenicity of the MV1-F4 candidate vaccine was not investigated in detail in the current study, as our study was primarely designed to assess the toxicology, biodistribution and shedding profiles of the candidate vaccine. However, immune responses directed against the F4 transgene were characterized in a separate study using the same animal model (MV-seronegative cynomolgus macaques). In the latter study, the MV1-F4 candidate vaccine was shown to induce F4specific CD4 + and CD8 + T cell responses, as well as antibody responses to F4 (manuscript in preparation). In conclusion, while viral dissemination was observed in various organs, no shedding of infectious viral particles was noted, and no toxic effect in relation to the MV vaccination was found following one or three IM injections with a clinically relevant dose of MV1-F4, with the same outcomes for the (Schwarz) measles comparator vaccine. Moreover, either vaccine virus replicated predominantly in secondary lymphoid organs and, to a lesser extent, in epithelium-rich tissues. Thus, as expected, introduction of the F4 transgene did not change the toxicological profile, shedding capacity or tropism of the parental strain.

Forty-Five Years of Marburg Virus Research

In 1967, the first reported filovirus hemorrhagic fever outbreak took place in Germany and the former Yugoslavia. The causative agent that was identified during this outbreak, Marburg virus, is one of the most deadly human pathogens. This article provides a comprehensive overview of our current knowledge about Marburg virus disease ranging from ecology to pathogenesis and molecular biology.

Marburg virus (MARV) first appeared in August 1967, when laboratory workers in Marburg and Frankfurt, Germany and Belgrade, Yugoslavia (now Serbia) were infected with a previously unknown infectious agent. The 31 patients (25 primary, six secondary infections) developed severe disease that progressed to a fatal outcome in seven of the cases. An additional case showing symptoms of disease was diagnosed retrospectively (reviewed in [1] ). The source of infection was traced back to African green monkeys (Chlorocebus aethiops) that had been imported from Uganda and were shipped to all three locations. The primary infections ironically occurred when the monkeys were necropsied for the purpose of obtaining kidney cells to culture poliomyelitis vaccine strains. In the remarkable period of less than three months the etiologic agent was isolated, characterized, and identified by the joint effort OPEN ACCESS of scientists in Marburg and Hamburg [2] and was later confirmed by Kunz and colleagues [3] and Kissling and colleagues [4] . The pathogen was named Marburg virus after the city with the most cases and represented the first isolation of a filovirus. Erroneously, a study published in 'The Lancet' claiming that the mysterious disease was caused by rickettsia or chlamydia has frequently been cited as the first report on the causative agent of Marburg virus disease (MVD) [5] . It was not until 1976 that the now better-known member of the family, Ebola virus (EBOV), first emerged in Africa [6, 7] . Shortly thereafter marburgviruses and ebolaviruses were classified together in a newly established family termed Filoviridae, so-named after their distinctive thread-like structure (filum being Latin for thread). MARV had not been heard of for eight years, when a young Australian who had traveled throughout Zimbabwe was admitted to a hospital in Johannesburg, South Africa with symptoms reminiscent of those observed during the 1967 outbreak in Europe [8] . When he died and the infection spread to his travel companion and later also to a nurse, Lassa fever was initially suspected resulting in strict barrier nursing techniques and isolation of the patients and their primary contacts. This lead to a quick containment of the outbreak, and while the secondary cases recovered, MARV was identified as the causative agent of the disease. In the following years from 1975 through 1985, only sporadic outbreaks that affected small numbers of individuals were caused by MARV on the African continent (Table 1, Figure 1a ). As the case fatality rates associated with MVD were also lower than those seen in the devastating outbreaks associated with EBOV disease that reached up to 90%, MARV was long thought to be less threatening (Table 1) . However, this view had to be revised as MARV reemerged in two large outbreaks occurring in the Democratic Republic of the Congo (DRC) in 1998-2000 [9] and then, for the first time also in Western Africa, in Angola, in 2004-2005 [10] . The total number of 406 cases and the high fatality rates (83% in DRC and 90% in Angola) revealed that MARV was as big of a threat for public health as EBOV [1, 11] . The variation observed in disease severity and case fatality rates between these outbreaks versus the initial one in 1967 may depend on many complicating/mitigating factors. These include quality and availability of medical care, infectious dose and route of infection, differences in host population susceptibility (depending on immune and nutritional status) and genetics, inherent differences in viral variant virulence, and the prevalence of co-infections (particularly malaria and AIDS in patients from sub-Saharan Africa) [9] . The assumption that MARV Angola might be inherently more virulent than other MARV variants has been proposed mainly based on infection studies with nonhuman primates (NHP) [12] [13] [14] but is a matter of debate [15] . The genomes of the Angolan isolates differ about 7% at nucleotide level from the majority of the East African MARV isolates, including the ones from 1967 [10] . There is no evidence so far that the observed genetic differences result in higher virulence in humans. The DRC outbreak was unique, as there were at least nine different virus variants circulating in the tested patients indicative of several different spillover events from the natural reservoir to the human population [9] . In contrast, sequence data from the Angolan outbreak suggested a single introduction of MARV to an unidentified index patient and subsequent spread via person-to-person contact. The viral genomes showed a remarkably high genetic stability within this outbreak. Identical MARV genomes were isolated from patients even after two to three human to human transmissions [10] . MVD is considered a zoonotic disease that is thought to persist in a healthy reservoir host in the endemic areas in Africa. Humans and NHPs are spillover hosts and show a high rate of fatal disease outcomes. Several large-scale attempts to identify the natural host of filovirus infection throughout sub-Saharan Africa had been undertaken in the years since filoviruses first emerged with frustratingly little success [28] [29] [30] [31] . Consistent with ecologic niche modeling of outbreaks and epidemiological patterns, isolated cases have suggested that EBOV is endemic in the rain forests of central and western Africa while MARV is more prevalent in open, dry areas of eastern, south-central Africa [32, 33] . Almost all of the primary infections of natural MVD outbreaks so far have been linked to human entry into caves inhabited by bats (e.g., cave visitors, mine workers) (Table 1) . Thus, bats have long been suspected to play an important role in the transmission cycle of the disease [31, 32, 34] . In 2007, evidence was detected for MARV infection of the common Egyptian fruit bat (Rousettus aegyptiacus) [35, 36] (Figure 2 ), and MARV was isolated from healthy infected R. aegyptiacus bats caught in the same year in Uganda [22] . The bats were collected in Kitaka cave around the same time as human infections occurred that had been linked to the cave (Table 1 and see above, 1. Epidemiology). Genomic analysis of the few isolates of MARV acquired from bats showed that the sequences matched closely to the MARV genomes isolated from patient samples (Figure 1b) . This was also the case for partial MARV sequences isolated from bats inhabiting the Goroumbwa mine in the DRC that was suspected to be the major location for several independent spillover events to gold miners between 1998 and 2000. The bat MARV sequences were closely related to the distinct isolates that had been reported during these outbreaks in humans [36] . A study analyzing MARV prevalence in bat populations in Gabon found MARV-specific nucleic acids in R. aegyptiacus bats in several local caves [37] . Together with previous data showing a high prevalence of MARV-specific antibodies in Gabonese bat populations [38] and with an observed relation of the isolated sequences with previously reported Gabonese bat isolates [35] this study Viruses 2012, 4 1886 suggests that MARV is enzootic in Gabon and raises the concern of further spread of MARV into other countries. Therefore, close ecological as well as serological surveillance of the bat populations in sub-Saharan Africa could help to predict and prevent further MVD outbreaks especially in areas where bats are still used as a food source. It is not currently clear whether R. aegyptiacus bats are the exclusive reservoir for MARV or if other bat species reported to be positive for viral antibodies and RNA are also natural reservoirs or merely intermediate hosts [36] . The genus Marburgvirus includes a single species, Marburg marburgvirus (formerly referred to as Lake Victoria marburgvirus) [39, 40] . Phylogenetic analysis based on genomic sequence data suggests that the known members of this species can be assigned to at least five different lineages of which four are very closely related (nucleotide sequences differ up to 7%) while the fifth is divergent (a nucleotide difference of 21%) ( Figure 1b) [10, 22, 39] . As the genomic divergence between all isolates is below 30%-the cutoff for the classification of the five different ebolaviruses into five different species-the five marburgvirus lineages were recently reclassified as two viruses. Ravn virus (RAVV) is represented by the Ravn isolates from 1987, one isolate from the DRC outbreak in 1998-2000, and one human and several bat isolates from infections that took place in Uganda in 2007. Marburg virus (MARV) is represented by all other sequenced isolates ( Figure 1b , Table 1 ) [39] . For the sake of simplicity, in this review the abbreviation "MARV" is used for all marburgviruses and the abbreviation "EBOV" for all ebolaviruses. Initial MVD patients are believed to contract the virus via exposure to an infected animal: either a reservoir host (several bat species) or a spill-over host such as NHPs as described in the first MVD outbreak (see above, 1. Epidemiology) [23, 41] . Following transmission to humans, spread of the virus between individuals is the result of direct contact with blood or other body fluids (saliva, sweat, stool, urine, tears, and breast milk) from infected patients. Typical risks of exposure include administration of medical care to infected individuals as well as handling of corpses without use of proper protection [34] . Of particular note, virus has been found in tears, semen, and in a liver biopsy weeks to months following the onset of symptoms highlighting the importance of monitoring convalescent patients [20, [42] [43] [44] . Much of what we know about typical MVD symptoms comes primarily from clinical data obtained during the three largest recorded MVD outbreaks: the 1967 outbreak in Germany and Yugoslavia, the 1998-2000 outbreak in the DRC, and the 2004-2005 outbreak in Angola. Although the case fatality rates were significantly higher in the latter outbreaks, most of the clinical symptoms observed were similar. Based on the most reliable documented cases of exposure and subsequent illness, MVD has an incubation period ranging from 3 to 21 days (typically 5 to 10 days), which is likely modulated by factors such as infectious dose and possibly by route of infection. The course of MVD has conventionally been broken down into three phases [45] : an initial generalization phase, an early organ phase, and either a late organ phase or convalescence phase depending upon disease outcome [45] . A summary of MVD symptoms is reviewed below [16, [46] [47] [48] [49] . The onset of illness begins with generic flu-like symptoms; a characteristic high fever (typically 39-40 o C), severe headache, chills, myalgia, prostration, and malaise. For many patients (50-75%) this is followed by rapid debilitation characterized by gastrointestinal symptoms including anorexia, abdominal pain, severe nausea, vomiting, and watery diarrhea. Starting on day four to five patients commonly develop enanthem, dysphasia, and pharyngitis. Additionally, a characteristic maculopapular rash is typically the first distinctive feature indicating a filovirus infection versus influenza or malaria. Other common symptoms include lymphadenopathy, leukopenia, and thrombocytopenia. Many of the initial symptoms may persist in the early organ phase, and patients may sustain a high fever. They may additionally display neurological symptoms including encephalitis, confusion, delirium, irritability, and aggression. Patients can also develop dyspnea and abnormal vascular permeability, particularly conjunctival injection and edema. During the latter part of this phase more than 75% of patients present with some form of clear hemorrhagic manifestation such as petechiae, mucosal bleeding, melena, bloody diarrhea, hematemesis, and ecchymoses. Due to the unusualness of hemorrhagic symptoms, diseases caused by filoviruses have sometimes been referred to as hemorrhagic fevers (Marburg Hemorrhagic Fever (MHF) and Ebola Hemorrhagic Fever (EHF)), although these terms are currently disfavored since not all patients display hemorrhagic symptoms. At this stage, multiple organs are affected including the pancreas, kidney, and liver. Elevated serum activity of a number of liver enzymes including SGOT and SGPT have been observed in most patients sampled. The late stages of MVD result in one of two potential outcomes: patients either succumb to the disease or enter a prolonged phase of recuperation. Typical preagonal symptoms include restlessness, obtundation, confusion, dementia, convulsions, reduced circulation due to severe dehydration, metabolic disturbances, severe diffuse coagulopathy, multiorgan failure, shock, and coma. Fatalities typically occur 8-16 days following the onset of symptoms, with death usually the resulting of shock and multiorgan failure. Non-fatal cases are typified by an extensive convalescent period during which myalgia, exhaustion, sweating, peeling of the skin at the sites of rash, partial amnesia, and secondary infections are all common. Prevention of newly emerging MARV infections and effective containment during ongoing outbreaks is both essential and challenging, as there is currently no licensed vaccine or treatment available for general use. Following the 1967 outbreak of MVD in Europe and cases of infection with ebolavirus Reston in imported crab-eating macaques (Macaca fascicularis) in the USA in 1989/1990 and 1996 as well as 1992 in Italy (reviewed in [50] ), strict quarantine procedures have been put in place that have so far prevented infections acquired by imported NHPs into non-endemic countries [51, 52] . To avoid the spread of filoviruses by tourists, Python cave was closed to the public following the diagnosis of the Dutch patient in 2008. The prevention and control of outbreaks and infections in endemic countries is much more challenging. In the past, joint efforts of teams from the WHO, Doctors Without Borders, the Red Cross, the CDC and others in collaboration with the local ministries of health have been undertaken to cease the spread of MVD. The main focus of outbreak control is the prevention of secondary transmission and further primary infections. The first measures in response to a MVD outbreak include setting up isolation wards in hospitals to assure rapid isolation of MARV-infected patients and prevent person-to-person transmission (Figure 3b ). Proper and fast laboratory diagnosis of suspected cases is key to eliminate further spread. Nosocomial infections were commonly seen in earlier outbreaks [8, 9, 23] . However, reinforcement of barrier nursing techniques and education of health care workers have limited these infections in recent outbreaks ( Figure 3a ). Epidemiological surveillance has been crucial in the identification of index cases as well as the predominant modes of transmission. In endemic areas, secondary infections mainly occurred while taking care of ill patients and family members or during traditional burial practices involving close contact to corpses [53] . Therefore, the execution of safe burial and disinfection techniques and information campaigns to educate the local population are essential in order to contain the spread of infections in endemic areas ( Figure 3a ) [27, 53, 54] . Biosafety and epidemiological efforts alone were not sufficient for efficient outbreak control during large outbreaks, emphasizing the need for additional psychosocial support of the affected communities [53] . The fast progression and high lethality rates associated with MVD even-and especially-after hospital admission resulted in a high level of fear and suspicion by the resident population. The fact that health care workers wearing recommended personal protective equipment (PPE) were fully masked and not identifiable further increased anxiety (Figure 3b ). This resulted in the hiding of infected family members and verbal and in some cases physical aggression towards members of aid organizations [53] . Communicating necessary protective measures while respectfully considering the affected families' and communities' traditions and culture during ongoing outbreaks is therefore essential for successful outbreak management. The recent identification of bats as the potential reservoirs for MARV as well as EBOV [22, 35, 36, 55, 56] will help to increase not only the public awareness, but also the effectiveness of the preventive measures taken in endemic areas to minimize contact with infected animals (i.e. closing of bat inhabited caves for the public, serosurveillance of bat populations) [18, 19] . This is a challenging task, emphasized by the fact that during the last cluster of MARV infections linked to a gold mine in Uganda, the miner hired to enforce the restricted access to the mine got infected. The mine had been closed in response to the ongoing outbreak and even though he was aware of the risk, he had entered the mine without the suggested PPE [19] . Later, the bat population of this mine was eliminated by the owner by means of fumigation [19] . As bats of most species are endangered, this does not seem a viable option and educational campaigns aimed at villagers living close to bat-inhabited caves as well as tourist groups and tour operators might prove more sustainable in the future. In 1967, during the first reported filovirus disease outbreak in Europe, the identification of the previously unknown causative agent of the deadly disease was performed by electron microscopy (EM) (Figure 4 ). The unusual filamentous structure of the particles led to some confusion and it was even suggested that the causative agent of the disease might be related to the spiral-shaped Leptospira, a genus of the spirochaetes bacteria [58] . Others concluded that the observed particles were viruses morphologically related to rhabdoviruses and named the newly discovered pathogen Marburg virus [23, 59] . Marburg virions are pleomorphic particles, which appear as rod-or ring-like, crook-or sixshaped, or branched structures. Cryo-EM analysis of purified virions showed that about 30% of viral particles released from infected Vero cells were filamentous, 37% were six-shaped, and 33% were round [60] . The same study revealed a mean particle length of 892 nm and a mean diameter of 91 nm. Previous conventional EM studies showed that the MARV particles were uniformly 80 nm in diameter, whereas the length varied widely with virions measuring up to 14,000 nm. The average particle length was 790 nm [61] [62] [63] . The reported differences in particle size might be due to experimental differences between cryo-EM and conventional EM [60] . Notably, MARV particles are considerably shorter than EBOV virions, although MARV genomes are slightly longer than EBOV genomes [62, 63] . MARV particles are surrounded by a host-derived membrane that is coated with spikes of 5-10 nm in length, which are formed by trimers of the viral glycoprotein (GP) ( Figure 5 ) [60] [61] [62] [63] [64] [65] . The central core of the viral particles is the ribonucleoprotein complex (nucleocapsid) formed by the viral RNA genome and tightly associated nucleocapsid proteins ( Figure 5 ). The nucleocapsids are highly organized tubular structures with an outer diameter of 45-50 nm and an electron-dense central axis of 19-25 nm. The central axis is surrounded by a helical capsid with cross-striations at a 5 nm interval [61] [62] [63] . A recently published detailed cryo-electron tomography analysis of MARV virions has shed some light on the structural organization of the nucleocapsids. Reconstructions of virion-associated nucleocapsids using subtomogram-averaging analysis revealed that the MARV nucleocapsids form a left-handed helix with a pitch of 7.5 nm and a flexible average symmetry of 14.96 protrusions per turn with two inner lobes of density per protrusion. The inner lobes represent the nucleoprotein (NP), suggesting that the MARV nucleocapsid contains an average number of 29.92 NP molecules per turn with each NP molecule packaging six RNA bases [60] . MARV nucleocapsids show directionality, having a pointed and a barbed tip [60] . The nonsegmented negative-sense (NNS) RNA genomes of the various MARV isolates range in size from 19,111 to 19,114 nts and contain seven monocistronic genes in a linear order ( Figure 5 ) [66, 67] . Each gene is composed of a highly conserved transcription start and stop signal, an unusually long 3' and 5' untranslated region, and the open reading frame (ORF). The genes are either separated by short intergenic regions that range from 4 to 97 nts, or the transcription stop signal of the upstream gene and the transcription start signal of the downstream gene overlap, sharing five highly conserved nts ( Figure 5 ). The structure of this gene overlap is found among all filoviruses and is unique among members of the order Mononegavirales (for review see [68] ). The 3' and 5' genome ends are extracistronic regulatory regions that contain cis-acting signals essential for transcription and replication, including transcription and replication promoters. There are generally two types of genomic replication promoters for NNS RNA viruses: a bipartite promoter found in members of the paramyxovirus subfamily Paramyxovirinae and one continuous more compact replication promoter for rhabdo-and pneumoviruses [69] . The bipartite promoter structure of the Paramyxovirinae subfamily is associated with the "rule of six", i.e., the total genome length must be a multiple of six [70] . Given that filoviruses do not obey the rule of six, it was surprising that mapping of the MARV genomic replication promoter revealed a bipartite structure. The 3' genome end, the leader, comprises 48 nts and contains the first promoter element of the bipartite genomic replication promoter. The second promoter element consists of a (UNNNNN) 3 motif with three conserved uridine residues separated from each other by five non-conserved nucleotides. The UN (5) hexamers are located within the 3' untranslated region of the first MARV gene, the NP gene, and are separated from the first promoter region by the 12 nts long transcription start signal. Substitutions in the NP transcription start signal do not affect replication activity but do interfere with transcription initiation [71] . The 5' extracistronic region, the trailer, spans the last 75 nts of the MARV genome and contains the complement of the antigenomic replication promoter (see below, 8.2. Transcription and Replication). The structure of the MARV antigenomic promoter has not yet been determined. However, due to the presence of UN (5) hexamers it is likely that it is of bipartite nature, similar to the genomic promoter. A common feature of the leader and trailer regions of all NNS RNA viruses is a high degree of complementarity of the 10-15 most terminal 3' and 5' nucleotides [72] . Although filoviruses share this feature, both the leader and the trailer also have the capability to form an internal secondary structure, which is not the case for the leaders and trailers of other NNS RNA viruses [71, [73] [74] [75] . The MARV genome encodes seven structural proteins listed in Table 2 . MARV has a single surface protein, GP, which is encoded by the fourth gene and mediates attachment to target cells and virus entry [76] . GP is a Type I transmembrane protein which is inserted into the viral envelope in the form of homotrimeric spikes [65] . In contrast to ebolaviruses, which use transcriptional editing to express the membrane-bound GP and at least two nonstructural glycoproteins [77] [78] [79] [80] , the MARV GP gene contains a single open reading frame (ORF) encoding the full-length GP. During its transport from the endoplasmic reticulum (ER) to the plasma membrane via the secretory pathway, the precursor GP is the target of various posttranslational modifications including glycosylation [65, 81] , acylation [82] , and phosphorylation [83] . GP is heavily glycosylated by complex and high mannose-type N-linked glycans as well as by mucin-type O-linked glycans, with the carbohydrates contributing about 50% of the apparent molecular weight of the protein [65, 84, 85] . Similar to EBOV GP, the O-linked glycans and many of the N-linked oligosaccharides are clustered in a mucin-like domain [76] . After synthesis in the ER, the precursor GP is cleaved at amino acid 435 by furin or a furin-like protease in the trans Golgi network, resulting in two disulfide-linked subunits, GP 1 (160 kD) and GP 2 (38 kD) [86] . While the ectodomain, which is mainly formed by GP 1 , mediates binding to entry factors and receptors [87] [88] [89] [90] [91] [92] [93] [94] [95] , the transmembrane subunit GP 2 contains the fusion peptide and is presumed to mediate fusion of the viral and the cellular membrane based on similarity to EBOV GP 2 both at the amino acid and structural level [96, 97] . The 30 amino acid long transmembrane domain of GP 2 is required for the incorporation of GP into virions [98] . In addition, the cytoplasmic tail of GP 2 is involved in enhancing the efficiency of viral entry by maintaining the structure of the ectodomain [99] . The receptor binding domain of MARV GP was mapped to the aminoterminal region of GP 1 spanning amino acids 38 to 188 [100] , whereas the highly glycosylated mucin-like domain is not essential for virus entry [101] . An important step in MARV entry is the proteolytic activation of GP 1 by endosomal proteases, facilitating binding of the receptor binding region to the endosomal entry factor Niemann-Pick C1 protein (see below, 8.1. Entry) [94, 95] . Besides its function in entry and budding, GP may also play a role in immune evasion. The IFN-inducible antiviral protein tetherin was shown to block the release of VP40-induced virus-like MARV and EBOV particles, suggesting that tetherin might act as a restriction factor for filovirus release [102, 103] . However, co-expression of GP was sufficient to counteract the antiviral activity of tetherin by a yet unknown mechanism [104, 105] . It is possible that GP not only subverts innate immune responses but also suppresses the adaptive immune response. Filoviral GP 2 subunits, including MARV GP 2 , contain a domain resembling an immunosuppressive motif found in retroviral envelope proteins [106] . A 17-mer peptide corresponding to the putative immunosuppressive domain of MARV GP was shown to induce lymphocyte death and suppression of cytokine responses [107] . It is not yet known if this motif plays a role in the induction of lymphocyte apoptosis observed in MARV infection. Finally, it has been suggested that shedding of the ectodomain of membrane-bound EBOV GP by tumor necrosis factor α-converting enzyme (TACE) may play a role in blocking the activity of neutralizing antibodies during infection [108] . It has been reported for MARV that considerable amounts of GP shed from infected cells, although it is not clear if MARV GP is a target for TACE cleavage [108, 109] . The matrix protein VP40 is encoded by the third MARV gene and is the counterpart of the M proteins of other NNS RNA viruses. VP40 plays a major role in the formation of virions by redistributing nucleocapsids from the perinuclear region to the plasma membrane, recruiting GP to the sites of budding, and mediating particle release [110] [111] [112] . Overexpression of VP40 led to reduced reporter gene expression of MARV minigenomes, suggesting a regulatory role of VP40 in transcription and/or replication [113] . As a peripheral membrane protein, VP40 coats the inner side of the virion's membrane ( Figure 5 ) [114] . Cryo-EM tomography studies suggest that VP40 associates with the nucleocapsid through flexible interactions [60] . It can be easily removed from the nucleocapsid by salt dissociation, indicating that it is only loosely connected to the nucleocapsid [115] . After synthesis in the cytoplasm of the infected cell, VP40 associates rapidly with cellular membranes and accumulates in membranous structures of the late endosomal compartment, the multivesicular bodies. A minor portion of VP40 is also found in association with viral nucleocapsids and in inclusions. Additionally, VP40 appears in patches beneath the plasma membrane where it is transported via the retrograde late endosomal pathway [62, 114, 116] . Similar to EBOV VP40, MARV VP40 is the major factor in particle formation and budding. Expression of VP40 in the absence of other viral proteins leads to the formation and release of filamentous virus-like particles (VLPs) resembling authentic virions. This process is enhanced in the presence of GP [113, [117] [118] [119] . The role of VP40 during budding is described in more detail below (see 8.3. Budding). Compared to EBOV VP40, little is known about the structure of MARV VP40. The N-terminal domain of MARV VP40 folds into ring-like structures, which have the tendency to polymerize into rod-like structures. While EBOV VP40 has been shown to form hexamers and octamers, the stoichiometry of MARV VP40 oligomers is not known [120] . MARV VP40 is phosphorylated at several tyrosine residues located in the N-terminal region of the protein. A non-phosphorylatable mutant of VP40 is impaired in its ability to recruit nucleocapsids to the sites of budding, but is still able to efficiently induce particle release [112] . VP40 also possesses a PPPY late domain motif in its amino terminus which is important for its interaction with components of the Endosomal Sorting Complex Required for Transport (ESCRT) machinery in order to mediate budding, including Tumor susceptibility gene 101 (Tsg101) and the membrane-bound E3 ubiquitin ligase Nedd4.1 [119, 121, 122] . Besides the PPPY motif, other motifs and single amino acids have been found to be important for particle release [123, 124] . Besides its role as a classical matrix protein, MARV VP40 also acts as a virulence factor by counteracting the innate immune response and determining the host tropism for MARV [125, 126] . MARV VP40 blocks the phosphorylation of Janus kinases, which play an important role in multiple signaling pathways by phosphorylating and activating STAT proteins ( Figure 6 ). When MARV-infected cells were treated with various stimuli, including IFNα, IFNγ, and IL6, it was shown that the STAT proteins were neither phosphorylated nor translocated into the nucleus [126, 127] . It was then shown that in MARV-infected cells treated with exogenous stimuli, Janus kinases were also not phosphorylated and VP40 was identified as the viral protein inhibiting IFN signaling. It is believed that Jak1 is the target for VP40, however, the mechanism of VP40-induced inhibition is not completely understood [126] . Intriguingly, EBOV is also able to block IFN signaling by employing a completely different mechanism. EBOV VP24 blocks the nuclear translocation of phosphorylated STAT proteins by binding to STAT1 and importins involved in the nuclear transport of specific STAT proteins ( Figure 6 ) ( [128] , reviewed in [129] ). When MARV was adapted to non-or less permissive animals, such as mouse and guinea pig, the adapted viruses showed mutations in VP40. Two of the amino acid changes in the mouse-adapted MARV VP40 have been shown to be essential for the inhibition of IFN signaling in mouse cells, underlining the importance of IFN suppression for the virulence and host specificity of MARV [125, 130, 131] . The protein product of the sixth gene, VP24, is unique to the filovirus family. VP24 is generally addressed as a second, minor matrix protein. However, cryo-electron tomography analysis of viral particles showed that VP24 is located in close proximity to the nucleocapsid proteins, suggesting that it might be part of the nucleocapsid complex [60] . VP24 can easily be released from virion-associated nucleocapsids by treatment with increasing salt concentrations, indicating that it is only loosely connected to the nucleocapsid [115] . Intracellular localization studies of VP24 showed that a minor part of the protein (approx. 10%) is weakly bound to cellular membranes, including filopodia enriched with VP40. VP24 is also distributed diffusely in the cytoplasm, relocalizes to nucleocapsid-containing inclusions, and is found in association with free nucleocapsids. Co-expression of NP and VP24 is sufficient to direct VP24 to NP inclusions in the cytoplasm [132] . Functional studies on MARV VP24 suggest that the protein is important for the release of viral particles in the context of infection, although it influences neither the morphology of VP40-derived VLPs nor the efficiency of VLP release. In addition, RNAi-mediated knockdown of VP24 in MARV-infected cells had no impact on viral genome replication, indicating that VP24 is involved in a step after replication and before budding [132] . According to the model that has been proposed based on these data, VP24 is involved in the maturation of transport-competent nucleocapsids and/or mediates the interaction between nucleocapsids and budding sites at the plasma membrane [132] . There is also evidence that MARV VP24 affects transcription and replication in a transcription and replication competent VLP system [113] . Structural information about MARV VP24 is very limited. It has been shown that it forms oligomers, preferentially tetramers [132] . Structure prediction studies have proposed an ancestral link between VP24 and the Armadillo repeat family [133] . The MARV nucleocapsid complex consists of the genomic RNA and four tightly associated proteins, NP, VP35, VP30, and L ( Figure 5 ). Encapsidation of the viral RNA by the nucleocapsid proteins protects it from both RNase degradation and detection by cellular pattern recognition receptors. Similar to the genomic RNA, the antigenomic RNA, a replicative intermediate, is also encapsidated by the nucleocapsid proteins (see below, 8.2. Transcription and Replication). In contrast, the viral mRNAs are not encapsidated [134] . The nucleocapsid rather than naked RNA serves as the template for viral transcription and replication. In a MARV minigenome system, NP, VP35, and L are essential for transcription and replication [113, 134] . The role of VP30 in MARV transcription and replication is not well understood and the steps in genome amplification that require, or do not require, VP30 are not yet defined. The nucleoprotein NP enwraps the genomic and antigenomic RNAs. Replication and transcription activity in a MARV minigenome system depends on the presence of NP [134] . When NP is expressed in the absence of other nucleocapsid proteins, it self-assembles into highly organized helical tubular structures that resemble the nucleocapsids in infected cells, indicating that it is the driving force for nucleocapsid formation [60, 135, 136] . Recently, it has been shown that the conserved 390 N-terminal residues of MARV NP are sufficient to form the helical structure of the nucleocapsid core [60] . Indeed, NP serves as a viral hub protein. It forms interactions with most of the other viral proteins, leading to the subcellular redistribution of these proteins. The strong binding of NP to the nucleocapsid proteins VP35 and VP30 redirects both proteins into NP-derived inclusions [115] . A bipartite coiled-coil motif in the central part of NP has been shown to play an important role for self-assembly and NP-VP35 interaction [137] . As mentioned above, there is also a weak interaction between NP and VP24, leading to the partial relocalization of VP24 into NP-containing inclusions [60, 132] . In addition, NP interacts with VP40, which is important for the transport of newly synthesized nucleocapsids to the plasma membrane [110, 111, 138, 139] . Interestingly, NP contains a C-terminal late domain motif, PSAP, which has been shown to be required for budding. NP recruits Tsg101, a component of the ESCRT I complex, through its late domain motif, leading to enhanced VP40-induced budding [111] . NP is heavily phosphorylated at serine and threonine residues clustered in seven regions in the C-terminal part of the protein. Only the phosphorylated form of NP is incorporated into virions [140, 141] . Recent studies suggest that the phosphorylation level in Region II modulates transcription and/or replication activity [142] . VP35 is a polymerase cofactor and essential for transcription and replication. Together with the catalytic subunit L, VP35 forms the RNA-dependent RNA polymerase complex [134, 143] . VP35 is tightly associated with NP and serves as a bridging protein between the nucleocapsid complex and L. Without VP35, L is not associated with the nucleocapsids which serve as the templates for viral transcription and replication [115, 134] . VP35 forms homo-oligomers mediated by a coiled-coil motif located in the N-terminal part of the protein. Homo-oligomerization of VP35 is essential for its interaction with L but not needed for redistribution of VP35 into NP-derived inclusions [144] . VP35 shares many features with the phospho (P) proteins of other NNS RNA viruses, including its position as the second gene in the viral genome and its role in transcription and replication. However, in contrast to the P proteins, VP35 is either not or only very weakly phosphorylated [145] . Besides its function in transcription and replication, MARV VP35 acts as an IFN antagonist. While the impact of EBOV VP35 on the host's antiviral response has been intensively investigated (reviewed in [129] ), much less information is available about similar functions of MARV VP35. When we tested MARV VP35 for its ability to block IFN induction in a reporter gene assay, it blocked reporter gene expression as efficiently as EBOV VP35 (unpublished data). In addition, Bosio and colleagues [146] reported that expression of MARV VP35 in the absence of other viral proteins was sufficient to completely block the induction of IFNα in stimulated human dendritic cells. Besides its ability to inhibit the induction of Type I IFN, EBOV VP35 has been shown to block the activation of the antiviral protein PKR and to interfere with RNA silencing pathways. Importantly, EBOV VP35 is a dsRNA binding protein. The C-terminus of EBOV VP35 contains a domain with patches of basic amino acids which is important for dsRNA binding and the protein's inhibitory functions (for review see [129] ). This C-terminal region, the so-called IFN inhibitory domain, is conserved in MARV VP35 [147] , suggesting that MARV VP35 possesses similar inhibitory functions. MARV and EBOV VP30 proteins show many structural similarities. Both MARV and EBOV VP30 proteins are tightly associated with the nucleocapsid via their binding to NP ( Figure 5 ) [115, 148] . Both are highly phosphorylated at N-terminally located serine and threonine residues, and phosphorylation is crucial for their interaction with NP [148, 149] . Both contain an unusual C3H1 Zn binding domain, which is essential for the function of EBOV VP30 as transcription initiation factor, but whose functional relevance for MARV VP30 is not known [150] . It has also been shown that EBOV VP30 forms hexamers [151, 152] , binds single-stranded RNA [153] , and interacts with L [154] . However, to date, similar data for MARV VP30 are not available. The role of MARV VP30 in viral transcription and replication is not well understood. In contrast to EBOV VP30, which plays an important role in regulating transcription initiation [68, 143, [155] [156] [157] , MARV VP30 is not essential for transcription or replication activity in a MARV minigenome system [113, 134] . Nevertheless, it seems to play an important role in viral amplification, since rescue of a full-length MARV clone is only successful in the presence of VP30 [158] . In addition, down-regulation of VP30 by RNA interference in MARV-infected cells led to the reduction of both viral protein synthesis and virion production [159] . Among the NNS RNA viruses, only the members of the subfamily Pneumovirinae possess a protein similar to VP30, M2-1, which functions as a transcription processivity factor [160] . The major component of the MARV polymerase complex, L, has an estimated molecular weight of 267 kD [67] . It is essential for transcription and replication and together with VP35 forms the RNA-dependent RNA polymerase complex (see above, 7.3. Viral Proteins, VP35). L contains the enzymatic functions of the polymerase. The binding site for VP35 has been mapped to the N-terminal 530 amino acid residues of L [115, 134] . The L proteins of the NNS RNA viruses are highly conserved multifunctional proteins, which are organized in functional domains [161] . Based on this conservation with other NNS RNA polymerases, MARV L is believed to carry out RNA synthesis, capping, and polyadenylation of viral mRNAs although these functions have not been shown experimentally. To date most of the studies characterizing the MARV replication cycle have utilized recombinant systems, allowing for these experiments to be performed in a biosafety level 2 (BSL-2) context, unfettered by the restrictions of a BSL-4 setting. Surrogate systems mimicking specific steps in the MARV replication cycle include MARV GP-pseudotyped retroviruses or recombinant vesiculoviruses expressing GP to study entry, VLPs to study budding, and minigenome systems to study replication and transcription. While such experiments allow for the more facile examination of the MARV replication cycle, the findings must be recapitulated with infectious MARV since all surrogate systems lack elements of the infectious virus such as the distinct morphological features and virion protein composition of MARV. Marburg virus entry consists of three distinct phases: cellular attachment, endocytosis, and fusion ( Figure 7) . Based on the sequence similarity between EBOV and MARV GPs many investigators have presumed identical functions and characteristics between the filovirus glycoproteins. This is presumptuous given the differences in glycosylation and sialic acid linkages [85] and dependence upon endosomal proteases (see below, 8.1.2. Endocytosis). Despite the existence of a number of detailed studies and structural analyses of EBOV GP [162] [163] [164] [165] , relatively few mechanistic studies of MARV GP have been performed [101] , although one recent post-fusion structure of MARV GP 2 has been reported [97] . The structure of MARV GP 2 is nearly identical to that of EBOV GP 2 , indicating that the mechanisms of fusion between the two viruses is likely conserved [97] . (2), GP 1 is cleaved by endosomal proteases (3) facilitating binding to NPC1, the entry receptor (4). Fusion is mediated in a pH-dependent manner by GP 2 . Following release of viral nucleocapsid into the cytosol (5), transcription of the viral genome takes place (6) . mRNA is subsequently translated by the host cell machinery (7) . Synthesis of GP takes place at the ER and undergoes multiple posttranslational modifications on its way through the classical secretory pathway (8) . Positive sense antigenomes are synthesized from the incoming viral genomes (9) . These intermediate products then serve as templates to replicate new negative sense genomes (10) . After cleavage in the Golgi, GP is transported to multivesicular bodies (MVB) and to the cell membrane where budding takes place (11) . Nucleocapsids and VP24 are also recruited to sites of viral budding (12) , which is driven by VP40 (13). MARV GP mediates both cell attachment and fusion of the virus. There is convincing evidence that initial virus attachment at the cell surface can occur via the binding of GP carbohydrates to various cellular C-type lectins, including the hepatocyte-specific ASGP-R [87] , DC-SIGN and DC-SIGNR (also known as L-SIGN) [89, 92, 166] , hMGL [91, 92] , and LSECtin [166, 167] . Other cell surface proteins have also been implicated in facilitating MARV entry including the TAM receptor protein kinases Ax1, Dtk, and Mer [90] , and TIM-1 [93] . However, although these proteins may play a role in attachment or entry of certain cell types, the ability of MARV to infect cells lacking these receptors [92, 93, 168] indicates that there might be redundancy in cellular molecules required for MARV attachment to cells. A number of key residues of EBOV GP that are involved in virion incorporation and GP-mediated entry have been identified [163] and found to play a similar role in MARV GP [101] , indicating that the viruses might utilize similar mechanisms to enter the cell. Following attachment, Marburg virions undergo endocytosis mediated through a mechanism that currently remains undetermined. (Figure 7 ) [16, 62] Initial studies investigating caveolin-mediated endocytosis showed that depletion of host cell cholesterol reduced viral infectivity but presented no direct evidence of caveolae involvement [169] . In addition, studies examining the role of caveolae in EBOV endocytosis are conflicting [169, 170] . A major role for clathrin in MARV entry has also been proposed based upon the ability of chlorpromazine (an inhibitor of both clathrin-mediated endocytosis and macropinocytosis) as well as RNAi-knockdown of clathrin heavy chain to inhibit MARV GP-pseudotyped HIV-1 entry [171] . A caveat to these analyses of MARV endocytosis is that they were performed only in the context of MARV GP-pseudotyped retroviruses which lack the characteristic filamentous morphology and size of Marburg virions. While other reports have verified that cholesterol is important for live MARV particle uptake [172] , canonical caveolae-and clathrin-mediated mechanisms are unlikely to be the primary mechanism of MARV entry due to steric issues. The typical MARV particle size (average 790 nm) is much larger than canonical caveolae (50-100 nm) or clathrin-coated pits (up to 200 nm) whereas pseudotyped murine leukemia virus (MLV) (100 × 100 nm) and vesicular stomatitis virus (VSV) (70 × 180 nm) are not [173] . These findings indicate that involvement of the caveolae-and clathrin-mediated endocytic pathways for virus entry may therefore be the result of the artificial nature of the pseudotyped virions and highlights the need to confirm such experiments with live MARV. Macropinocytosis has been identified as a major entry pathway of EBOV by research using the morphologically more relevant VLPs and live EBOV [174] [175] [176] [177] . Although none of these analyses examined the role of this pathway during MARV entry, it remains an intriguing possibility given the cholesterol-dependence and large size of macropinocytotic vesicles (up to 3-5 µm) [178] . Another important process in MARV entry is believed to occur while virions are being trafficked within endocytic vesicles; the proteolytic cleavage of GP 1 . Endosomal cleavage of GP has been shown to be critical for the efficient entry of MARV [179, 180] . The current model for MARV entry involves the cleavage of GP 1 by host endosomal cysteine proteases. This removal of a large portion of GP 1 (including the mucin-like domain) is believed to expose the putative receptor-binding domain based on studies conducted with EBOV GP [181, 182] . Studies examining the roles of endosomal proteases on the entry of MARV and EBOV have produced mixed results. Experiments analyzing recombinant VSV expressing EBOV GP indicate a primary role for Cathepsin B (CatB) and minor role for Cathepsin L (CatL) [181] . Entry of recombinant VSV particles containing MARV GP was inhibited when cells were treated with an inhibitor of both CatB and CatL [92] . These reports are confounded by a report conducted with infectious Marburg and Ebola viruses in which CatB and CatL inhibitors greatly reduced EBOV infection but showed mixed results with MARV [172] . Yet two other, more recent analyses determined that CatB was not required for MARV entry (although over-expression did enhance infectivity) and that CatL was required for entry into mouse embryonic fibroblasts but not Vero cells, 293T cells, or human macrophages [179, 180] . These data as well as the ability of other proteases to greatly diminish MARV infectivity [179, 180] , indicate that although CatB and CatL likely play a role in cleavage and activation of GP 1 in certain cell types, other endosomal proteases may also be able to facilitate GP 1 activation via cleavage. Recently, two independent studies elegantly showed the requirement of the endosomal cholesterol transporter Niemann-Pick C1 (NPC1) for the entry of both MARV GP-pseudotyped viruses (VSV and MLV) as well as infectious MARV [94, 95] . It was also shown that NPC1 catalytic activity is not required for EBOV infection indicating that specific binding to NPC1 rather than its role in cholesterol transport is required, although this was not tested for MARV [95] . In one of the studies identifying NPC1 as the MARV entry receptor, it was also determined that members of the homotypic fusion and vacuole protein-sorting (HOPS) complex were important for EBOV entry, although they appeared to play a less important role in MARV entry [94] . The current model of EBOV and MARV fusion is that GP 1 cleavage by endosomal proteases removes heavily glycosylated domains, exposing the receptor binding domain on GP 1 and enabling binding to NPC1 [95] . The membrane-bound fusogenic GP 2 undergoes a low pH-dependent rearrangement to an extended conformation resulting in the fusion of virion and endo-lysosomal membranes [96] . In support of the pH-dependence of GP-mediated fusion, pre-treatment of cells with ammonium chloride prevented entry of a MARV GP-pseudotyped virus [183] . A recent report with live MARV showed that ammonium chloride inhibited entry and replication, but that Bafilomycin A 1 , which specifically inhibits vacuolar-type H(+) ATPase and prevents re-acidification of vesicles of the central vacuolar system, surprisingly had no effect [172] . Following viral fusion with the endosomal membrane, the nucleocapsid is released into the cytoplasm (Figure 7 ). After the nucleocapsid is released into the cytoplasm of the infected cell, transcription and replication of the viral RNA genome takes place (Figure 7 ). The first morphological sign of viral replication observed by EM analysis is the appearance of granular material containing RNA and viral proteins in the cytoplasm of the infected cells at 12 h post infection. Later on, tubular structures can be detected in the granular material representing the newly synthesized nucleocapsids embedded in the viral inclusions [63] . While experimental data on the sites of MARV replication and transcription are not available, recent studies on EBOV have shown that viral replication takes place in the inclusions, while transcription was observed prior to inclusion formation [184] . The encapsidated negative-sense RNA genome is transcribed resulting in seven monocistronic mRNAs by the viral polymerase. They are co-transcriptionally capped and polyadenylated and subsequently translated by the cellular machinery (Figure 7) . The genomic RNA also serves as the template for the production of positive-sense antigenomes, which are complementary copies of the genomes. The antigenomes are encapsidated by the nucleocapsid proteins and are in turn used as templates for genome synthesis (Figure 7 ) (for review see [68] ). As mentioned above, NP, VP35, L, and probably VP30 are needed for viral transcription and replication. Analogous to EBOV, it is conceivable that VP40 and VP24 inhibit transcription and replication [113, 185, 186] . It is hypothesized that negative regulators of replication convert the active polymerase complex into an inactive state, resulting in mature and transport-competent nucleocapsids. Following assembly, newly synthesized nucleocapsids are recruited to the sites of virus budding (Figure 7) . Release of viral particles is mainly mediated by VP40 via recruitment of nucleocapsids from the inclusions to the plasma membrane, recruiting GP to the sites of budding, and inducing the formation and release of filamentous VLPs. VP40-induced budding is enhanced by NP, GP, and VP24 [98, 111, 132] . As is the case with many other viruses, MARV exploits the vesicular transport machinery of the infected cell for viral egress, including the COPII vesicular transport system and the ESCRT machinery. The COPII vesicular transport system is used by VP40 for its intracellular trafficking to the multivesicular bodies, where MARV budding takes place [118, 187] . Cellular proteins that promote particle release and are linked to the ESCRT machinery include Tsg101, Vps4A/B, and Nedd4.1 [111, 119, 121, 122] . MARV budding not only takes place at internal membranes but also at the plasma membrane [63, 118, 188] . In cell culture, MARV particles are preferentially released at filopodia, filamentous cellular protrusions [60, 188, 189] . Filopodia are used by cells to explore the extracellular environment, which includes neighboring cells, and it is believed that viral particles can bud directly into adjacent cells via filopodia-mediated cell-to-cell contact [188, 190] . Budding at filopodia depends on actin and is not sensitive to the depolymerization of microtubules [188] . MARV budding was observed at the basolateral membrane of polarized epithelial cells and hepatocytes [109, 191] , whereas viral particles were predominantly released from the apical membrane of infected endothelial cells [64] , suggesting that cell-type specific components determine the sites of virus release. Electron tomography studies of MARV-infected cells led to the following model for MARV particle release: The budding process is initiated when intracellular nucleocapsids associate laterally with the plasma membrane. Starting from one end, the nucleocapsids are then subsequently wrapped by the plasma membrane until the viral particles protrude vertically from the cell surface. The release of infectious filamentous MARV from cultured cells peaked at 1-2 days post infection, when the cells were still intact. At 4 d post infection, when most of the cells were vesiculated, the released virions were round or bent and infectivity was decreased [189] . Determination of the nucleocapsid orientation at the sites of budding by 3-D reconstructions revealed that the pointed tip of the budding nucleocapsids is oriented towards the membrane, indicating that MARV budding is directional [60] . MARV infections usually occur by direct contact with infected body fluids or direct personal contact with infected animals or humans. The viruses enter the body through small skin lesions or mucosal membranes (reviewed in [47] ). Cells of the mononuclear phagocyte system, including monocytes, macrophages and dendritic cells, are early target cells of MARV, as shown in different experimental animal models [14, [192] [193] [194] [195] [196] [197] . MARV replication was observed as early as 24 hours post infection in macrophages of infected guinea pigs [63] , and infected monocytes have been found in cynomolgus macaques at 2 days post infection [193] . Monocytes and macrophages were also identified as early target cells in human patients [197] . This has been confirmed by cell culture experiments showing that primary human monocytes and macrophages are highly susceptible to MARV infection and produce infectious virus [198] [199] [200] . In addition, primary human monocyte-derived dendritic cells (mDCs) and endothelial cells support MARV replication [64, 146, 201] . Early sites of virus replication are the lymph nodes, liver, and spleen where the most severe necrotic lesions are observed [8, 13, 14, 41, 130, 194, 202] . These organs contain high numbers of monocytes and macrophages. Migration of infected monocytes and macrophages into surrounding tissues or transport of free virus via the lymph-or bloodstream is believed to facilitate the dissemination to multiple organs, resulting in a systemic infection [203, 204] . Cell-free virus has been observed in the tissue and organs of infected animals, and high levels of virus have been detected in the blood [12] [13] [14] 130, 194, 195, 205, 206] . Besides monocytes, macrophages, and dendritic cells, a wide range of cell types including hepatocytes, adrenal cortical and medullary cells and fibroblasts are permissive to MARV infection [12, 14, [192] [193] [194] [195] 197, 201, 207] . Endothelial cells are late target cells during MARV infection in multiple tissues. Whether or not replication of MARV in endothelial cells is associated with the observed vascular impairment during MVD is a matter of debate [64, 194] . Only low numbers of infected endothelial cells are observed in NHP infection and therefore changes in the endothelium are likely caused by paracrine effects of cytokines [14] . In late stages of infection MARV particles can be isolated from nearly every organ [12, 14, 130, 194, 208] . Despite high viral load and necrotic lesions, only minor inflammation is observed in infected tissues and organs, indicating a dysregulated immune response [8, 14, 41, 194] . Strong liver pathology is observed, including increased serum activity of liver enzymes. This might influence synthesis of clotting factors and contribute to the observed coagulation defects in MVD [12, 41, 130, 194] . These factors together with systemic virus replication and associated pathology probably trigger the multiorgan failure associated with fatal cases. Although lymphocytes are not susceptible to MARV infection [14, 193, 194, 201] , massive bystander lymphocyte apoptosis is a hallmark of MVD [8, 14, 130, 194, 195, 201] . However, the molecular mechanisms for lymphocyte depletion and the role it may play in the pathogenesis of MVD are far from being understood. Cytokine secretion may play a role in the induction of lymphocyte apoptosis, since MARV-infected cells secrete cytokines known to induce apoptosis, including TNFα [194, 198, 200, 209] . Increased levels of TNFα have been observed in infected rhesus macaques [210] and mice [130] , although no increase was observed in infected cynomolgus macaques [14] . Elevated TNFα levels may also play a role in the formation of endothelial gaps in the context of MARV infection [198, 199] . In addition, increased survival of MARV-infected guinea pigs treated with anti-TNFα antibodies suggests that TNFα indeed plays an important role in MVD pathogenesis [209] . Increased serum levels of additional proinflammatory cytokines and chemokines have been observed in infected NHPs and in mice, but the reported data are not completely consistent [12, 14, 130, 193, 210] . Cytokine and chemokine secretion has also been observed in infected primary human monocytes and macrophages [200, 211] . However, data about the cytokine levels in the serum of MARV-infected patients are not available, but high levels of cytokines have been observed in EBOV-infected patients [212] [213] [214] [215] . Upregulation of the proinflammatory cytokines IL6 (mediator of fever and acute inflammatory response) and IL8 (chemoattraction of neutrophils and macrophages) is consistently found in infected NHPs, with macrophages and plasmacytoid dendritic cells (pDCs) serving as the main sources of IL6 secretion in the spleen [12, 14, 193] . Primary human monocytes and macrophages produce both IL6 and IL8 after infection [200] . Elevated levels of IL6 have also been detected in MARV-infected mice [130] . Increased levels of IL1β mRNA and secreted protein were detected in primary human cells [200, 211] , but contradictory data have been reported for the NHP model. One study reported elevated IL1β levels in final disease stages [12] , whereas no change was observed in another study [14] . IFNα levels were elevated in infected NHPs and mice [130, 193, 210] . However, no change in IFNα levels was detected in another study of infected NHPs [14] . It is unclear whether or not the observed differences are due to different MARV variants being used for the studies. Serum levels of several chemokines were also found to be elevated during MARV infection of NHPs and mice, including macrophage inflammatory proteins (MIP) and monocyte chemotactic protein 1 (MCP-1) [12, 14, 130] . The involvement of multiple cell types along with the possible role of non-infected cells in the secretion of cytokines further complicates the analysis of existing data. Primary human monocytes and macrophages are activated by MARV infection inducing the secretion of cytokines. Induction of cytokines has also been described using UV-inactivated MARV, suggesting that viral replication might not be needed for the observed cytokine increase [200] . In contrast, MARV-infected mDCs show no upregulation of activation markers, do not secrete cytokines, and fail to stimulate T cells [146] . mDCs treated with VLPs containing MARV VP40 and GP show functional mDC responses including cytokine secretion indicating that MARV replication is required to inhibit mDC activation [216] . However, infection of mDCs with MARV did not prevent LPS-induced TNFα production whereas dsRNA-dependent IFNα secretion was inhibited [146] , suggesting differential regulation of cytokines by MARV. Interestingly, pDCs in the spleen were identified as the major source of secreted IFNα in MARV infected NHPs, but secretion most likely occurs from non-infected cells [193] . It has been shown for EBOV that pDCs are not productively infected due to impairment of viral entry [217] . These results suggest that secretion of cytokines by non-infected bystander cells might play an important role during MARV pathogenesis. Taken together, MARV infection induces both an increase in the production of proinflammatory cytokines and high levels of chemokines, but the molecular mechanisms causing these changes are not well understood. To date four different animal models have been established for MARV infection: NHPs, mice, guinea pigs, and hamsters. The NHP model best reflects the symptoms and pathology observed in human cases (described in 5. Clinical Manifestations and reviewed in [46, 48] ) with uniform lethality in cynomolgus and rhesus macaques as well as African green monkeys [12, 14, 26, 193, 194, 206, [218] [219] [220] . The disease symptoms are generally the same for all types of NHPs. The animals develop febrile illness with high fever, anorexia, weight loss and unresponsiveness. Death is observed after 6-13 days and thrombocytopenia, lymphopenia, blood coagulation abnormalities and hemorrhages are observed. Squirrel monkeys have also been successfully infected with MARV, showing typical disease symptoms [206] . Recently, a small NHP model using marmosets has been developed recapitulating the features of human infections except for the typical maculopapular rash development that is observed in other NHPs and humans infected with MARV [221] . Rodents with an intact immune system do not develop disease after infection with MARV. MARV variants Musoke and Ci67 and RAVV variant Ravn were adapted to severe-combined immunodeficiency (scid) mice by serial passaging, reducing the time to death from 50-70 days to 7-10 days for all tested virus variants [195] . Further passaging of the scid mouse-adapted marburgviruses in immunocompetent mice was used to establish mouse models for RAVV Ravn and MARV Ci67 [130, 131] . Successful adaptation by serial passaging was also used to generate lethal infection models for both guinea pig [218, 222] and hamster [206, 208] . Coagulation abnormalities, typical rash development, and hemorrhagic manifestations (especially in mice) are not as pronounced as in the NHP model [130] ; (reviewed in [223] ). Neuropathogenicity, recapitulating the CNS involvement described during the first human MVD outbreak in Germany [41, 224, 225] , has only been observed in the hamster model [208] . It is not clear if CNS pathology is developed in other animal models as no brain pathology has been observed in mice [195] and cynomolgus macaques [14] . Nevertheless, virus has been isolated from brain from MARV-infected marmosets, showing micro-hemorrhages [221] . Sequence comparison of the rodent-adapted viruses to the human MARV isolates revealed several mutations. Sixty-one nucleotide changes in the mouse-adapted RAVV Ravn variant were detected in the ORFs of NP, VP35, VP40, and VP30 (14 amino acid changes in total) or untranslated regions (VP35, VP40, GP, VP30) [130] . In a second study analyzing genome changes during mouse adaptation, 75 nucleotide changes during adaptation of RAVV Ravn and 33 changes for MARV Ci67 were described, with most amino acid changes occurring in VP40 [131] . During guinea pig adaptation, only 11 nucleotide changes were observed, resulting in four amino acid changes. One of these changes was located in VP40 and the other mutations were detected in the viral polymerase L [222] . The only amino acid exchange detected in both the mouse-and the guinea pig-adapted MARV is amino acid 184 in VP40 (Asp to Asn). This was also the first mutation detected during mouse adaptation for both RAVV Ravn and MARV Ci67, as analyzed by sequencing of serial passages [131] . This is of particular interest because VP40 has been shown to function as an inhibitor of IFN signaling (see above, 7.3. Viral Proteins, VP40) [125, 126] . Both IFN receptor-and STAT1-deficient mice develop disease using non-adapted MARV, highlighting the importance of the IFN pathway for the control of MVD [226] [227] [228] . Virological, serological, and molecular diagnostic methods for the detection for MARV are available, including virus isolation, ELISA, RT-PCR, EM, and immunohistochemistry (summarized in [16, 46] ). During outbreak settings, mobile laboratories commonly use PCR and/or ELISA analysis for rapid screening. Sensitive ELISA assays have been developed for detection of viral antigen or virus-specific antibodies using overexpressed MARV NP or GP [229] [230] [231] [232] . Detection of filoviruses by PCR is the only assay currently available to distinguish between different virus variants for a variety of tissue and fluid specimens. The use of a combination of virus-specific primer sets for conventional RT-PCR makes the detection of all know filoviruses in a single assay possible [233] . For more sensitive and quantitative detection, real-time RT-PCR-based assays have been developed for the detection of MARV [10, [234] [235] [236] [237] [238] . Feasibility of real-time RT-PCR in the field was successfully proven during the MARV outbreak in Uíge, Angola using an improved field laboratory-adapted RNA isolation protocol [239] . A network of European BSL-4 facilities in collaboration with a company (QIAGEN) developed the first commercial prototype of a real-time-RT-PCR assay for the detection of filoviruses [240] . A recently developed assay, RT loop-mediated isothermal amplification (LAMP), has the potential to significantly improve field diagnosis of MARV infections, by eliminating the need of PCR machines [241] . Initial approaches using inactivated virus to develop a vaccine against MARV were unsuccessful or had contradictory results [16] . In addition, successful protection of rodents did not always translate into protection of NHPs. For example, inactivated MARV protects guinea pigs from lethal MARV challenge but only 50% of challenged NHPs survived [210, [242] [243] [244] . A panel of different approaches has been used in order to develop successful vaccines for MARV (reviewed in [245, 246] ). Recombinant GP expressed from insect cells or a DNA vaccine based on GP only partially protected guinea pigs, but use of a combination of both vaccines resulted in 100% survival of guinea pigs [243] . In another study, complete protection of guinea pigs using a different GP DNA vaccine was reported, but only four of six vaccinated NHPs survived the challenge with MARV, showing incomplete protection [247] . Increased doses of a codon-optimized DNA vaccine resulted in 100% survival of NHPs, although some animals developed symptoms before recovering. In comparison to other vaccine candidates, a poor induction of virus-specific antibodies was observed using a DNA vaccine [248] . A codon-optimized DNA vaccine elicited a strong antibody response and resulted in complete protection of mice with no clinical symptoms observed [249] . Vaccine candidates based on the Venezuelan equine encephalitis virus (VEEV) replicon system expressing either MARV GP along with NP or GP alone completely protected guinea pigs and NHPs [250] . A vaccine based on VLPs represents an additional candidate for protection against MARV [251] . Complete protection of guinea pigs has been demonstrated with a VLP-based vaccine containing MARV GP, with induction of virus-specific antibodies. Protection with this vaccine relied on a functional CD4+ T cell response, whereas depletion of CD8+ T cells did not ablate the protective response [244] . VLPs containing MARV Musoke GP provided cross-protection in animals challenged with MARV Ci67 or RAVV Ravn in both guinea pigs and NHPs [252, 253] . Another approach to MARV vaccines is the use of viral vectors expressing MARV GP. To date, two different systems have been established based on replication-defective adenoviral vectors or recombinant VSV expressing MARV GP. The adenovirus-based vaccine successfully protects guinea pigs and NHPs, and provides cross-protection. High levels of cross-reactive MARV-specific IgG and T cell responses are induced, indicating an induction of an immune response [248, 254] . Preexisting immunity against the adenovirus strain Ad5 might pose a problem for its successful use in humans (reviewed in [245] ). The VSV-based vaccine completely protects NHPs and additionally has proven successful in post-exposure treatment (reviewed in [255] ). A single immunization with recombinant VSV expressing MARV Musoke GP resulted in 100% protection of cynomolgus macaques challenged by intramuscular injection or aerosol exposure and protected against RAVV Ravn and MARV Angola [13, 256, 257] . Although MARV-specific IgG were produced, only low levels of neutralizing antibodies were detected [13, 257] . Surprisingly, T cell-mediated responses were not observed in NHPs vaccinated with recombinant VSV expressing MARV GP [13, 256] . Safety is a concern for this vaccine, especially for immunocompromised individuals, as it is a replication-competent VSV vector. However, in all VSV-based filovirus vaccine studies VSV viremia was observed only shortly after immunization. Additionally, the VSV-based filovirus GP vaccine was well tolerated and protective in immunocompromised mice and NHPs and lacked neurovirulence in NHPs [258] [259] [260] (reviewed in [255] ). Cross-protection has not been observed in animals vaccinated with MARV-based vaccines and subsequently challenged with EBOV, while combined MARV and EBOV vaccines have been successful in protection against both viruses [252, 261, 262] . To date no approved treatment is available for MARV infection. Supportive care (fluids, anti-microbials, blood transfusions) has been the primary treatment of patients during MVD outbreaks. In the guinea pig model various treatments had some success as reflected by prolonged survival or increased survival rates. Applied treatments included cytokine inhibition, IFN treatment, or antibody transfer. The tested treatments were unsuccessful in the NHP model (reviewed in [16, 47] ). A third of EBOV-infected NHPs survived, however, following treatment with recombinant nematode coagulant protein 2, while only one of six MARV-infected animals survived [12, 263] . Treatment using antisense technology to block viral protein expression using phosphorodiamidate morpholino oligomers (PMO) beginning 30 to 60 minutes after MARV infection completely protected NHPs [264] . Additionally, a small molecule inhibitor showed complete protection of MARV-infected mice when administered 24h after infection but has not been tested in NHPs [265] . The VSV-based vaccine expressing MARV GP has also been demonstrated to be effective as a post-exposure treatment. A hundred percent survival of NHPs was observed when the vaccine was administered 20 to 30 minutes after MARV infection [266] . Delaying the time before treatment results in incomplete protection, although five of six animals or two of six animals still survived when given the treatment 1 or 2 days after MARV exposure, respectively [205] . These post-exposure treatments may be useful to prevent disease after known exposure to MARV, such as a laboratory accident, but the effective time frame during an outbreak might be too short and alternatives are needed.

The Human Cytomegalovirus DNA Polymerase Processivity Factor UL44 Is Modified by SUMO in a DNA-Dependent Manner

During the replication of human cytomegalovirus (HCMV) genome, the viral DNA polymerase subunit UL44 plays a key role, as by binding both DNA and the polymerase catalytic subunit it confers processivity to the holoenzyme. However, several lines of evidence suggest that UL44 might have additional roles during virus life cycle. To shed light on this, we searched for cellular partners of UL44 by yeast two-hybrid screenings. Intriguingly, we discovered the interaction of UL44 with Ubc9, an enzyme involved in the covalent conjugation of SUMO (Small Ubiquitin-related MOdifier) to cellular and viral proteins. We found that UL44 can be extensively sumoylated not only in a cell-free system and in transfected cells, but also in HCMV-infected cells, in which about 50% of the protein resulted to be modified at late times post-infection, when viral genome replication is accomplished. Mass spectrometry studies revealed that UL44 possesses multiple SUMO target sites, located throughout the protein. Remarkably, we observed that binding of UL44 to DNA greatly stimulates its sumoylation both in vitro and in vivo. In addition, we showed that overexpression of SUMO alters the intranuclear distribution of UL44 in HCMV-infected cells, and enhances both virus production and DNA replication, arguing for an important role for sumoylation in HCMV life cycle and UL44 function(s). These data report for the first time the sumoylation of a viral processivity factor and show that there is a functional interplay between the HCMV UL44 protein and the cellular sumoylation system.

Most replicative DNA polymerases include a catalytic subunit, responsible for DNA polymerization, and a processivity factor that holds the catalytic subunit on DNA to allow continuous DNA synthesis. One of the best-studied processivity factors is proliferating cell nuclear antigen (PCNA) of eukaryotic DNA polymerases d and e [1] . PCNA, which belongs to the family of so-called ''sliding clamps'', has no inherent DNA-binding capacity, but with the aid of clamp loader proteins is assembled onto DNA as a toroidal homotrimer [2] . In addition to DNA replication, PCNA has been implicated in DNA recombination and repair, as well as in DNA methylation, chromatin remodeling, and cell cycle regulation [1] . Consistent with its pleiotropic functions, it interacts with a plethora of proteins [3] and undergoes a number of posttranslational modifications, including phosphorylation, acetyla-tion, ubiquitination and sumoylation, which are believed to regulate its subcellular localization, stability and protein binding specificity [4, 5, 6] . The human cytomegalovirus (HCMV) DNA polymerase includes a catalytic subunit, UL54 (the UL54 gene product), and an accessory, homodimeric subunit, UL44 (the UL44 gene product), that binds DNA without the aid of clamp loaders [7] yet wraps around DNA akin to PCNA [8] . While UL44 shows no apparent sequence homology with PCNA, there is striking structural similarity between UL44 and PCNA monomers [2, 9] . Similarly to PCNA, UL44 is a phosphoprotein [10] . Intriguingly, the phosphorylation state of UL44 has been shown to regulate its nuclear import rate by controlling its interaction with host cell factors [11, 12, 13] . The best-characterized function of UL44 during HCMV infection is that of binding to UL54 through a region named connector loop [14, 15, 16] , stimulating its activity and conferring processivity to the holoenzyme [17, 18] . However, UL44 continues to accumulate to strikingly high levels at late times after infection, when DNA replication is accomplished [19, 20] . Its early-late kinetics of transcription and the high level of expression suggest that UL44 might play additional roles during the viral life cycle. To investigate this possibility, we conducted yeast two-hybrid (Y2H) screenings to search for cellular partners of UL44. To our surprise, Ubc9, an enzyme involved in the sumoylation process, was identified as a UL44 protein interaction partner. Sumoylation is a post-translational protein modification analogous to ubiquitination. It consists of reversible and covalent conjugation of SUMO (Small Ubiquitin-related MOdifier) to a protein target [21, 22] . In the sumoylation cascade, the C-terminus of SUMO is activated by an activating enzyme (E1), transferred to a conjugating enzyme (E2, that is Ubc9), and linked to a lysine residue of the substrate protein with the aid of a ligase (E3). Mainly, three SUMO paralogs (SUMO-1, -2, -3) have been identified so far [23, 24] . SUMO-2 and SUMO-3 are highly homologous to one another (95% identity) while they differ from SUMO-1 by 50%. Conjugation of SUMO-1 has been shown to play a functional role in a number of biological processes, ranging from nucleocytoplasmic transport to transcription, the maintenance of genome stability, nucleic acid DNA metabolism, cell signaling, and many others [21] , whereas the role of SUMO-2/23 modification is less clear. Here we report that the association of Ubc9 and UL44 leads to conjugation of SUMO molecules on multiple lysine residues. Both SUMO-1 and SUMO-2/3 were found to be conjugated to UL44. Sumoylation of UL44 was detected not only in vitro and in transiently transfected cells but, more importantly, also in HCMVinfected human cells during virus replication. Interestingly, we observed that binding of UL44 to DNA greatly stimulates SUMO conjugation to the protein both in vitro and in cells. In addition, we show that overexpression of SUMO-1 alters the intranuclear distribution of UL44 in HCMV-infected cells, and enhances both viral DNA replication and virus production in an Ubc9-dependent manner. These data represent the first report of sumoylation of a viral processivity factor and show that there is a complex interplay between the HCMV UL44 protein and the cellular sumoylation system. The Y2H plasmids expressing LexA-UL44 and LexA-Ubc9 were generated by cloning the UL44 and Ubc9 coding sequences from pRSET44 (a gift of P. F. Ertl, GlaxoSmithKline, UK) and pACT2-Ubc9 (from G. Gao, Chinese Academy of Sciences, Beijing, China) respectively, in pBTMK, derived from pBTM116 [25] . The pACT-UL44 and pACT2-Ubc9 plasmids, encoding GAD-UL44 and GAD-Ubc9 fusions, respectively, have been described in [26, 27] . The plasmid expressing GAD-UL54 was created by cloning the UL54 coding sequence from pRSET-Pol (a gift of P. F. Ertl) in pACT2 (Clontech). Plasmid pRSET44 was used to express 6His-UL44 in Escherichia coli. Plasmid pRSET-Ubc9 was constructed by cloning the Ubc9 coding sequence from pACT2-Ubc9 in pRSET (Invitrogen). Plasmid pCDNA3-PB1, used for in vitro transcription of the PB1 subunit of influenza A virus RNA polymerase, was described previously [28] . Plasmids pD15-GST and pD15-UL44, which express GST and GST-UL44, respectively, have been described in [29] . Plasmid pTE1E2S1 [30] was provided by H. Saitoh (Kusamoto University, Japan). Plasmids GFP-UL44, pDESTnV5-UL44, and pDESTnV5-UL44DNLS have been described previously [11, 31] . Plasmids pDsRed2-Ubc9 [27] and pDsRed2-UL53 [32] were kindly provided by G. Gao (Chinese Academy of Sciences, Beijing, China) and D. Camozzi (University of Bologna, Italy), respectively. Plasmid pCDNA3.1-UL44-FLAG (from M. Marschall, Universitat Erlangen-Numberg, Germany) was used to express C-terminally FLAG-tagged UL44 [33] , while plasmid pDEST-nFLAG [34] was used to express N-terminally FLAG-tagged UL44 [35] , UL44Dloop [36] and UL44L86A/L87A in mammalian cells. Plasmids pCDNA3-Ubc9, pCDNA3-Ubc9C93S, pCDNA3-HA-SUMO-1, pCDNA3-HA-SUMO-2, and pCDNA3-HA-SUMO-3 used in overexpression experiments in mammalian cells were a gift of R. T. Hay (University of Dundee, UK). The deletion mutants LexA-UL44 1-100 , LexA-UL44 1-200 , LexA-UL44 1-300 , LexA-UL44 1-350 , LexA-UL44 1-390 , and LexA-UL44 1-420 were generated by PCR amplification of plasmid pBTMK-UL44 with appropriate primers (Table S1 ). The LexA-UL44 114-433 , LexA-UL44 201-433 , and LexA-UL44 313-433 constructs were generated by deleting part of the UL44 coding sequence from pBTMK-UL44 with restriction enzymes. Plasmids pDESTnFLAG-UL44(1-300) and pDESTnFLAG-UL44(313-433) were generated using the Gateway Technology (see Supplementary Material and Methods in Text S1). All other UL44 mutants and the Ubc9C93S mutant were obtained by using the QuikChange mutagenesis kit (Stratagene) with primers containing appropriate nucleotide change(s). More details on plasmid construction and mutagenesis are given in Supplementary Material and Methods in Text S1. The sequences of all primers used in this work are reported in Table S1 . All DNA sequences were confirmed by sequencing. Y2H screenings and interaction assays. Growth media and standard methods for manipulating yeast cells were as described [37] . Saccharomyces cerevisiae strain L40 was transformed [38] with the bait plasmid pBTMK-UL44 and subsequently with either of two cDNA libraries fused to GAD (see Supplementary Material and Methods in Text S1). Primary transformants were selected for growth on -His-Leu-Trp dropout plates. His + colonies were thereafter analyzed for b-galactosidase activity by filter lift experiments [39] . Double positive clones were subjected to another cycle of screening (for further details see Supplementary Material and Methods in Text S1). cDNA inserts of interactor plasmids were sequenced and analyzed with BLAST (www.ncbi/ blastn). To quantify b-gal expression, the method of Breeden and Nasmyth [40] was used. Proteins. E. coli-expressed, purified GST and GST-or 6Histagged UL44 proteins were obtained as previously described [29] , with modifications (Supplementary Material and Methods in Text S1). In some preparations, samples were treated with polymin P as described [41] to eliminate residual bacterial nucleic acids. Preparation of UL44 SUMO-modified in E. coli was accomplished as described in Supplementary Material and Methods in Text S1. GST-pulldown assays. Assays were performed using GST and GST-UL44 and in vitro-translated UL54, PB1, or Ubc9 as previously described [29] , with modifications (Supplementary Material and Methods in Text S1). In vitro transcription-translation of proteins was performed from the appropriate plasmid by using the TNT T7 coupled transcription-translation system (Promega). The translation products were labeled with [ 35 S]methionine (Amersham Pharmacia Biotech). In vitro sumoylation assays. The assays to test in vitro sumoylation of UL44 by SUMO-1, -2, and -3 were performed using purified wild-type and mutant 6His-UL44 or GST-UL44 fusion proteins and the SUMOlink SUMO-1 kit from Active Motif or the SUMOylation Kit from Enzo Life Science according to the manufacturer's suggestions. In some experiments, double stranded (ds) DNA (e.g., activated calf thymus DNA from Amersham Pharmacia Biotec or salmon testes DNA from Sigma) or singlestranded (ss) DNA (e.g., single-stranded calf lung DNA, from Crinos, Como, Italy) was added to the reaction mixture at a final concentration of 500 nM. Human foreskin fibroblasts (HFF; from the American Type Culture Collection [ATCC]), HeLa (from ATCC), eco Phoenix (a generous gift from G. P. Nolan, Stanford, USA; [42] ), and Human Embryonic Kidney 293T (HEK 293T; from ATCC) cells were maintained in Dulbecco's modified Eagle's medium (DMEM, Life Biotechnologies) supplemented with 10% fetal calf serum (FCS), 100 U/ml penicillin and 100 mg/ml streptomycin (P/S). COS-1 cells (from ATCC) were maintained in DMEM with 5% FCS and P/S. The U373-SUMO-1 cell line, which constitutively expresses FLAG-tagged SUMO-1 [43] , and the control U373-Neo cell line, stably transfected with an empty vector carrying a Neomycin resistance marker [44] , were kindly provided by G. S. Hayward (Johns Hopkins University School of Medicine, Baltimore, USA) and were maintained in medium containing 0.5 mg/ml Neomycin (G418, Gibco-BRL). HCMV strain AD169 was purchased from ATCC. To analyze UL44 sumoylation in transient expression assays, HeLa or Phoenix cells were transfected for 48 h with appropriate plasmids using the calcium phosphate precipitation method. For analysis of UL44 sumoylation during HCMV replication, HFF, U373-Neo, and U373-SUMO-1 cells were mock-infected or infected with HCMV at a multiplicity of infection (MOI) of 5 PFU/cell. Cells were harvested at different time points after infection and analyzed by western blotting as described below. Transfected or infected cells were lysed in an appropriate volume of buffer I (5% SDS, 0.15 M Tris-HCl pH 6.8, 30% glycerol) diluted 1:3 in buffer II (25 mM Tris-HCl pH 8, 50 mM NaCl, 0.5% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with Complete protease inhibitors (Roche Molecular Biochemicals) and 5 mM N-ethylmaleimide (NEM). Lysates were then incubated on ice for 20 minutes and boiled at 95uC for 10 min. Proteins were separated by SDS-PAGE, electroblotted onto a polyvinylidene fluoride membrane (Bio-Rad), and analyzed by western blotting with indicated antibodies (details are given in Supplementary Material and Methods in Text S1). For immunoprecipitation analysis, lysates were diluted 1:5 in E1A buffer (50 mM Hepes pH 7.5, 250 mM NaCl, 0.1% NP-40) supplemented with Complete protease inhibitors and 5 mM NEM. Immunoprecipitation was performed with 2-3 mg of total lysate using a ratio of 3-7 mg antibody/mg of total proteins (see Supplementary Material and Methods in Text S1 for details on antibodies) and protein A-Sepharose beads. For co-immunoprecipitation analysis, cells were lysed in E1A buffer supplemented with Complete protease inhibitors and 5 mM NEM, and successively co-immunoprecipitations were performed with 1.5-5 mg of total lysate and 50 ml of 50% slurry of anti-FLAG-M2-Agarose beads (Sigma). Mass spectrometric identification of sumoylated lysine residues within UL44 was performed after in-gel-digestion of E. coliexpressed and SUMO-modified UL44 with endoproteinase Trypsin. Extracted peptides were analyzed by LC-MSMS on an Orbitrap Velos (ThermoFisherScientific) exactly under the conditions described in Hsiao et al. [45] . Data analysis was performed by the use of software ''ChopNSpice'' [45] in combination with MASCOT as search engine. See Supplementary Material and Methods in Text S1 for details. For confocal laser-scanning microscopy (CLSM) analysis, COS-1 were transfected using the Arrest-IN TM (Biosystems) reagent, according to the manufacturer's recommendations. At 24 h posttransfection, cells were fixed with 4% paraformaldehyde. Cells were imaged using a Leica TCS-SP2 confocal microscope equipped with a 636 oil immersion objective. For analysis of UL44 intranuclear localization in HCMVinfected U373-SUMO-1 and U373-Neo cell lines, cells were seeded at 2.5 6 10 5 /well on glass coverslips in 6-well plates and allowed to attach. The next day, cells were infected with HCMV AD169 at an MOI of 1 or of 5 PFU/cell. Cells were fixed in 4% paraformaldehyde in PBS for 15 min at room temperature, and then permeabilized with acetone for 2 min at -20uC. After washing extensively with PBS, cells were incubated first with 4% FBS in PBS for 1 h at room temperature and then with a primary mouse monoclonal antibody against UL44 (10-C50, Fitzgerald Industries International) at a dilution of 1:100 in FBS 4% in PBS for 1 h at 37uC. Cells were then washed extensively with 4% FBS in PBS and incubated with a secondary goat anti-mouse fluorescein-conjugated antibody (Ig-FITC, Chemicon International) at a dilution of 1:1000 for 1 h at 37uC. Cells were successively washed with PBS and mounted in 70% glycerol in PBS. For better visualization, cells were counterstained with Evans Blue and analyzed also for red fluorescence. CLMS analysis was then performed as described above. To analyze the effects of SUMO-1 overexpression on viral DNA synthesis, U373-Neo and U373-SUMO-1 cells transduced with either shUbc9 or non-silencing lentiviral particles (see below) or non-transduced, were seeded at a density of 5610 4 per well in 24well plates. The next day, cells were infected with HCMV AD169 at an MOI of 1 PFU/cell. At 72 h post-infection (p.i.), cells were collected and total DNA was extracted using the QiAmp DNA Extraction Kit (Qiagen). The levels of viral DNA were then determined by qPCR and normalized to the cellular b-globin gene copies as described [46] . To analyze the effects of SUMO-1 overexpression on virus production, virus yield assays were performed as described previously [47] , with some modifications. Briefly, U373-Neo and U373-SUMO-1 cells transduced with either shUbc9 or nonsilencing lentiviral particles (see below) or non-transduced, 5 610 4 cells per well were seeded in 24-well plates, incubated overnight, and infected with HCMV AD169 at an MOI of 1. At 120 h p.i., cells were subjected to one cycle of freezing and thawing, and titers were determined by transferring 100-ml aliquots from each of the wells to a fresh 96-well monolayer culture of HFF cells followed by 1:5 serial dilution across the plate. After incubation at 37uC for 7 days, cell monolayers were stained with crystal violet and plaques were counted. Lentiviral particles were produced by transient transfection of HEK 293T cells with packaging plasmids helper D8.9 (Addgene) and helper Ampho (kindly provided by the Tissue Culture Facility at the IEO, Milan, Italy) and either pTRIPZ-shUBC9 vector (Open Biosystems) or non-silencing pTRIPZ control vector (Open Biosystems) by the calcium phosphate precipitation method. Supernatants were collected at 48 h post-transfection and used for the transduction of U373-SUMO-1 and U373-Neo target cells in the presence of polybrene (8 mg/ml, Sigma). Selection was done in puromycin (0.5 mg/ml, Invitrogen) for two weeks prior to 4-day doxycycline (1 mg/ml, Sigma) induction to obtain Ubc9 silencing. Ubc9-knocked-down cells were screened for Red Fluorescent Protein expression and used for further experiments. To identify cellular proteins that interact with UL44, Y2H screens were carried out with a bait consisting of full-length UL44 protein (amino acids 1-433) fused to the E. coli LexA protein. Control experiments demonstrated that the LexA-UL44 protein did not activate expression of either HIS3 or lacZ reporter gene by itself and, as expected [9, 14] , could both interact with UL54 and dimerize (Table 1 and data not shown). Thus, this bait was used to screen two different cellular cDNA libraries fused to S. cerevisiae GAL4 activation domain (GAD), one derived from human B lymphocytes [48] and the other from promyelocytic HL-60 cells [49] . A total of 167 or 85 colonies, respectively, were positive for both HIS3 and lacZ reporter genes. Plasmids encoding putative interactors of UL44 were isolated from double-positive clones and retransformed into yeasts expressing LexA-UL44 in order to confirm the interaction. The 28 and 13 positive clones after this retransformation, respectively, were sequenced. In total, from the two screenings we identified 7 cellular UL44binding proteins; here we report the identification of human Ubc9 as a specific interaction partner of UL44. In particular, 15 out of 28 clones of B lymphocytes library and 6 out of 13 clones of the HL-60 cells library contained the whole Ubc9 coding sequence, plus 59 and 39 untranslated regions which varied in length in different clones. By co-transformation experiments with each individual interactor clone expressing Ubc9 and the LexA vector, it was excluded that Ubc9 could activate the reporter genes in the absence of the bait protein (Table 1) . Ubc9 specifically interacted with UL44 in Y2H assays also in the reverse combination, i.e., with Ubc9 fused to LexA and UL44 fused to GAD (Table 1 ). In addition, in quantitative assays the b-galactosidase activity of yeasts expressing both UL44 and Ubc9 turned out to be comparable to that of yeasts expressing UL44 and UL54 (Table 1) , which served as a positive control. Having identified Ubc9 as a potential interaction partner of UL44 in Y2H screens, we wished to confirm their physical interaction by an independent experimental approach. For this purpose, pulldown assays with a purified GST-UL44 protein and in vitro-translated, 35 S-labeled Ubc9 were performed. As positive and negative controls, we also assayed the interaction between UL44 and in vitro-translated UL54 or the PB1 subunit of influenza A virus RNA polymerase. As expected, we could detect the interaction of UL54 [14, 29] , but not of PB1, with GST-UL44 (Fig. 1A) . Consistent with the Y2H results, Ubc9 specifically associated with GST-UL44, while no interaction with GST was observed (Fig. 1A) . Since our data suggested that UL44 can interact with Ubc9 both in vitro and in yeast cells, we sought to examine whether UL44 and Ubc9 could also co-localize in mammalian cells. To this end, aggregates distributed on a diffuse background fluorescence throughout the nucleus (Fig. 1B) . As previously reported [50] , when expressed alone Ubc9 exhibited a nuclear punctate pattern (Fig. 1B) . Upon co-expression of GFP-UL44 and DsRed2-Ubc9, co-localization of UL44 we analyzed the subcellular localization of GFP-UL44 when transiently expressed either alone or in the presence of DsRed2-Ubc9. As a negative control, we also expressed the DsRed2-UL53 fusion protein, which localizes to the cell nucleus but does not interact with UL44 [32] . In GFP-UL44-transfected cells, UL44 localized in a large number of discrete nuclear with Ubc9 was observed (Fig. 1B) . UL44 also colocalized with a catalytically impaired Ubc9 mutant (Ubc9C93S) [51] , but not with UL53 [32] . In addition, co-immunoprecipitation experiments confirmed that UL44 could physically interact with endogenous Ubc9 in mammalian cells. The specificity of the interaction was confirmed by the inability of UL44 to coimmunoprecipitate with Cyclin D1 (CycD1; Fig. 1C ). To further explore the interaction between UL44 and Ubc9, we sought to map the domain of UL44 that interacts with Ubc9. To this aim, several N-and C-terminal deletion mutants of UL44 fused to LexA were generated and tested for the ability to interact with GAD-Ubc9 by Y2H assays. Control western blot experiments with an anti-LexA antibody evidenced protein bands of the expected molecular mass for all mutants (data not shown). As shown in Fig. 1D , the truncated protein UL44 1-300 , lacking most of the C-terminal disordered region of UL44, exhibited interaction with Ubc9. Further C-terminal truncation of UL44 revealed that a protein fragment corresponding to the first 200 amino acids of UL44 (UL44 ) was still capable of interacting with Ubc9 ( Fig. 1D ). In contrast, the N-terminal 100 residues of UL44 (UL44 1-100 ) exhibited no interaction with Ubc9 (Fig. 1D ). Although unfolding of the UL44 1-100 mutant protein cannot be excluded, these results suggested that this region of UL44 may not contain sequences important and/or sufficient for Ubc9 binding. Therefore, we analyzed the effects of N-terminal truncations. Deletion of the N-terminal 113 residues of UL44 (UL44 114-433 ) did not impair the ability of UL44 to bind Ubc9. Similarly, the UL44 201-433 mutant, lacking the N-terminal 200 amino acids, interacted with Ubc9. Interestingly, a mutant that only expresses the C-terminal 121 residues of UL44 (UL44 313-433 ) still retained the ability to bind Ubc9 (Fig. 1D) . Control Y2H experiments showed that none of the truncated UL44 proteins was able to activate transcription by itself (data not shown). Mapping studies in mammalian cells expressing UL44 deletion mutants showed that both the UL44 1-300 and UL44 313-433 mutants could immunoprecipitate endogenous Ubc9, similarly to wild-type UL44 (Fig. S1 ). Thus, our results suggest that UL44 contains two domains capable of independently binding to Ubc9, located at the N-and Cterminus of the protein (likely within residues 1-200 and 313-433, respectively). The observation that UL44 interacts with the SUMO-conjugating enzyme Ubc9 prompted us to investigate the possibility that UL44 may be sumoylated. This hypothesis was first tested in a cellfree system by incubating a purified 6His-UL44 fusion protein with purified Aos1/Uba2 (E1), Ubc9 (E2), and SUMO-1. As a control, sumoylation of human p53 was also examined by the same assay. As expected [52, 53] , p53 was readily modified to give mainly a single mono-sumoylated product that reacted with both anti-p53 ( Fig. 2A , left bottom panel) and anti-SUMO-1 ( Fig. 2A , right bottom panel) antibodies (the bands appearing at high molecular weight, which are particularly visible in the anti-SUMO-1 panel, correspond to SUMO-E1 and -E2 enzymes conjugates). In a western blot analysis with an anti-UL44 antibody ( Fig. 2A , left top panel), three main slower migrating forms of UL44 were observed (lane 3). The appearance of the ,65, 80, and 95-kDa forms of UL44 was strictly dependent on the presence of SUMO-1, as the substitution of wild-type SUMO-1 with a mutant form which bears the Gly97-to-Ala change (SUMO-1 mut) and hence cannot be attached to target proteins, eliminated formation of these products ( Fig. 2A, left top panel, lane 4 ). In addition, SUMO-1 modification was abolished if either SUMO-1, Aos1/ Uba2, or Ubc9 was omitted from the reaction (not shown). A western blot analysis with an anti-SUMO-1 antibody confirmed that the slower migrating UL44 bands contained SUMO-1 ( Fig. 2A, right top panel) . Taken together, the above results established that UL44 is a substrate for in vitro SUMO-1 conjugation. Having shown that UL44 can be modified by SUMO-1, we wondered whether it could be conjugated also to SUMO-2 and SUMO-3. Thus, the in vitro sumoylation system was applied to purified 6His-UL44 in the presence of sumoylation enzyme components and activated forms of SUMO-2 or SUMO-3. In these assays, the RanGTPase-activating protein RanGAP1 [54] was used as a positive control and as previously observed [55] , was modified in the presence of SUMO-2 and SUMO-3 (data not shown). As shown in Fig. 2B , UL44 was also modified by either of the two SUMO peptides. To verify whether UL44 could also be sumoylated in mammalian cells, we expressed UL44-FLAG and HA-SUMO-1 fusion proteins, in the presence of either wild-type Ubc9 or the catalytically impaired mutant Ubc9C93S in Phoenix cells and analyzed cell lysates by western blotting with anti-FLAG, anti-HA, and anti-Ubc9 antibodies. As expected, several bands corresponding to SUMO-1-conjugated proteins that reacted with an anti-HA antibody were detected upon co-expression of wild-type Ubc9, while less SUMO-1 conjugation was observed in the presence of the Ubc9C93S mutant (Fig. 3A) . This was not due to differences in expression of the two Ubc9 variants, as evidenced by western blot analysis with the anti-Ubc9 antibody (Fig. 3A) . When UL44-FLAG was expressed in the absence of HA-SUMO-1 and Ubc9, a single band with an apparent molecular mass of ,50 kDa was detected using the anti-FLAG antibody. In contrast, slower migrating bands similar to those observed in in vitro sumoylated products ( Fig. 2A) were observed upon co-expression of UL44-FLAG with HA-SUMO-1 and wild-type Ubc9 (Fig. 3A) . These bands were significantly reduced in the presence of the Ubc9C93S mutant (Fig. 3A) , demonstrating that they were dependent on the catalytic activity of Ubc9, with the residual slower migrating bands due to the activity of wild-type endogenous Ubc9. To confirm that the observed products were indeed sumoylated forms of UL44, we expressed UL44-FLAG either in the absence or in the presence of HA-SUMO-1 and Ubc9, immunoprecipitated UL44-FLAG using an anti-FLAG antibody and analyzed the immunoprecipitated proteins by western blot. As expected, three slower migrating bands were detected by the anti-FLAG antibody from lysates of cells expressing UL44-FLAG in the presence of both HA-SUMO-1 and Ubc9 (Fig. 3B, left panel) . The ,50-kDa non-sumoylated form of UL44 could be immunoprecipitated both from cells expressing UL44-FLAG alone and from cells expressing UL44-FLAG together with HA-SUMO-1 and Ubc9, but not from Table 1 . UL44 interacts with human Ubc9 in yeast two-hybrid assays. LexA-UL44 / 2(,1) UL44, UL54, and Ubc9 proteins were fused to the C-terminus of LexA protein and/or of GAL4 activation domain (GAD). Fusion proteins were then assayed for interaction by qualitative b-galactosidase (b-gal) filter assays and by quantitative b-gal liquid assays. Ubc9, but not in immunoprecipitates obtained from cells expressing only UL44-FLAG (Fig. 3B, right panel) . Similar results were obtained in HeLa cells (Fig. S2) . Altogether, these results demonstrate that Ubc9 can mediate the conjugation of SUMO-1 to UL44 in mammalian cells. Moreover, sumoylation of UL44 in mammalian cells by both SUMO-2 and SUMO-3 could also be detected (Fig. S3) . Since sumoylation mainly occurs in the nucleus of mammalian cells [56] and UL44 is translocated to the host cell nucleus during HCMV infection [11] , we decided to investigate whether nuclear localization might be a prerequisite for conjugation of SUMO-1 to UL44. We therefore analyzed the ability of UL44bDNLS, a derivative of UL44 bearing point mutations within the nuclear localization signal (NLS, 425 PNTKKQK 431 ) which prevent UL44 nuclear accumulation [11] , to be modified by SUMO-1. Interestingly, mutations of UL44 NLS impaired the sumoylation of a V5-UL44 fusion protein (V5-UL44DNLS, Fig. 3C ), suggesting that SUMO-1 modification of UL44 most likely occurs into the nucleus or during nuclear import. We next sought to identify the SUMO-1 acceptor sites of UL44. A prediction analysis with the SUMOplot program (Abgent) identified seven residues in UL44 with a certain probability to be sumoylated: K73, K172, K224, K339, K371, K410, and K431. To test whether one or more of these lysines could be a SUMO-1 acceptor site, each of the candidate residues was conservatively mutated to arginine, both individually and in combination, and the mutant proteins were tested for in vitro sumoylation. None of the single point mutants exhibited a consistently altered SUMO-1 modification pattern as compared to wild-type UL44 (Fig. S4A ). Mutants carrying two or three K/R substitutions also showed a SUMO-1 modification pattern identical to that of wild-type UL44 ( Fig. S4A and data not shown) . Similar results were obtained when in vitro sumoylation reactions with SUMO-2/23 were performed (data not shown). We decided to also test the ability of these K/R mutants to be modified by SUMO-1 in mammalian cells. Consistent with the in vitro data, none of the tested mutations significantly affected the ability of UL44 to undergo SUMO-1 conjugation ( Fig. S4B and data not shown) . These results suggested that UL44 might possess multiple lysine residues that could alternatively serve as SUMO-1 acceptors, as reported for other proteins [57, 58] , and/or that UL44 might be sumoylated on lysine residues other than those predicted by SUMOplot. UL44 contains 31 lysines, most of which are solventexposed in the crystal structure [9] and therefore potentially accessible to SUMO molecules, which makes it difficult to identify the target lysines by mutational approaches. Therefore, we attempted to map the sites of UL44 where SUMO-1 is conjugated by mass spectrometry analysis. To this end, we expressed UL44 in an E. coli expression/modification system that produces SUMOconjugated proteins [30] . The 6His-tagged UL44 construct was co-expressed in E. coli with the pTE1E2S1 plasmid, which contains a linear fusion of genes for E1 and E2 enzymes and SUMO-1 under the control of an IPTG-inducible promoter [30] . To confirm sumoylation of UL44 with this system, UL44 was purified from the bacterial cultures expressing UL44 alone or in combination with the SUMO conjugation system and analyzed by western blotting with both the anti-UL44 (Fig. S5, left panel) and the anti-SUMO-1 (Fig. S5, right panel) antibody. Shifted bands with an apparent molecular mass of ,65, 80, and 95 kDa, similar to those detected in in vitro reactions ( Fig. 2A) and in cells (Fig. 3) , were observed in UL44 purified from bacteria cotransformed with pTE1E2S1, but not in UL44 purified from bacteria expressing only UL44 (Fig. S5) . Then, to identify lysine acceptor sites by mass spectrometry, sumoylated UL44 was separated by SDS-PAGE, in-gel-digested with trypsin and peptides were analyzed by LC-MSMS. Upon application of the software ''ChopNSpice'' for database search, we identified 16 sumoylation sites in UL44, including the predicted sites K172, K339, K371, K410, and K431. Table 2 summarizes the UL44 peptides that have been found to be sumoylated. The corresponding MS spectra are shown in Fig. S6 . These results indicated that UL44 possesses multiple SUMO target lysines that are located throughout the protein, in accordance with the observation that Ubc9 could bind to both N-and C-terminal portions of UL44 ( Fig. 1D and Fig. S1 ). Mutagenesis of several of these lysines in combination caused a strong decrease of UL44 expression (data not shown), likely due to protein misfolding and/or instability, making impossible to analyze the sumoylation state of the mutated protein and to compare it to that of wild-type UL44. UL44 sumoylation is stimulated by binding to DNA. As mentioned above, UL44 possesses a structural fold similar to that of the eukaryotic processivity factor PCNA [9] . In addition, like UL44, PCNA is SUMO-conjugated and its sumoylation involves both a consensus and a non-consensus site [4] . Since it has been shown that PCNA needs DNA to be sumoylated efficiently [59] , we wished to investigate whether UL44 might behave similarly and its sumoylation could be stimulated by the presence of DNA. Thus, we performed in vitro sumoylation experiments in the absence and in the presence of DNA using as a substrate a purified 6His-UL44 fusion protein treated with polymin P to eliminate residual bacterial nucleic acids. In a western blot analysis with an anti-UL44 antibody (Fig. 4A, left panel) , only a faint band corresponding to mono-sumoylated UL44 was observed in the absence of DNA (lane 2). Upon addition of dsDNA (e.g., activated calf thymus DNA) to the reaction mixture, a marked increase of the mono-sumoylated product and the appearance of bi-and trisumoylated forms of UL44 were observed (Fig. 4A , lane 4 of left panel). A western blot analysis with an anti-SUMO-1 antibody confirmed that these bands indeed contained SUMO-1 (data not shown). Similar results were obtained when GST-UL44 was used as a substrate (Fig. S7) . Thus, like PCNA sumoylation, UL44 sumoylation is strongly stimulated by the presence of DNA. To investigate whether the nature of DNA could influence the stimulation of UL44 sumoylation, we also performed in vitro sumoylation reactions in the presence of different DNA substrates. Similar stimulation levels were obtained when different dsDNAs were added, regardless of their sequence (data not shown), in keeping with previous observations that UL44 binds DNA in a sequence-independent manner [7] . In contrast, consistently less stimulation of UL44 sumoylation was observed in the presence of ssDNA (Fig. 4A, right panel) , for which UL44 has been shown to possess an apparent affinity lower than for dsDNA [7] . This suggested that UL44 sumoylation could depend on binding to To analyze UL44 sumoylation in vitro, purified 6His-UL44 was incubated in the absence or the presence of sumoylation enzymes and either wild-type SUMO-1 (SUMO-1 wt) or a mutant form of SUMO-1 (SUMO-1 mut) which cannot be covalently linked to substrates. The reaction products were analyzed by western blotting with anti-UL44 and anti-SUMO-1 antibodies. As a positive control, in vitro sumoylation of p53 was also analyzed. (B) Purified 6His-UL44 was incubated in the absence or the presence of sumoylation enzymes and either SUMO-2 or SUMO-3 and analyzed by western blotting with anti-UL44 and anti-SUMO-2/23 antibodies. For all panels, the arrowhead indicates the unmodified form of UL44 or p53 and the asterisks indicate the respective sumoylated forms. doi:10.1371/journal.pone.0049630.g002 DNA. To test this hypothesis, we analyzed the ability of two UL44 mutants, FLAG-UL44Dloop and FLAG-UL44L86A/L87A, which are defective for DNA binding [36] , to be modified by SUMO-1 in mammalian cells. FLAG-UL44Dloop contains three point mutations within a UL44 flexible loop ( 163 HTRVKRNVKKAP 174 ) involved in UL44-DNA interaction [8, 9] , and is therefore impaired in DNA binding [36] . FLAG-UL44L86A/L87A carries two point mutations preventing dimerization of UL44 and strongly impairing the UL44-DNA interaction both in vitro and in vivo [9, 36, 60] . Both FLAG-UL44Dloop (Fig. 4B) and FLAG-UL44L86A/L87A (Fig. 4C) , when coexpressed with HA-SUMO-1 and Ubc9, exhibited strongly reduced sumoylation levels when compared to FLAG-UL44. Importantly, co-immunoprecipitation experiments demonstrated that the reduced sumoylation of FLAG-UL44Dloop and FLAG-UL44L86A/L87A was not due to an impairment of binding of the mutant proteins to Ubc9, since the two mutants precipitated with endogenous Ubc9 at levels comparable to those of the wild-type protein (Fig. 4D ). In addition, the presence of DNA did not stimulate the sumoylation of the UL44Dloop or UL44L86A/L87A mutants in in vitro reactions (Fig. 4E) . Finally, since the UL44Dloop mutant contains a substitution (K167N) involving a potential SUMO target lysine, we wished to exclude the possibility that the reduced sumoylation levels of FLAG-UL44Dloop might be due to alteration of a putative SUMO acceptor site rather than an impairment of DNA binding. To this end, the K167 residue was conservatively mutated to arginine and the mutant protein was tested for in vitro sumoylation in the presence of DNA. The K167R mutant showed a SUMO-1 modification pattern identical to that of wild-type UL44 (Fig. 4F) . Altogether, these results suggest that UL44 is preferentially modified by SUMO-1 when it is bound to DNA as a dimer. Having demonstrated that UL44 is sumoylated by Ubc9 in vitro and in transfected cells, we sought to investigate whether a similar modification also occurs naturally in HCMV-infected cells. Protein lysates of HFFs infected with HCMV and collected at different times post-infection (p.i.) were analyzed by western blotting with an anti-UL44 antibody. Two main bands of 65 and 80 kDa were observed above the primary UL44 band of 50 kDa during the whole time course of lytic infection, being detectable already at 24 h p.i. (Fig. 5A) . A third band of ,95 kDa was also visible from 48 h p.i. These subforms were similar in electrophoretic mobility to the UL44 bands covalently modified by SUMO-1 in transfected cells (Fig. 3A) . A densitometric analysis of the protein bands (Fig. 5B ) revealed that the relative amount of the sumoylated forms increased during the course of HCMV infection, becoming ,50% of total UL44 protein at 120 h p.i. To confirm that the slower migrating forms indeed represent UL44 molecules conjugated to SUMO-1, lysates from HCMVinfected cells at 120 h p.i. or from mock-infected cells were immunoprecipitated with an antibody against UL44. Subsequent- ly, the immunocomplexes were analyzed by western blotting with an anti-SUMO-1 antibody. Three main bands of the expected molecular mass (,65, 80, and 95 kDa) were recognized in the anti-UL44 immunoprecipitate (Fig. 5C, left panel) . Finally, the same immunocomplexes were analyzed by western blotting with the anti-UL44 antibody to demonstrate that the SUMO-1 crossreactive proteins were indeed modified UL44 forms (Fig. 5C, right panel) . Thus, these results clearly indicate that UL44 is covalently modified by SUMO-1 in HCMV-infected cells. Considering the difficulties in expressing a UL44 mutant completely impaired in sumoylation, whose activities could be compared to that of wild-type UL44, to gain some insight on the functional role of UL44 sumoylation in the context of HCMV replication we sought to undertake a different approach. We overexpressed SUMO-1 in virus-infected cells and analyzed the effects on the intracellular distribution of UL44, as the targeting to specific subcellular domains is one of most common biological effect exerted by the conjugation of SUMO to a substrate protein. It has been previously shown that during HCMV replication UL44 localizes to large globular intranuclear structures that correspond to viral DNA replication compartments [61, 62, 63] . A U373-MG cell line that constitutively overexpresses FLAG-SUMO-1 and control U373-Neo cells were mock-infected or infected with HCMV at an MOI of 1 or of 5 PFU/cell and the intracellular localization of UL44 was successively analyzed by indirect immunofluorescence with an anti-UL44 antibody. Control western blotting experiments (Fig. S8A ) confirmed that UL44 is sumoylated in the HCMV-infected U373 cells. In fact, slower migrating bands of the expected molecular mass and similar to the UL44 sumoylated forms observed in infected HFFs (Fig. 5A) were detected. Furthermore, as expected, they increased upon SUMO-1 overexpression. In immunofluorescence assays, the nuclei of mock-infected cells were oval-shaped with no anti-UL44 staining (Fig. 6A, upper panels) . Control HCMV-infected U373-Neo cells showed deformed nuclei, many of which exhibited a kidney shape (Fig. 6A, upper panels) . Indeed, it has been observed that infection by HCMV causes this kind of distortions in nuclear shape [64, 65] . Moreover, UL44 showed a globular fluorescent pattern consistent with previously described viral replication compartments in HCMV-infected cells [61, 66] . In contrast, in HCMV-infected U373-SUMO-1 cells UL44 staining was more distributed throughout the nucleus and, especially at the lower MOI (MOI 1), also failed to coalesce into any recognizable globular structures (Fig. 6A, upper panels) . Therefore, overexpression of SUMO-1 during HCMV replication appears to alter the intranuclear distribution of UL44, likely leading to significantly decreased localization of UL44 in viral DNA replication compartments. Importantly, the intranuclear distribution of another HCMV protein localizing to the replication compartments, i.e. UL57, the single-stranded DNA (ssDNA)-binding protein [61, 67] , appeared not to change upon SUMO-1 overexpression (Fig. 6A, lower panels) . To investigate whether sumoylation is indeed involved in the altered intranuclear distribution of UL44, an RNAi approach was employed to suppress the sumoylation system of the cells by silencing Ubc9 since Ubc9 is the only unique and essential enzyme in the SUMO-conjugating pathway [21] . U373-Neo and U373-SUMO-1 cells were transduced with a lentivirus expressing shUbc9, followed by selection with puromycin and induction with doxycyline to establish Ubc9-knocked-down cell lines (U373-Neo shUbc9 and U373-SUMO-1 shUbc9, respectively). As a control, U373-SUMO-1 cells were also transduced with a lentivirus expressing a non-silencing shRNA sequence (U373-SUMO-1 NS). As shown in Figure 6B , Ubc9 was almost completely silenced in cells infected with the shUbc9 lentivirus (U373-SUMO-1 shUbc9) as compared to the cells transduced with the control lentivirus (U373-SUMO-1 NS) and to nontransduced cells (U373-Neo and U373-SUMO-1). To examine the effect of Ubc9 knock-down on intranuclear distribution of UL44, these cell lines were infected with HCMV. As shown in Fig. 6A (upper panels) , upon doxycycline induction, the nuclear UL44 staining in Ubc9-knocked-down U373-SUMO-1 cells was similar to that of U373 cells not overexpressing SUMO-1 (U373-Neo and U373-Neo shUbc9; Fig. 6A , upper panels, and Fig. S8B ). In contrast, the U373-SUMO-1 cells transduced with the non-silencing lentivirus (U373-SUMO-1 NS, Fig. 6A , upper panels) retained an altered intranuclear distribution of UL44 similar to that observed in the non-transduced cells (U373-SUMO-1) or in the shUbc9-transduced U373-SUMO-1 cells with no doxycycline induction (data not shown). Altogether, these results established that the altered intranuclear distribution of UL44 in HCMV-infected cells upon SUMO-1 overexpression depends on Ubc9-mediated sumoylation. These observations raised the question whether the altered intranuclear distribution of UL44 observed upon SUMO-1 overexpression affected the viral replication efficiency. To examine the effects of SUMO-1 overexpression on viral DNA replication, U373-Neo and U373-SUMO-1 cells were infected with HCMV at an MOI of 1 and viral DNA levels were measured by quantitative real-time PCR. Viral DNA production from U373-SUMO-1 cells was ,two-fold higher than that from the control cells (U373-Neo and U373-Neo shUbc9; Fig. 6C ). This increase was not observed in Ubc9-knocked down cells (U373-SUMO-1 shUbc9), while the U373-SUMO-1 cells transduced with the non-silencing lentivirus (U373-SUMO-1 NS) exhibited augmented viral DNA levels like the non-transduced U373-SUMO-1 cells (Fig. 6C) . We also examined the effects of SUMO-1 overexpression on virus yield. The titers of viral particles produced from nontransduced U373-SUMO-1 and from transduced U373-SUMO-1 shUbc9 and U373-SUMO-1 NS cells after infection with HCMV at an MOI of 1 were determined and compared to those produced from infected U373-Neo and U373-Neo shUbc9 control cells. A 2-3-fold increase in viral progeny titers was observed in U373-SUMO-1 and U373-SUMO-1 NS with respect to U373-Neo, while the U373-SUMO-1 shUbc9 cells exhibited yields of Ubc9, and anti-vinculin antibodies (left panel). Cell lysates were incubated with anti-FLAG-M2-Agarose beads and the immunoprecipitated samples were analyzed by western blotting with anti-UL44 and anti-Ubc9 antibodies (right panel). (E) The sumoylation in vitro of wild-type 6His-UL44 and mutant 6His-UL44Dloop and 6His-UL44L86A/L87A proteins was carried out as in (A) and analyzed by western blotting with an anti-UL44 antibody. (F) The sumoylation in vitro of a UL44 mutant bearing the K167R substitution in the flexible loop of UL44 involved in DNA binding was carried out in the presence of DNA and compared to that of wild-type UL44. For all panels, the arrowhead indicates the unmodified form of UL44 or free SUMO-1 and the asterisks indicate the sumoylated forms. doi:10.1371/journal.pone.0049630.g004 infectious virus similar to those of the U373-Neo and U373-Neo shUbc9 cells (Fig. 6D) . Thus, the altered intranuclear distribution of UL44 upon SUMO-1 overexpression appears not to compromise HCMV replication, but conversely, SUMO-1 overexpression causes a positive effect on virus production. In this study we report that UL44, a viral ortholog of PCNA, is sumoylated on multiple lysines by the cellular factor Ubc9. Importantly, a consistent portion of UL44 is SUMO-modified in HCMV-infected human cells, resulting in ,50% of the protein being modified at late times during virus replication. From a structural point and functional of view, UL44 and PCNA share blotting with anti-SUMO-1 (left panel) and anti-UL44 (right panel) antibodies. For all panels, the arrowhead indicates the unmodified form of UL44, the arrow indicates the immunoglobulin G heavy chain (IgG hc) and the asterisks indicate the sumoylated UL44 forms. doi:10.1371/journal.pone.0049630.g005 some remarkable similarities and some differences. Monomers of UL44 and PCNA are structurally very similar, despite having extremely different primary sequences [2, 9] . However, while PCNA forms toroidal-homotrimers, UL44 binds to dsDNA as a head-to-head homodimer [7, 9] . In addition, PCNA must be loaded onto DNA in an ATP-dependent process by so-called clamp loaders [68] ; in contrast, UL44 directly binds DNA without the need for ATP hydrolysis or accessory proteins [7, 14, 18] . Similarities and differences also emerge from the comparison of the sumoylation processes of UL44 and PCNA. The most striking similarity is the DNA-dependence of such post-translational modification: in the case of PCNA, clamp loading rather than the mere presence of DNA was shown to be important for stimulation, implying a change in the properties of PCNA upon loading that enhances its capacity to be sumoylated [59] . This could also be the case for UL44: in fact, point mutations preventing DNA binding [8, 9, 36, 60] strongly impaired UL44 sumoylation in cells (Fig. 4C) and abolished the ability of dsDNA to stimulate UL44 sumoylation in vitro (Fig. 4E) . Furthermore, the ability of DNA to stimulate UL44 sumoylation appears to correlate with the efficiency of the UL44-DNA interaction. In fact, addition of dsDNA -for which UL44 has a ,3-to 8-fold higher affinity than for ssDNA [7] -caused a much stronger increase of UL44 sumoylation as compared to ssDNA (Fig. 4A) . In this context, it is important to note that point mutations impairing DNA binding and sumoylation in cells did not compromise the UL44-Ubc9 interaction (Fig. 4D) , suggesting that binding to DNA does not promote UL44 sumoylation by facilitating its binding to Ubc9. Another similarity between UL44 and PCNA sumoylation is that both proteins are modified on canonical (K127 for PCNA, and K410 for UL44) and non-canonical residues (K164 for PCNA, and all other sumoylation sites for UL44, see Table 2 ). However, while PCNA is exclusively sumoylated at the Nterminus, both N-and C-terminal residues of UL44 can be modified. In addition, according to our MS-analysis UL44 can be alternatively sumoylated at 16 different sites, while K127 and K164 appear to be the only target sites in PCNA. This flexibility of UL44 in terms of sumoylation target sites, which is reminiscent of the ones described for Daxx and the small hepatitis delta antigen [57, 58] , is arguably the main difference from PCNA. This makes it extremely difficult to study the physiological importance of UL44 sumoylation, also because mutation of several of these lysines caused protein instability. Currently it is therefore impossible to test if, like in the case of budding yeast PCNA, SUMOmodification of UL44 also plays a role in DNA repair/ recombination [69] . In terms of functional effects, sumoylation is known to regulate protein activity and/or intracellular location [21, 22] . As for the latter, the targeting to specific subcellular domains is one of the best-characterized biological effects exerted by the conjugation of SUMO to a substrate protein. This effect is exemplified by the targeting of cellular protein RanGAP1 to the cytosolic side of the nuclear pore complex upon sumoylation [54, 70] . As a first, preliminary attempt to characterize the role of UL44's sumoylation, here we show that overexpression of SUMO-1 in the context of HCMV replication alters the intranuclear distribution of UL44 as it appears to result in a more diffuse pattern and in decreased localization of UL44 in viral DNA replication compartments (Fig. 6A) , suggesting that sumoylation of UL44 may retarget the protein to other nuclear site(s). Importantly, Ubc9 knock-down studies confirmed that sumoylation is responsible for such altered intranuclear pattern of UL44. The observation that SUMO-1 overexpression causes a positive effect on HCMV replication suggests that sumoylation of UL44 could be important for its function(s) in the context of the virus life cycle, although effects on HCMV replication mediated by sumoylation of other viral or cellular proteins cannot be excluded. However, at the moment understanding the molecular details of how SUMO alters the intranuclear distribution of UL44 is rather difficult. Most likely, sumoylation does not affect the functions of UL44 as a DNA polymerase processivity factor -the only role currently well established for UL44. In fact, several reports have shown that the UL44 protein expressed in E. coli, which is non-sumoylated, is capable of performing all known biochemical activities related to this role (e.g., [7, 14] ). In addition, our observation that UL44 sumoylation peaks at late times during virus replication, once the viral DNA replication has been accomplished (Fig. 5) , suggests that SUMO-conjugation might enable UL44 to fulfill some role(s) in HCMV replication other than that of conferring processivity to the viral DNA polymerase. A role of UL44 in late gene expression has been previously suggested [71, 72] ; however, no conclusive demonstration has been provided yet. Intriguingly, higher molecular mass forms of the Epstein-Barr virus DNA polymerase processivity factor BMRF1 compatible with sumoylated products have been observed [73] , suggesting that such post-translational modification could be a general feature of the DNA polymerase accessory subunit in herpesviruses. In addition, sumoylation of Vaccinia Virus G8R protein has been recently predicted on the basis of structural similarities to PCNA [74] , but not yet experimentally demonstrated. Thus, our findings could stimulate further studies on sumoylation of DNA polymerase subunits in other herpesviruses or, more in general, in other viral systems. Figure S5 Sumoylation of UL44 in E. coli. The pRSET44 plasmid, encoding 6His-tagged UL44, was introduced into E. coli together with the pTE1E2S1 plasmid, which expresses E1 and E2 sumoylation enzymes and SUMO-1. As a control, bacteria were also transformed only with pRSET44. The 6His-tagged UL44 was purified from bacterial cultures expressing UL44 alone or in combination with the SUMO conjugation system and analyzed by western blotting with anti-UL44 and anti-SUMO-1 antibodies. (TIF) Figure S6 Mass spectrometry analysis of sumoylation sites of UL44. MSMS analysis of tryptic peptides conjugated to a tryptic peptide of SUMO-1 (ELGMEEEDVIEVYQEQTGG or IADNHTPKELGMEEEDVIEVYQEQTGG, 1 tryptic miscleavage) derived from E. coli-expressed, sumoylated UL44. Sequence of the conjugate is listed in each table above the spectra. Y-type and b-type fragment ions are assigned in the spectra. The table below each spectrum summarizes the database search. Highlighted are the m/z values that match the fragment ions obtained from in silico fragmentation of the conjugate. Conjugated lysine residues to SUMO-1 are highlighted in red and the position in UL44 is listed. Modifications, measured (observed) m/z values, the actual mass (in Da), the charge state, and the mass error (ppm, parts per million) are listed as well in the table above each spectrum. The different colors represent the various measured and calculated fragment ions of each conjugate in the spectrum and table underneath, respectively. The question marks in some of the spectra indicate fragment ions that do not match to any calculated y-and b-type ions of the conjugate. (PDF) Figure S7 Sumoylation in vitro of a GST-UL44 fusion is stimulated by DNA. E. coli-expressed, purified GST-tagged UL44 was incubated with purified sumoylation enzymes in the absence or presence of SUMO-1 and/or DNA. The samples were analyzed by western blotting with an anti-UL44 antibody. (TIF) Figure S8 UL44 is sumoylated in HCMV-infected U373 cells. (A) U373-Neo and U373-SUMO-1 cells were either mockinfected or infected with HCMV at MOI of 5 PFU/cell for 72 h. Cell lysates were then analyzed by western blotting with an anti-UL44 antibody. The arrowhead indicates the unmodified form of UL44 and the asterisks indicate the sumoylated UL44 forms. (B) Control U373-Neo cells, and U373-Neo cells transduced with lentiviral particles expressing either a Ubc9-silencing shRNA (U373-Neo shUbc9) or a non-silencing shRNA sequence (U373-Neo NS) were mock-infected or infected with HCMV at an MOI of 5 or 1 PFU/cell. At 72 h p.i., cells were fixed and stained with a primary antibody against UL44 and successively with a secondary fluorescein-conjugated antibody (green) which contained Evans Blue to counterstain cells (red). Cell samples were then analyzed by CLSM. (TIF)

Fas-deficient mice have impaired alveolar neutrophil recruitment and decreased expression of anti-KC autoantibody:KC complexes in a model of acute lung injury

BACKGROUND: Exposure to mechanical ventilation enhances lung injury in response to various stimuli, such as bacterial endotoxin (LPS). The Fas/FasL system is a receptor ligand system that has dual pro-apoptotic and pro-inflammatory functions and has been implicated in the pathogenesis of lung injury. In this study we test the hypothesis that a functioning Fas/FasL system is required for the development of lung injury in mechanically ventilated mice. METHODS: C57BL/6 (B6) and Fas-deficient lpr mice were exposed to either intra-tracheal PBS followed by spontaneous breathing or intra-tracheal LPS followed by four hours mechanical ventilation with tidal volumes of 10 mL/kg, respiratory rate of 150 breaths per minute, inspired oxygen 0.21 and positive end expiratory pressure (PEEP) of 3 cm of water. RESULTS: Compared with the B6 mice, the lpr mice showed attenuation of the neutrophilic response as measured by decreased numbers of BAL neutrophils and lung myeloperoxidase activity. Interestingly, the B6 and lpr mice had similar concentrations of pro-inflammatory cytokines, including CXCL1 (KC), and similar measurements of permeability and apoptosis. However, the B6 mice showed greater deposition of anti-KC:KC immune complexes in the lungs, as compared with the lpr mice. CONCLUSIONS: We conclude that a functioning Fas/FasL system is required for full neutrophilic response to LPS in mechanically ventilated mice.

Acute lung injury (ALI) and its more severe form, the acute respiratory distress syndrome (ARDS) remain important clinical problems in the United States, with an incidence rate of 38.3 cases per 100,000 person-years and a mortality rate of 45% [1] . ALI/ARDS is characterized clinically by sudden respiratory failure with impaired oxygenation and non-cardiogenic pulmonary edema [2] . Pathologically ALI/ARDS is associated with an early inflammatory phase with neutrophilic alveolitis and destruction of the alveolar/capillary permeability barrier, followed by a late fibroproliferative phase with abnormal repair and collagen deposition. There are no specific treatments for ARDS and the main supportive treatment, mechanical ventilation, can be harmful to the lungs when delivered at high tidal volumes [3] . A growing body of experimental evidence suggests that in addition to the injury caused by high tidal volumes, even moderate or low tidal volumes markedly enhance injury when the lungs are exposed to pathogens or their products [4] [5] [6] . For example, mechanical ventilation synergistically enhances lung injury in response to low doses of bacterial lipopolysaccharide (LPS) and this is associated with expression of specific sets of genes that aren't expressed with LPS or ventilation alone [4, 5, 7] . A computational analysis has mapped the biological processes that are activated by the combination of mechanical ventilation and LPS [8] . One of these biological processes is apoptosis, and a known mediator of apoptosis in injured lungs is the Fas/FasL system. The Fas/FasL system is composed of the membrane surface receptor Fas and its cognate ligand, FasL. FasL exists in a membrane-bound form, and also in a soluble form that is present in the lungs during acute lung injury [9] . Binding of Fas to sFasL in alveolar epithelial cells independently activates apoptotic and inflammatory pathways that result in death of the cells but also in release of pro-inflammatory cytokines [10] . Although the Fas/ FasL system is better known for its pro-apoptotic function, the pro-inflammatory function is also important in the development of lung injury, and mice deficient in Fas (lpr) have an impaired neutrophil recruitment in response to LPS installation and bacterial infections [11] . Several lines of evidence suggest that the Fas/FasL system plays a role in ALI/ARDS. First, bioactive sFasL is present in the lungs of patients with ARDS, and this is associated with increased mortality [9, 12] . Second, genetic variations in the Fas gene are associated with increased risk for ALI/ARDS in humans [13] . Third, activation of the Fas/FasL system in mammals leads to acute inflammatory injury followed by fibrosis [14] [15] [16] [17] . And finally, mice lacking Fas are protected in a number of models of lung injury [11, [18] [19] [20] . Thus, the apoptotic and inflammatory responses induced by Fas activation in the lungs are one important factor in the development of ALI/ARDS. One of the most important neutrophil chemoattractants in mice is the chemokine KC (CXCL1), which is the murine functional homologue of the human chemokine IL8 (CXCL8). Activation of Fas in alveolar epithelial cells is followed by a marked increase in KC expression [10] . Interestingly, recent studies suggest that in injured lungs, IL8 in humans and KC in mice form immune complexes with autoantibodies, and these immune complexes can in turn bind Fc receptors such as FcγRIIa and FcγRIII that amplify the inflammatory response [21, 22] . These studies suggest that, in addition to the absolute levels of chemokines, the formation of autoimmune complexes and their association with Fc receptors is important for the development of inflammatory responses in the lungs. In this study, we ask whether the Fas/FasL system plays a role in the amplification of inflammatory responses that occur early in the course of mechanical ventilation. The specific hypothesis to be tested is that the Fas/FasL system is required for the development of lung injury in mechanically ventilated mice exposed to LPS. To test this hypothesis, we investigate whether the combination of a non-injurious mechanical ventilation strategy with a minimal dose of intratracheal LPS results in an acute lung injury, and whether this injury is attenuated in Fas-deficient lpr mice. We further investigate whether the development of injury is associated with formation and deposition of anti-KC:KC immune complexes. All of the animal experiments were approved by the institutional animal research committee of the University of Washington. Mice were housed in a pathogen-free environment according to University of Washington animal use guidelines. Male C57BL/6 mice ("B6") and mice carrying a spontaneous mutation in the Fas gene that impairs Fas signaling (B6. MRL-Fas lpr /J, "lpr") were obtained from the Jackson Laboratories and studied at 7-13 weeks of age. Briefly, the mice were anesthetized with 5% inhaled isoflurane and then intubated endotracheally with a 20-gauge angiocatheter. Placement of the catheter in the trachea was verified by visualizing the movement of a 100 μl bubble of water in response to respiratory efforts. After confirming intubation, the trachea was instilled with either E. coli LPS, 15 ng/kg or PBS. The instillate was suspended in 2.5% colloidal carbon to allow later confirmation of the extent and distribution of the instillation macro and microscopically. After the installations some of the mice were extubated, returned to their cages, and allowed free access to food and water; other mice where kept intubated and subjected to mechanical ventilation with the following settings: tidal volume (Tv) 10 ml/kg; respiratory rate (RR) 150 breaths/minute; fraction of inspired oxygen (FiO 2 ) 0.21; and positive end-expiratory pressure (PEEP) of 3 cm H 2 O. The heart rate, airway pressures, rectal temperatures and EtCO 2 were monitored continuously using a computerized monitoring system (Chart 4, AD Instruments, Colorado Springs, CO). The RR was adjusted to maintain the EtCO 2 between 30 -40 mmHg. The body temperature was maintained between 37 and 38°C with external heating. The mice were hydrated with a continuous intraperitoneal infusion of lactated ringer solution at 500 μl/hour. Muscle relaxation was attained with pancuronium bromide, 1 μg/g i.p followed by 0.5 μg/g i.p. every hour. After four hours of mechanical ventilation the mice were euthanized with 0.30 ml/kg i.p. of Beuthanasia-D (Schering-Plough Animal Health, Union, NJ). The thorax was rapidly opened and the mouse was exsanguinated by direct cardiac puncture. The left lung was removed and flash-frozen in liquid nitrogen. The right lung was lavaged with 0.6 mM EDTA in PBS; an aliquot of the bronchoalveolar lavage fluid (BALF) was removed for cell counts and differentials, the remaining fluid was spun at 1200 x g, and the supernatants stored at -80°C in individual aliquots. Following the BAL, the right lung was fixed in 4% buffered paraformaldehyde at an inflation pressure of 15 cm H 2 O for histological evaluation. We studied four groups of mice. Two groups consisted of B6 and lpr mice instilled with PBS and allowed to breathe spontaneously ("SB") (n = 7 for B6 mice, 4 for lpr). The other two groups consisted of B6 and lpr mice instilled with LPS and exposed mechanical ventilation (MV + LPS) (n = 10 for B6 mice, 6 for lpr). The main experimental comparison was between the B6 and lpr mice instilled with LPS and exposed to ventilation. Total cell counts in the BALF were performed with a hemacytometer. Differential counts were performed on cytospin preparations using the Diff-quick method (Fisher Scientific Company L.L.C., Kalamazoo, MI). BALF total protein was measured with the bicinchoninic acid method (BCA assay, Pierce, Rockford, IL). BALF IgM (Bethyl Laboratories, Montgomery, TX) and α-macroglobulin (Life Diagnostics, West Chester, PA) were measured with immunoassays. Lung homogenate TNF-α, KC, IL1β, and IL6 were measured using a multiplex fluorescent bead assay (R&D Systems, Minneapolis, MN). As a measurement of the total content of PMN in the lungs we measured myeloperoxidase (MPO) activity in lung homogenates prepared in 50 mM K 2 HPO 4 , pH 6.0 with 5% CH 3 (CH 2 ) 15 N(Br)(CH 3 ) 3 , 5 mM EDTA. Active caspase-3 and Poly ADP ribose polymerase (PARP) activity were measured in lung homogenates prepared on a 1:20 ratio of a lysis buffer (0.5% Triton-X-100, 150 mM NaCl, 15 mM Tris, 1 mM CaCl, and 1 mM MgCl, pH7.4). The lung homogenate was spun at 10,000 x g for 20 min at 4°C and the supernatant used for measurements of active caspase-3 and PARP using the CPP32/ Caspase-3-Fluorometric Protease Assay Kit (BioVision, Mountain View, CA) and a PARP activity kit (Cell Signaling, Boston MA). Serum Creatinine, ALT and bilirubin were measured at a commercial laboratory. Anti-KC autoantibody:KC immune complexes were measured in BAL fluids using an ELISA assay according to a previously described protocol [21] . Briefly, 96-well microtiter plates were coated with antibody against KC (Peprotech). After blocking, the plates were incubated with BAL fluid samples obtained from mice. Then, the plates were washed and incubated with biotinylated horse antibody against mouse immunoglobulins (Vector Laboratories) followed by HRP-conjugated streptavidin and the substrate tetramethyl benzidine (Sigma). C57BL/6 and lpr mice were euthanized by exposure to CO 2 followed by cervical dislocation. The femur and tibia of both hind legs were isolated and freed of all soft tissue, and then the ends of both bones were removed. The femur and tibia were placed proximal end down in a 0.6 mL Eppendorf tube, which had been punctured at its lower tip with an 18-gauge needle and placed inside a 1.5 mL Eppendorf tube. The tubes were spun at 2000 X g for 30 seconds and neutrophils were isolated as previously described [23] . After isolation, neutrophils were labeled with calcein-AM (5 μg/ml; Molecular Probes, Eugene, OR) for 30 minutes at 37°C, washed two times in PBS and resuspended at a concentration of 1 x 10 6 /mL. Neutrophil chemotaxis was assessed using the Neuro Probe ChemoTx W Disposable Chemotaxis system (Neuro Probe Inc. Gaithersburg, MD). The wells of the 96-well plate were filled with various concentrations of KC. A polycarbonate filter (8 μm pores) with a hydrophobic ring around the area over each well was placed on the 96-well plate and calcein-labelled neutrophils were added to each ring. The chemotaxis chamber, consisting of the polycarbonate filter and 96-well plate, was incubated for 30 min at 37°C in 5% CO 2 and then non-migrating neutrophils were removed from the upper side of the filter. The chemotaxis chamber was placed in a multi-well fluorescent plate reader (Synergy 4, BioTek, Winooski, VT) and the migrated cells were measured using the calcein fluorescence signal (excitation -485 nm, emission -530 nm). Neutrophil migration was expressed as a percent of the total number of neutrophils that were placed on the topside of the filter (% Total). Lung sections were embedded in paraffin, cut into 4 μm sections, and stained with hematoxylin and eosin. Lung tissue sections were deparaffinized, washed in xylene, rehydrated, permeabilized with proteinase K, and incubated with the TUNEL reaction mixture according to manufacturer instructions (In Situ Cell Death Detection kit, AP, Roche Applied Science, Indianapolis, IN). Negative controls were treated with labeling solution without terminal transferase. Immediately after TUNEL labeling the slides were washed three times in PBS, blocked with Protein Block (Dako, Carpinteria CA) and incubated for 2 hr in the dark at 37°C with the mouse monoclonal pan-cytokeratin antibody C11 (Abcam, Cambridge, UK) previously labeled with Alexa Fluor 555 (Invitrogen, Eugene, OR). Negative control slides were incubated with an isotype control mouse IgG1k labeled with Alexa Fluor 555 (BD Pharmingen, San Diego, CA). The slides were washed and immediately visualized with a fluorescent microscope. Lung tissue sections from B6 and lpr mice exposed to MV + LPS were processed as previously described [21] . The sections were incubated with anti-KC antibody (Peprotech, Rocky Hill, NJ) followed by chicken antirabbit secondary antibody (Alexa 647, pseudocolor red) (Invitrogen), and then with anti-FcγRIII antibody (R&D Systems) followed by chicken anti-rat secondary antibody (Alexa 488, green), and finally with biotynylated anti-Ly-6 G antibody (eBioscience, San Diego, CA), used as a neutrophil marker, followed by streptavidin (Alexa 568, pseudocolor magenta). Lung tissues were counterstained with Hoechst 33342 (Calbiochem, Gibbstown, NJ). The slides were evaluated using a PerkinElmer Ultra VIEW LCI confocal imaging system with Nikon TE2000-S fluorescence microscope using PlanApox20 objective and PlanApox60 or x100 immersion oil objective (numerical aperture [NA] 1.4) at room temperature. Ultra VIEW Imaging Suite software (version 5.5.0.4) was used for image processing. Statistical analysis was performed using two-factor ANOVA followed by the Bonferroni post-hoc analysis. One factor was "treatment" (SB or MV + LPS) and the other factor was "strain" (B6 or lpr). The analysis was designed to determine the overall effect of each of the factors, the presence of an interaction effect, and the comparison of "strain" for each level of "treatment". Data was generated using GraphPad Prism. A p value of less than 0.05 was considered significant. The neutrophil response to MV + LPS was attenuated in the lpr mice There were significantly less PMN in the BAL of lpr mice exposed to LPS + MV than in the B6 mice (2.2 x 0.5 x 10 3 vs 1.3 ± 0.2 x 10 4 cells, p < 0.05) (Figure 1-A) . A similar pattern was seen for lung MPO activity (Figure 1-B) . However, the lung concentrations of the PMN chemoattractants KC and MIP2 (CXCL2), although increased in response to MV + LPS, were similar in the lpr and B6 mice; this was also true for the pro-inflammatory cytokine TNFα (Figure 1 , C-F). Interleukin 6, GM-CSF, and VEGF were not affected by the administration of MV + LPS and were similar in the B6 and lpr mice (data not shown). These data suggest that both lung PMN recruitment and airspace PMN migration were impaired in the absence of functional Fas, but the difference in neutrophil recruitment seen in this model cannot be explained by differences in KC or MIP-2 release. Other parameters of lung injury were similar in B6 and lpr mice Permeability response We and others have postulated that the Fas/FasL system leads to lung injury by inducing apoptosis of pneumocytes, resulting in disruption of the alveolar-epithelial barrier and non-cardiogenic pulmonary edema. However, in the present study the BAL concentration of total protein and of the serum protein α-macroglobulin were similar in the lpr and B6 mice, despite the difference in PMN numbers ( Figure 2 ). The lack of difference may be due to a low level of injury, as supported by the lack of overall change in the MV + LPS group as compared with the SB group; indeed the model was designed so that the ventilatory pattern and the LPS dose would cause minimal or no injury by themselves, so as to determine the very first components of the injury response. The data suggest that neutrophil changes precede permeability changes in ventilated mice exposed to LPS, and that the role of the Fas/FasL system on neutrophil recruitment precedes its disruptive effect on the alveolar-capillary barrier. Interestingly, the activity of caspase-3, often used as a surrogate for apoptosis, was increased in the spontaneously breathing lpr mice, and this increase reached significance in the MV + LPS group (185 ± 24 vs 480 ± 137, p < 0.05) (Figure 3-A) . PARP, a downstream target of active caspase-3, showed similar activity in the lpr and B6 mice (Figure 3-B) . Double labeling of tissue sections for TUNEL and cytokeratin revealed that in both the lpr and the B6 mice, the apoptotic cells were located primarily in the alveolar walls, but were cytokeratin-negative (Figure 3-C) . Thus, contrary to our expectations, there was increased caspase-3 activity in the lpr mice exposed to MV + LPS, and the apoptotic cells were mostly cytokeratin-negative cells localized to the alveolar walls. The lungs of spontaneously breathing B6 and lpr mice showed normal architecture (Figure 4 ). The lungs of B6 mice exposed to MV + LPS showed thickening of the alveolar walls, intra-alveolar neutrophilic infiltrates, and intra-alveolar fibrin strand deposition. In contrast, the lungs of lpr mice exposed to MV + LPS retained their normal architecture. The LPS was administered with inert colloidal carbon, which on the tissue sections was taken up by macrophages and appears as black granulate material in the cytoplasm. This indicates that the Figure 1 Inflammatory response. C57BL/6 (B6) and Fas-deficient lpr mice received intratracheal installations of either PBS followed by four hours of spontaneous breathing (SB), or E. coli LPS, 15 ng/kg, followed by 4 hrs of mechanical ventilation (MV) with tidal volumes of 10 mL per kilogram, FiO 2 of 0.21, PEEP = 3 and respiratory rate = 150 breaths per minute. In response to the combination of MV and LPS, the B6 mice showed significantly more total neutrophils (PMN) in the BAL fluid than the lpr mice (A). A similar pattern was seen for the lung MPO activity, which is a measure of the total neutrophil content in the lung (B). MV + LPS was associated with increases in the lung homogenate concentrations of the cytokinesTNF-α, MIP-2, and KC, and these similar in the B6 and the lpr mice; MV + LPS had no effect on IL-1β (C-F). n = at least 6/group. sections depicted were all exposed to instillate -an important point as intratracheal instillations are patchy and normal tissue can simply reflect a non-instilled area. We have previously reported that the combination of MV + LPS for 6 hours (2 hr longer than the present study) is associated with distal organ injury, in particular biochemical and histologic evidence of kidney and liver damage [24] . In this study, we also noticed an increase in serum creatinine in the mice exposed to MV + LPS, but this increase was strain-independent ( Figure 5-A) . Interestingly, the lpr mice had increased serum AST concentrations at baseline, and this was not further increased by the addition of MV + LPS (Figure 5-B) . Total bilirubin was similar in lpr and B6 mice, and was not affected by MV + LPS (Figure 5-C) . Thus, at this early time the evidence of distal organ injury was limited to increased creatinine, and this increase was not dependent on the Fas/FasL system. The differences in BAL neutrophils are not explained by differences in B6 and lpr neutrophil chemotaxis An impairment in lpr neutrophil chemotaxis towards KC could explain the attenuation of BAL neutrophils seen in the lpr mice exposed to MV + LPS. Therefore, we compared the chemotactic ability of neutrophils isolated from B6 and lpr mice to KC in vitro. However, we noticed no impairment of chemotaxis in the lpr neutrophils; if anything, chemotaxis was slightly increased in the lpr neutrophils as compared with the B6 neutrophils ( Figure 6 ). A new line of research suggests that mammals form autoantibodies against chemokines such as IL8 in humans and KC in mice, and immune complexes derived from these autoantibodies can induce inflammation by binding to Fcγ receptors on leukocytes, endothelial cells and other cells. Measurement of anti-KC:KC complexes in the BALF revealed an increase in B6 mice treated with MV + LPS, which was not seen in the lpr mice (Figure 7) We also used immunofluorescence staining and confocal microscopy to investigate if there were differences in the deposition of anti-KC:KC complexes in ventilated B6 and lpr mice. The alveolar walls of the ventilated B6 mice contained abundant leukocytes, identified based on Gr-1 positivity ( Figure 8 ). The leukocytes expressed the receptor FcγRIII on their membrane surface. KC co-localized with the FcγRIII receptors. In contrast, there were, in general, Caspase-3 activity was significantly higher in the lpr mice exposed to MV + LPS. n = at least 6/group. Double-labeling for TUNEL (green) and cytokeratin (red) reveals that the TUNEL positive cells are located in the alveolar wall, but most of them are cytokeratin negative. less Gr-1 positive cells in the lungs of the lpr mice, and there was also decreased deposition of KC in the tissue. Thus, the formation of anti-KC:KC IC was attenuated in lpr mice, and this was associated with decreased KC deposition in the tissues. To confirm the specificity of the anti-KC:KC labeling, mouse PMN were purified and incubated with either anti-KC-KC immune complexes (Figure 9 , first two columns) or with KC alone (Figure 9 , third column). Then, the cells were labelled with either the Peprotech antibody, which detected the immune complexes, or an R&D antibody, which detected only KC. Finally, we tested expression of FcγRIII in WT and lpr mice treated with MV + LPS, and found it to be lower in the lpr mice ( Figure 10 ). The goal of this study was to determine whether Fas deficiency is protective in mechanically ventilated mice exposed to intratracheal LPS. The main finding was that Fas-deficient lpr mice showed decreased neutrophil recruitment in response to combined mechanical ventilation and LPS, even though the cytokine, permeability, and apoptotic responses were similar to those of B6 mice. Interestingly, despite the presence of similar concentrations of KC in the BAL of lpr and B6 mice, only the B6 mice showed extensive KC deposition in the lungs, and this was associated with the presence of anti-KC:KC complexes in the BALF and lung tissue. In this study, we focused on the initial events of the injury response by studying mice four hours after the initiation of injury, which was induced by combining a very low dose of endotoxin with mechanical ventilation. The main finding was that a functional Fas/FasL system was required for neutrophil migration into the airspaces of the lungs of mechanically ventilated mice exposed to LPS. The changes in BAL neutrophils seen in this study were associated with changes in lung myeloperoxidase (MPO) activity, and because lung MPO activity is a measurement of the total content of neutrophils in the lungs, the data suggests that the Fas/FasL system is required for both neutrophil migration into the airspaces and for neutrophil recruitment into the lungs. This observation confirms other studies that suggest a key role for the Fas/FasL system in neutrophil migration into the airspaces; for example, activation of Fas in the lungs of mice and rabbits results in a neutrophilic alveolitis (1, 2) , whereas pharmacologic blockade of the Fas/FasL system attenuates the neutrophilic response to bacteria and bacterial products (3) (4) (5) (6) . Tissue respons. C57BL/6 or Fas-deficient lpr mice were exposed to either intratracheal installations of PBS followed by spontaneous breathing or to intratracheal installation of LPS, 15 ng/kg, followed by four hours of mechanical ventilation (MV) with tidal volumes of 10 mL per kilogram. Lung tissue sections were stained with hematoxylin and eosin to demonstrate the pattern of injury. The instillate contained 2.5% colloidal carbon to identify the instilled areas. Spontaneously breathing mice instilled with PBS showed normal lung architecture, regardless of strain (A, C). The B6 mice exposed to MV and LPS showed intra-alveolar and interstitial neutrophilic infiltrates, as well as thickening of the alveolar walls and occasional fibrin deposition (B). One potential explanation for the decreased migration of neutrophils into the airspaces of the Fas-deficient mice is that Fas ligation induces release of neutrophilic cytokines such as KC in macrophages and in lung epithelial cells in vitro (7-9). Thus, we had expected to see lower concentrations of KC in the mechanically ventilated lpr mice exposed to LPS, as compared with the B6 animals. However, this was not the case, and the difference in neutrophil migration was not due to differences in soluble KC concentrations. Another potential mechanism that would explain differences in neutrophil migration with similar concentrations of soluble KC is a difference in neutrophil chemotaxis of lpr and B6 neutrophils. However, our chemotaxis studied showed that, if anything, neutrophils from lpr mice have slightly increased chemotaxis to KC, strongly suggesting that differences in chemotaxis are not the reason for the attenuation in the neutrophilic response seen in the lpr mice. Neutrophil recruitment and migration appears to be partly dependent on the formation of anti-KC:KC immune complexes, which can bind Fcγ receptors in local leukocytes, enhancing the inflammatory response [22] . For example, the generation of anti-KC:KC immune complexes in the lungs of mice is followed by acute inflammatory lung injury, and this injury requires the presence of Fcγ receptors [21] . Furthermore, mice lacking γ chains show attenuated injury in response to LPS, suggesting that this process is relevant for inflammation secondary to bacterial products [21] . The human equivalent of anti-KC:KC complexes are anti-IL8:IL8 complexes, and these are present in the lungs of patients with ARDS [25, 26] . In our study, we found that there are less anti-KC:KC complexes in the lungs of the lpr mice, even though the concentrations of KC were similar to those in the B6 mice. Thus, it is possible that the lower numbers of neutrophils seen in the lpr mice were due to decreased formation and deposition of anti-KC:KC complexes in the lungs. Additional studies are necessary to confirm this hypothesis, and it remains possible that differences in other unmeasured chemotactic agents account for the differences in neutrophil recruitment. The mechanism linking the Fas/FasL system with impaired formation and deposition of immune complexes remains unclear. To date, lpr mice, particularly the MRL/ lpr strain are known to generate autoantibodies that can result in a lupus-like syndrome [27] . A surprising finding in the present study was that there was no increase in the markers of permeability or apoptosis in the B6 mice exposed to mechanical ventilation and LPS, as compared to the lpr mice; instead, the injury response was limited to neutrophilic inflammation and cytokine release. One explanation is that neutrophil recruitment precedes the development of tissue injury in our model, in which the mice were studied relatively early, four hours after the onset of ventilation. Less clear is the finding that the lpr mice actually had increased caspase-3 activity and TUNEL (See figure on previous page.) Figure 8 Anti-KC autoantibody:KC complex deposition in the lungs. FcγRIII (green) and KC (red) in tissue sections from B6 (A) and lpr (B) mice treated with intratracheal E. coli LPS, 15 ng/kg, and exposed to 4 hours of mechanical ventilation (MV) at Tv = 10 cc/kg, FiO2 = 0.21, PEEP = 3. FcγRIII is shown in green, KC is shown in red, and Gr1 (staining leukocytes) is shown in pink. Nuclei are shown in blue. There is an increase signal for both KC and FcγRIII in the B6 mice as compared with the lpr mice. Note the colocalization of KC and FcγRIII. Figure 9 The detection of anti-KC:KC immune complex is specific. Mouse PMN were purified and incubated with either anti-KC-KC immune complexes (first two columns) or with KC alone (third column). Note that the Peprotech antibody, detected only anti-KC-KC immune complexes, whereas the R&D antibody detected only KC. The FcγRIII was labeled green and KC was labeled red. positive cells compared with the B6 mice. We do not have a clear explanation for this finding, but they seem to be specific to the LPS + MV model, because in another model using mechanical ventilation in which WT and lpr mice were exposed to pneumonia virus of mice (PVM) prior to four hours of MV, we found a decrease in caspase-3 activity in the lpr mice [28] . It is important to emphasize that the decrease in neutrophils seen in the lpr mice does not imply "protection". It is unclear whether the lung neutrophilic response is directly associated with negative outcomes in ALI/ARDS or not. Most studies of lung injury presume that neutrophilia or increased concentrations of cytokine are deleterious, but all of these studies are limited in that they do not effectively reproduce the series of events seen in the clinical setting, where patients with multiple comorbidities are intubated for prolonged periods of time. Our study should be interpreted as knowledge on the mechanisms of that initiate lung injury, but not extrapolated to the ultimate results of that injury. In summary, we report that in B6 mice, the combination of mechanical ventilation and LPS is associated with recruitment of neutrophils to the lungs and with deposition of anti-KC:KC immune complexes. In comparison, mice deficient in Fas recruited lower numbers of neutrophils, and this was associated with less deposition of immune complexes in the lungs. We conclude that a functioning Fas/FasL system is required for a full neutrophilic response to LPS in mechanically ventilated mice.

Evolutionary Dynamics of the Interferon-Induced Transmembrane Gene Family in Vertebrates

Vertebrate interferon-induced transmembrane (IFITM) genes have been demonstrated to have extensive and diverse functions, playing important roles in the evolution of vertebrates. Despite observance of their functionality, the evolutionary dynamics of this gene family are complex and currently unknown. Here, we performed detailed evolutionary analyses to unravel the evolutionary history of the vertebrate IFITM family. A total of 174 IFITM orthologous genes and 112 pseudogenes were identified from 27 vertebrate genome sequences. The vertebrate IFITM family can be divided into immunity-related IFITM (IR-IFITM), IFITM5 and IFITM10 sub-families in phylogeny, implying origins from three different progenitors. In general, vertebrate IFITM genes are located in two loci, one containing the IFITM10 gene, and the other locus containing IFITM5 and various numbers of IR-IFITM genes. Conservation of evolutionary synteny was observed in these IFITM genes. Significant functional divergence was detected among the three IFITM sub-families. No gene duplication or positive selection was found in IFITM5 sub-family, implying the functional conservation of IFITM5 in vertebrate evolution, which is involved in bone formation. No IFITM5 locus was identified in the marmoset genome, suggesting a potential association with the tiny size of this monkey. The IFITM10 sub-family was divided into two groups: aquatic and terrestrial types. Functional divergence was detected between the two groups, and five IFITM10-like genes from frog were dispersed into the two groups. Both gene duplication and positive selection were observed in aquatic vertebrate IFITM10-like genes, indicating that IFITM10 might be associated with the adaptation to aquatic environments. A large number of lineage- and species-specific gene duplications were observed in IR-IFITM sub-family and positive selection was detected in IR-IFITM of primates and rodents. Because primates have experienced a long history of viral infection, such rapid expansion and positive selection suggests that the evolution of primate IR-IFITM genes is associated with broad-spectrum antiviral activity.

First discovered by cDNA library screening in 1984 [1] , the interferon-induced transmembrane (IFITM) gene family plays critical roles in a variety of cellular processes and contains IFITM1, IFITM2, IFITM3, IFITM5, IFITM6, IFITM7, IFITM10 and some IFITM-like genes [2] . Except for IFITM5 that is specifically expressed in bone cells in an interferon (IFN)independent way [3, 4] , all IFITM genes can be stimulated by IFN [5, 6] , and are widely expressed in tissues and organs [2] . IFITM family members contain a conservative CD225 domain and two terminal hypervariable regions [2] . The CD225 domain accounts for more than half of the protein in length, containing one intact transmembrane domain (TMD), two S-palmitoylation sites regions and partial TMD in the C-terminus of the protein. The S-palmitoylation sites have been demonstrated to play important roles in post-translational processing and stability of IFITM proteins [7] . The N-terminal hypervariable region generally contains 21 amino acid residues and the C-terminal one includes a TMD (Fig. 1) [8] . In different vertebrates, the functions of different IFITM members diverge. IFITM1, IFITM2 and IFITM3 are involved in cell adhesion [9] , antiproliferation [9] , tumor suppression [10, 11] , and germ cell and embryonic development [12] . More recently, these genes were identified as novel types of antiviral restriction factors with a wide spectrum of antiviral activity against influenza A viruses (e.g. H1N1 viruses), West Nile virus, dengue virus, filoviruses, HIV-1, HCV, venezuelan equine encephalitis virus (VEEV), chikungunya virus (CHIKV), vesicular stomatitis virus (VSV) and even SARS-coronavirus [8, [13] [14] [15] [16] [17] [18] . The main function of IFITM5 is associated with bone development in vertebrates [4, 19, 20] . IFITM6 seems to be involved in macrophage functions in tumor suppression [20] . To date, however, there is no information about the functions of IFITM7 and IFITM10. Several antiviral restriction factors (e.g. APOBEC3G, Tetherin, and SAMHD1) have been demonstrated to evolve under positive selective pressure from viruses [21] [22] [23] [24] [25] [26] [27] [28] [29] . As important virus inhibitors, IFITM1, IFITM2 and IFITM3 may have also undergone a similar co-evolutionary process, such as other antiviral restriction factors do. Despite this connection, relationships between antiviral functions and adaptive evolution in IFITM family have seldom been reported and although previous reports had illustrated the phylogenetic history of IFITM family in some eukaryotic species [2, 30, 31] , there has been no detailed information about IFITM genes in vertebrates. In this study, we performed detailed evolutionary analyses not only to test whether the primate IFITM genes evolved under positive selection throughout primate evolution, but also to unravel the evolutionary history of vertebrate IFITM family. To characterize the IFITM gene repertoires in vertebrates, we searched 27 vertebrate genome sequences with high genome coverage ($66) or representing the major evolutionary lineages in vertebrate phylogeny (such as opossum, lizard, platypus, etc.), using previously described IFITM sequences as queries. The taxa included ten non-mammalian vertebrates (five fishes: stickleback, tetraodon, medaka, fugu, zebrafish; one amphibian: frog; one nonavian reptile: lizard; three birds: chicken, turkey, zebra finch) and 17 mammals covering six primates (human, chimpanzee, gorilla, orangutan, macaque, marmoset), four glires (mouse, guinea pig, rat, and rabbit), five other mammals (tree shrew, cow, horse, dog and elephant), 1 metatherian (opossum) and 1 prototherian (platypus). We divided the newly identified genes into three types based on the following criteria: (i) functional genes, which contain full-length ORFs with intact CD225 domain and C-terminal TMDs; (ii) putative functional genes, which have CD225 domains but contain incomplete ORFs; and (iii) pseudogenes, which are sequences with pre-mature stop codons. We identified a total of 286 IFITM-related sequences ( Table 1 ). The number of IFITM genes varies considerably between mammals and non-mammalian vertebrates. A total of 27 functional genes were identified from 10 non-mammalian species, ranging from one gene in stickleback or tetraodon to six genes in frog. By contrast, 134 functional genes were identified from 17 mammalian species, ranging from one in platypus or rabbit to 26 in marmoset. Additionally, 14 putative functional genes including 13 in mammals and one in non-mammalian vertebrates were identified. Among the 10 non-mammalian and 17 mammalian species, 10 and 102 pseudogenes were identified, respectively. Interestingly, among 10 non-mammalian vertebrates, frog has the highest number (six functional genes and eight pseudogenes) of IFITM-related sequences. Among the 17 mammals, marmoset has the highest number (40) of IFITM-related sequences, of which 26 are functional genes. Human has the second highest number (29) of IFITM-related sequences, of which 18 are pseudogenes, the maximum in vertebrates. To understand the phylogenetic relationship of IFITM gene family, 160 functional IFITM genes identified from 27 vertebrate species (Table S1 ) were subjected to phylogenetic analyses using Bayesian inference, maximum likelihood (ML) and maximum parsimony (MP) methods. The Bayesian, ML and MP trees show consistent topological structures (Figs. 2A and S1). All functional IFITM genes were well divided into three clades I, II and III with Bayesian posterior probabilities of $97. These analyses indicate that the IFITM family most likely originated from three progenitors. The clades I, II and III contain 114, 26, and 20 IFITM genes, respectively. All IFITM10 and IFITM5 genes are clustered in clades II and III, respectively. All other functional IFITM genes, including IFITM1, IFITM2, IFITM3, IFITM6 and IFITM7, are grouped in clade I, forming the biggest sub-family in the IFITM family. Because the expression of the IFITM genes in clade I can be induced by IFN and their functions are associated with immunity [8, 13, 15] , they are defined as immunity-related IFITM (IR-IFITM) sub-family. In each clade, eutherian IFITM genes form a separate group from opossum and bird genes, consistent with the species phylogeny. Besides IFITM5 and IFITM10, IR-IFITM genes also have orthologs in both marsupials and eutherians, arguing against previous observation that only IFITM5 and IFITM10 orthologs could be identified in marsupials and eutherians [30] . In addition, clade II and III contain homologous IFITM sequences from teleosts and amphibians, but clade I does not ( Fig. 2A) , indicating that IR-IFITM originated later than IFITM5 and IFITM10. Phylogenetic Analyses of the Vertebrate Immunityrelated IFITM Sub-family IR-IFITM genes can be divided into two groups: one consisting of eutherian homologs and the other including homologs from metatheria and bird. We further constructed a sub-tree to show the phylogenetic relationship of 109 IR-IFITM genes from eutheria (Fig. 3) . The sequences from elephant are located on the basal position. All IR-IFITM genes from the primate lineages form a sub-clade, and those from rodents form another sub-clade. Four genes from tree shrew form a species-specific cluster located between the sub-clades of the primates and the rodents. Three mammal species, dog, horse and cow, form several species-specific IR-IFITM gene clusters, which further compose a sub-clade in accordance with the phylogeny of these three species. These suggest that IR-IFITM genes evolved via gene duplication after species separation. Interestingly, the majority of IR-IFITM genes from rodents do not form species-specific clusters as each of IFITM1, IFITM2 and IFITM3 clusters together in a lineage-specific manner. This clustering indicates that IR-IFITM genes have diverged into different IFITM isoforms prior to the split of rodents from other mammals. Additionally, in the lineage-specific clusters, more than one IR-IFITM gene was observed from certain species, indicating that gene duplication of IR-IFITM genes continued until after species separation of the rodents. Furthermore, rat IFITM7 clusters closely with rat IFITM3, indicating that they are a pair of duplicated genes. Similarly, mouseIFITM7, mouseIFITM1 and mouseIFITM-like3 (IFITMac) are another group of duplicated genes, suggesting that IFITM7 might have similar biological function to IFITM3 or IFITM1. Within the primate sub-clade, three separate clusters of IFITM1, IFITM2, and IFITM3 were observed. The IFITM1 cluster contains the sequences from all analyzed primates, excluding marmoset and orangutan, and is located at the basal position of the primate sub-clade, indicating that IFITM1 separated earlier than other IR-IFITM genes, including IFITM2 and IFITM3, during the primate evolution. The IFITM2 and IFITM3 clusters only contain sequences from three hominids (human, chimpanzee and gorilla), indicating that IFITM2 and IFITM3 originated prior to the separation of these three hominids. Likewise, similar to the rodent sub-clade, primate IR-IFITM genes also form different clusters, which contain orthologous IFITM genes from different species, indicating that most IR-IFITM members diverged prior to species separation. Furthermore, some IR-IFITM genes from same species cluster together and form species-specific sub-clusters, indicating that the IR-IFITM subfamily experienced a rapid expansion through gene duplications after the divergence of primates. Interestingly, a species-specific cluster was formed by the 25 IR-IFITM genes from marmoset, suggesting a rapid expansion of IR-IFITM genes by gene duplication. This speculation is supported by the four pairs of marmoset IFITM genes (IFITM-like8 and IFITM-like24, IFITM-like10 and IFITM-like29, IFITM-like14 and IFITM-like26, and IFITM-like11 and IFITM-like27) that exhibit very close genetic relationships, possibly indicating relatively recent gene duplication events. Additionally, the two genes in each pair are located in different chromosomes (Fig. 4B ), indicating their origination by segmental duplication rather than tandem duplication. In total, 20 IFITM5 genes were identified in 20 vertebrates, covering species from teleosts to eutherians (Fig. 2B ). No IFITM5 gene was identified in gorilla, marmoset, tree shrew, rabbit, dog, fugu, or medaka. Due to low sequencing coverage, we cannot rule out the possibility that the lack of IFITM5 in these genomes should be ascribed to relatively low quality of the genome sequences. The phylogenetic relationships of the available IFITM5 genes are consistent with the known species phylogeny. Only one IFITM5 gene was identified in each species, indicating that no gene duplication occurred in IFITM5 gene during the evolution of vertebrates. In 19 vertebrates, 26 IFITM10 genes were identified, covering species from teleosts to eutherians (Fig. 2C ). Each terrestrial vertebrate we surveyed possesses one IFITM10 gene, implying that no gene duplication occurred during the evolution of terrestrial vertebrates. In semi-aquatic frog, one IFITM10 and four IFITM10-like genes were identified. In aquatic vertebrates, three and two IFITM10 or IFITM10-like genes were identified in fugu and medaka, respectively. These indicate that species-specific gene duplications occurred in lower vertebrates. Within the IFITM10 clade, two groups were observed. One includes four IFITM10-like genes from frog and various numbers of IFITM10like genes from teleosts. The other one comprises the frog IFITM10 gene and all IFITM10 genes from terrestrial vertebrates. Such division might distinguish aquatic-and terrestrial-type of IFITM10 and accordingly the amphibian frog possesses both types. These findings suggest that functional divergence of IFITM10 in the evolution of vertebrates may have occurred during the transition from an aquatic to a terrestrial environment. In order to further understand the evolutionary scenario of IFITM family, we investigated the chromosomal distribution of well-defined IFITM genes including IFITM1, IFITM2, IFITM3, IFITM5, IFITM6, and IFITM10. Because IFITM7 has chromosome location independent from any other IFITM genes, and only mouse and rat have IFITM7 gene, we did not take it into account in this analysis. Orthologous relationships between IFITM family members were well confirmed with conserved syntenies (Fig. 4A ). All these genes are located in one chromosome and form two loci in terrestrial vertebrates, except in cow that has two loci in two different chromosomes. With the exception of IFITM10, that is located at one locus, all other IFITM genes gather together and form a gene-cluster in the other locus. In teleosts (e.g. zebrafish), the two loci are dispersed in two different chromosomes. Although we identified two loci in two scaffolds of frog, whether they are dispersed in two different chromosomes is still unknown. We investigated genes flanking both sides of the two IFITM loci, and found that the two loci have almost completely same flanking genes from lower (e.g. zebrafish) to higher (e.g. human) vertebrate species. These findings suggest that chromosomal fusion might have occurred during the vertebrate evolution from aquatic to terrestrial species. Furthermore, all these IFITM genes have two exons, and the IFITM gene-clusters exhibit consistent gene order in four hominids, two rodents and three birds, supporting the syntenic relationship of IFITM genes. The IFITM gene-clusters of different vertebrate species contain a variety of IFITM gene numbers (Fig. 4B) . Due to incomplete genome information, the gene-cluster of some species including tree shrew, platypus and some teleosts are not determined. In zebrafish, the orthologous IFITM gene-cluster only contains IFITM5 in chromosome 25. In bird and lizard, two IFITM genes IFITM3 and IFITM5 are included in the gene-cluster. The hominid IFITM gene-clusters contain 3-4 IFITM genes, and the two rodent (mouse and rat) clusters contain five IFITM genes. In other mammals (cow, dog, horse, and elephant), gene-clusters have 4-6 IFITM or IFITM-like genes, including IFITM5 and various number of IFITM-like genes. The IFITM gene family was divided into three clades. Clade I was further divided into several sub-clades. To test whether there is functional divergence between different clades or between different sub-clades, we estimated type I divergence using DI-VERGE v2.0 [32] and detected significant functional divergence between IR-IFITM and IFITM5, and between IFITM10 and IFITM5 (P,0.01). However, functional divergence signal was not detected between the IR-IFITM genes and IFITM10 (P = 0.0655) ( Fig. 5 and Table 2 ). IFITM10 genes are divided into two subgroups, terrestrial-type and aquatic-type and significant functional divergence was also detected between the two sub-groups (Fig. 5D ). Among IR-IFITM genes, although IFITM2 and IFITM3 might originate from IFITM1 via gene duplication, there is no functional divergence observed between IFITM1 and IFITM2&3 genes (data not shown). Conversely, crucial amino acid residues responsible for the functional divergence of IFITM genes among the three clades were predicted using a posterior-based site-specific profile (Fig. 5) . Surprisingly, almost all sites of IFITM CD225 domains are crucial for the functional divergence between IR-IFITM and IFITM5. Some residues located in the CD225 domain and the C-terminal To investigate whether positive selection drove the evolution of the vertebrate IFITM gene family, we calculated the nonsynonymous substitution (dN) and synonymous substitution (dS) distances [33] between each pair of the sequences from the three clades. To exclude false signals caused by recombination, we first evaluated the effect of gene conversion using GENECONV [34] . Gene conversions were found in some species which are under species-specific duplication including dog, cow, horse, etc. (Table S2 ). Those sequences were removed from our datasets for subsequent analyses. There is no significantly higher dN than dS in the pairwise comparisons of the sequences from clade III, suggesting that no positive selection acted on IFITM5 (Fig. 6A) . Further site-specific analyses using PAML (Table 3 ) and HyPhy (data not shown) confirmed no positive selection acting on IFITM5 genes. There are three pairs of IFITM10 genes (in fugu: IFITM10-like3 vs. IFITM10-like2, and IFITM10-like1 vs. IFITM10-like2; in medaka: IFITM10-like1 vs. IFITM10-like2) with dN/dS.1 (Fig. 6A) , indicating positive selection acting on aquatic IFITM10 gene. Because gene expansion of IFITM10 was observed in frog (Fig. 2C) , we analyzed the five IFITM10 related genes (including one IFITM10 and four IFITM10-like genes) using the site-specific model. A strong signal of positive selection was detected. Eight sites were identified as under positive selection at the level of posterior probability.0.8 (Table 3) . We further constructed the phylogeny of frog IFITM10-related genes and counted the numbers of non-synonymous (n) and synonymous (s) substitutions on each branch (Fig. S6) . The gn/gs ratio (1.75) is significantly smaller than the N/S ratio (2.96) (P = 0.0112, chi-square test), showing no positive selection. However, a very strong signal of positive selection (v = 999, P = 0.0077, chi-square test) was Within clade I, 12 of 5996 pairwise comparisons exhibit significantly higher dN than dS (Fig. 6B and Table S3 ), indicating the presence of positive selection. Among 12 pairwise comparisons with significantly higher dN than dS, eight occur between the primate sequences (Fig. 5C) , three between the rodent sequences (Fig. 5D) , and one between cow sequences. Further site-specific analysis detected two positively selected sites (PSS) in the primate IR-IFITM genes and one in the rodent IR-IFITM genes ( Table 3) . No PSS was detected in the primate IFITM1 and IFITM2&3 subgroups. Because positive selection is generally associated with the occurrence of gene duplication [35] , and a large number of gene duplications were observed in the marmoset IR-IFITM genes, we also performed site-specific analysis on marmoset IR-IFITM genes. As expected, one significant PSS was detected in marmoset IR-IFITM genes. To confirm the results from PAML, similar positive selection analyses were performed using MEME on the DATAMONKEY server. The results reveal that there are five, four, seven and one sites underwent significant positive selection in primate, rodent, marmoset IR-IFITM groups and frog IFITM10 group, respectively (Table S4) . PSSs identified by MEME method are consistent with those identified by PAML. Furthermore, positive selection can be also confirmed by branch-site REL method in both IR-IFITM and IFITM10 clades (Fig. S7) . The major lineages that have undergone positive selection in the IR-IFITM clade are primate and rodent (Fig. S7A) . Additionally, IFITM10 in lower vertebrate lineages also experienced strong selective pressure. Positively selected branches were detected in both teleost and frog lineages (Fig. S7B ). IFITM family contains seven members (IFITM1, IFITM2, IFITM3, IFITM5, IFITM6, IFITM7 and IFITM10), as well as some IFITM-like genes. All vertebrate IFITM genes are divided into three clades ( Fig. 2A) , implying origins from three progenitors. Clades I, II and III contain IR-IFITM, IFITM5 and IFITM10 genes, respectively. Substantial functional divergences occurred between IR-IFITM, IFITM5 and IFITM10 genes, indicating that IR-IFITM, IFITM5 and IFITM10 experienced individual evolutions. IR-IFITM, IFITM5 and IFITM10 genes are usually located in two loci (Fig. 4) . One locus contains only the IFITM10 gene, and the other locus contains various numbers of IR-IFITM genes with a syntenic location with IFITM5, forming an IFITM genecluster. The two loci can be used as good markers to trace the evolutionary history of IFITM family. They are located in two different chromosomes in lower aquatic vertebrates, and evolved to lie in one chromosome by chromosomal fusion in higher mammals (Fig. 4A) . The syntenic relationship of IR-IFITM genes and the presence of more IR-IFITM genes in mammals than other vertebrates suggest that IR-IFITM gene sub-family experienced a rapid expansion via tandem duplication during evolution from lower vertebrates to mammals. Different IFITM members exhibit various functions. IFITM5 is specifically expressed in bone cells, but could not be induced by IFN stimulation [3, 19] . IFITM5 is involved in bone formation and considered as a bone-specific modulator of mineralization. The vertebrate IFITM5 genes form an independent clade in IFITM family ( Fig. 2A) . Neither gene duplication nor positive selection was identified in IFITM5 sub-family ( Fig. 6B and Table 3 ), implying the functional conservation of IFITM5 in vertebrate evolution. Interestingly, no IFITM5 gene was found in the genomic data of two primate species, gorilla and marmoset. Gorilla has a close phylogenetic relationship with human and chimpanzee. In human and chimpanzee, IFITM5 is located upstream of the IFITM1-IFITM2-IFITM3 gene-cluster in chromosome 11. IFITM1, IFITM2 and IFITM3 in gorilla are also located in chromosome 11, and form a similar gene-cluster to those of human and chimpanzee (Fig. 4) . Therefore, gorilla is also presumed to have an IFITM5 gene upstream of the IFITM1-IFITM2-IFITM3 gene-cluster. In fact, we found a region with incomplete sequencing at the corresponding locus of human or chimpanzee IFITM5 genes in chromosome 11 of gorilla, explaining why we could not identify IFITM5 gene locus in current gorilla genome dataset. In contrast, marmoset has special IFITM gene organization and experienced rapid gene expansion, thus forming unique species-specific gene cluster (Fig. 3) . Considering that marmoset is one of the smallest monkeys in the world, as well as that IFITM5 plays a crucial role in bone formation [19, 36] , we inferred that marmoset most likely lost its IFITM5 during the long period of evolution. Nevertheless, whether the loss of IFITM5 gene contributes to the tiny size of the marmoset needs to be determined by future studies. So far, there is no study reporting the function of IFITM10. All various vertebrate IFITM10 genes can be divided into two subgroups, one of which contains IFITM10 from terrestrial vertebrates with frog IFITM10 at the basal position, suggesting that terrestrial vertebrates and frog share a common IFITM10 ancestor. The other one contains IFITM10 from aquatic vertebrates, as well as four frog IFITM10-like genes. Significant signal of functional divergence was observed between the two subgroups (Table 2) , possibly suggesting an association with terrestrial and aquatic environments. In particular, both gene duplication and positive selection can be detected in IFITM10 or IFITM10like genes from the aquatic vertebrates ( Fig. 6D and Table 3 ), indicating that IFITM10 is associated with the adaptation to aquatic environments. There are one IFITM10 gene and four IFITM10-like genes in frog (Fig. 2C ). An episodic adaptive evolution was found on the branch leading to three frog IFITM10like genes (Fig. S6) , supporting the association of IFITM10 with terrestrial and/or aquatic environments. However, what the function of IFITM10 is and how it helps the aquatic vertebrates to adapt to aquatic environments still need to be determined. Distinct from IFITM5 and IFITM10, the IR-IFITM sub-family contains IFITM1, IFITM2, IFITM3, IFITM6, IFITM7, as well as a large number of IFITM-like genes [37, 38] . This sub-family forms a large clade in the phylogenetic tree (Fig. 3) . The IR-IFITM genes from same mammalian species, such as dog, horse, cow, elephant, guinea pig and tree shrew, cluster together and form species-specific IR-IFITM gene sub-clusters (Fig. 3) , in-dicating that gene duplication occurred after the separation of these mammalian species in a species-specific pattern. The same IFITM member from mouse and rat cluster together, indicating that the gene duplication there occurred prior to the separation of the two species in a lineage-specific pattern. The rodent IFITM1, IFITM2 and IFITM3 sub-groups cluster together (Fig. 3) , suggesting a close phylogenetic relationship and similar functions. IFITM6 and IFITM7 are specific for the rodents. The rat and mouse IFITM6 genes cluster together and further group with the guinea pig IR-IFITM gene sub-cluster. Mouse IFITM7 clusters closely with mouse IFITM1 gene and rat IFITM7 clusters closely Evolutionary Dynamics of IFITM Gene Family with rat IFITM3 gene (Fig. 3) , which might indicate IFITM7 has similar functions to IFITM1 or IFITM3. Lineage-specific gene duplications were also observed in the primate IR-IFITM genes. The IFITM1 genes from some primate species form an individual sub-group located at the basal position of the primate IR-IFITM sub-clade (Fig. 3) . IFITM1, IFITM2 and IFITM3 have miscellaneous functions including cell adhesion, antiproliferation, tumor suppression and embryonic development. Apart from these biological functions, human IFITM1, IFITM2 and IFITM3 have a broad-spectrum of antiviral activity, possibly brought on by inhibiting the viral entry processes [14, 18, 39] . In particular, human IFITM3 and IFITM2 appear to have higher antiviral activity than IFITM1 [14] and IFITM3 has been reported to inhibit virus replication in other mammals [40] . These findings suggest that after IFITM1 occurrence, the generation of IFITM2 and IFITM3 might be associated with host defense against various virus infections. Primate and rodent IFITM1, IFITM2 and IFITM3 have similar functions [2] , but do not form a monophyletic cluster (Fig. 3) , indicating that they do not share the most recent common ancestor (MRCA) and moreover suggesting convergent evolution of IFITM1, IFITM2 and IFITM3 in primates and rodents. Convergently evolved amino acids between primates and rodents were found in the C-terminus of IFITM2 and IFITM3 (data not shown), a crucial region for antiviral activity, supporting the association between viral infections and the evolution of IR-IFITM genes. The largest scale of gene expansion of IR-IFITM genes was observed among the primates in a complex pattern that includes both lineage-and species-specific gene duplication events. The species-specific pattern mainly seems to have occurred in macaque and marmoset, giving rise to seven IFITM2&3-like genes in macaque and 29 IR-IFITM genes in marmoset (Fig. 3) . Large scale of duplication and pseudogenization events suggest that the IR-IFITM clade evolved under birth-and-death model [41, 42] . Interestingly, however, marmoset does not possess any one of IFITM1, IFITM2 or IFITM3, and macaque possesses IFITM1 but not IFITM2 and IFITM3. Why marmoset and macaque evolved so many IR-IFITM genes remains unclear and should be explored in more detail in future studies. The Red Queen hypothesis presumes that the antagonistic coevolutionary dynamics of host-virus systems can generate selection Table 3 . Site-specific tests for positive selection on different IFITM sub-families. for accelerated evolution of host antiviral restriction factors, just like the observations on the primate antiviral restriction factors APOBEC3G, Tetherin, and SAMHD1 [21] [22] [23] [24] [25] [26] [27] [28] [29] . We detected positive selection acting on marmoset IR-IFITM genes (Table 3) . Macaque and marmoset are susceptible to infection by many contemporary viruses and are often used as suitable non-human primate models for viral infectious disease studies [9, [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] . Cell lines (e.g. kidney cells) from marmoset can be infected by most primate viruses, such as flaviviridae family of viruses (GBV-B), lassa virus, peste des petitis ruminants (PPR) virus and so on [59, 60, 61] . These imply that marmoset and macaque might be able to be infected by the ancestors of contemporary primate viruses and/or some unknown viruses during early evolution of primates. Additionally, relative to pig IFITM3 that has been demonstrated to have an antiviral activity [40] , IFITM1, IFITM2 and IFITM3 genes from macaque and marmoset have higher similarity to human IFITM1, IFITM2 and IFITM3 in sequence and domain organization. These imply that although lacking experimental support, the IR-IFITM genes from macaque and marmoset might have similar antiviral activity to human IFITM1, IFITM2 and IFITM3. Therefore, the rapid expansion of IR-IFITM genes might be ascribed to the infection of marmosets by viruses. In this study, we demonstrated the evolutionary dynamics of IFITM genes that diversifies in different sub-clades, probably in accordance with their distinct functions. Future studies on immunology, developmental biology and comparative biology to determine IFITM functions would better clarify the relationship between divergence and functions and likewise extend our knowledge on IFITM function and evolutionary mechanisms. Functional IFITM gene sequences were gained based on orthologous and paralogous relationships by querying the Ensembl genome assemblies (http://www.ensembl.org/index.html) using known authentic IFITM genes [17] . Collected IFITM genes were used as queries to search against known IFITM gene datasets using tBLASTn or BLASTn searches to make sure the best hit is an functional IFITM genes with E value,10 210 . The presence of CD225 domain in each obtained IFITM protein sequence was confirmed using P-fam database (http://pfam.sanger.ac.uk). Annotated IFITM genes were retrieved from Ensembl databases (http://www.ensembl.org/). To identify additional putative IFITM functional genes and pseudogenes, tBLASTn searches were performed in Ensembl using human IFITM1, IFITM2, IFITM3 and IFITM5 as queries against the available genome sequences of species listed in Table 1 with E value,10 25 [62] . After deleting the redundancies and merging overlapping sequences, the remained sequences with.150 nt were analyzed by GENSCAN to identify the putative coding sequence. Each candidate IFITM gene was used as a query to BLAST against GenBank non-redundant protein database to make sure the best hit is an IFITM gene. The exons and intron of the remained sequences were detected with GeneWise. The presences of CD225 domain and C-terminal TMD in candidate sequences were identified using the online tools SMART (http://smart.emblheidelberg.de) and SOSUI (http://bp.nuap.nagoya-u.ac.jp/sosui/ ), respectively. If the sequence did not have complete CD225 domain and the C-terminal TMD, or its open reading frame (ORF) was disrupted, this sequence was regarded as a pseudogene. To avoid the possible error in pseudogene identification, the PSEDOPIPE approach (http://pseudofam.pseudogene.org/ pages/main/about.jsf) [63] was further used to confirm identified IFITM pseudogenes. On the other hand, because authentic IFITM proteins are in a range of 102 to 157 amino acids [2] , the candidates that are outside above range and have complete open reading frame, complete CD225 domain and the C-terminal TMD were identified as functional genes and referred to as IFITM-like genes. Conserved motifs in the complete amino acid sequences of the mammalian IFITM proteins were analyzed by MEME/MAST software (http://meme.sdsc.edu/meme/website/intro.html). The secondary structure of IFITM proteins was predicted using SMART (http://smart.embl-heidelberg.de/smart). The logo pictures were generated by Weblogo (http://weblogo.berkeley.edu). Multiple sequence alignments were performed using Muscle in MEGA 5.0 [64] and were refined manually in Bioedit (http:// www.mbio.ncsu.edu/BioEdit/BioEdit.html). Unambiguously aligned positions (Figs. S2, S3 and S4) were used for subsequent phylogenetic analyses. Maximum likelihood (ML) tree of IFITM gene family was reconstructed by PHYML2.4 implemented in Jmodeltest 0.1 package with the best-fitting model of F81+I+G that is selected using JmodelTest in the same package [65, 66] . The bootstrap analysis was performed with 1000 replications. Bayesian inference (Bayes) tree was reconstructed using MrBayes v3.1.2 [67] . Four independent Markov Chain Monte Carlo (MCMC) chains were used with the default temperature of 0.01. Four repetitions were run for 8,000,000 generations with tree and parameter sampling occurring every 1,000 generations. The first 25% of trees were discarded as burn-in, leaving 750 trees per run. Posterior probabilities for internal node were calculated from the posterior density of trees. Maximum parsimony (MP) tree was reconstructed by PAUP 4.0 with a bootstrap value of 1,000 repetitions [68] . To investigate whether gene conversion occurred in mammalian IR-IFITM genes, an analysis was performed using the GENECONV program [34] . To detect whether positive selection acted on IFITM family, the CODEML program implemented in PAML 4.2 package was used. The site-specific model was performed by comparing the models M2a (positive selection) and M8 (beta & v) vs. the null models M1a (nearly neutral) and M7 (beta), respectively. Likelihood ratio tests (LRT) of different models were used to find the best fit model for the data [65, 66, 69] . We also used the MEME method and the branch-site REL model implemented in DATAMONKEY (http://www.datamonkey.org/) to confirm the results by PAML analyses [70, 71] . MEME is the latest method to identify PSS and can find signatures of episodic selection even when the majority of lineages are subject to purifying selection. In the phylogenetic tree, IR-IFITM, IFITM5 and IFITM10 genes from different species form three independent clusters. We investigated type I functional divergence among IR-IFITM, IFITM5 and IFITM10 using Diverge v2.0 with the Maximum-Likelihood Estimation (MLE) and Model-Free Method (MFE) [32] . Type I sites represent amino acids conservation in one cluster, but high variability in another, suggesting that these residues have been subjected to different functional constraints. The coefficient of functional divergence, h, ranging from 0 to 1, was used to test the statistical significance of functional divergence that has occurred between different clusters. A null hypothesis of h = 0 indicates that the evolutionary rate is virtually the same between two clusters at each site. When h.0.5, the null hypothesis is considered to be significantly rejected. The important amino acid residues most likely to be responsible for functional divergence were then predicted by calculating the site-specific profile based on posterior analysis for all pairs of clusters with functional divergence. Figure S2 Sequence alignment of the vertebrate IFITM5 genes. Alignment was used to reconstruct the Bayesian tree in Figure 2B . (PDF) Figure S3 Sequence alignment of the vertebrate IFITM10 genes. Alignment was used to reconstruct the Bayesian tree in Figure 2C . (PDF) Figure S4 Sequence alignment of the mammalian IR-IFITM genes. Alignment was used to reconstruct the phylogenetic trees shown in Figure 3 . The abbreviations were adopted from the species names in Table 1

Filovirus Research in Gabon and Equatorial Africa: The Experience of a Research Center in the Heart of Africa

Health research programs targeting the population of Gabon and Equatorial Africa at the International Center for Medical Research in Franceville (CIRMF), Gabon, have evolved during the years since its inception in 1979 in accordance with emerging diseases. Since the reemergence of Ebola virus in Central Africa, the CIRMF “Emerging Viral Disease Unit” developed diagnostic tools and epidemiologic strategies and transfers of such technology to support the response of the National Public Health System and the World Health Organization to epidemics of Ebola virus disease. The Unit carries out a unique investigation program on the natural history of the filoviruses, emergence of epidemics, and Ebola virus pathogenesis. In addition, academic training is provided at all levels to regional and international students covering emerging conditions (host factors, molecular biology, genetics) that favor the spread of viral diseases.

The International Centre for Medical Researches of Franceville (CIRMF) was founded in 1974 by His Excellency El Hadj Omar Bongo Ondimba, President of the Gabonese Republic, and Mr. Pierre Guillaumat, the chairman of the petroleum company, Total Gabon. The Centre was inaugurated on December 5th, 1979 with the participation of numerous internationally eminent scientists (Figure 1) . In the 1990s, viral hemorrhagic fevers became a focus of attention in Equatorial. The decision to develop a high security laboratory for the study of Ebola virus disease came after a 1996 Ebola virus disease outbreak in Mayibout area, Gabon. The main objective was to develop the potential for rapid and specific diagnosis on viral hemorrhagic fevers and to have a backup for field investigation of severe viral hemorrhagic fever epidemics. Because the unique expertise and interaction of the CIRMF team along with the international World Health Organization (WHO) teams for Ebola virus disease, the Gabonese government agreed to such a project. The first BSL3+ (including negative pressure and glove box) laboratory was built in 1997, mostly financed by the Foreign Ministry of France ( Figure 2 ). This laboratory was built in a period of quasi "emergence" of successive Ebola virus disease outbreaks in Gabon, and the plans to upgrade the BSL3+ laboratory were not efficient. A specific research unit was founded in 1998 to study emerging infectious diseases: The Emerging Viral Disease Unit (UMVE). Thanks to work done between 1996 and 1998, both in the field and in research, CIRMF became a National Reference Laboratory and a WHO Collaborating Center in the Equatorial African region. A second high security laboratory for Risk Group 3/4 Agents, mostly funded by the TOTAL Gabon oil company and the Gabonese Government, was built between 2003 and 2008 on CIRMF campus ( Figure 3 ). This BSL-4 laboratory was commissioned by a combined team of experts from the Pasteur Institute of Paris, National Institute of Health and Medical Research, France, and Jean Mérieux P4 laboratory of Lyon. On 45 hectares, the CIRMF campus has a working space of 2,500 square meters composed of a main building, laboratories, service buildings, and living accommodations (Figure 1 ). The present high containment and high security laboratory, operated by the Emerging Viral Diseases Unit (UMVE), is one of 2 laboratories in Africa that can manipulate Risk Group 3/4 Agents (i.e., Ebola, Marburg, and Crimean-Congo hemorrhagic fever viruses). Research, including isolation and characterization of these highly pathogenic viruses is performed in accordance with international rules defined by WHO on the handling Risk Group 3/4 Agents Updated equipment includes a double door autoclave, thermo regulated cabinet, a high security centrifuge system, a conventional photonic microscope, a virus isolation unit, and two independent rooms the can be shut down alternatively after decontamination when necessary. An uninterruptable controlled electrical supply for refrigeration, computer systems, and other systems is ensured by two back-up power plants. The telecommunication network consists of mobile phones and the Internet through a dedicated satellite antenna. Other service units consist of a Primate Center, the Gorilla and Chimpanzee Study Station in Lopé National Park, and the Dienga Health Observatory. Investigators from this observatory conduct field studies on the prevalence of viral and parasitic diseases and their implications for public health [1] Dedicated to medical research, the Primate Center houses more than 450 primates belonging to ten different African species (e.g., chimpanzees (Pan troglodytes), gorillas (Gorilla sp), mandrills (Mandrillus sphinx), guenons (Cercopithecus sp.), collared mangabeys (Cercocebus torquatus), greater spot-nosed monkeys (Cercopithecus nictitans), vervet monkeys (Chlorocebus pygerythrus) and an Asian macaque (Macaca sp.) colony. One of the largest primate centers in Africa, the Primate Center is equipped with level A2 and A3 animal facilities for scientific research protocols. The Great Apes are housed in large open-air facilities. Semi-free living colony of twelve forested hectares harbor about half of the primates at the Primate center. At the Gorillas and Chimpanzee Study Station, researchers study ecological approaches to the emergence of zoonotic diseases, inter-species transmission of pathogens, and disease outbreaks in humans and wild animals [2] . Running costs are funded by the Ministry of Economy, Gabon, the national petroleum company of Total-Gabon, and the Ministry of Foreign and European Affairs, France. Several international agencies participate in a variety of financial supports including scientists' salaries, equipment, research projects, and academic grants (e.g., IRD, WHO, United States Agency for International Development, National Center for Scientific Research, France). Technical laboratory training support of Gabonese teams and other African countries has been one of the major aims of CIRMF. The UMVE actively participates in the academic training at the Regional Graduate School and the different State universities of Equatorial Africa. A special relationship with the "Health Sciences University" of Libreville and the "Sciences and Technology University" of Masuku in Franceville encourages collaborative research projects with teachers and supports students of the Faculties of Medicine and Sciences in the preparation of doctoral theses. CIRMF receives doctoral and post-doctoral scientists from other universities of developed countries (e.g., Bonn, Marseille, Montpellier, and Tübingen Universities). Continuing medical education in the form of post-doctoral workshops are held at the CIRMF for discussion and demonstration of modern techniques. As a National reference laboratory, CIRMF has the following roles: diagnosis of suspected cases during outbreaks of viral hemorrhagic fevers or severe clinical infectious syndromes; development of new methods for diagnosing such infections; surveillance of animal fatalities in reservoir or intermediate hosts; and intervention during outbreaks of unknown etiology. CIRMF diagnosed infections of more than 70 pathogens that could not be identified in other biology laboratories throughout the country. CIRMF maintains close ties to several components of the National Healthcare system, such as Amissa Bongo Regional Hospital in Franceville or the Sino-Gabonese Friendship Hospital. In order to facilitate national and international scientific exchanges including scientists, equipment, biological specimens, the capital of Gabon, Libreville, is part of CIRMF operational system. Hosted by the University of Health Sciences, Libreville, one laboratory is now operational. Tight connections with other scientific teams in Libreville are under development (i.e.,: Military Hospital, Libreville; General Hospital, Libreville; A. Schweitzer Lambarene Foundation). The Emerging Viral Diseases Unit, CIRMF, proposes forming a research partnership to study infectious diseases transmitted by animals of the tropical rain forests regions of Equatorial Africa. The proposed partnership builds on such existing collaborations of several years between the major research centers of 2 other French speaking Equatorial African countries, namely the National Public Health Laboratory in Brazzaville, Republic of the Congo, and the Institute for Biomedical Research in Kinshasa, Democratic Republic of Congo. An international partnership with the IRD, Marseille, France, and the Institute of Virology, Bonn, Germany, will assist in the development of this regional partnership. With the WHO Regional Office, Brazzaville, Republic of the Congo, the UMVE-CIRMF field team participates along with other international partners (e.g., Centers for Disease Control and Prevention, USA; Laboratory Centre for Disease Control and National Microbiology Laboratory, Winnipeg, Canada; P4 Laboratory of the National Institute for Communicable Diseases, South Africa) to respond to all Ebola virus disease epidemics in Africa. CIRMF aims to use laboratory and field expertise be a regional focal resource in conjunction with local health authorities to organize epidemic responses. CIRMF expertise is also offered from by entering into laboratory-to-laboratory agreements. Also, an informal international laboratory network for the diagnosis and surveillance of severe infectious clinical syndromes includes: the Institut National de Recherche Biomédicale, Democratic Republic of the Congo; Laboratoire National de Santé Publique de Brazzaville, Republic of the Congo; Metabiota/Laboratoire des Maladies Emergentes, Yaoundé, Cameroon; Pasteur Institute, Bangui, Central African Republic; Institute of Virology, Bonn University, Germany; and P4 Jean Mérieux Lyon, France. Exchange of materials, equipment, and personnel is facilitated through memorandums of understanding. CIRMF holds more than 250,000 samples of various origins in a biological repository, which is accessible to the International scientific community. The UVME assists the National Public Health System in consolidating and formalizing microbiological monitoring of the Equatorial African sub-region. Ultimately, CIRMF will be positioned as a Center of Excellence for microbiological surveillance and research in a global network. Developing diagnostic tools and strategies is the main driver to improve surveillance and research of emerging viral diseases. A strategic choice was made to link syndromes to an etiological agent, including hemorrhagic syndromes. To isolate and diagnose highly pathogenic viruses, a progressive and diversified methodology was applied. The first approach used real-time virus-specific PCR (qRT-PCR). If the first approach was not successful, conventional RT-PCR was implemented using degenerate consensus primers targeting conserved regions of the genome. Ultimately, random amplification of nucleotide sequences was directly applied to the original biological material (DNA chip re-sequencing, meta-genomic pyrosequencing (454 Life Sciences, Branford, CT)). UMVE studied the factors implicated in the three steps that led to Ebola virus and Marburg virus diseases emergence in humans. These steps include: the identification of reservoir species, the circulation within the natural host, the crossing to intermediary animal species, and finally the direct transmission to humans from great apes and fruit bats. Antibodies and nucleotide sequences specific for Ebola virus were detected in the liver and spleen of fruit bat belonging to three species (Hypsignathus monstrosus, Epomops franqueti, Myonycteris torquata) in Gabon and Republic of the Congo (Figure 4 ). Antibodies and nucleotide sequences specific for Marburg virus were found in the Egyptian fruit bat (Rousettus aegyptiacus) in Gabon, suggesting that bats might be reservoirs for filoviruses [3] [4] [5] . We showed that Ebola virus caused extensive epizootics among gorillas and chimpanzees, killing thousands of animals during the last decade in parts of Gabon and Republic of the Congo [4] . We characterized the viral variants associated with all Ebola virus disease outbreaks that occurred between 1996 and 2008 and developed new epidemiological models of Ebola virus disease epidemics, based on the identification of several independent epidemic chains. The identification of multiple variants during the 2001 Gabon/Republic of Congo outbreak and two phylogenetically divergent lineages suggest independent introductions into great ape and human populations following multiple viral spillovers from a reservoir host [6] [7] . In this "multi-emergence" hypothesis, Ebola virus disease outbreaks would occur episodically during certain ecological conditions caused by habitat disturbances or climatic phenomena. This hypothesis also implicitly assumes that Ebola virus was present in Equatorial Africa long before the first documented disease outbreak in 1976, as supported by various serological surveys. Furthermore, we recently showed that the 2007 Luebo outbreak in the Democratic Republic of the Congo was linked to massive fruit bat migration, strongly suggesting that humans could be infected directly by bats or by consumption of bats [8] . In the study of immunological mechanisms of Ebola virus disease humans, we showed that fatal infection is associated with aberrant innate immunity and global suppression of adaptive immunity [, [9] [10] [11] [12] . The innate immune reaction is characterized by a 'cytokine storm', with a hyper secretion of numerous pro-inflammatory cytokines, chemokines and growth factors, and by the noteworthy absence of antiviral interferon (IFN)-α [9, 13, 14] . Immunosuppression of adaptive immunity is characterized by very low levels of circulating cytokines produced by T lymphocytes and by massive loss of peripheral CD4 and CD8 lymphocytes, probably through Fas/FasL-mediated apoptosis. Finally, we hypothesized that a viral protein with super-antigen activity might be involved in the massive T cell apoptosis [15] . In striking contrast with fatal outcome, effective control of Ebola virus infection is associated with balanced immune responses in survivors. Asymptomatic Ebola virus infection was demonstrated in humans during the 1996-97 disease outbreak in Gabon [16] . Asymptomatic infection was associated with an early strong inflammatory response that may be involved in the early inhibition of viral replication [16, 17] . Consistent with this discovery, we showed a decade later that a large fraction of the human population living in forested areas of Gabon has both humoral and cellular immunity to Ebola virus [18] . In the absence of identified risk factors, the high prevalence of 'immune' individuals suggests a common source of human exposure such as fruits contaminated by bat saliva. Initially focused on Ebola virus disease, UMVE science policy was redirected since 2007 by expanding the main research themes to other emerging viral diseases that could threaten public health in the Congo basin (Table 1) [19] [20] [21] [22] . CIRMF is geographically isolated from the capital of Gabon, Libreville. The Libreville office is essential to the Franceville headquarters as it coordinates visits from staff on field missions, and forwards imported equipment to headquarters. The capital is accessible by a 12-hour ride in a four-wheel drive vehicle or in a train (three times/week schedule) covering 641 km. Due to tropical weather, four weekly domestic plane rotations often fly on an inconsistent schedule. Ultimately CIRMF needs to be largely autonomous in term of electrical power (i.e.,: unexpected fuel supply disruption), cold chain with the necessity to maintain in situ a unit of liquid nitrogen production (repository), and purified water supply. CIRMF is uniquely suited to study infectious diseases of the Congolese tropical rain forest, the second world's largest rain forest. As a central point of a North-South transect of the rain forest, the Center is able to study the biodiversity of Africa including animal species, microbes, and parasites. CIRMF is dedicated to conduct medical research of the highest standard, and is the only facility of its type in Equatorial Africa. With unrivalled infrastructure, multiple sites, and multidisciplinary teams, the Center promotes a modern healthcare system in Gabon. CIRMF teams are engaged in trans-disciplinary projects bringing together specialists from the health sciences, biological sciences, veterinary medicine, conservation, the humanities, and environmental sciences. The Center welcomes partnerships from around the world to work on global human health issues.

Reproductive Number and Serial Interval of the First Wave of Influenza A(H1N1)pdm09 Virus in South Africa

BACKGROUND/OBJECTIVE: Describing transmissibility parameters of past pandemics from diverse geographic sites remains critical to planning responses to future outbreaks. We characterize the transmissibility of influenza A(H1N1)pdm09 (hereafter pH1N1) in South Africa during 2009 by estimating the serial interval (SI), the initial effective reproductive number (initial R(t)) and the temporal variation of R(t). METHODS: We make use of data from a central registry of all pH1N1 laboratory-confirmed cases detected throughout South Africa. Whenever date of symptom onset is missing, we estimate it from the date of specimen collection using a multiple imputation approach repeated 100 times for each missing value. We apply a likelihood-based method (method 1) for simultaneous estimation of initial R(t) and the SI; estimate initial R(t) from SI distributions established from prior field studies (method 2); and the Wallinga and Teunis method (method 3) to model the temporal variation of R(t). RESULTS: 12,360 confirmed pH1N1 cases were reported in the central registry. During the period of exponential growth of the epidemic (June 21 to August 3, 2009), we simultaneously estimate a mean R(t) of 1.47 (95% CI: 1.30–1.72) and mean SI of 2.78 days (95% CI: 1.80–3.75) (method 1). Field studies found a mean SI of 2.3 days between primary cases and laboratory-confirmed secondary cases, and 2.7 days when considering both suspected and confirmed secondary cases. Incorporating the SI estimate from field studies using laboratory-confirmed cases, we found an initial R(t) of 1.43 (95% CI: 1.38–1.49) (method 2). The mean R(t) peaked at 2.91 (95% CI: 0.85–2.91) on June 21, as the epidemic commenced, and R(t)>1 was sustained until August 22 (method 3). CONCLUSIONS: Transmissibility characteristics of pH1N1 in South Africa are similar to estimates reported by countries outside of Africa. Estimations using the likelihood-based method are in agreement with field findings.

During 2009, the emergence and worldwide spread of influenza A(H1N1)pdm09 (pH1N1) was observed [1] . While a rapid and timely estimation of the transmission parameters of this novel virus played an important role in informing transmission potential and mitigation interventions during the 2009 pandemic period, the post-pandemic documentation of these parameters is equally important as many previous estimates were established from analyses conducted during the early stages of epidemics and often from preliminary data [2, 3] . Additionally enhancing our knowledge of past pandemics assists in providing greater insight to prepare and respond in future outbreaks. Four key measures are typically used to describe the transmissibility of an infectious disease. First, the serial interval (SI) describes the mean time between illness onset of two successive cases in the chain of transmission. Second, the secondary attack rate (SAR) describes the proportion of susceptible contacts that acquire infection from an infectious person. Third, the basic reproductive number (R 0 ) is defined as the average number of secondary cases per primary case in an idealised entirely susceptible population in the absence of control measures. Finally, the effective reproductive number (R t ) at any given time point represents the actual average number of secondary cases per primary case observed in a population. R t reflects the impact of control measures and the depletion of susceptible persons over time. The initial R t may approximate R 0 in pandemic situations. [2] [3] [4] [5] . Previously published estimates of pH1N1 transmission parameters vary by study setting and methods employed. The majority of studies found the mean SI of pH1N1 to range from 2.5-3.3 days [2, [6] [7] [8] [9] [10] [11] ; however, Canada and Texas reported a longer SI of 4-5 days, respectively [12, 13] . Estimates of the R 0 of pandemic influenza from the USA range from 1.3-2.3 [2, 9, 11] . Estimates from Mexico range from 1.4-2.9 [2, 14, 15] . Outside of North America, R 0 estimates include: Australia (mean 2.4) [16] , Canada (mean 2.62) [12] , Thailand (mean 2.07) [17] , Peru (range 1.2-1.7) [18] and New Zealand (mean 1.96) [19] . Finally, Japan revised their mean R 0 estimates from 2.3 to 1.28 after repeating analyses later in the pandemic [20] ; thus demonstrating a need to revisit revised and more complete datasets. A variation in R t with progression of the pandemic was observed in Mexico, averaging at 1.47 (based on a negative binomial model) [14] , but peaking between 2.1-4.0 depending on the generation interval chosen [21] . In a previous work, we estimated the SAR and SI of pH1N1 among the first 100 cases detected in South Africa by prospectively examining virus transmission between household contacts [22] . We found a SAR of 10% and a mean SI of 2.3 days (SD 61.3, range 1-5) between successive laboratory-confirmed cases in the transmission chain. When additionally including suspected secondary cases into the analysis, the SAR increased to 17% and the SI to 2.7 days (SD 61.5, range 1-6). In this work we incorporate data collected on all laboratory-confirmed cases detected during the 2009 pH1N1 epidemic in South Africa with the aim of describing the transmissibility characteristics (initial R t and temporal variation of R t ) of the epidemic in the country and compare its dynamics with those observed in other countries in the same year. During 2009, the National Institute for Communicable Diseases (NICD), of the National Health Laboratory Service (NHLS), South Africa, maintained a central registry of all pH1N1 laboratory-confirmed cases detected throughout the country. The methodology of collating this data has previously been described in detail [23] . Briefly, we collated individual case-based data from all laboratories offering pH1N1 testing throughout South Africa, which included patient age, sex, dates of illness onset and specimen collection, and the administrative location (province) of the healthcare facility where the patient presented. Testing was performed by accredited laboratories, including: the National Influenza Centre (NICD-NHLS), NHLS public-sector laboratories or private-sector laboratories. All testing laboratories performed detection and characterisation of pH1N1 virus by real-time PCR by either the protocol developed by the WHO Collaborating Centre for Influenza, U.S. Centers for Disease Control and Prevention [24] , or using commercially available kits. Wherever the date of symptom onset was missing, we estimated it from the date of specimen collection using a multiple imputation approach. Firstly, we modelled the lag time from date of symptoms onset to date of specimens collection from cases with complete data via a Poisson regression model using predictors significant at p,0.05. The covariates assessed in the model were patient age, gender, province, date of specimen collection, and collection of a specimen on a weekend day (i.e. Saturday or Sunday). Secondly we obtained an estimated lag-time for each observation with missing date of symptoms onset using a random sampling process from a Poisson distribution centred on the predicted value from the Poisson regression model. A Poisson distribution was selected to model count data. Thirdly we imputed missing dates of symptoms onset by subtracting the estimated lag-time from the date of specimen collection. The imputation process was repeated 100 times for each missing value, creating 100 datasets with information on the onset date (imputed or observed) for 12,630 laboratory-confirmed cases. We based the estimation of initial R t and temporal variation of R t on date of symptoms onset (observed and imputed). In all analyses we modelled the SI via a multinomial distribution. When estimating initial R t , we focus our analysis on the exponential growth phase of the epidemic in South Africa (i.e. the period from the first occurrence of five consecutive days with confirmed cases reported to the epidemic peak). The parameters were estimated using three methods: Method 1. We make use of the likelihood-based method for the simultaneous estimation of initial R t and the SI described by [25] . This method is well suited for estimation of initial R t and SI in real-time with observed aggregated daily counts of new cases, denoted by N = (N 0 , N 1 …,N T ) where T is the last day of observation and N 0 are the initial number of seed cases that begin the outbreak. The N i are assumed to be composed of a mixture of cases that were generated by the previous k days, where k is the maximal value of the serial interval. We denote these as X j , the number of cases that appear on day i that were infected by individuals with onset of symptoms on day j. We assume that the number of infectees generated by infectors with symptoms on day j follows a Poisson distribution with parameter R t N j . Additionally, X j = (X j,j+1 , X j,j+2 …,X j,j+k+1 ), the vector of cases infected by the N j individuals, follows a multinomial distribution with parameters p, k and X j . Here p is a vector of probabilities that denotes the serial interval distribution. Using these assumptions, the following likelihood is obtained: where m i~Rt ( P k j~1 p j N i{j ). Parameter estimates are obtained using maximum likelihood methods. For this method we used 6 days as the maximal value of the SI (k), which is consistent with the length of the SI observed in field investigations in South Africa [22] . In addition we implemented a sensitivity analysis to assess the variation of the initial R t estimates vis-à-vis k values of 4 days and 8 days, respectively. Method 2. We assume a known distribution of the SI in South Africa and we estimate the initial R t using the maximum likelihood estimator for known SI described by [9, 25] . The estimator of initial R t in this case is a modification of Method 1 and is given by: For this analysis we use the two SI distributions observed from investigations of the first 100 pH1N1 cases in South Africa [22] : (1) the SI distribution between primary cases and laboratoryconfirmed secondary cases only (39%, 24%, 14%, 17%, 3% and 3% for day 1 to 6 respectively), and (2) the SI distribution between primary cases and suspected plus laboratory-confirmed secondary cases (30%, 17%, 20%, 23%, 7% and 3% for day 1 to 6 respectively). We consider suspected secondary cases, individuals that developed ILI symptoms within 14 days from the symptom onset of a confirmed index case within the same household. Method 3. We make use of the Wallinga and Teunis' method for estimation of R t from the imputed data [26] . This method uses the daily case counts of cases and assumes the serial interval is known. We make the same assumptions for the serial interval as in method 2. The method calculates the relative probability a case on day i infects a case on day j as: Table 1 . Observed lag-time between date of symptom onset and date of specimen collection, incidence rate ratio (IRR) and significance value of the covariates significant in the Poisson regression model. where p k is the probability of a serial interval of length k. Then the estimate for the reproductive number for case i, is: This method requires that we make use of the entire epidemic curve. We calculate R t as the average of the R i when i is in the epidemic period, as previously defined. Estimates are reported as the means across the 100 imputations. For all estimates, we calculate bootstrap confidence intervals as has been described previously [9, 26] . We combine the results from all 100 imputations to obtain a confidence interval that incorporates both imputation error, as well as random error [27] . All analyses were performed using R version 2.14. 12,630 laboratory-confirmed pH1N1 cases were captured by the South African central registry during 2009. The overall demographic, spatial and temporal distribution of these cases has been previously described [23] . Data on date of symptom onset was available for 758 (6%) cases and date of specimen collection for 12,500 (99%) cases. The first case reported illness onset of June 12, 2009 The lag-time between symptom onset and specimen collection was significantly associated with the provincial location of specimen collection, as well as the collection of a specimen on a weekend day (Table 1) . We used these two covariates in the multiple-imputation to predict the date of symptom onset where missing for all cases (Figure 1 ). Other available variables, including date of specimen collection (period during the epidemic), patient age and sex were not significantly associated with the lag-time between symptom onset and specimen collection and, therefore, not included in the final model. Analyses to simultaneously estimate initial R t and serial interval, and estimate initial R t given a known serial interval, were performed over the exponential growth phase of the epidemic from June 21 to August 3, 2009. Using the likelihood-based method to simultaneously estimate initial R t and the SI across 100 imputations of the dataset (Method 1), we estimated aR R t of 1.47 (95% CI: 1.30-1.72) and a mean SI of 2.78 days (95% CI: 1.80-3.75) (Figure 2 ).R R t estimates ranged from 1.31 (95% CI: 1.21-1.48) to 1.54 (95% CI: 1.37-2.03) when the maximal value of the SI ranged from 4 to 8 days. We first utilised the SI established from the aforementioned field investigations of the initial 100 cases in estimating R t , as described in method 2. When performing the analysis using the SI distribution observed for laboratory-confirmed pH1N1 secondary cases only (mean 2.3 days, SD 61.3, range 1-5) [22] , we found an initialR R t of 1.43 (95% CI: 1.38-1.49) ( Figure 3A) . When performing the analysis using the SI distribution observed for both confirmed and suspected secondary cases (mean 2.7 days, SD 61.5, range 1-6) [22] , we found an initialR R t of 1.49 (95% CI: 1.44-1.55) ( Figure 3B ). Figure 4 shows the variation inR R t with the progression of the outbreak over time. We observed relatively highR R t values following the introduction of pH1N1 virus into South Africa, corresponding to high rates of transmission and exponential growth of the local epidemic during this period.R R t peaked on the first day of the epidemic growth period (June 21) at 2.91 (95% CI: 0.85-3.99).R R t began to drop from July 27 onward and remained consistently below one after August 22. This corresponds with the decline in the daily incidence of new cases detected. Averaging the R t values obtained during the epidemic growth period (June 21 to August 3, 2009), we estimate initial R t to be 1.42 (95% CI: 1.20-1.71). Utilising temporal data on illness onset and specimen collection, and the epidemic curve derived from these data, we provide estimates of the transmissibility parameters of pH1N1 during the first wave experienced in South Africa. Our results focus primarily on the use of analytical techniques to estimate initial R t and SI without incorporating contact tracing or household transmission studies. However, when parameters from field studies are available, we show that these can be incorporated to provide robust estimates of transmission parameters. We found that initial R t estimates established using the likelihood-based method for the simultaneous estimation of R t and SI (method 1: initialR R t : 1.47, SI: 2.78 days) are in agreement with those obtained using SI observed in field investigations [22] (method 2: initialR R t : 1.43 and 1.49 using observed SI for laboratory confirmed or laboratory confirmed and suspected cases respectively). In addition, the mean SI estimate obtained with method 1 (2.78 days) is in agreement with field findings (SI: 2.3-2.7 days using observed SI for laboratory confirmed or laboratory confirmed and suspected cases respectively). Previous estimates of initial R t and the mean SI for pH1N1 have ranged between 1.3-2.9 and 2.5-3.3 days, respectively [2, [6] [7] [8] [9] [10] [11] [14] [15] [16] [17] [18] [19] . Our estimates are consistent with these findings, regardless of the method used for the analysis and despite difference in climate, demography and health systems across these countries. It appears that once established, the transmission characteristics of pH1N1 are very consistent. Differences in transmission rates may occur within smaller subgroups of the overall population; however, this has not been well-studied. Previous estimates of the epidemiological parameters of seasonal influenza epidemics found a SI = 2-4 days [28] [29] [30] , and a R t a little over 1 with slight variation between climates; R t = 1.03 in Brazil [31] versus R t = 1.1-1.3 in more temperate climates [32] . A number of studies have retrospectively estimated the transmissibility of influenza pandemics. During the 1918 Spanish influenza A(H1N1) pandemic, when assuming a SI = 4 days, R 0 estimates range from 2.0-4.3 in community settings [33, 34] , and even higher values (R 0 = 2.6-10.6) in confined settings such as ships and prisons [34] . A separate analysis predicted a slightly lower SI of 3.3 in community settings and a SI of 3.81 in confined settings during the 1918 pandemic, and subsequently estimated R 0 values of 1.34-3.21 and 4.97 in these respective settings [35] . R 0 estimates from the 1957 Asian influenza A(H2N2) pandemic range from 1.65-1.68 [36, 37] . During the first wave of the 1968-1969 Hong Kong influenza A(H3N2) pandemic, estimates of R 0 range from 1.06-2.06 and increased to 1.21-3.58 during the second wave [38] . Given our findings, the overall transmissibility of pH1N1 in South African during 2009 was more similar to that of seasonal influenza strains than the 1918 pandemic, and comparable to lower end estimates of the latter pandemics. However, by showing variation in transmissibility with time, we demonstrate that shortly after introduction of pH1N1 into the country, transmission of the virus reached anR R t of 2.9, resulting in exponential growth of the local epidemic and widespread illness. Nonetheless, we show that after a period of less than 2 months of heightened transmission,R R t dropped below 1, corresponding to a decline in the incidence of new cases; likely a result of a combination of herd immunity, There are several limitations in this analysis which merit discussion. First, we assume that all cases are known and reported. It has been shown previously that, if cases are not reported, this may bias estimates generated using this method [39] . If the proportion of cases reported remains consistent over the study, then the estimates of transmissibility will not be biased; however, if the reporting fraction varies through time, then biased estimates of the reproductive number and serial interval may result. Likewise, variation in case ascertainment with time may bias our estimates of the temporal variation of R t . Generally higher reporting rates may be anticipated in the early phase, with reporting fatigue later becoming a factor. Secondly, data for this study are derived from laboratory-based surveillance data from several regions across South Africa; a large and diverse country. Our findings do not incorporate heterogeneities (such as spatial and demographic differences) that likely exist in transmission patterns, or assess the degree to which these impact aggregate measures of initial R t . Methodologies that incorporate heterogeneities inherent in public health data warrant further study. Despite these limitations, the post-pandemic estimates presented here add to the body of knowledge of pH1N1 transmissibility parameters, which were previously dominated by estimates from developed nations and often based on preliminary data. It remains important that revised parameters, from complete datasets and diverse geographies, are incorporated into planning mitigation strategies for future pandemics. Nonetheless, the methods used in this study would be adaptable to generating real-time estimates during future epidemics. As we continue to build epidemiological capacity in developing nations, including South Africa, we must keep in mind the need for rapid assessments of transmissibility of novel pathogens, in addition to disease severity, to better inform public health interventions.

Measuring healthcare preparedness: an all-hazards approach

In a paper appearing in this issue, Adini, et al. describe a struggle familiar to many emergency planners—the challenge of planning for all scenarios. The authors contend that all-hazards, or capabilities-based planning, in which a set of core capabilities applicable to numerous types of events is developed, is a more efficient way to achieve general health care system emergency preparedness than scenario-based planning. Essentially, the core of what is necessary to plan for and respond to one kind of disaster (e.g. a biologic event) is also necessary for planning and responding to other types of disasters, allowing for improvements in planning and maximizing efficiencies. While Adini, et al. have advanced the science of health care emergency preparedness through their consideration of 490 measures to assess preparedness, a shorter set of validated preparedness measures would support the dual goals of accountability and improved outcomes and could provide the basis for determining which actions in the name of preparedness really matter.

Despite years of planning and billions of dollars spent on disaster preparedness and response activities worldwide, the science of preparedness is in its infancy. The empirical evidence for much of health emergency preparedness is scant. As a result, it is challenging to define what it means to be "fully prepared." Collectively, the world has dealt with numerous disasters, but only a handful of nations have confronted many. This has made it challenging to convincingly link the structures and processes for health preparedness to outcomes, particularly mitigation of morbidity and mortality. Israel, in part because it has confronted numerous mass casualty events, has been at the forefront of medical preparedness planning, and has developed sophisticated structures and processes for dealing with such medical emergencies. As described by Adini, et al. [1] , Sarpy, et al. [2] , and Einav, et al. [3] , elements of this system include standard operating procedures, drills and exercises for most conceivable events, and measures to inform continuous improvement following each drill, exercise, and actual event. Given the state of the evidence, these authors have relied on a rigorous use of expert opinion to develop their measures; many of the experts involved have had frequent and direct response experience. In a paper appearing in this issue, Adini et al. [4] describe a struggle familiar to many planners-the challenge of planning for all scenarios. They contend that allhazards, or capabilities-based planning is a more efficient way to achieve general health care system preparedness than scenario-based planning. The authors describe a systematic investigation of the components of their preparedness system that impact hospital preparedness. The paper represents an important advance on several fronts. First, it describes a system of measurement which is consistently applied to assess and improve the preparedness of hospitals. Second, it uses thoughtful analytic methods to answer the question of whether an all-hazards approach is an appropriate methodology for preparedness. Supporting this, the authors found moderate to strong correlations between preparedness measures for various kinds of disasters-including mass casualty, toxicologic, and biological events. In other words, the core of what is necessary to plan for and respond to one kind of disaster (e.g. biologic event) assists with for planning and responding to other types of disasters. This finding allows for improvements and greater efficiencies in planning and time-consuming and expensive drills and exercises; an important consideration given the current fiscal climate in countries worldwide. Third, the authors found that SOPs, training and drills made more of a contribution to overall preparedness than equipment and preparedness knowledge of personnel-an important observation that facilitates effective resource allocation. The authors' findings are consistent with recent releases of a US Presidential Policy Directive [5] , and related guidance from the US Department of Health and Human Services [6, 7] , which shift the focus of preparedness planning in the US toward an all-hazards, capabilities-based approach. One challenge faced by most countries, including Israel and the US, is finding a parsimonious set of measures to assess and improve preparedness. While Adini, et al., initially considered 490 measures to assess preparedness, we recognize that countries and systems may not be capable of consistently deploying and analyzing such a large volume of measures. Interestingly, the findings by Adini, et al., support a shift to an all-hazards approach which, enables a reduction in the number of measures. Future work should examine the correlations between measures, and could make use of techniques such as factor analysis to identify a short set of measures, or even scales, that could serve as proxys for a longer, more complex measurement set. This paper does not address another critical issuemeasuring the preparedness of the public health system. As public health and medical care are so interdependent, we hope that similarly rigorous work regarding the public health system is ongoing. Adini, et al., have advanced the measurement of health care emergency preparedness. With limited resources, the necessity to find commonalities of approach and efficiency in all we do, including measurement, is critical. In the end, a set of validated preparedness measures would support the dual goals of accountability and improved outcomes and could provide the basis for determining which actions in the name of preparedness really make a difference.

Presentation of hemophagocytic lymphohistiocytosis due to a novel MUNC 13–4 mutation masked by partial therapeutic immunosuppression

Hemophagocytic lymphohistiocytosis is a potentially fatal disease characterized by excessive macrophage and lymphocyte activity. Patients can be affected following immune activation after an oncologic, autoimmune or infectious trigger. An associated gene mutation may be found which impairs cytolytic lymphocyte function. We describe a pediatric case of hemophagocytic lymphohistiocytosis with a novel mutation of MUNC 13–4 whose diagnosis was confounded by concurrent immunosuppression. Clinical reassessment for hemophagocytic lymphohistiocytosis is necessary in persistently febrile patients with laboratory derangements in the setting of immunosuppressive agent exposure.

Hemophagocytic lymphohistiocytosis (HLH) is a clinical syndrome of abnormal immune activation causing excess inflammation. There is increased ectopic migration and proliferation of T cells, tissue infiltration by activated macrophages (histiocytes), hyper-activation of lymphocytes, and prolonged release of pro-inflammatory cytokines [1] . This leads to uncontrolled inflammation manifesting as fever, cytopenias and organ dysfunction. The syndrome is fatal without prompt symptom recognition and treatment. Primary HLH, also known as familial HLH, is often considered a pediatric disease. Affected individuals have mutations in genes leading to impaired lytic activity of lymphocytes, including NK cells and cytotoxic T lymphocytes [2] . These individuals are at risk of developing HLH. Mutations are inherited in an autosomal recessive pattern, but can also result from de novo mutations, making family history less reliable for diagnosis. Acquired HLH, also known as secondary HLH, occurs more commonly in adults. Nonetheless, it is also described in children. There are presently no identifiable genetic mutations linked to this phenotype. Currently, secondary HLH is attributed to an abnormal immune response triggered by an infectious, oncologic or autoimmune etiology. As such, other names for secondary HLH include Virus-Associated Hemophagocytic Syndrome and Malignancy-Associated Hemophagocytic Syndrome [3] . The diagnosis of HLH is based on clinical criteria. The most recent diagnostic guidelines were revised in 2004 [4] . HLH is diagnosed if either a molecular diagnosis consistent with HLH is made, or five of the eight following diagnostic criteria are met: 1.) fever; 2.) splenomegaly; 3.) cytopenias affecting at least two of three lineages in the peripheral blood (haemoglobin <90 g/L, platelets <100 × 10 9 /L, or neutrophils <1 × 10 9 /L); 4.) hypertriglyceridemia (fasting triglycerides ≥ 3 mmol/L (≥ 265 mg/ dL)) and/or hypofibrinogenemia (fibrinogen ≤ 1.5 mg/dL); 5.) hemophagocytosis in bone marrow, spleen, or lymph nodes (excluding signs of malignancy); 6.) low or absent NK-cell activity; 7.) hyperferritinemia (ferritin >500 μg/L); and 8.) high levels of sIL-2R (sIL-2R ≥ 2400 U/ml). Here we present a pediatric case of HLH with a novel mutation in MUNC 13-4 whose diagnosis of HLH was confounded by low dose treatment with immunosuppressive agents thereby complicating her clinical picture. Our patient is a 3 year-old female who was previously healthy. Four months prior to hospitalization she began experiencing fatigue, recurrent fevers, progressive muscle weakness, and behavioral changes. The week prior to admission, she had decreased urine output, increased abdominal girth and respiratory distress. She was admitted to a community hospital for possible pneumonia. After 3 days, she was transferred to a tertiary center for oncologic assessment. There, her physical examination was concerning for a tender right axillary lymph node and hepatosplenomegaly. A chest radiograph demonstrated diffuse airspace opacities. PCR studies identified both rhinovirus and mycoplasma in nasal pharyngeal swabs. Twenty-four hours after arrival she developed hypoxemic respiratory failure requiring tracheal intubation and mechanical ventilation. This was complicated by an aspiration event and subsequent cardiopulmonary arrest, requiring resuscitation. Multiple subspecialties convened to facilitate an underlying diagnosis. HLH was considered given her hepatosplenomegaly, cytopenias, elevated CRP and normal ESR in the context of documented infection. Table 1 presents key HLH laboratory markers ordered at onset and their results. MAS, possibly associated with sJIA, was also considered despite the absence of arthritis, because of an intermittent, pink, net-like rash in the preceding weeks. Bone marrow evaluation was performed soon after her admission and demonstrated only rare hemophagocytes, similar to those seen in children with an underlying infection [5] . Over the next month our patient continued to be dependent upon mechanical ventilation and developed acute respiratory distress syndrome (ARDS). Other complications included respiratory syncitial virus (RSV) infection and central line-related deep vein thrombosis requiring a six-week course of heparin sulfate. A brain MRI showed diffuse patchy white matter lesions most prominent in the right frontal lobe, suggestive of ischemia. Concern arose for adrenal insufficiency and she received several courses of stress dose corticosteroids with intermittent tapers. Intermittent episodes of fever and worsening cytopenias continued over the next month and HLH, possibly MAS, remained a consideration. For this reason an 18-day trial of anakinra was undertaken. However, fevers continued and abnormalities in liver enzymes along with hematologic abnormalities continued. At this point, genetic testing results became available and identified two mutations in MUNC13-4: a known pathogenic mutation 1389(+1) G > A, a splice donor site of intron 15, and a second previously unidentified mutation, 1847 A > G. This second mutation was located in the splice donor site of exon 20, and the A > G change at this position would likely cause splicing error. It was unclear if this represented a compound heterozygous mutation or if the two variants were on a single chromosome. Parental testing was initiated. Given the genetic results and her persistent symptoms, the laboratory and bone marrow evaluations for HLH were reconsidered. Investigations relevant to HLH before, during and after anakinra use were evaluated (Tables 1 and 2) . When compared to the initial assessment, there was subsequent marked reduction in natural killer (NK) cell function, CD107a upregulation and an elevation in soluble IL-2 (soluble CD25) receptor. Importantly, a drop in ANC ( Figure 2a ) and platelet count (Figure 2b ) was identified during the end of the anakinra trial. Children with active HLH have been noted to have elevated white cell and platelet counts initially that decrease over time [6] , as illustrated in our case. As a confounding feature, however, cytopenia is an uncommon but known side effect of anakinra. It is thus possible that addition of anakinra in our patient, with underlying MUNC 13-4 mutations, magnified the drop in platelet count and ANC that would have been seen with HLH alone. In addition, when ferritin was measured on the 15 th day of anakinra therapy, the level was 629 ng/mL (above the aforementioned acceptable limit of 500 ng/mL), whereas one day after anakinra treatment was terminated, the ferritin level increased to 7129 ng/mL ( Figure 2c ). IL-1 is known to increase the synthesis of ferritin subunits in vitro [7] . We hypothesize, therefore, that anakinra administration inhibited the surge of ferritin that would have otherwise been identified in association with clinical progression of HLH. A repeat bone marrow biopsy and the first and second bone biopsies were compared ( Figure 2 and Figure 3 ). The first and second bone marrow biopsies were also critically compared (Figure 3) . The latter showing a substantive increase in hematophagocytes compared to the former. In addition, a lumbar puncture showed an elevated protein level (109 mg/dL) with a lymphocytic pleocytosis (12 WBC with 100% lymphocytes). These findings suggested central nervous system HLH involvement. Table 1: Initial laboratory values Table 2 : Evolution of laboratory values Figure 1 : Illustration of key laboratory value evolution Figure 2 : Bone marrow aspirate; Wright stain postanakinra Figure 3 : CD 163 immunohistochemical stain of bone marrow biopsy With this collective evidence, the HLH-2004 clinical protocol was initiated [4] . Our patient was treated with dexamethasone and etoposide. Cyclosporine was not initially utilized because of concern of abnormal kidney and liver function, but was added later as she improved. Parental genetic testing was performed in order to determine if her case was likely primary HLH with familial genetic etiology. Testing confirmed that each parent carried one of the MUNC 13-4 mutations found in the patient. One parent possessed the novel variant identified while the other carried the known splice site mutation thus defining the patient as a compound UNC13D heterozygote consistent with FHL3. Our patient was hospitalized for almost 4 months prior to transfer to a rehabilitation institution. She received a tracheostomy tube prior to discharge because of her inability to be weaned from respiratory support. She developed mild hypertension which was attributed to cyclosporine. Her liver enzymes continue to decrease although she has a persistently enlarged liver and spleen. She has not had any recent fevers and has an age-appropriate neurologic exam. However, cognitive limitations cannot be predicted. An unrelated HLA-matched donor has been identified and she will be proceeding to bone marrow transplant. Hemophagocytic lymphohistiocytosis is a potentially fatal disease characterized by excessive macrophage and lymphocyte activity. The onset of HLH in a susceptible individual typically follows either a documented or presumed viral infection. Presently, HLH is often classified into primary (familial) and secondary (acquired) HLH. The incidence of primary HLH is approximately 1:50,000 live born children [8] . However, HLH incidence varies in clinical studies, most likely due to a difference in prevalence across ethnic groups and/or emerging awareness [9] . Although HLH often presents between birth and 18 months of age, onset in older age groups is possible and has been shown to be a feature of particular gene mutations [10] . Familial HLH has a median survival of less than 2 months after diagnosis if it remains untreated [9] . It is likely that many cases are misdiagnosed as severe fatal infection. In both primary and secondary HLH, Epstein Barr virus is the most commonly identified inciting infection, although cases associated with cytomegalovirus, human herpes virus 8, influenza, parvovirus, enterovirus and human immunodeficiency virus have also been described [11] . In secondary HLH, associated malignancies include neuroblastoma, Non-Hodgkin's lymphoma and Langerhans' cell histiocytosis. Macrophage activation syndrome (MAS) can be considered a form of secondary HLH syndrome associated with autoimmune disease. MAS complicates an estimated 10% of Systemic Juvenile Idiopathic Arthritis (sJIA) cases [12] , and approximately 1-5% of Systemic Lupus Erythematosus (SLE) cases [13] . Clinical features of MAS similar to primary HLH include high unremitting fever, hepatosplenomegaly, hepatic dysfunction, lymphadenopathy, encephalopathy, cytopenia, elevated ferritin and coagulopathy [12, 14] . Laboratory similarities include depressed natural killer (NK) cell cytotoxic function, elevated soluble IL-2 receptor levels and soluble CD163. In its early stages, MAS can be a diagnostic challenge due to the overlap of symptoms with the underlying autoimmune disease. Factors potentially differentiating HLH from MAS include: 1.) change from quotidian to persistent fever pattern; 2.) sudden change from cytosis to cytopenia; 3.) coagulopathy; and 4.) decreasing ESR. Preliminary diagnostic criteria have been examined for MAS in sJIA, which may help to improve diagnosis of this condition in these patients [13, 15] . As mentioned before, HLH is diagnosed through clear clinical criteria. Despite awareness of the diagnostic criteria, HLH diagnosis can be challenging because of the variability in presentation. Symptoms, especially if infection associated, may spontaneously remit [3] . Individual diagnostic criteria can be observed at distinct points in the disease course and can also remit and recur. Moreover, many of the criteria are non-specific for HLH, making it a diagnosis of exclusion in many cases. Finally, criteria such as NK cell function and genetic testing can take weeks to be finalized. Given all these challenges, it is important to consider HLH as a diagnosis both at onset and in the early stages of the disease process. In a review of familial and acquired HLH, key laboratory findings in establishing a diagnosis were identified as negative or decreased NK function as well as elevated soluble CD25 levels in all patients [16] , suggesting these studies must be evaluated early on in the clinical course. In some centers, CD107a upregulation, a marker of NK cell degranulation is used as a surrogate for NK cell cytotoxicity [17] . However, as depressed NK cell cytotoxicity is seen in many conditions secondary to inflammation or immunosuppressive factors, other symptoms must be considered. Fever and splenomegaly occur in approximately 70% of HLH patients [16] . Fevers may be protracted, variable and may even resolve spontaneously. About 50% of patients initially present with elevated triglycerides, high ferritin, high LDH and/or a combination of anemia, neutropenia and thrombocytopenia [8, 16] . Of note, ferritin is rarely > 200 μg/L in pediatric patients with infections outside of the context of HLH [16] and levels higher than this should raise suspicion for HLH. Lastly, hemophagocytosis is seen in less than 40% of patients at onset but is present in >80% of patients at time of diagnosis [16] . There are presently five types of familial HLH (FHL) [9] , all of which impair the lytic activity of cytotoxic lymphocytes. Specifically, cytotoxic lymphocytes mediate contact-dependent elimination of cells perceived as dangerous by secreting preformed destructive molecules contained within specialized organelles termed lytic granules [2] . In order to mediate cytotoxicity, lytic granules must contain appropriate effector molecules and be localized to the contact site with the cell targeted for destruction. Once localized to that intercellular interface, the lytic granules dock at the cell membrane. The membrane of the granule and the cytotoxic lymphocyte are fused, allowing the release of the lytic effector molecules onto the targeted cell. Familial HLH type 1 (FHL1) is linked to chromosome 9q21, but the exact gene remains unknown [9] . Perforin (PRF-1) was the first identified FHL gene (located on chromosome 10q21-22), and is responsible for familial HLH type 2 (FHL2). Perforin protein is an effector molecule found in lytic granules. Once secreted, perforin inserts into the membrane of the target cell and facilitates the uptake of granzyme B and other cytolytic molecules contained in the lytic granules, into the target cell to induce cell death [1, 9] . In perforin-deficient mice infected with high doses of lymphocytic choriomeningitis virus (LMCV), haemophagocytic lymphohistiocytosis could be induced, similar to human FHL [18] . This suggests that failure to clear virus will lead to persistence of viral antigens and prolonged CD8 T cell activation and cytokine production. Other animal model studies demonstrate perforin's role in regulating lymphocyte number in autoimmunity [19] , after microbial infection [20] and when other cell-death pathways are impaired [21] . The extent that these mechanisms are involved in the control of immune responses, however, is still speculative. PRF-1 mutations account for approximately 20-40% of FHL cases. It is important to note however that CD107a up-regulation is usually normal in FHL2 as CD107a represents a measure of degranulation [22] . FHL3 is caused by mutations in the UNC13D gene located on chromosome 17q25. UNC13D encodes the protein MUNC 13-4. Found on the surface of lytic granules, MUNC 13-4 is required for priming the lytic granules for docking at the cytotoxic cell membrane [9, 12] . FHL 4 is caused by mutations in STX11 on chromosome 6q24 that encodes the syntaxin 11 protein. Syntaxin-11 is a member of the SNARE protein family and facilitates the fusion of the lytic granule membrane with that of the cytotoxic lymphocyte [1, 23] . FHL5 is caused by mutations in MUNC 18-2 located on chromosome 19p13. MUNC 18-2 encodes the syntaxin binding protein 2. It is a partner of syntaxin 11 and is required for SNARE complex-mediated fusion of the lytic granule with the cytolytic cell membrane [9, 22] . In addition to genetic defects associated with FHL, there are also immunodeficiency syndromes associated with HLH that impair the secretion of lytic granule contents. These include Griscelli syndrome type 2, Hermansky Pudlak type II, Chediak Higashi and X-linked lymphoproliferative disease Type 1. Griscelli syndrome type 2 and The orange triangles indicate the first and last day that daily anakinra was given. In Figure 2a and 2b the red line denotes the threshold below which diagnostic criteria is met for HLH while in Figure 2c this line denotes the threshold above which HLH criteria is met. Chediak Higashi syndrome are typically associated with albinism due to an effect upon melanocyte pigment secretion. Griscelli syndrome type 2 is identified by mutations in RAB27A while Chediak Higashi syndrome is associated with LYST mutations. Mutations of both these gene impair proteins important for formation and/or trafficking of secretory lysosomes [24] . In these syndromes, a gene defect interferes with lytic granules reaching the cytotoxic lymphocyte membrane, thus leading to impaired NK cell cytotoxicity. Patients with X-linked lymphoproliferative disease have difficulty clearing Epstein-Barr virus (EBV) infected B-cells, with subsequent extensive lymphocytic expansion into multiple organs. The SH2D1A mutation, which encodes for the signaling lymphocyte activation molecule (SLAM)-associated protein SAP, is identified in XLP1. Impaired cytolytic function in XLP1 is thought to cause accumulation EBV-infected B cells and persistence of reactive inflammatory cells, which combine to produce an exaggerated immune response [25] . Treatment for HLH is described in the HLH 2004 revised guidelines [14] and is divided into acute and long-term management. Initial treatment with an immunomodulatory regimen is recommended. Patients with primary HLH who fail to reach disease resolution within 8 weeks of treatment should continue on this regimen for an additional treatment course. Hematopoetic stem cell transplantation should be pursued as soon as a suitable donor is available for all patients with primary HLH, relapsed HLH, or those failing to progress on therapy [4] . There is no standard for the treatment of secondary HLH or MAS in the context of rheumatologic disease. High dose corticosteroids, biologic agents, cyclosporine and high dose intravenous immunoglobulin have been used with varying success. Anakinra, a recombinant IL-1 receptor antagonist, has been increasingly used for sJIA patients with MAS [24] . If initial treatment fails, HLH salvage therapy may be pursued according to the HLH 2004 chemotherapeutic regimens [14, 12] . Confirmation of a genetic mutation is not needed for immediate management but is important in differentiating familial HLH and secondary HLH for long-term patient management and family genetic counseling. While anakinra has been successfully utilized in MAS, it has not been formally studied in HLH associated with bonafide mutations. In our patient with two identified UNC13D mutations, anakinra was utilized as an immunomodulator while the diagnosis was evolving. Although the progression of our patient's disease was not florid while receiving anakinra, there were decreased platelet and neutrophil counts during therapy. However, when anakinra was stopped, there was a clear surge in ferritin levels. It is unclear from our experience if anakinra magnifies certain abnormalities found in primary HLH while moderately quelling others. A more likely explanation is that the anakinra functioned in partially blocking the HLH-associated inflammation in our patient thus allowing for the masking of certain phenotypic associations of HLH such as ferritin, but not others (i.e., platelet counts). While we were not situated to immunologically prove the partial blockade of inflammation in our patient while receiving anakinra, our experience suggests that its use be reserved for the more mild secondary forms of Figure 3 Comparison of CD163 immunohistochemical stain of bone marrow one month before trial of anakinra trial (3a), and the day after discontinuation (3b). Shown at 40x magnification. CD163 is an immunohistochemical stain that is specific for monocytes and macrophages and the positive cells show brown surface staining. Prior to anakinra, there are slit-like macrophages and some with a few ingested cells (blue arrow). On repeat biopsy and stain done after anakinra, there is an increase in the number of macrophages and the CD163 stain demonstrates numerous hemophagocytic macrophages containing multiple ingested cells (blue arrows). HLH and that therapy of primary HLH be limited to more wide-ranging immunosuppression such as that provided through the HLH-2004 protocol [4, 14] . HLH is a clinical syndrome that remains difficult to diagnose. Our patient's case demonstrates that use of immunosuppressive agents can cloud diagnosis. As clinicians, it is important to be aware of this in order to avoid delay in diagnosis and life-saving therapy. "Written informed consent was obtained from the parent of the patient for publication of this Case Report and any accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal."

Relative cost and outcomes in the intensive care unit of acute lung injury (ALI) due to pandemic influenza compared with other etiologies: a single-center study

BACKGROUND: Critical illness due to 2009 H1N1 influenza has been characterized by respiratory complications, including acute lung injury (ALI) or acute respiratory distress syndrome (ARDS), and associated with high mortality. We studied the severity, outcomes, and hospital charges of patients with ALI/ARDS secondary to pandemic influenza A infection compared with ALI and ARDS from other etiologies. METHODS: A retrospective review was conducted that included patients admitted to the Cleveland Clinic MICU with ALI/ARDS and confirmed influenza A infection, and all patients admitted with ALI/ARDS from any other etiology from September 2009 to March 2010. An itemized list of individual hospital charges was obtained for each patient from the hospital billing office and organized by billing code into a database. Continuous data that were normally distributed are presented as the mean ± SD and were analyzed by the Student’s t test. The chi-square and Fisher exact tests were used to evaluate differences in proportions between patient subgroups. Data that were not normally distributed were compared with the Wilcoxon rank-sum test. RESULTS: Forty-five patients were studied: 23 in the H1N1 group and 22 in the noninfluenza group. Mean ± SD age was similar (44 ± 13 and 51 ± 17 years, respectively, p = 0.15). H1N1 patients had lower APACHE III scores (66 ± 20 vs. 89 ± 32, p = 0.015) and had higher Pplat and PEEP on days 1, 3, and 14. Hospital and ICU length of stay and duration of mechanical ventilation were comparable. SOFA scores over the first 2 weeks in the ICU indicate more severe organ failure in the noninfluenza group (p = 0.017). Hospital mortality was significantly higher in the noninfluenza group (77 vs. 39%, p = 0.016). The noninfluenza group tended to have higher overall charges, including significantly higher cost of blood products in the ICU. CONCLUSIONS: ALI/ARDS secondary to pandemic influenza infection is associated with more severe respiratory compromise but has lower overall acuity and better survival rates than ALI/ARDS due to other causes. Higher absolute charges in the noninfluenza group are likely due to underlying comorbid medical conditions.

The spread of a novel H1N1 strain of the Influenza A virus represents the first pandemic of the 21 st century and the first influenza pandemic since 1968 [1] . Compared with seasonal influenza, this strain was more prevalent in younger-aged individuals, obese patients, and pregnant women [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] . Severe cases of pandemic H1N1 resulted in respiratory failure thought to be secondary to direct cell damage and systemic cytokine release resulting in acute lung injury (ALI) or acute respiratory distress syndrome (ARDS) requiring prolonged ventilatory assistance and the frequent use of rescue therapies [4, 5, 8, [13] [14] [15] [16] [17] . Limited data exist that compare the clinical differences between ALI in H1N1 patients and ALI arising from other etiologies. Furthermore, whereas a number of studies have assessed different aspects of the economic impact of the recent pandemic [18] [19] [20] [21] , few have focused on the health care cost of the pandemic, particularly the utilization of limited ICU resources. We report the severity, clinical outcomes, and hospital charges of ALI/ARDS secondary to pandemic influenza A infection compared with ALI/ARDS from other etiologies during a similar period of time. Based on clinical bedside observations and published reports [4, 5, 8] , we hypothesize that ALI/ARDS secondary to pandemic influenza is associated with similar ICU outcomes but increased resource utilization and higher hospital charges due to the frequent need for rescue interventions and prolonged ventilatory assistance. The study was approved by the Human Investigation Committee of the Cleveland Clinic Foundation (CCF) (Institutional Review Board approval # 10-229) as a retrospective, single-center study at the CCF Medical ICU. Patients were identified from a unit-based acute lung injury screening database (Cleveland Clinic is one of the centers participating in the ARDSnetwork) and the H1N1 patient log maintained during the fall-winter season of 2009-2010. Patients were included if they met criteria for ALI (PaO 2 /FiO 2 ≤300; acute bilateral infiltrates; positive pressure ventilation via endotracheal tube; and no clinical evidence of left atrial hypertension or congestive heart failure) between the months of September 2009 to March 2010-the time that influenza infection was most prevalent. Diagnostic methods for influenza A virus detection consisted of rapid antigen testing, polymerase chain reaction (rtPCR), and viral culture from nasopharyngeal swabs, tracheal aspirates, and bronchioalveolar lavage specimens. The patients were grouped into two categories: those with laboratoryproven H1N1 infection; and those in whom H1N1was not clinically suspected. Only patients with confirmed infection were included in the influenza group to ensure that the clinical course of the disease was accurately captured. Patients were excluded from the study if they did not meet the above criteria for ARDS, or if clinical suspicion pointed to a likely pandemic viral infection with negative diagnostics. A Research Electronic Data Capture (REDCap) database was constructed with a complete listing of the patient's demographic and clinical information, including age, gender, height, weight, body mass index (BMI), presenting symptoms, past medical history, primary reason for admission to the ICU, vital signs, presence of vasopressors, laboratory values, ventilator settings and respiratory parameters, Acute Physiology and Chronic Health Evaluation (APACHE) III and Sequential Organ Failure Assessment (SOFA) scores on admission to the MICU, number of intubated days, duration of ICU and hospital stay, mortality, and rescue therapies (namely inhaled nitric oxide, proning, high-frequency oscillatory ventilation, and extracorporeal membrane oxygenation [ECMO]) [22] . The data collection was de-identified and collected in accordance with HIPAA guidelines. As part of the routine MICU respiratory therapy protocol, mechanical ventilation parameters are recorded every 4 hours. All patients are managed according to a mechanical ventilation protocol that incorporates the use of nonconventional modes when a lung protective strategy on conventional modes failed to provide adequate oxygenation. The following criteria were used to define the analyzed parameters: 1) mode of ventilation: the mode of ventilation that was used for the longest time for a given day; 2) PaO 2 /FiO 2 : worst daily ratios were recorded; 3) plateau pressure (Pplat): for patients on volume control ventilation the airway pressure was measured after a 5-second inspiratory hold without concomitant active inspiratory efforts, and for patients on pressure control ventilation (PCV) the highest total system pressure (PEEP + inspiratory pressure) was recorded; 4) positive end expiratory pressure (PEEP): the value corresponding to the highest PEEP for the day was recorded; 5) tidal volume (Vt): the largest daily volume was recorded. Respiratory data were captured on the first day of intubation (day 1) and then on subsequent days 3, 7, and 14 of mechanical ventilation. There were no differences in ventilator protocols or management between the two groups. An itemized bill of individual charges for each patient was obtained from the hospital billing office and was organized by billing code into the following categories: room/board, pharmacy, supplies, laboratory, radiology, surgical (including procedures performed under general anesthesia), blood products, respiratory services, dialysis, and miscellaneous (which included some professional fees, nonsurgical procedures and phlebotomy, and diagnostics not included in the other categories, such as electroencephalograms, electrocardiograms, echocardiograms, cardiac catheterizations, and vascular studies). The values represent the hospital charges for the aforementioned services rather than the actual reimbursement, which may be subject to more variability. The single-center nature of the study removes interfacility differences in clinical and billing practices. Continuous data that were normally distributed are presented as the mean ± SD and were analyzed by the Student's t test. The chi-square and Fisher exact tests were used to evaluate differences in proportions between patient groups. In instances where the data were not normally distributed, the groups were compared with the Wilcoxon rank-sum test. Differences were considered statistically significant if the p value was <0.05. Fifty-one patients were identified in the acute lung injury screening database between September 2009 and March 2010. Twenty-two met criteria for ALI and did not have confirmed or suspected H1N1 infection and were thus included in the noninfluenza group (ALI/ARDS secondary to noninfluenza etiologies). Thirty-six patients in the H1N1 patient log had confirmed influenza A testing. Of those, 23 had ALI requiring mechanical ventilation (MV) during their MICU stay and were included in our analysis. Demographics, presenting symptoms, past medical history, and acuity on admission are shown in Table 1 . Patients in the influenza group tended to be younger with a higher BMI. Patients in the influenza group presented more often with lower respiratory infection (100 vs. 73%, p = 0.135) and had increased requirement for mechanical ventilation on admission to the ICU (96 vs. 68%, p = 0.022). On the other hand, the noninfluenza group had a higher propensity to present with shock requiring vasopressors (45 vs. 22%, respectively, p = 0.07). The primary cause of ALI in the H1N1 group was pneumonia (n = 23), whereas in the noninfluenza group the etiologies were more varied, including pneumonia (n = 9), sepsis (n = 5), aspiration of gastric contents (n = 2), transfusion reaction (n = 1), and other (n = 5). Whereas seven patients (30%) in the H1N1 group were considered healthy, only one patient (5%) in the noninfluenza group had no comorbid medical conditions on admission to the ICU (Table 1) . This difference is reflected in the lower mean APACHE III score on admission to the ICU in the H1N1 group (66 ± 20 vs. 89 ± 32, p = 0.015), despite similar SOFA scores (8.3 ± 3.4 and 9.2 ± 4.1, p = 0.44). There were no statistically significant differences between the two groups for initial laboratory data, including white blood cell count, platelets, serum creatinine, bilirubin, and creatinine kinase. The number of patients who developed acute renal failure that required dialysis throughout their ICU stay was the same (n = 8) in both groups. SOFA scores on days 1, 3, 7, and 14 of mechanical ventilation indicate that patients in the noninfluenza group had more severe organ failure during their ICU stay (p = 0.017; Table 2 ). Table 3 shows oxygenation index and mechanical ventilation related parameters on days 1, 3, 7, and 14. There was a nonsignificant trend toward worsening hypoxia in the H1N1 group, despite significantly higher PEEP and Pplat on days 1, 3, and 14. Tidal volumes were comparable throughout. Plateau pressures in the H1N1 group were high due to the relative decrease in pulmonary compliance in H1N1-related lung injury. Four patients in both groups were ventilated with airway pressure release ventilation (APRV). More patients in the influenza group required rescue therapies on day 1 of mechanical ventilation (4 vs. 0, respectively, p = 0.108); however, similar numbers of patients in both groups required rescue therapies over the duration of MV (7 and 5 patients, respectively). Rescue therapies in the H1N1 group included inhaled NO (n = 4), ECMO (n = 2), prone ventilation (n = 3), and high-frequency ventilation (n = 1), and in the noninfluenza group only inhaled NO (n = 3) and prone ventilation (n = 2). Mechanical ventilation days were comparable between groups (22 ± 17 vs. 19 ± 15 days for groups I and II, respectively, p = 0.53) as were 28-day ventilator-free days (5 ± 7.6 and 4.6 ± 9, p = 0.88). Four patients in the H1N1 group and seven in the noninfluenza group underwent a tracheostomy procedure. Hospital and ICU LOS were comparable (median ± IQR: 16 ± 22 vs. 24.5 ± 26.5 and 12 ± 15 vs. 17 ± 25.5 days for the influenza group and II, respectively, Wilcoxon p = 0.17 and 0.45). Mortality was significantly higher for patients in the noninfluenza group (77 vs. 39%, p = 0.016). Interestingly, a Kaplan-Meier curve of ICU mortality (Figure 1) indicates that patients in the H1N1 group were more likely to be discharged alive from the ICU when the length of stay was greater than 25 days, despite a trend toward higher mortality within the first 2 weeks. Even though all charges were higher in the noninfluenza group, only the difference in blood products utilized in the ICU was significant (4 ± 6 vs. 21 ± 25 thousands of U.S. dollars, Wilcoxon p < 0.001; Table 4 ). Differences in ICU charges in pharmacy (p = 0.23), supplies (p = 0.09), radiology (p = 0.08), and miscellaneous (p = 0.09) were large but not significant due to considerable variation. The proportion of charges in each of the major categories was similar between the groups (Figure 2 ). The average total ICU cost per patient (253 ± 193 vs. 350 ± 270 thousands of U.S. dollars, Wilcoxon p = 0.19) and the average ICU cost per patient per day (13 ± 4 vs. 15 ± 6 thousands of U.S. dollars, Wilcoxon p = 0.06) tended to be higher in the noninfluenza group. The fall of 2009 heralded the influx of patients suffering from severe hypoxic respiratory complications secondary to the pandemic H1N1 influenza to ICUs across the country. Due to the severity of pulmonary disease that many of these patients experienced, perception among treating clinicians was that these patients would have a All values expressed as mean ± SD. Using mixed models, the overall p value comparing the influenza and noninfluenza groups is 0.017. The trend over time was not significant (p = 0.1). worse outcomes and consume more resources, as measured by hospital charges, than patients who developed ALI from other etiologies. We demonstrated that, contrary to what was perceived, pandemic influenza A ALI/ ARDS was associated with a lower acuity and, consequently, lower hospital mortality that ALI/ARDS from other etiologies, and had a similar ICU and hospital LOS. ICU and total hospital charges reflected a trend toward higher overall charges for room and board, blood products, pharmacy, and overall charge per patient in the noninfluenza group. In accordance with other descriptive reports of pandemic influenza [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] , patients who tested positive for H1N1 infection, tended to be young (no patients >64 years old), obese (15 had BMI >30 kg/m 2 ), and in relatively good health (30% with no comorbid medical conditions). There were no pregnant patients in either group. Compared with other studies of pandemic influenza patients who required mechanical ventilation, SOFA scores (mean 8.3) were similar, although APACHE II (25 ± 9) scores were higher [5] [6] [7] [8] 14, 16, 17, 23] . The degree of respiratory compromise in our patients was more severe than other reports judging by the higher PEEP requirements and longer duration of mechanical ventilation, which was roughly double that reported in other studies [4] [5] [6] 8, 11, 13, 14, 16, 23] . Plateau pressures in these studies were not consistently reported. However, despite significantly longer ventilation duration and prolonged ICU and hospital stays, the mortality in our cohort was not higher than that seen in other studies, which ranged from 22-41% in patients who required mechanical ventilation [4] [5] [6] 8, 11, 13, 14, 16, 23] . Looking at the different patient characteristics between groups, it may be tempting to postulate that the higher rate of patients with pulmonary ARDS in the H1N1 group, in contrast to prevalent nonpulmonary ARDS in the noninfluenza group, would correlate with a higher PEEP response among the latter [24] . Our findings suggest the contrary. Patients in the H1N1 group had higher mean plateau pressure, likely indicative of lower compliance. The similarity of PaO 2 /FiO 2 ratios in the two groups may be a reflection of higher PEEP values used in the H1N1 group for lung recruitment, rather than being indicative of comparable degrees of lung injury. Although assessing recruitability from this retrospective analysis is difficult and may be inaccurate, the higher PEEP used and the implication of lower compliance observed are predictors of potentially recruitable lung [24] . These observations support the recent call for a reevaluation of the ALI and ARDS criteria to account for this heterogeneity in the patient population [25] . A number of important differences between the two cohorts emerged as well. As expected, the noninfluenza group was older, had more comorbid medical conditions, and less often presented to the ICU with respiratory failure. The degree of ventilator support was significantly higher in the H1N1 group on days 1, 3, and 14, and there was a trend to more severe hypoxemia during that time as well. Nevertheless, the use of use of APRV and rescue therapies was comparable in both groups. Despite more severe respiratory compromise, H1N1 patients did not have longer time on the ventilator, longer ICU or hospital stays, or higher mortality. Although SOFA scores were similar, the noninfluenza group had significantly higher APACHE III scores, likely secondary to points assigned to comorbid medical conditions. The high acuity of illness, as well as the presence of severe comorbidities, such as solid and hematologic oncologic conditions (7 patients), chronic renal insufficiency (6 patients), and cirrhosis of the liver (4 patients), likely contributed to the poor outcomes in the noninfluenza group. Conversely, despite more severe respiratory compromise, patients in the H1N1 group were more likely to recover due to their younger age and better overall health histories. The 77% mortality in the noninfluenza group was much higher than typically reported in clinical trials, with one notable exception [26] . However, reports from tertiary care centers involving patient cohorts with similar underlying comorbid conditions have reported equally high mortality rates [27] . Our observation brings up an interesting point, namely the difference between the reported mortality in clinical trials and the observed mortality in a similar clinical condition affecting patients that would have been excluded from such trials due to coexisting comorbidities. A Kaplan-Meier plot of ICU mortality (Figure 1) indicates that although patients in the H1N1 group were less likely to survive the first 14 days of ICU care, those that did survive past day 25 were more likely to be discharged alive from the hospital. Patients in the noninfluenza group were unlikely to survive if their ICU length of stay exceeded 3 weeks. ARDS is among the most expensive conditions encountered in the ICU [28] . In 1984, Bellamy and Oye described the charges of patients with ARDS, with the most expensive being room and board (30%), clinical laboratory (24%), pharmacy (14%), and inhalation therapy and ventilation (8%) [27] . Twenty-five years later, our study indicates that the aforementioned categories continue to represent the most expensive charges incurred by ARDS patients in the ICU. The overall similarity of charges in room and board and respiratory therapy between the two groups is likely indicative of the comparative durations of hospitalization and mechanical ventilation. Interestingly, despite higher ventilatory requirements and more severe hypoxemia in the H1N1 group, respiratory charges were similar between the two groups, suggesting that the high cost of maintaining a patient on mechanical ventilation is independent of the degree of ventilator support necessary. Thus, respiratory charges are more likely a reflection of duration of mechanical ventilation rather than the degree of ventilator support necessary. Absolute ICU charges for room and board, blood products, pharmacy, radiology, average daily charge, and overall charge per patient were larger in the noninfluenza group. ICU charges for blood products in the noninfluenza group were greater by a factor of four, and pharmacy charges double that of the H1N1 group. This finding is likely a reflection of the higher prevalence of underlying comorbid medical conditions in the noninfluenza group, such as malignancy and cirrhosis, which require expensive medications and predispose to anemia. Moreover, the high mortality in this cohort likely precluded even higher hospital charges. Nevertheless, the H1N1 cohort amassed charges of similar magnitude to the most ill and expensive patients in the ICU, indicating the abundant health care resources consumed by severe pandemic influenza infection. There are a number of limitations to our study. As a retrospective chart review rather than a prospective investigation, the information was culled from sources that were at times incomplete. Second, the study contained a relatively small number of patients, and measures taken to ensure internal validity of each group, such as limiting the influenza group to confirmed H1N1 infection and the noninfluenza group to the duration of the influenza season, further limited its size. Additionally, whereas our study provides descriptive information relevant to the patient population of our institution and tertiary referral centers with similar acuity, other ICUs may be exposed to a different cohort of patients. On the other hand, as a single-center study, potential differences in clinical and billing practices could be minimized. Although a comprehensive charge profile of each patient was generated, trends in the timing of charges could not be obtained. Finally, the hospital charge data were mined from an extensive database divided by charge coding, and therefore, some charges may have been mislabeled or inappropriately categorized. Our study provides interesting observations about the clinical course, outcomes, and cost of the H1N1 influenza pandemic. Although patients with severe pulmonary complications of pandemic influenza infection have poor oxygenation and require significant ventilatory support and rescue therapies, their younger age and tendency to have fewer comorbid medical conditions contribute to their improved prognosis compared with patients with ALI from other causes. Both groups of patients consume enormous amounts of hospital resources, and physicians and policy makers must be aware of this when future pandemics arise.

Identification of a Conserved B-cell Epitope on Reticuloendotheliosis Virus Envelope Protein by Screening a Phage-displayed Random Peptide Library

BACKGROUND: The gp90 protein of avian reticuloendotheliosis-associated virus (REV-A) is an important envelope glycoprotein, which is responsible for inducing protective antibody immune responses in animals. B-cell epitopes on the gp90 protein of REV have not been well studied and reported. METHODS AND RESULTS: This study describes the identification of a linear B-cell epitope on the gp90 protein by screening a phage-displayed 12-mer random peptide library with the neutralizing monoclonal antibody (mAb) A9E8 directed against the gp90. The mAb A9E8 recognized phages displaying peptides with the consensus motif SVQYHPL. Amino acid sequence of the motif exactly matched (213)SVQYHPL(219) of the gp90. Further identification of the displayed B cell epitope was conducted using a set of truncated peptides expressed as GST fusion proteins and the Western blot results indicated that (213)SVQYHPL(219) was the minimal determinant of the linear B cell epitope recognized by the mAb A9E8. Moreover, an eight amino acid peptide SVQYHPLA was proven to be the minimal unit of the epitope with the maximal binding activity to mAb A9E8. The REV-A-positive chicken serum reacted with the minimal linear epitopes in Western blot, revealing the importance of the eight amino acids of the epitope in antibody-epitope binding activity. Furthermore, we found that the epitope is a common motif shared among REV-A and other members of REV group. CONCLUSIONS AND SIGNIFICANCE: We identified (213)SVQYHPL(219) as a gp90-specific linear B-cell epitope recognized by the neutralizing mAb A9E8. The results in this study may have potential applications in development of diagnostic techniques and epitope-based marker vaccines against REV-A and other viruses of the REV group.

Reticuloendotheliosis viruses (REVs) are a group of viruses in the family Retroviridae, specifically gammaretroviruses in the same genus as mammalian C-type retroviruses [1] . The REV group includes defective REV-T [2, 3] , non-defective REV-A [4, 5] , chick syncytial virus (CSV) [6] , duck infectious anemia virus [7] , and spleen necrosis virus (SNV) [8] . Except for the defective REV-T, all isolated REV strains belong to a single serotype [5] and their genetic sequences show little variation [9] . REV genome consists of three structural genes (gag, pol and env) flanked by long-terminal repeats (LTRs) [10] . The major mature env gene products of REVs are the surface glycoprotein (gp90) and the transmembrane protein (gp20) [11, 12] . The gp90 protein containing both continuous and discontinuous epitopes functions as the immunodominant protein [13] and is responsible for eliciting REV antibodies. Previous studies indicated that the Cterminal epitope of gp90 was exposed on the outer surface of the REV-A-infected cell [12] . However, the epitope identified in REV gp90 protein has not been finely mapped, and the core sequence of the epitope needs to be determined. Detailed analysis of epitopes is important for the understanding of immunological events, and the development of epitopebased marker vaccines and diagnostic tools for various diseases [14, 15] . In this study, we prepared a neutralizing monoclonal antibody (mAb) against gp90 protein from the REV-A strain HLJ07I, and used it to screen a phage-displayed random 12mer peptide library for the linear B-cell epitope. This study describes the first identification of the precise location of the epitope on gp90 protein. The information provided in this study will facilitate the development of specific serological diagnosis of REV infection, and will contribute to the rational design of vaccines by further understanding of the antigenic structure of gp90. Care of laboratory animals and animal experimentation were performed in accordance with animal ethics guidelines and approved protocols. All animal studies were approved by the Animal Ethics Committee of Harbin Veterinary Research Institute of the Chinese Academy of Agricultural Sciences (SYXK (H) 2006-032). REV-A Strain HLJ07I (GenBank accession No. GQ375848) was isolated from Heilongjiang Province in China in 2007. Chicken embryo fibroblasts (CEFs) were prepared as primary cultures from 10-day-old chicken embryos as previously described [16] and were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum plus antibiotics. Viruses were grown in CEFs and incubated at 37uC with 5% CO 2 for 5 days. The suspension was frozen and thawed three times to disrupt cells and release virus, and then clarified by two centrifugation steps (2000 g for 15 min, and 10,000 g for 60 min). Virus present in the upper phase was precipitated with 10% (w/v) polyethylene glycol 6000 (PEG 6000) for 4 hours at 4uC. Precipitates were collected by centrifugation at 9,000 g for 30 minutes and resuspended in TNE buffer (50 mM tris-HC1, pH 7.5; 0.1 M NaC1, 10 mM EDTA). Finally, they were centrifuged through a 30% (w/v) sucrose cushion for 90 minutes at 200,000 g and resuspended in TNE buffer. The purified virus was analyzed in SDS-PAGE. Six-week-old female BALB/c mice were subcutaneously immunized with 100 mg of the purified recombinant gp90 protein emulsified with an equal volume of Freund's complete adjuvant (Sigma, St. Louis, MO, USA). Two boosters of the Freund's incomplete adjuvant (Sigma, St. Louis, MO, USA) emulsified antigen were given at two week interval. Two weeks after the third immunization, the mice were intraperitoneally boosted with 100 mg antigen alone. Three days later, the spleen cells from immunized mice were fused with myeloma cells SP2/0 (SP2/0-Agl4; ATCC CRL 1581) [17] , using 50% (wt/vol) polyethylene glycol and 10% dimethyl sulfoxide (DMSO) (vol/vol) (Sigma, St Louis, MO, USA). Hybridomas were screened by indirect enzyme-linked immunosorbent assay (ELISA) and indirect immunofluorescence assay (IFA). The hybridomas producing mAbs were cloned three times by limiting dilution of the cells. Antibody subtype identification was performed using SBA Clonotyping TM System/HRP Kit (Southern Biotech, Birmingham, AL, USA). Plates were coated with 100 mL/well of purified REV gp90 antigen diluted in carbonate-bicarbonate buffer (pH 9.6) for incubation overnight at 4uC. Following 4 washes with 200 mL/ well of PBS/0.05% Tween-20, the plates were blocked with 200 mL/well of blocking buffer (PBS containing 5% skim milk) for 1 h at 37uC. The supernatant of hybridoma culture (100 mL/well) was added in duplicate and the plates were incubated for 1 h at 37uC. After washing three times with PBS, 100 mL of horseradish peroxidase (HRP)-conjugated goat anti-mouse immunoglobulin G (IgG, 1:5,000 dilution,Sigma, St Louis, MO, USA) was added to each well and incubated for 1 h at 37uC. Plates were washed three times with PBS and incubated with 100 mL/well of o-phenylenediamine dihydrochloride (OPD, Sigma, St Louis, MO, USA) containing 0.3% H 2 O 2 for 5 minutes at room temperature in the dark. The reaction was stopped with 50 mL/well of 2 M H 2 SO 4 and the absorbance measured at 492 nm. About 70-80% confluent CEF cells in 96-well plates were infected with REV-A HLJ07I at a MOI of 0.2. At 5 days postinfection, the infected cells were fixed with icy cold ethanol absolute for 15 min at 4uC, and air dried. The fixed cells were incubated with mAb A9E8, REV-A-positive chicken serum, antiporcine IFN-c mAb (Sigma, St Louis, MO, USA), or REV-Anegative chicken serum for 1 h at 37uC. After washing three times with PBS, 50 mL/well of FITC-conjugated goat anti-mouse IgG or FITC-conjugated rabbit anti-chicken IgG (Sigma, St Louis, MO, USA) at 1:100 dilutions were added and incubated for 1 h at 37uC. The cells were rinsed three times with PBS and once with deionized water, and mounted in 50 mL of 90% glycerol in PBS, and then observed under the Nikon Eclipse Ti-E microscope equipped with NIS-Elements AR software. The micro-neutralization assay was modified from a previously described procedure [18] . The ascitic fluid was heat inactivated for 30 min at 56uC, and two fold serial dilutions were incubated with 2610 3 tissue culture infective doses 50% (TCID 50 /mL ) of REV-A in a 96-well micro-plate. Four uninfected control wells were included on each plate as control wells. After 2 h incubation at 4uC, 100 mL of CEF cells at 1.5610 5 cells/mL was added to each well. The plates were incubated for 5 days at 37uC and 5% CO 2 . The monolayers were washed with PBS and fixed in icy cold ethanol for 15 minutes. The presence of viral gp90 protein was detected by ELISA with the mAb A9E8. The absorbance was measured at 492 nm with an ELISA microplate reader. The average A492 was determined for quadruplicate wells of virusinfected and uninfected control wells, and a neutralizing endpoint was determined by using a 50% specific signal calculation. The endpoint titer was expressed as the reciprocal of the highest dilution of ascitic fluid with A492 value less than X, where 6= [(average A492 of infected wells) 2 (average A492 of control wells)]/2+ (average A492 of control wells). The Ph.D.-12 TM Phage Display Peptide Library Kit was purchased from New England BioLabs Inc. The dodecapeptide library consisted of 2.7610 9 electroporated sequences (1.5610 13 pfu/mL). The mAb was purified from the ascites uid of mice inoculated with the hybridma cells secreting A9E8 by affinity chromatography using rProtein G Agorose (Invitrogen, Carlsbad, CA,USA) according to the manufacturer's instructions. The concentration of the purified protein was determined using the Bradford Protein Assay Kit (Beyotime, Shanghai, China). Three successive rounds of biopanning were carried out according to the manufacturer's instruction manual. Briey, one well of a 96well microtiter plate was coated with 10 mg/mL of mAb A9E8 in coating buffer (0.1 M NaHCO 3 , pH 8.6) overnight at 4uC, followed by blocking with blocking buffer (0.1 M NaHCO 3 , pH 8.6, 0.02% NaN3, and 5 mg/ml BSA) for 2 h at 4uC. The phage library (1.5610 11 phages/100 mL) was added to the blocked wells and the plate incubated for 1 h at room temperature. The unbound phages were removed by successive washings with TBS buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl) containing gradually increased concentrations (0.1%, 0.3%, and 0.5%) of Tween-20, and the bound phages were eluted by 0.2 M glycine-HCl containing 1 mg/mL BSA (pH 2.2) and immediately neutralized with 1 M Tris-HCl (pH 9.1). The eluted phages were amplified by infecting E. coli (ER2738), and were titered on LB/IPTG/Xgal plates for the subsequent rounds of selection. The output to input ratio was calculated as follows: (titer of the amplified eluent phages/titer of the input phages (1.5610 11 ))6100%. After three rounds of biopanning, eight individual phage clones were selected for target binding in ELISA as described in the manufacturer's instructions. Briey, 96-well plates were coated with 100 ng of purified mAb A9E8, or anti-porcine IFN-c mAb (Sigma, St Louis, MO, USA) as negative controls overnight at 4uC. The coated wells were blocked for 2 h at room temperature and then the phages (10 10 pfu/100 mL/well) diluted in blocking solution were added. The plates were incubated for 1 h at room temperature followed by washing ten times with TBST. Bound phages were subjected to reaction with horseradish peroxidase (HRP)-conjugated sheep anti-M13 antibody (Pharmacia, Piscataway, NY, USA), followed by color development with substrate solution containing o-phenylenediamine (OPD). The positive phage clones identified by phage ELISA were sequenced with the 296 gIII sequencing primer 59-TGA GCG GAT AAC AAT TTC AC-39 as described in the manufacturer's instructions. A series of complementary oligonucleotides (Table 1) coding for wild-type and truncated motif SVQYHPL were synthesized, annealed, and cloned into the BamHI/XhoI sites of the prokaryotic expression vector pGEX-6p-1 (Pharmacia, Piscataway, NY, USA), producing a group of recombinant plasmids. All the resulting recombinant plasmids were validated by restriction analysis and DNA sequencing. Expression plasmids were transformed into BL21 (DE3) competent cells, followed by the addition of 1 mM isopropyl-D-thioga-lactopyranoside (IPTG; GE Healthcare, USA) for induction. Approximately equivalent amount of each GST fusion protein was subjected to 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (12% SDS-PAGE). The gel was either stained with commassie blue staining solution or electrophoretically transferred to nitrocellulose membrane. After being blocked with 5% skim milk in PBS overnight at 4uC, the membrane was incubated with mAb A9E8 (diluted 1:2,000 in PBS) or REV-A-positive chicken serum (diluted 1:100 in PBS) at 37uC for 1 h. After being washed three times with PBST, the membrane was probed with a 1:5,000 dilution of HRP-conjugated goat anti-mouse IgG To investigate the conservation of the epitope among REV viruses, sequence alignment of the epitope and the corresponding regions on gp90 proteins of 32 REV-A strains, one REV-T strain, four SNV strains and one CSV strain was performed using the DNASTAR Lasergene program (Windows version; DNASTAR Inc., Madison, WI, USA). Purified gp90 protein was used to immunize BALB/c mice. After cell fusion and screening, several hybridoma cell lines were generated, which produced gp90-reactive mAbs. One monoclonal antibody produced by the line designated as A9E8 was selected for strong reactivity with recombinant gp90 protein in Western blot ( Figure 1A ) and in an indirect ELISA (data not shown). It also showed strong reactivity with purified whole virus in Western blot ( Figure 1B) and could be used to detect REV-A antigen by an indirect immunofluorescence assay (IFA; Figure 1C ). The mAb A9E8 was compose of an IgG2b heavy chain paired with a k-type light chain, as determined using the SBA Clonotyping TM System/ HRP Kit. The titers of antibody in hybridoma cell culture supernatants and in ascites were measured by indirect ELISA and determined to be 1:3,200 and 1:128,000, respectively. The neutralizing activities of the mAb A9E8 were then determined by a micro-neutralization assay on CEF cells using REV-A HLJ07I. The mAb A9E8 neutralized the virus with a neutralization titer (NT 50 ) of 100. To determine the epitope recognized by mAb A9E8, biopanning of a phage displayed 12-mer random peptide library was performed using the affinity purified mAb A9E8. After three rounds of biopanning, an enrichment of phages bound to the mAb A9E8 was obtained. The output to input ratios of the three rounds of biopanning were 0.00008%,0.038% and 0.79%. Eight phage clones were selected for reactivity with the mAb A9E8 after three rounds of biopanning and enrichment of the phages binding to the mAb A9E8. These selected clones were further evaluated by Phage ELISA for reactivity with the mAb A9E8 and a negative control mAb (anti-porcine IFN-c). As shown in Figure 2 , all the selected eight phage clones (A1-A8) showed specific reactivity with A9E8 (OD492 nm .1.10), but not with anti-porcine IFN-c mAb (OD492 nm ,0.15). The eight phage clones were sequenced, and were shown to display a consensus sequence SVQYHPL, which was identical to the motif 213 SVQYHPL 219 at the C-terminus of the gp90 protein of REV-A strain HLJ07I (Table 2) . To verify whether the identified motif represented an epitope recognized by the mAb A9E8, a DNA fragment coding for the motif SVQYHPL was expressed as a GST fusion protein (GST-H7wt) in E. coli. Western blot analysis showed that the fusion protein was recognized by the mAb A9E8 ( Figure 3A ) and REV-A infected chicken antiserum ( Figure 3B ), indicating that the motif represented a linear B-cell epitope. To define the epitope precisely, four mutants with deletions at C-and N-termini of the motif SVQYHPL (Table 1) were constructed to express the GST fusions GST-H7DS, GST-H7DL, GST-H7DSV, and GST-H7DPL representing -VQYHPL, SVQYHP-, -QYHPL and SVQYH-(deletions were shown as dashes) in E. coli, respectively. We found that only the full-length SVQYHPL polypeptide (GST-H7wt) was recognized by the mAb A9E8 ( Figure 3A ). Removal of one or more amino acids at either the amino or carboxyl terminus of the peptide abolished antibody binding, indicating that the peptide SVQYHPL represented the minimal requirement for the reactivity of the epitope with A9E8. Minimal Unit of the Epitope with the Maximal Binding Activity to mAb A9E8 To investigate minimal unit of the epitope with the maximal binding activity to mAb A9E8, a series of GST-fusion proteins were expressed with extended amino acid residues at both N and C termini of the motif SVQYHPL (Table 1) . These GST-fusion proteins were subjected to SDS-PAGE and testing for reactivity with mAb A9E8 in Western blot. Fusion proteins GST-R1 (SVQYHPLA), GST-R2 (SVQYHPLAL) and GST-R3 (SVQYH-PLALP) reacted strongly with mAb A9E8 in Western blot ( Figure 4) . The GST-R2 and GST-R3 showed similar binding activity to the GST-R1, indicating that alanine alone significantly increased binding activity of the core epitope to mAb A9E8. In contrast, GST-fusion proteins with extended amino acid residues at the N terminus of the motif SVQYHPL showed no increased binding activity compared with GST-H7wt in Western blot (data not shown). Taken together, these results showed that SVQYH-PLA was the minimal unit of the epitope with the maximal binding activity to mAb A9E8. To investigate the conservation of the SVQYHPL epitope, we aligned the epitope identified in this study with REVs gp90 coding regions available in GenBank. The alignment results showed that all amino acids in the motif were identical among all REV strains ( Figure 5 ), indicating that the motif represented a conserved epitope on the gp90 protein of REVs. Figure 2 . Detection of the selected phages for antibody binding by Phage ELISA. Eight phage clones selected after three rounds of biopanning were added to the microplate wells (10 10 pfu/100 mL/well) coated with the mAb A9E8 or anti-porcine IFN-c mAb (negative control) (100 ng/well ), and incubated for 1 h at room temperature. Bound phages were subjected to reaction with horseradish peroxidase (HRP)-conjugated anti-M13 antibody, followed by color development with substrate solution containing o-phenylenediamine (OPD). Three independent assays were performed for each selected phage. doi:10.1371/journal.pone.0049842.g002 Table 2 . Sequence comparison of random peptide inserts displayed on the positive phages. Amino acid sequence of the insert a Conservative amino acid motifs are bold and underlined. doi:10.1371/journal.pone.0049842.t002 The gp90 protein of REV is an important antigenic protein and is associated with virus neutralization, which is the major candidate antigen for vaccine development and disease serological diagnosis [12, 13] . Studies showed that recombinant gp90 protein expressed in Pichia pastoris induced a protective immune response against REV in chickens [19] . Precise mapping of epitopes in gp90 is important for understanding antibody-mediated protection and developing epitope-based marker vaccines and diagnostic tools. Cui et al. [20] reported the generation and partial characterization of a panel of 11 mAbs against the nondefective REV Strain T, and showed that the epitope was on the viral envelope glycoprotein. However, they only identified the relative regions in REV envelope glycoprotein recognized by the mAbs, and did not map the fine locations of the epitopes. To our knowledge, there has been no report on linear epitope mapping of the gp90 of REV. Mapping epitopes using monoclonal antibodies has become a powerful tool to study protein structure and has been used to diagnose diseases and design marker vaccines [21, 22, 23] . In this study, we described the generation and epitope mapping of a gp90 protein specific mAb, and demonstrated that the epitope was conserved among the REV group. Precise analysis of REV-A gp90 protein epitope will provide the fundamental information for development of epitope-based vaccines and diagnostic tools for REV-A and/or other REV group infection. Phage display is an in vitro selection technique in which a peptide or protein is genetically fused to a coat protein of bacteriophage and the fused peptide or protein is displayed on the exterior surface of the phage virion. The phage displayed random peptide library is a powerful and high throughput tool for rapid mapping of epitopes [24] . In this study, we generated a gp90-specific mAb A9E8 using recombinant gp90 protein expressed in E. coli. The mAb A9E8 showed strong reactivity against purified whole virus in Western blot and could be used to detect REV-A antigen by an indirect immunofluorescence assay. The linear epitope recognized by the mAb A9E8 was defined as SVQYHPL by screening a random phage display peptide library. This peptide sequence was identical to 213 SVQYHPL 219 of the gp90 protein of REV-A. N-or Cterminal deletions of amino acids of this epitope demonstrated that 213 SVQYHPL 219 is the minimal requirement for recognition by A9E8. Fusion proteins GST-R1 with extended amino acid residues at the C terminus of the motif SVQYHPL showed increased binding activity compared with that of GST-H7wt in Western blot, indicating that alanine alone significantly increased binding activity of the core epitope to mAb A9E8. Thus, the peptide SVQYHPLA was determined to be the minimal unit of the epitope with the maximal binding activity to mAb A9E8. The peptide was also recognized by REV-A-positive chicken serum, revealing the importance of the eight amino acids of the epitope in antibody-epitope binding reactivity. Sequence alignments of REV-A strains, REV-T strain and five other REV strains demonstrated that the motif was highly conserved among REV viruses, indicating that it is a broad group-specific epitope. Since A9E8 was identified as a neutralizing mAb, the epitope identified with A9E8 in this study was a neutralizing epitope. Many neutralizing epitopes have been mapped in the variable regions of the proteins of viruses, including infectious bursal disease virus [25] , infectious bronchitis virus [26] , hepatitis C virus [27] , and HIV [28] . Some neutralizing epitopes, however, are highly conserved across most of the viruses in the same group [29, 30] . A novel epitope was mapped within the highly conserved flavivirus fusion loop peptide 98 DRXW 101 by phage-display biopanning and structure modeling using mAb 2A10G6 that had broad cross-reactivity with dengue virus (DENV) 1-4, yellow fever virus (YFV), West Nile virus (WNV), and Japanese encephalitis virus (JEV) viruses. This mAb potently neutralized DENV 1-4, YFV, and WNV and conferred protection against lethal challenge with DENV 1-4 and WNV in murine model. Further functional studies revealed that 2A10G6 blocked infection at a step after viral attachment. These results show that the broad cross-reactivity epitope recognized by neutralizing mAb 2A10G6 is highly conserved among DENV 1-4, YFV and WNV [29] . An epitope recognized by mAb 51 belonging to isotype IgM was mapped to 215 KQEKD 219 of the VP1 capsid protein of Enterovirus 71 (EV71), which possessed neutralizing activity in vitro and provided 100% in vivo passive protection against lethal challenge with EV71 strain HFM 41. BLAST analyses of the neutralizing epitope revealed that it was highly conserved among all EV71 strains, but not coxsachievirus 16 [30] . In this study, the epitope recognized by neutralizing mAb A9E8 was mapped to a highly conserved region of the gp90 protein among REVs, which would be useful for development of REV marker vaccines and diagnostic techniques. In summary, a highly conserved neutralizing linear B-cell epitope on the gp90 protein of REV-A was identified in this study. The identified conserved epitope may have potential for development of REV specific diagnostic assays and epitope-based marker vaccines. Identification of an Epitope on REV gp90 Protein PLOS ONE | www.plosone.org

Lower Respiratory Tract Infection in a Renal Transplant Recipient: Do not Forget Metapneumovirus

Human metapneumovirus (hMPV) is emerging as a cause of a severe respiratory tract infection in immunocompromised patients. hMPV pneumonia has only been seldom reported in nonpulmonary solid organ transplanted patients, such as renal transplant recipients. We report here a case of a 39-year-old patient presenting with fever, cough, and interstitial opacities on CT scan diagnosed as a nonsevere hMPV pneumonia 11 years after a renal transplantation. Infection resolved spontaneously. Differential diagnosis with Pneumocystis pneumonia was discussed. We review the medical literature and discuss clinical presentation and detection methods that can be proposed in solid organ transplant recipients.

Respiratory viruses such as respiratory syncytial virus (RSV), influenza, parainfluenza viruses (PIV), and adenovirus are commonly associated with mild to severe symptoms, depending on the immune status. Human Metapneumovirus (hMPV) was the sixth most frequent viral infection in patients hospitalized for respiratory illness [1] . hMPV is a nonsegmented, enveloped, negative single-stranded RNA virus [2] responsible for lower respiratory tract infections (LRI), especially in extreme ages [3, 4] . It has a seasonal distribution and occurs mainly in winter and spring, with an incubation period usually between 4 and 6 days [5] . hMPV is now widely recognized as an opportunistic infection in immunocompromised hosts such as hematopoietic stem cell transplant (HSCT) and pulmonary transplant recipients, leading to a significant respiratory morbidity [6] . Although its detection is not yet routinely performed, hMPV appears to account for 9% of acute pneumonia in patients with haematologic malignancies (including HSCT), in a similar proportion to RSV [7] . This rate is close to that reported in lung transplant recipients, ranging from 6% to 12% of LRI [6, 8] . In contrast, it has been seldom reported in other SOT settings such as renal transplantation [9] . A 39-year-old patient with an 11-year history of kidney transplantation for severe amyloidosis was referred to the Centre d'Infectiologie Necker Pasteur for acute fever for 2 days. After 8 years of transplantation, he was treated for graft rejection by corticosteroids. Clinical course was uneventful, except for recurrent prostatitis. His current immunosuppressive regimen consisted of mycophenolate mofetil except for sore throat and rhinorrhea. Biological analyses showed an elevated C-reactive protein blood level (157 mg/L, normal <5 mg/L) but normal blood leukocytes and neutrophil counts. Blood lymphocyte count was low (0.94 G/L), with CD4+ T cells accounting for 39.5% of total lymphocytes (0.372 G/L). HIV serology was negative. Blood and urine cultures were sterile and initial chest radiograph was normal. Nonproductive cough without dyspnea or chest pain appeared on day 3 of hospitalization. Oxygen saturation in ambient air was 92%. Chest auscultation was normal. As the cough increased, a thoracic computed tomography (CT) scan was performed on day 6 and revealed bilateral ground glass infiltrates mainly located in subpleural and peripheral areas, associated with bilateral pleural effusion ( Figure 1 ). No mediastinal adenopathy was seen. Because of clinical and radiological presentation suggesting Pneumocystis jirovecii pneumonia, trimethoprim-sulfamethoxazole was initiated the same day. Nasopharyngeal aspirates were screened by direct immunofluorescence with specific monoclonal antibodies to RSV, influenza virus A and B, PIV, adenovirus, and hMPV (Argène, Verniolle, France) on day 7. Immunofluorescence was strongly positive for hMPV and negative for other viruses. Blood cultures and S. pneumoniae and L. pneumophila urinary antigen detections were negative. As Pneumocystis pneumonia was initially suspected, bronchoalveolar lavage (BAL) fluid analysis performed on day 11 demonstrated 450.10 3 cells/mL (macrophages 72%, neutrophils 17%, and lymphocytes 11%). Microbiological studies did not reveal any bacterial or fungal microorganisms. Gomori-Grocott staining for Pneumocystis jiroveci detection, indirect immunofluorescence, and polymerase chain reaction (PCR) for P. jiroveci were negative. hMPV was also detected in BAL fluid by direct immunofluorescence. As all microbiological investigations were negative except for hMPV, antibiotics were discontinued; respiratory symptoms spontaneously improved within 6 days. Thus, decreased immunosuppression or other medications such as ribavirin or intravenous immunoglobulin were not considered. The patient was discharged on day 14. This is the second hMPV pneumonia in a kidneytransplanted recipient described in the literature. The first reported case was a severe LRI requiring transient intensive care unit stay [9] . It occurred three years after kidney transplantation, while receiving immunosuppressive regimen consisting of ciclosporine (125 mg b.i.d), azathioprine (75 mg/d), and prednisone (10 mg/d). Compared to this case, our patient had mild symptoms, mainly cough and upper respiratory symptoms. He was also less immunosuppressed without corticosteroids regimen. In solid organ transplanted patients, hMPV is responsible for LRI and may lead to hospitalization and significant respiratory illness in up to 63% of cases [6, 8] . As initial clinical symptoms are nonspecific, thoracic CT scan can be more helpful than chest X-ray, which is less sensitive. Consolidation, nodular infiltrates, and pleural effusions may be seen. Subpleural and basal areas are usually observed, and bilateral locations are seen in 50% of cases, as in our case [10] . Whereas crazy paving, network of a smooth linear pattern superimposed on an area of ground-glass opacity, is unusual, bronchiectasis is common, up to 68% in the series by Wong et al. [10] . Of note, lymphopenia, as noticed in our patient, is the most common feature reported in HSCT patients with hMPV, accounting for 73% of patients in one series [7] . This illustrates that although innate immune responses are stimulated upon hMPV exposure, adaptive immunity also appears important to control hMPV. As for other paramyxoviruses, the matrix proteins are involved in the induction of proinflammatory and Th1 responses by dendritic cells and macrophages (i.e., production of interleukin-2 and interferon-γ) [11] . Inflammation may cause diffuse alveolar damage and hyaline membrane formation as shown by histopathology investigations [12] . Apart from other respiratory viral infections occurring in SOT recipients, differential diagnoses of hMPV-associated LRI include severe bacterial and fungal pneumonitis, particularly Pneumocystis pneumonia. Ribavirin, previously shown active in a mouse model of infection [13] , has been suggested as a potential antiviral therapy in HSCT and lung transplant recipients with hMPV-associated LRI [14, 15] . In our case and in the other case of the literature [9] , ribavirin was not used because the diagnosis was made retrospectively after the patient's spontaneous clinical improvement. In conclusion, hMPV has to be considered as a potential cause of LRI in kidney transplant recipients and may mimic Pneumocystis pneumonia. A prompt recognition would have avoided antibiotic use and further diagnostic studies such as bronchoscopy. Its early detection using immunofluorescence and/or RT-PCR must be proposed routinely in transplantation settings. In addition, early recognition could improve the implementation of appropriate infection control practices to prevent viral spread of this potential lifethreatening infection in immunocompromised patients.

What was the primary mode of smallpox transmission? Implications for biodefense

The mode of infection transmission has profound implications for effective containment by public health interventions. The mode of smallpox transmission was never conclusively established. Although, “respiratory droplet” transmission was generally regarded as the primary mode of transmission, the relative importance of large ballistic droplets and fine particle aerosols that remain suspended in air for more than a few seconds was never resolved. This review examines evidence from the history of variolation, data on mucosal infection collected in the last decades of smallpox transmission, aerosol measurements, animal models, reports of smallpox lung among healthcare workers, and the epidemiology of smallpox regarding the potential importance of fine particle aerosol mediated transmission. I introduce briefly the term anisotropic infection to describe the behavior of Variola major in which route of infection appears to have altered the severity of disease.

Controversy exists regarding the best method of protecting the public against the potential release of smallpox as a biological weapon (Bicknell, 2002; Fauci, 2002; Halloran et al., 2002; Kaplan et al., 2002; Mack, 2003) . Infectious disease modeling plays an important role in this dialog, and the biology of the transmission pathway, the focus of this review, is critical to producing appropriate predictive models and understanding which controls will work best under varying conditions (Ferguson et al., 2003) . The rapidity with which smallpox would spread in a developed nation is not known and is a major source of uncertainty in models used for public health planning (Ferguson et al., 2003) . The basic reproductive number (R 0 ), which describes the tendency of a disease to spread, has been estimated for smallpox from historical data and outbreaks in developing countries (Gani and Leach, 2001; Eichner and Dietz, 2003) . Because R 0 is a function of the contact rate between individuals, it can be affected by changes in the environment (Anderson and May, 1991) . A potentially important difference between contemporary environments and those used to estimate R 0 is that today many buildings, including hospitals, mechanically recirculate air. If smallpox was almost entirely transmitted by mucosal contact with large droplets (aerodynamic diameters >10 µm), which can only occur following "face-toface" exposure over distances of a few feet, then change in the built environment would not change the contact rate between individuals. If, however, smallpox was frequently transmitted from person-to-person by airborne droplet nuclei [fine particles with aerodynamic diameters of ≤2.5 µm capable of remaining suspended in air for hours and of depositing in the lower lung (Hinds, 1999) ] then mechanically recirculated air systems would increase the contact rate, R 0 , the risk of epidemic spread, and the difficulty of hospital infection control. Unfortunately, leading authorities disagree regarding the relative importance of fine and large particle routes of transmission; some state that smallpox was transmitted primarily via airborne droplet nuclei, (Henderson et al., 1999) while others emphasize "face-to-face" contact and state that, airborne transmission was rare (Centers for Disease Control, 2002; Mack, 2003) . This paper reviews the evidence for each of these modes of transmission. Prior to Jenner, variolation, (Fenner et al., 1988) inoculation of variola into the skin or nasal mucosa, was used to reduce the risk of smallpox. Jenner himself was variolated as a child. Skin inoculation with a small amount of fresh pustule fluid, likely to have contained large numbers of infectious virions, produced a local lesion with satellite pustules, but generalized rash was reported to be less severe and mortality rates were usually 10-fold lower than with naturally acquired disease (Fenner et al., 1988) . In China, variolation was frequently performed by inoculation of the nasal mucosa. Some accounts describe blowing carefully aged scabs compounded with plant material into the nose (MacGowan, 1884). Other reports suggests that nasal insufflation was considered relatively ineffective and that nasal insertion of cotton pledgets impregnated with powdered scabs or smeared with vesicle contents was preferred (Wong and WU, 1936; Miller, 1957) . Descriptions of the latter method do not include ageing infectious material before use. Because natural infection was thought to occur via large droplets deposited on the upper respiratory mucosal, the success of nasal inoculation in producing low mortality rates has been hard to understand. A theory suggested by Henderson to the author of a smallpox history, (Hopkins, 1983, p. 114 ) "is that virus inhaled naturally was in sufficiently small particles to be deposited deep within the lung, whereas particles inoculated by nasal insufflation may have been much larger and were likely to implant in the nose or throat where [only] a local lesion might be produced." The relative importance of age and health of inoculated subjects, infectious dose, and route of exposure are not known. However, it appears that inoculation via the skin or nasal mucosa tended to produce modified disease. If true, this would indicate that natural transmission did not occur via direct skin or mucosal contact. Figure 1 shows graphically a how these different routes of exposure may have produced altered patterns of viral replication within the host and resulted in different risks of extensive viremia and severe disease. If natural smallpox was initiated through the upper respiratory mucosa, then an early asymptomatic mucosal infection would be expected. To investigate this, Sarkar and colleagues performed pharyngeal swab surveys of household contacts (Sarkar et al., 1973a 4-8 days following onset of rash in the index cases. They found that contacts with positive throat cultures often did not develop smallpox. In one survey, (Sarkar et al., 1973a) 10% (Westwood et al., 1966) of 328 contacts had positive swabs, but only 12% (Kaplan et al., 2002) of those with positive swabs developed smallpox. Among 59 unvaccinated contacts 27% (Miller, 1957) were culture positive, but only one developed smallpox. All subjects were vaccinated at the time of examination. However, vaccination four or more days after exposure is usually considered to be too late to prevent disease. The observation that disease did not develop in 94% of persons with mucosal infection suggests that, even in unvaccinated contacts, mucosal infection may not have been sufficient to initiate disease. Sarkar and colleagues also showed that the oropharyngeal excretion of virus was greatest during the first days after the rash erupted and generally resolved at most 2 weeks following onset of rash (Sarkar et al., 1973b) . Rao et al. found that oropharyngeal excretion was greatest in the most severe, hemorrhagic cases and corresponded with the period of infectiousness (Rao et al., 1968) . In contrast to oropharyngeal excretion, scabs contained large quantities of virus regardless of disease severity and were shed for another week or more after throat cultures were negative. Scabs alone, however, were not associated with further cases (Rao et al., 1968; Mitra et al., 1974) . The apparent lack of infectiousness of scab associated virus has been attributed to encapsulation with inspissated pus (Fenner et al., 1988 ). Henderson's theory about the importance of small particles may provide a straightforward mechanism for why encapsulated virus, simply by entrapment in large particles, had low infectious potential. Sarkar et al. (1973a) were concerned that asymptomatic contacts could have been infectious because their throat swab viral titers were similar to those of milder smallpox cases. A paradox arose from these data because there was never evidence of infection arising from asymptomatic household contacts. Yet, oropharyngeal secretions were thought to be the primary source of infectious virus particles. An explanation may be that oropharyngeal excretion of virus was merely temporally correlated with excretion of virus from elsewhere in the respiratory tract and not the actual source of fine particles virus aerosols. The large spray of particles from sneezing visualized by high speed photography consists of particles down to about 10 µm in diameter (Papineni and Rosenthal, 1997) . Smaller particles may also be dislodged from the upper airways by the turbulence of sneezing, coughing, and talking, but will mostly be larger than 2.5 µm in diameter. Recent studies, however, show that the healthy lung generates abundant fine particles (100-1000/l with size <0.3 µm diameter) during normal breathing (Fairchild and Stampfer, 1987 ) that do not arise from the oropharynx; condensates of these particles are the subject of recent reviews (Mutlu et al., 2001; Hunt, 2002) . Such particles could carry variola virus (0.2-0.3 µm diameter), would remain airborne in indoor air for many hours, and would be deposited primarily in the lower airways after inhalation. There is some evidence that variola was present in the lung and potentially available for aerosolization. Animals infected by inhalation produced high concentrations of variola in the lung (Hahon and Wilson, 1960) . Fenner et al. (1988) regarded bronchitis and pneumonitis as a part of the normal smallpox syndrome, especially in the more severe cases which were also the most infectious, (Rao et al., 1968) although specific lesions were less frequent in the lower trachea and bronchi. Systematic evaluations of viral excretion in the lower respiratory tract of non-fatal cases were not reported. Thus, if some degree of pneumonitis with pulmonary excretion of virus and exhalation of fine particle variola aerosols was a feature of clinical smallpox but was not a feature asymptomatic household contact with positive throat cultures, then the paradox would be resolved. Air sampling for viruses is a difficult undertaking and the literature on the subject remains sparse in comparison with that for bacteria and fungi (Sattar and Ijaz, 2002) . Only three attempts to detect airborne variola were published. The earliest attempt used highly inefficient methods and was negative (Meiklejohn et al., 1961) . In a subsequent study, Downie and colleagues used short duration, low volume air sampling with liquid impingers and obtained 5 positive samples out of 47 attempts to sample exhaled breath of patients (Downie et al., 1965) . Assuming that each positive sample represented a single infectious particle, the concentration of airborne infectious particles was 0.85/m 3 ; higher concentrations were observed close to shaken bed sheets. Concentrations were likely to have been underestimated because of several frequently encountered problems with air sampling for viruses including failure of impingers to retain particles less than 1 µm in diameter that represent the majority of particles in exhaled breath, culture of only a portion of the impinger fluid, uncertain suitability of sampling fluid for virus survival, and loss of infectivity due to sampling trauma (Spendlove and Fannin, 1982) . In the 1970s, Thomas Figure 3 .1] appears to have frequently been less extensive after dermal inoculation and nasal insufflation compared with naturally acquired infection. This may have been due to less extensive lymphatic replication of virus and limited viremia by dermal and nasal routes as compared with infection via lower respiratory tract deposition. The size of the arrows represents the historically reported proportions of cases following each pathway. The size of the X on each image represents the reported mortality rate from each pathway. For natural infection, the ordinary-type rash and flat and hemorrhagic rashes are shown. efficiency for submicrometer particles) for long duration large air volume viral sampling (Thomas, 1970a) . He showed that 23% of naturally airborne rabbit pox particles were ≤2.5 µm and 71% were between 2.5 and 10 µm (Thomas, 1970b) . Both Thomas and Westwood et al. (1966) in a room supplied with 10 ACH containing 7-9 infected rabbits. Westwood et al. probably obtained higher concentrations because they used an electrostatic precipitator allowing higher efficiency collection of submicrometer particles compared with Thomas's slit sampler. Thomas also studied convalescent cases of variola minor (Thomas, 1974) . One patient with relatively active lesions produced an average concentration of approximately 1 PFU/m 3 . Unfortunately the samples were collected late in the disease when the patient was probably minimally infectious, based on comparison with epidemiological data (Rao et al., 1968; Eichner and Dietz, 2003) . The airborne virus observed appears to have been due to resuspension and is unlikely to be representative of the airborne concentration of respirable variola earlier in the course of the infection. The method used would also not have been able to collect submicrometer viral aerosol particles. Overall, the air sampling studies suggest that animals and people infected with poxviruses generated respirable aerosols, but that air concentrations may have been low, or airborne virus was present in submicrometer particles that could not be collected the instruments available. Because detection of virus aerosols is subject to potentially large losses in sampling equipment, especially when sampling dilute natural aerosols over extended periods, and because plaque assays may not accurately represent the infectivity of virus deposited in human airways at 100% relative humidity, (Spendlove and Fannin, 1982; Ijaz, 1987, 2002) the available data can be considered a lower limit on concentration of infectious natural poxvirus aerosols. Experimental aerosol data suggested that poxvirus, which survived the trauma of artificial aerosolization, remained infectious for significant periods of time. Aerosols of vaccinia demonstrated a half-life of about 6 h at 22 • C and relative humidity ≤50% with reduced stability at higher relative humidity and temperature (Harper, 1961) . Variola appeared to have a similar half-life and not to be affected by relative humidity at 26.67 • C (Mayhew and Hahon, 1970) . Other experiments demonstrated that airborne vaccinia is highly sensitive to inactivation by germicidal ultraviolet light (Edward et al., 1943; Jensen, 1964) . Westwood et al. (1966) demonstrated that inhalation of a single PFU of a submicrometer vaccinia aerosol was sufficient to infect rabbits. Airborne rabbit pox was similarly infectious. They demonstrated rabbit-to-rabbit airborne transmission of rabbit pox in each of seven trials by placing uninfected rabbits in separate cages in the same room with infected animals. They also infected rhesus monkeys using submicrometer aerosols of variola. In one of the earliest extensive animal models of smallpox, Brinckerhoff and Tyzzer (1906) reported the effect of inoculating cynomologus monkeys with variola at different sites. Inoculation of mucus membranes of the lip, palate, and nose produced local lesions, but generalized rash occurred in only 10% of animals. Inoculation through the skin produced a local lesion and a generalized eruption in 70-80% of animals. Animals inoculated by scratching the tracheal mucosa through a rigid bronchoscope all developed a generalized rash, and one developed a variolous bronchitis and pneumonia. Laryngeal instillation of dry pustule contents produced infections while instillation of powdered crusts did not. Inhalation exposures to an atomizer spray of vesicle contents infected only one of five monkeys; however, the particle size distribution and type of atomizer were not reported. Hahon and Wilson demonstrated that infection of Macaca irus with high dose [5 × 10 5 PFU] fine particle (<5 µm) variola aerosols produced a disease that simulated human smallpox (Hahon and Wilson, 1960; Hahon, 1961) . The initial site of virus replication was the lung, with subsequent appearance of virus in the nasopharynx and nares. Peak concentrations of virus per gram of tissue were higher in the lung than in the upper respiratory tract; the peak in lung tissue occurred during the incubation period and lung levels declined during the secondary viremia and exanthem. Whether the time course and viral concentrations in lung in this animal model produced by inhalation of high dose aerosols mimicked that in humans with natural infection is doubtful. However, it may be relevant to the first generation of cases exposed to concentrated aerosols in a biological attack. In a relatively recent experiment, (Kalter et al., 1979) a female chimpanzee became infected with variola while housed in the same room, but without direct contact, with two infected chimpanzees. She developed a generalized rash and was reported to have had more severe constitutional symptoms than the other chimpanzees infected by dermal inoculation or direct contact. The authors concluded that she was infected via aerosol. The animal data show that artificial respirable aerosols were effective means of producing poxvirus infections, that the infectious dose by the airborne route could be very low, and that animal-to-animal airborne transmission of rabbitpox and variola was observed. They also suggest that inoculation of mucus membranes was less effective at producing a generalized rash than was exposure of the lower respiratory tract. Two reports, one from the 1940s and one from the 1960s showed that, during epidemics, staff in smallpox hospitals who had been repeatedly vaccinated sometimes developed malaise, fever, and pneumonitis without evidence of infection with smallpox or other viruses, and without evidence of allergic reaction to other agents (Howat and Arnott, 1944; Morris Evans and Foreman, 1963) . In one outbreak, after investigation of other possible causes, the authors attributed the phenomena to an allergic reaction to inhaled variola. The pulmonary focus of the reaction suggests that there were significant concentrations of respirable variola in the vicinity of smallpox patients. Concentrations of respirable variola high enough to elicit allergic reactions, if true, raise a significant concern for the likelihood of airborne transmission. Fomites, particularly exposure of laundry workers to contaminated bedding, were implicated in a few reported outbreaks (Cramb, 1951) . However, during the eradication campaign careful epidemiologic investigation rarely implicated fomites as a source of infection (Fenner et al., 1988) . Laundry was contaminated by scabs containing large amounts of virus, and with respiratory secretions containing virus in smaller particles (Downie et al., 1965) . Very large particles with diameters greater than 50-100 µm are easily reaerosolized. Thus, the rarity of clear evidence of transmission due to fomites would be surprising, if exposure of upper respiratory mucosa to virus in large particles were an efficient means of initiating infection. However, the probability of reaerosolizing particles ≤10 µm from surfaces is extremely low because surface forces tend to bind particles more avidly the smaller the particle (Hinds, 1999) . Thus, the rarity of smallpox transmission via fomites suggests that mucosal exposure was not the primary means of transmission and is consistent with a preference for infection via the lower respiratory tract. The rarity of transmission on crowded buses and trains could be evidence that airborne transmission was not important. However, Fenner et al. (1988) state that transmission on public transport was rare because patients seldom traveled after becoming ill. They showed that transmission did occur on public transport by reporting a case of confluent smallpox who traveled early in her illness and infected five persons on a bus. If most patients who traveled were convalescent so that they no longer had virus in respiratory secretions and only shed virus in large particles from scabs, which were rarely associated with transmission of infection, (Rao et al., 1968 ) then lack of transmission on buses and trains was consistent with a preference for airborne transmission. Mack (1972) emphasized that 85% of cases had clear-cut exposures to known cases. However, the remaining 15% had no obvious exposure suggesting that a small number of more distant or casual contacts transmitted infection as would be expected if smallpox were transmitted by dilute virus aerosols. For example, in the 1947 New York outbreak one secondary case was seven floors away in the hospital (Weinstein, 1947) . Dispersal of smallpox downwind of hospitals was the only obvious explanation for a small number of cases in a British outbreak (Bradley, 1963; Westwood, 1963) . Unexplained introductions of smallpox into Pakistani towns was greatest in towns with facilities for treatment of smallpox, (Thomas et al., 1972) which may suggest that relatively casual contact, or down wind dispersal were capable of occasionally spreading infection. Some well-known hospital-associated outbreaks make it clear that airborne transmission at a distance of more than a few feet did occur occasionally (Wehrle et al., 1970) . But, these examples were rare. However, because highly infectious disseminators are rare in other airborne infectious diseases, (Riley, 1980; Olsen et al., 2003) the rarity of superspreaders in smallpox is not an indication that transmission by less infectious cases was necessarily by a different route. To examine whether the available data on variola aerosols is consistent with Mack's observation regarding known contacts, we can apply a standard Poisson probability model of airborne infection to estimate how long a susceptible person would need to be in a patient's room to have a reasonably high probability of contracting disease (Riley et al., 1978; Rudnick and Milton, 2003) . If, we assume that inhalation and lower respiratory deposition of one PFU of variola was sufficient to cause infection, as for rabbits exposed to vaccinia and rabbit pox, (Westwood et al., 1966) and if a patient's room contained between 0.5 and 5 PFU/m 3 in particles with a 25% lower respiratory deposition fraction (consistent with the literature discussed above), a susceptible individual breathing at 8 l/min would have needed to spend between 1.7 and 16.7 h in the patient's room to have a 63% probability of becoming infected. Outside of the patient's room, aerosol concentrations would have been much lower. If most patients stayed at home in small buildings or in hospitals without mechanically recirculated air, the risk of infection would have been significantly lower outside of patients' rooms, consistent Mack's (1972) observation that 85% of cases arose from identifiable contacts. Thus, a predominance of identifiable face-to-face contacts among cases is not strong evidence against transmission by fine particle aerosols. The weight of evidence suggests that fine particle aerosols were the most frequent and effective mode of smallpox transmission because this would explain the relatively low mortality after variolation, the rarity of transmission by fomites, resolve the paradox of mucosal infection, and be consistent with "smallpox handler's lung" and with animal and virus aerosol experimental data. Certainly other modes of transmission occurred; fullblown disease could result from inoculation through the skin, the nasal mucosa, or the conjunctiva. Thus, smallpox cannot be classified as an "obligate" airborne infectious disease, such as tuberculosis (Riley et al., 1995) (sometimes referred to as a "true" airborne infection), because it was capable of initiating disease via infection of tissues outside of the lower respiratory tract. However, smallpox also cannot be classified as an isotropic infection (formerly termed "opportunistically" airborne infectious disease) because it appeared not to have been transmitted with equal effectiveness and virulence by all routes, whether aerosol, large droplet, or direct contact and skin inoculation. Smallpox appears to have been most effectively and virulently transmitted by fine particle aerosols and therefore should be classified as an anisotropic infection; an infection where route of transmission influences either virulence and or probability of infection, formerly called a "preferentially" airborne infectious disease. Current recommendations for control of secondary smallpox infections emphasize transmission "by expelled droplets to close contacts (those within 6-7 feet)" (Centers for Disease Control, 2002 Control, , 2003 . Recommendations include vigilant maintenance of standard, droplet, and airborne precautions. However, emphasis on spread via large droplets may reduce the vigilance with which more difficult airborne precautions are maintained. High concentrations of variola in the lung during the incubation and prodromal periods in monkeys after simulated use of variola as a bioweapon (Hahon, 1961 ) may indicate that first generation cases after an attack with a concentrated aerosol may be more infectious than expected based on historical data. Moreover, because airborne precautions are not routine for all hospitalized patients, and because first generation cases will probably not be initially suspected to have smallpox, it is likely that they will not be placed on airborne precautions until well into their infectious period. Therefore, the extent of transmission to a second generation in the contemporary hospital environment may be greater than expected based on historical estimates. These considerations suggest that models of a potential smallpox attack should incorporate an aerobiological perspective to predict how the infection might propagate in the modern environment. It is particularly important to examine smallpox transmission in hospitals because hospitals have previously been identified as the major site of transmission in developed countries and ill patients will inevitably gravitate to hospitals, at least early in the outbreak before alternatives exist (Mack, 1972 (Mack, , 2003 . Additional attention to prevention of airborne transmission in hospitals from unrecognized cases may not only be an important aspect of public health preparedness for smallpox, but may also benefit society by reduced morbidity and disruption from SARS and other emergent airborne infections.

Chitinase Dependent Control of Protozoan Cyst Burden in the Brain

Chronic infections represent a continuous battle between the host's immune system and pathogen replication. Many protozoan parasites have evolved a cyst lifecycle stage that provides it with increased protection from environmental degradation as well as endogenous host mechanisms of attack. In the case of Toxoplasma gondii, these cysts are predominantly found in the immune protected brain making clearance of the parasite more difficult and resulting in a lifelong infection. Currently, little is known about the nature of the immune response stimulated by the presence of these cysts or how they are able to propagate. Here we establish a novel chitinase-dependent mechanism of cyst control in the infected brain. Despite a dominant Th1 immune response during Toxoplasma infection there exists a population of alternatively activated macrophages (AAMØ) in the infected CNS. These cells are capable of cyst lysis via the production of AMCase as revealed by live imaging, and this chitinase is necessary for protective immunity within the CNS. These data demonstrate chitinase activity in the brain in response to a protozoan pathogen and provide a novel mechanism to facilitate cyst clearance during chronic infections.

The brain has unique structures in place to limit access of immune cells and molecules. Although this can provide protection against an overambitious inflammatory response it may also lead to the high prevalence of latent and chronic infections that can persist at this site. Removal of such pathogens has its own particular problems in an organ dense with sensitive neurons and stringent gateways for immune cell infiltration. Toxoplasma gondii is a common intracellular protozoan parasite that forms a chronic infection in the brain for the lifetime of the host. The infection is controlled, in part, through the effector mechanisms of macrophages that result in the conversion of fast replicating tachyzoites to the slow replicating, cyst forming bradyzoites [1] [2] [3] . Cysts can form in all tissues but exist predominantly in the brain for the lifetime of the host requiring a continuous immune response to prevent cyst reactivation and Toxoplasmic encephalitis, a common cause of AIDS related fatalities [4, 5] . The infection-induced immune response in the brain consists of activated CNS resident cells including astrocytes and microglia, infiltrating CD4+ and CD8+ T cells, peripheral macrophages and substantial tissue remodeling [6] [7] [8] . Such immune activity in the brain is often associated with a pathological outcome yet despite the high prevalence of infection Toxoplasma is seemingly controlled without adverse neurological damage. The mechanisms that are involved in the trafficking and control of such a potentially pathological immune response within the CNS are only beginning to be understood [6, [8] [9] [10] [11] . The cyst and cyst-forming bradyzoites are poorly immunogenic [12, 13] and although we have known for some time that T cells are required to prevent cyst reactivation [4, 5, 14] , very little is understood about the biology of this structure in the brain. Although anti-Toxoplasma drugs are available that efficiently control the tachyzoite, there are as yet no therapies available that can effectively remove the cyst form of the parasite. Thus, the continuous presence of Toxoplasma cysts in the brain presents a critical and constant danger for the immune compromised patient. It is widely believed that cysts remain intracellular within neurons possibly minimizing their contact with host defense systems [15] . However it has been known for some time that cyst burden reaches a peak, declines and becomes stable over time pointing to some form of effector mechanism that can target this stage of the parasite [16] . Studies have implicated CD8+ T cell production of perforin in cyst clearance with perforin deficient mice exhibiting higher cyst burden and susceptibility at the chronic stage of infection [17, 18] . Nevertheless, histological analysis from these studies as well as recent live imaging of cell interactions in the CNS [19] demonstrates monocyte accumulation and contact with cysts. In recent years, our understanding of macrophages has expanded and we now appreciate these cells' remarkable plasticity. Thus, although whole populations of macrophages can become polarized to classical or alternative phenotypes associated with protection against protozoan and helminth pathogens respectively, the ability to respond and adapt to local stimuli in the environment is paramount [20] [21] [22] [23] [24] . The role of classically activated macrophages in the control of T. gondii infection is well documented. These cells are a source of IL-12, reactive oxygen and nitrogen species, and GTPases that enable the direct killing of the parasite [6, [25] [26] [27] [28] [29] [30] . However, here we describe a population of CXCR3+ macrophages in the brain following T. gondii infection. These cells express the scavenger receptors MMR and stabilin-1 and produce arginase in response to the presence of Toxoplasma cysts. In addition to these traditional signs of alternative activation, these studies demonstrate that macrophages respond to chitin present in the cyst wall and produce the true mammalian chitinase, AMCase. Finally we show that this chitinase activity destroys cysts and is essential for the control of cyst burden within the chronically infected brain. Recent studies have identified a substantial increase in tissue remodeling in the brain during chronic T. gondii infection [8] . Additionally, there is a continuous need for the clearance of debris from ruptured cysts and dead cells in the brain [31] . To investigate if AAMØ, known for their role in tissue remodeling and homeostatic clearance, are present during such an event in the CNS, macrophage populations from the infected brain were phenotypically analyzed for the expression of known markers of alternative activation. One of the key molecules that has been associated with a tissue remodeling macrophage phenotype in the CNS is the expression of CXCR3 on microglia [32, 33] . CXCR3 is required for protective immune responses to T. gondii primarily due to its role in Th1 cell recruitment and most recently for T cell search strategies in the brain [34] [35] [36] [37] . Indeed CXCR3 and its ligands are significantly upregulated in the brain at a timepoint associated with significant T cell influx into the CNS following infection ( Figure S1A , B) with ,35% of T cells expressing CXCR3 ( Figure 1A ). However, in addition to this well characterized role on T cells, there is a small but distinct population of macrophages that express high levels of CXCR3 (,10% of total macrophages)( Figure 1B ). There is also constitutive, although lower, expression of CXCR3 by CNS resident microglia, which remains unchanged following infection ( Figure S1C ). To confirm that expression of CXCR3 is associated with alternative activation of macrophages the expression of the scavenger receptor 'macrophage mannose receptor' (MMR; also known as Mrc1 and CD206), a key indicator of the AAMØ phenotype [38] was analyzed. Here we show that MMR expression is limited to macrophages and microglia that also express CXCR3 (Figures 1C and S1D ). In contrast these cells did not express IL-10 ruling out an anti-inflammatory phenotype ( Figure S1E ) [39] . Depletion using blocking antibodies to CXCR3 or its ligand, CXCL10 led to a significant decrease in T cell recruitment and a reciprocal increase in parasite burden ( Figure S2 ). However, in addition, the proportion of macrophages in the brain was significantly reduced ( Figure 1D ) despite no defect in macrophage-attracting chemokines ( Figures 1E, F) , confirming a role for CXCR3 in the maintenance of this cell population. To quantify MMR expression by macrophages and microglia in the infected brain, qRT-PCR was performed on magnetically isolated CD11b+ cells from the brains of naive and infected animals. Our results show an approximate 3-fold increase in MMR expression in macrophage populations from infected mice over naïve ( Figure 1G ). Confirmation of this population in the brain was revealed by immunohistochemical analysis. MMR+ macrophages were observed as small and discrete populations of IBA-1+ or tomato lectin+ cells confirming the source of MMR on macrophages or microglia ( Figures 1H, I) . A further functional marker of alternative activation is the scavenger receptor stabilin-1 [40] . Stabilin-1 is involved in the clearance of cell corpses as well as the uptake of extracellular matrix components [41, 42] . Expression of MMR co-localized with stabilin-1 and microglia/ macrophage markers, confirming that these cells display an alternatively activated phenotype ( Figures 1H, I) . These cell populations were frequently found in close proximity with intact and degrading T. gondii cysts in the CNS ( Figure 1J and S3 ). An important feature of AAMØ is the cell's ability to produce arginase-1, which acts on its substrate, L-arginine to produce Lornithine, a precursor to collagen [43] . L-arginine is also the substrate for NO synthase and the two enzymes compete for substrate availability and are regulated by Th1 and Th2 type cytokines [44, 45] . Previous studies have demonstrated that direct infection of macrophages by T. gondii tachyzoites can induce arginase expression via STAT-6 dependent and independent pathways [46] [47] [48] . Furthermore these studies imply that such an induction is a survival strategy enlisted by the parasite to inhibit killing via NO. To assess whether or not macrophages and microglia in the infected brain produce arginase, CD11b+ BMNCs from infected mice were isolated and analyzed for arginase-1 expression by qRT-PCR. Our results show almost a 2fold increase in arginase-1 expression in cells from infected brains over naïve ( Figure 1K ). Thus, during chronic T. gondii infection there is a population of AAMØ in the CNS characterized by expression of CXCR3, MMR, stabilin-1 and the production of arginase-1. Alternatively activated macrophages secrete an active chitinase in the CNS in response to chitin in the cyst wall During chronic infection there are several forms of the parasite that could be the source of the infection-associated stimulus for Described here is a novel mechanism of protozoan cyst clearance in the CNS during chronic infection. These data show the presence of a population of alternatively activated macrophages in the brain that secrete the active chitinase, AMCase, in response to chitin in the cyst wall. Using both chemical and genetic inhibition in vitro, it is revealed that this enzyme is required for efficient degradation and destruction of the cyst. The necessity for AMCase is demonstrated in vivo, as the absence of the enzyme resulted in a significant increase in cyst burden and decrease in survival during chronic infection. Together, these data identify an important mechanism of parasite control and cyst clearance in the CNS. Currently, no therapies exist that lead to the total clearance of this parasite from the brain. Therefore, developing an understanding of the natural mechanisms of cyst clearance has the potential to lead to new and effective therapies for this and other chronic infections. alternative activation of macrophages in the CNS. Since latent cysts are the most prevalent form of infection in the brain, an attractive candidate for the source of this stimulus is the presence of chitin in the cyst wall [49, 50] as it has been shown that the presence of chitin induces the recruitment of macrophages that have an alternatively activated phenotype [51, 52] . To determine if sources of T. gondii can induce alternative activation, tachyzoites, bradyzoites, and cysts were added to bone marrow derived macrophage (BMDM) cultures and the production of urea, a downstream product of arginase activity, was measured [53] . In addition, soluble antigen derived from freeze-thawed tachyzoites (sTAg) and whole cysts (cystAg) was assessed for their ability to induce urea (Figure 2A) . Our results show a baseline production of urea in unstimulated (media) macrophage cultures, possibly due to the presence of M-CSF [54] . This significantly increased (p,0.001) during AAMØ polarization with IL-4. Despite the known ability of tachyzoites to induce arginase production [46] , tachyzoite infection of macrophages did not lead to significant production of urea ( Figure 2A ). This can be attributed to the use of a type II strain which is a weak inducer of arginase-1 [47, 48] . Importantly, addition of cysts or cyst antigen but not ''naked'' bradyzoites, did lead to a significant increase in urea production although not as great as induction of alternative activation by IL-4 [38, 55] . This points to components of the cyst wall as the stimulus for AAMØ. Taken together, these data demonstrate that macrophages can be alternatively activated by the presence of T. gondii cysts, but not free replicating parasites. Chitinase activity has been demonstrated in certain populations of AAMØ in both mice and humans [52, 56, 57] . Chitinolytic activity by macrophages has also been implicated in host defense against chitin-containing fungal pathogens [56, 58] . To test whether or not chitinase activity is induced by Toxoplasma infection, a chitinase assay was performed on whole brain lysates from naïve and infected animals. Three chitin substrates labeled with 4-methylumbelliferone (4MU) were used to assess the type of chitinase activity present. Upon hydrolysis, 4MU is released and can be measured fluorometrically to determine chitolytic activity. Our data reveal that chitinase activity is significantly increased in the brains of infected mice as compared to the naïve group in only one of the three substrates ( Figure 2B ). This substrate, 4MU Nacetyl-b-D-glucosaminide, is suitable for detection of exochitinase activity where the enzyme degrades the non-reducing end of the chitin [59] . Several studies have linked chitinases and chitinase-like proteins to inflammation [58] [59] [60] . This family of 18 glycosyl hydrolases is typically induced during Th2 type immune responses and plays a role in tissue remodeling, fibrosis, and the modulation of both the innate and adaptive immune response [60] . Acidic mammalian chitinase (AMCase) and chitotriosidase (CHIT1) are unique members of this family in that they possess an enzymatically active domain that hydrolyzes the b 1-4 linkages that exist in chitin [56, 58] . Analysis using qRT-PCR demonstrated a significant upregulation of AMCase but not CHIT1. In addition, the chitinase-like protein, Ym-1 (Chi313) was also upregulated following infection ( Figure S4 ). This molecule is known to inhibit IL-12 production and induce alternative activation in macrophages [25, 61, 62] . In contrast to Ym-1, that has been associated with AMCase production by macrophages in the lung and airway [63] , Ym-2 and BRP-39 were not upregulated in infected brains ( Figure 2C , S4). Since chitin has been shown to activate and recruit AAMØ, it is possible that the cyst wall may serve as the stimulus for chitinase activity in this population of cells. To test this further, BMDM were co-cultured with different forms of the parasite. The addition of tachyzoites, bradyzoites or sTAg did not lead to chitinase production ( Figure 2D ). In contrast, live cysts and cyst antigen led to a significant increase in chitinase activity that was abolished following chitinase treatment of cysts ( Figure 2D , S4). Furthermore, treatment with IL-4 to induce alternative activation in macrophages did not lead to increased chitinase activity. Indeed measurement of IL-4, IL-4Ra and the IL-4dependent RELM-a [64] [65] [66] in the brains of chronically infected mice showed no significant increase over naïve mice ( Figure S4 ). These data suggest that the presence of chitin in the cyst wall induces a phenotype of macrophage characterized by the production of the enzymatically active chitinase, AMCase and is distinct from IL-4 induced activation. Previous work has shown macrophages in close association with rupturing cysts [17, 31] and the presence of an active chitinase could point to a role for these cells in the breakdown of cysts within the brain. Recognition of chitin by macrophages is size dependent and likely contact dependent [49, 58, 67] . To test this, we cocultured cysts separated from macrophages using 5 mm transwell inserts and assayed for urea and chitinase activity as previously described ( Fig. 2A, 2D ). Our results show no increase in either urea production or chitinase activity from macrophages that have been separated from cysts, confirming that the observed alternative activation is dependent on contact with cysts or cyst antigen. Immunohistochemical analysis of the location of AMCase secreting macrophages in the infected brain shows them in close proximity with tissues cysts (Figures 3A-C) . As a proportion of macrophages and microglia in the brain, alternatively activated cells are in the minority and it was not possible to find such cells in the naïve brain. However, cysts are easily identifiable with a highly spherical distinct morphology, can stain non-specifically and specifically with antibodies and individual bradyzoites within the cyst are visible by DAPI staining. Examination of chitinase localization in the infected brain revealed distinct cytoplasmic staining of several cells, nearly all of which were within 75 mm of a cyst (Fig. 3A) . Although there were cells that were AMCase positive yet did not stain positively for macrophage/microglial markers, there were clearly several macrophages in close association with cysts that displayed chitinase activity polarized to the cyst wall ( Figure S5 , arrows). AmCase activity was also observed in macrophages surrounding cysts that seemed to be in the process of lysing or cysts that had been lysed (Figures 3B,C and Video S1). The examples provided show the destruction of the spherical cyst (Video S1) and escaping parasites visualized using anti-Toxoplasma antibodies. Directly at the point of rupture there are AMCase expressing macrophages (Figures 3B). Taken together, these data suggest that the induction of chitinase activity in macrophages occurs in close proximity with the cyst wall and that this distinct population of macrophages is responsible for attacking the long-term chronic cyst stage of Toxoplasma via chitinase activity. The prevailing view is that cysts in the brain remain intracellular within neurons and that CD8+ T cell production of perforin is responsible for cyst clearance in the brain although the exact mechanism of cyst destruction has yet to be described [17, 18] . In order to determine whether or not macrophage chitinase activity could be responsible for the direct lysis of cysts, BMDM were co-cultured with cysts; with and without the chitinase inhibitor allosamidin. Cultures were observed for 14 hours capturing images every 10 minutes. Cysts observed in the absence of macrophages remained intact for the entire time course suggesting no parasite intrinsic mechanism of cyst destruction ( Figure 4A ; Video S2). In contrast, the addition of 20 mg/ml trichoderma chitinase to cyst cultures led to rapid rupture of the cysts within an average of 4 hrs, releasing bradyzoites into the media ( Figure 4B ; Video S3). Strikingly, cysts that were cultured with macrophages came under vigorous attack. This involved efficient and rapid migration of macrophages toward the cyst creating clusters of macrophages that could be seen pulling at the cyst wall ( Figure 4C ; Video S4). In these cultures most of the cysts were destroyed during the observation period with the average survival time of 9.5 hours ( Figure 4E ). In contrast cysts cultured in the presence of macrophages and the chitinase inhibitor allosamidin survived significantly longer than in untreated cultures. Although there appeared to be no defect in the recruitment and activity of macrophages to cysts with similar clustering and 'pulling' of the cyst wall, the majority of cysts survived the entire 14 hour period when treated with either 100 mM or 10 mM allosamidin ( Figure 4D ; Video S5). Decreasing concentrations of allosamidin led to a dose dependent decrease in cyst survival time ( Figure 4E ). These results demonstrate that macrophages can induce cyst lysis in a chitinase dependent manner. Although both AMCase and CHIT1 are upregulated in certain bacterial and nematode infections [68, 69] only AMCase was significantly increased in the brain following Toxoplasma infection ( Figure 2C ). To confirm that AMCase is responsible for the observed chitinase activity, we performed a chitinase assay similar to that in Figure 1D using BMDM from WT and AMCase null mice ( Figure 5A ). Our results reveal a severe defect in chitinase production by AMCase null macrophages. Indeed, these cells showed a significantly lower baseline level of chitinase and were unresponsive to the addition of cysts. To test whether the ability to destroy cysts is dependent on this enzyme, BMDM from WT and AMCase2/2 mice were fluorescently labeled and cultured with Me49-RFP expressing cysts and cyst lysis time imaged as before. Using fluorescently labeled parasites enhanced the ability to see escaping parasites from lysing cysts. Results, as before, demonstrated that WT non-polarized macrophages were able to lyse cysts in ,10 hours ( Figures 5B and 5D ; Videos S6 and S10). In contrast, cysts cultured with AMCase2/2 macrophages had a significantly increased survival time over WT macrophages consistent with AMCase being the source of chitinase activity required to lyse cysts ( Figures 5C and 5D ; Videos S7 and S11). To determine the requirement for macrophage polarization in their ability to lyse cysts, macrophages were treated with cytokines to polarize them to classical or alternative phenotypes prior to cyst addition. In line with the lack of chitinase induction, IL-4 priming had no effect on cyst survival time, suggesting that cytokineinduced alternative activation does not enhance the ability to destroy cysts ( Figures 5D; Video S9) . In contrast, macrophages that were classically polarized with LPS and IFN-c showed a defect in chitinase activity and cyst destruction (Figures 5D and 5E ; Video S9). Suggesting that polarization of macrophages is required but that chitin is the most likely source of alternative activation and not IL-4. Taken together, these data demonstrate that macrophages lyse cysts in an AMCase-dependent manner in vitro. The consequences of chitinase dependent cyst lysis in the CNS could potentially benefit either the host or the parasite. If the escaping bradyzoites were quickly killed by macrophages or associated immune cells, we would expect this mechanism to benefit the host and result in a lower parasite burden. Conversely, if bradyzoites are able to propagate and infect new cells, this could be a mechanism that promotes the persistence of the parasite in the brain. To investigate the role of AMCase in the brain in vivo, we infected WT and AMCase deficient mice and analyzed the immune response and parasite burden in the absence of chitinase activity. To determine if AMCase is required during the acute stage of infection, tissue samples from lungs were taken at day 7 and analyzed for parasite burden by qPCR. No significant differences in lung parasite load were found and serum cytokine concentrations were equivalent throughout acute infection ( Figure 6A-F) . Thus a lack of AMCase does not lead to deficient immune responses early on during infection in the periphery. At 5 weeks post infection, when systemic inflammation has been controlled and parasites are located solely in the brain predominantly as cysts containing bradyzoites [67, 70] parasite burden was evaluated. In the absence of AMCase, there was a significant increase (p = 0.0014) of approximately 2-fold in the total number but not the size of cysts in the brain (Figures 6G and 6H) . Differences in cyst burden were not observed at 3 weeks post infection ( Figure S6 ), a period representing the transition between acute and chronic infection, further suggesting that the increase in cyst burden is occurring due to events within the CNS during chronic infection. In addition, total parasite burden in the CNS as measured by qPCR was significantly greater by more than 2-fold (p = 0.0055) ( Figure 6I ) correlating with the appearance of more cysts histologically ( Figure 6J ). In addition, parasite burden was evaluated using RT-qPCR with stage-specific primers identifying tachyzoite (SAG1), bradyzoite (SAG4), and cyst (MAG1) specific transcripts [71] (Figure S6 ). Our results show similar increases in parasite burden for all three transcripts, suggesting that cyst lysis is also an important mechanism to control the cell invasive forms of the parasite. Flow cytometric analysis revealed no differences in infiltrating CD4+, CD8+ T cells, or macrophage populations ( Figure 6K) . Therefore, the increase in parasite burden is not due to a defect in infiltrating effector immune cells. Furthermore, AMCase2/2 mice failed to survive and succumbed to infection beginning at six weeks (p = 0.0177) ( Figure 6L ). Although some acute mortality was noted over several experiments significance was only achieved when chronic mortality was included. These results demonstrate that AMCase activity is required for the protective immune response to T. gondii during chronic infection in the brain and that AMCase mediated cyst lysis in the CNS is a beneficial mechanism for the host to control parasite burden at non-lethal levels. Chronic infections represent a continuous battle between the host's immune system and pathogen replication. Many protozoan parasites and fungal pathogens have evolved a cyst lifecycle stage that provides it with increased protection from environmental degradation as well as endogenous host mechanisms of attack [72] [73] [74] . In the case of Toxoplasma, these cysts are predominantly found in the immune protected brain making clearance of the parasite more difficult and resulting in a lifelong infection. Here we describe three novel findings 1) despite a dominant Th1 immune response during Toxoplasma infection there exists a population of macrophages in the infected brain which display a distinct alternatively activated phenotype; 2) these cells are responsible for chitinase dependent lysis of Toxoplasma cysts and 3) this chitinase activity is through the production of AMCase which is required for protective immune responses. Multiple studies have demonstrated the role of CXCR3 and its ligands in the migration of activated T cells during Th1 immune responses including to sites of infection [75, 76] . It is also known that the chemokines CXCL9 and CXCL10 are induced by the presence of the proinflammatory cytokine, IFN-c [34, 36, 37, 77] . More recently, however, the function of this family of chemokines has expanded to include neural-glial signaling following brain lesion where injured neurons upregulate CXCL10 and recruit CXCR3 expressing microglia to phagocytose denervated dendrites [33] . Consistent with this, another recent study has implicated CXCR3 in the function of perivascular macrophages and their ability to remodel the vasculature during stress [78] . We noted upregulation of CXCR3, CXCL9 and CXCL10 in the brain during chronic Toxoplasma infection. Furthermore, CXCR3 was preferentially expressed on macrophages expressing the scavenger receptors MMR and stabilin-1, suggesting an alternatively activated phenotype for these cells. Previous studies have established important functions of AAMØ in the context of helminth infection and wound healing [43, 79, 80] however not during an infection that generates such a polarized Th1 immune response such as Toxoplasma. T. gondii is known to exploit the arginine metabolic pathway and induce arginase-1 expression in macrophages thereby suppressing nitric oxide production but this does not lead to the alternative activation of these cells [46] [47] [48] . Instead our data point to a role for the cyst being the source of alternative activation of macrophages and the subsequent ability of these cells to lyse cysts via destruction of the chitin in the cyst wall. Thus, we observed a contact dependent significant increase in arginase activity following treatment with cysts and cystAg suggesting that this induction is not a result of infection by the replicating parasite, but rather by the presence of chitin in the cyst wall. The weak induction of arginase activity observed also points to a limited role for arginase-1 in the chitin-induced phenotype. Chitin is found in the exoskeletons of insects, fungal cell walls, sheaths of parasitic nematodes, and is a component of the T. gondii cyst wall [49, 50] . The presence of this exogenous molecule can induce the recruitment of AAMØ, basophils, neutrophils, and eosinophils [52, [81] [82] [83] [84] . Active chitinases such as AMCase and chitotriosidase are secreted by macrophages in response to chitincontaining pathogens and has been shown to inhibit hyphal growth of chitinous fungi such as Candida and Aspergillus [56, 58] . Despite the link between chitin and the recruitment of type 2 inflammation in the lung [52] , a recent study has demonstrated no role for AMCase in the generation of allergic airway pathology [85] . In this study we demonstrate for the first time, macrophage chitinase activity in response to a protozoan pathogen. Chitin recognition is thought to be a size dependent process and involve a combination of TLR2 and scavenger receptors such as MMR and dectin-1 [49, 58] . Here we have demonstrated the presence of such scavenger receptors in association with cysts and it will be of interest in future studies to determine the role of these molecules in cyst containment during Toxoplasma infection. Independent of the receptors involved it is likely that this is a contact-dependent process and indeed, cysts were unable to induce urea production or chitinase activity in macrophages when separated by transwell membranes. Furthermore, analysis of the location of AMCase producing cells in the brain finds them reliably close to cysts and often in direct contact with reactivating or rupturing cysts. These images show that despite the presence of many DAPI positive cells surrounding the cyst structure, the escape of parasites through the cyst wall occurs juxtaposed to the macrophage or AMCase activity. Our data suggest that the presence of cyst antigens induces alternative activation of macrophages and that these antigens are required for macrophages to produce chitinase even in the presence of IL-4. Thus alternative activation of macrophages is not sufficient for AMCase production and chitinase activity. The significant increase of the non-enzymatically active chitinase-like molecule YM-1 in infected brains is consistent with previous reports of AMCase and YM-1 being co-expressed specifically in macrophages and not epithelial cells [63] . Live imaging in vitro demonstrated AMCase dependent degradation of cysts as shown using both a chitinase inhibitor and AMCase2/2 macrophages. Although there is no evidence of an active chitinase produced by T. gondii (ToxoDB), the similar cyst survival times observed for AMCase-null macrophages and chitinase-inhibited macrophages exclude the possibility that bradyzoites are the source of enzymatic activity and are breaking out of the cyst. The chitin dependent induction of chitinase activity implies that macrophages have access to the chitin in the cyst wall Figure 6 . AMCase2/2 mice have a higher parasite burden in the brain and succumb to infection during the chronic stage. (A-F) C57Bl/6 (WT) and AMCase2/2 mice were infected with the Me49 strain of T. gondii. Serum was isolated from whole blood samples at days 3, 7 and 14 post infection and analyzed for (A) IFN-c, (B) IL-6, (C) MCP-1, (D) TNF-a, (E) IL-12p70 (F) At day 7 DNA was isolated from the lungs and analyzed for parasite burden using qPCR. Results are displayed as parasites per mg tissue. (G-K) C57Bl/6 (WT) and AMCase2/2 mice were infected with the Me49 strain of T. gondii and sacrificed at 5 weeks following infection. Brains were harvested and analyzed for cyst burden, cellular composition and histology. (G) Cyst counts were obtained from homogenized whole brain samples. (H) Cyst area, 20 cysts from each mouse were photographed microscopically and cyst area was determined using ImageJ software. (I) DNA from brains of WT and AMCase2/2 was isolated and analyzed for parasite burden using qPCR. (J) Whole brains were fixed, frozen and stained for H&E to examine cyst burden and pathology. (K) BMNCs were isolated and analyzed for expression of CD4+ T cells, CD8+ T cells, macrophages (CD45 hi /CD11b+) and microglia (CD45 hi /CD11b+) by flow cytometry. Significance was determined using log rank test with p = 0.0177. Data are representative of at least 2 individual experiments with a minimum of n = 4 and are represented as mean 6 SEM, ns = not significant, * p,0.05, ** p,0.01. (L) Survival data from C57Bl/6 (squares, n = 7) and AMCase2/2 (triangles, n = 7). Data are representative of 4 individual experiments with C57Bl/6 (n.40) and AMCase2/2 (n = 40) and significance tested using Logrank (Mantel-Cox) and Gehan-Breslow Wilcoxan test * p,0.05. doi:10.1371/journal.ppat.1002990.g006 prior to chitinase-mediated cyst destruction. The prevailing view from ultrastructure studies is that cysts remain intracellular within neurons [15, 86, 87] (A. Koshy and J. Boothroyd personal communication) yet analysis of cyst burden over time shows a reduction in cyst numbers implying some form of effector mechanism in place [16] . Several studies have demonstrated perforin dependent control of cyst burden during chronic infection [17, 18] . We suggest that instead of a direct effect of perforin on cysts, it is most likely that perforin production by CD8+ T cells may initiate this process by lysing the cyst infected cell, thus exposing the cyst wall to chitinase activity from macrophages. This model would explain the many observations of macrophages in close association with rupturing cysts [8, 17, 19] . Of note, we found that BALB/c macrophages that are more easily alternatively activated had enhanced cyst lysis activity when compared to C57Bl/6 ( Figure S7 ). This may be one explanation for the increased resistance to toxoplasmic encephalitis in BALB/c mice [30, 88] . Although AMCase activity is not required for protective immunity during acute infection it is required for protection during the chronic stage of infection. Our observation of a higher cyst count and parasite burden as well as decreased survival in AMCase-null mice points to a specific and important role for chitinase mediated cyst lysis in the brain. Thus, within the brain, cyst containment seems as important as the killing of free parasites in the control of pathology. In addition, continuous chitinmediated attack by macrophages and the release of parasites from latent cysts will provide a constant source of antigenic stimulation for the immune response. This latter discovery may provide an explanation for the continuous recruitment of T cells into the brain. It has been apparent for some time that cyst numbers in the brain can be controlled endogenously yet identification of the exact effector mechanisms has not been so apparent. In these studies we demonstrate the presence of a distinct population of macrophages in the brain during chronic Toxoplasma infection, which express CXCR3, MMR, stabilin-1 and arginase-1. Furthermore these macrophages have chitinase activity, are localized to cysts and are observed in association with cyst degradation. The mechanism of cyst lysis is dependent on AMCase and this enzyme is required for survival during chronic infection to reduce parasite burden. The continuous presence of Toxoplasmic cysts in the brain and other tissues presents a constant threat of reactivation to the immune compromised patient. Mechanisms that enhance cyst removal or prevent their reactivation during Toxoplasma or other protozoan infections would provide a novel line of anti-parasitic therapies. The experiments in this study were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the Institutional Animal Care and Use Committee at University of California, Riverside. All efforts were made to minimize animal suffering during the course of these studies. T. gondii Pruigniund and RH strains were maintained in vitro as previously described [89] . Soluble toxoplasma antigen (sTAg) was prepared from RH strain tachyzoites as previously described [90] . The Pruigniud strain was used for in vitro tachyzoite infections at a ratio of 3:1. The Me49 strain of T. gondii was maintained in infected Swiss Webster and CBA/CaJ mice. For infection, brains from infected mice were removed placed in 3 ml sterile 16PBS and passed 3-5 times through an 18.5 gauge followed by 20.5 and 22.5 gauge needle. The number of cysts in a 30 ml aliquot was determined microscopically. Brain suspensions were adjusted to 100 cysts/ml and mice were infected each with 20 cysts intraperitoneally. Infection studies of C57Bl/6 and AMCase null mice were conducted at least four times with a minimum of 7 biological controls. C57Bl/6, CBA/CaJ (Jackson, Bar Harbor, ME) and Swiss Webster mice (Charles River, Wilmington, MA) were maintained in a pathogen free environment under IACUC established protocols at the University of California Riverside. AMCase-null mice were generated by targeting exon 5 using loxP/CRE recombination as previously described [85] . The AMCase gene deleted mice were of a mixed background, C57BL/6NTac:129SvEvBrd, and were backcrossed to C57/BL6 for at least 10 generations. These mice were generated and maintained under IACUC protocols established by Pfizer. A single cell suspension from spleens was prepared by passing through a nylon 40 mm cell strainer (BD, San Jose, CA). Suspensions were washed with RPMI complete (10% FCS, 1%Pennicilin/Streptomycin, 1% Glutamine, 1% Sodium Pyruvate, 1% nonessential amino acids, 0.1% B-mercaptoethanol) (Life Technologies, Grand Island, NY) and centrifuged for 5 minutes at 1200 RPM at 4uC. Red Blood Cells were lysed using 0.86% ammonium chloride solution, centrifuged and resuspended in RPMI complete. BMNCs were prepared as previously described [89] . BMNCs or splenic cells were incubated with various conjugated antibodies against CXCR3, CD3, CD4, CD8, CD11b, IL-10, and CD45, (eBioscience, San Diego, CA) and MMR (Biolegend, San Diego, CA). Cells were analyzed using the BD FACSCanto II flowcytometer (BD Biosciences, San Jose, CA) and FlowJo analysis software v.8.7.3 (Treestar Software, Ashland, OR). Cell populations were determined by gating on CD4+, CD8+, CD45 hi / CD11b+ (macrophages) and CD45 int /CD11b+ (microglia) from live cell gate. Total RNA from brain tissue samples was extracted with TRIzol reagent (Life Technologies, Grand Island, NY) according to manufacturer's instructions. DNase1 treatment and first strand cDNA synthesis was performed using cDNA synthesis kit (BioRad, Hercules, CA) according to the manufacturer's instructions. CXCR3, CXCL9, CXCL10, CCL2, CCL5, AMCase, Arg1, and Chit1 specific primers for Real Time PCR were purchased from IDT's primer Quest (http://www.idtdna.com/Scitools/ Applications/Primerquest/). Primer sequences were as follows: Immediately following excision, brains were bisected sagitally and flash-frozen in cold isopentane. Frozen brains were then put into standard Tissue-Tek cryomold and filled with Optimal Cutting Temperature (OCT) solution (Tissue-Tek, Torrance, CA) and put on dry ice and subsequently stored at 280uC. Serial sections of 10-20 mm were prepared on a standard Cryostat machine (LEICA/CM1850, Simi Valley, CA). Frozen tissue sections were fixed 75% acetone/25% ethanol then blocked in 10% donkey serum prior to incubation with purified antibodies. Purified primary antibodies for Iba-1 (Wako, Richmond, VA), CXCR3 (Life Technologies, Grand Island, NY), MMR (AbD Serotec, Raleigh, NC) arginase-1, AMCase and stabilin-1 (Santa Cruz Biotechnology, Santa Cruz, CA) as well as biotinylated tomato lectin (Sigma-Aldrich, St. Louis, MO) were incubated with tissue samples for 2 h at RT or overnight at 4uC, and followed with appropriate secondary antibodies conjugated to Alexa 488, Alexa 568, or Alexa 647 at 2 mg/mL (Life Technologies, Grand Island, NY). Samples were mounted in Prolong Gold with DAPI (Life Technologies, Grand Island, NY) for nuclear counterstaining. Images were collected on a Leica SP2 scanning confocol microscope (Leica Optics, Germany), and analyzed using Improvision Volocity 5.0 (Perkin-Elmer, Waltham, MA). Distance of cells from cysts were calculated from confocal images of at least 12 cysts and at least 6 AMCase expressing cells per cyst. Parasite burden was measured by amplifying the T. gondii genes B1, SAG1, SAG4, or MAG1 by real-time PCR as previously described [71, 89] . In vivo peptide blocking C57BL/6 mice were infected i.p. with 10 4 Pruigniund tachyzoites. At day 21, 23, 25, and 27 post infection the animals were injected i.p. with either 0.5 ml a-CXCL10 (0.5 mg/mL), 0.5 mL a-CXCR3 (polyclonal), or 0.5 ml PBS as previously published [76, 91] . The mice were sacrificed on day 28 p.i. and brains were excised for flow cytometric analysis, and parasite burden as described above. Blocking studies were conducted twice with at least 5 biological replicates. Supernatants from infected macrophage cultures were added to a 96 well UV plate at 50 ml per sample in triplicates. Urea reagents A and B were mixed from quantichrom urea assay kit (Bioassay systems, Hayward, CA) and 200 ml of mixture added to each well. Included standard was used starting at 50 mg/ml and diluted 2 fold. Samples were incubated for 30 min at room temperature and plates were read at 520 nm to determine urea concentration. Femurs and tibias were obtained from 6-12 week old C57BL/6 mice. After euthanasia, the mice were sprayed with 70% ethanol and the femurs and tibias were dissected using scissors. Muscles connected to the bone were removed using scissors, and the femurs were placed into a 50 mL tube containing sterile DMEM on ice. In a tissue culture hood, the bones were washed in sterile DMEM and then both epiphyses were removed using sterile scissors and forceps. The bone marrow was flushed out with a 10 ml syringe filled with BM20 differentiation media (DMEM supplemented with 10% fetal bovine serum, 20% L929 supernatant, 5% horse serum, 100 U/ml penicillin, 100 mg/ml streptomycin, and 2 mM L-glutamine) (Life Technologies, Grand Island, NY) into a 50 mL sterile tube. The tube was vortexed gently and topped off to 50 mL with fresh BM20. 10 mL of cell suspension was plated out on 100 cm untreated dishes and incubated for 7 days at 37uC, 5% CO 2 with fresh media added at day 4. Cells were then washed, counted and plated at 10 6 cells/mL in BM10 media (DMEM supplemented with 10% fetal bovine serum 10% L929 supernatant, 5% horse serum, 100 U/ml penicillin, 100 mg/ml streptomycin, and 2 mM L-glutamine) (Life Technologies, Grand Island, NY) into a 50 mL sterile tube and allowed to rest for 3 days. Macrophages were stimulated overnight with either recombinant IL-4 (10 ng/ml), LPS (50 ng/ml) or IFN-c (100 U/ml) (all from R&D Systems, Minneapolis, MN), stAg (100 mg/ml) or cystAg (100 mg/ml) in complete DMEM. To observe the interaction of macrophages and cysts in vitro, cysts were isolated from the brains of chronically infected mice as described above. 50 cysts were added per well to 96 well plates containing 2610 5 bone marrow derived macrophages. Cysts and cells were viewed using a BD HT Pathway 855 microscope (BD Biosciences, San Jose, CA) in a climate-controlled chamber (37uC, 5% CO 2 ). Nine cysts were identified per condition and photographed every 10 minutes for 14 or 16 hours. Movies were compiled using ImageJ software (NIH, Bethesda, MD) and cyst survival time was determined. For statistical analysis of survival data, the log-rank and Gehan-Breslow Wilcoxon test was used and involved over 40 C57Bl/6 and 40 AMCase2/2 mice. Acute (0-14 days) and chronic deaths (.14days) were analyzed individually. For all other data, an unpaired, two-tailed Student's t test, or ANOVA test with a 95% confidence interval was used (Prism; GraphPad Software, Inc., La Jolla, CA). All data are represented as means 6 SEM. Figure S3 AAMØ associated with cyst lysis. Confocal fluorescence microscopy of 20 mm brain slices taken from mice at 4 weeks post infection. Imunohistochemical analysis of alternatively activated macrophage (Iba-1, red) as judged by its expression of stabilin-1 (green), adhering closely to a large round cyst. Polarized, to the site of macrophage 'attachment', bradyzoites are seen escaping in an organized fashion towards or into the AAMØ (arrows). (TIF) Figure S4 Chitinase activity is dependent on the presence of chitin and is independent of IL-4 activation. A) Bone marrow derived macrophages were analyzed for chitinase activity. Macrophages were cultured with whole cysts, cysts treated with trichoderma chitinase or media alone. Data are representative of at least 2 individual experiments with a minimum of n = 3 and are represented as mean 6 SEM. B) qRT-PCR was conducted on BMNC to measure YM-1, YM-2, RELM-a, BRP39, IL-4 and IL4Ra. Data are presented as fold increase over naïve. (TIF) Figure S5 AMCase activity associated with cyst. Confocal fluorescence microscopy of 20 mm brain slices taken from mice at 4 weeks post infection. Immunohistochemical analysis of macrophage (Iba-1, green) and AMCase (red), arrows point to AMCase polarized to the cyst wall. (TIF) Figure S6 AMCase2/2 polarization and infection studies. BMDM from WT and AMCase2/2 mice were polarized to A) M2 and B) M1 phenotype as measured by Urea and Greiss assays respectively. C57Bl/6 (WT) and AMCase2/2 mice were infected with the Me49 strain of T. gondii and sacrificed at C) 3 weeks for cyst counts and D) 5 weeks following infection. RNA was isolated from infected brains, reverse transcribed, and the resulting cDNA was analyzed for SAG1, SAG4, and MAG1 transcript levels using qRT-PCR to measure gene expression from tachyzoites, bradyzoites and cysts, respectively. Results are shown as absolute quantitation of copy number using standard curve. Data are representative of at least 3 individual experiments with a minimum of n = 3 and are represented as mean 6 SEM, ** p,0.01, *** p,0.001. (TIF) Figure S7 BALB/c macrophages lyse cysts more quickly than C57Bl/6 macrophages. BMDM from BALB/c and C57Bl/6 mice were cultured with cysts and imaged using an HT pathway microscope for 16 hours. Images were collected every 10 minutes and cyst survival time was calculated. (TIF) Video S1 3D colocalization of AMCase-secreting macrophages and parasite cysts. Three dimensional z-plane progression of 20 mm confocal image taken from mice at 4 weeks post infection. Green: tomato lectin labeling macrophages; red: AmCase; blue: DAPI labeling nuclei; white: anti-Toxoplasma. (MOV) Video S2 Cysts do not lyse in the absence of macrophages. 14 hour time-lapse movie of cysts cultured in the absence of BMDM. Images were collected every 10 minutes and movies were compiled using ImageJ. (MOV) Video S3 Rupture of T. gondii cysts in the presence of chitinase. 14 hour time-lapse movie of cysts co-cultured with WT BMDM and pretreated with 10 mg/ml trichoderma chitinase. Images were collected every 10 minutes and movies were compiled using ImageJ. Video S4 Untreated Cysts cultured with bone marrowderived macrophages. 14 hour time-lapse movie of cysts co-cultured with untreated BMDM. Images were collected every 10 minutes and movies were compiled using ImageJ. Video S5 Cysts cultured with BMDM and treated with allosamidin. 14 hour time-lapse movie of cysts co-cultured with WT BMDM and treated with 100 mM allosamidin. Images were collected every 10 minutes and movies were compiled using ImageJ. Video S6 Cysts cultured with untreated WT BMDM. 16 hour time-lapse movie of cysts co-cultured with BMDM from WT mice. Images were collected every 10 minutes and movies were compiled using ImageJ. (MOV) Video S7 Cysts cultured with untreated AMCase-null BMDM. 16 hour time-lapse movie of cysts co-cultured with BMDM from AMCase-null mice. Images were collected every 10 minutes and movies were compiled using ImageJ. Video S8 Cysts cultured with WT BMDM pretreated with LPS/IFN-c. 16 hour time-lapse movie of cysts co-cultured with BMDM from WT mice, pre-treated with LPS/IFN-c. Images were collected every 10 minutes and movies were compiled using ImageJ. Video S9 Cysts cultured with WT BMDM pretreated with IL-4. 16 hour time-lapse movie of cysts co-cultured with BMDM from WT mice, pre-treated with IL-4. Images were collected every 10 minutes and movies were compiled using ImageJ. Video S10 Me49-RFP cysts cultured with WT BMDM labeled with CellTracker green. 16 hour time-lapse movie of Me49-RFP cysts co-cultured with CellTracker green labeled BMDM from WT mice. Images were collected every 10 minutes and movies were compiled using ImageJ. (MOV) Video S11 Me49-RFP cysts cultured with AMCase-null BMDM labeled with CellTracker green. 16 hour time-lapse movie of Me49-RFP cysts co-cultured with CellTracker green labeled BMDM from AMCase-null mice. Images were collected every 10 minutes and movies were compiled using ImageJ. (MOV)

Automatic Detection and Quantification of Tree-in-Bud (TIB) Opacities from CT Scans

This study presents a novel computer-assisted detection (CAD) system for automatically detecting and precisely quantifying abnormal nodular branching opacities in chest computed tomography (CT), termed tree-in-bud (TIB) opacities by radiology literature. The developed CAD system in this study is based on 1) fast localization of candidate imaging patterns using local scale information of the images, and 2) Möbius invariant feature extraction method based on learned local shape and texture properties of TIB patterns. For fast localization of candidate imaging patterns, we use ball-scale filtering and, based on the observation of the pattern of interest, a suitable scale selection is used to retain only small size patterns. Once candidate abnormality patterns are identified, we extract proposed shape features from regions where at least one candidate pattern occupies. The comparative evaluation of the proposed method with commonly used CAD methods is presented with a dataset of 60 chest CTs (laboratory confirmed 39 viral bronchiolitis human parainfluenza CTs and 21 normal chest CTs). The quantitative results are presented as the area under the receiver operator characteristics curves and a computer score (volume affected by TIB) provided as an output of the CAD system. In addition, a visual grading scheme is applied to the patient data by three well-trained radiologists. Interobserver and observer–computer agreements are obtained by the relevant statistical methods over different lung zones. Experimental results demonstrate that the proposed CAD system can achieve high detection rates with an overall accuracy of 90.96%. Moreover, correlations of observer–observer [Formula: see text] , [Formula: see text] and observer–CAD agreements [Formula: see text] , [Formula: see text] validate the feasibility of the use of the proposed CAD system in detecting and quantifying TIB patterns.

I NFECTIOUS lung diseases, such as novel swine-origin H1N1 influenza, tuberculosis (TB), etc., are among the leading causes of disability and death all over the world [1] - [3] , [5] . Computed tomography (CT) examination of the lungs during acute respiratory tract infections has become an important part of patient care, both at diagnosis and monitoring progression or response to therapy. Although CT examination serves as a primary (imaging) diagnostic tool for assessing lung infections, visual analysis of CT images is restricted by low specificity for causal infectious organisms and a limited capacity to assess severity and predict patient outcomes [2] . Common CT findings associated with respiratory tract infections include tree-in-bud (TIB) nodularity, ground-glass opacities (GGO), random distribution of pulmonary nodules, linear interstitial/bronchovascular thickening, and consolidations [6] . Although none of these visual patterns are specific for one pathogen, the amount of lung volume exhibiting these features could provide insights into the extent or severity of infection. Among these patterns, TIB opacities, represented by thickened bronchial structures surrounded locally by clusters of 2-3 mm micronodules, are associated with inflammation of the small airways (bronchioles), such as in viral or bacterial bronchiolitis, and the increasing sizes of abnormal regions on CT can suggest the progression of disease [6] . Often considered to have a limited differential diagnosis-M TB infection, infection with nontuberculous mycobacteria, viral infection, cystic fibrosis, this pattern is recognized as a CT appearance of many different entities. Unlike the other imaging patterns such as GGO and consolidations, it is an extremely challenging task to detect and quantify the regions with TIB opacities due to interobserver variations and inconsistent visual scoring methods [2] . Therefore, an accurate method for detecting TIB is a critical in computer-assisted detection (CAD) schemes from chest CT. Although the correct diagnosis for TIB pattern is very important, it is also one of the most difficult tasks for radiologists because the contrast of lesions is often low and the disease patterns are very complex. All these limitations suggest that CAD could make a valuable contribution to the management of respiratory tract infections by assisting in the early recognition of pulmonary parenchymal lesions and providing quantitative measures of disease severity. Respiratory tract infections, caused by viruses, bacteria, fungi, and parasites, are a major component of global infectious disease mortality. TIB patterns, in particular, usually represent the disease of the small airways such as infectious-inflammatory bronchiolitis as well as bronchiolar luminal impaction with mucus, pus, cells, or fluid causing normally invisible peripheral airways to become visible on CT [7] . Fig. 1 shows typical TIB patterns in a chest CT ( Fig. 1 shows labeled TIB patterns with blue). As its name implies, this pattern resembles a budding tree in CT due to the branching opacities with adjacent centrilobular nodularity [2] . It is not specific for a single disease entity, but suggests pathology in the peripheral airways, which can be associated with air trapping or subsegmental consolidation in the surrounding alveolar airspaces. Because any organism that infects the small airways can cause a TIB pattern, pulmonary infections are its most common cause. TIB is difficult to be detected with conventional CAD systems due to high complexity of their irregular shapes, as well as strong textural similarity of micronodules and thickened airways to other normal and abnormal lung structures. Currently, no reported CAD system is capable of automatically detecting a TIB pattern, therefore, which warrants a need for the development of such a system to improve the diagnostic decision process and quantitative measurement of respiratory tract infections. In this paper, we develop a new CAD system to evaluate respiratory tract infections by automatically detecting and quantifying TIB patterns on CT images. The main contributions of this study are twofold. 1) A candidate selection method that locates possible abnormal patterns in the images. This process comes from a learning perspective such that the size, shape, and textural characteristics of TIB patterns are learned a priori. The candidate selection process removes large homogeneous regions from consideration which results in a rapid localization of candidate TIB patterns. The local regions enclosing candidate TIB patterns are then used to extract shape and texture features for automatic detection; 2) another novel aspect in this study is to extract Möbius invariant local shape features (i.e., Willmore energy-based features). Extracted local shape features are combined with statistical texture features to classify lung tissues. In addition, we also investigate the extraction and use of different local shape features as compared to the proposed shape features to facilitate local structure analysis. To the best of our knowledge, this is the first study that uses automatic detection of TIB patterns for a CAD system in infectious lung diseases. Since there is no published work on automatic detection of TIB patterns in the literature, we compare our proposed CAD system on the basis of different feature sets previously shown to be successful in detecting lung diseases in general. Early version of this study appeared in [3] , and can be accessed in [8] . This paper is organized as follows. Section II explains the methods of the proposed CAD system. We discuss our proposed and conventional feature extraction methods in Section III. Next, we present the feasibility of the proposed CAD system by evaluating the detection and quantification performances in Section IV followed by a discussion and conclusion in Section V and Section VI, respectively. The proposed CAD methodology is illustrated in Fig. 2 . First, lungs are segmented from chest CTs. Second, we use locally adaptive scale-based filtering method to detect candidate TIB patterns. Third, segmented lung is divided into local patches in which we extract Möbius invariant shape features and statistical texture features followed by support vector machine (SVM) classification. We extract features from local patches of the segmented lung only if there are candidate TIB patterns in the patches. The details of the proposed methods are described in the following. Prior to detection, segmentation is often the first step in CAD systems. In this study, fuzzy connectedness (FC) image segmentation algorithm is used to achieve successful delineations [9] . In FC framework, as illustrated in Fig. 3 , left and right lungs are "recognized" by user-defined or automatically assigned seeds, which initiate FC segmentation. In this study, one seed per lung volume (i.e., left or right) is automatically set by only considering the locations of small intensity valued voxels inside the body region (see [9] for a detailed description of the use of FC in anatomy segmentation). Fig. 3 (middle and right) shows the resulting segmentation of the chest CT given on the left. Although we use FC algorithm to segment lung regions, there are many well-established lung segmentation methods in the literature [9] - [11] , [29] , and [41] such that they could possibly be used as well to accomplish the delineation step. In that sense, we do not have any strict restriction on the choice of segmentation algorithm prior to detection system as long as it successfully segments the lung regions. The size/volume of a region occupied by a typical TIB pattern does not usually exceed a few mm 2 /mm 3 . Together with the fact that TIB pattern has a complex shape with varying intensities over discontinues branches (i.e., buds), TIB patterns have intensity characteristics with high variation toward nearby voxels (see Fig. 1 ). In other words, TIB patterns do not constitute sufficiently large homogeneous regions. Thus, TIB patterns are localized only in the vicinity of small homogeneous regions, and their boundaries have high curvatures due to the nature of its complex shape. In the next section, we use these two observations to extract novel characteristic features to detect TIB patterns. Our candidate selection method comes from a learning perspective such that we assign every internal voxel of the lung a membership value reflecting the size (i.e., scale) of the homogeneous region that the voxel belongs to. To do this, we use a locally adaptive scale-based filtering method called ball-scale (or b-scale for short) [9] , [16] , [17] . The b-scale is the simplest form of a locally adaptive scale where the scene is partitioned into several scale levels. Every voxel in each scale is assigned the size of the local structure it belongs. For instance, voxels within the large homogeneous objects have highest scale values, and the voxels nearby the boundary of objects have small-scale values. Voxels on the boundary of objects have smallest scale values. Because of these observations, we conclude that TIB Fig. 4(c) . We describe the computation of b-scale patterns and the details of the candidate selection process in the next section. There are several advantages to the local scale-based approach. For instance, boundary-and region-based representations of objects are explicitly contained in the scale-based methods. Based on continuity of homogeneous regions, geometric properties of objects (i.e., size information) can be identified, and this new representation is called scale images, i.e., b-scale, tensor-scale (t-scale), generalized-scale (g-scale) images [16] , [18] , [19] . The b-scale model has been shown to be extremely useful in object recognition [17] , image segmentation [9] , [41] , filtering [16] , inhomogeneity correction [20] , and image registration [20] , [21] . In this study, on the other hand, we show how to use b-scale encoding together with a proper scale selection method for detecting candidate abnormality patterns. The main idea in b-scale encoding is to determine the size of local structures at every voxel as the radius of the largest ball centered at the voxel within which intensities are homogeneous under a prespecified region-homogeneity criterion. Although the conventional b-scale encoding method is well established for nD images (n ≥ 2), we use 2-D b-scale encoding method in this study because low-resolution CT data do not allow continuous analysis of TIB patterns through lowresolution imaging direction. In the 2-D digital space (Z 2 , ν), a scene C = (C, f ) is represented by a pair where C is a rectangular array of voxels, ν = (ν 1 , ν 2 ) indicates the size of the voxels, and f is a function that assigns to every voxel an image intensity value. A ball B k (c) of radius k ≥ 0 and with center at The fraction of object is denoted by F O k (c) and indicates the fraction of the ball boundary occupied by a region which is sufficiently homogeneous with c. and W ψ is a homogeneity function [9] . In all experiments, we use a zero-mean unnormalized Gaussian function for W ψ . The size of the local structure is estimated using appearance information of the gray-level images, i.e., region-homogeneity criterion; b-scale scenes contain rough geometric information. A detailed description of W ψ and F O k,ν is presented in [9] . The b-scale algorithm works as follows: the ball radius k is iteratively increased starting from one, and the algorithm checks for F O k,ν (c), the fraction of the object containing c that is contained in the ball. When this fraction falls below a predefined threshold, it is considered that the ball contains an object region different from that to which c belongs [16] . This process is repeated for every voxel within the scene. Voxels are assigned their b-scale values discreetly from 1 to r max . 1 In principle, b-scale partitions the scene into several levels based on the size of local structures from 1 to r max . Computing b-scale values for every voxel leads b-scale scenes as shown in Fig. 4 (a). Note also that locally adaptive scale in regions with fine details or in the vicinity of boundaries is small, while it is large in the interior of large homogeneous objects. Developing a successful CAD system for infectious lung diseases requires acquisition of representative features characterizing shape and texture of TIB patterns efficiently. Since TIB is a complex shape pattern consisting of curvilinear structures with nodular structures nearby, we propose to use local shape features (derived from geometry of the local structures) combined with gray-level statistics derived from a given local patch (i.e., local window with a predefined size). The shape operator is the second-order invariant (or curvature) which determines the original surface. Since it is usually more convenient to work with scalar quantities rather than vectorial shape quantities, symmetric functions of local Hessian matrices are usually used to extract geometric meaning of the surface/shape of interest. Therefore, curvatures play an important role in the representation and recognition of intrinsic shapes. However, similarity of curvature values may not necessarily be equivalent to intrinsic shape similarities, which causes a degradation in recognition and matching performance. To overcome this difficulty, we propose to use Willmore energy functional [22] and several different affine invariant shape features parametrically related to the Willmore energy functional. While local shape features characterize the curvilinear and small nodular structures (via Willmore energy), gray-level features characterize background and foreground intensity variation with objects' pose and size for a given local window. Moreover, for comparison purpose, we use different feature sets previously shown to be successful in detecting lung diseases in general. Fig. 5 enlists all the features that we extracted for the proposed CAD system and for the experimental comparison. Details of extracted features are defined in the following. The Willmore energy of surfaces plays an important role in digital geometry, elastic membranes, and image processing [23] . It is closely related to Canham-Helfrich model [24] , where a surface energy is defined as where α, β, and γ are some constants, H is the mean curvature vector on Σ (area space), K is the Gaussian curvature on ∂Σ (boundary space), and dA is the induced area metrics on Σ. This model is curvature driven, invariant under the group of Möbius transformations (in particular, under rigid motions and scaling of the surface) and shown to be very useful in energy minimization problems [25] . Invariance of the energy under rigid motions leads to conservation of linear and angular momenta, and invariance under scaling plays a role in setting the size of complex parts of the intrinsic shapes (i.e., corners, wrinkles, folds, etc.). In other words, the position, gray-level characteristics, size, and orientation of the pattern of interest have minimal effect on the extracted features as long as the suitable patch is reserved for the analysis. In order to have simpler and more intuitive representation of the given model, we simply set α = 0 and β = γ = 1, and the equation turns into Willmore energy functional where ds is the length metric on ∂Σ. The resultant energy of a surface can be regarded as a function H and K, and captures the deviation of a surface from local sphericity [22] such that a sphere has zero Willmore energy. Note also that the Willmore energy is always nonnegative. Since a homogeneity region that a typical TIB pattern appears is small in size, total curvature (or energy) of that region is high and can be used as a discriminative feature. The main motivation in describing intrinsic shapes by Willmore energy is due to its ability to encode surface (i.e., image area in 2-D) with Möbius invariant features (translation, contrast, rotation, and inversion invariant). In addition to Willmore energy features that we adapt from Canham-Helfrich surface model, we have included seven different local shape features, which are parametrically related to Willmore energy formulation, into the proposed CAD system due to their some invariant properties and discriminative powers. Assume κ 1 and κ 2 indicate eigenvalues of the local Hessian matrix H e for any given local patch L , the following shape features are extracted: 1) shape index (SI), 2) Gaussian curvature, 3) mean curvature, 4) elongation, 5) distortion, 6) shear, 7) compactness. The SI is a statistical measurement and used to define intrinsic shape of the localized structure within the image [26] , [27] . SI values are encoded as a continuous range of values between −1 and 1, with zero SI indicates saddle-like local structures, +1 and −1 SI values indicate umbilical minima and maxima (i.e., cap and cup, respectively), and midpoints of the two half-intervals (+0.5 and −0.5) indicate concave and convex parabolic or line-like structures (i.e., rut and ridge, respectively). SI can simply be computed through principal curvatures (κ 1 , κ 2 ) as follows: where κ 1 ≥ κ 2 . As suggested in [26] , we obtain principal curvatures from the eigenvalues of the local Hessian matrix (H e ) as where L xx , L xy = L y x , and L y y are second-order derivatives of local image patch L , and eig() denotes eigenvalue decomposition. We choose to use SI because of its invariance property with respect to rotation, absolute gray value, and translation. 2) Gaussian Curvature: Gaussian curvature (K) is an intrinsic measure and simply the product of the principal curvatures as K = κ 1 κ 2 for a given point on a surface, equivalent to the determinant of local Hessian matrix H e . Note that K is unchanged even by bending the surface without stretching it, meaning that the Gaussian curvature is independent of the choice of unit normal and it gives three types of classified local shapes: elliptic shape (K > 0), hyperbolic shape (K < 0), parabolic shape (K = 0) with one of the κ is equal to zero, planar shape (K = 0) with both κ are equal to zero. Gaussian curvature is translation and rotation invariant, but not scale invariant. 3) Mean Curvature: Mean curvature (H) is an extrinsic measure and it describes the curvature as H = (κ 1 + κ 2 )/2. Unlike K, H is defined in the distributional sense. Note that mean curvature measure is the trace of local Hessian matrix H e . Mean curvature can be thought as a negative gradient (as a Laplacian) of the area functional due to its nice variational interpretation over the surface. This does not only give insights into the size of the local shape but also into the total symmetrical deviation from the sphere. Mean curvature is translation and rotation invariant, but not scale invariant. 4) Elongation: Shape elongation is one of the basic shape descriptors and it indicates flatness measure of the local shape [28] . In this paper, we used the ratio of principal curvatures to measure elongation as κ 2 /κ 1 with κ 2 ≤ κ 1 . Elongation measure is invariant with respect to a similarity transformation, and therefore, it is a robust feature that helps to identify curvilinear shapes. Elongation varies from −1 to +1, from hyperbolic to elliptic points. Distortion is an algebraic quantity defined as the difference of eigenvalues (i.e., |κ 1 − κ 2 |) of the local Hessian matrix H e . Distortion is a valuable image analysis property revealed by magnitude difference of principal curvatures. Distortion measure captures the deviation of principal curvatures, thus nonplanarity of a region. Together with Gaussian or mean curvature, distortion measure brings further information into encoding of local shape. Distortion measure is translation and rotation invariant, but not scale invariant. 6) Shear: The shear is another algebraic distortion quantity defined as proportional to the normalized distortion: (κ 1 − κ 2 ) 2 /4. The physical information contained in the shear is basically the same as that of the distortion; it is related to distortion with powers of the difference of principal curvatures. Different than distortion, shear descriptor captures higher degree of nonplanarity of a region due to having more robustness against noise. 7) Compactness: Compactness feature measures the similarity between shape of interest and a perfect ellipse, and is defined as 1/(4π √ κ 1 κ 2 ). Note that this ratio is a dimensionless ratio between the area of the shape (1 for a normalized shape) and the area of the best ellipse fitting the shape. Note also that the compactness measure is invariant to affine transformations and parametrically related to Gaussian curvature. Given a single-axial CT slice of left lung, Fig. 6 (b) indicates a thresholded (i.e., selected candidate patterns) b-scale scene encoded from the corresponding gray-level CT slice shown in Fig. 6(a) . Furthermore, Fig. 6 (c) and (d) shows mean and Gaussian curvature maps from which all the other local shape features are extracted, respectively. In addition, Fig. 6 (e) and (f) shows Willmore energy maps using both mean and Gaussian curvature maps as formulated in (4) and shown in Fig. 6(c) and (d) . Based on the observation in training step where we analyzed the appearance and shape of TIB patterns, TIB patterns most likely occur in the regions inside the lung with high variability of intensity values over a small number of voxels and with certain size (i.e., a few millimeter in length). These observations (size and high intensity variation) facilitate one practically useful fact in the algorithm that, in the feature extraction process, we only extract features if and only if at least "one" small b-scale pattern exists in the local regions (i.e., blue local regions in Fig. 4) . We also explore the use of alternative local shape features as a comparison to Willmore energy-based features. Based on the observations of spatial properties of the selected candidate patterns, it becomes apparent that instead of using conventional high-dimensional feature extraction methods such as Gabor wavelets, steerable wavelets, etc., one may extract much fewer and more reliable statistical features to discriminate the pattern of interest. Motivated from the fact that TIB patterns consist of numerous small (or micro-) nodules nearby the main curvilinear structure and those small structures have varying opacities, the location and distribution of those small structures can be obtained by simple thresholding method which has been popular in estimation for more than two decades [30] . However, since the opacities are varying through different nodular structures, it is challenging to find an optimum threshold value. Therefore, instead of using one single threshold level, we empirically choose n = 10 different threshold levels (λ j ) to obtain local statistics of those structures in a hierarchical way, where λ j = 10j, 1 ≤ j ≤ 10 [31] . This process is named LGS because we extract different statistical measurements in gradient of the images. Note also that we confine ourselves into the local patches where at least one b-scale pattern occupies. To obtain shape statistics over local patches, we use gradient fields because boundary information can be used much more effectively in that sense. Fig. 7 shows an example thresholding process over a candidate TIB pattern centered at c (only for four levels are shown for demonstration purpose). After different threshold levels are applied over the local regions of b-scale images, resultant thresholded local patches are used to extract the following features: mean SI values of the local patch for each thresholding level (one feature), and the number of bscale patterns left after thresholding process (one feature). Since we use ten different thresholding levels, we extract 20 features totally. Moreover, for a local region centered at a voxel c of a candidate TIB pattern, we extract one global feature as an SI value of the voxel c, three features as the maximum, minimum, and mean SI values over the local region prior to thresholding. Therefore, a total of 24 features (LGS+SI) are extracted from a typical local patch to be used in CAD system [4] . Although n and λ j are chosen empirically based on the observations of shape and textural characteristics of normal and TIB patterns during the training step, one may propose to use cross validation, control of the global and local false discovery rate, and uncertainty principles to decide those parameters near-optimally [30] . Steerable Features: It has been well documented in the literature that decomposition of images by using basis functions localized in spatial position, orientation, and scale (e.g., wavelets) are extremely useful in object recognition and detection [32] , [33] . Since steerable filters are rotation and translation invariant, they accurately represent the underlying image structure [34] . In this study, we use steerable derivative of Gaussian filters to decompose local regions around each candidate pattern into several oriented basis. These basis are used as features in voxelwise classification for TIB identification. We extract steerable features (i.e., directional derivatives) from one scale and six different orientations. Gray-Level Co-Occurrence Matrix (GLCM) Features: Spatial statistics based on GLCM [35] are shown to be very useful in discriminating and quantifying patterns pertaining to lung diseases. As texture can give a lot of insights into the classification and characterization problem of poorly defined lesions, regions, and objects, we combine our proposed shape-based invariants with Haralick's popular GLCM-based features [35] . We extract 18 features from each local patch including autocorrelation, contrast, entropy, variance, dissimilarity, homogeneity, cluster shade, energy, maximum probability, sum of squares of variance, sum of averages, sum of variance, sum of entropy, difference of entropy, difference of variance, normalized inverse difference moment, cluster prominence, and mutual information. Readers are encouraged to refer to [35] for further details on these well-established features in machine learning, and [12] - [15] for particular CAD systems in identification of lung abnormalities from CT scans in general. Laboratory confirmed (with pathology identification tests) 39 CTs of human parainfluenza (HPIV) infection and 21 normal lung CTs were collected for the experiments. All patients were imaged at our institution using a 64-detector row Philips Brilliance or a 320-detector row Toshiba Aquilion CT scanner. The noncontrasted chest CT studies were performed at end inspiration with 1.0 or 2.0 collimation obtained at 10-or 20-mm intervals from the base of the neck to upper abdomen with a tube voltage of 120 kV and a current of 200-320 mA depending on the subject's weight. Imaging data were constructed to 512 × 512 matrices with slice thickness of 5 mm. The in-plane resolution was affected by patients' size and varied from 0.62 to 0.82 mm. All 60 CT scans (both HPIV and normal) were collected from different subjects (no multiple scans from subjects). A well-trained radiologist [with more than nine years experience (DMJ)] carefully examined the complete scan (i.e., 60 CTs) and labeled the lung regions as normal and abnormal (with TIB patterns) (see Fig. 1 ). As many regions as possible showing abnormal lung tissues from 39 HPIV patients were labeled (see Table I for details of the number of regions used in the experiments). Those 39 patients do not include only TIB opacities, but also GGO, nodules, consolidations, and linear thickening such that only TIB regions are labeled in training step. Note also that the control group consisting of 21 subjects with no observed lung abnormalities was constructed and lung tissues pertaining to this group were labeled carefully. In the training step, we also explored how the number of b-scale patterns change for normal and diseased subjects. Our observations from detail analysis in candidate selection part showed that only 21-40% of the segmented lung volumes were chosen as candidate TIB patterns. This interval was subject to change based on the severity of the diseases. For patients without having infections (i.e., control group), for instance, the percentage of the candidate regions was smaller than the patients with infections; therefore, an increase in the amount of small-sized b-scale patterns is observed. In any case, local scale could be used as a quantitative measure validating the sensitivity and specificity of the classification rates as we describe it in Section IV-E. Occurrences of TIB abnormality and normality of subjects were noted for each CT scan. To analyze existence and severity of abnormality as well as normality of subjects, a visual grading system was adapted from studies examining CT findings in other infections [36] - [38] . Each lung was divided into three zones (for a bilateral total of six) as shown in Fig. 8 . Zone 1 included the apex to the carina. Zone 2 extended from the tracheal carina to the left atrium's junction with inferior pulmonary veins. Zone 3 included the remainder of the lungs below the level of the inferior pulmonary veins atrial junction. A severity score (0 to 5 such that 0 indicates no abnormality) was assigned to each zone based on the percentage of the zone occupied as listed in Fig. 8 (second row). A total score was also extracted by considering all zones during visual grading. Consensus visual scores 2 from participating radiologists [one with more than nine years of experience (DMJ) and one with more than one year of experience (AW)] on a scale of 0-5 over lungs were recorded and compared with computer scores (of the proposed CAD system). Following the same visual scoring scheme, another participating radiologist [with more than seven years of experience (OA)], who was blinded to the consensus visual scores previously obtained, was involved in the visual grading process to provide information on interobserver variability. To measure and evaluate the detection capabilities of a CAD system quantitatively, the area under the receiver operator characteristic (ROC) curves is often used [39] . After the proposed CAD system was tested via twofold cross validations with labeled dataset, we presented ROC curves of the system performances. Table I summarizes the performance of the proposed CAD system as compared to other feature sets. The performances are reported as the areas under the ROC curves (A z ). Note that proposed shape features (i.e., Willmore energy and parametrically related local shape features) alone are superior to other methods even though the dimension of the proposed shape feature is only 8. The best performance is obtained when we combine the proposed shape and GLCM features. This is to be expected because spatial statistics are incorporated into the shape features such that texture and shape features are often complementary to each other. On the other hand, compared to the proposed shape features, the LGS and SI features have lower detection rates because they are not affine (and Möbius) invariant and eventually having difficulty in appreciating the large amount of details of TIB patterns. Another reason is that there is no optimal choice of thresholding process and this may yield less remarkable statistical measurements over local patches. However, the LGS and SI features alone perform better than the high-dimensional conventional features similar to the proposed shape features. This result itself suggests the use of local shape features and their adapted extensions in detection of TIB patterns. In what follows, we selected the best window size for each feature set and plotted their ROC curves all in Fig. 9 . Superiority of the proposed shape features is clear in all cases. To have a valid comparison, we repeated candidate selection step for all the methods because we observed that the CAD performances of compared conventional feature sets had much lower accuracies if the candidate selection part was not applied (i.e., proposed method's accuracy was decreased to A z = 0.6803, while the best result of all compared methods were decreased to A z < 0.5281). To show whether the proposed method was significantly different than the other methods, we compared the performances through paired t-tests. p-values of the tests indicate that none of the feature set are significantly correlated with the proposed CAD features such that highest and smallest pvalues are reported as 0.0195 (p < 0.05) and 0.0053 (p < 0.01), respectively. Visual scoring by radiologists still lies at the heart of diagnostic decisions, and often used as a validation tool. In this section, we explore the correlation between computer score (i.e., CAD score) and visual scores by participating radiologists. Furthermore, we investigate the effectiveness of the proposed method's ability to roughly discriminate normal and diseased patients by only considering the size of the structures pertaining to lung anatomy. Based on the visual grading scheme explained in Section IV-C, we compared the consensus reading of two expert observers (AW and DJM) to another expert observer (OA), who was blinded to the consensus scores. We used Pearson product-moment correlation coefficients to determine interobserver agreement over each zone, left, right, and all lung volumes. The reported correlation ratios are shown in Fig. 10(a) . Note that interobserver agreement correlation values for all TIB measurements were high for all zones and the lung. The lowest agreement seen on the zone 1 may be because subtle abnormalities in this zone may have been given greater visual assessment variance among the observers. Nevertheless, an overall correlation coefficient of R 2 = 0.8848 (p < 0.01) indicates an excellent agreement on the existence of TIB patterns. We further analyzed the variability of change of scores of expert radiologists for each subject. For this, we constructed Bland-Altman plot [40] where the limits of observer agreements were indicated by bias ± 1.96 std (bias: average difference, std: standard deviation). In Bland-Altman plot, the difference of the performances was plotted against the average of the performances as shown in Fig. 10(b) . It was noted from this figure that the largest disagreement of the scoring between observers To obtain an overall computer score from the proposed CAD system, on the other hand, TIB regions detected by the CAD system were first labeled automatically during the detection process. Then, a computer score was calculated by averaging the volume occupied by the labeled TIB regions over the whole lung volume. Calculated computer score was then normalized to fit the visual grading scheme explained in Section IV-C. Linear regression model was fitted to all subjects' scores both from computer and the consensus scores of the participating radiologists (DMJ and AW) and Pearson product-moment correlation coefficient was computed for this model. A scatter plot of the linear regression model and the computer-observer agreement correlation is shown in Fig. 11 . It is clear from this plot that visual and quantitative assessments correlate well as indicated by the Pearson product-moment correlation of R 2 = 0.824 (p < 0.01). Finally, we illustrate an example of TIB and non-TIB region classification by expert annotation and computer quantification by our proposed method in Fig. 12(a) and (b) , respectively. Scale-based analysis: In addition to visual scoring scheme, we also show the effectiveness of the proposed scale-based method on quantification of the disease extent and identification. Scale-based analysis of the regions occupied by TIB patterns is illustrated in Fig. 13 . A CT slice of a patient with HPIV shows fewer large homogeneous regions (green) with respect to a normal control. It also shows a greater number of small homogeneous regions (yellow and red). (from 1 to 10), we recorded the average number of b-scale patterns. As readily seen from both curves, the existence of TIB patterns was indicated through the small number of highly homogeneous regions (i.e., small number of large b-scale patterns) and large number of less homogeneous regions (i.e., large number of small b-scale patterns). This figure validated the qualitative Fig. 13 . The difference between two curves was at statistically significant level (p < 0.01). All programs used in this study were developed using gcc 4.5 (Copyrigth (C) 2010 Free Software Foundation) on a Linux platform (Pardus), and all statistical computations were processed in R (Version 2.12.2) and MATLAB (Copyright (C) 2010 Mathworks). All the programs were executed on an Intel (R) Core(TM) i7 CPU 930 at 2.80 GHz with 12 GB RAM workstation. While segmentation of lung regions from CT scans takes only about 10 s, the b-scale encoding algorithm takes a couple of minutes (average 2 min, at most 5 min). The time required to compute b-scale scenes changes from patient to patient due to different number of slices in CT scans. Details of the computational cost analysis for segmentation of lungs, and feature extractions for particular algorithms are enlisted in Table II. A further feature selection method such as a principal component analysis might be used to reduce the dimension of steerable features that we used only for comparison purposes. Note that the proposed features are having a small number of dimension per local patch; there is not necessarily an additional feature selection method needed; hence, it is outside the scope of this paper. Briefly, the whole dataset was randomly divided into training and test sets of 30 CT scans (20 HPIV-10 Normal versus 19 HPIV-11 Normal). Parameters of the SVM classifier were learned based on the CT scans pertaining to training set. SVM regression was based on pixel-wise classification [42] . Followed by feature extraction step, the trained SVM classifier was applied to the test set. Note also that we have used twofold cross-validation technique for training and testing; therefore, we changed the role of training and testing dataset in the second fold. We also noticed that there was no significant changes in training and test performances of SVM classifications if twofold cross validation was changed into n-fold cross-validation system with n > 2. In addition, we have used Efron's bootstrap [43] method (i.e., repeating the experiments 100 times based on the actual data) to assess the variability of the estimated classifications derived from SVM regressions, and provide confidence intervals for ROC curves. We used radial basis functions as kernel of SVM, and set to epsilon parameter of SVM as 0.1 [42] . Resulting SVM values of pixels are ranging from 0 to 1. This value indicates the likelihood of a local patch belonging to a certain class (TIB or non-TIB); low ratings indicate a non-TIB region, and high ratings indicate a TIB region. Soon after the SVM values were computed for the entire lung, we changed the cutoff values of SVM (0.5 as default) several times to obtain ROC curves. In this paper, we studied a very particular, yet important, pattern of lung abnormality observed in chest CTs. Our proposed detection system is tuned to detect TIB regions from non-TIB regions; therefore, a multiclass classifier (with specifically tuned detection filters for each abnormality class) might be needed as an extension of this study to detect as much abnormality as possible in a whole system. Although such a system will bring its unique challenges into the CAD platform, it would be a valuable second opinion tool for radiologists. As a further step, we are currently investigating combining different imaging patterns pertaining to lung abnormalities as well as clinical laboratory information into our CAD system. One question arises as to the use of high-resolution CT (HRCT) scans instead of conventional CT scans in detecting TIB patterns, as well as the effect of using HRCT scans in this process. Although HRCT scans appreciate detection of small nodular patterns, they have more noise and lungs might not be fully covered due to large gaps between slices (i.e., 10-30 mm). Furthermore, at our institution and in many other institutions, the protocol for acute pulmonary infection is 5 mm contiguous slice images of the chest without IV contrast, for which we adapted our CAD method. Nevertheless, the method we present is not data dependent and can be used for HRCT scans as well. Considering 2-D computation of b-scale scenes, one may doubt if the algorithm can be extended into 3-D. Based on our observations on appearance and location of TIB patterns over the lung regions and experiences on feature extraction in 3-D, as we stated previously, TIB patterns rarely extend in depth direction for more than a few slices due to constraints of lowresolution imaging direction. Therefore, there is no significant classification rate changes in 3-D; however, there is an increase in computational cost. Nevertheless, 3-D b-scale encoding and feature extraction for a similar pattern detection problem or the same problem with high-resolution images (with thinner slice thickness compared to low-resolution CT images) can readily be combined and used with similar accuracies reported in this study. Number of large and small b-scale patterns might perhaps be used to identify other type of abnormality patterns such as GGO and consolidations where we expect to have more large b-scale patterns than small b-scale patterns. Therefore, as an extension of this study, we will tune our proposed methodology with different types of abnormalities to generalize the CAD systems for infectious lung diseases in general. Our proposed method is capable of detecting and quantifying TIB patterns very accurately as validated by the statistical tests compared to the expert annotations (i.e., ground truth). Therefore, both in detection and quantification steps, the proposed CAD system will highly possibly be helpful for clinicians as a second opinion tool in routine clinical examinations. In this study, we have proposed b-scale-based binary classification approach for automatic TIB pattern detection and quantification from chest CTs. The proposed system integrates 1) fast localization of candidate TIB patterns through b-scale filtering and scale selection, and 2) combined shape and textural features to identify TIB patterns. Note that texture-based recognition methods offer a complementary view to shape-based methods; therefore, the integration of spatial information and the proposed shape features achieves high detection rates. Moreover, our proposed local shape features illustrate the usefulness of the invariant properties, Willmore energy in particular, to analyze TIB patterns in chest CT. We have also compared computer scoring of the proposed CAD system with subjective visual grading. A high correlation between objective (CAD) and subjective (visual grading) scores is obtained, which implies highly satisfactory accuracy of the proposed CAD system.

Mannose-binding lectin deficiency and acute exacerbations of chronic obstructive pulmonary disease

BACKGROUND: Mannose-binding lectin is a collectin involved in host defense against infection. Whether mannose-binding lectin deficiency is associated with acute exacerbations of chronic obstructive pulmonary disease is debated. METHODS: Participants in a study designed to determine if azithromycin taken daily for one year decreased acute exacerbations had serum mannose-binding lectin concentrations measured at the time of enrollment. RESULTS: Samples were obtained from 1037 subjects (91%) in the trial. The prevalence of mannose-binding lectin deficiency ranged from 0.5% to 52.2%, depending on how deficiency was defined. No differences in the prevalence of deficiency were observed with respect to any demographic variable assessed, and no differences were observed in time to first exacerbation, rate of exacerbations, or percentage of subjects requiring hospitalization for exacerbations in those with deficiency versus those without, regardless of how deficiency was defined. CONCLUSION: In a large sample of subjects with chronic obstructive pulmonary disease selected for having an increased risk of experiencing an acute exacerbation of chronic obstructive pulmonary disease, only 1.9% had mannose-binding lectin concentrations below the normal range and we found no association between mannose-binding lectin concentrations and time to first acute exacerbation or frequency of acute exacerbations during one year of prospective follow-up.

Mannose-binding lectin (MBL) is a pattern-recognition collectin that is related to surfactant proteins A and D and has two roles in host defense. MBL activates complement via serine proteases, particularly MBL-associated serine protease-2, with which it circulates. 1,2 MBL is also involved in opsonophagocytosis, binding several types of pathogens to phagocytes via its carbohydrate recognition domain, triggering release of a number of proinflammatory cytokines [3] [4] [5] and facilitating clearance of apoptotic cells. 6 MBL is secreted primarily by the liver and circulates in the serum. Low MBL concentrations occur as a result of one of three single nucleotide polymorphisms on exon 1, but the most common cause of deficiency in Caucasians is the LXP haplotype resulting from polymorphisms of the promoter region of the gene, presenting either as a homozygous mutation or in combination with other haplotypes. [7] [8] [9] Several observations suggest that MBL deficiency may compromise host defense in the lungs. MBL binds to carbohydrates on the surface of a number of respiratory pathogens that are associated with acute exacerbations of chronic obstructive pulmonary disease (COPD), including Haemophilus influenzae, 10, 11 Mycoplasma pneumoniae, 12 and influenza. 13 Patients with cystic fibrosis and low MBL concentrations have decreased lung function, an increased prevalence of Burkholderia cepacia infection, and reduced survival relative to patients with cystic fibrosis and normal MBL concentrations. The difference in survival is particularly notable in those with chronic Pseudomonas aeruginosa infection and those with homozygous mutations in the MBL-2 gene. [14] [15] [16] [17] Low MBL concentrations have been associated with an increased risk of respiratory infections in immunocompetent subjects, [18] [19] [20] [21] with an increased frequency of respiratory syncytial virus infections, 22 and with worse outcomes in patients with community-acquired pneumonia and Streptococcus pneumoniae infections. [23] [24] [25] [26] MBL concentrations in bronchoalveolar lavage fluid obtained from a small number of current and former smokers with COPD are lower than those in fluid from healthy controls and tend to be higher in former smokers than in current smokers, 27, 28 but no association has been found between genotypes producing low MBL concentrations and the prevalence of COPD. 29 However, studies assessing the association of low MBL concentrations with acute exacerbations of COPD (AECOPDs) report conflicting results. [30] [31] [32] Accordingly, we prospectively measured MBL concentrations in a large cohort of subjects with COPD who were at increased risk of experiencing acute exacerbations and followed them for one year while tracking the number of acute exacerbations that occurred. Our hypothesis was that subjects with COPD who had an increased risk of experiencing an AECOPD would have more frequent acute exacerbations during one year of follow-up if they were deficient in MBL than if they were not. Subjects were men and women enrolled in a multicenter randomized trial designed to determine if azithromycin, taken daily for one year, decreased the frequency of AECOPDs. 33 Eligibility criteria included age $ 40 years, a clinical diagnosis of COPD, and an increased risk of experiencing an AECOPD based on criteria defined by Niewoehner et al. 34 Patients had to be free of AECOPDs for a minimum of 4 weeks prior to enrollment. Exclusion criteria included a diagnosis of asthma or bronchiectasis, among others. 33 Acute exacerbations were defined as "a complex of respiratory symptoms (increased or new onset) of more than one of the following: cough, sputum, wheezing, dyspnea, or chest tightness with a duration of at least three days requiring treatment with antibiotics or systemic steroids". 34 Serum was collected at the time subjects were enrolled in the study when they had not experienced an AECOPD for a minimum of 4 weeks and was stored at −80°C until MBL concentrations were assayed by enzyme-linked immunosorbent assay (R & D Systems, Minneapolis, MN). Samples were diluted 1/500 for this assay and assayed in duplicate wells. At this dilution, the range of the standard curve corresponds to concentrations ranging from 78 ng/mL to 5000 ng/mL. When concentrations were extrapolated above 5450 (n = 7) or below 60 (n = 6), the samples were reassayed at either a 1/2500 or a 1/20 dilution and these new values used as MBL serum concentrations. In one case, a sample was still less than the detection limit at 1/20 dilution and this sample was assigned the concentration of less than 3 pg/mL. The MBL concentration that def ines MBL def iciency is debated. Some define deficiency as a serum concentration , 500 ng/mL. 24, 31, 35, 36 Others def ine it as #100 ng/mL, and still others define severe deficiency as #50 ng/mL and partial deficiency as .50 ng/mL but ,1000 ng/mL. 9, 10, 21, 25, 31, [37] [38] [39] [40] The normal value reported by the manufacturer of the assay is 1135 ng/mL with a range of 103-3308 ng/mL (R&D Systems). Because of these uncertainties, we defined MBL deficiency in four ways, ie, #50 ng/mL, #100 ng/mL, #500 ng/mL, and .50 but #1000 ng/mL. Azithromycin increases expression of the mannose receptor, and uptake of apoptotic cells by human alveolar macrophages, and decreases recovery of apoptotic bronchial epithelial cells. 27 Accordingly, MBL concentrations from patients randomized to receive azithromycin or placebo were analyzed both separately and together. A Cox proportional-hazards model analysis was used with time-to-first-exacerbation as the outcome variable and MBL group (ie, below versus above specified limits) as the primary variable of interest. Bootstrap methods were used to compute confidence intervals for median survival times. Rates of AECOPDs were determined by dividing the number of AECOPDs by person-years of follow-up and were analyzed as a function of MBL concentration using a negative binomial model. MBL concentrations are presented as medians and interquartile ranges. P , 0.05 was considered to be statistically significant. The study (ClinicalTrials.gov number NCT00325897) was approved by the institutional review boards at all participating centers. submit your manuscript | www.dovepress.com MBL assays were performed at baseline in 1037 (91%) of the 1142 subjects enrolled in the azithromycin trial. Of these, 909 had experienced one or more AECOPDs in the year prior to enrollment. Duplicate measurements of MBL had coefficients of variations ,15% in all cases and ,5% in most. The median MBL concentration for all patients was 918 ng/mL (interquartile range 508-1683 ng/mL, inclusive range 0-8194 ng/mL). The median MBL concentration in subjects randomized to receive azithromycin or placebo was 1008 ng/mL (95% confidence interval [CI] 909-1082) and 850 ng/mL (95% CI 776-929), respectively (P = 0.017). Patient demographics and clinical characteristics, stratified by MBL concentration, are summarized in Table 1 . The prevalence of MBL deficiency was 0.5%, 1.9%, 24.2%, or 52.3%, when deficiency was defined as #50, #100, #500, or .50 and #1000 ng/mL, respectively (Table 1) . Regardless of the concentration of MBL used to define MBL deficiency, no difference was observed with respect to the prevalence of MBL deficiency by gender, race, or age (with the exception that a greater fraction of women had MBL deficiency defined as .50 ng/mL and #1000 ng/mL than was seen with the other definitions), smoking status, chronic bronchitis, or steroid use, and there was no suggestion that airflow limitation was worse or that GOLD (Global Initiative for Chronic Obstructive Lung Disease) stage was higher in subjects with MBL deficiency ( Table 1 ). The same findings were also observed for the subgroups of subjects receiving azithromycin or placebo (data not shown). When analyzing time-to-first AECOPD using a model that included treatment group (ie, azithromycin versus placebo) and log-transformed MBL concentration stratified by clinic, treatment group was significant (P , 0.0001) and log-transformed MBL concentration was not (P = 0.629). The hazard ratio for a one-unit increase in log-transformed MBL concentration was 1.02 (95% CI 0.94-1.12). For the rate per person-year of AECOPDs, a negative-binomial analysis of a model that included treatment group and log-transformed MBL concentration found that treatment group was significant (P = 0.010) but log-transformed MBL concentration was not (P = 0.470). The coefficient for log-transformed MBL concentration in this model was 0.031 (95% CI, −0.053 to +0.115). The median time to first AECOPD and the rate of AECOPD per patient-year are shown in Figures 1-3 Table 2 relative to the various definitions of MBL deficiency (only five patients had MBL concentrations # 50 ng/mL, precluding life table analyses for patients in this subgroup). Regardless of the MBL concentration used to define deficiency, no association between the time to first AECOPD, or the rate of AECOPDs and MBL concentration was observed in the population as a whole, or in either treatment subgroup. Nonsignificant trends favoring a longer time to first AECOPD were seen in the subgroup of subjects with MBL concentrations , 100 ng/ mL compared with those with concentrations $ 100 ng/mL ( Figure 1A and C). No difference was observed in MBL concentrations in subjects experiencing no, one, two, or at least three AECOPDs during the course of the study in the population as a whole, or in either treatment group (Table 3) . Of the 1037 subjects in the study, 220 (21%) required hospitalization for AECOPDs. The median MBL concentration in these 220 subjects was 1055 (95% CI 861-1213) ng/mL [1091 (958-1450) for those on azithromycin and 891 (816-1276) for those on placebo]. The median MBL concentration in the 817 subjects who were not hospitalized was 904 (833-980, P = 0.17, Table 4 ). The important findings of this study are that, in this large sample of subjects with COPD selected for having an increased risk of experiencing an AECOPD within one year, only 1.9% had MBL concentrations below the normal range reported by the manufacturer of the assay, and regardless of the MBL concentration used to define MBL deficiency, we found no association between deficiency and time to first AECOPD, rate of AECOPDs, or need for subjects to be hospitalized for AECOPDs. Two studies concluded that MBL deficiency was associated with an increased incidence of AECOPDs and one concluded that it was not. The age, spirometry, and smoking histories of the subjects in these three studies were similar to those we evaluated, with the exception that none of the three studies selected patients who were at increased risk of experiencing acute exacerbations as we did. Yang et al 30 found that 24 of 82 (29%) subjects with COPD had MBL-deficient genotypes. These subjects had lower MBL concentrations than those with the wild-type genotype (107 ng/mL, IQR 30-246, range 21-7675 versus 920 ng/mL, IQR 398-1355, range 21-2256, P , 0.001). MBL concentrations were not presented in a fashion that allowed determination of the prevalence of MBL deficiency based on the definitions used in the literature. Forty of the 82 patients (49%) had one or more admissions for AECOPDs during a two-year follow-up period and 18 of these (45%) had MBL-deficient genotypes. Forty-two patients had no AECOPDs and only six (14%) of these had MBL-deficient genotypes (odds ratio 4.9 95% CI 1.7-14.4, P = 0.0037). We did not determine MBL genotypes but genotypes do not COPD (P = 0.23 and P = 0.10, respectively). No association between MBL concentration and a history of AECOPDs was observed, but no information was provided with respect to how AECOPDs were defined. Eagan et al 31 also noted that subjects with GOLD stage 3 disease had a higher prevalence of MBL deficiency (defined as #100 ng/mL). We found no demographic or COPD severity indicators that were more or less common in subjects with MBL deficiency, regardless of how deficiency was defined ( Table 1) . The strengths of our study include the large sample size, the multicenter design, and the prospective ascertainment of AECOPDs using an event-based (ie, health care utilization) definition. Our study population was enriched by enrolling subjects whom we anticipated would be at increased submit your manuscript | www.dovepress.com risk of experiencing an AECOPD within the one-year follow-up period, based on previous predictors identified by Niewoehner et al. 34 This should have increased the likelihood of finding a higher prevalence of MBL deficiency, regardless of how deficiency was defined, compared with the prevalence in healthy controls. Our study has a number of limitations. First, we only measured MBL concentrations on one occasion. Several groups have demonstrated that MBL is an acute phase reactant. [41] [42] [43] However, in clinically stable patients, MBL concentrations are constant over time, 42, 44 and our patients had to be free of AECOPDs for at least 4 weeks before meeting inclusion criteria. In addition, even during acute phase responses, MBL concentrations in patients with low concentrations only increase by 1.5-4.3-fold and do not reach normal values. 45, 46 Accordingly, we suggest that there is a low likelihood of our data being confounded by spuriously elevated concentrations of MBL resulting in an underestimate of the prevalence of MBL deficiency. We did not confirm by genotyping that low MBL concentrations were the result of variant alleles. Approximately 30% and 4%-8% of the normal population have heterozygous or homozygous genetic mutations, respectively, associated with low MBL concentrations. 35 Although MBL concentrations are well correlated with genotypes, (eg, MBL concentrations , 50 ng/mL are 100% sensitive and 83% specific for variant exon-1 polymorphisms 43 ) subjects with wild-type MBL genes may still have low concentrations of MBL, 35 and MBL concentrations may vary as much as 10-fold in patients with identical genotypes for all of the MBL variants described to date. 38 Accordingly, assessing relationships between MBL concentrations and AECOPDs is likely to be a more sensitive approach than genotyping. 21 Similarly, we did not assess MBL function using a complement deposition assay. Bouwman et al 47 and Eisen 48 suggested that while MBL function was a better way of defining MBL deficiency than determining MBL concentrations, assessing MBL concentrations was "most appropriate" for defining MBL deficiency in studies seeking associations with infections. Differences between MBL binding and complement activation may vary depending on the method of assessment, 49 and MBL concentrations from 500 ng/mL to 1000 ng/mL are associated with decreases in function by up to 90%. 9 While we saw no association between MBL deficiency and AECOPDs, MBL deficiency could still be associated with AECOPDs that result from infection with one or more specific pathogens if the frequency with which these specific pathogens affected our subjects was too low to discern the association. However, we think this possibility is unlikely, because the two studies documenting the association between MBL deficiency and AECOPDs found an association with AECOPDs that were not otherwise defined by cause or potential infecting organism. 30, 32 In addition, Lin et al 32 found no difference in the distribution of pathogens in those subjects with AECOPD and without MBL-deficient genotypes. Because we only found 20 subjects (1.9%) with MBL concentrations below the lower range of normal reported by the manufacturer of the assay, we cannot exclude the possibility that very low concentrations of MBL have an association with AECOPDs. However, the infrequency of this finding implies that even if this association were found, it would pertain only to a small minority of patients suffering AECOPDs. In conclusion, we found a very low prevalence of MBL deficiency in subjects who were at increased risk for experiencing AECOPDs and no association between MBL deficiency and time to first acute exacerbation, frequency of acute exacerbations, or percent of subjects requiring hospitalization for acute exacerbations in subjects with MBL deficiency regardless of how deficiency was defined. Accordingly, our data do not support the idea that MBL might be a therapeutic target to reduce the incidence of AECOPDs. Rather, they imply that, while COPD is an inflammatory disorder with systemic manifestations, the fundamental pathophysiology of COPD differs from conditions in which MBL deficiency seems to be a clear risk factor (ie, childhood pneumonia, rheumatoid arthritis, systemic lupus).

The changing phenotype of microglia from homeostasis to disease

It has been nearly a century since the early description of microglia by Rio-Hortega; since then many more biological and pathological features of microglia have been recognized. Today, microglia are generally considered to be beneficial to homeostasis at the resting state through their abilities to survey the environment and phagocytose debris. However, when activated microglia assume diverse phenotypes ranging from fully inflamed, which involves the release of many pro-inflammatory cytokines, to alternatively activated, releasing anti-inflammatory cytokines or neurotrophins, the consequences to neurons can range from detrimental to supportive. Due to the different experimental sets and conditions, contradictory results have been obtained regarding the controversial question of whether microglia are “good” or “bad.” While it is well understood that the dual roles of activated microglia depend on specific situations, the underlying mechanisms have remained largely unclear, and the interpretation of certain findings related to diverse microglial phenotypes continues to be problematic. In this review we discuss the functions of microglia in neuronal survival and neurogenesis, the crosstalk between microglia and surrounding cells, and the potential factors that could influence the eventual manifestation of microglia.

I. Introduction II. The origin of microglia III.Microlgia the dual natures of neurotoxicity and neuroprotection IV.Crosstalk between microglia and other brain cells 1. Cross talk between microglia and neurons: neurons as regulators of microglial activation 2. Cross talk between astrocytes and microglia: reciprocal influences 3. Microglia-T cell crosstalk:key determinants for the Introduction Microglia are generally considered the immune cells of the central nervous system (CNS) and account for 10% of the total glial cell population in the brain. In a normal physiological environment, they work as sentinel cells by continually screening the brain tissue; they actively participate in pathological processes by changing morphology, expressing various antigens and becoming phagocytic. During the past 20 years, thousands of papers have been published describing both the detrimental and beneficial roles of microglia in various brain disorders, from acute infection or stroke to the long and chronic process of neurodegeneration. Microglia have been firmly established as a key cellular component involved in the eventual outcome of inflammation and eventually contribute to the chronic neurodegeneration; The physiology and signaling of microglia have been comprehensively reviewed by Kettenmann's series papers [1] [2] [3] [4] [5] [6] , however, the regulation of microglial activity is a highly complex system, and the responses of microglia are tailored in a multi-factor dependent manner, and which are the focus we try to review in this paper. The precise origin and cell lineage of microglia has been a long time debate. So far two most important hypotheses for microglial origin have been held: "neuroectodermal" or "myeloid-monocytic". Even though the latter has been more widely accepted now, the neuroectodermal hypothesis remains interesting. Skoff [7] detected "multipotential glia cell" with a rat model of optic nerve degeneration and optic nerve development, these cells were demonstrated to originate from neuroectodermal matrix cells, and Kitamura later confirmed this result by describing a continuous morphological transition between glioblasts and ramified microglia in the developing gray matter of hippocampus [8] . The hematopoietic origin of microglia also received a lot attention, the presence of bone marrow Mac-1 positive cells were demonstrated in the brain of embryonic and adult mice, and these cells were proved to be the progenitors for microglial cells [9] , also transplantation of GFP + mice bone marrow cells in GFP-host mice revealed the presence of many GFP + microglia throughout developing and/or inflamed CNS [10, 11] , which strongly suggest the hematopoietic stem cells as one of the origins for replenishment of microglia in the neuropathology. Additionally due to the high similarity in marker expression and phagocytosis behavior between circulating monocytes and microglia, people speculate the monocytic origin of microglia, and a couple of experiments have been performed to show the appearance of labeled monocytes in the developing [12] or inflamed brain [13] . In many cases, the peripheral macrophages are considered to be the orthologue [14, 15] or backup of microglia and infiltrate the brain to supplement microglia, thus to some extent peripheral macrophages mirror the behavior of microglia in the brain and Monocytederived Macrophages (MDMs) from patients have been used as a substitute of microglia in many studies [16] [17] [18] . Neuroinflammation has long been considered a mediator of secondary damage following a small injury to the CNS. As the primary immune cells in the brain, microglia are expected to take active roles in the damage process. The presence of activated microglia within injured brain regions and in post-mortem tissue from patients having various neurodegenerative disorders has led to the assumption that all reactive microglia contribute to an adverse and degenerative process. Further studies describe destructive roles for microglia by demonstrating the release of a range of neurotoxins from microglia that includes pro-inflammatory cytokines [19] [20] [21] , nitric oxide [22, 23] and reactive oxygen species [24, 25] ; the inhibition of microglial activation in various experiments results in the attenuation of neurotoxic events and improves neuronal survival. In various neurodegenerative disorders, the over-activation of microglia is considered to be a key causative factor in the process or, at a minimum, to promote the neuropathology. For example, in Alzheimer's disease, microglia activated by amyloid-β(Aβ) protein, the hallmark of the disease, release neurotoxins and potentiate neuronal damage, and this microglial over-activation is an early event that precedes neuropil destruction [26] . The activated microglia cluster around or penetrate the neuritic plaques [27] , supporting a critical role of microglial activation in the pathogenesis and progression of the disease. In Parkinson's disease (PD), an increased number of activated microglia are present in the vicinity of degenerating neurons [28] in the substantia nigra [29] , which is particularly deleterious to dopaminergic neurons due to their glutathione deficiency [30] . A single injection of lipopolysaccharide (LPS) to activate microglia in the substantia nigra region led to a progressive, preferential and irreversible loss of dopaminergic neurons [31] [32] [33] , even though LPS itself did no direct harm to the neurons, indicating that the over-activation of microglia is capable of inducing neuronal death in the absence of other pathological stimulation. All of the evidence described above supports the hypotheses of the neurotoxic features of microglia. However, as the sentinel and essential cells of the CNS, it is unlikely that microglia would function to damage neurons in all scenarios. Once stimulated the microglia migrate rapidly to the injury site along the chemokine gradients in vitro [34] and also in response to chemoattractants including ATP and NO released directly or indirectly by the injury [35] to exert effect on the survival of neurons. In fact, some specifically designed experiments have begun to uncover the neuroprotective roles of microglia, and more studies are emerging to show beneficial functions of microglia. Firstly, studies have demonstrated instructive roles for microglia in the developing brain for neuronal differentiation [36, 37] and in the regulation of neuronal apoptosis [38] through the production of neurotrophins [39] . Secondly, in the adult brain, resting microglia, which are characterized by many fine perpendicular processes extending from a few long prolongations, have been regarded as sensor cells for the detection of abnormalities or changes in the brain [40] and help to maintain environmental homeostasis. Lastly but most importantly, activated microglia have also been shown to perform neurotrophic functions following neuronal injury. One compelling study supporting this finding involves the axotomy of peripheral nerves (facial or optic), where a rapid microglial response is exhibited with the efficient clearance of myelin debris that contained inhibitory molecules of axon growth, finally leading to successful axonal regeneration [41] ; the inhibition of this microglial response to facial nerve axotomy impairs neuronal survival [42] . In addition, in neonatal mice administered MPTP, highly activated microglia show neurotrophic potential towards dopamine neurons [43] and after traumatic injury, clear glutamate without evoking inflammatory mediators [44] . The benefits of microglial activation are further demonstrated by the exacerbation of neuropathology in inducible mouse models that are deficient in microglia [45, 46] , the finding of protective microglia in cases of cerebral ischemia [47] and multiple sclerosis [48] and the fact that transplantation of microglia can help to enhance neurite growth and functional recovery after CNS injury [49, 50] . The bunch of factors that can activate microglia and the differential behavior of microglia in various conditions have been listed in Table 1 & 2. The above studies clearly demonstrate that microglia can be neurotrophic in the proper situations; there might be a third possibility that microglia are activated by simply reacting to pathogenic stimulation and takes very limited roles in the neurological disorders, in such case the activation of microglia is solely a result of pathogenic stimulation and work as a bystander that either involved passively during the whole process or even go to apoptosis by some other signals. Thus These activated microglia might have different phenotypes. However, the details of what conditions induce microglia to take beneficial phenotypes remain unknown. Many factors are likely involved in determining the eventual outcome of the manifestation of microglia, including their interaction with neurons or astrocytes in the same environment, age-related dysfunction of microglia, activation timing, and the activation state of the microglia, which we will be discussing below. Microglia have been considered to be the first line of defense in the CNS [91] , a hypothesis that has been supported by the finding that microglia actively screen their microenvironment with highly motile processes; thus, the brain is under continual surveillance by microglia. To do this with high efficiency, microglia must be variable, adaptive to their environment and capable of integrating various inputs and responding appropriately [92, 93] . All of these processes require significant interactions with other components within the same environment, including neurons and astrocytes. When we talk about whether microglia are neuroprotective or neurotoxic, we only refer to the influence of microglia on neurons. However, many studies indicate that neurons are not merely passive targets of microglia but rather exert control over microglial activities [94] . There are considerable interactions between neurons and microglia. For example, Polazzi hypothesized that activation of microglia as a consequence of neuronal injury is primarily aimed at neuroprotection, with the loss of specific communications between neurons and microglia leading to the neurotoxic behavior of microglia [95] . Accumulating evidence demonstrates that there is significant information exchange between neurons and microglia. Depending on whether they are healthy or injured, neurons send "on" or "off" signals to influence microglial activation. On one hand, the activation of [81, 82] microglia by neuronal injury or degeneration has been widely reported [91, 96] . On the other hand, in the healthy brain, microglial activation is tightly restricted by signaling from neurons. CD200-CD200R has been identified as one of the critical pathways in attenuating microglial activation. CD200 is a member of the immunoglobulin superfamily and is expressed on the neuronal membrane surface, while the CD200 receptor (CD200R) is primarily present in the macrophage lineage, which includes microglia [97] . The disruption of CD200-CD200R interactions results in an accelerated microglial response, whereas intensified CD200-CD200R interactions contribute to an attenuation in neurodegeneration [98] . In mice that have had CD200 selectively removed from neurons, microglia exhibited an activated phenotype and were numerous upon facial nerve transaction; damaged CD200-deficient neurons elicited an accelerated microglial response, which demonstrated a loss of the neuronal inhibitory signal for microglial response [97] . Apart from direct interactions through receptor-ligand combinations, electrical activity and soluble factors released from intact neurons also maintain microglial quiescence. In a neuron-glia co-culture, the blockade of neuronal electrical activity by tetrodotoxin or a glutamate receptor antagonist facilitated microglial activation induced by IFN-γ [99] . Soluble molecules from neurons such as neurotrophins and anti-inflammatory agents downregulate antigen expression on cultured rat microglia [99, 100] . Additionally, released factors from neurons can also influence the survival of microglia. Fukui et al. demonstrated that treatment with conditioned media from mature neurons significantly induced the death of microglial cells independent of LPS, while heated neuron-conditioned media or low-calcium-ion media prevented the death of microglia [101] , indicating that specific factors released from neurons exert detrimental effects on microglia. It has been demonstrated that microglial cells undergo apoptosis following peripheral nerve injury [102] [103] [104] or in cases of experimental autoimmune encephalomyelitis(EAE) [105] Injured neurons induced either neuroprotective or neurotoxic behaviors in microglia depending on the manner of injury [91, [106] [107] [108] [109] , providing strong evidence to support the hypothesis of crosstalk between neurons and microglia. Thus, microglia are not merely surveyors of brain tissue but also receive and actively respond to signals from neurons. Although less obvious than the crosstalk with neurons, the interactions between microglia and astrocytes are far from simple and are also crucial for our understanding of how microglia respond to their environment and exert influence on neuronal degeneration or regeneration. Several studies have demonstrated the substantial influence of astrocytes on microglial activation [110] . The induction of microglia by Trimethyltin or Borna disease virus-infected neurons is dependent on the presence of astrocytes [111, 112] . Astrocytes play neuroprotective roles by modulating microglial cell activity and decreasing their cytotoxicity [113, 114] . The expression of IL-12 and the production of inducible nitric oxide synthase (iNOS) in activated microglia have been reported to be suppressed by astrocytes or conditioned media from astrocytes [82, 111, [115] [116] [117] , delineating the signals from astrocytes that affect the activities of microglia. Furthermore, the communication between these two types of cells is two-way; microglia both receive and give signals, as proinflammatory cytokines released from microglia inhibit gap junctions and down-regulate connexin 43 expression in astrocytes [118] [119] [120] , which enhances astrocyte survival. In another study, comparative proteome analysis was performed on astrocytes that were treated with conditioned media from quiescent or activated microglia. Following culture in activated-microglial media, the anti-oxidative enzymes expressed in astrocytes were up-regulated, and these astrocytes were protected against oxidative stress. This result gave insight into the complex intercellular events that take place during neurological disorders [121] . Alzheimer's Disease Internalize and degrade amyloid beta [87] Multiple sclerosis Secrete soluble mediators that trigger neural repair and usually contribute to the creation of an environment conductive for regeneration [48] As in many pathological conditions in the central nervous system such as in neurodegeneration [122] , microglia, activated earlier than astrocytes, promote astrocytic activation through IL-1which is mostly from microglia [123] . On the other hand, activated astrocytes not only facilitate activation of distant microglia via calcium wave [124, 125] , but also inhibit microglial activities [126] . Additionally, it was observed that activated-microglial-conditioned media increased astroglial proliferation [127] , down-regulated the astroglial metabotropic glutamate receptor [128] and induced astroglial brain-derived neurotrophic factor (BDNF) and IL-6 gene expression [129] . Taken together, the importance of microglial activities lies in that they not only exert direct effects on neuronal survival, but they also affect the responses of other supporting cells in the same environment. Microglia-T cell crosstalk: key determinants for the trend of immune response The entire immune response consists of the cooperation of the innate and adaptive immune systems. In the brain, it has been postulated that the beneficial or destructive outcome of the local microglial (innate) response is determined by a well-controlled dialogue between the innate and the adaptive immune players, which are, in most cases, the microglia and T cells. Activated T cells can cross the bloodbrain barrier and interact with resident microglia in the parenchyma [130] ; these microglia have been characterized as myeloid progenitor cells that can differentiate into macrophage-like or dendritic-like cells [131] and thus work crucially as the principal APCs [85] in the CNS. Monsonego et al. demonstrated that IFN-γ-treated microglia serve as efficient Aβ antigen-presenting cells (APCs) of both Aβ1-40 and Aβ1-42, mediating CD86-dependent proliferation of Aβ-reactive T cells [132] . The activated T cells then exert effects in the injured neural tissues by altering the reactive microglial phenotypes and inducing the astrocytic expression of growth factors or modulating microglia to act as glutamate scavengers [44] to improve neuronal survival [133, 134] . In a model for optic nerve injury, the passive transfer of regulatory CD4 + CD25+ T cells was either destructive or beneficial depending on the genetic background of the mice tested, which determines the differential interaction of T cells with microglia and thus the different T cellmediated microglial phenotypes [133] . Kipnis even observed that both the suppressor and the effector activities of T cells could be mediated through dialogue with microglia in the condition of neurodegneration [135] , The entire scenario of crosstalk between T cells and microglia could be described as the following: microglia are initially activated by pathological stimuli during acute or chronic injury to the brain; if the activation occurs with the proper timing and mode and is well-controlled, the activated microglia will work as APCs [133] to stimulate Treg cells that eventually modulate the microglial activation directly or indirectly and affect the milieu balance between neurotrophism and cytotoxicity [44, 136, 137] . Whether microglial activation is neurotrophic or neurotoxic is context-dependent After considerable time and research, we have recognized the "double-edged sword" nature of microglial cells. On one hand, significant evidence from in vitro and in vivo studies has associated neuronal injury with microglial activation [138] [139] [140] [141] . This evidence results from an inflammatory phenotype of microglia releasing neurotoxic factors, mediators and reactive oxygen species [138] [139] [140] [141] . On the other hand, several other studies have highlighted the beneficial and important roles of microglia in neuronal regeneration, repair and neurogenesis [142] [143] [144] [145] [146] . These seemingly paradoxical results cannot be directly compared, because they come from different experimental sets that vary in terms of the stimulus, timing of microglial activation and age of animals. Thus, whether microglia have positive or negative effects on neuronal survival is contextdependent. There are studies suggesting that senescence in microglia causes them to function abnormally and that the destructive roles of activated microglia in the aged neurodegenerative brain may result from age-associated microglia senescence, causing a failure of the aged microglia to respond correctly to stimuli [147, 148] and eventually promoting neurodegeneration [149] (Figure 1 ). The most prominent and also the initially identified feature of microglial senescence is the morphological alteration described as "dystrophy" [150] . Characteristics of "dystrophic" microglia observed in the aged brain include de-ramification (the loss of finely branched cytoplasmic processes), cytoplasmic beading/ spheroid formation, shortened and twisted cytoplasmic processes, and instances of partial or complete cytoplasmic fragmentation [150] . Such dystrophic microglia were prevalent and extensively distributed in the brain of older human subjects [150, 151] , whereas normally ramified microglial morphology with only rare instances of dystrophic microglia is observed in the young brain [148] . These observations provide initial evidence for the age-associated changes in microglia in the healthy elderly brain. Telomere shortening, a marker of aging, has also been demonstrated in microglia in the aged brain in Flanary's study, who reported that microglial cells in rats exhibit significant telomere shortening and a reduction in telomerase activity during normal aging [152] . More importantly, microglial senescence is also manifested by functional alterations, such as an altered inflammatory profile, increased immunophenotypic expression, and the switch from neuroprotective Luo and Chen Translational Neurodegeneration 2012, 1:9 Page 5 of 13 http://www.translationalneurodegeneration.com/content/1/1/9 in the young brain to neurotoxic in the aged brain upon activation [147] . Also, the timing of microglial proliferation and presentation in the injured aged brain is distinct from that in the young brain. For example, Conde et al. reported that microglial proliferation rates in the aged rat brain were significantly higher than in the young rat brain four days after axotomy of the facial nerve [148] . The distinct pattern of the microglial response to injury in the aged brain has also been recorded in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced model of neurotoxicity [153] , the model of controlled cortical impact (CCI) [154] , cortical stab injury [155] and transient retinal ischemia [156] . Although more attention has been paid to the dysfunction of aged microglia, many critical questions remain unanswered. Some of these questions are: whether the activated state of microglia in the aging brain is concurrent with or secondary to microglial dystrophy; which specific function of microglia is primarily affected by microglial dystrophy, how it is affected and what is the direct consequence of the affected function; and whether the deterioration of a specific microglial function is more related to neurodegeneration than others. Clearly, more research is needed to answer these questions. Another important element that critically determines the destructive or neuroprotective role of microglia is the timing of their activation. Because large and very complicated communications pathways exist between immunocompetent cells and cytokines in the CNS, the timing of microglial activation leads to diverse trends and outcomes related to the entire inflammation event. In a model of optic nerve crush injury, Shaked et al. found that an earlier onset of phagocytic activity and antigen presentation by microglia results in a resistance to injury and neurons survived [133] ; the early, moderate, transient and well-controlled activation of local microglia caused them to function as APCs, leading to the communication with Treg cells that subsequently proves to be neuroprotective through the modulation of microglial activation states [133] . In a multiple sclerosis (MS) Luo and Chen Translational Neurodegeneration 2012, 1:9 Page 6 of 13 http://www.translationalneurodegeneration.com/content/1/1/9 model of experimental allergic encephalomyelitis (EAE) [157] , the inhibition of microglial activation through tPA knockout (tissue plasminogen activator, an essential element for microglia activation) leads to a delayed onset of the disease but increased severity and delayed recovery from the neurological dysfunction, which suggests that microglial activation is harmful during the onset of the disease but beneficial in the recovery phase [157] . Furthermore, when microglial activation was either stimulated or inhibited at different stages, the disease progression was attenuated or exacerbated accordingly [158] . For example, the inhibition of microglial activation at EAE onset, rather than prior to EAE induction, markedly decreased EAE progression, while the stimulation of microglial activation prior to the onset of EAE promotes lower-level EAE and an earlier recovery from symptoms. Together, these findings suggest different roles for microglial activation during various phases of the disease and that different timing of microglial activation dramatically affects whether microglia will be neuroprotective or deleterious [158] . Similarly, in an oxygen-glucose deprivation model, the time window of microglial neuroprotection has been estimated to up to 48 hour after injury, while the pre-stimulation of microglia with LPS before the injury fails to induce microglial-mediated neuroprotection [86] . It has been proposed that the effects of the early activation of microglia on disease progression could be beneficial through phagocytic activity and antigen presentation, recruitment and interactions with the adaptive immune response and the induction of protective autoimmunity [133] . Furthermore, the balance between protective autoimmunity and autoimmune disease may be determined by the timing and intensity of microglial activation [133] . As the immuno-competent cells in the CNS, microglia are critical determinants of the outcome of injury, and the timing of microglial activation appears to be crucial to the outcome of the injury. Thus, any interference with microglial activation in an attempt to affect the disease course clearly must be temporally-restricted. Two distinct phenotypes of macrophages have long been known to play different roles in the inflammatory context. Classically-activated macrophages, characterized by the involvement of T Helper type 1 (Th-1) cytokines such as interferon-γ, promote the release of various proinflammatory cytokines and thus exacerbate the inflammation. Alternatively, activated macrophages predominate in the T Helper type 2 (Th-2) microenvironment and tend to soothe the inflammation. Thus, the behavior of macrophages is dictated by their phenotype, which may eventually affect the beneficial or detrimental roles of macrophages during inflammation. Similarly, research over the past few years has established that microglia do not constitute a single, uniform cell population, but rather comprise a family of cells with diverse phenotypes; some are neuroprotective while others are destructive [92] . So far, three distinct functions have been proposed for microglia. The first is the classical activation state of microglia, which, accompanied by the induction of receptors that participate in the innate immune response [159] , is responsible for the pro-inflammatory milieu, and has been linked to neurotoxic effects in the brain. The second is alternatively activated microglia, which are associated with the production of anti-inflammatory cytokines in the resolution phase of the inflammatory response. Recently, the third activation state of microglia has been identified: it overlaps with and is complementary to the alternative activation and is called acquired deactivation [160, 161] . This is another activation state that promotes immunosuppression and is associated with the anti-inflammatory and functional repair phenotype .Both alternative activation and acquired deactivation down-regulate innate immune responses and have similar gene profiles; the most prominent difference is that acquired deactivation is induced by the exposure of microglia to apoptotic cells or to TGF-β or IL-10, while IL-4 and IL-13 induce alternative activation [160, 161] . It has been observed that multiple activation states of microglia coexist in certain chronic inflammations due to parasitic disease [162] , in which the balance between classical activation and alternative activation/acquired deactivation states is of "benefit" to both host and parasite: the host benefits from reduced self-damage, and the parasite eventually survives within the host. Neurodegenerative disorders are also associated with chronic inflammation and the coexistence of various activation states. For example, in AD, some levels of classical activation may be required to limit the brain levels of Aβ despite the risk of self-damage [163] , while alternative activation of microglia in AD may foster the protection of the surrounding tissue from immune damage even though it may facilitate Aβ deposits. Similar studies [164] [165] [166] have shown that the immune cells in the vicinity of amyloid deposits in AD express mRNA and proteins for pro-inflammatory cytokines, leading to the hypothesis that microglia demonstrate classical activation in AD, while Colton et al. found increased mRNA expression of alternative activation-associated gene profiles in microglia in both the AD brain and an AD mouse model [167] , suggesting the presence of multiple activation states of microglia during neurodegeneration. However, the recognition of heterogeneous phenotypes of microglia only raises more questions: what instructs microglia to acquire a particular phenotype; can any conversion occur between these phenotypes; and is it possible to avoid or at least change the commitment to a destructive phenotype? All of these questions are difficult to answer with our current knowledge of microglia; more extensive work is warranted before we can reach a conclusion. As an active sensor in the brain, microglia respond to even minor stimuli; however, different types of stimulation may also lead to different actions of microglia and thus be either harmful or beneficial to neuronal survival. In a neonatal mouse MPTP-induced brain injury model, microglia activated by systemic administration of LPS were shown to be neuroprotective. In contrast to the MPTP model, LPS-activated microglia in neonatal mice receiving a stereotaxic injection of ethanol into the striatum were shown to be neurotoxic, and systemic LPS administration in the ethanol-injury model caused a marked increase both in the volume of necrotic lesions and in the number of degenerating neurons in the striatum [168] . Even with the same stimuli, the degree can also determine microglial release of toxic versus protective effectors [169] ; neurotoxic cytokines and ROS were released from microglia only in response to mild neuronal injuries, while trophic microglial effectors such as BDNF and GDNF were up-regulated in response to all degrees of neuronal injury [169] . Additionally, different types of pain resulted in differing activations of microglia [170] . So far, what we know is that not all microglia respond in the same way, even to the same stimulus, and microglial function is tailored in a context-specific manner [171] . Numerous elements are involved in this context; most likely there are many more beyond what we have discussed here. Identifying these elements and clarifying their interactions or crosstalk with microglia is essential before we are able to design a strategy to control inflammation through the manipulation of microglia. The simple therapy of inhibiting all microglia without differentiating their function in a context-dependent manner surely should be abandoned. It has been long recognized that the birth of new neurons within the postnatal brain continues throughout life and remains as a potential source of replacement cells in the CNS for the treatment of disease. The microenvironment or the niche in which neural progenitor cells live critically influences the process of neurogenesis, which spans several steps including the proliferation of stem or progenitor cells; the survival of immature or mature neurons; the migration of new neuroblasts to their appropriate locations; and the differentiation of neuroblasts to a neuronal phenotype and the construction of synaptic connectivity [172] . As an important component of the brain microenvironment and due to their invariant participation in most pathological processes in the CNS, microglia are increasingly implicated as a potential non-neural regulator of neurogenesis, as demonstrated by circumstantial evidence [144, 172] . However, just as in the debate over the neuroprotective or neurotoxic nature of microglial activation, whether microglia support or damage the survival and development of neural progenitor cells also remains controversial. On one hand, microglia were shown to play instructive roles during postnatal neurogenesis in the neurogenic niche either by influencing the differentiation of stem cells toward a neuronal phenotype or by directing their migration [144, [173] [174] [175] . On the other hand, multiple studies have demonstrated the deleterious effect of microglial activation on neurogenesis [176, 177] and the effective restoration of neurogenesis though the blockade of microglial activation. In the two situations of neurogenesis and neuronal survival, similar factors are shared, leading microglia to take supportive or detrimental roles. Among these factors, the most prominent is the microglial activation phenotype that is associated with different cytokine profiles. When acutely activated by either LPS or injury, microglia that release the pro-inflammatory cytokines IL-6, TNF-α or IL-1β usually down-regulate the differentiation or proliferation of neural stem cells or induce the aberrant migration of newborn neurons [178] . This group of inflammatory cytokines has been proven to inhibit neurogenesis [176, 177, 179] ; conversely, blocking antibodies to these pro-inflammatory cytokines (such as IL-6 [177] ) or the use of monocycline to mitigate the microglial activation simply restores neurogenesis [176] . In contrast, microglia that are activated by anti-inflammatory cytokines such as IL-4 or TGF-β increase neurogenesis in vitro or the differentiation of neural stem cells (NSCs) in vivo [180, 181] . Neurotrophins, such as IGF-1, were identified [181] in anti-inflammatory cytokineactivated microglia and were proposed to be one of the mechanisms underlying this pro-neurogenic activity of microglia [182, 183] . However, just like the dual roles in neuroprotection, whether a specific cytokine-activated microglial cell will take a pro-or anti-neurogenic role is also context-dependent. For example, microglial cells activated by IFN-γ, a pro-inflammatory cytokine can be neurotoxic or supportive of neurogenesis, depending on the concentration of IFN-γ [184] . TGF-β, which is considered to be beneficial to neurogenesis, can actually exert a negative influence on neurogenesis when it is chronically produced in the aged brain [185] . Additionally, if other cytokines exist in the same niche simultaneously, the outcome will be determined by the balance among the various cytokines; some authors have concluded that activated microglia are not pro-or anti-neurogenic per se, but the balance between pro-and anti-inflammatory secreted molecules influences the final effect of microglial activation [172, 180] . However, in which situations the microglia will release pro-or anti-inflammatory cytokines is complicated and is affected by multiple factors such as the injury type, the phase of disease or inflammation, and crosstalk with other regulating components, including neural precursors; this is similar to the question of whether microglia will be neuroprotective or neurotoxic. Most likely the same inflammatory scenario that induces neurodegeneration would also inhibit neurogenesis, while a situation that favors neuronal survival would also support neurogenesis. Interestingly, even in a high-inflammation environment, such as two days after a Trimethyltin-induced acute injury in the hippocampus, significant neurogenesis can be detected [186, 187] , suggesting a complicated system of neurogenesis regulation beyond the inflammation scenario. Cumulative studies have found an age-related decline in neurogenesis, both in the aged adult and in the diseased brain. Because aging may contribute to microglial dysfunction and neurotoxicity, as we discussed previously in this review, one could assume that microglial dysfunction may also be involved in the downregulation of neurogenesis in the aged or diseased brain [188, 189] . Even though very few studies have focused on the effect of microglial dysfunction on neurogenesis, we can still find a clue from Zhu's study that the difference in microglia function patterns between the immature and juvenile brain might be related to a decrease in neurogenesis in the juvenile brain [190] ; however, stronger evidence from the direct comparison of microglia-associated neurogenesis between aged and young brains is needed to support this view. Another important element regulating the activities of microglia is the T cell, which comes from the peripheral adaptive immune system and enters the CNS by extravasating across the endothelium of the choroid plexus into the cerebrospinal fluid [191] . The interaction of T cells with microglia in the injured spinal cord correlates with enhanced neuronal survival [184] , and rapidly recruited T cells in the middle cerebral artery obstruction (MCAO) model increased hippocampal and cortical neurogenesis by modulating the microglial response and through the production of IGF in the sub-acute phase [192] . Hippocampal neurogenesis was associated with the recruitment of T cells and microglial activation. Immune-deficient mice show impaired neurogenesis in the hippocampus, but this deficiency was attenuated and neurogenesis boosted by T cells recognizing a specific CNS antigen [193] . The cellular source of IFN-γ and IL-4 in vivo is likely to be T cells, therefore it is reasonable to assume that the T cell-mediated immune response is an integral part of the regulation of microglial phenotype or function, and thus can influence neuronal survival or neurogenesis directly or indirectly. From an increasing number of studies of diverse microglial activity in different experimental sets, we are beginning to appreciate the heterogeneity of microglial functions that have either beneficial or detrimental roles in specific physiological or pathological environments. Whether microglia are committed to one function from the very beginning or if there is any conversion between different phenotypes remains elusive and the factors that initiate this commitment or promote its conversion are far from being clarified. Due to the invariant critical participation of microglia in most diseases, ongoing research to uncover these questions is warranted; before we are sure about the answer, any potential strategies targeting microglia to manipulate inflammation and modify a disease course are unrealistic.

Diversity of Salmonella spp. serovars isolated from the intestines of water buffalo calves with gastroenteritis

BACKGROUND: Salmonellosis in water buffalo (Bubalus bubalis) calves is a widespread disease characterized by severe gastrointestinal lesions, profuse diarrhea and severe dehydration, occasionally exhibiting a systemic course. Several Salmonella serovars seem to be able to infect water buffalo, but Salmonella isolates collected from this animal species have been poorly characterized. In the present study, the prevalence of Salmonella spp. in water buffalo calves affected by lethal gastroenteritis was assessed, and a polyphasic characterization of isolated strains of S. Typhimurium was performed. RESULTS: The microbiological analysis of the intestinal contents obtained from 248 water buffalo calves affected by lethal gastroenteritis exhibited a significant prevalence of Salmonella spp. (25%), characterized by different serovars, most frequently Typhimurium (21%), Muenster (11%), and Give (11%). The 13 S. Typhimurium isolates were all associated with enterocolitis characterized by severe damage of the intestine, and only sporadically isolated with another possible causative agent responsible for gastroenteritis, such as Cryptosporidium spp., Rotavirus or Clostridium perfringens. Other Salmonella isolates were mostly isolated from minor intestinal lesions, and often (78% of cases) isolated with other microorganisms, mainly toxinogenic Escherichia coli (35%), Cryptosporidium spp. (20%) and Rotavirus (10%). The S. Typhimurium strains were characterized by phage typing and further genotyped by polymerase chain reaction (PCR) detection of 24 virulence genes. The isolates exhibited nine different phage types and 10 different genetic profiles. Three monophasic S. Typhimurium (B:4,12:i:-) isolates were also found and characterized, displaying three different phage types and three different virulotypes. The molecular characterization was extended to the 7 S. Muenster and 7 S. Give isolates collected, indicating the existence of different virulotypes also within these serovars. Three representative strains of S. Typhimurium were tested in vivo in a mouse model of mixed infection. The most pathogenic strain was characterized by a high number of virulence factors and the presence of the locus agfA, coding for a thin aggregative fimbria. CONCLUSIONS: These results provide evidence that Salmonella is frequently associated with gastroenteritis in water buffalo calves, particularly S. Typhimurium. Moreover, the variety in the number and distribution of different virulence markers among the collected S. Typhimurium strains suggests that within this serovar there are different pathotypes potentially responsible for different clinical syndromes.

Salmonella spp. found in water buffalo (Bubalus bubalis) herds are a matter of concern since they are responsible for serious economic losses in livestock and are a zoonotic agent responsible for foodborne illness [1] . As for bovine calves, Salmonella-induced diseases in water buffalo calves are characterized by severe gastrointestinal lesions, profuse diarrhea, and severe dehydration [1] . Acute salmonellosis generally induces diarrhea, mucous at first, later becoming bloody and fibrinous, often containing epithelial casts. Ingestion is the main route of infection, although it can also occur through the mucosa of the upper respiratory tract and conjunctiva. The major source of infection in the herd is represented by asymptomatic older animals shedding heavy loads of bacteria through feces. Other sources of infection are contaminated forages and water, as well as rodents, wild winged animals, insects and man [1, 2] . The disease can also cause sudden death without symptoms. Occasionally, the infection is systemic, affecting joints, lungs and/or the central nervous system (CNS) [1] . Moreover, several Salmonella serovars seem to be able to infect water buffalo, mainly affecting 1-12 week old calves, even though reports on salmonellosis in B. bubalis are scarce [1, 3] . Water buffalo calves are more frequently affected by gastroenteritis than bovine calves, with mortality rates as high as 70% in water buffalo species vs. 50% in bovine [1, 4] . This difference might be due to a greater susceptibility of water buffalo to gastroenteric pathogens, although it also may reflect the lack of appropriate management practices for this animal species. Therefore, water buffalo represents a suitable model to study causative agents of gastroenteritis. In water buffalo, S. enterica serovar Typhimurium can induce a variety of clinical syndromes with different anatomopathological lesions [1, 3] . The severity of the disease can depend on several factors, including host-pathogen interactions, which is highly influenced by the route of infection, the infectious dose, natural or acquired host resistance factors, and the possible presence of other pathogens. Moreover, specific Salmonella virulence factors, frequently located on Salmonella pathogenicity islands (SPIs), prophage regions or virulence plasmids, play a key role in the pathogenesis of the gastroenteritis [5] . The current study investigated the intestinal contents collected from 248 water buffalo calves affected by gastroenteritis with lethal outcome to: (i) evaluate the prevalence of Salmonella spp., and (ii) perform a polyphasic characterization of the collected isolates of S. Typhimurium. Salmonella spp. were isolated from 25% of the intestinal contents collected from 248 water buffalo calves affected by gastroenteritis with lethal outcome. Positive samples were detected in subjects bred in 37 of 58 farms (interherd prevalence, 64%). The S. enterica serovars most frequently isolated were Typhimurium (n=13), Muenster (n=7) and Give (n=7). Other recovered serovars were: Derby (n=5), 4 Bovismorbificans (n=4), Newport (n=4), monophasic S. Typhimurium (B:4,12:i:-; n=3), Blockley (n=2), Meleagridis (n=2), Umbilo (n=2), Altona (n=1), Anatum (n=1), Bredeney (n=1), Enterica (−;i;1,2; n=1), Gaminara (n=1), Haardt (n=1), Hadar (n=1), Infantis (n=1), Isangi (n=1), Kottbus (n=1), London (n=1), Muenchen (n=1), and S.II:41;z;1,5 (n=1). Phage-typing of the S. Typhimurium and monophasic Typhimurium strains (Table 1 ) indicated a variable distribution of phage types among strains with nine different phage types of 13 Typhimurium strains, and three different phage types out of three monophasic Typhimurium strains. This study reports a significant prevalence of Salmonella spp. (25%) in diarrheic water buffalo calves, that are more relevant than those reported in previous studies (11 and 0.8%) [3, 6] . Moreover, in contrast with bovine species where salmonellosis results primarily associated with serovars Dublin and Typhimurium [5] , the extremely variable distribution of the observed serovars confirms the absence of a serovar specifically adapted to water buffalo, as previously suggested [1] . These data provide therefore evidence that Salmonella, particularly S. Typhimurium, can be potentially considered an important pathogen for this animal species. The definitive phage type 104 (DT104), which has often been associated with multiple-antibiotic-resistant strains with ascertained zoonotic potential and, in many countries, has increased over the past two decades [5] , does not seem to be widely spread in water buffalo. Three monophasic S. Typhimurium (B:4,12:i:-) isolates were also found that are S. Typhimurium lacking phase two flagellar antigens that have a rapid emergence and dissemination in food animals, companion animals, and humans. More significantly, the public health risk posed by these emerging monophasic S. Typhimurium strains is considered comparable to that of other epidemic S. Typhimurium [7] . The diagnostic investigation indicated that non-Typhimurium Salmonella isolates were detected with at least another potential pathogen in 78% of cases ( Figure 1A ). In 35% of cases Salmonella was linked with pathogenic Escherichia coli that were characterized for the presence of virulence factors. Other frequent associations were found with Cryptosporidium spp. (20%) and Rotavirus (10%) ( Figure 1A) . Remarkably, S. Typhimurium was never associated with pathogenic E. coli, while it was isolated sporadically with Clostridium perfringens (strain #82280), Rotavirus (strain #107025), and Cryptosporidium spp. (strain #112) ( Figure 1B) . The presence of more pathogens in the same subject might suggest that, as for other animal species [5] , diarrhea in water buffalo calves can be characterized by a multifactorial etiology. Data from necroscopic examinations of tissues indicated that the lesions caused by S. Typhimurium were characterized by severe damage of the intestine, ranging from congestive to necrotic-ulcerative enterocolitis. In particular, the strains isolated from animals exhibiting the most severe lesions were #16, #92, #233, and #83528. Among these strains, the two DT104 strains were also found, thus supporting the pathogenic role of this phage type. The other Salmonella serovars were instead isolated from subjects exhibiting a variety of different lesions, mostly minor lesions confined to the jejunum, and often (78% of cases) associated with other pathogens. Similarly, the monophasic S. Typhimurium strains were detected either with Rotavirus (strain #154) or st-positive E. coli (strains #175 and #188). These data confirm the pathogenic potential of the serovar Typhimurium for water buffalo calves. On the other hand, the scarcity of observed lesions and the frequent presence of more than one microorganism in the same subject hamper a clear understanding of the potential pathogenic role of the non-Typhimurium Salmonella serovars included in this study. S. Typhimurium and monophasic S. Typhimurium strains were further characterized by the molecular detection of 24 genes coding for virulence factors. The genetic characterization (Table 2) included five loci (avrA, ssaQ, mgtC, siiD, and sopB) located on SPI 1-5, respectively [8] , eight loci (gipA, gtgB, sopE, sodC1, gtgE, gogB, sspH1, and sspH2) of prophage origin [9] [10] [11] [12] [13] , the gene spvC, located on a virulence plasmid [12] , and nine genes (stfE, safC, csgA, ipfD, bcfC, stbD, pefA, fimA, and agfA) coding for bacterial fimbriae, involved in surface adhesion and gut colonization [5] . As a positive control for the PCR assay, amplification of the chromosomal gene invA was carried out for each strain. All the S. Typhimurium and monophasic Typhimurium isolates displayed the presence of avrA, ssaQ, mgtC, siiD, sopB, sspH2, stfE, ipfD, bcfC, stbD, and fimA genes, and the absence of the sopE gene. Other loci were variably distributed among the strains, with frequency values ranging from 38-92% (Table 1) . On the basis of the presence or absence of the 24 loci included in the study, the 13 strains of S. Typhimurium were subdivided into 10 different genotypes (Table 1) ; however, the isolates with identical genotype displayed different phage types suggesting the presence of 13 different strains. Interestingly, the three monophasic S. Typhimurium strains exhibited three different genotypes (Table 1) . The following loci: invA, sspH2, stfE, ipfD, bcfC, stbD, fimA, avrA, ssaQ, mgtC, siiD, sopB were present in all the strains; the sopE gene was not found in any of these strains. b NT = not typeable. The 24 loci-genetic characterization was also extended to the S. Muenster and S. Give isolates to investigate their pathogenic potential because of their large presence in water buffalo calves. In addition they have already been reported to cause saepticemic salmonellosis in cattle and calves [14, 15] . The molecular results (Table 3) indicated that the loci invA, safC, bcfC, fimA and ssaQ were present in all the strains, the genes gipA, gogB, sspH2, sodC1, gtgE, spvC, stfE, ipfD and pefA were not found in any of these isolates, while the remaining loci were variably distributed, with frequency values ranging from 14-86%. In particular, the prophage genes were scarcely present (2 loci in the Muenster serovar, 1 locus in the Give serovar), the plasmidic spvC locus was absent in all the analyzed isolates, while the fimbrial genes and the SPI 1-5 genetic markers were discretely represented (6 loci for the former genes in both serovars, 5 and 4 loci for the latter genes in the serovar Muenster and Give, respectively). Moreover, the molecular profiles allowed to identify 6 different genotypes out of the 7 S. Muenster isolates, and 5 different genotypes out of the 7 S. Give isolates (Table 3) . Our data confirm the high variability of the Typhimurium serovar [9, 10] , mostly related to virulence factors, and highlight the high discriminating potential of the genotyping technique performed. Our data also suggest that monophasic Typhimurium strains are likely to possess a similarly high degree of genetic variability, particularly linked to virulence markers. Moreover, the presence of virulence markers in the isolated strains of monophasic S. Typhimurium, S. Muenster and S. Give could further support their pathogenic potential. The products of the genes included in the virulotyping assay performed here are known to be important during different stages of infection (Table 2) . However, the distribution of these factors among the tested strains highlights the complexity and the variety of potential mechanisms used by Salmonella to induce disease in the host. The avrA, ssaQ, mgtC, siiD, and sopB genes are genetic markers for the presence of the SPI 1-5 in all S. Typhimurium strains tested, although their presence does not necessarily implicate the presence of the entire SPI. SPIs are clusters of genes on the chromosome, likely to be horizontally acquired, and variably associated with enhanced invasion and intracellular survival within both phagocytic and non-phagocytic cells. In particular, SPI-5 has been largely associated with the ability to produce enteritis [5] . The S. Typhimurium strains included in this study all displayed the presence of the investigated SPI markers. Interestingly, these loci appeared widely distributed also among the serovars Muenster and Give. The sopE gene is known to favor the entry of Salmonella into host cells and its presence has been correlated with disease in humans [16] and with the epidemic potential of S. Typhimurium strains in cattle [17] . This gene was absent in all the S. Typhimurium strains included in the present study, while was present in all the S. Muenster strains analyzed. The pefA (plasmid encoded fimbria), agfA (aggregative fimbria A) and spvC (Salmonella plasmid of virulence gene C) genes are all located on plasmids [18] . Five S. Typhimurium isolates tested in the current study possessed both pefA and spvC, two isolates were positive for only spvC, and three isolates were positive for only agfA (Table 1) . These results confirm the presence of more than one virulence plasmid among S. Typhimurium strains isolated from diarrheic water buffalo calves, and suggest horizontal exchange of virulence factors. However, the loci pefA and spvC were absent in all the monophasic S. Typhimurium, S. Muenster and S. Give strains tested. Prophage genes are known to account for most of the variability of closely-related S. Typhimurium strains. Moreover, lysogenic bacteriophages promote changes in the composition of genomic DNA often altering the phenotype of the host [9, 10] . The prophage virulence genes included in this study exhibited a variable distribution among the isolates tested, thus suggesting synergistic and/or redundant effects of these loci on the pathogenicity of Salmonella, likely contributing to the ACTGCGAAAGATGCCACAGA phenotypic variability of this pathogen. These loci were mostly present in S. Typhimurium and monophasic S. Typhimurium rather than in S. Muenster and S. Give isolates. Fimbrial genes appeared widely distributed among all the serovars tested, particularly in S. Typhimurium strains, with frequency values ≥92%, except for the plasmid-borne pefA and agfA genes (with frequency values of 38% and 54%, respectively). These data are consistent with the essential functions of adhesion factors for the attachment and internalization processes that occur during pathogenesis. To better characterize in vivo virulence, three strains representative of all S. Typhimurium isolates were chosen to perform mixed infections in mice. Animal experiments included the two strains exhibiting the highest and the lowest number of virulence factors (strains #92 and #112, respectively), and strain #16, carrying the same virulotype as strain #92, but that does not harbor the agfA locus (Table 1 ). In the competition assay, strain #92 outcompeted both strains #112 and #16 (CI 0.004; P<0.001, and CI 0.031; P<0.001, respectively). These results were confirmed in a gastrointestinal mouse model of infection, which better resembles the clinical form of salmonellosis in livestock. Using oral inoculation, in the competition assay, again strain #92 outcompeted both strains #112 and #16 (CI 0.009; P<0.001, and CI 0.186; P<0.01, respectively). Our data indicate that among those strains included in the experiment, strain #92 was the most virulent in mice. These competition assays in mice suggest a key role of the agfA gene coding for a thin aggregative fimbria involved in the colonization of host intestinal epithelial cells by attachment to glycoprotein or glycolipid receptors on epithelial cell surfaces. Indeed, the strain which was more virulent in in vivo experiments was characterized by a high number of virulence factors and by the presence of the agfA locus. Moreover, it was isolated from one of the subjects with necrotic-ulcerative enterocolitis. The presence of this type of fimbria has been reported in clinical human and animal isolates of Salmonella + + ----+ ---2 15228 -+ --------3 66761 -+ --------3 72827 - Freq. a The following loci: invA, safC, bcfC, fimA and ssaQ were present in all the strains; the genes gipA, gogB, sspH2, sodC1, gtgE, spvC, stfE, ipfD and pefA were not found in any of these strains. [19, 20] . The data presented here suggest that agfA might increase bacterial pathogenicity. Nevertheless, we cannot reject the hypothesis that the mouse model chosen for in vivo experiments could have influenced the virulence phenotype of the tested strains originally isolated from water buffalo calves. Therefore, future studies will be necessary to exclude the possibility that the phenotypic differences observed among the tested Salmonellae are dependent on the animal model or on other virulence factors not included in this study. However, in vivo experiments carried out in mouse models represent a good preliminary source of information on the expression of traits associated with pathogenicity of Salmonella in mammalian species. This study showed a significant (25%) prevalence of Salmonella spp. in water buffalo calves affected by gastroenteritis with lethal outcome. However, our results did not indicate the existence of a Salmonella serovar specifically adapted to water buffalo and highlighted that S. Typhimurium is the most frequently found serovar. The molecular and phenotypic characterization of the S. Typhimurium isolates provided evidence that within this serovar there are different pathotypes potentially responsible for different clinical syndromes, therefore requiring prophylaxis protocols including the use of specific vaccines for the effective control of salmonellosis in water buffalo calves and possible contamination of the food chain. This study was carried out in the Campania region, Southern Italy, during 2008-2009, using samples taken from 248 water buffalo calves bred in 58 different farms. The animals were aged between 1-12 weeks old and were all affected by gastroenteritis with lethal outcome. During necropsy, the intestinal lesions were evaluated and the intestinal content of the involved sections was collected and tested for the presence of Salmonella spp. In addition, the presence of E. coli, Eimeria spp., Cryptosporidium spp., Giardia spp., Coronavirus, Rotavirus, and C. perfringens were also determined to investigate their association with Salmonella spp. The isolation of Salmonella spp. was performed according to ISO 6579:2002 [21] . The isolated Salmonella spp. were serotyped according to the Kaufmann-White scheme [22] . Phage-typing of the isolated S. Typhimurium strains was performed by the Italian National Reference Centre for Salmonellosis (Istituto Zooprofilattico Sperimentale delle Venezie). The presence of Rotavirus and Coronavirus was detected by polymerase chain reaction (PCR) amplification [23, 24] . Cryptosporidium spp. and Giardia spp. antigens were detected by chromatographic immunoassay (Oxoid, Basingstoke, UK). The presence of Eimeria spp. was examined by flotation technique using saturated saline [25] . E. coli and C. perfringens were isolated according to the protocol reported by Quinn et al. [2] . E. coli hemolytic activity was evaluated by growing colonies on blood agar base, while virulence factors (lt-heat-labile toxin, st-heatstable toxin, stx1-Shiga toxin 1, stx2-Shiga-toxin 2, eaeintimin, cnf-cytotoxic necrotizing factor, and cdt-cytolethal distending toxin) were detected by molecular assays, as previously reported [26] [27] [28] . Bacterial DNA was extracted from 1 mL of overnight cultures using Chelex 100 Resin (BioRad, Hercules, CA) and used as the template for the PCR detection of genes listed in Table 2 , as described previously [8] [9] [10] [11] [12] [13] 18] . The primers used to amplify the genes sspH1, sspH2, ssaQ, sopB, siiD, stfE, safC, csgA, ipfD, bcfC, stbD, and fimA were designed using the Primer3 software (version 0.4.0; http:// frodo.wi.mit.edu/), and PCR was performed in a final volume of 25 μL containing HotStar Taq Master Mix (Qiagen, Valencia, CA) 1×, 0.4 μM each primer and 1 μL of extracted DNA. The thermal profile included an initial denaturation step at 95°C for 15 min, followed by 35 cycles at 95°C for 30 s, 58°C for 30 s, and 72°C for 1 min, and a final extension step at 72°C for 5 min. Amplification products were visualized under ultraviolet (UV) light after electrophoresis on 3% agarose gels and staining with SYBRsafe (Invitrogen, Carlsbad, CA). Groups of five age matched (8-10 weeks old) female BALB/c mice used in this study were purchased from Charles River (Calco, Italy). Three strains (S. Typhimurium #16, S. Typhimurium #92, S. Typhimurium #12), representative of the 13 genotypically characterized S. Typhimurium isolates, were selected for an in vivo analysis of virulence by using the Competitive Index (CI) resulting from mixed infections [29] . In particular, two strains were selected that exhibited the highest and lowest number of virulence factors (strains #92 and #112, respectively), and strain #16, carrying the same virulotype as strain #92, but without the locus agfA (Table 1) . Bacteria were grown overnight at 37°C in Brain Heart Infusion medium (Oxoid, Basingstoke, UK), washed, and diluted in sterile saline. Cultures were alternatively combined in a mixture of equivalent numbers (1:1 ratio) of two of the three selected strains (input). Mice were inoculated intraperitoneally (IP) with a dose of 2×10 4 bacteria or received 20 mg of streptomycin orally (200 μL of sterile solution or sterile saline) 24 h prior of being intragastrically administered with 2×10 7 bacteria. The number of colony-forming units (CFU) contained in the inocula were confirmed by plating serial dilutions and counting colony growth. At 4 (IP) or 7 (os) days after infection, mice were sacrificed, spleens were aseptically removed, and bacteria were counted by plating serial dilutions (output). The ratio of two strains in the input and in the output was evaluated by picking and transferring 200 colonies on selective plates. Antibiotics used were streptomycin and sulfonamide, for which strain 92 and strains 16 or 112 were naturally resistant. The CI was calculated using the formula: CI = output (strain A/strain B)/inoculum (strain A/strain B). Statistical differences between outputs and inputs were determined by Student's t test. All animal handling and sampling procedures were performed under the conditions of the local ethics committee meeting the requirements of Italian legislation.

Severe Childhood Malaria Syndromes Defined by Plasma Proteome Profiles

BACKGROUND: Cerebral malaria (CM) and severe malarial anemia (SMA) are the most serious life-threatening clinical syndromes of Plasmodium falciparum infection in childhood. Therefore it is important to understand the pathology underlying the development of CM and SMA, as opposed to uncomplicated malaria (UM). Different host responses to infection are likely to be reflected in plasma proteome-patterns that associate with clinical status and therefore provide indicators of the pathogenesis of these syndromes. METHODS AND FINDINGS: Plasma and comprehensive clinical data for discovery and validation cohorts were obtained as part of a prospective case-control study of severe childhood malaria at the main tertiary hospital of the city of Ibadan, an urban and densely populated holoendemic malaria area in Nigeria. A total of 946 children participated in this study. Plasma was subjected to high-throughput proteomic profiling. Statistical pattern-recognition methods were used to find proteome-patterns that defined disease groups. Plasma proteome-patterns accurately distinguished children with CM and with SMA from those with UM, and from healthy or severely ill malaria-negative children. CONCLUSIONS: We report that an accurate definition of the major childhood malaria syndromes can be achieved using plasma proteome-patterns. Our proteomic data can be exploited to understand the pathogenesis of the different childhood severe malaria syndromes.

Human malaria caused by Plasmodium falciparum has an estimated annual global disease burden of 300 million clinical episodes, leading to one million deaths [1] [2] [3] [4] . Eighty-five per cent of the cases and 90% of the mortality occurs in sub-Saharan Africa, mostly amongst children [5, 6] . Recent reports point to a reduction of malaria cases in parts of Africa [7] . However, Nigeria, the most populous country of Africa, accounts for a quarter of the global cases and a third of the malaria-attributable childhood deaths [2, 8, 9] . Cerebral malaria (CM) and severe malarial anemia (SMA) are the major severe disease syndromes in African children with a high level of mortality in the under-five age group. The current WHO case definitions for severe malaria combine P. falciparum blood stage parasitemia with coma, severe anemia or respiratory distress [10] , and it is well documented that there is significant overlap across these syndromes [11] . Despite the fact that these WHO case definitions are sensitive and useful for clinical diagnosis, the pathogenesis of severe disease is not well understood. One disadvantage of the WHO clinical definitions is that they lack the specificity required to carry out studies aimed at understanding the pathogenesis of clinically different forms of childhood malaria. Previous studies have attempted to define malaria syndromes by studying plasma correlates of severity using reductionist approaches with variable success [12] [13] [14] . Small sample sizes, a lack of validation cohorts and a focus on a small selection of host plasma proteins have limited these studies. To overcome such limitations we use a systems approach to define the plasma proteome profile during malaria infection and identify distinctive patterns that are characteristic of different disease states. Contrary to other proteomic approaches, high-throughput plasma proteome profiling enables simultaneous analysis of a large number of samples. Therefore plasma proteome profiling allows the use of statistical pattern-recognition methods to discover and validate proteome-patterns that discriminate disease states. We hypothesized that the plasma proteome during malaria infection reflects the molecules that are modulated as the severe status is established. In the present study we show that distinctive plasma proteome-patterns distinguish the different severe presentations of P. falciparum childhood malaria from the uncomplicated cases and also from well or unwell children without malaria. Parents or guardians of study participants gave informed written consent. This research was approved by the joint ethics committee of the College of Medicine of the University of Ibadan and the University College Hospital Ibadan. All study participants were recruited under the auspices of the Childhood Malaria Research Group (CMRG) at the 600-bed tertiary hospital University College Hospital (UCH) in the city of Ibadan, Nigeria in west sub-Saharan Africa. Ibadan is a densely populated urban setting with a population of 2.5 million inhabitants. Ibadan has a lengthy 8 months rainy season from March to October with malaria transmission and severe disease present all year round (holoendemic). The study site is located in the UCH Ibadan Department of Paediatrics. We screen about 12,000 children attending the hospital (ill and well) for malaria parasites per year. Our studies report 11.3% SMA and 19.7% CM admissions in the parasitized children under five years of age [9] . The participants in this study were recruited during 2006 to 2009 as part of a larger prospective case-control study of childhood severe malaria currently ongoing under the auspices of the CMRG. This case-control study was divided into a Discovery Cohort consisting of those patients recruited during 2006 to 2008 and a Validation Cohort made up of those recruited in the 2008 to 2009 period. Malaria parasites were detected and counted by microscopy following Giemsa staining of thick and thin blood films [15] . Children with severe malaria were recruited on admission from the Otunba Tunwase Children's Emergency Ward (OTCHEW). Children with uncomplicated malaria were recruited as part of a daily routine malaria parasite screening at the Children's Outpatient Clinics (CHOP). Malaria-negative ill children were recruited either at admission from OTCHEW or from the Department of Paediatrics In-patient wards. Malaria-negative healthy community control children were recruited from local vaccination clinics as well as during school visits across several Ibadan districts. We recruited children aged from 6 months to 13 years using five participant definitions. The malaria-positive children, the cases, are Uncomplicated Malaria (UM), Severe Malarial Anemia (SMA) and Cerebral Malaria (CM). The malaria-negative children, the controls, are Disease Control (DC) and Community Controls (CC). We followed the WHO criteria for severe P. falciparum malaria [10] . Cerebral malaria cases were defined as children in unrousable coma for at least one hour in the presence of asexual P. falciparum parasitemia with normal cerebrospinal fluid. A Blantyre coma score less than 2 was used to assess coma status. Children with hypoglycemia were excluded from the study. Added to the Table 1 . Characteristics of discovery and validation study groups. strict clinical and laboratory definitions of CM, our study patients recover consciousness after effective antimalarial therapy. We excluded from this study those CM patients who died. Our overall mortality rate for CM is of the order of 10%. Severe malarial anemia cases were defined as conscious children with Packed Cell Volume (PCV) less than 16% in the presence of P. falciparum parasitemia. We excluded from this study those SMA patients who died. Our overall mortality rate for SMA is less than 1%. Uncomplicated malaria cases were defined as febrile children with P. falciparum parasitemia who did not require hospital admission. Our study was designed to discover and validate plasma proteome changes in dichotomous cases for which we only included those children with CM and UM with PCV greater than 20% (Table 1) . We excluded from the study blood culture positive cases. Although we did not carry out blood cultures in all severe malaria patients, the cases recruited into this study are those in whom septicemia was not suspected and who were successfully treated with antimalarial alone. The DC group consists of malaria-negative children with infectious diseases such as meningitis, otitis media, diarrhea and upper respiratory tract infections. It also includes mild to moderately anemic children and children admitted for surgery. Participants's clinical data were collected using a malariatailored questionnaire designed by the CMRG. A 2.5 ml blood sample was obtained from each participant in an EDTA blood collection tube for subsequent plasma separation. Blood samples were kept on ice and transferred to the central malaria laboratory. Plasma for this study was harvested by centrifugation (1000 g, 10 Packed cell volume (PCV) was measured using the microhaematocrit method [15] . Briefly, Blood was obtained in capillary tubes. Tubes were centrifuged at 12,000 g for 5 minutes. The percentage cell volume compared to the whole tube volume was calculated (i.e. PCV). Mean (6 standard deviation, sd), minimum and maximum PCV for each clinical group are tabulated in Table 1 . For discovery and validation cohort, these data were compared using a one-way multiple ANOVA test (p,0.05). Malaria parasites were detected and counted by microscopy following Giemsa staining of thick and thin blood films [15] . Malaria Parasite (MP) densities were calculated as follows MP/ Table 1) . The microscopic criterion for declaring a participant to be free of malaria was the absence of parasites in 100 high-power (1000X) fields. One in 10 thick blood films were randomly selected and independently reviewed by local experienced microscopists not part of the research team. Crude plasma was profiled using Surface Enhanced Laser Desorption/Ionization-Time Of Flight (SELDI-TOF) mass spectrometry. All plasma samples underwent two freeze-thaw cycles prior to analysis. Plasma samples were coded, blinded and randomized before application onto the following solid-phase fractionation surfaces (ProteinChipH arrays Bio-Rad): weak-cation exchange (CM10), strong-anion exchange (Q10) and reverse phase (H50) as previously described [16] . Liquid handling steps were automated using a Biomek 3000 Laboratory Automation Workstation (Beckman Coulter) and a 96 well BioprocessorH (Bio-Rad). Each ProteinChipH 96 well BioprocessorH included 1 quality control plasma standard derived from a single healthy individual, placed at random. Mass spectra were generated on a System 4000 Bio-Rad ProteinChipH mass spectrometer. Spectral peaks corresponding to mass/charge (m/z) clusters were detected and clustered using ProteinChipH Datamanager Client 4.1 software (BioRad). Mass spectrometer calibration was performed using Allin-1 Peptide and Protein calibrants (Bio-Rad). Reproducibility was determined by measuring the inter-ProteinChipH coefficient of variation (CV) for the quality control spectra, based on all peaks in the spectrum with intensity .1 mA. Overall interchip CV for the quality control sample was 20%, consistent with similar studies. Liquid-phase anion-exchange fractionation of plasma samples was carried out using the ProteinChipH Fractionation Kit (Bio-Rad) according to the manufacturer's instructions with a Biomek 3000 Laboratory Automation Workstation. Six fractions were obtained from each sample eluting at pH 9.0 (f1), pH 7.0 (f2), pH 5.0 (f3), pH 4.0 (f4), pH 3.0 (f5) and organic phase (f6). We selected subsets of the most relevant mass clusters in the discovery cohort groups using the weighted Kernel-based Iterative Estimation of Relevance Algorithm [17] (wKIERA) that combines a stochastic-search estimation of distribution algorithm with a kernel pattern-recognition method. We then used discovered relevant subsets of mass clusters to build discriminatory predictive models. We adopted a supervised learning approach to derive a classification rule using the Support Vector Machine (SVM) method [18] . Briefly, we used 10-fold cross validation to select parameters for the SVM. For the final model parameters, we selected those that gave the overall highest accuracy across the whole 10-fold cross validation. To obtain robust accuracy estimates for the classifier on the discovery data, we took 100 random re-samplings of the data, using 80% for training and 20% for testing. We selected as a final classifier the one that produced the highest accuracy and was then tested on the validation cohort data. Results were expressed as sensitivity, specificity and accuracy (proportion of correct classifications) and plotted on Receiver Operator Characteristic (ROC) space plots. Our multivariate statistical tests included testing against age or sex to ascertain that significant pattern changes in the proteome were not dependent on those variables in the population studied. To visualize the covariance within the mass spectral profiles we used Principal Component Analysis (PCA). PCA encapsulates the covariance within a set of variables by extracting a ranked set of independent factors or principal components. The first 3 components encompass a high proportion (,95%) of the informational content of a multivariate dataset. We plotted each patient with respect to the first 3 components, in 3-dimensional space, color-coding according to patient group. A total of 946 children participated in this study as part of the discovery and validation case-controlled cohorts. The discovery cohort comprised of 367 malaria-positive children with either Cerebral Malaria (CM), Severe Malarial Anemia (SMA) or Uncomplicated Malaria (UM), and 289 malaria-negative children who were either Disease Controls (DC) or Community Controls (CC) ( Table 1 ). The validation cohort was prospectively recruited after the discovery cohort and comprised 160 malaria-positive children with either CM, SMA or UM, and 130 malaria-negative DC or CC children (Table 1) . PCV and malaria parasite (MP) densities are presented in Table 1 . Consistent with the recruitment criteria, both discovery and validation SMA groups had PCVs below 16% (Table 1 ). There was mild anemia across CM, UM and DC groups in both cohorts, whereas CC had normal mean hematocrit (Table 1) . Parasite densities across all the infected groups were similar (Table 1) . To compare the proteome-patterns of the study groups, we fractionated plasma samples by three different chromatography procedures on solid-phase surfaces (weak-cationic and stronganionic ion-exchange, and reverse-phase) followed by Time-Of-Flight mass spectrometry. The resulting mass spectra from each of the surfaces contained a series of mass/charge ratio (m/z) peak clusters, each representing a protein of a particular mass. A set of proteins that are present, absent or at a different level in the samples defines a proteome-pattern that may discriminate between two or more of the study groups. To discover such patterns we applied statistical pattern recognition algorithms to the profiles and the selected number of discriminating proteins for each of the pairwise group comparisons is shown in Figure 1 , as the numbers in parentheses (Data S1). We built predictive models with the selected proteome-pattern for each study group comparison using a non-parametric supervised learning statistical framework. The discriminatory accuracy of these predictive models in the discovery cohort groups is shown in Figure 1a . To determine differences for malaria-positive children from healthy malaria-negative children we compared individually the plasma proteome of CM, SMA and UM groups with that of the CC group. Overall, 22 to 33 proteins composed the discriminatory patterns with accuracies above 90% across the three comparisons (Figure 1a, blue bars) . Twenty-six proteins discriminated healthy from ill (hospital admitted) malaria-negative children (CC vs. DC) with similar accuracy (Figure 1a, green bar) . To examine proteins that are specific to malaria infection we compared each of the malaria-positive groups (CM, SMA, UM) to the DC group, obtaining discrimination accuracies above 80% (Figure 1a . orange bars). Finally, to assess differences between defined malaria syndromes we compared the malaria-positive groups (Figure 1a . yellow bars). In the comparison between CM and SMA, the two major severe syndromes, the accuracy was 70% (24 proteins). Higher accuracies between 70 to 80% were observed when samples from either CM or SMA groups were compared to UM children, using 36 and 54 proteins, respectively. To validate the accuracy of the discrimination for the discovered plasma proteome-patterns, we tested the predictive models on the validation cohort groups (Figure 1b) . The best predictive model for each group comparison in the discovery cohort was asked to predict the group class in the validation cohort. We found that the predictive models obtained using the discovery cohort had similar accuracy for discrimination in the different group comparisons for the validation cohort (Figure 1b) . We compared the sensitivity and specificity of the predictive models for both discovery and validation cohort groups in ROC space and found them to be similar ( Figure 2) . We then used Principal Component Analysis (PCA) on the selected proteins to visualize the separation of patient groups. The CC group clustered tightly together (Figure 3 , green spheres). Individual malaria-positive groups showed good separation from the malaria-negative CC group (Figure 3a-c) indicating that regardless of disease severity there are significant differences in the proteomes of the groups. The heterogeneous DC group had a more dispersed cluster pattern with little overlap with the CC group (Figure 3d ). The DC group, despite being distinct, showed different degrees of overlap with the malaria-positive groups (Figure 4a-c) . Of these comparisons, the CM vs. DC patient groups showed the greatest level of cluster dispersion (Figure 4a ) indicating greater covariance in the proteins that define these groups. We then compared the malaria-positive patient groups among themselves (Figure 4d-f ). CM and SMA groups showed overlap at the cluster interface and clearer segregation at the periphery; in the comparison of both severe forms (CM and SMA) with UM we observed that the severe patient groups had compact center clusters surrounded by a more disperse cluster of the UM patient group. We simplified further the complexity of the plasma proteome by high-throughput liquid-phase anion-exchange fractionation followed by solid-phase weak-cation exchange fractionation prior to protein mass determination in the spectrometer on a subset of the samples. We assessed the discriminatory accuracy of relevant proteins obtained from each of the six anion-exchange fractions ( Figure 5 , f1 to f6) (Data S2). The reduction in the complexity of each fraction of the plasma samples resulted in a larger subset of proteins that improved discrimination between the malaria syndromes. Sets of proteins that distinguish between SMA and CM groups (Figure 5a , f1 to f6 in brackets) slightly outperformed the proteome-pattern from non-fractionated plasma. Sets of proteins differentiated the CM and UM groups with accuracies ranging from 70 to 80% (Figure 5b , f1 to f6 in brackets) and distinguished between SMA and UM with comparable accuracy (Figure 5c, f1 to f6 in brackets) . We carried out an overall analysis of plasma proteome pattern overlap by comparing the discovered sets of proteins that discriminate UM, CM, SMA (malaria-positive) and DC (malaria-negative) ill children from the malaria-negative well children CC (Figure 6 , f1 to f6). We show that each plasma fraction (f1 to f6) contains a set of proteins that clearly define both the malariapositive and malaria-negative ill children to those malaria-negative well children in the community. Furthermore, we also show that the set of proteins that discriminate SMA and CM from UM have very little overlap across the six plasma fractions (Figure 6 , f1 to f6). In the present study we carried out a large case-control study of severe childhood malaria, using a discovery cohort to define discriminatory plasma proteome-patterns and a second cohort to validate our findings, at the main tertiary hospital of the city of Ibadan, Nigeria. We show that proteome-patterns from both crude and prefractionated plasma samples accurately define childhood malaria syndromes in the discovery cohort. We confirmed these findings using a prospectively collected validation cohort. Malaria infection introduces distinguishable changes in the plasma proteome of children as seen by the striking differences between the malarianegative CC and the malaria-positive children groups. The plasma proteome differences are specific for the malaria disease process and not surrogate markers of acute illness, as we are able to accurately distinguish between malaria-negative ill children and malaria-positive groups independently of their disease severity. We have also discovered plasma proteome differences that are specific to each of the childhood malaria syndromes assessed in the present study. Our findings provide a starting point to refine the current WHO definitions of these syndromes, which lack the necessary specificity to further study severe malaria pathogenesis. We show that assessing the plasma proteome of the major malaria syndromes provides an unbiased discovery of combination of proteins that could be used to deepen our understanding of the pathogenesis of childhood malaria. This is supported by the finding that we can discriminate children with uncomplicated malaria from those with severe malarial anemia or cerebral malaria in both discovery and validation cohorts. These proteomepatterns encapsulate what changes differentiate uncomplicated malaria from the severe cases. Overall, accuracy of discrimination between the CM and SMA was lower than that in the comparison of each of these syndromes with the UM group. The degree of overlap between CM and SMA goes beyond that expected from strict application of the WHO case definitions used in this study. Nevertheless, the plasma proteome-pattern discriminated with over 70% accuracy between the severe groups. This suggests that beyond common underlying mechanisms, such as acute inflammation, there are significant differences in the pathogenesis of the severe syndromes studied. Our large cohorts allowed us to statistically validate the patternbased proteome definitions of the major childhood malaria syndromes. Although the mass spectrometry platform used in our study does not provide direct molecular identification, the chromatographic chemistry used and the mass-to-charge (m/z) ratio can be exploited to guide the identification of the set of proteins relevant for discrimination between syndromes. Plasma proteome profiling has been used to define a variety of disease states [16, [19] [20] [21] [22] [23] [24] as there is growing recognition of the advantages of using 'omics'-based methods to achieve sufficient levels of accuracy [24] . Our study showed that complex plasma protein patterns were necessary to discriminate between the different malaria syndromes. This further underlines the advantage of using unbiased high-throughput pattern recognition based methods. In many infectious diseases, there are clinically important distinctions to be made between different manifestations associated with the same underlying pathogen and malaria clinical syndromes are a clear case in point. The pathogenesis of malaria due to its erythrocytic cycle occurs in the cardiovascular system and it is plausible that proteome changes in organs such as brain, spleen, kidney and bone marrow can be reflected in the plasma proteome. Our study confirms that there are proteome changes characteristic of the clinical malarial syndromes with different level of accuracy. Furthermore, host modulation by the pathogen is likely to generate changing patterns of protein expression associated with the progression of severe malaria syndromes and our current studies are designed to address such changes. The lack of specific childhood malaria definitions has limited the progress on understanding the pathology of the major severe syndromes. To the best of our knowledge this study is the first to show that a panel of proteins, defined as a proteome-pattern, dissects clinical malaria syndromes. Further identification of the proteins that comprise the proteome-patterns will provide hints to the underlying pathogenesis on each of the syndromes. Furthermore, these proteome-patterns provide a reference point to facilitate the identification of other complex and overlapping severe childhood malaria syndromes. Data S1 Solid-phase fractionation data. (XLS) Data S2 Liquid-phase fractionation data. (XLS)

Predicting pseudoknotted structures across two RNA sequences

Motivation: Laboratory RNA structure determination is demanding and costly and thus, computational structure prediction is an important task. Single sequence methods for RNA secondary structure prediction are limited by the accuracy of the underlying folding model, if a structure is supported by a family of evolutionarily related sequences, one can be more confident that the prediction is accurate. RNA pseudoknots are functional elements, which have highly conserved structures. However, few comparative structure prediction methods can handle pseudoknots due to the computational complexity. Results: A comparative pseudoknot prediction method called DotKnot-PW is introduced based on structural comparison of secondary structure elements and H-type pseudoknot candidates. DotKnot-PW outperforms other methods from the literature on a hand-curated test set of RNA structures with experimental support. Availability: DotKnot-PW and the RNA structure test set are available at the web site http://dotknot.csse.uwa.edu.au/pw. Contact: janaspe@csse.uwa.edu.au Supplementary information: Supplementary data are available at Bioinformatics online.

Macromolecules such as DNA, RNA and proteins have the ability to form diverse tertiary structures, which enable functionality and thus, life. For many decades, proteins were deemed the global players in the cell until RNA entered the spotlight. For example, RNA structures have been found to be catalytically active, which was assumed to be the privilege of proteins. Furthermore, small RNAs are known to regulate gene expression and RNA viruses employ a plethora of structure elements to invade the host cell. To gain insight into macromolecule function, one must investigate the structure. The first step in RNA folding is stable base pairing that leads to a secondary structure. As RNA structure formation is of hierarchical nature, secondary structure is the basis for the tertiary fold that produces the functional structure. Especially for RNAs, structure determination by experimental means is an intricate and expensive task. Computational RNA structure prediction is therefore an invaluable tool for biologists. Comparative structure prediction is considered the most reliable approach for computational RNA structure prediction. Single sequence structure prediction is always limited by the accuracy of the underlying folding model. Three main streams have been identified for comparative RNA secondary structure prediction: (i) predict a structure from a pre-computed sequence alignment; (ii) simultaneously compute an alignment and a structure and (iii) alignment-free methods (Gardner and Giegerich, 2004) . Tools for multiple sequence alignments such as ClustalW (Thompson et al., 1994) are readily available and thus, structure prediction from an alignment is a tempting approach [e.g. RNAalifold (Hofacker et al., 2002) ]. Such methods heavily depend on the sequence conservation and quality of the underlying alignment. However, ncRNAs are conserved rather on the structure level than on the sequence level. The gold standard of RNA comparative structure prediction is the Sankoff approach as it does not rely on a high-quality sequence alignment and captures the structural conservation of ncRNAs. Sankoff (1985) introduced a theoretical dynamic programming algorithm for simultaneous folding and aligning for a set of N sequences that takes O(n 3N ) time and O(n 2N ) space. Practical variants have been derived which more or less retain the Sankoff principle by sacrificing optimality. Alignment-free methods aim to avoid the pragmatic restrictions made in a practical Sankoff approach as well as the reliance on a high-quality alignment [e.g. CARNAC (Perriquet et al., 2003) ]. Note that all of these comparative structure prediction methods exclude the prediction of RNA pseudoknots. RNA pseudoknots are crossing structure elements with diverse functions. The principle of pseudoknot formation is that bases within a loop region pair with complementary unpaired bases outside the loop. From an algorithmic point of view, even the simplest type of pseudoknot adds considerable computational demands due to crossing base pairs. In fact, the majority of comparative RNA structure prediction methods exclude pseudoknots. Biologists have delivered a wealth of studies, which show that pseudoknots have an astonishing number of diverse functions and occur in most classes of RNA (Staple and Butcher, 2005) . RNA viruses use pseudoknots for hijacking the replication apparatus of the host (Brierley et al., 2007) . A limited number of RNA comparative structure prediction methods can handle pseudoknots due to the computational complexity. Several of these methods take a sequence alignment as an input. ILM is an algorithm that takes as an input either individual sequences or a sequence alignment (Ruan et al., 2004) . A base pair score matrix is prepared initially and helices are added to the structure in an iterative fashion. In the approach hxmatch, a maximum weighted matching algorithm with combined thermodynamic and covariance scores is used (Witwer et al., 2004) . This program gives the option to be combined with RNAalifold. KNetFold is a machine learning method, which takes a sequence alignment as an input and outputs a consensus structure allowing pseudoknots (Bindewald and Shapiro, 2006) . Simulfold takes an alignment as an input and simultaneously calculates a structure including pseudoknots, a multiple-sequence alignment and an evolutionary tree by sampling from the joint posterior distributions (Meyer and Miklos, 2007) . Tfold combines stem stability, covariation and conservation to search for compatible stems and subsequently for pseudoknots for a set of aligned homologous sequences (Engelen and Tahi, 2010) . Several comparative structure prediction methods including pseudoknots do not rely on an initial sequence alignment. The graph-theoretical approach comRNA computes stem similarity scores and uses a maximum clique finding algorithm to find pseudoknotted structures (Ji et al., 2004) . SCARNA performs pairwise structural alignment of stem fragments with fixed lengths derived from the probability dot plot (Tabei et al., 2008) . In the following, a novel comparative approach for predicting structures including H-type pseudoknots called DotKnot-PW will be introduced. The input consists of two unaligned, evolutionarily related RNA sequences. Similarity scores between structure elements will be calculated. Statistically significant pairs will be used to find the set of conserved structure elements common to two sequences, which maximize a combined thermodynamic and similarity score. Using a hand-curated test set of pseudoknotted structures with experimental support, the prediction accuracy of DotKnot-PW will be compared with methods from the literature. Pseudoknots are functional elements in RNA structures and therefore, the most promising approach for comparative prediction is a structure comparison with less focus on exact sequence matching. In fact, perfect conservation on the sequence level can be more of a curse than a blessing. Especially ncRNAs are known to evolve quickly and so-called consistent and compensatory base pairs in both sequences will give much more confidence for structure conservation than a sequence alignment. One strong point of the DotKnot method for single sequence pseudoknot prediction (Sperschneider and Datta, 2010; Sperschneider et al., 2011) is that the set of possible H-type pseudoknot candidates (and secondary structure elements) is explicitly computed and thus readily available for further investigation. The main steps in the pairwise pseudoknot prediction approach DotKnot-PW are as follows ( Fig. 1) : (1) Run DotKnot for two unaligned sequences Seq x and Seq y . This returns secondary structure element and Htype pseudoknot candidate dictionaries. (2) Calculate pairwise base pair similarity scores for secondary structure elements and H-type pseudoknot candidates. Keep significant pairs that have a low estimated P-value. (3) Use significant pairs to calculate the set of conserved structure elements and pseudoknots for the two sequences that maximizes a combined free energy and similarity score. The key point of the DotKnot-PW approach is how to score the similarity of stems, secondary structure elements and H-type pseudoknot candidates derived from sequences Seq x and Seq y . Related work has been done for stem finding in unaligned sequences, where stem candidates are assigned a matching score across unaligned sequences, e.g. in SCARNA. Another point is how to assess the significance of a similarity score using Pvalues. These points will be explained in detail in the following section. For two unaligned RNA sequences Seq x and Seq y , the single sequence prediction method DotKnot (Sperschneider and Datta, 2010; Sperschneider et al., 2011) returns two stem dictionaries D s (x) and D s (y) derived from the probability dot plot. It also returns secondary structure element dictionaries D L s ðxÞ, D L s ðyÞ and D M s ðxÞ, D M s ðyÞ and H-type pseudoknot candidate dictionaries D p (x) and D p (y) (Fig. 1) . To detect conserved structure elements for the two sequences, a pairwise structural comparison is performed. Instead of a full structure-tostructure alignment, which takes O(n 4 ) time and O(n 3 ) space, pairwise base pair similarity scores are calculated using the RIBOSUM85-60 matrix for base pair substitutions (Klein and Eddy, 2003) . For two given stems s i (x) and s j (y) with fixed lengths in sequences Seq x and Seq y , respectively, the base pair similarity score sim[s i (x), s j (y)] is calculated using an ungapped local structure alignment of the base pairs with the RIBOSUM85-60 matrix. As an example, consider the following optimal ungapped local structure alignment of the two stems with base pair similarity score of sim[s 1 (x), s 2 (y)] ¼ 22.04 using the RIBOSUM85-60 matrix. )))))--To evaluate the significance of base pair similarity scores instead of the raw score, one has to find out what the underlying probability distribution is. Similar to the case of ungapped local sequence alignments (Karlin and Altschul, 1990) , it is assumed here that the base pair similarity scores follow an extreme value distribution. However, the main difference is that a comparison between fixed-length stem fragments is made. It is important to remember that parameters and K describe the extreme value distribution of optimal local alignment scores in the asymptotic limit of long sequences (Altschul et al., 2001) . Here, the parameters for the generalized extreme value distribution are pre-calculated using maximum Fig. 1 . For two unaligned RNA sequences Seq x and Seq y , DotKnot-PW produces structure element dictionaries derived from the probability dot plot. Similarity scores and P-values are computed to detect conserved elements likelihood fitting of a distribution to the histogram of a large sample of random base pair similarity scores. The maximum likelihood fitting was performed using the ismev package of the R statistical language for a range of stem lengths (see Supplementary Material). The P-value is defined as the probability to obtain a score greater than or equal to the observed score strictly by chance. A stem s i (x) in sequence Seq x and a stem s j (x) in sequence Seq y are a significant pair if the score sim[s i (x), s j (y)] has an estimated P-value less than . Stem pairs with a P-value larger than are not considered in the following. For two interrupted stems, the base pair similarity score is calculated by deleting bulges and internal loops and scoring stems as consecutive base pairs. Base pair similarity scores for regular and interrupted stems are also calculated if the difference in number of base pairs is less than 5. For example, a stem with one bulge might be a conserved match with a regular stem. A stem s i (x) in sequence Seq x and a stem s j (y) in sequence Seq y are a significant pair if the score sim[s i (x), s j (y)] has an estimated P-value less than . Calculating the base pair similarity score for two multiloop structures is complex due to the variety of inner loop elements, which may be regular or interrupted stems. A multiloop s M i ðxÞ can be decomposed into an outer stem s o i ðxÞ and a set of inner structure elements The base pair similarity score sim½s o i ðxÞ, s o j ðyÞ for the outer stems of two multi-loops can be easily obtained from the previously calculated base pair similarity scores. If the outer stem is a conserved match, a local alignment on the set of inner structure elements is used to find the base pair similarity score. Here, gaps are allowed in the local alignment of inner structure elements; however, no gap penalty is used. Let two sets of inner structure elements S i (x) ¼ [s 1 (x),. . ., s n (x)] and S j (y) ¼ [s 1 (y),. . ., s m (y)] be given. Let H(i, j) be the maximum similarity score between a suffix of S i (x) and a suffix of S j (y). The optimal local alignment is calculated as follows: Hði, 0Þ ¼ 0, 0 i n Hð0, jÞ ¼ 0, 0 j m Hði, jÞ ¼ max 0 Hði À 1, jÞ Hði À 1, j À 1Þ þ sim½s i ðxÞ, s j ðyÞ Hði, j À 1Þ A multiloop s M i ðxÞ in sequence Seq x and a multiloop s M j ðyÞ in sequence Seq y are a significant pair if the similarity score sim½s M i ðxÞ, s M j ðyÞ has an estimated P-value less than . A H-type pseudoknot has two pseudoknot stems S 1 and S 2 . The prerequisite for a conserved pseudoknot pair is that both core H-type pseudoknot stem pairs [S 1 (x), S 1 (y)] and [S 2 (x), S 2 (y)] are significant. The base pair similarity score for two H-type pseudoknots p i (x) and p j (y) in sequences Seq x and Seq y , respectively, is the sum of base pair similarity scores for the core pseudoknot stems as well as the base pair similarity score from a gapped local alignment of the recursive secondary structure elements in the loops (as described for multiloops). A pseudoknot p i (x) in sequence Seq x and a pseudoknot p j (y) in sequence Seq y are a significant pseudoknot pair if the similarity score sim[p i (x), p j (y)] has an estimated P-value less than . The base pair similarity score calculated in the previous sections might not be powerful enough to distinguish true positive conserved structure element pairs from false-positive structure element pairs due to the finite lengths of stems and exclusion of loop sequences in the alignment. Therefore, a dissimilarity score is also used to confirm whether a pair is significant. The dissimilarity for two given structure elements s i (x) and s j (y) in sequences Seq x and Seq y is defined as: where dissim 1 is the difference in the stem lengths and dissim 2 is the difference in the number of loop lengths. As an example, consider the pseudoknot pair p 1 (x) and p 1 (y) in sequences Seq x and Seq y , respectively, with stems S 1 , S 2 and loops L 1 , L 2 , L3. The pseudoknot pair has dissimilarity of 6. [[[[[[.) )))))) [.[[[[[[[.) ))))). . A weight is assigned to a significant pair, which is a combination of the free energy, covariation and dissimilarity. The overall weight s of a significant structure element pair [s i (x), s j (y)] in sequences Seq x and Seq y is a combination of the free energy weights w[s i (x)] and w[s j (y)], base pair similarity score sim[s i (x), s j (y)] and dissimilarity dissim[s i (x), s j (y)]: s½s i ðxÞ, s j ðyÞ ¼  sim½s i ðxÞ, s j ðyÞ À  fw½s i ðxÞ þ w½s j ðyÞg À  dissim½s i ðxÞ, s j ðyÞ Only structure element pairs with positive score s are allowed in the following dynamic programming algorithm. Here, and are set to 0.5 and is set to 1. Let p x 1 , . . . , p x n be the number of structure elements in the first sequence Seq x and p y 1 , . . . , p y m be the number of structure elements in the second sequence Seq y . Each structure element has a left and right endpoint in the sequence and is a stem, interrupted stem, multiloop or H-type pseudoknot. Structure elements can also be represented as nodes in a graph. In each sequence, the structure elements are ordered by their right endpoints. An edge is drawn between two structure elements in the first and the second sequence if their base pair similarity score has a P-value less than . Given the set of edges between nodes p x 1 , . . . , p x n and p y 1 , . . . , p y m , the goal is to find the set of edges with maximum weight that are non-crossing. This relates to finding the set of non-overlapping structure elements in the two sequences that maximize the score under the requirement that the interval ordering is preserved. A set of structure elements in the first and second sequence, which preserves the interval ordering is called a feasible structure element alignment and must satisfy the following two requirements. Each structure element can be aligned with at most one other structure element in the other sequence. The order of structure elements must be preserved with respect to the alignment. That is, if structure elements p x i and p x j in the first sequence are aligned with p y a and p y b in the second sequence, respectively, the pairs may never overlap: p x i 5p x j^p x a 5p x b (Fig. 2) . Given nodes p x 1 . . . , p x n in the first sequence Seq x and p y 1 , . . . , p y m in the second sequence Seq y , let f(i, a) be the maximum sum of edge weights for nodes between 1 and i in the first sequence and 1 and a in the second sequence such that the edges are non-crossing (i n and a m). The nodes that maximize the sum of edge weights are called an optimal structure element alignment for the two sequences. The optimal structure element alignment is calculated using dynamic programming. For a given structure element p i with start point a i and end point b i , let pre(i) be the non-overlapping predecessor. For each structure element, its predecessor is pre-computed using the sorted list of structure elements. The recursion for calculating the optimal structure element alignment is as follows: Furthermore, nested structures are taken into account for significant outer stem pairs, which have estimated P-value less than . For each significant outer hairpin loop pair, the optimal structure element alignment of inner elements is computed. For two unaligned RNA sequences Seq x and Seq y , the single sequence prediction method DotKnot returns structure element dictionaries derived from the probability dot plot. Let n and m be the number of structure elements in sequences Seq x and Seq y , respectively. Calculating the similarity scores and the optimal structure element alignment takes O(nm) time. Furthermore, nested structures are taken into account for significant outer stem pairs, which have estimated P-value less than . Let a be the number of significant stem pairs, where both stems are hairpin loops. For each significant outer hairpin loop pair, the optimal structure alignment of inner elements is computed. In the worst case, this increases time requirements to O(a  nm). The number of structure elements depends on the base composition of the sequence. Empirically, n and m can be observed to grow linearly with the length of the sequence for uniform base distribution (see Supplementary Material). In practice, DotKnot-PW can be expected to run in the order of minutes for sequences shorter than 500 nt. Many pseudoknot prediction programs have been evaluated using all the entries in the PseudoBase database (van Batenburg et al., 2000) . There are several caveats in this approach. First, the sequences given in PseudoBase are those which exactly harbor the pseudoknot. However, in practice structure prediction algorithms will be applied to longer sequences without prior knowledge of the pseudoknot location. Second, long-range pseudoknot entries appear in a truncated version in the database. Third, some classes of pseudoknots have a large number of entries (such as short H-type pseudoknots in the 3 0 -untranslated regions of plant viruses), whereas more complex types of pseudoknots only have one representative (such as long-range rRNA pseudoknots). Therefore, a hand-curated dataset of pseudoknot structures will be used here. When it comes to pseudoknots, many structures have been published based on a secondary structure predicted by free energy minimization. These predicted secondary structures are used as a working model and refined using experimental techniques such as chemical and enzymatic probing. However, the native structure remains unsolved unless tertiary structure determination methods such as X-ray crystallography are used. Testing structures that are based on computer predictions with no experimental support creates a bias in the benchmark and will be avoided in this evaluation. A total of 16 pseudoknotted reference structures from different RNA types were collected, which have strong experimental support. For each reference structure, a supporting set of 10 evolutionarily related sequences was obtained from the RFAM database (Gardner et al., 2010) . Note that for the vast majority of supporting sequences, no experimentally determined structures are available. The average pairwise sequence identities vary from 55% to 99%. Given a reference structure, the performance of prediction algorithms is evaluated in terms of sensitivity (S), i.e. the percentage of base pairs in the reference structure, which are predicted correctly, as well as positive predictive value (PPV), i.e. the percentage of predicted pairs, which are in the reference structure. The Matthews correlation coefficient (MCC) is also reported and is in the range from À1 to 1, where 1 corresponds to a perfect prediction and À1 to a prediction that is in total disagreement with the reference structure. The performance of each method for predicting the reference structure was evaluated as described in Gardner and Giegerich (2004) . DotKnot-PW was compared with methods that are freely available and use standard input and output formats. The comparative methods are CARNAC, Tfold and hxmatch (with the -A option using RNAalifold). All of these methods return structure predictions for only the reference structure with regards to the support set of evolutionarily related sequences. Tfold and hxmatch take a sequence alignment as the input. ClustalW with the default parameters was used to produce the initial sequence alignment. DotKnot-PW and CARNAC take a set of unaligned sequences as the input. Furthermore, prediction results for the reference sequence (not the supporting sequences) were obtained from the single sequence methods DotKnot (Sperschneider and Datta, 2010; Sperschneider et al., 2011), ProbKnot (Bellaousov and Mathews, 2010) , IPknot (Sato et al., 2011) and RNAfold (Hofacker et al., 1994) . Note that all methods except CARNAC and RNAfold allow pseudoknot prediction. The results are shown in Table 1 . DotKnot-PW has the highest average MCC of 0.75 for the test sequences. For each reference structure with the 10 support sequences from the corresponding RFAM family, 10 predictions are returned ordered by the combined free energy and similarity score. If only the Fig. 2 . A set of edges with positive scores is given between nodes p 1 ,. . .,p 7 in the first sequence and p 1 ,. . .,p 6 in the second sequence. The goal is to find the best set of non-overlapping structure elements in the two sequences such that the interval ordering is preserved. The optimal structure element alignment, which preserves the interval ordering includes structure elements p 1 , p 4 , p 7 in the first sequence and p 1 , p 4 , p 6 in the second sequence pairwise prediction with highest combined free energy and similarity score is taken, DotKnot-PW has an improved average MCC of 0.81. Tfold and hxmatch have average MCC of 0.6 and 0.59, respectively. CARNAC has average MCC of 0.45 with much higher average specificity than sensitivity. The prediction results for single sequence structure prediction for each of the reference sequences with experimentally determined structures are also shown in Table 1 . Note that this does not include the prediction for the support sequences from RFAM, as no experimentally determined structures are available. All single sequence pseudoknot prediction methods show improved results over using RNAfold. DotKnot has the highest average MCC of 0.76, followed by IPKnot and ProbKnot. As an example, consider the S15 mRNA pseudoknot that binds to specific proteins in the autoregulation mechanism of ribosomal protein S15 synthesis (Philippe et al., 1995) . For the reference sequence S15 and 10 support sequences from the corresponding RFAM family, DotKnot-PW returns pairwise predictions ordered by the combined free energy and similarity score. The top two pairwise predictions with the highest scores are shown in Figure 3 . We presented DotKnot-PW for prediction of structures common to two RNA sequences, including H-type pseudoknots. Both DotKnot and DotKnot-PW have been designed as dedicated pseudoknot prediction tools. In the following, important aspects of pseudoknot prediction will be discussed. Single sequence prediction methods are always limited by the underlying RNA folding model. This may be the set of free energy parameters used by free energy minimization methods or the underlying methodological framework such as maximum expected accuracy methods. DotKnot-PW shows excellent results on H-type pseudoknots with short interhelix loops. For this type of pseudoknots, DotKnot-PW uses free energy pseudoknot parameters by Chen (2006, 2009 ) based on polymer statistical mechanics. Improvements of the accuracy of free energy parameters, both for secondary structures and pseudoknots, will lead to more accurate prediction methods. However, one has to keep in mind that the algorithms themselves must be designed in such a fashion that novel parameters can be efficiently incorporated. The heuristic framework of DotKnot-PW has been designed such that it can incorporate sophisticated free energy parameters for pseudoknots, secondary structures and coaxial stacking. In the future, DotKnot-PW could also use contributions from basic tertiary structure elements such as base triples around the pseudoknot junction or stem-loop interactions. Pseudoknot prediction algorithms come in two flavors: either they can predict a certain, restricted class of pseudoknots or they do not have a restriction on the type of pseudoknot that can be predicted. For methods using free energy parameters, the inclusion of general types of pseudoknots might be more of a Each reference structure is given by its ID (see Supplementary Material for dot-bracket notation). The following column gives the method of experimental support (NMR, NMR spectroscopy; X-ray, X-ray crystallography; SC, sequence comparison; MG, mutagenesis; SP, structure probing), length of the sequence and number of pseudoknots. For each reference structure, the corresponding RFAM family ID, average sequence length and average pairwise sequence identity is shown. The * symbol means that the method failed to run. The 'first' prediction for DotKnot-PW is the pairwise prediction with highest combined free energy and similarity score. curse than a blessing, as no reliable free energy parameters for complex pseudoknots are available. DotKnot-PW has restrictions on the type of pseudoknot that can be predicted. However, this does not always lead to poor prediction results in practice. For example, DotKnot-PW shows the best result for the HDV ribozyme, which is a complex double nested pseudoknot. The results from the benchmark for structure prediction in Table 1 must be interpreted with care. First, the tested methods can be run with different parameters, possibly producing better results. However, as a typical user has no prior knowledge about the structure, the default parameters for each method are used. Of course, a comprehensive benchmark should include a larger number of structures to obtain a more reliable evaluation. However, in this study, the focus has been on a test set where the structures are supported experimentally. Many structures have been published, which were determined using computational tools and this will inevitably create a bias in a benchmark, and thus they were excluded here. Here, an extension of the single sequence prediction method DotKnot was presented based on the pairwise comparison of structure elements. This approach called DotKnot-PW is designed as an algorithm for finding the structure including H-type pseudoknots common to two sequences. As shown in Table 1 , DotKnot-PW can greatly improve structure predictions for RNA families when compared with the single sequence prediction using DotKnot. In some cases, a comparative approach might have lower sensitivity than a single sequence prediction; however, this should not generally be judged as 'inferior'. For example, ncRNAs might preserve some integral base pairs throughout evolution and only these will be detected by a comparative approach, which returns the set of base pairs common to a set of evolutionarily related sequences. DotKnot-PW uses a set of unaligned sequences as the input; therefore, no expert user intervention is required. In the future, DotKnot-PW will be extended to include intramolecular kissing hairpins. Furthermore, constrained folding will be implemented to predict a structure subject to constraints, e.g. enforce certain base pairs or regions, which must remain unpaired. DotKnot-PW has been designed as a dedicated pseudoknot prediction tool and should be applied to RNA sequences where pseudoknotted interactions are suspected in the structures. Prediction accuracy will inevitably decrease for sequences, which are longer than say 400 nt for any single sequence structure prediction method (Reeder et al., 2006) . To achieve reliable results, short sequences should be folded using DotKnot and predictions should be compared with results from other methods from the literature. To gain confidence in predictions, subsequent comparative prediction using DotKnot-PW and other comparative methods is highly recommended. Ideally, experimental verification of computationally predicted pseudoknots should be sought. Fig. 3 . Pairwise prediction results for the S15 mRNA pseudoknot (RFAM family RF00114) with the top two combined free energy and similarity scores. The reference structure is shown at the top and folds into two conformations in dynamic equilibrium: a H-type pseudoknot or a series of hairpins. For the pairwise prediction with highest score, the pseudoknot structure is returned. For the second-best pairwise prediction, the alternative hairpin loop structure is returned