content
stringlengths 0
113k
| doi
stringlengths 14
40
| pagenum
stringlengths 9
9
|
|---|---|---|
Figure 9.
Radial distribution, g(r), for oxygen-oxygen atom pairs in liquid methanol at 298K.Ren et al. Page 33
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0032 |
Figure 10.
Radial distribution, g(r), for oxygen-hydrogen atom pairs in liquid methanol at 298K.Ren et al. Page 34
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0033 |
Figure 11.
Radial distribution, g(r), for nitrogen-nitrogen atom pairs in liquid ammonia at 277K.Ren et al. Page 35
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0034 |
Figure 12.
Comparison of solvation free energies of 27 small molecules calculated with AMOEBA
force field with the experimental values. Signed average error = −0.11 kcal/mol; Unsigned
average error = 0.56 kcal/mol; RMSE = 0.69 kcal/mol.Ren et al. Page 36
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0035 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 37
Table 1
vdW parameters and atomic polarizabilities for AMOEBA atom classes.
Atom Description R0
(Å)ε
(kcal/mol)Polarizability
(Å3)
C Alkane (CH 3– or –CH 2–) 3.820 0.101 1.334
H Alkane (CH 3–) 2.960 0.024 (0.92) 0.496
H Alkane (–CH 2–) 2.980 0.024 (0.94) 0.496
C Alkane (–CH<) 3.650 0.101 1.334
H Alkane (–CH<) 2.980 0.024 (0.94) 0.496
O Hydroxyl (water, alcohol) 3.405 0.110 0.837
H Hydroxyl (water, alcohol) 2.665 0.0135 (0.91) 0.496
O Carbonyl (aldehyde, amide, acid) 3.300 0.112 0.837
H Acid (HO) 2.665 0.0150 (0.91) 0.496
C Carbonyl (aldehyde, amide, acid) 3.820 0.106 1.334
C Aromatic carbon 3.800 0.091 1.750
H Aromatic (HC) 2.980 0.026 (0.92) 0.696
N Amine nitrogen (ammonia, amine) 3.710 0.105 1.073
H Amine (HN) 2.700 0.020 (0.91) 0.496
N Amide nitrogen 3.710 0.110 1.073
H Amide (HN) 2.590 0.022 (0.90) 0.496
S Sulfur 4.005 0.355 2.800
H Sulfhydryl (HS) 2.770 0.024 (0.96) 0.496
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0036 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 38Table 2
Comparison of experimental and computed molecular polarizabilities (Å3). Experimental data are taken from tables V and VI of Applequist , et al.146
Where available, more recent experimental values for αavg from Bosque and Sales147 are reported in parentheses.
αavg αx αy αz
Methane AMOEBA 2.48 2.48 2.48 2.48
Thole 2.55 2.55 2.55 2.55
Expt 2.62 2.62 2.62 2.62
Ethane 4.25 4.66 4.05 4.05
4.46 4.93 4.24 4.24
4.48 4.99 4.22 4.22
Propane 6.01 6.75 5.78 5.51
6.29 7.18 5.98 5.68
6.38 7.66 5.74 5.74
Formaldehyde 2.44 2.77 2.55 2.01
2.54 3.07 2.70 1.86
2.45 2.76 2.76 1.83
Formamide 3.65 4.32 3.87 2.74
3.79 4.86 4.04 2.50
4.08 (4.22) 5.24 (αy + αz = 7.01)
Acetamide 5.43 6.26 5.72 4.30
5.71 6.70 6.30 4.13
5.67 6.70 (αy + αz = 10.3)
Methanol 3.19 3.61 3.02 2.93
3.35 3.92 3.13 2.99
3.32 (3.26) 4.09 3.23 2.65
Ethanol 4.94 5.44 4.84 4.54
5.08 5.76 4.98 4.50
5.26 (5.13) 6.39 4.82 4.55
Propanol 6.73 7.63 6.53 6.03
7.21 8.42 6.89 6.30
6.97 (6.96)
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0037 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 39αavg αx αy αz
NH 3 1.92 2.07 2.07 1.62
1.95 2.17 2.17 1.52
2.22
Dimethylether 4.99 5.92 4.55 4.52
5.24 6.55 4.58 4.57
5.24 6.38 4.94 4.39
Benzene 9.68 11.42 11.42 6.20
9.71 11.70 11.70 5.72
9.01 (10.44) 11.03 11.03 4.97
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0038 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 40
Table 3
Molecular polarizability (Å3) of aromatic systems. Experimental data are taken from Table VIII of
Applequist.148
αx αy αz
Benzene Expt 11.70 11.70 5.72
Expt 12.26 12.26 6.66
AMOEBA 12.30 12.30 6.64
Naphthalene Expt 20.20 18.80 10.70
Expt 22.20 18.20 7.30
AMOEBA 21.78 18.51 9.77
Anthracene Expt 35.20 25.60 15.20
Expt 44.70 25.80 9.80
DFT(B3LYP/6-31G*) 38.65 21.65 6.51
AMOEBA 32.85 24.67 12.63
Nanotube Armchair (3,3) DFT(B3LYP/6-31G*) 59.65 39.06 39.06
DFT(B3LYP/6-31+G*) 64.45 47.33 47.33
AMOEBA 61.68 41.20 41.20
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0039 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 41
Table 4
Relative conformational energies (kcal/mole) of n-butane.
AMOEBAab initioaExperimental b
anti 0.00 0.00 0.00
syn 5.61 5.50 3.95
gauche 0.51 0.62 0.67
120° 3.55 3.31 3.62
aRef 149.
bRef 150.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0040 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 42Table 5
Gas phase dimer equilibrium structure and binding energy from QM and AMOEBA.
DimerBond Dist (Å) a /Angle (degree) b Binding Energy (kcal/mol)
MP2/aug-cc-pVDZ AMOEBAMP2/aug-
cc-pVTZcBSSE
CorrectedAMOEBA
Methane-Water 3.49/86.90 3.48/84.90 −1.18 −0.92 −1.21
Methane-Methane 4.01/179.96 3.93/165.00 −0.52 −0.36 −0.53
Methanol-WDd 2.84/165.73 2.83/174.14 −6.10 −5.51 −5.85
Methanol-WAe 2.90/177.05 2.99/178.30 −5.30 −4.78 −4.77
Methanol-Methanol2.85/167.78,179.80f2.88/179.42 −6.33 −5.26 −5.66
Ethanol-WD 2.84/161.12 2.87/175.48 −6.32 −5.70 −5.65
Ethanol-WA 2.91/177.12 2.93/179.81 −5.27 −4.71 −4.67
Ethanol-Ethanol 2.86/166.46 2.89/167.63 −6.43 −5.62 −5.80
Isopropanol-WD 2.84/161.68 2.89/168.37 −6.85 −6.11 −5.87
Isopropanol-WA 2.92/177.72 2.89/176.15 −5.47 −4.85 −5.28
Dimethylether-WD 2.81/159.96 2.84/174.51 −6.68 −5.93 −6.28
Phenol-WD 3.01/163.55 2.94/167.73 −4.78 −3.98 −4.58
Phenol-WA 2.84/176.64 2.88/176.95 −7.45 −6.71 −6.32
p-Cresol-WD 3.00/162.90 2.92/167.62 −4.91 −4.14 −4.79
p-Cresol-WA 2.85/176.76 2.88/176.85 −7.28 −6.55 −6.46
H2S-WD 3.49/164.81 3.36/170.65 −3.35 −2.89 −3.48
H2S-WA 3.54/176.38 3.60/173.92 −3.04 −2.70 −2.78
H2S-H 2S 4.09/172.62 4.04/167.90 −2.18 −1.81 −2.09
Methylsulfide-WD 3.33/150.58 3.28/164.20 −4.94 −4.34 −4.84
Methylsulfide-WA 3.57/170.47 3.62/176.74 −2.73 −2.38 −2.52
Dimethylsulfide-WD 3.25/150.17 3.23/166.87 −6.07 −5.36 −5.19
Methylamine-WD 2.86/162.25 2.86/175.87 −8.09 −7.45 −8.46
Methylamine-Methylamine 3.16/153.85 3.20/157.57 −4.74 −4.14 −4.09
Ethylamine-WD 2.87/162.09 2.91/173.27 −8.17 −7.49 −7.57
Imidazole-WA 2.87/160.21 2.93/177.67 −8.34 −7.05 −7.68
Indole-WA 2.96/179.99 3.04/171.55 −6.52 −5.78 −5.58
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0041 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 43DimerBond Dist (Å) a /Angle (degree) b Binding Energy (kcal/mol)
MP2/aug-cc-pVDZ AMOEBAMP2/aug-
cc-pVTZcBSSE
CorrectedAMOEBA
Ethylsulfide-WD 3.32/151.03 3.23/165.79 −5.52 −4.84 −5.46
Ethylsulfide-WA 3.57/168.35 3.66/172.76 −2.69 −2.33 −2.10
Methylethylsulfide-WD 3.24/151.66 3.20/168.66 −6.46 −5.68 −5.78
Formamide-WD 1.91/99.52 1.86/115.00 −7.32 −6.75 −6.94
Formamide-Formamide 1.84/174.28 1,87/176.92 −16.86 −15.62 −16.00
NMA-WD 1.85/165.32 1.82/172.52 −8.71 −7.98 −8.38
aHeavy atom distance in the hydrogen bond, link O-O or O-N.
bHydrogen bond angle N(O)-H---N(O) except for methane.
cSingle point with MP2/aug-cc-pVTZ after structural optimization with MP2/aug-cc-pVDZ.
dWD denotes water as the hydrogen bond donor in dimer structure.
eWA denotes water as the hydrogen bond acceptor in the dimer structure.
fMP2/6-31+G* optimization result.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0042 |
A&A 641, A54 (2020)
van Dishoeck & Blake 1998; Tan et al. 2014), similarly to
what occurs during low-mass star formation (see e.g. Caselli
& Ceccarelli 2012; Yamamoto 2017). Hot cores, in particu-
lar, show a rich chemistry and the biggest variety of complex
organic molecules (COMs, e.g. Caselli 2005; Fontani et al.
2007; Bisschop et al. 2007; Choudhury et al. 2015; Rivilla et al.
2017a,b), which are defined as molecules with six or more atoms
including carbon (Herbst & van Dishoeck 2009). High-mass star-
forming regions are therefore a very suitable laboratory to study
astrochemistry, and particularly the formation of COMs.
COMs are expected to have an important role in prebiotic
chemistry, as keys to the formation of basic ingredients of life
such as aminoacids, sugars and lipids (Caselli & Ceccarelli 2012;
Rivilla et al. 2017a). About 70 different COMs have been iden-
tified to date in the interstellar medium (ISM) and circumstellar
shells1. Since the molecular transitions (at radio-mm as well as
FIR and sub-mm wavelengths) are sensitive to the local physical
parameters of the gas (temperature and density), the detection
of different species allows us to trace zones with different physi-
cal conditions within molecular clouds, thus getting considerable
information about the formation and destruction pathways of
COMs, and the local evolving physics of star-forming regions.
However, the mechanisms responsible for the formation of
COMs are still under debate. Two main pathways have been
proposed: (i) gas-phase chemical reactions (see e.g. Duley &
Williams 1984; Caselli 2005; Vasyunin & Herbst 2013; Balucani
et al. 2015; Skouteris et al. 2018), and (ii) surface chemistry on
the surface of interstellar dust grains (see e.g. Hasegawa et al.
1992; Ruffle & Herbst 2000; Caselli 2005; Bisschop et al. 2007;
Garrod et al. 2008; Ruaud et al. 2015). These processes are not
completely independent, but rather complementary, considering
how the evolving physical conditions during the star forma-
tion affect local chemistry. For example, the grain composition
can influence the surrounding gas-phase chemical complexity
through desorption. Hence, molecular abundances at different
phases of star formation should be strictly related (Garrod &
Herbst 2006; Garrod et al. 2008), and a trend with the evolu-
tionary stage of the sources is expected. Investigations about the
chemical evolution of star-forming regions at different evolu-
tionary stages have been conducted in several works (e.g. Doty
et al. 2002; Beuther et al. 2009; Hoq et al. 2013; Fontani et al.
2011, 2015b; Gerner et al. 2014; Choudhury et al. 2015; Colzi
et al. 2018a) using observations of selected simple or complex
molecules (and their isotopologues). These works show that the
varying physical conditions in the molecular environment during
the star-forming process can significantly affect the molecular
abundances and their emission lines strength. However, a sys-
tematic study of the evolution of COMs within the star formation
process in high-mass star-forming regions is still missing.
In this paper we present a study of four COMs in a sam-
ple of 39 high-mass star-forming regions representing different
evolutionary stages, from HMSCs to UCHIIs. The analysis has
two main goals: (i) to compare the physical parameters obtained
from the emission lines of each molecule (e.g. column density,
molecular abundance and excitation temperature), in order to
find potential correlations and links between these COMs (such
as common pathways or similar physical conditions for their for-
mation); and (ii) to evaluate the variation of their abundance with
source luminosities and evolutionary stages, in order to find if it
can be used as an evolutionary tracer within the star formation
process.
1CDMS Catalogue Oct. 2019: https://cdms.astro.uni-koeln.
de/classic/moleculesIn particular, we analyse single-dish observations of the
oxygen-bearing molecules CH 3OCHO (methyl formate, here-
after MF) and CH 3OCH 3(dimethyl ether, hereafter DE) (see
e.g. Garrod & Herbst 2006; Peeters et al. 2006; Brouillet et al.
2013; Skouteris et al. 2019), and the nitrogen-bearing molecules
NH 2CHO (formamide, hereafter F) and C2H5CN(ethyl cyanide,
hereafter EC) (see e.g. Johnson et al. 1977; Saladino et al. 2012;
Adande et al. 2013; López-Sepulcre et al. 2015, 2019; Allen et al.
2018).
Several authors have searched for correlations between
the abundances of various O-bearing and N-bearing complex
molecules, reporting different, sometimes conflicting results (see
e.g. Blake et al. 1987; Caselli et al. 1993; Fontani et al. 2007;
Bisschop et al. 2007; Suzuki et al. 2018). In particular, a chemical
link between MF and DE is suggested by both recent theoreti-
cal models (Garrod & Herbst 2006; Garrod et al. 2008; Garrod
2013; Balucani et al. 2015) and observations (e.g. Bisschop et al.
2007; Brouillet et al. 2013; Jaber et al. 2014; Rivilla et al. 2017a),
while a correlation between DE and EC is observed by Fontani
et al. (2007) in six HMCs. Bisschop et al. (2007) instead find
no correlation between DE and N-bearing species abundances.
Moreover, interferometric observations (e.g. Sutton et al. 1995;
Blake et al. 1996; Wyrowski et al. 1999; Liu 2005) suggest
that O- and N-bearing molecules trace different portions of a
molecular star-forming clump (see also Csengeri et al. 2019 and
references therein). However, several details are still unclear.
In Sect. 2 we present our sample, and in Sect. 3 we describe
the observations and the data reduction. The molecular line fit-
ting procedure through which we derived the physical parameters
is illustrated in Sect. 4. The results are reported in Sect. 5. In
Sect. 6 we present an extensive analysis of the results and dis-
cuss their potential implications, mainly focusing on correlations
among the molecules and with the evolutionary stage of the
sources. Lastly, Sect. 7 summarises the main results of this work
and draws the conclusions.
2. Source sample
Our sample consists of 39 high-mass star-forming regions,
selected to represent different evolutionary stages within the star
formation process (from HMSCs to UCHIIs), in order to eval-
uate the variation of measured molecular parameters through
different phases. These sources are part of the sample studied
by Fontani et al. (2011, 2014, 2015a,b, 2016, 2018, 2019), Colzi
et al. (2018a,b), and Mininni et al. (2018).
The sources for which we have detected at least two molec-
ular transitions of at least one of the COMs studied in this
work are 20, and are listed in Table 1. The other 19 sources are
listed in Appendix A (Table A.1). We focus the analysis only
on the sample of sources with detections. This sample covers a
wide range of distances from the Sun ( 1 9kpc), luminosities
(103 107L ) and masses (10 104M ). Sources have been
divided into four groups: 1 HMSC, 5 HMPOs, 5 Intermediate
(hereafter INTs) and 9 UCHIIs. We based our evolutionary clas-
sification on the one made by Fontani et al. (2011) for the sources
included in their paper, and on the references listed in Tables 1
and A.1 for the others. HCHII sources 18089-1732 and G75-
core, listed as HMPOs in Fontani et al. 2011, have been here
included in the INT group. We have defined the INT group to
include HCHIIs and high-mass sources in between the HMPO
and UCHII phase for which we found uncertain or discordant
classifications among different works of literature (see specific
references). From an observational point of view, in fact, it is
often difficult to clearly differentiate between these kinds of
A54, page 2 of 25 | 10.1051_0004-6361_202038212 | page_0001 |
A. Coletta et al.: Evolution of COMs in star-forming regions
Table 1. Sources with detections of at least one of the COMs studied in this work (MF, DE, F, and EC), sorted by evolutionary stage.
Source (J2000) (J2000) d L M References
(h : m : s) (:0:00) (kpc) ( L ) ( M )
HMSC
05358-mm3 05: 39: 12.5 +35 : 45 : 55 1.8 103:8101:9(1, 6, 31, 32, 33, 34, 35, 36)
HMPO
AFGL5142-MM 05: 30 : 48.0 +33 : 47 : 54 1.8 103:6101:8(1, 6, 31, 32, 33, 34, 35, 36)
18182-1433M1 18 : 21 : 09.2 14 : 31 : 49 4.5 104:0102:5(2, 3, 4, 34)
18517+0437 18 : 54 : 14.2 +04 : 41 : 41 2.9 104:1102:1(1, 6, 30, 31, 32, 33, 34, 35, 36)
I20293-MM1 20 : 31 : 12.8 +40 : 03 : 23 2.0 103:6101:6(1, 6, 30, 31, 32, 33, 35)
I23385 23 : 40 : 54.5 +61 : 10 : 28 4.9 104:2102:1(1, 6, 30, 31, 32, 33, 35)
INT
18089-1732 18 : 11 : 51.4 17 : 31 : 28 3.6 104:5102:4(1, 6, 9, 30, 31, 32, 33, 34, 35, 36)
G24.78+0.08 18 : 36 : 12.6 07 : 12 : 11 7.7 105:3103:5(2, 10, 11, 34)
G31.41+0.31 18 : 47 : 34.2 01 : 12 : 45 3.8 104:6102:9(2, 7, 8, 29, 34, 38)
20126+4104M1 20 : 14 : 25.9 +41 : 13 : 34 1.7 104:0102:9(2, 12, 13, 29, 34, 37, 39)
G75-core 20 : 21 : 44.0 +37 : 26 : 38 3.8 104:8102:1(1, 2, 5, 6, 30, 31, 32, 33, 34, 35, 36)
UCHII
W3(OH) 02 : 27 : 04.7 +61 : 52 : 25 2.0 105:0101:5(16, 20, 21, 22, 34)
G5.89-0.39 18 : 00 : 30.5 24 : 04 : 01 1.3 105:1102:3(1, 6, 29, 30, 31, 32, 33, 34, 35, 36)
G10.47+0.03 18 : 08 : 38.0 19 : 51 : 50 5.8 106:1103:2(17, 18, 19, 29, 34)
G14.33-0.65 18 : 18 : 54.8 16 : 47 : 53 2.6 104:3102:5(2, 23, 24, 34)
G29.96-0.02 18 : 46 : 03.0 02 : 39 : 22 8.9 105:8103:9(10, 15, 16, 29, 34)
G35.20-0.74 18 : 58 : 13.0 +01 : 40 : 36 2.2 104:5102:2(2, 25, 26, 27, 28, 34)
W51 19 : 23 : 43.9 +14 : 30 : 32 5.4 106:7102:8(7, 14, 34)
19410+2336 19 : 43 : 11.4 +23 : 44 : 06 2.1 104:0102:1(1, 2, 6, 30, 31, 32, 33, 34, 35, 36)
ON1 20 : 10 : 09.1 +31 : 31 : 36 2.5 104:3102:5(1, 2, 6, 30, 31, 32, 33, 34, 35, 36)
Notes. Distances from the Sun and bolometric luminosities are taken from literature (see references), while masses have been derived from the
molecular hydrogen column densities of the sources from the literature (see Table 4 and Sect. 6.3 for details).
References.(1)Fontani et al. (2011);(2)Colzi et al. (2018b);(3)Rosero et al. (2016);(4)Beuther et al. (2006);(5)Murphy et al. (2010);(6)Fontani et al.
(2015a);(7)Rivilla et al. (2017a);(8)De Buizer (2003);(9)Lackington (2011);(10)Cesaroni et al. (2017);(11)Beltrán et al. (2007);(12)Beltrán & de Wit
(2016);(13)Cesaroni et al. (1999);(14)Etoka et al. (2012);(15)Kirk et al. (2010);(16)Hoare et al. (2007);(17)Pascucci et al. (2004);(18)López-Sepulcre
et al. (2009);(19)Hatchell et al. (2000);(20)Fish & Sjouwerman (2007);(21)Mueller et al. (2002);(22)Wyrowski et al. (1997);(23)Liu et al. (2010);
(24)Walsh et al. (1997);(25)Caratti o Garatti et al. (2015);(26)Zhang et al. (2014);(27)Sánchez-Monge et al. (2013);(28)De Buizer (2006);(29)Fontani
et al. (2007);(30)Fontani et al. (2014);(31)Fontani et al. (2015b);(32)Fontani et al. (2016);(33)Fontani et al. (2018);(34)Fontani et al. (2019);(35)Colzi
et al. (2018a);(36)Mininni et al. (2018);(37)Cesaroni et al. (1997);(38)Immer et al. (2019);(39)Fontani et al. (2006).
sources, for example due to their structural complexity (see e.g.
Beuther et al. 2007).
3. Observations and data reduction
This work uses data obtained with the IRAM-30m2Telescope
(Pico Veleta, Spain) during three observing sessions: August
2014, June 2015 and December 2016.
We obtained spectra of the 39 high-mass star-forming
regions with the EMIR (Eight MIxer Receiver, Carter et al.
2012; Kramer 2016) receiver in bands E090 (E0, at 3mm), E150
(E1, at 2mm), and E330 (E3, at 0:9mm). We investigated nine
spectral windows (three at 3mm, four at 2mm, and two at
0:9mm) within the lower side band (LSB) of the receivers. Spec-
tra were obtained with two fast fourier transform spectrometers
(FTS, Klein et al. 2012) (see Kramer 1997, 2016): (i) FTS200
spectrometer (aggregate bandwidth of 8GHz), with a 195
kHz frequency resolution, corresponding to 0:2 0:7km s 1;
2IRAM-30m Documentation: http://www.iram.es/IRAMES/
mainWiki/FrontPage(ii) FTS50 spectrometer ( 1:8GHz), with 49kHz resolution
corresponding to0:1 0:2km s 1. Table 2 reports the observed
frequency ranges (or spectral windows) and the main properties
of the setups used for each one. At the observed frequencies,
the angular resolution of the telescope (half power beam width),
which can be expressed as HPBW (00)=2460=(GHz) (Kramer
2018) is2700 2900,1600 1800, and900, for the 3mm, 2mm,
and0:9mm bands, respectively. In more detail, the following
setups were employed in each observing run: (i) August 2014:
E0, E1 receivers and FTS200 spectrometer; (ii) June 2015: E0,
E1 receivers and FTS50 spectrometer; (iii) December 2016: E1,
E3 receivers and FTS50 spectrometer. Session (i) was performed
in position-switching mode, while sessions (ii) and (iii) were
done in wobbler-switching mode, with a maximum wobbler
throw of 24000.
The data were reduced using the CLASS software from
the GILDAS3package (see Pety 2005). First, we converted the
3The GILDAS software is available at: http://www.iram.fr/
IRAMFR/GILDAS
A54, page 3 of 25 | 10.1051_0004-6361_202038212 | page_0002 |
A&A 641, A54 (2020)
Table 2. Spectral windows (first column) observed with the IRAM-30m telescope, with the relative setup used.
Frequency Receiver 0 Spectrometer Vmax HPBW Tsys
(GHz) (mm) (kHz) (km s 1) (00) (K)
85.3–87.1 E0 (LSB) 3 FTS50 49 0.2
27 29 100 200 85.6–93.4 FTS200 195 0.7
88.6–90.4 FTS50 49 0.2
140.0–141.8 E1 (LSB) 2 FTS50 49 0.1
16 18 100 500141.1–148.9 FTS200 195 0.4
143.3–145.1 FTS50 49 0.1
151.8–153.6 FTS50 49 0.1
280.9–282.7 E3 (LSB) 0.9 FTS50 49 0.059 400 1000284.2–286.0 FTS50 49 0.05
Notes. In order: EMIR receiver with its wavelength, FTS spectrometer with its frequency and velocity resolutions, HPBW of the beam of the
telescope at the observed frequencies, and system temperatures for each waveband.
measured intensity, originally expressed in antenna temperature
units T
A, into main beam brightness temperature TMB, using
the relation T
A=TMBMB, whereMB=Be 4=Fe is the ratio
between the main beam efficiency and the forward efficiency
of the telescope. Then, baselines were all removed by fitting
the line-free channels with first order polynomial functions, and
subtracting them from the spectra.
4. Data analysis: molecular line fitting
Baseline subtracted spectra of the sources were exported from
CLASS to MADCUBA5(MAdrid Data CUBe Analysis, Martín
et al. 2019) to perform the molecular line fitting procedure, in
order to estimate the physical parameters of MF, DE, F, and
EC. We identified the transitions of each molecule using the
SLIM (Spectral Line Identification and LTE Modelling) tool
of MADCUBA, which searches the JPL6(Pickett et al. 1998)
and CDMS7(Müller et al. 2005) catalogues for all rotational
transitions of the molecules within the spectral windows cov-
ered by the data. In particular, the JPL catalogue was used for
MF lines, while CDMS for DE, F, and EC lines (Ilyushin et al.
2009; Endres et al. 2009; Kryvda et al. 2009; Brauer et al. 2009
and refs. therein, respectively). Molecules were considered as
clearly detected if we could identify at least two of their tran-
sitions with peak intensity TMB3(whereis the rms noise
of the spectrum). SLIM generates a synthetic spectrum of the
source, based on the assumption of local thermodynamic equi-
librium (LTE) conditions. The LTE assumption is a reasonably
good approximation for these star-forming regions, since at their
high typical densities ( n>105cm 3) the molecular energy levels
populations are thermalised. The synthetic spectrum considers
five input physical parameters: total molecular column density
(N), excitation temperature ( Tex), radial systemic velocity of the
source with respect to the local standard of rest ( VLSR), full-width
half-maximum (FWHM) of the lines, and angular size of the
4http://www.iram.es/IRAMES/mainWiki/
Iram30mEfficiencies
5MADCUBA is a software developed in the Madrid Center of Astro-
biology (INTA-CSIC) which enables to visualise and analyse single
spectra and data cubes; MADCUBA is available at: http://cab.
inta-csic.es/madcuba/MADCUBA_IMAGEJ/ImageJMadcuba.html
6Jet Propulsion Laboratory catalogue: http://spec.jpl.nasa.
gov/
7Cologne Database for Molecular Spectroscopy: https://cdms.
astro.uni-koeln.de/cdms/portalemission (). SLIM assumes that all the transitions of a certain
species have the same VLSR, FWHM, and Tex. By varying the
values of the parameters we can model the theoretical profile
of the spectrum until the best fit to the observed one is found.
The AUTOFIT function of SLIM automatically compares the
two spectra, performing a non-linear least-squares fitting via the
Levenberg-Marquardt algorithm (see Press et al. 2007; Martín
et al. 2019; Rivilla et al. 2019) to provide the optimal combi-
nation of the above-mentioned parameters with the associated
uncertainties. Other quantities like integrated intensity and opac-
ity ( ) are also computed for each detected transition. In our case,
all transitions proved to be optically thin ( 1).
In the molecular line fitting procedure, a fit including all
the three wavebands was first attempted. However, it was not
possible to properly fit all the lines simultaneously, due to the
differential attenuation caused by dust absorption at each wave-
length. This effect is further discussed in Sect. 6.1. Therefore,
for each source and molecule, we fitted the three observed wave-
bands ( 3,2, and 0:9mm) separately. The input parameters have
been left free when possible. In some cases, leaving all five
parameters free did not allow the algorithm to converge. Hence,
we fixed one or more among velocity, FWHM, (and, if nec-
essary, Tex) to the values that best reproduced the observed
spectrum. As initial guesses for the parameters, we used Tex=
100 K, the VLSRknown from the observations, and approx-
imate estimates of the source sizes ( 0, ranging from0:600
to3:800) derived from their distance ( d) assuming a diameter
D0'5000 au.
We have applied the beam dilution factor taking into account
and the frequency-dependent beam size (HPBW), thus obtain-
ing source-averaged molecular column densities. To derive the
source size for each molecule, we have left free and run AUT-
OFIT in the 2mm band, being the frequency range in which
we report the most numerous detections (see Sect. 5.1). We have
then used the obtained values to perform the fits at 3mm and 0:9
mm. The source size was left free to vary between molecules,
as they might trace different regions within the same source.
A selection of the fits of the detected molecular lines
performed in different wavebands and sources is shown in
Figs. B.1–B.4.
5. Results
In this section we present the results obtained for the sources,
listed in Table 1, which showed enough transitions to derive the
A54, page 4 of 25 | 10.1051_0004-6361_202038212 | page_0003 |
A. Coletta et al.: Evolution of COMs in star-forming regions
Table 3. Number of sources with detections per band per molecule,
within the sample of 39 sources.
0 # of sources
(mm) MF DE F EC
0.9 7 14 3 6
2 13 17 4 9
3 4 3 2 3
TOT. 13 19 5 9
physical parameters for MF, DE, F, and EC. Additional results
(see below) are available in Appendix C.
5.1. Detection summary
We have detected at least one of the four molecules in 20 of
the 39 sources. DE was found in 19 sources, MF in 13, EC
in 9, and F in 5 sources. DE has been detected at all stages
(1 HMSC, 5 HMPOs, 4 INTs, and 9 UCHIIs), while MF in 2
HMPOs, 4 INTs, and 7 UCHIIs, EC in 3 INTs and 6 UCHIIs,
and F in 4 INTs and 1 UCHII. In general, the highest num-
ber of detections has been reported in UCHII regions (45%,
9 sources), followed by HMPOs and INTs ( 25% each, 5 sources),
and HMSCs ( 5%, 1 source). This result is not affected by a
distance-induced observational bias, as the average distances of
the sources of the different evolutionary groups are consistent
(see Table 1). A possible interpretation of the distribution of
detections among the different groups is discussed in Sect. 6.4.
Table 3 shows, for each observed band, the number of sources
in which the molecules were identified. The detected rotational
transitions considering all the sources are listed in Tables D.1
(2mm band), D.2 ( 0:9mm), and D.3 ( 3mm). The highest num-
ber of detected transitions for all COMs has been reported in
the2mm band. Being also less affected by dust absorption (see
Sect. 6.1) than the 0:9mm band (the band with the second-
highest number of detections), we considered the 2mm data
to be the most reliable, and decided to take them as a refer-
ence for our analysis, for instance for the derivation of molecular
abundances, as we see in Sect. 5.5.
5.2. Molecular source sizes
Columns 2–5 of Table 4 show the source angular sizes obtained
for each molecule from the fitting procedure. Reported values
were derived at 2mm (as explained in Sect. 4), except for a few
cases (see caption of Table 4) in which the molecule was detected
only at 0:9mm. Derived values are consistent with direct
high-resolution measurements (interferometric maps) such as the
ones presented by Zhang et al. (2002) for AFGL5142-MM ( '
1 200), Rivilla et al. (2017a) for G31.41+0.31 ( 1 200), and Olmi
et al. (2003) and Beltrán et al. (2011) for G29.96-0.02 ( 200). In
addition, Col. 7 of Table 4 reports the overall ranges of the corre-
sponding linear size (diameter D) among the molecules detected
in each source. We note that the sizes obtained from different
molecules in the same source are similar, and in only two cases
they differ by a factor 2at most. This ensures that the derived
molecular column densities can be consistently compared. More-
over, the average linear sizes ( 5800 7300 au ) obtained from each
molecule considering all the sources are consistent. A more
detailed discussion of the molecular source sizes is addressed
in Sect. 6.2.3.5.3. Excitation temperatures, FWHM, and systemic velocities
Excitation temperatures ( Tex) obtained for each molecule assum-
ing LTE conditions (see Sect. 4) are shown in Table C.1. Best-fit
values in the three observed wavebands are called T1(0:9mm
band), T2(2mm), and T3(3mm). A high variability between
the three bands can be noticed in all molecules. Although one
could expect on average higher excitation temperatures at higher
frequencies (i.e. T1>T2>T3) because of the higher average
energy of the detected transitions (see e.g. Tables D.1–D.3), our
results do not show any clear trend with frequency. The small
number of transitions detected at 0:9and3mm, particularly
compared to the 2mm band, could have prevented a more accu-
rate determination of Texin those bands. For this reason, we
decided to take the more reliable T2as reference values for our
sources, rerunning the fits at 0:9and3mm with Texfixed to
T2. The values of T2cover a wide range: 20 220K for MF,
30 170K for DE,90 115K for F, and30 200K for EC.
Further considerations on excitation temperatures are made in
Sect. 6.2.3.
The FWHM of the lines (assumed to be unique for each
species in a given source, even when multiple transitions are
detected) obtained from the 2mm fits is listed for each molecule
in Table 5. Potential correlations between the FWHM of the
molecules are discussed in Sect. 6.2.3.
The other physical parameter derived from the molecular line
fitting, the LSR source velocity ( VLSR), is listed for the 2mm
waveband in Table C.2 of Appendix C.2. The derived values are
consistent with what found by other authors, such as Rivilla et al.
(2017a) for G31.41+0.31 ( VLSR'96 98km s 1), Olmi et al.
(2003) for G29.96-0.02 ( 98km s 1), and Fontani et al. (2019,
Table 1, and refs. therein) for the other sources.
5.4. Molecular column densities
Source-averaged total column densities ( N) measured for each
detected molecule are given in Tables 6–9. As done for temper-
atures, they are listed as N1(0:9mm band), N2(2mm), and
N3(3mm). We assumed the same source size and excitation
temperature (those obtained at 2mm, see Sects. 5.2 and 5.3,
respectively) for all wavebands. This made N1andN3more con-
sistent with N2, also reducing their uncertainties (Cols. 2 and
4 of Tables 6–9). Measured column densities range from 1015
to1018cm 2for MF, DE, and EC, with G31.41+0.31 (INT),
G10.47+0.03, and W51 (UCHIIs) reporting the highest values,
and from1014to1017cm 2for F, with G31.41+0.31 show-
ing the highest value. For all molecules and sources we observe
thatN3>N2>N1. This trend is discussed and explained in
Sect. 6.1 and Appendix E, where comparisons between the
column densities measured at different wavebands are made.
5.5. Molecular abundances
Molecular abundances with respect to molecular hydrogen ( H2)
have been derived from the total column densities N2(Col. 3
of Tables 6–9). Molecular hydrogen column densities ( N(H2))
and their corresponding angular sizes ( H2) were taken from
literature (see references in Table 4). All these values are beam-
averaged (H2'2000 6000). Therefore, our source-averaged
molecular column densities ( '100 300, see Table 4) have been
rescaled to the respective H2by multiplying them by the factor
(=H2)2:
N0=N2(=H2)2: (1)
A54, page 5 of 25 | 10.1051_0004-6361_202038212 | page_0004 |
A&A 641, A54 (2020)
Table 4. Best-fit values of the source angular size for each molecule ( , obtained at 2mm except when differently specified(a), see Sect. 4), and
average value .
Source (00) D N (H2)H2 Ref.
MF DE F EC (00) (103au) (cm 2) (00)
05358-mm3 2:4(a)2.4 4.3 1.1 102328 (1)
AFGL5142-MM 2.4 2.4 2.4 4.3 1:0102328 (1)
18182-1433M1 2.2 2.2 9.9 3:9102236.6 (5)
18517+0437 1.2 1.6 1.4 3.5–4.6 7:9102228 (1)
I20293-MM1 2.4 2.4 4.8 4:9102228 (1)
I23385 1.4 1.4 6.9 2:4102228 (1)
18089-1732 1:10:3 1:20:1 1.2 1.4 1.2 4.0–5.0 9:6102228 (1)
G24.78+0.08 1.9 1.6 1.2 1:10:21.5 8.5–14.6 1:4102336.6 (5)
G31.41+0.31 1:30:1 1:70:1 1.3 0:90:11.3 3.4–6.5 1:4102336.6 (5)
20126+4104M1 2.8 2.8 4.8 2:8102418 (4)
G75-core 1.2 1.0 1.1 3.8–4.6 4:4102228 (1)
W3(OH) 2.5 2.5 2.4 2.5 4.8–5.0 0:5102323 (3)
G5.89-0.39 0:90:2 1.2 1.1 1.2–1.6 5:5102328 (1)
G10.47+0.03 1.6 1.6 2.4 1.9 9.3–13.9 5:2102259 (2)
G14.33-0.65 2.0 2.0 2.0 2.0 5.2 1:3102336.6 (5)
G29.96-0.02 1.7 1.4 1.6 1.6 12.5–15.1 9:5102259 (2)
G35.20-0.74 2.4 2.4 2.4 5.3 9:8102236.6 (5)
W51 1:70:1 1:70:1 1:1(a)1:70:11.6 5.9–9.2 2:0102319 (3)
19410+2336 1:7(a)1.7 3.6 1:4102328 (1)
ON1 0.8 0.8 0.8 2.0 5:2102259 (2)
average D(103au) 6:71:1 6:00:7 5:80:8 7:31:4
Notes. Values without error come from fits performed with the parameter fixed. We also report the overall range of source linear sizes ( D)
corresponding to , and for each molecule the average Dconsidering all the sources. For each source, molecular hydrogen column densities
(N(H2)) with their angular scale ( H2) and reference are also listed. Together with , these parameters have been used to rescale the source-averaged
molecular column densities and derive the abundances (see Sect. 5.5 and Eq. (1)). Here and in the following tables, the horizontal black lines
subdivide the sources according to their evolutionary classification (see Table 1).(a)Source size obtained from the fit at 0.9 mm.
References.(1)Fontani et al. (2018);(2)Liu et al. (2010);(3)Rivilla et al. (2016);(4)Fontani et al. (2006);(5)Mininni et al. (in prep.).
By doing this, we balanced potential discrepancies between col-
umn densities corresponding to different angular scales. Then
we computed the fractional abundances of the molecules ( X)
through the formula X=N0=N(H2). The parameters used to
rescale the column densities and obtain the abundances are listed
in Table 4 for the 20 sources with detections. Sources 05358-
mm3 (HMSC) and 19410+2336 (UCHII) are included for com-
pleteness, although for these sources the only species detected,
DE, has only been detected at 0:9mm (see Table 7), so they were
not considered for the abundance calculation.
Molecular abundances derived for the 18 sources with detec-
tions in the 2mm band are listed in Table 10. We obtained
molecular abundances for 5 HMPOs, 5 INTs, and 8 UCHIIs.
In the error estimates we included the molecular column den-
sity uncertainties from AUTOFIT, and assumed a reasonable
20% error on the N(H2) values from literature. Derived frac-
tional abundances range from 10 10to10 7for MF and DE,
from10 12to10 10for F, and from10 11to10 9for
EC. G10.47+0.03 and W51 (UCHII regions) show the highest
abundances of MF, DE, and EC, whereas F is most abundant in
G31.41+0.31 ( X'10 10). The abundances of MF, DE, and EC
are consistent with the ones recently predicted for hot cores by
Bonfand et al. (2019) through chemical models. DE abundances
are also comparable to those observed by Fontani et al. (2007) in
several high-mass star-forming regions. MF and EC abundancesare consistent with those found by Allen et al. (2018) in G35.20-
0.74, while the abundances of F agree with the ones found by
Kahane et al. (2013) and López-Sepulcre et al. (2015) in several
low- to high-mass star-forming regions.
Correlations between the molecular abundances of the dif-
ferent COMs, and their behaviour during different evolutionary
stages, are investigated and discussed in Sects. 6.2.1 and 6.3,
respectively.
6. Discussion
In this section we discuss the main physical and chemical impli-
cations of the results presented in Sect. 5. The discussion mainly
focuses on MF, DE, and EC, since F presents poor statistics, hav-
ing been detected in only five sources (see Table 3) with a limited
number of transitions (see Tables D.1–D.3).
6.1. Dust absorption effect on molecular column densities
Figure 1 shows the total column density of DE (Table 7) as
a function of the observed waveband, for sources in which
the molecule was detected in more than one band. Equiva-
lent plots for MF, F, and EC are shown in Fig. E.1. A clear
trend of the derived column densities with the wavelength is
observed: All molecules show N3>N2>N1in all sources,
A54, page 6 of 25 | 10.1051_0004-6361_202038212 | page_0005 |
A. Coletta et al.: Evolution of COMs in star-forming regions
Table 5. FWHM of the lines obtained for each molecule in the 2 mm
waveband.
Source FWHM (km s 1)
MF DE F EC
AFGL5142-MM 4:80:2 4:60:3
18182-1433M1 3.0
18517+0437 6:10:9 3.0
I20293-MM1 5:40:7
I23385 3.0
18089-1732 4:10:1 3:70:1 5:20:7 5:20:2
G24.78+0.08 5.0 5:10:2 5:00:6 5.7
G31.41+0.31 5:20:1 4:80:1 9:50:3 7:20:2
20126+4104M1 8:41:2
G75-core 3:60:3 3:90:4
W3(OH) 9:60:1 7:90:5 7 :70:3
G5.89-0.39 4:10:4 8 :70:9
G10.47+0.03 9:30:3 9:40:8 10 :20:2
G14.33-0.65 1.9 4:00:3 2.6
G29.96-0.02 4.6 4:60:3 6 :50:2
G35.20-0.74 3.7 3.0
W51 6.0 6:20:1 10 :80:1
ON1 4.0 6:10:8
Notes. Values without error come from fits performed with the FWHM
parameter fixed.
Table 6. Source-averaged total column densities of MF obtained from
the fits (see Sects. 4 and 5.4) in the three observed wavebands:
N1(0.9 mm), N2(2 mm), and N3(3 mm).
Source N(MF) (cm 2)
N1 N2 N3
AFGL5142-MM (4:10:8)1015
18517+0437 (1:30:1)1016(4:41:1)1016
18089-1732 (3:80:5)1016(2:50:9)1017(6:40:8)1017
G24.78+0.08 (1:20:1)1016(1:70:2)1017(4:80:4)1017
G31.41+0.31 (1:00:1)1017(1:80:2)1018(3:90:1)1018
G75-core (1:00:1)1016
W3(OH) (2:40:3)1016(1:00:1)1017
G10.47+0.03 (1:90:1)1017(1:70:1)1018
G14.33-0.65 (1:91:5)1016(6:61:1)1016
G29.96-0.02 (3:70:9)1016
G35.20-0.74 (2:60:6)1016
W51 (4:90:1)1017(2:70:1)1018
ON1 (3:21:7)1016
with a considerable gap between 3mm and 0:9mm values
(up to2orders of magnitude). Table E.1 reports the column
density ratios N3=N2andN2=N1in the three sources where
the molecules were detected in all the three wavebands. It is
N2=N1>N3=N2in all cases, by factors of 3 5on average for
all molecules (see Appendix E.2). These significant discrepan-
cies between N1,N2, and N3cannot be due to differences in
excitation temperature or source size, since the fits were per-
formed with Tex=T2and=(2mm)for all wavebands (see
Sects. 4 and 5.3). We interpret these results as an effect of dust
opacity ( d, see e.g. Ossenkopf & Henning 1994; Chandler &
Sargent 1997; Draine 2011; Palau et al. 2014; Rivilla et al. 2017a;
De Simone et al. 2020), which causes an attenuation of theTable 7. Same as Table 6, but for DE.
Source N(DE) (cm 2)
N1 N2 N3
05358-mm3 (1:00:8)1015
AFGL5142-MM (3:20:1)1015(8:11:2)1015
18182-1433M1 (1:90:5)1016
18517+0437 (1:00:1)1016(1:60:4)1016
I20293-MM1 (7:11:4)1015
I23385 (65)1015
18089-1732 (6:00:1)1016(2:90:3)1017(5:90:5)1017
G24.78+0.08 (1:60:1)1016(1:80:2)1017(3:70:3)1017
G31.41+0.31 (5:90:2)1016(8:10:6)1017(1:50:1)1018
G75-core (1:20:1)1016(2:30:5)1016
W3(OH) (1:90:1)1016(6:80:7)1016
G5.89-0.39 (3:20:3)1016(1:30:4)1017
G10.47+0.03 (2:60:1)1017(1:80:2)1018
G14.33-0.65 (7:40:8)1016
G29.96-0.02 (1:40:2)1016(2:00:6)1017
G35.20-0.74 (1:80:4)1016
W51 (6:20:1)1017(1:80:1)1018
19410+2336 (3:10:9)1015
ON1 (3:80:2)1016(8:11:5)1016
Table 8. Same as Tables 6 and 7, but for F.
Source N(F) (cm 2)
N1 N2 N3
18089-1732 (5:00:8)1014(2:61:1)1015
G24.78+0.08 (4:10:4)1015(4:00:7)1016
G31.41+0.31 (9:51:2)1014(3:50:4)1016(1:00:1)1017
20126+4104M1 (73)1014
W51 (3:30:3)1016
Table 9. Same as Tables 6–8, but for EC.
Source N(EC) (cm 2)
N1 N2 N3
18089-1732 (2:30:1)1015(8:60:3)1015(2:30:1)1016
G24.78+0.08 (4:40:4)1015(2:01:0)1016(4:00:3)1016
G31.41+0.31 (1:40:1)1016(3:30:5)1017(1:00:1)1018
W3(OH) (1:10:1)1015(4:10:4)1015
G5.89-0.39 (7:91:7)1015
G10.47+0.03 (5:40:1)1015(4:60:1)1016
G14.33-0.65 (83)1014
G29.96-0.02 (5:90:3)1015
W51 (1:10:1)1016(7:30:5)1016
molecular emission (resulting in a lower measured line inten-
sity) of e d. The dust opacity depends on frequency according
to d/ , where is the opacity spectral index of the source.
This leads to a total column density underestimation, which
becomes more and more important as the frequency increases
(for instance, in our case, going from 3to0:9mm), so that
N2=N1>N3=N2.
These results highlight that the effect of dust absorption can-
not be neglected when studying young and dust-rich regions such
A54, page 7 of 25 | 10.1051_0004-6361_202038212 | page_0006 |
A&A 641, A54 (2020)
Table 10. Abundances with respect to H2of MF, DE, F, and EC, derived (see Sect. 5.5) from the total column densities in the 2mm band
(Tables 6–9).
Source X=N0=N(H2)
MF DE F EC
AFGL5142-MM (3:01:2)10 10(62)10 10
18182-1433M1 (1:70:8)10 9
18517+0437 (1:00:5)10 9(73)10 10
I20293-MM1 (1:10:4)10 9
I23385 (66)10 10
18089-1732 (42)10 9(5:51:7)10 9(53)10 11(2:20:5)10 10
G24.78+0.08 (3:31:1)10 9(2:40:7)10 9(3:10:9)10 11(1:30:9)10 10
G31.41+0.31 (1:70:5)10 8(1:30:3)10 8(3:21:0)10 10(1:40:5)10 9
20126+4104M1 (64)10 12
G75-core (4:41:1)10 10(73)10 10
W3(OH) (2:40:6)10 8(1:60:5)10 8(93)10 10
G5.89-0.39 (2:51:2)10 10(2:61:1)10 11
G10.47+0.03 (2:50:7)10 8(2:60:8)10 8(1:50:3)10 9
G14.33-0.65 (44)10 10(1:70:5)10 9(1:91:2)10 11
G29.96-0.02 (3:31:4)10 10(1:20:6)10 9(4:61:2)10 11
G35.20-0.74 (1:20:5)10 9(83)10 10
W51 (1:10:3)10 7(72)10 8(2:90:8)10 9
ON1 (1:10:8)10 10(2:91:1)10 10
Fig. 1. Total molecular column densities of DE as a function of the
observed waveband, in sources where the molecule was detected in
more than one band.
as massive star-forming cradles, in particular when comparing
observations at different wavelengths. This, together with the
considerations made in Sects. 4 and 5.3, brought us to concen-
trate our analysis on the 2mm data. It has to be noted nonetheless
that these data are still affected by dust opacity. An estimation of
this attenuation is made in Appendix E.2, where a more in-depth
and quantitative analysis of the dust effect on column densities,
especially on their ratios in the Table E.1 sources, is performed.
6.2. Correlations between the molecules
In this section we compare the derived physical parameters of
the different molecules and discuss possible correlations.6.2.1. Molecular abundances
Investigating relations between molecular abundances might
give us important clues about the formation processes of COMs
(see e.g. Yamamoto 2017). In Fig. 2 we compare the abundances
relative to H2of MF, DE, and EC (Table 10), derived from the
respective column densities at 2mm (see Sect. 5.5). For each
pair of tracers, we have performed a linear regression fit to the
data to check a possible correlation between the different abun-
dances. A very strong correlation emerges between each pair of
molecules (linear correlation coefficient r0:92), spanning 2–
3 orders of magnitude in abundance ( 10 10 10 7for MF and
DE,10 11 10 9for EC), which are uniformly covered by our
source sample. We also compare our results with measurements
obtained in different interstellar environments, including other
high-mass star-forming regions (HMSFRs), intermediate- and
low-mass star-forming regions (IMSFRs and hot corinos, respec-
tively), a protostellar shock region (PS shock), pre-stellar cores
(PCs), and Galactic centre (GC) clouds. Individual sources and
respective references can be found in Table F.1. These sources
agree with the correlations found in our sample, regardless of
their nature. HMSFRs, in particular, show the highest abun-
dances for all tracers, thus expanding the correlation range by
2orders of magnitude. HMSFR Sgr B2(N) N2 (Belloche et al.
2016; Bonfand et al. 2017, 2019) does not appear in the plots
including EC abundances (middle and lower panels of Fig. 2,
see also Sect. 6.3), since for this molecule its data points dif-
fer considerably from the distribution of all the other sources
(see Sect. 6.2.2) and thus they fall out of the range shown. MF
and DE (Fig. 2, upper panel) present the strongest correlation
(r=0:99) and rather similar abundances (i.e. a nearly constant
ratio) in almost all sources, denoted by the fact that the lin-
ear best-fit to the data and the x=yline nearly coincide. A
strong abundance correlation between MF and DE is also found
by Bisschop et al. (2007) in seven high-mass YSOs ( r=0:90),
A54, page 8 of 25 | 10.1051_0004-6361_202038212 | page_0007 |
A. Coletta et al.: Evolution of COMs in star-forming regions
Fig. 2. Comparison between the observed molecular abundances ( X,
Table 10) of MF and DE ( upper panel ), MF and EC ( middle panel ),
and DE and EC ( lower panel ). The sources analysed in this work are
drawn with filled coloured circles, while literature ones (see Sect. 6.2.1
and Table F.1) with empty coloured circles. All abundances are relative
toH2. Error bars are shown whenever available. The solid black line
corresponds to the linear best-fit to the data of the sources studied in this
work, while the dashed grey line to the identity. The linear correlation
coefficient between the two molecules ( r) is also given.
Brouillet et al. (2013) in Orion-KL, Jaber et al. (2014) in vari-
ous objects (including hot corinos, clouds, comets) ( 0:95), and
El-Abd et al. (2019) in the massive star-forming region
NGC 6334I. This result may suggest the existence of a tight phys-
ical or chemical link between these two molecules, which wefurther explore in Sects. 6.2.2 and 6.2.3, and thoroughly dis-
cuss in Sect. 6.4. The strong correlation we find between DE
and EC ( 0:95, Fig. 2, lower panel) is even stronger than the one
derived by Fontani et al. (2007) in six HMCs ( 0:86). It dis-
agrees instead with Bisschop et al. (2007), who find uncorrelated
abundances between DE and N-bearing species (including EC).
We also find strong correlations comparing the abundances of
MF, DE, and EC to F ( r>0:9in all cases), although these rela-
tions are less reliable due to the poor statistics (only three sources
within1–2orders of magnitudine in molecular abundance). For
MF and F, this would agree with what found by Jaber et al. (2014)
(r=0:92).
6.2.2. Molecular ratios
Molecular ratios are considered one of the main tools to investi-
gate potential chemical links between COMs (see e.g. Bisschop
et al. 2007; Fontani et al. 2007; Rivilla et al. 2017a; El-Abd et al.
2019). Table 11 shows the 2mm column density ratios MF/DE,
MF/EC, and DE/EC, in all sources for which at least two of the
species have been detected. In Fig. 3 we report the molecular
ratios derived from our analysis together with literature values
from other types of sources (see Table F.1) as a function of
source luminosity. The MF/DE ratio (Fig. 3, upper panel) is
remarkably constant within the errors, with values within 1
order of magnitude ( 0:2 3) across almost 9 orders of magnitude
in luminosity (10 2 107L ) with a rather uniform coverage,
from hot corinos to high-mass sources. Figure 4 shows the aver-
age MF/DE ratio for each type of source, extending the analysis
to a protostellar shock, GC clouds, and comets (respectively
from Lefloch et al. 2017; Requena-Torres et al. 2006; Biver &
Bockelée-Morvan 2019, see Table F.1 for details). All sources
are consistent with a constant ratio of 1, even though PCs and
comets report slightly higher average values ( 2). A nearly con-
stant MF/DE ratio of 1is also found by Rivilla et al. (2017a),
but for only six sources (hot corinos, IMSFR and HMCs) in sep-
arate limited ranges of luminosity ( 10 102and105 106L ),
and by Ospina-Zamudio et al. (2018) in seven low- to high-mass
sources. The MF/EC ratio (Fig. 3, middle panel) is nearly con-
stant (20on average) for the high-mass sources (black and blue
circles), with values within 1order of magnitude ( 4 40).
Hot corinos show instead a higher dispersion (a factor of 50)
between2and102L . Lastly, in the DE/EC ratio (Fig. 3,
lower panel) high-mass sources show a slightly greater disper-
sion (2 92, averaging30) than hot corinos ( 4 50). In the
bottom two panels of Fig. 3, the blue data points clearly deviating
from the trend of the other high-mass sources (molecular ratios
<0:5) belong to Sgr B2(N) N2, as already noted in Sect. 6.2.1.
6.2.3., FWHM, and Tex
In order to explore further similarities and correlations between
the different molecules, we also compared other physical param-
eters derived in the sample from the line fitting procedure of
Sect. 4 at 2mm, such as molecular source size ( , Table 4),
FWHM of the lines (Table 5), and excitation temperature ( T2,
Table C.1).
In agreement with what already found with abundances (see
Sect. 6.2.1), MF and DE show the strongest correlation ( r=
0:92) also in terms of the estimated angular size of the emis-
sion, as depicted in Fig. 5 (upper panel), and overall cover the
same range of (0:800 2:500). The pairs MF-EC and DE-EC show
moderate correlations instead ( 0:69and0:59, respectively). As
noted in Sect. 5.2, however, all molecules share nearly the same
A54, page 9 of 25 | 10.1051_0004-6361_202038212 | page_0008 |
A&A 641, A54 (2020)
Table 11. Relative column densities of MF, DE, and EC, using the
column densities derived at 2mm, N2(Tables 6, 7, and 9, respectively).
Source MF/DE MF/EC DE/EC
AFGL5142-MM 0:50:2
18517+0437 2:71:4
18089-1732 0:90:4 2912 335
G24.78+0.08 1:00:2 85 95
G31.41+0.31 2:30:4 5:61:3 2:50:5
G75-core 0:50:1
W3(OH) 1:50:2 243 163
G5.89-0.39 179
G10.47+0.03 1:00:2 384 405
G14.33-0.65 0:30:2 2428 9247
G29.96-0.02 0:20:1 6:31:8 3412
G35.20-0.74 1:40:7
W51 1:50:2 374 244
ON1 0:40:3
average 1:10:7 2112 3025
Notes. Average values and standard deviation considering all sources
are also given.
range of source sizes, differing by a factor of 2 at most within
the same source. It has to be noted nonetheless that high angular
resolution observations are needed to resolve potentially differ-
ent nearby emission zones within a star-forming region and infer
spatial correlations between molecules (see e.g. Mookerjea et al.
2007; Allen et al. 2017; Guzmán et al. 2018; Bøgelund et al. 2019;
Belloche et al. 2020).
For FWHM, we find MF and DE (shown in Fig. 5, mid-
dle panel) to be again the most correlated ( r=0:78), followed
by MF-EC ( 0:75) and DE-EC ( 0:63). For both MF and DE,
sources W3(OH) and G10.47+0.03 show the highest FWHMs
(8 9km s 1). Overall MF, DE, and EC share almost the same
range of linewidths ( 2 10km s 1). This agrees with what
found by Fontani et al. (2007) for DE and EC in G31.41+0.31,
G10.47+0.03, and G29.96-0.02, and by Rivilla et al. (2017a) for
MF and DE in G31.41+0.31. These results suggest, especially for
MF and DE, that these molecules could trace similar gas within
star-forming regions across different evolutionary stages.
Excitation temperatures, conversely, show no significant cor-
relations among our molecules, the only one being 0:45between
MF and DE (Fig. 5, lower panel). Moreover, temperatures and
abundances of each molecule turn out to be independent, as
observed by Fontani et al. (2007) for MF, DE, and EC. This
attests that the strong abundance correlations found in Sect. 6.2.1
are not affected by any systematic effect due to excitation tem-
perature. The temperature distributions of MF and EC peak
at higher values ( T2>150 K) than the ones of DE and F
(T2<150K). However, the overall temperature ranges are sim-
ilar among all the molecules (especially MF, DE, and EC, see
Sect. 5.3).
6.3. Evolution of molecular abundances
In this section we evaluate the variation of the derived molec-
ular abundances (Table 10) with the evolutionary stage of the
sources, in order to draw an evolutionary sequence and poten-
tially infer the most likely formation pathways for the COMs. We
report detections of COMs at 2mm in sources at three different
Fig. 3. Molecular ratios MF/DE ( upper panel ), MF/EC ( middle panel ),
and DE/EC ( lower panel ) as a function of the luminosity of the sources.
The results found in this work (Table 11, black circles) are compared
with a sample of different star-forming regions from literature (non-
black coloured circles, see Table F.1 for references). Error bars are
shown when available. The dashed black and red lines are the linear
best-fits to the data of the sources included in this work and the hot
corinos, respectively. For MF/EC, the large error bar of the lowest lumi-
nosity source of our sample (G14.33-0.65) results from the propagation
of the high uncertainties of the individual column densities.
evolutionary stages (HMPO, INT, and UCHII, see Sect. 2 and
Table 1). MF and DE have been detected at all three stages,
while EC only in INT and UCHII sources, and F only in INTs,
so the analysis mainly focuses on the first three molecules. As
A54, page 10 of 25 | 10.1051_0004-6361_202038212 | page_0009 |
A. Coletta et al.: Evolution of COMs in star-forming regions
Fig. 4. Average MF/DE ratio compared among the sources of our
sample (black) and different interstellar environments from literature
(various colours, see Table F.1). Standard deviations are shown when
available. The dashed grey line marks MF/DE =1.
it can be noted from Table 1, the three groups represent differ-
ent luminosity ranges: 103 104L the HMPOs,104 105L
the INTs, and104 107L the UCHIIs. We can interpret this
distribution on the basis of the theoretical model developed by
Molinari et al. (2008) for young massive stellar objects, predict-
ing an increase in the total luminosity of the clump during the
protostellar phase (mainly due to accretion), and a gradual stabil-
isation following the ignition of the (proto)star. However, since
luminosity can depend not only on age but also on mass, we
use the ratio L/M as an evolutionary tracer, which is expected to
increase with evolution. For the sources of our sample, the mass
was estimated assuming a spherical shape via the formula:
M=4
30BBBB@D
21CCCCA3N(H2)
Dm(H); (2)
where Dis the linear source diameter corresponding to the
angular source size (see Sect. 5.2 and Table 4), N(H2) is the
molecular hydrogen column density, m(H)=1:710 24g is the
mass of the atomic hydrogen, and =2:8(Kauffmann et al.
2008) is the mean molecular weight per hydrogen molecule.
For each source, N(H2) has been rescaled to the respective
(see Sect. 5.5). Errors on L=Mwere computed assuming a 20%
uncertainty both on luminosity and mass.
Figure 6 shows the molecular abundances of MF, DE, F,
and EC as a function of L=Mfor the sources of our sample.
An increasing abundance trend (with a similar slope of the lin-
ear fit) is evident in all molecules, spanning up to 3orders
of magnitude both in abundance and L=M. MF and DE abun-
dances nearly coincide, consistenly with the 1constant ratio
found in Sect. 6.2.2. These trends are analysed in detail in Fig. 7
for individual molecules, where we introduce the evolutionary
classification of the sources of our sample and compare our
results with a sample of various interstellar environments (see
Table F.1). We assumed as abundance uncertainty, when not
available, a conservative factor 3above and below the value,
Fig. 5. Comparison between the source angular sizes ( ,upper panel ,
listed in Table 4), the FWHM of the lines ( middle panel , Table 5), and
the excitation temperatures ( T2,lower panel , Table C.1) of MF and DE,
obtained with the molecular line fitting procedure at 2mm. The dashed
lines are the linear best-fits to the data. The linear correlation coefficient
(r) is also given. Values without error come from fits performed with the
relative parameter fixed.
in order to cover1order of magnitude in total. The full sam-
ple shows increasing abundances going from lower to higher
L=M(Fig. 7). This behaviour is mainly dominated by the total
A54, page 11 of 25 | 10.1051_0004-6361_202038212 | page_0010 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 44
Table 6
Gas phase dimer association energy (kcal/mol) and structure (Å) from ab initio and AMOEBA calculations of
multiple configurations.
ab initio QM AMOEBAStruct. RMSEf
Formamidea
A (cyc)−16.1, −16.1d, −15.96e−16.0 0.03
B (s1) −10.6 −10.3 0.05
C (np3) −8.2 −8.9 0.07
D (np1) −7.2 −7.5 0.23
E (s2) −6.9 −7.3 0.04
F (HT) −5.4 −5.5 0.08
DMFb w/o BSSE w/BSSE
A −6.95 −5.35 −5.01 0.08
B −5.82 −4.14 −5.62 0.08
C −11.41 −8.34 −7.39 0.26
D −12.11 −8.90 8.25 0.15
NMA-waterc
A −8.07 −8.34 0.14
B −8.01 −8.13 0.05
C −5.18 −5.22 0.13
NH 3a
Linear 3.03 3.20 0.14
Asymmetrical 3.07 3.21 0.17
Benzeneg
T−2.57h−2.74i −2.16 0.10
TT −2.66 NA −2.61 0.15
PD −2.49 −2.78 −2.80 0.04
S −1.51 −1.81 −2.05 0.13
aData from this study. MP2/aug-cc-pVQZ with BSSE correction.
bRef 97. MP2/aug-cc-pVTZ.
cData from this study. MP2/aug-cc-pVTZ with BSSE correction.
dRef 97. Geometry optimized at MP2/aug-cc-pVTZ level. The binding energy was reported, which was adjusted to an association energy based on
a deformation energy of 1.6 kcal/mol total.
eRef 101. CCSD(T)/CBS results for association energy.
fThis study. RMS deviations between the AMOBEA and MP2/aug-cc-pVDZ optimized dimer structures, methyl hydrogen atoms excluded.
gBenzene dimer structures: T-shaped (T), T-shaped tilted (TT), parallel displaced (PD) and parallel sandwich (S).
hRef 109. DFT-D structures and association energy from CCSD(T) 2|T 70% results.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0043 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 45
iRef 108. Estimated CCSD(T)/CBS.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0044 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 46
Table 7
Formic acid gas phase dimer energy (kcal/mol) and structure (Å) from ab initio and AMOEBA calculations.
ab initio QMa AMOEBAStruct
RMSEEass Ebind Eass
A−18.6 −15.8 15.9 0.05
B−10.3 −9.6 −9.7 0.04
C−6.7 −6.1 −6.5 0.03
D−3.4 −3.2 −3.5 0.06
E−2.4 −2.2 −2.5 0.03
F−4.4 −4.2 −4.4 0.05
aMP2/aug-cc-pVQZ energy at the MP2/aug-cc-pVDZ optimized geometry.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0045 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 47Table 8
Heat of vaporization (kcal/mol) and pressure from NVT simulations of neat liquids. E is the potential energy in kcal/mol. ΔH is the heat of vaporization in
kcal/mol, calculated as ΔH = E gas − Eliq + RT. P is pressure in Atmospheres. T is temperature in Kelvin. ρ is density in g cm−3.
Eliq EgasΔHsim ΔHexpt Psim Tρexpt
Water −9.02 0.90 10.5110.49a−61 298.20.997a
MeOH −3.57 4.90 9.068.95b−40 298.20.786b
EtOH −3.73 5.80 10.1210.11b−40 298.20.785b
n-PrOH −1.41 9.26 11.2611.31b−42 298.20.800b
i-PrOH −2.68 8.07 11.3410.88b−21 298.20.781b
NH 3 −2.89 2.17 5.545.58d 146 239.80.682d
MeNH 2 −2.89 2.67 6.096.17e 177 266.90.694f
EtNH 2 1.94 8.24 6.886.70g−46 289.70.687h
PrNH 2 4.24 11.53 7.897.47a−184 298.20.711i
DiMe amine 3.73 9.45 6.286.33j−97 280.00.671h
TriMe amine 15.88 20.95 5.625.48k−231 276.00.653h
Formic acid −16.04 −5.17 11.4611.13l 163 298.21.214b
Acetic acid −22.88 −10.91 12.5612.49c 8298.21.044m
Formaldehyde −3.47 1.41 5.385.54a,n−65 254.00.812a,n
Acetaldehyde −6.73 −1.15 6.166.09a,n 184 293.20.783a,n
DiMe ether 3.00 7.39 4.895.14a,n−62 248.30.736a,n
H2S −3.28 0.67 4.384.39o−12 220.20.934o
MeSH −2.95 2.40 5.915.87p 180 280.00.891q
EtSH 2.00 7.90 6.496.58c 109 298.10.833r
DiMeS 0.15 6.14 6.586.61b 28 298.20.842s
DiMeS 2 −7.39 1.54 9.539.18c−15 298.21.057r
MeEtS 0.98 7.97 7.587.61b 20 298.20.837r
Benzene 10.90 18.38 8.078.09t,u 96 298.00.874t,u
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0046 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 48Eliq EgasΔHsim ΔHexpt Psim Tρexpt
Toluene 3.79 12.20 9.019.09a 142 298.20.865a
Phenol −5.87 7.22 13.6813.82b−52 298.21.058b
Phenol −4.44 8.14 13.2213.36a,n 270 323.01.050a,n
Ethylbenzene 10.68 20.17 10.0810.10a 93 294.00.863a
Cresol −2.98 11.27 14.8714.77v 173 313.21.019a
Formamide −17.96 −4.00 14.5514.70w−18 298.21.129x
N-MeForm −12.57 0.54 13.7113.43n 52 298.21.005n
DiMeForm −5.86 5.04 11.4911.21a,n−129 298.20.944a,n
DiMeForm −2.36 7.28 10.3810.40y 107 373.20.873z,aa,bb
Acetamide −18.88 −4.92 14.7014.23v 6373.00.984v
Acetamide −13.87 −2.42 12.4313.30v 198 494.20.867v
N-MeAcet −13.33 −0.62 13.4413.30cc 30 373.20.894cc
DiMeAcet −8.78 2.16 11.5311.75b 349 298.20.936b
Methane −0.86 1.00 2.081.95t,u 293 111.00.424t,u
Ethane 1.81 4.86 3.423.52t,u 98 184.00.546t,u
Propane 5.24 9.21 4.434.49t,u 19 230.00.581t,u
aRef 151.
bRef 152.
cRef 153.
dRef 154.
eRef 155.
fRef 156.
gRef 157.
hRef 158.
iRef 159.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0047 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 49jRef 160.
kRef 161.
lRef 162.
mRef 163.
nRef 164.
oRef 165.
pRef 166.
qRef 167.
rRef 168.
sRef 169.
tRef 170.
uRef 171.
vRef 172.
wRef 173.
xRef 174.
yRef 175.
zRef 176.
aaRef 177.
bbRef 178.
ccRef 179.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0048 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 50Table 9
Comparison of hydrogen sulfide liquid properties from experiment165 and AMOEBA simulation results.
T (K) Eliq EgasΔHexptΔHsim Pexpt Psimρexpt
220.2 −3.28 0.67 4.39 4.38 1.441 −12 0.934
239.7 −3.00 0.73 4.19 4.20 3.326 54 0.902
252.4 −2.82 0.76 4.03 4.08 5.321 83 0.878
281.2 −2.39 0.85 3.63 3.80 12.969 163 0.816
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0049 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 51Table 10
Static dielectric constant and self-diffusion coefficient (×10−9 m2 s−1). The uncertainty of the calculated static dielectric constant is given in the
parenthesis. The uncertainty in self-diffusion constants is less than 0.1×10−9 m2 s−1.
Dielectric Constant Self-Diffusion
T (K) Expt. AMOEBA Expt. AMOEBA
Water 298.278.4a 81.0 (3.1)2.3g 2.1
Formamide 298.2105.0b
84 (293 K)c97.8 (12.3)
DMF 298.21.6h 0.9
NMA 308.2170.0c 153 (15.0)1.2i 1.0
Ammonia 240.022.0c 28.6 (3.0)5.5j 5.0
Methylamine 266.910.5d
16.7 (215 K)a15.8 (0.6)4.5k 3.8
Dimethylamine 280.06.0c 7.3 (0.5)
Trimethyamine 276.02.4a 1.9 (0.4)4.7k 4.4
Methanol 298.233.0a 38.0 (5.8)2.4l 1.9
Ethanol 298.224.3e,f 22.1 (5.6)1.1g 0.8
Acetamide 373.259 (355 K)c 52.4 (6.6)
Benzene 298.22.3a 1.1 (0.5)2.2m 2.2
aRef 151.
bRef 180.
cRef 181.
dRef 182.
eRef 183.
fRef 184.
gRef 185.
hRef 186.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0050 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 52iRef 187.
jRef 186, 188.
kRef 189.
lRef 190.
mRef 191.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0051 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 53Table 11
Comparison of experimental and AMOEBA-optimized crystal structures and cell parameters of organic molecules. The cell lengths are in Å and angles
are in degrees.
Struct
RMSE Cell a b c α β γ Ref
Formamide Expt (90K) 3×1×2 10.812 9.041 13.988 90 100.5 90 192
Calc 0.3 10.643 9.340 13.497 90 104.3 90
Acetamide Expt (23K) 1×1×1 11.513 11.513 12.883 90 90 120 193
Calc 0.1 11.564 11.564 12.289 90 90 120
Acetic Acid Expt (83K) 1×3×2 13.214 11.772 11.532 90 90 90 194, 195
Calc 0.1 13.424 11.573 11.352 90 90 90
Imidazole Expt (293K) 2×3×2 15.464 16.374 19.558 90 117.3 90 196
Calc 0.3 14.756 15.718 20.100 90 117.2 90
1H-Indole 3-carbox-aldyhydeExpt (295K) 1×2×2 14.145 11.664 17.428 90 90 90 197
Calc 0.3 14.458 11.964 16.683 90 90 90
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0052 |
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author ManuscriptRen et al. Page 54
Table 12
Hydration free energies of small molecules (kcal/mol). Statistical uncertainties of AMOEBA calculations are
given in parenthesis.
Molecule AMOEBA Expt
Methane 1.73 (0.13)1.99a
Ethane 1.73 (0.15)1.83a
Propane 1.69 (0.17)1.96a
n-Butane 1.11 (0.21)2.08a
Methanol −4.79 (0.23)−5.11a
Ethanol −4.69 (0.25)−5.00a
Propanol −4.85 (0.27)−4.83a
Isopropanol −4.21 (0.34)−4.76a
Phenol −5.05 (0.28)−6.62a
p-Cresol −5.60 (0.31)−6.14a
Methylether −2.22 (0.38)−1.90a
Benzene −1.23 (0.23)−0.87a
Toluene −1.53 (0.25)−0.89a
Ethylbenzene −0.80 (0.28)−0.80a
Methylamine −5.46 (0.25)−4.56a
Ethylamine −4.33 (0.24)−4.50a
Dimethylamine −3.04 (0.26)−4.29a
Trimethylamine −2.09 (0.24)−3.24a
Imidazole −10.25 (0.30)−9.63b
N-Methylacetamide −8.66 (0.30)−10.00c
Acetic Acid −5.63 (0.20)−6.70a
Hydrogen sulfide −0.41 (0.17)−0.44a
Methylsulfide −1.43 (0.27)−1.24a
Ethylsulfide −1.74 (0.24)−1.30a
Dimethylsulfide −1.85 (0.22)−1.54a
Methylethylsulfide −1.98 (0.32)−1.50d
Water −5.86 (0.19)−6.32d
aRef 198.
bRef 199.
cRef 200.
dRef 201.
J Chem Theory Comput . Author manuscript; available in PMC 2012 October 11. | 92136f85b4781fcfbc37a73a4fa8f2230398d2ef | page_0053 |
Single molecule fluorescence experiments determine protein
folding transition path times
Hoi Sung Chung , Kevin McHale , John M. Louis , and William A. Eaton
Laboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney
Diseases, National Institutes of Health, Bethesda, MD, 20892-0520
Abstract
The transition path is the tiny fraction of an equilibrium molecular trajectory when a transition
occurs by crossing the free-energy barrier between two states. It is a single-molecule property that
contains all the mechanistic information on how a process occurs. As a step toward observing
transition paths in protein folding we determined the average transition-path time for a fast- and a
slow-folding protein from a photon-by-photon analysis of fluorescence trajectories in single-
molecule Förster-resonance-energy-transfer experiments. While the folding rate coefficients differ
by a factor of 10,000, the transition-path times differ by less than a factor of 5, showing that a fast-
and a slow-folding protein take almost the same time to fold when folding actually happens. A
very simple model based on energy landscape theory can explain this result.
Theory predicts that folding mechanisms are heterogeneous, so that an individual unfolded
molecule can self-assemble to form its biologically-active, folded structure via many
different sequences of conformational changes (1). The distribution of these folding
pathways can now be caluclated from atomistic molecular dynamics simulations (2-6).
Information on pathway distributions from experiments must come from measurements on
single molecules, since only average properties are obtained in experiments on the large
ensemble of molecules in bulk experiments. Fig. 1 shows a schematic of a single-molecule,
equilibrium protein folding/unfolding trajectory monitored by Förster resonance energy
transfer (FRET) spectroscopy and its relation to the free energy barrier crossing between the
folded and unfolded states. The most interesting part of the trajectory is contained in what
appears to be an instantaneous jump between the two states, called the transition path, which
contains all of the information on the mechanism of folding and unfolding. The first step
toward observing transition paths in protein folding, which we report here, is the
determination of its average duration (transition-path time) for a fast-folding, all- β protein
(39-residue FBP WW domain) shown to be two-state in ensemble studies (7, 8), as well as a
markedly-reduced upper bound compared to our previous study for the 56-residue, α/β
protein GB1 (9). In contrast to a rate coefficient, which measures the frequency of a
transition, the transition-path time is the duration of a successful barrier-crossing event (Fig.
1).
The strategy employed in this study is to illuminate dye-labeled protein molecules at very
high intensities to increase the number of detected photons per transition path, to discard the
majority of photons from the less interesting segments of the trajectories between
transitions, and to analyze the transition region with a maximum likelihood method using
simple models for the transition path.
Photon trajectories were measured for immobilized WW domain and protein GB1 molecules
with donor and acceptor fluorophores attached to cysteines incorporated in the proteins (Fig.
2). In these trajectories, two properties of each photon were recorded - the color, either
donor green or acceptor red, and the absolute time of arrival to within ~ 0.5 ns. As shown in
NIH Public Access
Author Manuscript
Science . Author manuscript; available in PMC 2014 January 02.
Published in final edited form as:
Science . 2012 February 24; 335(6071): . doi:10.1126/science.1215768.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0000 |
Figs. 3A and 3B, transitions between states are clearly resolved in the binned fluorescence
and photon trajectories, and the FRET efficiency distributions (Figs. 3C and 3D) are
bimodal, indicating the presence of two states. The photon trajectories were extracted from
the region near the transitions and analyzed using the Gopich-Szabo maximum likelihood
method (10).
For a given model, the Gopich-Szabo method calculates the parameters of a kinetic model
that can most accurately reproduce the photon trajectories (Fig 3). We adopt a one-step
model for the transition path, which may be viewed as the simplest discrete representation of
how the FRET efficiency changes along the path. This picture can be represented in a
kinetic model for a two-state system with a finite transition path by introducing a third
virtual state, S, for which the FRET efficiency is midway between the folded and unfolded
states (ES = (EF + EU)/2). In this model the lifetime of S ( τS) corresponds to the average
transition-path time, 〈tTP〉 (Fig. 4A). S has the property of a transition state, since the rate
coefficients from S to F and S to U ( kS) are the same and therefore the p-fold = ½.
The likelihood function for the jth photon trajectory is (10):
(1)
Here, K is the rate matrix (Eq. S6 (11)) containing the 3 rate coefficients ( kF’, kU’, kS), N is
the number of photons in the jth trajectory, ci is the color of the ith photon (donor or
acceptor), and τi is a time interval between the ith and (i-1)th photons as shown in Fig. S4B
(11). The photon color matrix F depends on the color of a photon as F(acceptor) = E and
F(donor) = I − E, where E is a diagonal matrix with elements that are FRET efficiencies of
the 3 states (F, S and U). n is a diagonal matrix with elements that are photon count rates of
the three states. vini and vfin are vectors that describe the state (folded or unfolded) at the
beginning and the end of the trajectory. Practically, log-likelihood functions were calculated
and the total log likelihood function of all trajectories were calculated by summing the log-
likelihood functions (
) of individual trajectories that contain a transition
between folded and unfolded states. In the likelihood function L, τS is the only variable
parameter (11).
The difference of the log-likelihood functions, ΔlnL, = lnL(τ) − lnL(0), as a function of τS is
plotted in Fig. 4B for the WW domain. This function was calculated from 527 transitions
between the folded and unfolded states. In this plot, the likelihood at τS = 0,
is the value for a two-state model where every transition between folded
and unfolded states is instantaneous, i.e. it occurs faster than the shortest photon interval.
Therefore, the plot displays how much better (or worse) a two-state model with a finite
transition-path time describes the photon trajectories than a two-state model with an
instantaneous transition. There is a highly significant peak in the likelihood function in Fig.
4B at 16 (±3) μs. (The error is the standard deviation obtained from the curvature of the
peak.) Simulations of photon trajectories show that if ΔlnL at the peak is higher than a
certain confidence level, the value of τS at the peak corresponds to the assumed τS and does
not arise from statistical fluctuations (Fig. S6 (11)). We use a confidence level that satisfies
a condition L(τS)/(L(τS) + L(0)) = 0.95, which assures 95% confidence in the significance of
the maximum and corresponds to Δln L ≈ 3 (the dashed horizontal lines in Fig. 4). The
value of 16 μs at Δln L = 7.8 is therefore a well-determined quantity, and corresponds in our
model (Fig. 4A) to the average transition-path time, 〈tTP〉. Fig S5 (11) shows that 〈tTP〉 is theChung et al. Page 2
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0001 |
same for folding and unfolding transitions, which is consistent with the requirement of
microscopic reversibility that 〈tTP〉 for a barrier crossing be the same in both directions (12).
To extrapolate the value of 〈tTP〉 to the viscosity in the absence of glycerol, we determined
the rate coefficients at different viscosities (see Table S1 (11)). Using a linear free energy
relation to account for the change in stability due to both the glycerol and GdmCl, we find
that the rate coefficients for folding and unfolding depend inversely on the first power of the
viscosity (11), so 〈tTP〉 should scale the same way (see Eqs. 2 and 3 below). Since the
viscosity of 3 M GdmCl/50% glycerol solution is found to be 10 times higher than that of 2
M GdmCl (11), our best estimate of 〈tTP〉 in the absence of a viscogen at 293 K is ~ 2 μs.
We have employed the simplest possible model for determining 〈tTP〉. However, more
realistic models that depict a more gradual change in the FRET efficiency along a transition
path, with 2 and 3 steps in the FRET efficiency in the transition path between states instead
of just one (Fig. 4A), yield very similar values for 〈tTP〉 (see Fig. S9 (11)). We also found
that the value of 〈tTP〉 is not sensitive to the choice of the FRET efficiency for S as long as
the value is between the two FRET efficiencies of the folded and unfolded states (0.6 ≤ ES ≤
0.7, Fig. S7 (11)).
For proteins with very low free energy barriers, it may be possible to estimate 〈tTP〉 from
ensemble measurements. Gruebele and coworkers have studied the kinetics of the ultrafast-
folding, 35-residue FiP35 WW domain, which has a very similar structure to our WW
domain (FBP), but a non-homologous amino-acid sequence (13). Prior to the ~ 10 μs
folding/unfolding relaxation at the melting temperature of ~ 350 K, a ~ 1.5 μs relaxation was
observed, which was called a “molecular phase” and attributed to a change in the small
population of molecules at the top of a low free energy barrier in response to the temperature
jump (importantly, no molecular phase has been observed for FBP WW domain, presumably
because it is a slower folder due to a higher barrier, and there is therefore no detectable
amplitude from the change in the barrier top population (7)). In this interpretation, the ~ 1.5
μs relaxation corresponds to the lifetime, τS, of our kinetic model for the transition path
(Fig. 4).
Shaw and coworkers have simulated equilibrium trajectories of the FiP35 WW domain
using all-atom molecular dynamics calculations (4). They found 〈tTP〉 to be 0.5 (±0.1) μs
360 K using the TIP3P explicit water model (6). After rescaling for the difference in
viscosity compared to real water, the simulated 〈tTP〉 becomes ~1.5 μs (14). Although the
sequences for the 2 WW domains are different, the finding of similar values for 〈tTP〉 from
the simulations and both ensemble and single molecule experiments provides support for the
accuracy of the simulations, Gruebele’s interpretation of the molecular phase, and our
interpretation of the single-molecule photon trajectories.
The folding time of protein GB1 in 4 M urea is ~ 1 s. This time is far too long to observe
folding transitions in trajectories simulated by atomistic equilibrium molecular dynamics,
making even an upper bound for the transition-path time an interesting quantity. In previous
work, we were able to determine an upper bound of ~ 200 μs, based on an analysis of
individual trajectories. The photon count rate in those experiments was only 50 ms−1, and
the average time before photo-bleaching was ~ 100 ms. In the current experiments the much
higher count rate of 350 ms−1 from the increased illumination intensity, together with the
collective analysis using the maximum likelihood method, has allowed us to determine a
much more accurate upper bound.
The penalty for the higher photon count rate is that the lifetime of the trajectories is
shortened to ~ 10 ms by the more intense illumination, and transitions, albeit clearlyChung et al. Page 3
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0002 |
resolved (Fig. 3B), are only observed in a very small fraction of the trajectories.
Measurement at 4 M urea (with no added glycerol) of trajectories for ~ 47,000 molecules
yielded just 114 transitions.
These 114 transitions were analyzed using the same model as for the WW domain. No peak
is observed in the Δln L vs τS plot (Fig. 4C), so 〈tTP〉 is too short to measure. Nevertheless
the analysis permits a determination of an upper bound for 〈tTP〉. By analogy to the
significance of the peak for the WW domain, we can set a confidence level for the answer to
the question: how long can 〈tTP〉 be before it becomes inconsistent with the data? The 95%
confidence level that τS in a two-state model with a finite transition path is less consistent
with the photon trajectories than a two-state model with an instantaneous transition path is
given by its value at Δln L ≈ −3. In other words, 〈tTP〉 cannot be longer than τS at Δln L =
−3, and is therefore an upper bound on 〈tTP〉. As shown in Fig. 4C, this upper bound is ~ 10
μs.
The major result of our experiments is that while the folding rate coefficients for the WW
domain and protein GB1 differ by 4 orders of magnitude, 104 s−1 and 1 s−1, the transition-
path times differ by less than 5-fold (~ 2 μs and <10 μs), showing that a fast- and a slow-
folding protein take almost the same time to fold when folding actually happens.
Interestingly, a very simple model by A. Szabo, based on describing the kinetics of folding
for a two-state system as diffusion over a barrier on a one-dimensional free energy surface
as in the energy landscape theory of Wolynes, Onuchic and coworkers (1, 15), can explain
this result. According to Kramers theory for such a barrier crossing (Fig. 1A), the folding
time (τF = 1/kF) is given by:
(2)
where D* is the diffusion coefficient at the barrier top, ω2 is the curvature of the unfolded
well (near x0 in Fig. 1A), -( ω*)2 is the curvature at the barrier top, β = 1/kBT, and ΔGB* is
the height of the folding free-energy barrier (16-20). For ω = ω* 〈tTP〉 is approximately
given by (9, 12):
(3)
The model predicts that 〈tTP〉 is insensitive to the barrier height and that fast and slow-
folding proteins will have similar transition-path times as long as there are only small
differences in the curvatures and the diffusion coefficients (i.e. small difference in τ0). The
diffusion coefficient depends on the roughness of the underlying energy landscape and could
therefore differ substantially among proteins (21-23). The best current estimate for τ0 of
fast-folding proteins is ~ 1 μs (24), which predicts a ratio of 〈tTP〉 for protein GB1 and the
WW domain of 1.4, compared to the experimental ratio of < 5, assuming the same τ0 for the
two proteins. This ratio varies from 1.3 to 1.8 for τ0 between 0.1 μs to 10 μs.
Our determination of an average transition-path time is a first step towards the goal of
obtaining information on the distribution of folding pathways from measurements of inter-
dye distance versus time trajectories during transition paths. However, the result of this first
step by itself has turned out to be extremely interesting. Folding involves a complex and
intricate rearrangement of a polypeptide chain to form a unique structure, yet the time forChung et al. Page 4
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0003 |
this non-trivial self-assembly process is almost the same for two proteins with different
topologies and vastly different folding rates.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
We thank Irina Gopich, Attila Szabo and Gerhard Hummer for numerous helpful discussions, and Annie Aniana for
technical assistance in the expression and purification of proteins. This work was supported by the Intramural
Research Program of the NIDDK, NIH.
References
1. Bryngelson JD, Onuchic JN, Socci ND, Wolynes PG. Proteins-Struct. Funct. Gen. 1995; 21:167.
2. Noe F, Schutte C, Vanden-Eijnden E, Reich L, Weikl TR. Proc. Natl. Acad. Sci. USA. 2009;
106:19011. [PubMed: 19887634]
3. Bowman GR, Pande VS. Proc. Natl. Acad. Sci. USA. 2010; 107:10890. [PubMed: 20534497]
4. Shaw DE, et al. Science. 2010; 330:341. [PubMed: 20947758]
5. Piana S, Lindorff-Larsen K, Shaw DE. Biophys. J. 2011; 100:L47. [PubMed: 21539772]
6. Lindorff-Larsen K, Piana S, Dror RO, Shaw DE. Science. 334:517. [PubMed: 22034434]
7. Nguyen H, Jager M, Moretto A, Gruebele M, Kelly JW. Proc. Natl. Acad. Sci. USA. 2003;
100:3948. [PubMed: 12651955]
8. Petrovich M, Jonsson AL, Ferguson N, Daggett V, Fersht AR. J. Mol. Biol. 2006; 360:865.
[PubMed: 16784750]
9. Chung HS, Louis JM, Eaton WA. Proc. Natl. Acad. Sci. USA. 2009; 106:11837. [PubMed:
19584244]
10. Gopich IV, Szabo A. J. Phys. Chem. B. 2009; 113:10965. [PubMed: 19588948]
11. SI, Information on materials and methods is available as supporting online material on Science
Online.
12. Hummer G. J. Chem. Phys. 2004; 120:516. [PubMed: 15267886]
13. Liu F, Nakaema M, Gruebele M. J. Chem. Phys. 2009; 131:195101. [PubMed: 19929078]
14. Yeh IC, Hummer G. J. Phys. Chem. B. 2004; 108:15873.
15. Socci ND, Onuchic JN, Wolynes PG. J. Chem. Phys. 1996; 104:5860.
16. In the case of protein GB1 there is the possibility of a sparsely populated intermediate between the
folded and unfolded states (17-20). In this study we have implicitly defined the transition-path
time for both the WW domain and protein GB1 in terms of just the two deep minima of the folded
and unfolded states.
17. Park SH, Shastry MCR, Roder H. Nat. Struct. Biol. 1999; 6:943. [PubMed: 10504729]
18. McCallister EL, Alm E, Baker D. Nat. Struct. Biol. 2000; 7:669. [PubMed: 10932252]
19. Morrone A, et al. Biophys. J. 2011; 101:2053. [PubMed: 22004760]
20. Krantz BA, Mayne L, Rumbley J, Englander SW, Sosnick TR. J. Mol. Biol. 2002; 324:359.
[PubMed: 12441113]
21. Clarke and coworkers (22) have found, for example, domains with similar structures and stability
that have folding rates which differ by ~3,000-fold. The slower-folding domains show very little
dependence on solvent viscosity, suggesting a large internal friction and therefore a much smaller
D* (22, 23).
22. Wensley BG, et al. Nature. 2010; 463:685. [PubMed: 20130652]
23. Cellmer T, Henry ER, Hofrichter J, Eaton WA. Proc. Natl. Acad. Sci. USA. 2008; 105:18320.
[PubMed: 19020085]
24. Kubelka J, Hofrichter J, Eaton WA. Curr. Opin. Struct. Biol. 2004; 14:76. [PubMed: 15102453]
25. Best RB, Hummer G. Proc. Natl. Acad. Sci. U.S.A. 2005; 102:6732. [PubMed: 15814618]Chung et al. Page 5
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0004 |
26. Kubelka J, Henry ER, Cellmer T, Hofrichter J, Eaton WA. Proc. Natl. Acad. Sci. USA. 2008;
105:18655. [PubMed: 19033473]Chung et al. Page 6
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0005 |
Fig. 1.
Schematic of a folding transition path for a two-state protein. ( A) The kinetics of protein
folding is described by energy landscape theory as diffusion on a one-dimensional free
energy surface with an order parameter ( x) as a reaction coordinate (1, 15, 25, 26). The
unfolded molecule spends the vast majority of time visiting a large number of conformations
in the free energy well of the unfolded state. A transition path is the part of the trajectory
that crosses the reaction coordinate x at x0 and reaches x1 on the other side of the barrier
without re-crossing x0 (12). The length of this trajectory is the transition-path time. ( B)
FRET efficiency trajectory. In the typical experiment the donor and acceptor FRET
fluorophores are attached to cysteine residues which are closer on average in the folded state
(higher FRET efficiency) than in the unfolded state (lower FRET efficiency). The duration
of the jump in the FRET efficiency trajectory is the transition-path time. The FRET
efficiency monitors reconfiguration of the polypeptide backbone to form the native fold, but
is most probably blind to annealing of side chains. Consequently, the transition path
measured by FRET is expected to be shorter than the transition path monitored by side chain
contacts, for example in a molecular dynamics simulation (6).Chung et al. Page 7
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0006 |
Fig. 2.
Schematic of immobilized folded proteins showing donor (green-emitting) and acceptor
(red-emitting) fluorophores. The proteins are attached to a polyethyleneglycol-coated glass
surface via a biotin-streptavidin-biotin linkage (11).Chung et al. Page 8
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0007 |
Fig. 3.
Representative fluorescence and photon trajectories, and FRET efficiency histograms of
WW domain and protein GB1. ( A and B) For the fluorescence trajectories - donor (green)
and acceptor (red) - photons were collected in 50 μs bins for the WW domain and in 100 μs
bins for protein GB1. Measurements were made at 293 K at high illumination intensity in 3
M GdmCl/50% glycerol for the WW domain ( A) (20 kW/cm2, ~ 650 photons/ms) and in 4
M urea for protein GB1 ( B) (10 kW/cm2, ~ 350 photons/ms). Strings of arrival times of
donor and acceptor photons (photon trajectories) in the transition region (the 80 μs yellow
shaded regions) are displayed below the binned fluorescence trajectories. Dashed vertical
lines in the photon trajectories indicate the most probable transition interval found by the
Viterbi algorithm (11). The absolute times refer to the start of data collection, which began
at ~ 100 ms before the laser was turned on. ( C and D) FRET efficiency histograms. The
mean FRET efficiencies for the WW domain were calculated for each of 50 μs bins for the
trajectories with the mean photon count rate > 400 ms−1 (C) and for folded and unfolded
segments of protein GB1 containing ~ 2500 photons ( D).Chung et al. Page 9
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0008 |
Fig. 4.
Determination of average transition-path times in a kinetic model. ( A) Schematic of a FRET
efficiency trajectory using a one-step model to describe the transition path from unfolded
(U) to folded (F) states for a protein exhibiting two-state kinetics and thermodynamics. The
average transition-path time, 〈tTP〉, is equal to the lifetime of a virtual intermediate state S
(τS = (2kS)−1 ). (B and C) The difference of the log likelihood, ΔlnL = lnL(τS) − lnL(0),
between the two-state model with a finite transition-path time and a two-state model with an
instantaneous transition-path time is plotted as a function of τS for the WW domain in 3 M
GdmCl/50% glycerol ( B) and protein GB1 in 4 M urea ( C). The horizontal dashed line at
ΔlnL = +3 represents the 95% confidence limit for the significance of the peak in B and the
intersection of the likelihood function with the horizontal dashed line at ΔlnL = −3 in C
yields the 95% confidence limit for the upper bound of τS.Chung et al. Page 10
Science . Author manuscript; available in PMC 2014 January 02.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 3758e75bdacf0bb9b3b1c65cf464d4a825104d37 | page_0009 |
Y,ty18is,1901.] EPITME(WF..URRiNT MRDICAL LITERATURE.
ANEPITOME
CURRENT'MEDICALLITERATURE.
MEDICINE.
(322)Acute Pulmonary Complications in
Malaria.
CRESPIN ANDMAILFERT haveinvesti-gatedsomeofthelesionspresent inthe'lungsincasesofmalaria, moreespecially;thebroncho-pneumonic (Arch.GMn.deM4d.,April,o90i).Theyfindthatacutebronchitis isfrequent inmanycasesofmalaria andthatthebronchial lesionismostmarked inthebases,especially ontheleftside.Thebronchial complica-tionseemstobeinproportion to'theamount oflesionpresent inthespleenandliver.Pulmonary congestion isalsocommon, butvariesverymuchinitsintensity indifferent cases.Pneumoniaisalsopresent inseveral cases,andtheprognosis isextremely uncertain, deathtakingplaceinalargenumber. Pneu-moniamaycomeonafterthemalarialattack, andsupervening incasesofmalaria oflongstanding isveryserious.Thetemperature inmalarial pneumoniaisdeceptive, andinmanycasesofpostmalaria theremaynotbeaveryhigh
degreeofpyrexia. Infactthewritersdescribe anapyrexic pneumonia ofaveryserious type.Pneumonia isduetothe
pneumococcus, nottothehaematozoa.Thereistherefore nothing specific in
thisformofpneumonia. Therelationofthislattertothepneumonic processisnotquiteeasytounderstand. Itis
suggested bythewriters thatitsten-dencytocausecongestion createsapre-disposition toapneumococcal infection.Thewriters alsodrawattention tothefact,whichhasbeenseveraltimesnoted,thatinmalaria theremaybeanapicalconsolidation simulating phthisis in
manyofitscharacteristics. Acuriouspointabouttheapicalconsolidation metwithinmalarial casesisitstendency to
passtotheopposite apexwithextremerapidity. Intwenty-four hoursthe
upperportion ofonelungmayhavecom-pletelyconsolidated. Thetreatment ofallthesecomplications isthefreeex-hibition ofquinine. Thisdoesnot,how-ever,mean,according totheauthors, thatthepulmonary condition isaspecificone.Theactionofquinine iswidespreadandseemstoreducecongestion intheorgansgenerally, andtherefore favourstheresolution ofthepneumonic process.
(323)Intermittent Tension oftheEpigas-
trium.
BOUVERET (LyonMMd.,March 31St,I90I)callsattention toanearlysignofpyloricobstruction. Thepatient beingrecumbent, withabdominal muscles re-lax'ed,theupperhalfoftheepigastriummaybeseenoncloseexamination tobeasymmetrical. Theleftsideprojectsmorethantheright.Ongentlepalpa-tionasenseofincreased resistance isdetected bythehand.Thereisnolossofresonance. Iftheexamination becontinued forashorttimetheprojec-tionwillbefelttodisappear moreorlesssuddenly andagaintoreappear,theimpression beingcomparable toa,balloon alternately filledandemptiedbyapump. Thesignisusually onlynoticed duringthefirsthourortwoafterameal.Itisamanifestation ofin-creased motility ofthestomach wall.Itisnotsoconstant asign'asepigastricundulation, butoccursmucheariierinthecourseofthedisease. Thusiiasomecasesofcancerofthepylorus itmaybedetected atatimewhenanorexia islittlemarked, andbeforetheoccurrenceofvomiting orpalpable tumour'. In-termittent tension, thetonicspasm,givesplacetotheclonicspasm,undula-tion,whenthedisease ismoderatelyadvanced, buttbhereisoccasionally atransition stage,atwhichforashorttimethetwosignsarepresent together.
(324)Curious Phenomena seeninTyphoid
Fever.
BACCARANI (LaRif.Med.,February23rd,24th,I90I),describing somemoreorlessunusual symptoms observed byhimintyphoid, mentions thepulsusparadoxus whichwasobserved inone
caseduring thefeverandinconva-lescence. Itwasnotregularly present,andwasapparently duetomyocardialweakness. Inthesamewasalsoob-served thesplenic cough-that is,paroxysmal cough-induced whenevertheenlarged spleenwashandled; whenthespleenreturned toitsnormal sizethisphenomenon couldnotbeexcited.Inanother casesciatorrhoea wasamarked symptom. Thediazo-reactionwasnegative, although verymarked intheurine. Nothing abnormal couldbedetected inthesalivary gland. Inthesamepatientslightbalanitis andureter-itiswerealsoobserved; nogonococciwerepresent. Intwocasesmoderatedesquamation, chieflyconfined totheabdomen andthorax, wasnoticed. Adystrophic pruritus waspresent onthefifteenth dayinanother case.Withre-
gardtotreatment, theauthorspeakswellofthepractice ofgivingsmallwaterenemata everydayin'theconsti-pationofearlyconvalescence aftertyphoid. Thepalmo-plantar signfirstde-scribedbyFilipovicz in1897andcon-
sistinginayellowdiscoloration ofthepalms ofthehandsandfeet,'wasnoticed inallthecases. Itappearedtohave norelation tothegravity of,theattack. Thecommon timeofoccurrence isonthefifteenth day.
(325)Bronzed SkininDiabebes.
MIMI(Rivist. Crit.diClin.Med.,Marchi6th,190I)reportsthecaseofawoman,aged59,whosefatherandthreebrothershaddiedofphthisis. InDecember, I898,shefirstnotiecd greatweakness, un-
usualthirst,andpigmentation onherarm.Thesesymptoms increased, andon
admission (March 6th,I899)almostthewholeoftheskinwasbronzed exceptthesolesofthefeetandthepalmsofthehand(thepalmar andinterphal-angeal sulciwere,however, deeplypigmented). Thebronzing wasveryslightonthedorsum ofthefootandrTDBan= 77Lkun.LJora_
overthepatellar region. Thepigmenta-tionwasverymarked overthedorsalsurface ofthefingers. Thetongue was
non-pigmented, buttheoralmucousmembrane wassparsely dotted withbrownish maculwe. Thegenital mucous
membrane wasnormal. Thewoman
waswasted, andtheinguinal andaxillary glandsslightly enlarged andhard.Nothing abnormal wasdetectedintheabdominal orthoracic viscera.OnMarch i8thtroublesome diarrhoea,withleftlumbo-abdominal pain,setin,which waschecked byadministeringsuprarenal capsule tabloids. Theurine(specific gravity 1025)contained from2to4percent.ofsugar;therewasno
albumen orpeptone. Theglycosuriadefinitely disappeared onApril26th.Therewasnosteatorrhoea. Inthebeginning ofi900thepatient leftthehospital quitewell;nobronzing, no
glycosuria.
SURGERY.
(326)Diagnosis orSunppurative ALtectilou&
oftheKidney.
BAZY(Bull.etM6m.delaSoc.deChir.deParis,No.I4,190o)directsattention tocertainsymptoms andsigns,particularlypyelo-vesical anduretero-vesical re-
flexes,whichhe'regards asvaluableaidsinthediagnosis ofsuppurativelesions ofthekidney. Special or,as
theauthorstylesthem,extraordinary,methods ofdiagnosis, ascystoscopy andureteral catheterism, are,hestates,oftenimpracticable andinadequateevenintrainedhands,andsoitisheldtobenecessary toperfect andmultiplythesimple andpractieal methods thatcanbereadilyapplied byanysurgeon.Theauthordealsinthiscommunicationwithcasesoffrequent micturition withterminal painandturbid urine, in
whichitisdifficult todetermine withcertainty whether thosesymptoms beduetoassociated pyelo-nephritis andcystitis, ortopyelo-nephritis withmerely reflexorsympathetic affectionofthebladder. Muchimp6rtance isattached topainsradiating towards thebladder, together withdesiretopassurinecausedbydigitalpressure ontheanterior abdominal wall,aboutaninchtotherightorleftsideofthemedianline.Thissign,whichiscalledpyelo-vesicalreflex, seems,however, tobeveryrare.Theuretero-vesical reflex,whichhasbeenmuchmorefrequentlymadeout,isapplicable forthemostparttofemalepatients, asitcannotbepractised without muchdifficulty inthemale.Onvaginalexamination ofthepos-teriorwalloftheemptybladder, tender-nessstrictlylocalised attheorificeoftheureterwillindicate diseaseofthecorre-
sponding kidney, andifmadeouton
bothsideswillindicate thatbothorgansareaffected. Iftheneckofthebladderbetender onvaginalpressure, andtheregioncorresponding totheorificesoftheureters befreefrompain,thebladder aloneisaffected, andnotthekidneys. Thediagnosis ofthesourceofpusinpyuriamay,itisstated,bemadeby.adding afewdropsofFehling'^ solu-
1220A .', ", .. 'i ---- | 8df8a85cfa67919aa1c91a85586192e14232eff7 | page_0000 |
M3IA ORA]EIO E O U R N EIA LIE AT R.MA 8 19.
tiontoaspecimen ofthepurulent urineinatesttube,andbyheating the
mixture. Ifthecoagulum thusformedfallstothebottomofthetubethepusisderived fromthebladder, andif,onthe
otherhand,itrisestothesurface, itmaybeassumed tobederived fromthekidney. Frequency ofmictirition outofproportion topainindischargingurine, andnocturnal pollakiuria inayoungsubjectarealsoregarded asindi-cations ofrenaldisease.
(327)TheArrest ofmemorrhage in
Operations ontheLiver.
CHAPOT-PREVOST (Arch.Prov.deChir.,April,igoi) finding thedifferent methodsemployed forarresting hmemorrhage afterremoving aportion ofthelivermoreorlessunsatisfactory, hascarefully studiedthesubject ofheemostasis inthisbranchofsurgerywith aviewtotheimprovementofitspractical details. Heregards asanidealmethod onebymeansofwhichthesurgeon isenabled toapplyforasufficient timetothehepatic woundaregular andmethodical compression suchasthatexercised bythefingersapplieddirectly overthesurface ofthegland,andonewhichatthesametimewillpre-ventcomplete closure oftheabdominalwall.Thiscompression, itisasserted,maybeeffected bybringing thecontig-uoussurfaces oftheparietal peritoneumintoclosecontact withthesurface oftheliveroneithersideofthewoundinthisgland,andalsobybringing together therawsurfaces ofthiswound, bymeansofquilled sutures passedthrough theab-dominal wallandliverandsecured ontheexternal surface oftheabdomen byrollsofgauze. Thethorough efficacy ofthisplanhasbeendemonstrated bytheresults notonlyofexperiments onani-
malsbutalsoofanoperation performedfortheseparation oftwolivinggirlswhowereunitedbyabroadbandofskinen-closing athickbridgeofhepatic struc-tureandacartilaginous rodformed by
coalescence ofthexiphoid processes.Theageofthisdoublemonster, judgingfromtheillustrations giveninthearticle, was6or7years.Theinterveningthickmassofhepatictissuehavingbeendivided therawsurfaces onwhichtheopenendsofthreelargebloodvesselscouldbeseenwereineachchildcom-pressed byquilled sutures. Theimme-diateresultsoftheoperation weresatis-factory asperfectandpersistent heemo-stasis waseffected bytheauthor'smethod. Inthecourseoftheoperationthreeserouscavities-the peritoneum,pleura, andpericardium-which werecommon tobothchildren, wereopenedandclosedbysutures. Onechildre-covered andwaslivingatthedateofthepreparation ofthispaper. Theotherchilddiedonthesixthdayfromthedateofoperation frompleurisy.
(308)Permanent Dislocation ofthePatella.
M'LAREN (Reprint fromtheTransactionsoftheMedico -Chirurgical Society ofEdinburgh 1899-1900) published there-sultsofastudyofacaseofpermanent dislocation ofthepatellarecently under hiscare,andalso
ofsomeothers thathavebeen
1220Brecently recorded. Thisrarelesionisadisplacement ofthepatellaoutwards, sothatthebonelies,eveninextension oftheknee,inmostcasestotheoutside ofthemiddleline,andinflexion itstepsoutwards andbackwards untilitliesontheoutersurfaceoftheexternal condyleofthefemur, itsanterior surfacedirected partlyoutwards. Thepatient'sgaitisthusrendered quiteinsecure, andfrequent fallsandsynovitis ofthekneearetheresult. TheauthoragreeswithAppel,whohasalsodirected muchat-
tention tothisformofdislocation, thatitdepends onacongenital defect,namely,adeformity oftheexternal condyle andthetrochlea ofthefemur. Thepatellaisusually ill-developed andsmall,andthepatellar fossaveryshallow. Thedis-location isoftencomplicated withgenuvalgum andoutward rotation oftheleg.Various methods ofoperation havebeenpractised inthetreatment ofthiscon-
dition. Intheoriginal caserecorded inthispaperthefollowing method oftreat-mentsuggested itselftotheauthor. Aflapofskinandfathaving beenturnedupinfrontoftheknee,theexpansion ofthequadriceps tendon andthecapsuleweredivided ontheoutersideofthepatella, without opening thecavityofthejoint,soastoallowthebonetobebrought easilytothemiddle line.Twoholeswerethenboredthrough theinneredgeofthepatella, through eachofwhichacatgutsuturewaspassed. Thebonewasthenretained initsnormalsituation bystitching ittotheinternallateralligament. Thecatgut sutures
gavewayanditwasfoundnecessary atasecondoperation tosubstitute suturesofstoutsilk.Thismethod ofdividingcontractions ontheoutersideandstitching thepatellabysuturespassedthrough thestructures ontheinnersideseemstotheauthor toberational andeffective, andshould, hethinks,succeedinmostcases.Itispossible, however,hesuggests, thattheseverer procedureofopening thejointandenlarging thegroove forthepatellamight insome
casesbenecessary; while,again,iftheoutward rotation ofthelegisverymarked, itmightbefoundadvisable todetachtheligamentum patellae andfixitfurtherinwards.
(329)Naso-orbital Hyperostosis dueto
Distension oftheFrontal binus.
ROLLET (LyonMed.,Marcll3ISt,I90j,believes thatmucocele ofthefrontalsinusresultsfromachronic inflamma-toryhypersecretion withretention ofmucuscausing over-distension ofthecavityofthesinus. InI896heshowedacasetotheSociWtd deMedecine ofa
youth,agedi8,whohadanold-standinghyperostosis obliterating thebridgeofthenose,followed bysymmetricalorbitaltumours, whichappeared eightmonths beforeoperation. Ontrephin-ing,thesinuswasfounddistended withamucous fluid.In3othercasesofmucousdistension thehyperostoses havebeennoted; inXofthem,however,thebonyovergrowth wasunilateral.Rollethasalsofoundthehyperostosesin2casesofold-standing empyemaofthefrontalsinusuponwhichhehasoperated. Thetumours areoftenfluctu-atinginparts,andthisfact,togetherwiththeirexactlimitation tothenaso-orbital region, should sufficetodis-tinguish themfromosteomata andsyphi-liticexostoses. Thecasesusually occurattheperiodofadolescence, butthetheories of'primary bonyovergrowthandprimary obstruction ofthecanalarenotasprobable explanations oftheoccurrence asisthetheoryofprimaryinflammation. Inoneofthecasesmen-tioned pressure caused theorbitaltumour toemptyitselfintothenasalfossa.
MIDWIFERY ANDDISEASES OFWOMEN.
(330)Local 'Uterine Hmemostatics.
TOFFwritesonthissubject(Rivista de
Chir.,1901,p.23).Withregard to(i)ferripyrin, hesaysthatin65cases
there wereonly6failures. Of4ofpuerperal bleeding Ifailed,andofi8afterabortion a.Of20endo-metritics Ifailed, whileI9casesofprofuse menstruation allreacted. Inoneof3myoma patients itfailed,butitgavegoodresultinacervicalcancer. Onec.cm.ofa15to20percent.solution wasinjected byaBraunsyringe, andinonlyonewerepainfulcontractions ofthewomb excited.Whenthebleeding isveryfreethetampon isusedaswell,buttheferri-pyrinstopped bleeding insomecases
wheretamponade, ergot, andothermeanshadfailed. (2)Antipyrin-salol.Spaeth (Monats. f.Geburts. u.Gyn.,I901,xiii,p.330).Thisisamixture ofantipyrin andsalolliquefied byheating,andisapplied bywadding onaprobe.Itisonlyfluidattemperatures from1200toI90°F.,andismeltedinatesttube.Itproduces aslough,especiallyatthecervix. Asallthepatients didnotreturn forobservation thefullnumber ofpermanent curesisnot
known. In70cases6showed no
result.andin7,thoughcessation ofa
menorrhagia ormetrorrhagia wasgot,noprotection againstfurtherprofusionoreffectonitwasgot.Ofthese5hadendometritis, and4ofthese were
known tobeglandular. Fourcases
associated withadnexal swelling were
alsofailures. Theotherswere2withmyoma, wheretheeffectwastransitory,Iwithapolypatthecornu, andI
constitutional. Ofthosewhichgavegoodresults I9wereendometric and17subinvolution postabortum, 8ofthemneeding tobecompleted, 3were
bleeding aftercurettage, Iwasclim-acteric, andIpubertal flooding. Theothers wereendometritic, withpara-
metritis ini,perimetritis in3,salpingo-obphoritis in4,andoOphoritis in5.Thenumber ofapplications maybeone
ortwo,andafewmayneedthreeor
four.Whenthesearenoteffectual more
aregenerally useless. Atableofthecasesisgiven.
(331)Version rorContracted Pelvis.
ALBERT (Munch. nzed.Woch.,April2nd,I901)giveshisexperience ofthetreatment ofcontracted pelvis. HehasTmBarrixz19t KIDICALJOUILMAJLJ EPITOME OFCURRENT MEDICAL LITERA.TURE. [M.iYi8,1901. | 8df8a85cfa67919aa1c91a85586192e14232eff7 | page_0001 |
MAY 18,1901.] EPITOME OFCURRENT MEDICAL LITERATURE.rTD_m 7LXeL JOU=N7
-collected hismaterial fromhisexternalmidwifery department inDresden, anddealswith1,187births. Hegeneralisestheteaching bysaying thatahighforcepsoperation mustneverbeunder-takenwhenversion ispo3sible, andthat-version andimmediate extraction shouldbepractised whzi'eareasonable chance,ofspontaneous delivryhasbeengivenwithout avail. Hisassistants arenotallowed touseforceps (highoperation)ona"movable"head, andespecially-whenthereisacontraction ofthepelvis.Headvisesthatnarcosis beadopted forversion, andespecially emphasises theutilityoftheVeit-Smellie method of-dealing withtheafter-coming head.Theassistant ornursemustapplypressuretotheabdominal wall(thatis,totheuterus) inadownward andbackwarddirection, withthepalmofthehandandnotwiththefingers. 'Further, he-findsthatthe"hanging" position,afterWalcher, inwhichthemother isplaced inadorsalposition, withthethighshyperextended andhanging overtheedgeofthebedortable,isparticu-larlyvaluable inthestageofextractionoftheafter-coming head.Themem-branesmustbepreserved intactaslongaspossible, andallowed toexerttheirinfluence indilating theos.Incasesinwhichthemembranes havealready rup-tured,orfailtoproduce thenecessarydilatation, artificial dilators mustbe;used,andhepointsoutthatacarefulintroduction ofasuitable "bag"isunattended withanyrisk.Heusesit(i)when,themembraneshaving ruptured prematurely, hecanretain theliquoramnii, dilatetheos,andstrengthen thepains,-especially inprimiparae; (2)when,withintactmembranes, theheaddoesnot,engage, onaccount ofcontracted pelvisandincompletely-dilated os,andalsoinobliqueand transverse presentations;'(3)whenaclearideacannotbeformedastohowthecourseofthecasewillbeincasesofmoderate contraction of-thepelvis. Heusually givesmorphineinsmalldoses(Igr.)whenheusesartificial dilatation. Inhiscasesver-'sionwasperformed 105times;ofthese,45turnings wereusedwithnormalpelvesforthenon-engagement ofthehead,transverse presentation, prolapseofthecord,placenta praevia, andother-auses. In6ocases,therewereflattenedpelves(14cases),generally contractedandflattened pelves(I7cases),generally-contracted pelves(14cases), and15casesofcontracted pelvesinwhichthemeasurements werenotrecorded. Thetrueconjugate variedbetween 2.7and3.5inches; ofthechildren, 9outof9easesofplacenta prievia died,and3werealready deadbeforeversion wasperformed. Thisleft93children, and5ofthemdied,giving82.8percent.of'livingchildren (thatis,livingwhenthemother wasdischarged as"well").Ofthecontracted pelvischildren, I8.7per-cent.diedand8I.3percent.lived.The,mothers allrecovered except i.In4casesatemporary pyrexia wasre-,corded. Inconclusion, hepointsout*thattheresultsmightbeevenbetter,ashisassistants wereofteninexperi-encedintheoperation ofversion, andthusmayhavefailedtosavethechildincertain casesinwhich amoreskilledpractitioner couldhavebeensuccessful.
(332)lnoontroliable Vomiting or
Pregnancy.
L.LAPEYRE (L'Obst6trique, vi,.44,I901)records acaseofhyperemesis ina
secundipara, aged 25,whohadalso a
tumour. Tintiltheperitoneal cavitywasopened intopervaginam, itwas
doubtful whether theswelling whichexisted inaddition totheuterine en-
largement wasanovarian cystoran
extrauterine gestation. Itturned out
tobeanovarian cystoma inthepouchofDouglas, andwaseasilyremoved byposterior colpotomy; atthesametimethediagnosis ofpregnancy wasclearlyestablished, theuterus wasfoundtobeinclined totheleftsideandwasofthesizeofatwo-months gestation. Not-withstanding theremoval ofthecystthevomiting continued tobeincoerci-ble,andtheinduction ofabortion bymeansoftheintroduction ofalaminariatent wasrendered necessary. Twinfoetuses wereextracted, andthenthevomiting ceased, orratherithadceasedwLththeoompletion ofthecervical dila-tation. Agood recovery wasmade.Lapeyre concludes fromthestudyofthiscaseandacomparison ofitwithothers thatinpregnancy complicatedbythepresence ofanovarian cysthyperemesis isnotmorefrequent thaninsimple pregnancy; thatifuncontrol-
lablevomiting doesexist,itisconnectedwiththepregnancy solely, andnotinanyway withthecyst;andthatthere-foretheremoval ofthecystisuselessinthetreatment ofthevomiting, whichindeedrequires theinduction ofabor-tion.Healsoregards thevaginal routefortheremoval ofthecystastheoneto
beadopted ifpregnancy coxist.
THERAPEUTICS.
(333)Gelatine Inleetions InAortic
Anenrysms.
K.BARTH (Munch. med.Woch.,April2nd,I901)describes acaseofaneurysm
oftheascending partoftheaorticarch,which hetreated withLancereaux'sgelatine ifjections. Thepatient was
50yearsofage, andshowed apast his-
toryofpertussis at4years,leavingchronic coughand-narrowness ofthechest; at17hehadsomeglands re-
moved; ati8hesuffered afracture ofhisright"leg"(bonenotnamed); attheageof24hewenttoAmerica, and
therehadtwoattacks ofmalaria, twoattacks ofpneumonia, and oneof
pleurisy (boththechestconditions on
theleftside),andattheageof48hefellfromhishorseandlayunderthehorseforashort spaceoftime. Thefrontandrightsideofhischest werehurt,andsubsequently "tumour" formed intherightlowerregion ofthechest,which wastreated surgically. Fromthedateofhisfallhehadsymptoms ofintrathoracic trouble, andayearbeforethetreatmentwas undertaken henoticed
aswelling intherightinfraclavicularregion. Thesignsandsymptoms atthe-timeoftreatment wereslightcyanosisofcheeksandlips,enlarged veinsintheupperthoracic wall;radialpulsesoft,andrightpulsealittlesmaller andlaterthantheleft;aswelling intherightinfraclavicular region waswell;marked, andshowedpulsations. Thedulness extended equally inalldirec-tions, andmeasured 4inches indiameter. Therewassomeindistinct-nessofthebreathsoundsintheneigh-bourhood ofthetumour. Bronchitis,tracheitis, andemphysema werefoundtobepresent, andintheheart,besidesamitralinsufficiency, withdilatation oftherightandleftventricle, thesecondsoundwasaccentuated everywhere.Therewasaloudsystolic murmur tobeheardalloverthecardiac areaandover
theswelling. Hecomplained ofpainin
chestandrightarm.TheinjLctions ofgelatine werecarriedoutwithallasepticprecautions, andcaused nolocalor
generalinconvenience; 15.4grainswere
injected insolution (ipercent.)everyotherdayatfirst.Twelve suchinjec-tionscompleted thefirstcourse,andrestinbed,sodium iodide, Leiter's"cooler," andrestricted dietwerecom-
binedwithit.Lateronasecond course
wascarriedout.Theexamination afterthelastshowed thatthedulness hadshrunk toanareameasuring 2.34inchesby1.95inch;theswelling feltfirmer,andwasnotsoprominent, nordiditpulsate asvisibly. Allothersignswere
lessmarked, andthesubjective sym-ptoms(pain,discomfort, inability tomakeanyexertion) hadcompletely dis-appeared. Barthproposes shortly to
applyathirdcourse.
(334)Treatment ofSciatica byIntra-
rachidean Injections orCocaine.
PULLE(LaRif.Med.,February 22nd,I90I)reportsthecaseofaman,aged47,admitted intohospitalOctober8th, I900,suffering fromsciatica ofaboutamonth'sduration. Blisters andetherspray,whichhadrelieved aprevious -attack,provinginefficacious, intrarachidean in-jection of2percent.aqueous solutionofcocaine waspractised. Theinjectionwasgiveninthefourthlumbar space,about Icm.totherightofthemedianline,thesolution being warm.Aboutthreepartsfullofanordinary hypo-dermicsyringe wasgivenwithstrictantiseptic preliminaries. Theresultwasverysatisfactory, andexcept forsomeslightfeverlasting threeorfour
daysandapatchoflabialherpes, no
secondary symptoms followed. Theanalgesia lastedabouttentotwelvehours. Thesciatica didnotreturn,andthepatientgotuponNovember i6thfeel-ingquitewell,andleftthehospital on
November 23rdinper,fecthealth.
(335)Q4ulic Acid.
STERNFELD ofMunich (Miunch. med.
Woch.,NO.7,I901)strongly recommendsquinicacidasaremedy forgout,owingtoitsstrongsolvent action onuricacidintheblood. Ithasnoneofthedis-agreeable effectsofquinine, andwheninthebodyitisconverted intobenzoicacid,which,unitedwithnitrogenous122002 1 | 8df8a85cfa67919aa1c91a85586192e14232eff7 | page_0002 |
.-- -.-. --AEPITOME -, CURREN---P 1-1
wasteproducts, getsexcreted inthe
urine ashippuric (amido-benzoic) acid.Thecombination ofquinicacidintoan
alkali, as,forexample, lithium quinate,hasbeenfoundeffective bothindissolv-inguricacidandinpromoting diuresis
andtheexcretion ofuricacid.Hegives
itintheformoftablets (prepared byLimmer andCo.ofFrankfort) of2gram
each(8grs.),andadministers from6to
iotablets aday.Astheresultoftreat-
ingaconsiderable number ofcasesheconcludes thatquinicacid isaspecific
forgout, asarethesalicylates foracute
articular rheumatism andquinine for
malaria. Theonlydrawback atpresent
asregards lithium quinate isitshigh
price, atubeof25tabletscosting about3.50marks. Theresults, however, are
sogratifying thatinprivate practice at
leastthistreatment shouldbepreferred
tothatbyotherdrugs.
(336) TheSterllisation ofCocaine and
Atropine Solutions.
SIDLER (Correspondenzbl. f.Schweiz.Aerzte, I900,No.6)hasfoundthatcocaine andatropine solutions sterilisedbyheatarechemically changed andren-
dereduseless, andthatsuchsolutions,oncesterilised, areveryeasilyinfected.Heexamined forty-seven suchsolu-tions, andfound various organismspresent, including thestreptococcus
pyogenes, staphylococci, various diplo.cocci,andmoulds. Suchorganisms are
communicated tothesolutions bytouching thelidsandlasheswiththepipette, andalsobythedropscoming
incontact withtheusually narrow neck
ofthebottles used. Theauthor de-vised aspecial bottle,butstillfound
contamination frequent. After atrialofvarious antiseptics, hefound a45per
cent.alcohol themostsatisfactory.
Solutions of3grams ofatropine or
cocaine in20grams of45percent.alcohol remained sterileforayear,andtheirchemical composition was un-
changed. Theywerekeptinasterilisedbottlewithgraduated pipette, andthedesired strength obtained bydilution
withtherequisite amount ofsterilewater.
(337)Protargol inGonorrha.
NIESSEN (Miinch. med.Woch., MarchI9th,l90I)hasgivenprotargol afair
trial,having useditin244casesof
soldiers. Heargues from 99ofthesecases,inwhichtheattack wasuncom-
plicated, anterior urethritis, and was
thefirstattack. -Inconjunction withtheinjections hekepthispatients in
bedonalowdietforfromloto14days,and,later,used everyprecaution to
assist thetreatment. In 34cases
(=34.34percent.)complications deve-lopedduring';thee treatment. Ofthese29'showed urethritis posterior, 13had
bladder symptms, 5inflammation ofthetestes. andvperiurethritis. Heusedah*t iiiyofai,percent.solution
at,firt,And,Aater, increased thisto
;iliiLdrpercent.solution. ThedstMi6iibf-the casesshowed anaverage
Of30!7'duYs, whilein4cases recurrence
tookplacelater. This average isslightly longer thantheaverage he
1220 Dobtained -withotherdrugs.. Incon-
clusion, hestatesthat:(i)Protargoltakesaboutthesametimeasarg.nit.
tokillthegonococci, thatis,threeweeks; (2)-toXpercent.solutions
produce nodiscomfort; (3)theoccur-
renceofcomplications, forexample,posterior urethritis, isnotsoserious as
witharg.nit.;(4)inthegleetystagesastringents seemtoworkbetterthan
protargol ;and(5)thedisease isnotcutshortbytheuseofthedrug.
(338) Santonin (Santonic Acid) InTabetic -
Palms
NEGRO(Giorn., dellR.Accad.diMed.diTorino,February, igoi)has triedsantoninwith success inthetreatment ofthe
lightning painsoftabes. Ofthe ii
casesinwhichthedrugwastried8were
decidedly relieved, 2temporarily re-
lieved, andIunaffected. Atfirsttheauthor gaveI5grs. inthreedosesatin-
tervalsofthreehours,andinsubseque-ntattacks beganwith iogrs.,5grs.fivehourslater. Thepaingotdecidedlylessinthreehoursafterthefirstdose,andcompletely ceased twohoursafter
thesecond dose.kSofar,theauthorhasonlyadministered itduring thecrisis,
notintheintervals. Inonecaseitgave
reliefwhen amixture ofantipyrin andphenacetin hadproved futile. Noneofthepatients hadthistreatment more
thanfourorfivetimesinthecourseoftwoorthreemonths.
PATHOLOGY.
(339) Cancerous Lymphangitis oftheLung.
TROIsIER ANDLETULLE (Arch.deMid.
Exp&r., March, I9OI)contribute anar-
ticeIeonthemicroscopic characters foundinthelungincasesofcancer. The
richness ofthelymphangitic network in
thelungenables ustorecognise with
great easethedifferent stagesinthein-
volvement ofthetissues bycancer.
Cancerous lymphangitis is'closely re-
latedtothewholequestion ofthegene-
ralisation oftheprocess bythelymph-atics,hencethevalueofcarefulstudyof
thatprocess inthelung.Thealveoliof
cancer areindirectcommunication with
thelymphatics ofthestructure in-
volved, andthustheirpenetration bytheneoplastic elements ismerely a
question cftime.Mostfrequently thecancercellscarriedbythelymphstream
arearrested inthefirstlymphatic gland,butinother casesthesecellsareen-
abledtoinsinuate themselves intothe
lymphatics andbecome engrafted on
theirwallsandproliferate intotheir
cavities. There aretwovarieties ofcan-
cerouslymphangitis: inoneallthe
neoplastic elements which fillupthelymphatic spaces arelargecellsshow-ingoneormorenuclei. They accumu-
lateandcompress eachother,anditbecomes impossible todistinguish en-
dothelial cellsandtheinternal coating
ofthevessel. Thisformoflymphang-itisismostusualinthelymphatics of
smallcalibre. Thesecond variety is
characterised bycaseous degenerationinthemiddle oftheneoplastic mass.
Thelymphatics arevery evident onthe
surface ofthelung,andhavethe.ap-pearance ofnoduled ormoniliformIesions ofayellowish-white colour.
Theyaremostevident intheneighbour-hoodofthesecondary cancerous nodules
scattered overorinthepulmonary sub-
space. Thewallofthelymphatie does.
notseemtobemuchaffected, thechief
lesion tobemadeoutbeingthedisap-pearance oftheendothelial cellswhich
Linethenormalvessel. Thesecells,how-
ever,donotseemto-playanypartin
theformation ofcancer cells,thelatter
beingduesolely toproliferation of
primitive cancercellscarried intothe
haticvessel. Thewriters state
thatthethoracic duct mayalsobe
involved inthesame process. Thelesionissimilar inallitsaspects, and
thevesselistransformed into,a hard
greyoryellowmoniliform cord.Trans-
versesection ahows thatitiscom-
pletelyobstructed byamassadherenttoitswall,andwhichisgreyatthe
periphery, caseous atthecentre. This
mass ismade upoftypical cancer cells.
Thewriters statethatthewallofthe
thoracic duct,wheninvaded bycancer,
shows somethickening, afeature not.
observed inthelymphatics ofthelung.
(340)Symmetrical Ripoinatesis.
DEBuCK ANDL.DEMOOR(BelgiqueMUd.,vii,64I,673,November, I900)describe acaseofsymmetrical lipoma-tosisduetofattymetaplasia ofmuscle.
Thepatient wasafarmlabourer, 30
yearsofage.Thedisease hadbegunto
manifest itselfwhenhewas14years.
old,andhadtakentheformofprogres-
siveincrease insizeoftherightflankandlumbar regions. Abouttwoyears.
later atumour, nature unknown, was
remolved fromtheflank. Sincethenam
increase insize.ofbothflankshasbeen
observed, extending symmetrically on
bothsidesofthevertebral column fronm
thecrestoftheiliumtothescapularregion. Atumour hadalsoformed in
thecoccygeal region. Ayearagohewas
operated onforaright-sided inguinalhernia, consisting entirely oflobules of
fat.Themanhasneverhadtogiveup
work onaccount ofthesetumours. On
palpation theygavethesensation of
lipomata. They were,however, deeplysituated. Nevertheless, anattempt was.
madetoremove them,butitfailed,for
thetumours wereevidently partsofthe
lumbar muscles. Asmallpiece was
extirpated formicroscopical examina-
tion;andinitwhile somemuscular
fibrescouldberecognised inbundles.andhaving anormal appearance, others.
had ayellowish, andyetothers a
whitish-yellow look.Microscopically all
thestagescouldbeseenbetween simplenon-degenerative atrophy andcompletedisappearance. Therehadbeen,further,
notaninterstitial proliferation offatat
theexpense ofpre-existing connective!tissue,butaretrogressive involution of
themuscular tissuewhichhadpassedthrough thesarcoblastic embryoniestageintotheultimate transformationintofat.Bothstriated muscle andfat,.
itwillberemembered, areembryo-logically mesodermic, andtherefore tbeexplanation givenabove,although novel,,isquitefeasible.WnICALJOUR'vA
fMAY18,1901. EPITOME, -,OF .,CURRENT, -MEDICAL 11TERA.T.VRE. | 8df8a85cfa67919aa1c91a85586192e14232eff7 | page_0003 |
Single-molecule FRET measures bends
and kinks in DNA
Anna K. Woz ´niaka, Gunnar F. Schro ¨derb, Helmut Grubmu ¨llerb, Claus A. M. Seidela,1, and Filipp Oesterhelta,1
aHeinrich-Heine-Universita ¨t Du¨ sseldorf, Institut fu ¨ r molekulare Physikalische Chemie, Universita ¨tsstraße 1, Geb. 26.32.02.44, 40225 Du ¨ sseldorf, Germany;
andbMax-Planck-Institut fu ¨ r biophysikalische Chemie, Theoretische und computergestu ¨ tzte Biophysik, Am Fassberg 11, D-37077 Go ¨ ttingen, Germany
Edited by Steven M. Block, Stanford University, Stanford, CA, and approved September 24, 2008 (received for review January 31, 2008)
We present advances in the use of single-molecule FRET measure-
ments with flexibly linked dyes to derive full 3D structures of DNAconstructs based on absolute distances. The resolution obtained bythis single-molecule approach harbours the potential to study indetail also protein- or damage-induced DNA bending. If one is togenerate a geometric structural model, distances between fixedpositions are needed. These are usually not experimentally acces-sible because of unknown fluorophore-linker mobility effects thatlead to a distribution of FRET efficiencies and distances. To solvethis problem, we performed studies on DNA double-helices bysystematically varying donor acceptor distances from 2 to 10 nm.Analysis of dye–dye quenching and fluorescence anisotropy mea-surements reveal slow positional and fast orientational fluoro-phore dynamics, that results in an isotropic average of the FRETefficiency. We use a nonlinear conversion function based on MDsimulations that allows us to include this effect in the calculationof absolute FRET distances. To obtain unique structures, we per-formed a quantitative statistical analysis for the conformationalsearch in full space based on triangulation, which uses the knownhelical nucleic acid features. Our higher accuracy allowed thedetection of sequence-dependent DNA bending by 16°. For DNAwith bulged adenosines, we also quantified the kink angles intro-duced by the insertion of 1, 3 and 5 bases to be 32° /H115506°, 56° /H115504°
and 73 /H115502°, respectively. Moreover, the rotation angles and shifts
of the helices were calculated to describe the relative orientationof the two arms in detail.
absolute distance measurements /H20841fluorescence energy transfer /H20841
multiparameter fluorescence detection /H20841nucleic acid structures
In recent years, fluorescence energy transfer experiments
(FRET) have shown great potential for subnanometer analysis
of biomolecular structures and their dynamics that can even beapplied to single molecules (1–5). Calculation of absolute FRETdistances between a donor and acceptor fluorophore is compli-cated because of several ‘‘calibration’’ factors such as detectionefficiencies, spectral cross-talk and fluorescence quantum yields(6) that are difficult to determine accurately. MultiparameterFluorescence Detection (MFD) (7) avoids most pitfalls (5) bysimultaneously collecting all fluorescence parameters (intensity,lifetime and anisotropy in both spectral ranges) at the single-molecule level.
However, determination of absolute distances remains a ma-
jor challenge because of the uncertainty in the fluorophorepositions. This uncertainty is due to the use of the long linkers
through which the fluorophores are attached to the biomol-ecules. Orientational freedom is a prerequisite to safely assumean orientation factor (
/H92602) of 2/3 (8); however, this prevents a
defined fluorophore position, which is needed for fitting data toa geometric model. Here, we present a new conversion functionfor experimental FRET efficiencies that considers the dynamicsof the movement of the fluorophore-linker system and takes intoaccount the fluorophores volume of occupancy derived frommolecular dynamics (MD) simulations.
To date, the handedness, helicity (9–11), and sequence-
dependent bending (12) of DNA has only been studied by FRETin a qualitative manner. In this work, we were able to determine
unique 3D structures of DNA by triangulation of a set ofaccurate FRET distance constraints in combination with arigorous analysis of errors. This allowed characterization ofsequence-dependent bending of double-stranded DNA and gavestructures of DNA kinks introduced by bulged A-loops. Weperformed MFD studies of various DNA constructs with sys-tematic donor-acceptor distances that varied between 2 and 10nm to benchmark the spatial resolution of our method.
Results and Discussion
MFD Measurements. We investigated 12 distances within a double-
stranded DNA sequence ranging from 5 to 27 bp in steps of 2bases. To obtain a set of 12 appropriately double-labeled samplemolecules, we combined each of 3 DNA oligonucleotides labeledwith the donor fluorophore Alexa Fluor 488 at different posi-tions with each of 4 complementary strands, labeled with theacceptor Cy5 at different positions. Taking a set of 3 sampleswith interfluorophore distances at 5, 11, and 19 bp, bursts offreely diffusing molecules measured by MFD were arranged intoone joint histogram. Fig. 1 Ashows a 2D frequency histogram of
the donor-acceptor fluorescence intensity ratio F
D/FAversus the
donor fluorescence lifetime (in presence of the acceptor) /H9270D(A)
calculated for all single-molecule events. The 3 FRET-speciesand the donor-only species can be clearly distinguished. Single-molecule MFD-FRET allows one to take into account allexperimental corrections that affect accurate distance measure-ments, such as detection efficiencies and cross-talk (5). The redline in Fig. 1 Ashows the expected relation between the donor-
acceptor intensity ratio and
/H9270D(A) when only FRET is changing.
The fact that all species fall on this red line, proves that theobserved differences are only due to a FRET change. For a givendonor-acceptor distance R, the transfer efficiency is given by E/H11005
R
06/(R06/H11001R6) where R0is the Fo ¨rster radius (in ångstroms), which
accounts for the system properties. It is calculated by R0/H11005
(cFTJ/H92602/H9021FD (0) n/H110024)(1/6), where Jis the overlap integral of the
donor emission spectrum with the acceptor absorption spectrumwith the units [M
/H110021/H18528cm/H110021/H18528nm4],/H92602accounts for the relative
orientation of donor and acceptor, /H9021FD(0) is the donor fluores-
cence quantum yield in absence of transfer, and nis the refractive
index of the medium ( n/H110051.33). For the given units, the constant
cFTequals 8.79 10/H110025mol (5). To check that no orientation effects
influence the FRET efficiency, the anisotropy ris recorded
simultaneously for each detected molecule (Fig. 1 B). The short
mean rotational correlation time /H9267D/H110050.63 ns obtained by a fit
Author contributions: H.G., C.A.M.S., and F.O. designed research; A.K.W. and G.F.S. per-
formed research; F.O. contributed new reagents/analytic tools; A.K.W., G.F.S., and F.O.analyzed data; and A.K.W., C.A.M.S., and F.O. wrote the paper.
The authors declare no conflict of interest.This article is a PNAS Direct Submission.
1To whom correspondence may be addressed. E-mail: filipp.oesterhelt@uni-duesseldorf.de
or cseidel@gwdg.de.
This article contains supporting information online at www.pnas.org/cgi/content/full/
0800977105/DCSupplemental .
© 2008 by The National Academy of Sciences of the USA
www.pnas.org /H20862cgi /H20862doi /H2086210.1073 /H20862pnas.0800977105 PNAS /H20841November 25, 2008 /H20841vol. 105 /H20841no. 47 /H2084118337–18342
BIOPHYSICS
| 9fa368fc44cd70a30ed489930b67ce05abc4c32b | page_0000 |
with the Perrin equation (red line) shows the high mobility of the
donor dye. Moreover, the mobility of Cy5 is also known to behigh, which is in agreement with the steady-state ( r
A/H110050) and
time-resolved anisotropy (see supporting information (SI) Fig.
S1e) of the sensitized acceptor fluorescence. This justifies as a
first approximation the assumption that the mean orientationfactor
/H92602/H110052/3 (for more details, see Comparison of Experimental
FRET Data with the MD-Simulation ), which allows one to
calculate a Fo ¨rster radius given by R0/H1100551.8 Å. Thus, MFD
confirms without need of additional ensemble measurementsthat the observed distinct FRET efficiencies Eare only due to
differences in distance and are not affected by orientationeffects.
Development of an Appropriate Geometric Model. In previous
FRET investigations on double-stranded DNA, a simple straighthelix was used as a geometric model that accounts for the helicityof DNA (9, 10). It was then assumed that both fluorophores arepositioned at a fixed mean distance from the axis of the DNA,on a helical path according to the DNA rise (3.38Å) and twist(ß/H1100536°) per base pair (see Fig. 2 BandSI Appendix , Eq. 13).
However, in this work, MD simulations provide individual dyepositions, which are used to test different averaging models.
MD Simulation of the Fluorophore-Linker System on a Straight DNA
Model. Here, the fluorophores were allowed to move within the
sterically allowed space confined by the straight B-DNA and thelength of the linkers. Because we are interested in the maximumrange of possible dye positions and not in the actual dynamics,
the DNA was kept fixed in the simulation by harmonic poten-tials, electrostatic interactions were neglected, and the samplingefficiency was increased by elevating the temperature to 2000 K.Fig. 2 Ashows the structural model of a double-stranded B-DNA
together with the computed clouds of 5,157 sterically allowedpositions of the donor and acceptor fluorophore, respectively, ata distance of 21 bp. For each fluorophore, the resulting mean dyepositions are marked by big spheres (mean position model).More complex models are described in Correct Averaging Pro-
cedure . Because of the long linkers, the
/H92602calculated for all
combinations of donor-acceptor positions ranged from 0.65 to0.69 with an average of 0.67 and a standard deviation of 0.01; thisis in close agreement with the value of 2/3 for fully flexiblelinkers.
Comparing the Experiment with Simple Models. The dependence of
the experimental FRET efficiencies on the number of separatingbase pairs /H9004
bpis given in Fig. 2 B(open green squares). Curves
of the mean position model of the MD data (black solid line) anda linear distance increase model (blue dashed line) are alsoshown. Comparison of the experimental data with the 2 basicmodels shows significant deviations. First, efficiencies measuredfor small distances are smaller than is predicted by the meanposition model for a straight helix, and second, DNA helicityalone is apparently not sufficient to produce the sinusoidalfeatures observed in the experimental curve. Especially at adistance of 17 bp, a much higher than expected efficiency ismeasured. Thus, both the mean position model and the assump-tion of a perfect shape of a B-DNA are questionable.
Correct Averaging Procedure. Knowing the /H9270D(0)without transfer,
2 molecular constraints influence the FRET rate constant kFT/H11005
1//H9270D(0) (R0/R)6(6, 13, 14): ( i) orientation fluctuations with the
reciprocal rotational correlation time kR/H110051//H9267, which are
reflected in the effective orientation factor /H92602;(ii) diffusion of
the dyes in the sterically allowed volume with characteristic
diffusion rate constant kd, which influences the actual distance
R. By averaging all fluorophore positions the information on the
width of the position distribution is lost, i.e., the distancebetween the centres of the clouds of donor and acceptorpositions converges to zero for zero base pair distance, whereasthe average distance between donor and acceptor does not (15).Thus, neglecting the positional distributions would overestimatethe short range distances. Moreover, it is k
FT/H11008/H92602/R6, and not the
distance, that is the relevant physical parameter for averaging.Considering individual distances R
iand individual orientation
factors /H9260i2, 3 extreme cases with distinct mean FRET efficiencies
/H20855E/H20856are computed using R0calculated with /H92602/H110052/3 (for sketches,
seeFig. S2 A–C).
Dynamic average: kR,kd/H11022/H11022kFT:
/H20855E/H20856dyn/H11005/H20855/H9260i2/Ri6/H20856/H185283⁄2/H18528R06
3⁄2/H18528R06/H18528/H20855/H9260i2/Ri6/H20856/H110011[1]
Static average: kR,kd/H11021/H11021kFT:
/H20855E/H20856stat/H11005/H20883/H92602/H185283⁄2/H18528R06
/H92602/H185283⁄2/H18528R06/H11001Ri6/H20884 [2]
Isotropic average: kR/H11022/H11022kFT/H11022/H11022kd:
/H20855E/H20856iso/H11005/H20855R06
R06/H11001Ri6/H20856[3]
To generate geometric models, distances between mean fluoro-
phore positions ( Rmp) are needed ( SI Appendix ). However,
Fig. 1. Two-dimensional frequency histogram of single-molecule data of a
mixture of 3 distinct samples with interfluorophore distances of 5, 11, and 19bp, where frequency increases in gray scale from white to black. The ratio ofdonor and acceptor fluorescence ( F
D/FA) and the donor anisotropy ( rD) are
both plotted against /H9270D(A).FDand FAare determined from green and red
signals by correcting for background counts, BGandBR, detection efficiencies,
gGandgR, and spectral cross-talk /H9251according to equation 3 in ref. 5. (Typical
values are BG/H110052.9 kHz, BR/H110050.9 kHz, gG/H110050.36 and gR/H110050.53 and /H9251/H110050.019).
The direct acceptor excitation is negligible. ( A) The red curve shows the FRET
relationship FD/FA/H11005/H9270D(A)/H9278FD/((/H9270D(0)/H11002/H9270D(A))/H18528/H9278FA) with /H9270D(0)/H110054.1 and the fluo-
rescence quantum yields of the donor and acceptor /H9021FD/H110050.8 and /H9021FA/H110050.43,
respectively. ( B) The red curve shows the Perrin equation r/H11005r0/(1/H11001/H9270//H9267) with
a fundamental anisotropy r0/H110050.375 and a mean rotational correlation time
/H9267/H110050.63 ns.
18338 /H20841www.pnas.org /H20862cgi /H20862doi /H2086210.1073 /H20862pnas.0800977105 Woz´niak et al.
| 9fa368fc44cd70a30ed489930b67ce05abc4c32b | page_0001 |
flexible fluorophore linkage prohibits direct measurement of
Rmpand the corresponding FRET efficiency Emp(Eq. 4).
Emp/H11005R06
R06/H11001Rmp6 [4]
Comparison of Experimental FRET Data with the MD Simulation.
Based on the distribution of fluorophore positions and their
dipole orientations obtained from the MD simulation, we cal-culated the 3 different FRET efficiency averages. They indeedshow a strong dependence on the fluorophore dynamics (Fig.2C). The slower the dynamics is with respect to the fluorescence
lifetime, the lower is the resulting efficiency. As the calculateddistances significantly depend on the choice of averaging model,we have to determine k
Randkd.
Subensemble and ensemble fluorescence spectroscopy consis-
tently revealed (see Fig. S1 a–e) that both dyes are very mobile. The
donor anisotropy decay is described by a sum of 3 exponentialdecays with the following rotational correlation times
/H9267xand the
components ( r0x): 0.28 ns (0.250), 1.36 ns (0.076) and 14.3 ns (0.014).
Using the fundamental anisotropy of Alexa Fluor 488 r0/H110050.375,
the decay is dominated by fast relaxation rate kR/H110051//H92671/H110053.6/H11003109
s/H110021. The slowest component reflecting mainly the rotation of the
DNA has only a fraction of 4% of the total donor anisotropy decay.The acceptor anisotropy decay under direct excitation can bedescribed by the sum of 2 exponential decays with
/H9267xand ( r0x): 0.682
ns (0.206) and 14.3 ns (0.126) indicating significant local mobility.The rapid rotation of the donor alone is evidence enough foradopting the value of 2/3 for
/H92602.Based on dye-dye quenching in experiments with varying
numbers of intervening base pairs, we can estimate the rateconstant of 2D diffusion on the sterically allowed area definedby a distance of approximately /H110064b pt ob e k
d/H110055/H11003105s/H110021
(A.K.W., C.A.M.S., and F.O., unpublished data). Based on the
FRET efficiencies shown in Fig. 2 B(and Table 2 ) and /H9270D(0)/H11005
4.1 ns, the computed FRET rate constants kFTrange from 7.9 /H11003
108to 8.3 /H11003106s/H110021. The order of the kvalues, kR/H11022kFT/H11022kd
indicates that the isotropic average /H20855E/H20856isoshould describe the
experimentally observed FRET efficiency best. Comparing the3 different averages with the experimental data, we indeed foundthe best agreement for the isotropic average.
FRET Detects Sequence-Dependent Bending of DNA. The ultimate
aim of MFD-FRET measurements is to obtain structural infor-mation. This can only be achieved by deriving distances R
mp
between 2 fixed points from the data, in our case between the
mean fluorophore positions. However, as stated in Comparing
the Experiment with Simple Models , this is not measured in a
FRET experiment. Based on the simulation data, we generatedan empirical polynomial function, that converts the isotropicaverage /H20855E/H20856
isointo Emp, which is needed for the geometric
description ( Fig. S3 ). This function now permits us to calculate
the correct distance between the average donor and acceptor
position from any measured FRET efficiency.
The corrected experimental efficiencies are in reasonable
agreement with the values calculated for a straight B-DNA helixwith a fixed position of the fluorophores, even at small distances(Fig. 2 D). However, systematic deviations centered at 17 bp are
Fig. 2. Comparison of modeled and measured FRET distances. ( A) MD simulation of fluorophores on B-DNA. The distribution of the donor (green) and acceptor
(red) fluorophores is shown for the 21 bp distance, as represented by 5,157 positions of the reference atoms (O7 in Alexa Fluor 488 and C27 in Cy5, respect ively).
The average positions of the fluorophores for all of the donor and acceptor labels are shown as big spheres. ( B) Transfer efficiencies of 12 measured distances
as calculated from the donor lifetime. The standard deviations were obtained from 10 subsequent single-molecule measurements. The helicity of the D NA is
reflected in the data and can be seen compared with a model of linear distance increase with 0.34 nm per base pair and to the mean position model calculatedfrom the MD-simulated fluorophore positions at a B-DNA (see SI Appendix Eq.13). (C) From the dataset of the MD simulation, transfer efficiencies were calculated
by applying the 3 different averaging regimes (Eq. 2–4): dynamic, isotropic, and static averages. On average the mean position model fits best to the data. ( D)
To obtain E
mpfor a straight DNA, we calculated a conversion function ( Emp/H110050.008 /H110010.679 x/H110011.470 x2/H110021.141 x3) by plotting the efficiencies as calculated from
the mean position model against the isotropic average that reflects the experimental data. Here, /H20855E/H20856iso/H11005x(see Fig. S3 ). (E) The bent DNA-structures 1, 2, and
3 are obtained from the program DIAMOD, using the parameters of Gabrielian and Pongor (19), Goodsell and Dickerson (20), and Ulyanov and James (21),respectively. By modeling the average positions ( SI Appendix ) of the fluorophores at the according bases in sequence, we obtain specific distance data
corresponding to the mean position model (bent DNA 1, 2, and 3). ( F) Overlay of B-helix (red) and sequence-dependent structure [Goodsell and Dickerson (21)],
(cyan) of the examined DNA. The bending can be approximated by a single kink angle of 16°.
Woz´niak et al. PNAS /H20841November 25, 2008 /H20841vol. 105 /H20841no. 47 /H2084118339
BIOPHYSICS
| 9fa368fc44cd70a30ed489930b67ce05abc4c32b | page_0002 |
still visible and much larger than the experimental errors. The
deviations are too large to result solely B-DNA helicity, indi-cating that additional structural features are contained in theFRET data.
It is well known that DNA can adopt different conformations.
The most prominent, the A, B, and Z helixes are deduced fromcrystals, and thus reflect average conformations. Beyond this,each DNA sequence shows an individual bending, which hasbeen investigated extensively by other techniques (16–18). Thesestudies resulted in different models providing sets of bendingparameters for any triplet or quartet of consecutive bases, and,thus, can be used to describe the bending of any longer DNAsequence. Based on the 3 alternative sets of bending parametersfrom Gabrielian and Pongor (19), Goodsell and Dickerson (18),and Ulyanov and James (20), we used the open source programDIAMOD (21) to generate 3 structures of our DNA showing thesequence-dependent bending. Although the authors above pro-vide different sequence related bending parameter sets, theseparameters result in a similar overall bending. The differencebetween the structure of our DNA sequence predicted byGoodsell and Dickerson (cyan) and a B-DNA structure (red) isdisplayed in Fig. 2 F. To compare the bent DNA structures with
our data, we modeled the fluorophore position clouds obtainedfrom the MD simulation to the respective bases in the bent DNA(see SI Appendix ). Subsequently, we calculated the FRET effi-
ciencies for the distance between the mean fluorophore posi-tions. Fig. 2 Eshows a superposition of the converted experi-
mental transfer efficiencies and E
mpcalculated for the 3 bent
DNA structures. Without any additional fitting all models within
their variation are in excellent agreement with the experimentaldata. Both the absolute values and the shape of the experimentaltransfer efficiency curve are well described by the mean positionmodel of the bent DNA. Small deviations at shortest distancesof 5 and 7 basepairs originate from acceptor quenching due tofluorophore collisions, which change their photophysical prop-erties as detected by fluorescence correlation spectroscopy (22).In conclusion, single-molecule FRET is sensitive enough todetect DNA bending that can be approximated by a single kink
angle of only 16°. The data presented here clearly show that bytaking into account the correct average for the fluorophoredynamics and their positional variability, molecular distances canbe calculated with an accuracy of a few percent, using the methodof single-molecule multi parameter fluorescence detection.
Kinked DNA. Our single-molecule MFD FRET technique was also
applied to measure kinked DNA of unknown structure. Weinduced kinks into the same DNA sequence as was used aboveby insertion of unpaired adenosines, so-called A bulges (seeMaterial and Methods ). Samples containing 1, 3, and 5 ad-
enosines (A
1,A3, and A 5bulge) in the donor strand of the DNA
above were investigated. By choosing proper donor and acceptor
positions the helical wheel could be studied in 1 or 2 base pairsteps (see Fig. 3 G). Fig. 3 A,C, and Eshow the converted transfer
efficiencies from single-molecule MFD measurements on DNAsamples containing the A
1,A3, and A 5bulges. All 3 samples show
a significantly higher efficiency for the distance of 17 bp com-pared with the 15-bp distance. Because of the cross combinationof the donor and acceptor labeling positions (Fig. 3 B,D, and F),
the relative location of the kink varies for each base pairseparation. If a kink is not located symmetrically between the 2fluorophores, its influence on the interfluorophore distance ismuch smaller than in the symmetric case. Here, donor d3 andacceptor a2 are closer to the kink than d1 and a3. Accordingly,the 15-bp distance (d1-a2) is hardly affected by the kink, whereasthe 17-bp distance (d3-a3) is significantly decreased. This effectbecomes more pronounced with increasing number of bulged As,indicating a strong increase of the kink angle. Finally, for the A
5
bulge, the 17-bp distance (d3-a3) is almost as short as the 11-bpdistance (d3-a2).
Quantitative 3D FRET Structure Analysis of Kinked DNA. To charac-
terize the structural arrangement of the helical segments, wefitted a detailed molecular model to sets of measured distances.This model takes into account the sequence-dependent bending
A
A
A
A
Aa3 a5d3d2 d1
a4a2F∆
bpd1 d2 d3
a2 15 13 11
a3 21 19 17
a5 22 20 18
a4 27 25 23A3bulge
C
10 15 20 250.00.20.40.60.8
101 52 02 54
Separation [base pairs]101 52 02 5EA5bulge
A1bulge
A exp. mean pos. data
kinked helix modelTransfer EfficiencyG
Aa3a5d3d2
d1
a4a2B
AAAa3 a5d3d2
a4a2D
d1
Fig. 3. Fitting DNA structures to the measured distances. ( A,C, and E) FRET-efficiencies of converted experimental data (black) from kinked DNAs. By varying
the torsion angles in the kink site (backbone opposite to the inserted adenosines, see Fig. S4 a) a kinked 3D helix model (gray) was fitted to the converted
experimental data (black). ( B, D, and F) Corresponding 3D structures resulting from the best fit (PDB files are given in SI Appendix ). The labeling positions for
the donor (d) and acceptor (a) are enumerated as described in Materials and Methods . The structures were modeled using bent DNA sequences according to
(18). ( G) Different DNA oligonucleotides labeled with the donor fluorophore (d3 to d1) and acceptor fluorophore (a2 to a5), respectively, are combined to yield
a set of sample molecules with a continuous distance increase (gray line, see Materials and Methods ). The interfluorophore distances in the table are given in
number of separating base pairs /H9004bp.
18340 /H20841www.pnas.org /H20862cgi /H20862doi /H2086210.1073 /H20862pnas.0800977105 Woz´niak et al.
| 9fa368fc44cd70a30ed489930b67ce05abc4c32b | page_0003 |
of the 2 helical arms (18) and fits the kink by the rotation of all
5 backbone torsion angles of the acceptor strand opposite to theposition where the A-bulges are inserted, while assuming thedonor strand as being cleaved at the respective position (see Fig.
S4a). For a geometric description of the overall fitted structure,
we calculated the kink angle,
/H9272, and the angles of rotation, /H9258and
/H9274, of the 2 helical arms around their individual axes. In addition,
we calculated 3 shifts of the acceptor-carrying arm, 1 parallel and2 perpendicular to the axis of the donor-carrying arm (Fig. 4 andFig. S4 b).
The use of a random search algorithm with a rigorous and
quantitative reduced
/H92732(/H9273r2) criterion (see Material and Methods
andSI Appendix ) allowed us to find the best molecular confor-
mations (Fig. 3 B,D, and F) for a fit of the detailed molecular
model to the converted experimental FRET distances. Despitenot having included steric hindrance in the fitting routine for the2 double-stranded DNA segments, structures without any stericconflicts are obtained. The
/H9273r2surfaces (Fig. 4 Aand Fig. S5 )
made it possible to calculate errors for the parameters (angles
and displacement) in the geometric model. For all parameterswell-defined single minima were found, i.e., the FRET dataresulted in unique structures for the different A bulges. The pdbfiles of these structures are provided in SI Appendix .
The complete sets of parameters that describe the kinks
according to the geometric model are listed in Table 1. The kinkangles increase with the number of bulged As. Our values are ingood agreement with conformations that have already beenmeasured with other techniques: Qualitative transient electricbirefringence studies estimate a kink angle of 10–20° in RNA per
unpaired base (23). The kink angle we measured for the A
5bulge
is in good agreement with 73 /H1100611° obtained by detailed NMR
measurements (24). Qualitative FRET measurements con-ducted by Gohlke et al. (25) were done under ensemble condi-
tions and led to an estimation of the kink angle between 81 and105°, using 1 and 2 measured distances (26).
In contrast, we generated a complete 3D model based on kink
angles that were calculated taking the bending of the helical armsinto account. This leads to different helix axes for the donor andacceptor arm. We have already seen bending of /H1101516° of the
linear helix (Fig. 2 f), which now acts as a ‘‘frame’’ for the bulge.
Depending on the relative orientation of the kinking and bend-ing direction, both angles may add up or partially compensate inthe experimentally determined values. Thus, in general, differ-ent sequences in the helical arms may lead to different measuredkink angles for the same number of bulged As.
The reference for rotation of the helical arms and for the
direction of kinking is defined as the direction of the phosphoratom P (see Fig. S4 ) in the strand opposite to the bulged As,
when seen from the Center of the helix. The resulting angles ofrotation reveal an average kink direction given by (
/H9258/H11001/H9274)/2 for
all bulges of /H1101535° with a variation of /H1100610° for the A 1and/H1100620°
for the A 5bulge. Interestingly, for the A 1and the A 5bulges, the
rotation of the donor arm is stronger than that of the acceptorarm, such that the bases before and after the kink overlapstronger than in the B-DNA conformation.
The resulting gap in the bulge-carrying strand is 8, 12, and 15
Å for the A
1,A 3, and A 5bulges, respectively. With 1 base
occupying /H110155 Å in the helical backbone, the gap is well sized to
harbor all of the bulged bases, implying that the backbone itselfcan bulge out when containing more than one A. The transferefficiencies corresponding to the distances calculated from thefitted DNA structures are superimposed in Fig. 3 a, c and e,
showing the high quality of the fit to the A
1and A 3bulge data,
which is also reflected in the resulting minimal /H9273r2of 2.72 and
0.94. This also justifies the assumption of our molecular modelthat only the backbone atoms opposite to the bulged As areforming the kink whereas the rest of the DNA structure remainsunchanged. However, the distances calculated for the A
5bulge
could not be fitted perfectly by our molecular model, as reflectedin the higher
/H9273r2. This indicates that the structures of the DNA
helical arms may be distorted by the inserted As. Such distortinginfluence of additional bases on the double helical structure waspreviously shown by NMR measurements on an A
5bulge in
another DNA sequence, where the bases were found to be tiltedrelatively to the helix axis (25).
Conclusion
Significant advances in absolute single-molecule distance mea-
surements via FRET have been achieved by considering posi-tional and orientational fluorophore dynamics and by takingadvantage of known helical nucleic acid features in combinationwith triangulation methods. These improvements in FRETanalysis allowed us to determine quantitatively DNA bendingand kinking in all 3 dimensions. We emphasize that the givenconversion function, which maps the measured to the correctedFRET efficiency may be applied to any other double-strandedDNA structure with the same fluorophore-linker system in acomparable environment. For that purpose we provide the filesof the simulated fluorophore positions and a detailed applicationprotocol (see SI Appendix ). In view of the fact that protein- or
damage-induced DNA bending has been shown to regulate theassembly and function of DNA-multiprotein complexes, struc-tural FRET studies harbor the potential to provide fascinatinginsights in structural and molecular biology. Moreover, the sameapproach can be directly transferred to the study of the arrange-ment of
/H9251-helices in proteins.
Fig. 4. Kink parameters. ( A) The /H9273r2surface is shown for the kink angle /H9272and
kink direction /H9258induced by the A 3bulge. ( B) Visualization of the parameters
describing the kink. /H9272denotes the kink angle between the donor and the
acceptor axis. /H9258and/H9274denote the angle of rotation around the helical axes of
the donor and the acceptor arm, respectively. The gray lines define the xand
ydirections of the initial coordinate system. (c–e) The shift of the acceptor
helix is given by the components of a trihedron, along the helical axis of thedonor arm (c), perpendicular to the donor and acceptor axis (d), and indirection of the kink (e). Note, that the kink direction is equivalent to the angleof rotation of the donor arm. For a more detailed description see Fig. S4 b.
Table 1. Parameters fitted to the kinked DNA
A1 A3 A5
/H9272,° 3 2 /H1100665 6 /H1100647 3 /H110062
/H9258,° 4 7 /H1100614 34 /H1100665 0 /H110064
/H9274,° 2 6 /H1100611 38 /H1100691 5 /H110065
c, Å 1.6 /H110062 3.7 /H110061.3 5.2 /H110061.1
d, Å /H110020.3/H110063 /H110023.2/H110062.5 1.4 /H110061.9
e, Å /H110023.7/H110061.8 /H110020.16/H110061.9 /H110021.9/H110060.7
/H9273r22.72 0.94 70.6
Determined kink parameters with errors and minimal /H9273r2values for the
analysed kinked DNA structures containing the one-, three- and five-A bulge.For a more detailed discussion of errors, see SI Appendix .
Woz´niak et al. PNAS /H20841November 25, 2008 /H20841vol. 105 /H20841no. 47 /H2084118341
BIOPHYSICS
| 9fa368fc44cd70a30ed489930b67ce05abc4c32b | page_0004 |
Materials and Methods
Samples. All oligonucleotides were synthesized and fluorescently labeled by
IBA. The dyes (Alexa Fluor 488 NHS as donor and Cy5 NHS as acceptor) werecoupled via 5-C
6-aminoallyl-deoxythymidines. The sequence and labeling po-
sitions (superscript numbers) for the donor are 5 /H11032-d(X GGA CTA GTC TAG GCG
AAC GTT TAA GGC (A 1,3,5) GAT CTC T3GT2TT1A CAA CTC CGA), whereas those
for the acceptor are 5 /H11032-d(TCG GAG TTG TAA ACA GAG AT1C GCC TT2A AAC
GT3T5CGC CT4A GAC TAG TCC). The single strands are referred to as d3, d2, and
d1 and a1, a2, a3, a5, and a4, respectively. X denotes over-hanging bases AATTin case of the sequences including the bulged As, which are given in brackets.Four different series of FRET labeled duplexes are generated by combinationof each donor labeled DNA strand with each acceptor carrying strand. Thedouble-strand series without bulged As is labeled at the acceptor positions 1,2, 3, and 4. Those including the A, A
3or A 5bulge are labeled at the acceptor
positions 2, 3, 5, and 4. The length of the DNA of 48 bp ensured a minimum of10 bases adjacent to all fluorophore sites and thus a comparable environmentfor all dyes. The measurements were performed in a buffer with 50 mMTris /H18528HCl, 150 mM NaCl, 1 mM EDTA, 5 mM MgCl
2, 0.4 mM ascorbic acid (pH 7.6).
Ensemble and single-molecule fluorescence spectroscopy. See ref. 7 and SI
Appendix .Simulations. The calculation of the cloud of possible fluorophore positions was
done via MD simulations, using the Gromacs simulation software (27) asdescribed in ref. 28. The model structures were assembled in VMD 1.8.2 (29).
The formulas for calculating the kink and rotation angles and the shifts
between the helical arms are given in SI Appendix . We calculated the errors of
the fitted parameters that correspond to the standard deviation of theexperimental distances by determining the maximal and minimal values for allconformations with a
/H9273r2less than the minimal /H9273r2/H110011. This was done by using
the same algorithm that we used to find the minimal /H9273r2with 2 differences: ( i)
all conformations with a /H9273r2less than the minimal /H9273r2/H110011 were collected and not
further varied and ( ii) the random variation was applied to next 10,000 best
conformations. This procedure was repeated until we obtained a set of at least50,000 conformations with a
/H9273r2less than the minimal /H9273r2/H110011(SI Appendix ).
ACKNOWLEDGMENTS. We thank Suren Felekyan, Volodymyr Kudryavtsev,
and Matthew Antonik for supporting the MFD measurements by writingspecialized data analysis software; Wajih Al-Soufi for helpful advice in datafitting; Stanislav Kalinin for help in the fluorescence anisotropy analysis; RolfWagner for supporting know-how on DNA bending; and Michael Levitt forcarefully reading the manuscript. This work was supported by Bundesminis-terium fu ¨ r Bildung und Forschung Nanotechnology Competition Project
03N8714.
1. Ha T, et al. (1996) Probing the interaction between two single molecules: Fluorescence
resonance energy transfer between a single donor and a single acceptor. Proc Natl
Acad Sci USA 93:6264–6268.
2. Rasnik I, Myong S, Cheng W, Lohman TM, Ha T (2004) DNA-binding orientation and
domain conformation of the E-coli Rep helicase monomer bound to a partial duplexjunction: Single-molecule studies of fluorescently labeled enzymes. J Mol Biol 336:395–
408.
3. Andrecka J, et al. (2008) Single-molecule tracking of mRNA exiting from RNA poly-
merase II. Proc Natl Acad Sci USA 105:135–140.
4. Mekler V, et al. (2002) Structural organization of bacterial RNA polymerase holoen-
zyme and the RNA polymerase-promoter open complex. Cell108:599–614.
5. Rothwell PJ, et al. (2003) Multi-parameter Single-molecule Fluorescence Spectroscopy
reveals Heterogeneity of HIV-1 Reverse Transcriptase:primer/template Complexes.Proc Natl Acad Sci USA 100:1655–1660.
6. van der Meer BW, Cooker G, Chen SY (1994) Resonance Energy Transfer: Theory and
Data (VCH Publishers, New York).
7. Widengren J, et al. (2006) Single-molecule detection and identification of
multiple species by multiparameter fluorescence detection. Anal Chem 78:2039–
2050.
8. Dale RE, Eisinger J, Blumberg WE (1979) Orientational freedom of molecular probes—
Orientation factor in intra-molecular energy transfer. Biophys J 26:161–193.
9. Clegg RM, Murchie AIH, Zechel A, Lilley DMJ (1993) Observing the helical geometry of
double-strand DNA in solution by fluorescence energy transfer. Proc Natl Acad Sci USA
90:2994–2998.
10. Jares-Erijman EA, Jovin TM (1996) Determination of DNA helical handedness by
fluorescence resonance energy transfer. J Mol Biol 257:597–617.
11. Lee NK, et al. (2005) Accurate FRET measurements within single diffusing biomolecules
using alternating-laser excitation. Biophys J 88:2939–2953.
12. Toth K, Sauermann V, Langowski J (1998) DNA curvature in solution measured by
fluorescence resonance energy transfer. Biochemistry 37:8173–8179.
13. Schuler B, Lipman EA, Steinbach PJ, Kumke M, Eaton WA (2005) Polyproline and the
’’spectroscopic ruler’’ revisited with single-molecule fluorescence. Proc Natl Acad Sci
USA 102:2754–2759.
14. Haas E, Katchalski-Katzir E, Steinberg IZ (1978) Effect of the orientation of donor and
acceptor on the probability of energy transfer involving electronic transitions of mixedpolarization. Biochemistry 17:5064–5070.15. Parkhurst KM, Parkhurst LJ (1995) Donor-acceptor distance distributions in a double-
labeled fluorescent oligonucleotide both as a single strand and in duplexes. Biochem-
istry 34:293–300.
16. Peck LJ, Wang JC (1981) Sequence dependence of the helical repeat of DNA in solution.
Nature 292:375–378.
17. Sarai A, Mazur J, Nussinov R, Jernigan RL (1988) Origin of DNA helical structure and its
sequence dependence. Biochemistry 27:8498–8502.
18. Goodsell DS, Dickerson RE (1994) Bending and curvature calculations in B-DNA. Nucleic
Acids Res 22:5497–5503.
19. Gabrielian A, Pongor S (1996) Correlation of intrinsic DNA curvature with DNA prop-
erty periodicity. FEBS Lett 393:65–68.
20. Ulyanov NB, James TL (1995) Statistical analysis of DNA duplex structural features.
Nuclear Magn Reson Nucl Acid 261:90–120.
21. Dlakic M, Harrington RE (1998) DIAMOD: Display and modeling of DNA bending.
Bioinformatics 14:326–331.
22. Widengren J, Schweinberger E, Berger S, Seidel CAM (2001) Two new concepts to
measure fluorescence resonance energy transfer via fluorescence correlation spectros-copy: Theory and experimental realizations. J Phys Chem A 105:6851–6866.
23. Zacharias M, Hagerman PJ (1995) Bulge-induced bends in RNA: Quantification by
transient electric birefringence. J Mol Biol 247:486–500.
24. Dornberger U, Hillisch A, Gollmick FA, Fritzsche H, Diekmann S (1999) Solution struc-
ture of a five-adenine bulge loop within a DNA duplex. Biochemistry 38:12860–12868.
25. Gohlke C, Murchie AIH, Lilley DMJ, Clegg RM (1994) Kinking of DNA and RNA helices
by bulged nucleotides observed by fluorescence resonance energy transfer. Proc Natl
Acad Sci USA 91:11660–11664.
26. Stu ¨ hmeier F, Hillisch A, Clegg RM, Diekmann S (2000) Fluorescence energy transfer
analysis of DNA structures containing several bulges and their interaction with CAP. J
Mol Biol 302:1081–1100.
27. Berendsen HJC, Spoel D, Drunen R (1995) GROMACS: A message-passing parallel
molecular dynamics implementation. Comp Phys Comm 91:43–56.
28. Wozniak AK, et al. (2005) Detecting protein-induced folding of the U4 snRNA kink-turn
by single-molecule multiparameter FRET measurements. RNA 11:1545–1554.
29. Humphrey W, Dalke A, Schulten K (1996) VMD: Visual molecular dynamics. JM o l
Graphics 14:33–38.
18342 /H20841www.pnas.org /H20862cgi /H20862doi /H2086210.1073 /H20862pnas.0800977105 Woz´niak et al.
| 9fa368fc44cd70a30ed489930b67ce05abc4c32b | page_0005 |
Improving Selection Detection with Population Branch
Statistic on Admixed Populations
Burak Yelmen1,2,*, Davide Marnetto1, Ludovica Molinaro1,2, Rodrigo Flores1, Mayukh Mondal1,†,a n d
Luca Pagani1,3,*,†
1Institute of Genomics, University of Tartu, Estonia
2Institute of Molecular and Cell Biology, University of Tartu, Estonia
3Department of Biology, University of Padova, Italy
†These senior authors contributed equally to this work.
*Corresponding authors: E-mails: burakyelmen@gmail.com; lp.lucapagani@gmail.com
Accepted: 22 February 2021
Abstract
Detecting natural selection signals in admixed populations can be problematic since the source of the signal typically dates back prior
to the admixture event. On one hand, it is now possible to study various source populations before a particular admixture thanks tothe developments in ancient DNA (aDNA) in the last decade. However, aDNA availability is limited to certain geographical regions and
the sample sizes and quality of the data might not be sufficient for selection analysis in many cases. In this study, we explore possible
ways to improve detection of pre-admixture signals in admixed populations using a local ancestry inference approach. We usedmasked haplotypes for population branch statistic (PBS) and full haplotypes constructed following our approach from Yelmen et al.
(2019) for cross-population extended haplotype homozygosity (XP-EHH), utilizing forward simulations to test the power of our
analysis. The PBS results on simulated data showed that using masked haplotypes obtained from ancestry deconvolution instead ofthe admixed population might improve detection quality. On the other hand, XP-EHH results using the admixed population were
better compared with the local ancestry method. We additionally report correlation for XP-EHH scores between source and admixed
populations, suggesting that haplotype-based approaches must be used cautiously for recently admixed populations. Additionally,we performed PBS on real South Asian populations masked with local ancestry deconvolution and report here the first possible
selection signals on the autochthonous South Asian component of contemporary South Asian populations.
Key words: natural selection, PBS, XP-EHH, local ancestry inference, admixed populations, South Asia.
Significance
Detecting natural selection in recently admixed populations can be difficult due to the obscurity of the source of
selection signals. In this study, we used local ancestry inference methods to obtain ancestry assigned haplotypes out ofsimulated admixed populations and reported improvement in detecting selection signals before the admixture event
using these haplotypes instead of the admixed ones. We additionally demonstrated that methods utilizing haplotype
structure to detect selection must be used cautiously for recently admixed populations. Finally, we applied our ap-proach to real South Asian genomes to report first possible selection signals for autochthonous South Asian
populations.
/C223The Author(s) 2021. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution.
ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommonsAttributionNon-CommercialLicense( http://creativecommons.org/licenses/by-nc/4.0/ ),whichpermitsnon-
commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please conta ct journals.permissions@oup.com
Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 1GBE | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0000 |
Introduction
Investigating signatures of selection in admixed populations is
challenging due to the fact that independent signals fromsource populations may be obscured by each other ( Huerta-
S/C19anchez et al. 2014 ;Galaverni et al. 2017 ;Pierron et al.
2018 ). Although recent advancements in ancient DNA
(aDNA) grants us with the opportunity to independently study
source populations prior to the admixture event, detectingselection signals in populations with admixture backgroundis still widely an unexplored and challenging endeavor, espe-cially in areas where aDNA preservation is limited. When con-ventional methods like population branch statistics (PBS; Yi et
al. 2010 ) or cross-population extended haplotype homozy-
gosity (XP-EHH; Sabeti et al. 2007 ) are utilized to detect pos-
sible signals on admixed populations, it is not easy to resolve
whether the candidate signals are due to selection acting afterthe admixture event or before and in the latter case, on whichsource population the selection event took place. Here, weparticularly concentrated on detecting selection signals beforeadmixture. We postulate that applying local ancestry infer-ence as a preliminary step and then searching for signatures
of selection within each set of source haplotypes may greatly
decrease false positives and may help assign the observedsweeps to the correct ancestral population. In a previousstudy, we used a local ancestry deconvolution-based ap-proach to create surrogates of the two main ancestral com-ponents of contemporary South Asian human genomes toshed light into the demographic history of these highly
admixed human populations ( Yelmen et al. 2019 )a n dt oa d -
dress patterns of selection after admixture. Here, we test ouridea through forward simulations and investigate the possibleuse of these surrogates for frequency based (PBS) and haplo-type based (XP-EHH) selection tests in humans or any otherdiploid organism.
Results
Comparison of PBS Scores Based on Forward Simulations
We used forward simulations to create an admixed humanpopulation by mixing available genomes from European(French) and East Asian (Han) individuals for 200 generations(assuming 30 years per generations, which will be around sixthousand years ago). Notably, European-East Asian popula-tion splits may mimic the split between the North and Southgenomic components found within contemporary South
Asian populations (see Yelmen et al. for a full description of
South Asian demography). For this reason, we decided torespectively label “N” and “S” the European and East Asianancestries retrieved from this simulation test, to keep consis-tency with the subsequent real case scenario applied on SouthAsian human genomes. Utilizing both PCAdmix ( Brisbin et al.
2012 )a n dE L A I( Guan 2014 ) and using German and Japanese
genomes (two populations deemed to be genetically close toFrench and Han, respectively) as reference populations for
local ancestry deconvolution, we formed MASK_S (German-assigned chunks masked out) and MASK_N (Japanese-assigned chunks masked out) populations from the simulateddata. Then, we performed PBS on both (masked and naive)and compared with the results obtained from the originalsource populations (French and Han), which were used as atrue dataset, set as our standard (see Materials and Methods).Overall SNP by SNP correlation showed significantly higher
correlation between the PBS scores of original source popu-
lations and masked populations compared to the unmaskedadmixed populations ( Supplementary figures S1 and S2 a).
However, when we only analyzed SNPs with PBS scores abovethe 99th percentile (with SNPs selected based on Han andFrench source population PBS scores), the ELAI approachretained higher correlation whereas PCAdmix performedpoorly ( figure 1 ,Supplementary figure S2 b). The poor perfor-
mance of PCAdmix was due to false negatives, as can be seen
from the low PBS values accumulating along the xaxis.
Additionally, both masking approaches performed bettercompared with the naive populations when we concentratedonly on the top 50 scoring SNPs except for MASK_N_ELAI(Supplementary table S1 a).
Aside from SNP by SNP comparison, we also compared
window-based mean PBS scores as widely used in selection
studies ( Huerta-S /C19anchez et al. 2013 ). Overall, precision and
true positive rate indicators were higher in masked popula-tions compared with the naive population ( table 1 ), showing
that our approach helps retrieving a lower fraction of falsesignals compared with simply studying the admixed genomes.Although setting signal threshold to 99.9% reduced the hitssignificantly which, in return, made the comparison difficult(Supplementary table S2 ), 99.5% threshold results mostly
showed improvement using masked populations.
Comparison of XP-EHH Scores Based on Forward
Simulations
Since XP-EHH cannot be performed with missing chunks,
masked populations could not be utilized in a straightforwardmanner. We, therefore, created ancestral random breeders(ARBs, see methods for more details), full genomes with nogaps made by combining together masked haplotypes frommultiple individuals, as described in Yelmen et al. (2019) ,
resulting in ARB_N and ARB_S individuals. This approach
assigns ancestries to the chunks of admixed genomes and
combines matching ancestries to create whole genomes,bypassing the lack of a proper imputation panel when eitherof the ancestry is no longer available in unadmixed form.Therefore, they can be seen as “Frankensteins” of stitchedhaplotypes of the same ancestry. Similar to the PBS analysis,we performed XP-EHH on ARB (filled in after local ancestrydeconvolution) and naive (prior to any type of processing)genomes, and compared these with the results from FrenchYelmen et al. GBE
2Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0001 |
and Han source populations (see Materials and Methods). This
time, ARB populations performed worse in comparison to
admixed naive population for position by position XP-EHH
scores ( Supplementary fig. S3 ), although both ARB and naive
populations performed very poorly when we checked top 50
scoring SNPs in comparison to the source populations
(Supplementary table S1b ).
PBS on Real South Asian Populations
Given that applying XP-EHH on admixed populations proved
to be imprecise using either ARB or admixed genomes, wethen resorted to PBS as our only viable approach to detect
signatures of selection in a case study: contemporary South
Asian human populations. Genomic composition of SouthAsians can be characterized in a broad perspective as an ad-mixture of West Eurasian and South Asian components ( Reich
et al. 2009 ;Chaubey et al. 2011 ;Metspalu et al. 2011 ;
Moorjani et al. 2013 ;Basu et al. 2016 ;Lazaridis et al. 2016 ;
Damgaard et al. 2018 ;Pathak et al. 2018 ;Yelmen et al.
2019 ). Even though recent studies shed more light into this
composition ( Narasimhan et al. 2019 ), there is still no available
unadmixed aDNA attributed to the South Asian componentfor a selection scan to be carried out. In our previous study, wechecked for local admixture imbalance between ancestralcomponents of South Asians and detected some possible se-
lection signals pointing to selective pressures acting after the
admixture event ( Yelmen et al. 2019 ). In this study, we instead
focused on events that took place prior to the admixture ofthe two genomic components, and reported the first time
possible selection signals for the South Asian component of
contemporary South Asians ( fig. 2). We additionally report the
genes present within 50-kb windows with possible positivesignals ( Supplementary table 3 ) and top 20 scoring SNPs with
related genes ( Supplementary table S4 ). We stress that the
PBS analysis could also detect postadmixture selection, forexample, by searching for overlap between the signals foundon each ancestry independently; however, postadmixture im-balance ( Yelmen et al. 2019 ) remains the strategy of choice
FIG.1 .—SNP by SNP PBS comparison for SNPs with PBS values above 99% threshold (SNPs selected based on Han and French source population scores)
using Spearman’s correlation. ( a) Han vs naive (correlation coefficient: 0.409, 95% confidence interval: 0.377–0.441, p-value<2.2e/C016,n¼2,529), Han vs
MASK_S_ELAI (correlation coefficient: 0.510, 95 percent confidence interval: 0.479–0.542, p-value<2.2e/C016,n¼2,529), Han versus MASK _S_PCAdmix
(correlation coefficient: 0.210, 95% confidence interval: 0.169–0.252, p-value<2.2e/C016,n¼2,529). ( b)French vs naive (correlation coefficient: 0.400,
95% confidence interval: 0.367–0.436, p-value<2.2e/C016,n¼2,530), French versus MASK_N_ELAI (correlation coefficient: 0.452, 95% confidence
interval: 0.422–0.483, p-value <2.2e/C016,n¼2,530), French versus MASK_N_PCAdmix (correlation coefficient: 0.394, 95% confidence interval:
0.354–0.432, p-value<2.2e/C016,n¼2,530).
Table 1
Comparison of Possible Selection Signals Based on Mean PBS Scores
(Above 99.5% Noted as Positive) for 50 kb Windows with TP, FP, FN,
TPR (Measuring the Fraction of Correctly identified positives), and FDR,Measuring Expected Fraction of FPs) Indicators for Each Tested Population
(Admixed Naive along with PCAdmix and ELAI-Masked Populations, see
Materials & Methods for details) Compared with True Source Population
TP FP FN TPR FDR
Naive 25 39 29 0.46 0.61
MASK_S_ELAI 27 23 27 0.50 0.46MASK_S_PCAdmix 25 18 29 0.46 0.42Naive 50 33 60 0.45 0.40MASK_N_ELAI 41 21 69 0.37 0.34MASK_N_PCAdmix 68 38 42 0.62 0.36Improving Selection Detection with Population Branch Statistic on Admixed Populations GBE
Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 3 | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0002 |
given the short evolutionary time elapsed after the admixture
event.
Discussion
In our study, we explored the possibility of improving selectiontests on admixed populations using local ancestry deconvolu-
tion approaches. Following our previous work ( Yelmen et al.
2019 ), we created surrogates of components from a simu-
lated admixed population both in the form of masked andreconstructed genomes. Knowing the true source populationswhich we merged to create the admixture, we were able tocompare selection scores and observed an increased perfor-
mance in the site-specific PBS using masked populations in-
stead of the admixed one. However, it is important to notethat even with this improvement, the scores are far frommatching the ones obtained from the true source populationin our simulations ( Supplementary Table 1 ). Additionally, al-
though we calculated the true positive rate and precision forwindow-based comparison, due to the vast number of true
negatives, it was not possible to obtain other meaningful
indicators such as accuracy or false positive rate.
On the other hand, haplotype aware XP-EHH analyses
revealed that both reconstructed and admixed genomes per-formed poorly compared with the benchmark(Supplementary table 1b ), with the naive approach beingslightly better than our proposed solution in the case of
haplotype-based methods. This might be an expected out-
come since combining chunks from different individual
genomes disrupts haplotype integrity ( Yelmen et al. 2019 ),
although it should still be seen as a strategy to be preferredover simple imputation in the absence of a suitable referencepanel. Nonetheless, we found some regions above the thresh-old level ( >2) in the admixed dataset, though there we never
simulated any selection (after admixture). This suggests thatthe selection signal found in an admixed population using XP-EHH (or similar selection tests) does not necessarily advocatefor the selection to have happened in that admixed popula-tion. It might be a remnant effect coming from source pop-ulations where the variant was already selected. Moreover,some signals detected in the admixed populations were notpresent in the source populations ( Supplementary fig. 4 ).
These findings suggest that XP-EHH and haplotype-awaremethods are probably not a good option for selection analysisin admixed populations.
In addition to the analyses on simulated data, we also
performed PBS on real masked genomes and reported thefirst possible selection signals for the South Asian componentof contemporary South Asian populations ( fig. 1 ,
Supplementary tables S2 and S3 ). Some possible signals in-
cluded high-scoring SNPs from chromosome 6 within 50 kb ofHLA-G andHLA-F andTRIM31 andTRIM40 genes related to
FIG.2 .—PBS on masked South Asian genomes (MASK_S) with the dashed line marking 99.9% threshold. Genes within 50-kb range of the highest
peaks are annotated.Yelmen et al. GBE
4Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0003 |
immune system ( Ishitani et al. 2003 ;Rajagopalan and Long
2012 ;Fu et al. 2017 ;Liu et al. 2017 ;Zhao et al. 2017 ;Lin and
Yan 2019 )a l o n gw i t h KCTD6 related to sweet taste signaling
pathway ( Liu et al. 2013 ),ACOX2 related to branched fatty
acid processing ( Bjørklund et al. 2015 ;Vilarinho et al. 2016 ),
DOK5 coding for adapter proteins involved in signal transduc-
tion ( Favre et al. 2003 ), and CDH4 that is thought to be in-
volved in brain segmentation and neuronal growth ( Babb
et al. 2005 ). These highlight for the first time putative selec-
tion signals that took place in the autochthonous South Asianpopulation, as yet unsampled in its unadmixed form.However, it is important to note that signals related to HLA
genes should be interpreted very cautiously due to difficultiesrelated to variant calling and genotyping for that region.
Furthermore, in our previous work ( Yelmen et al. 2019 ), we
found this region to be under unbalanced admixture between“North” and “South” South Asian components, with“North” haplotypes to appear preferred over “South”haplotypes.
In conclusion, our work shows that applying selection
scans on admixed populations of any diploid organism with
no prior deconvolution yields several off target results, while apreliminary local ancestry deconvolution step may help im-prove the detection of true signals, in the case of PBS. Onthe other hand, more research is required to assess the effectof different local ancestry methods on detecting true selectionsignals since we detected varying results between PCAdmixand ELAI applications. Our results also show that haplotype-
aware methods to detect selection may be severely impaired
in presence of recent admixture between highly divergentpopulations. Overall, our results should inform future studiesin the field to be cautious when reporting selection scans fromhuman or other organisms where recent admixture has beendetected, with particular reference to those populationswhere at least one of the admixing sources is no longer avail-
able in its unadmixed form due to complete assimilation or
extinction (in case of wild species) of that particular group.
Materials and Methods
Samples and Simulations
We used SNP array data of French, Han, Japanese, Yoruba ( Li
et al. 2008 ), and German ( Yunusbayev et al. 2015 ) popula-
tions along with masked South Asian populations (BrahminGujarati, Gujarati, Khatri, Maratha, Pallan, Chamar, Dharkar,
Kanjar, Gujjar, and Ror) ( Metspalu et al. 2011 ;Basu et al.
2016 ;Pathak et al. 2018 ) with same sample sizes as described
inYelmen et al. 2019 (seeSupplementary table S5 ). A simu-
lated admixed population consisting of 500 samples was cre-ated via admix-simu ( Williams 2016 ) using French and Han as
admixing groups. First, we only kept positions whose minorallele frequency is more than 1% (using vcftools –maf com-
mand), remove all the indels and kept only SNPs (usingvcftools –remove-indels) and kept only the first biallelic posi-
tion in case of multiple allele (using bcftools –norm -d all).
After filtering, we used admix-simu to simulate the admixedpopulations, which were admixed 200 generations ago with50/50 contributions from the two sources.
Local Ancestry Inference
Local ancestry inference was performed with PCAdmix(Brisbin et al. 2012 ) setting window size to ten SNPs after
default LD pruning ( r
2>0.8, based on a built-in window
size), a value suggested by the software’s developers as the
smallest size for reliable resolution. Additionally, we only usedwindows assigned to ancestral proxies with posterior proba-bilities (provided in the PCAdmix .fbk output) higher than0 . 9 5 ,b a s e do ns i m u l a t i o nt e s t sd e s c r i b e di n Yelmen et al.
2019 . For the simulated admixture events we used
Germans and Japanese as reference groups, given thatFrench and Han were used as mixing populations. For theSouth Asian real populations, we used French and Paniya asproxies for “North” South Asian (N) and “South” South Asian
(S) ancestry donors, respectively, following Yelmen et al.
2019 .F o rE L A I( Guan 2014 ), we used -mg 100 to define
the admixture event that happened 100 generations ago.We find that using -mg 200 gives spurious results, although
the true admixture in this case was 200 generations ago.
Further we used -s 20 for having 20 steps for EM. We ran10 times independently for ELAI and then averaged it out, andused the same source populations described for the PCAdmixapproach.
Haplotype Masking
For each local ancestry method, we kept the regions assignedto either ancestry with at least 95% confidence and marked
all the other regions as unknown or unassigned. Based on
ancestry assigned genomic chunks, two sets (PCAdmix andELAI) of 500 MASK_N and 500 MASK_S individuals werecreated to be used in PBS analysis. These haploid individualsretained information only for the confidently assigned sites
and showed gaps or “NA” for the rest of their genome.
Ancestral Random Breeders
Additionally, to overcome issues introduced by the missingdata and by the lack of suitable imputation panels for the
“South” South Asian component, we generated 20 ARBsfor each ancestry, using ELAI as the local ancestry method,to be used in XP-EHH analysis. ARBs were created for each Nand S ancestry, by taking all the MASK_N or MASK_S individ-
uals and replenishing the masked-out haplotypes by randomly
picking (with replacement) a nonmasked haplotype from an-other donor within the same ancestry assignation at that lo-cus. This process hence created a number of ARB haploidindividuals which feature the genetic makeup of the originalImproving Selection Detection with Population Branch Statistic on Admixed Populations GBE
Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 5 | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0004 |
individual used as a scaffold, and, where not available, a ran-
dom set of haplotypes from a given ancestry drawn from that
individual’s population. The reconstructed ARB populationcan therefore be seen as a set of random breeders, to beconsidered as the best available proxy to the actual ancestrysource within the studied population. To control for fluctua-
tions in allele frequencies introduced by this drawing ap-
proach, we worked only on a small number of recipientARBs (20 individuals, from the simulated populations orfrom the whole set of available South Asian genomes) byretaining only positions for which a minimum number of avail-
able donor haplotypes were available in a given ancestry sta-
tus to minimize the donation of the same haplotype tomultiple ARBs, and by maximizing the length of the donatedhaplotype and its affinity to the surrounding sequences of agiven receiving ARB, to minimize the number of ancestry
switches artificially introduced by the ARB making process.
This method has been initially described in Yelmen et al.
(2019) to which we refer for further details and testing.
Population Branch Statistic
We used scikit-allel package ( Miles et al. 2020 ) for calculating
PBS score for each available position using allel.pbs functionwith window_size ¼1 and window_step ¼1. Two different
sets of analyses were performed for simulated data. PBS for
MASK_N_ELAI, MASK_N_PCAdmix, naive (as an admixture ofFrench and Han), and French group was performed usingYoruba and Japanese outgroups [PBS(N, Japanese, Yoruba)].PBS for MASK_S_ELAI, MASK_S_PCAdmix, naive (as an ad-
mixture of French and Han) and Han group was performed
using Yoruba and German outgroups [PBS(S, German,Yoruba)]. For window-based analysis, genomes were dividedinto 50-kb regions and mean PBS was calculated for eachregion. We defined 99.9% and 99.5% thresholds as possible
selection signals for each region.
Cross -Population Extended Haplotype Homozygosity
We used scikit-allel package ( Miles et al. 2020 ) for calculating
XP-EHH score for each available position using default param-
eters with allel.xpehh function. Two different sets of analyseswere performed for simulated data. XP-EHH for ARB_N, naive(as an admixture of French and Han), and French group wasperformed using Japanese outgroup. XP-EHH for ARB_S, na-
ive (as an admixture of French and Han), and Han group was
performed using German outgroup.
Computing Correlation Coefficients and Sensitivity/
Specificity Measures
We calculated Spearman’s correlation coefficient with R ver-
sion 3.6.3 ( Development Core Team R 2020 ) to assess the SNP
by SNP correlation of PBS and XP-EHH scores. We used boot-strapping with 1,000 replicates to acquire 95 percentconfidence intervals. For 50 kb window analyses, we calcu-
lated the true positive rate (TPR) and false discovery rate (FDR)
as
TPR¼TP
TPþFN
FDR¼FP
FPþTP
where TP is true positives, FN is false negatives, and FP is false
positives.
Data Availability
All data used in this article are publicly available from the
original publications and on https://evolbio.ut.ee link. All
data used for the analyses were obtained from literature, no
new sample data were collected.
Acknowledgements
This work was supported by the European Union through theEuropean Regional Development Fund (Project No. 2014-
2020.4.01.16-0024, MOBTT53: L.P., L.M., D.M., B.Y.;
Project No. MOBEC008: M.M.); Horizon 2020 research and
innovation programme under grant no. 810645: MM; the
Estonian Research Council grant PUT (PRG243): L.P. Thanks
to High Performance Computing Center of the University of
Tartu for providing computational resources.
Literature Cited
Babb SG, et al. 2005. Zebrafish R-cadherin (Cdh4) controls visual system
development and differentiation. Dev Dyn. doi:10.1002/dvdy.20431.
Basu A, Sarkar-Roy N, Majumder PP. 2016. Genomic reconstruction of the
history of extant populations of India reveals five distinct ancestral
components and a complex structure. Proc Natl Acad Sci Usa.
113(6):1594–1599. doi:10.1073/pnas.1513197113.
Bjørklund SS, et al. 2015. Expression of an estrogen-regulated variant
transcript of the peroxisomal branched chain fatty acid oxidase
ACOX2 in breast carcinomas. BMC Cancer. 15(1):1–13.doi:10.1186/s12885-015-1510-8.
Brisbin A, et al. 2012. PCAdmix: principal components-based assignment
of ancestry along each chromosome in individuals with admixed an-
cestry from two or more populations. Hum Biol. 84(4):343–364.doi:10.3378/027.084.0401.
Chaubey G, et al. 2011. Population genetic structure in indian austroasiatic
speakers: the role of landscape barriers and sex-specific admixture.
Mol Biol Evol. 28(2):1013–1024. doi:10.1093/molbev/msq288.
Damgaard Peter de Barros, et al. 2018. The first horse herders and the
impact of early Bronze Age steppe expansions into Asia. Science (80-).
360(6396): eaar7711.
Development Core Team R. 2020. R: A language and environment for
statistical computing. Vienna (Austria): R Foundation for StatisticalComputing. Available from: https://www.r-project.org/ .A c c e s s e d
February 1, 2021.Yelmen et al. GBE
6Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0005 |
Favre C, et al. 2003. DOK4 and DOK5: new dok-related genes expressed
in human T cells. Genes Immun. 4(1):40–45. doi:10.1038/
sj.gene.6363891.
Fu B, et al. 2017. Natural Killer Cells Promote Fetal Development through
the Secretion of Growth-Promoting Factors. Immunity.
47(6):1100–1113.e6. doi:10.1016/j.immuni.2017.11.018.
Galaverni M, et al. 2017. Disentangling timing of admixture, patterns of
introgression, and phenotypic indicators in a hybridizing Wolf popu-
lation. Mol Biol Evol. 34(9):2324–2339. doi:10.1093/molbev/msx169.
Guan Y. 2014. Detecting structure of haplotypes and local ancestry.
Genetics. 196(3):625–642. doi:10.1534/genetics.113.160697.
Huerta-S /C19anchez E, et al. 2013. Genetic Signatures Reveal High-Altitude
Adaptation in a Set of Ethiopian Populations. Mol Biol Evol.
30(8):1877–1888. doi:10.1093/molbev/mst089.
Huerta-S /C19anchez E, et al. 2014. Altitude adaptation in Tibetans caused by
introgression of Denisovan-like DNA. Nature. 512(7513):194–197.
doi:10.1038/nature13408.
Ishitani A, et al. 2003. Protein Expression and peptide binding suggest
unique and interacting functional roles for HLA-E, F, and G inmaternal-placental immune recognition. J Immunol.
171(3):1376–1384.
Lazaridis I, et al. 2016. Genomic insights into the origin of farming in the
ancient Near East. Nature. 536(7617):419–424. doi:10.1038/
nature19310.
Li JZ, et al. 2008. Worldwide Human Relationships Inferred from Genome-
Wide Patterns of Variation. Science. 319(5866):1100–1104.
Lin A, Yan WH. 2019. The emerging roles of human leukocyte antigen-F in
immune modulation and viral infection. Front Immunol. 10:964.
doi:10.3389/fimmu.2019.00964.
Liu B, et al. 2017. The ubiquitin E3 ligase TRIM31 promotes aggregation
and activation of the signaling adaptor MAVS through Lys63-linked
polyubiquitination. Nat Immunol. 18(2):214–224. doi:10.1038/
ni.3641.
Liu Z, Xiang Y, Sun G. 2013. The KCTD family of proteins: structure,
function, disease relevance. Cell Biosci. 3(1):45.doi:10.1186/2045-
3701-3-45.
Metspalu M, et al. 2011. Shared and unique components of human pop-
ulation structure and genome-wide signals of positive selection in
South Asia. Am J Hum Genet. 89(6):731–744. doi:10.1016/
j.ajhg.2011.11.010.
Miles A, et al. 2020. cggh/scikit-allel: v1.3.2. Available from: 10.5281/
zenodo.3976233. Accessed December 1, 2020.Moorjani P, et al. 2013. Genetic evidence for recent population mixture in
India. Am J Hum Genet. 93(3):422–438. doi:10.1016/
j.ajhg.2013.07.006.
Narasimhan VM, et al. 2019. The formation of human populations in
South and Central Asia. Science. 365(6457):eaat7487. doi:10.1126/
science.aat7487.
Pathak AK, et al. 2018. The Genetic Ancestry of Modern Indus Valley
Populations from Northwest India. Am J Hum Genet. 103(6):918–929.
doi:10.1016/j.ajhg.2018.10.022.
Pierron D, et al. 2018. Strong selection during the last millennium for
African ancestry in the admixed population of Madagascar. NatCommun. 9(1):1–9. doi:10.1038/s41467-018-03342-5.
Rajagopalan S, Long EO. 2012. Cellular senescence induced by CD158d
reprograms natural killer cells to promote vascular remodeling. ProcNatl Acad Sci U S A. 109(50):20596–20601. doi:10.1073/
pnas.1208248109.
Reich D, Thangaraj K, Patterson N, Price AL, Singh L. 2009. Reconstructing
Indian population history. Nature. 461(7263):489–494. doi:10.1038/
nature08365.
Sabeti PC, The International HapMap Consortium, et al. 2007. Genome-
wide detection and characterization of positive selection in human
populations. Nature. 449(7164):913–918. doi:10.1038/nature06250.
Vilarinho S, et al. 2016. ACOX2 deficiency: a disorder of bile acid synthesis
with transaminase elevation, liver fibrosis, ataxia, and cognitive impair-ment. Proc Natl Acad Sci U S A. doi:10.1073/pnas.1613228113.
Williams A. 2016. admix-simu: admix-simu: program to simulate admix-
ture between multiple populations. Available from: 10.5281/zen-
odo.45517. Accessed November 1, 2020.
Yelmen B, et al. 2019. Ancestry-Specific Analyses Reveal Differential
Demographic Histories and Opposite Selective Pressures in Modern
South Asian Populations. Mol Biol Evol. 36(8):1628–1642.
doi:10.1093/molbev/msz037.
Yi X, et al. 2010. Sequencing of 50 Human Exomes Reveals Adaptation to
High Altitude. Science. 329(5987):75–78.
Yunusbayev B, et al. 2015. The Genetic Legacy of the Expansion of Turkic-
Speaking Nomads across Eurasia. PLoS Genet. 11(4):1–24.
Zhao C, et al. 2017. The E3 Ubiquitin Ligase TRIM40 Attenuates Antiviral
Immune Responses by Targeting MDA5 and RIG-I. Cell Rep.
21(6):1613–1623. doi:10.1016/j.celrep.2017.10.020.
Associate editor: Dr. Tanja SlotteImproving Selection Detection with Population Branch Statistic on Admixed Populations GBE
Genome Biol. Evol. 13(4) doi:10.1093/gbe/evab039 Advance Access publication 26 February 2021 7 | bb2d5edfd19092b51abf8f617825af3d6bbef83e | page_0006 |
BRITISHMEDICALJOURNALVOLUME2818NOVEMBER 1980 1267
Medicine andMathematics
Statistics andethicsinmedical research
Studydesign
DOUGLAS GALTMAN
Theterm"design" encompasses allthestructural aspects ofa
study,notably thedefinition ofthestudysample, sizeofsample,
method oftreatment allocation, typeofstatistical design
(randomised, cross-over, sequential, etc),andchoiceofoutcome
measures. Theimportance ofthisstage cannot beover-
emphasised since noamount ofcleveranalysis laterwillbe
abletocompensate formajordesignflaws.InthisarticleIwill
consider therelation between designandethicsinobservational
studies andclinicaltrials,butIwilldefertheproblem ofsample
sizeuntilthenextarticle.
Observational studies
Inobservational studies datafrom asample ofindividuals
areused,eitherimplicitly orexplicitly, tomakeinferences
aboutthepopulation ofinterest, such asmenaged20-65,
hypertensives, orpregnant women. Forthisextrapolation to
bevalid,itisessential thatthedataobtained areasrepresentative
ofthepopulation aspossible.
Thisusually entails sometypeofrandom sampling ofsubjects,
forwhich aready-made listofthewholepopulation ofinterest
(asampling frame)isneeded. Suchlists,however, maybeout
ofdate(electoral registers) orinaccurate (doctors' listsof
patients), inwhich casetheirusecanleadtomisleading results.
Furthermore, itisoftendesirable toimprove therepresentative-
nessofthesample bysampling separately fromdifferent
subgroups-for example, byageandsex-but thisadditional
information maynotbeavailable.
Formanypopulations, suchasthethreeexamples above, no
sampling frameexists, sothatitmaybeimpossible toobtain a
representative sample. Consider, forexample, trying toselect a
random sample ofallthepreschool children inanareato
estimate theprevalence ofvision orhearing defects. Yetfor
studies suchasthis,which setouttoestimate theprevalence or
incidence ofsomecondition, theneedforatrulyrepresentative
sample isparticularly great-otherwise theresults areof
uncertain value.
Evenwith agoodselection procedure thestudy maybe
ruinedbyapoorresponse rate.Although deemed non-invasive,
suchstudies mayentailvisiting people athome,expecting them
tocomplete andreturn aquestionnaire, ortoattend a
clinic,andthusmaybeliabletoconsiderable non-cooperation.Unfortunately, thosewhodonotparticipate oftentendtobe
somewhat different fromthosewhodo,bothinrespectoftheir
medical condition (ifthisisrelevant) andtheirsocialand
demographic characteristics. Thisproblem shouldbeanticipated
atthedesign stage,andplansmade to"chase up"non-
responders. Itisgenerally advisable tokeepquestionnaires and
otherprocedures shortandsimple tohelpreduce non-response.
Intheend,though, theresponse ratemaylargelydepend on
thesubjects' perception oftheimportance ofthestudy.
Itismuchlesscommon incase-control studies tofind
researchers concerned aboutdefining thesubjects whowillbe
eligible forastudy,although Sackett' hasdescribed 22biases
thatmayariseatthisstage.Oneofthemostinteresting is
Berkson's bias,whichMainland2 recently drewtotheattention
ofreaders ofthisjournal. Case-control studies ofhospital
patients areoftensetuptostudytherelation between aspecific
disease andexposure toasuspected causalfactor.Ifthehospital
admission ratesforexposed andunexposed casesandcontrols
differappreciably, thentheobserved association between the
factorandthedisease maybeseriously biased(ineither
direction).2 Indeed, thechoiceofcontrol group mayaffectthe
observed association between adisease andasuspected cause.A
consequence ofthisisthatsuchstudies mayneedtobesupported
byprospective studies.
Another ofSackett's catalogue' isthemembership (or"self-
selection") bias.Hecitestheexample ofanapparent association
between lackofexercise aftermyocardial infarction andthe
increased riskofrecurrent attacks. Thisresult wasfoundin
twoobservational studies whereexercise wastakenvoluntarily,
butwasnotsubstantiated byaprospective randomised study.
Sothemajorproblem ofallobservational studies isthe
selection ofsubjects forstudy.Thisaspect mustbegiven con-
siderable attention atthedesign stage,because ifthesampleis
notrepresentative ofthepopulation thentheresultswillbe
unreliable andofdubious worth.
Clinical trials
Whatever one'sviewonthebesttypeofdesign,clinicaltrials
ofsome sortareclearlyimportant fornewtreatments. As
May4says:"Theethicaljustification forsuchexperimentation,
which isoutside thepurephysician-patient relationship, is
based onajudgment thatincertaincircumstances itislegitimate
toputasubject atrisk,withhisorherconsent, because ofthe
overriding needofsociety forprogress incombating certain
diseases."
Arevealing example concerns theepidemic ofretrolental
fibroplasia inthe1950s.5 6Thetreatment ofinfants withearly
eyechanges withadrenocorticotrophic hormone wasthoughtDivision ofComputing andStatistics, Clinical Research Centre,
Harrow, MiddxHAl3UJ
DOUGLAS GALTMAN, BSC,medical statistician (member ofscientific
staff)BRITISH MEDICAL JOURNAL VOLUME 281 8NOVEMBER 1980 1267 | 98b7c1fd986b61bf682a38ca2f6fea32ef5da68c | page_0000 |
BRITISH MEDICAL JOURNAL VOLUME 281
tobeasuccess astherewasacurerateof7500,.Aclinicaltrial,
however, wouldhaveshownthatadrenocorticotrophic hormone
wasineffective since7500ofsuchinfants returntonormal
without treatment. Thewidespread useofthistreatment meant
thathundreds ofinfants wereexposed tounnecessary risk,and
thatdiscovery ofthecauseoftheepidemic (anoxygen-rich
environment) wasdelayed.
Thedebateabouttheethicsofclinical trialsisstillvery
active.Someauthors havesuggested thatitisunethical notto
carryoutaclinical trialonanewtreatment, whereas others
believethatsuchtrialsareunethical, atleastinthewaythey
areusuallyconducted.
ISITETHICAL TORANDOMISE ?
Inmostclinical trialssubjects areallocated tothenew
treatment atrandom, othersreceiving eitherastandard treatment
oraplacebo. Themainethicalproblem isthebalancing ofthe
welfare oftheindividuals inthetrialagainst thepotential
benefit tofuturepatients.
Itistherandom allocation ofsubjects thatcomesinformost
criticism. Itisarguedthatevenifatthebeginning ofatrial
onemaynotknowifatreatment iseffective, asthestudy
progresses itisunethical tocontinue torandomise ignoring the
resultssofar.7AsMeier"hasobserved, however, thisattitude is
basedonthequestionable premise "thatitisunethical todeny
anindividual anyexpected benefitoftreatment Aovertreatment
B,regardless ofhowsmallthatbenefitmaybeorhowuncertain."
Because ofthedifficulty ininterpreting interim resultsof
randomised studies, twotypesofnon-randomised studyhave
recently foundsomefavouranddeserve acloserlook.
HISTORICAL CONTROLS
Isitreallynecessary tohaveaconcurrent control group
whencarrying outaclinical trial?Cranberg9 hasrecently
arguedthatinstead onecanuseretrospective or"historical"
controls-that is,previously collected dataonpatients whohad
received whatwouldbethecontrol treatment. Although widely
practised, andperhaps ofvalueinsomecircumstances,"° this
canbeextremely risky.
Themainproblem ofstudiesusinghistorical controls istheir
insensitivity tosecularchanges, mostimportantly inselection
criteria.1' Theworsthistorical datatouseareotherpeople's
published results,perhapspartlybecause ofthepublication bias
towardspositive results.Pocock" givesasanexample 20studies
offluorouracil foradvanced cancerofthelargebowelwith
reported success ratesrangingfrom8%to8500.Butdatafroma
previous studyinthesameinstitution mayalsobeunreliable.
Pocockreportsthatin19instances wherethesametreatment
wasusedintwoconsecutive trialsofcancerchemotherapy in
oneorganisation thechanges indeathratesfromonetrialtothe
nextrangedfrom-46%to+240', fourofthedifferences
beingsignificant atthe200level.
Theuseofhistorical controls isoftenadvocated asbeing
moreethicalthanusingaconcurrent randomised controlgroup.
Theresultsofstudiesusinghistorical controls areextremely
unreliable, however, sothatunlessthereissoundjustification
fortheirusesuchdesigns shouldthemselves berejected as
unethical.
ADAPTIVE DESIGNS
Designs wheretheproportion ofsubjects allocated toeach
treatment depends ontheaccumulated results sofarmay
appearpreferable torandomised trials.7Itmustberealised,
however, thatwithsuchdesigns somesubjects arestillallocated
tothetreatment thatislesssuccessful sofar,notsomanyas
withrandomised studiesbutstillessentially atrandom. Further-more,because oftheunequal samplesizesforthetwotreatments,
thestudymayrequire moresubjects thananequalallocation
study.'1 3
Suchdesigns require thattheresultforeachindividual is
knownquickly, whichisoftennotthecase.Itisimplicitly
assumed thatthereisasingleoutcome ofinterest, whereas
theremaybeseveralpossible methods ofassessment, aswellas
aspects suchassideeffectstobeconsidered. Theyarealso
insensitive toanysecular changes duringthecourseofthe
study.Forthesereasons, although appealing inprinciple,
adaptive designs haverarely,ifever,beenused.'
SEQUENTIAL DESIGNS
Sequential designs'4 mayseemthebestcompromise inthat
theycombine themanyadvantages ofarandomised study
withthedesirable featureoftakingaccount oftheresultssofar
indetermining thelengthofthetrial.
Themainadvantage overanordinary randomised studyis
thattherequired samplesizewillbesmaller ifthetreatment
"effect" islarger. Sothebigger thedifference between
treatments, thefewersubjects receive thelesssuccessful
treatment.
Theirmaindisadvantages arethesameasforadaptive
designs, especially theneedfortheresultsforeachsubjecttobe
available quickly. Sequential designs areclearlyofnovaluein
long-term studies, whereallthesubjects willberecruited before
anyresultsareobtained. Nevertheless, intherightcircumstances
theycanbeusefulandshouldprobably beusedmoreoften.
CONSENT
Another problem ofclinical trialsistheneedtoobtainthe
"informed consent" ofthesubjects. Insomecasesthismaybe
impossible because oftheageorcondition ofthesubjects, or
because ofthedifficulty ofexplaining thescientific issues.
Zelen"5hasrecently proposed anewdesignforcomparing a
newtreatment withastandard onethatneatlyavoidsthe
problem. Heproposed that,ofthesubjects entering atrial,half
arerandomly assigned toreceivethestandard treatment (group
1).Thesesubjects aretreated asiftheywerenotinthetrial
apartfromtheneedsofstandardised assessment andrecord
keeping. Theotherhalf(group2)aregivenachoice:theyare
offeredthenewtreatment B,whichisunderinvestigation, but
theymayhavethestandard treatment Aiftheywish.The
important pointisthatthesubjects choose-this isquitedifferent
fromagreeing toberandomised-so thattheproblems associated
withinformed consent donotarise.
Ifmostofthesecondgroupelecttohavethenewtreatment B,
asisquitelikely,thenthisdesignwillprobably bemoreefficient
overall. Itisofcourseessential tocompare group1withgroup
2,notallthoseundergoing treatment Awiththoseundergoing
B.Inthiswaytworandomly selected groupswillbecompared.
Therewillbesomelossofefficiency because group2is"con-
taminated" byaminority undergoing treatment A,butthis
effectislikelytobeoutweighed bytheadvantage ofhaving
virtually norefusers.
Thisdesign,whichseemsperfectly ethical(Zelen"5 discusses
manyoftheissues), hastwoadvantages overordinary
randomised trials-the ability toinclude alleligiblesubjects
andtheavoidance ofthetrickyproblem ofinformed consent.
PLACEBOS
Toomanystudiescompare anewtreatment withaplacebo
ratherthananexisting treatment, andthusyieldresultsthat
areofnopractical importance. Itissometimes necessary to
includeplacebos, butwhenever possible theyshouldbeused1268 8NOVEMBER 1980 | 98b7c1fd986b61bf682a38ca2f6fea32ef5da68c | page_0001 |
BRITISH MEDICAL JOURNAL VOLUME 281 8NOVEMBER 1980 1269
onlywhenthereisnoappropriate treatment forcomparison.
Invasive placebo treatment isunlikely evertobejustified.
CONCLUSIONS
Thereisnoonebestdesignforallclinicaltrials.Thechoice
foraspecific trialmustdepend ontheseriousness ofthecon-
ditionbeingtreated, thenatureofthetreatments, theresponse
time,themeasures ofoutcome, andsoon.Themainethical
problem isbalancing theinterests oftheindividuals inthestudy
withthoseofthemuchlargernumber whomaybenefitinthe
longterm.Butitisalsovitalthattheresearch shouldprovide
usefulresults, andthismayoftenbeachieved bestbya
randomised study(double-blind ifpossible). Ifitisthought
likelythathighlyfavourable earlyresultsorahighincidence of
sideeffectswouldargueinfavourofpremature termination of
thestudy,thentheseconsiderations maybebuiltin,usinga
sequential design.
Theethicaldifficulties associated withthewidespread useofa
newtreatment without atrialarefargreaterthanthoseassociated
withthetrialitself.Theimportance ofgooddesign,however, is
reflected inthemanyexamples ofconflicting resultsthatmaybe
foundinseriesofcase-control studiesofthesametopic.16Asa
notableexample, after32studiesover25yearsthereisstillno
consensus ontheefficacyofanticoagulants following myocardial
infarction.11References
1SackettDL.Biasinanalytic research. 7ChronDis1979;32:51-63.
2Mainland D.Berkson's fallacyincase-control studies. BrMedJ1980;
280:330.
3Roberts RS,SpitzerWO,Delmore T,SackettDL.Anempirical demon-
stration ofBerkson's bias.JChronDis1978;31:119-28.
4MayWW.Thecomposition andfunction ofethicalcommittees. JfMed
Ethics1975;1:23-9.
Silverman WA.Thelessonofretrolental fibroplasia. SciAm1977;236(6):
100-7.
6Herbert V.Acquiring newinformation whileretaining oldethics.Science
1977;198:690-3.
7Weinstein MC.Allocation ofsubjects inmedical experiments. NEngl
JMed1974;291:1278-85.
8MeierP.Terminating atrial-the ethicalproblem. ClinPharmacol Ther
1979;25:633-640.
9Cranberg L.Doretrospective controls makeclinical trialsinherently
fallacious ?BrMedJ31979;ii:1265-6.
10GehanEA,Freireich EJ.Non-randomized controls incancerclinical
trials.NEnglJ'Med1974;290:198-204.
1DollR,PetoR.Randomised controlled trialsandretrospective controls.
BrMedJ71980;280:44.
12PocockSJ.Allocation ofpatients totreatment inclinicaltrials.Biometrics
1979;35:183-97.
3ByarDP,SimonRM,Friedewald WT,etal.Randomized clinicaltrials.
Perspectives onsomerecentideas.NEnglJfMed1976;295:74-80.
14Armitage P.Sequential medicaltrials.2nded.Oxford: Blackwell, 1975.
15ZelenM.Anewdesignforrandomized clinicaltrials.NEnglJ'Med
1979;300:1242-5.
16Horwitz BI,Feinstein AR.Methodologic standards andcontradictory
resultsincase-control research. AmJMed1979;66:556-64.
Thisisthesecondinaseriesofeightarticles.
Noreprintswillbeavailable fromtheauthor.
LessonoftheWeek
Endoscopic assessment ofoesophageal disease
GLITTLE, HRMATTHEWS
Thewidespread availability offibreoptic endoscopy hasledto
significant changes intheinvestigation oftheuppergastro-
intestinal tract.Wehavenotedinparticular atendency fordoc-
torstodispense witharadiological examination completely
andrelysolelyonendoscopic findings forboththediagnosis
andthemanagement ofoesophageal disorders. Thishasledto
severalseriouserrorsinthetreatment ofpatients withoesopha-
gealdisease,andwedescribe threecases.Wewanttoemphasise
thatradiological contrast examination mustprecede endoscopy
intheinvestigation ofcasesofoesophageal diseaseifsimilar
errorsaretobeavoided.
Casereports
Case1-A76-year-old manpresented withafive-month
historyofprogressive dysphagia andahistory ofheartburn.
QueenElizabeth Hospital, Birmingham B152TH
GLITTLE, FRCS,seniorsurgicalregistrar
EastBirmingham Hospital, Bordesley GreenEast,Birmingham
B95ST
HRMATTHEWS, FRCS,consultant thoracic surgeonEndoscopic findings alonedonottellthewhole
storyintheinvestigation ofoesophageal disease.
Nobariumstudiesweredone,butendoscopy showed astricture
ofthemiddlethirdoftheoesophagus. Thecausecouldnotbe
determined, butwhendilatation wasattempted (Eder-Puestow
technique) theoesophagus wasperforated. Thepatient was
transferred toourunitandtheunderlying pathology was
unknown.
AGastrografin swallowfailedtooutlinetheoesophagus, allthe
contrast medium havingpasseddirectly intothepleuralcavity.
Because ofthepossibility ofabenignstricture thoracotomy
wasperformed, butanunresectable carcinoma ofthemiddle
thirdoftheoesophagus wasdiscovered. Despite palliative
intubation thepatientdied.Hadaradiological diagnosis been
madeonthispatientfirstauselessandpainfulthoracotomy
couldhavebeenavoided.
Case2-Awomanaged68presented withretrosternal dis-
comfort andvomiting. Hergeneralpractitioner hadrequested
thatbariumstudiesbearranged. Instead, endoscopy wasper-
formed andapossible paraoesophageal herniadiagnosed. | 98b7c1fd986b61bf682a38ca2f6fea32ef5da68c | page_0002 |
Citation: Bouillon-Minois, J.-B.;
Dutheil, F. Biomarker of Stress,
Metabolic Syndrome and Human
Health. Nutrients 2022 ,14, 2935.
https://doi.org/10.3390/
nu14142935
Received: 23 June 2022
Accepted: 27 June 2022
Published: 18 July 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
nutrients
Editorial
Biomarker of Stress, Metabolic Syndrome and Human Health
Jean-Baptiste Bouillon-Minois1,2,*
and Fr édéric Dutheil1,3
1Universit éClermont Auvergne, CNRS, LaPSCo, Physiological and Psychosocial Stress,
F-63000 Clermont-Ferrand, France; fdutheil@chu-clermontferrand.fr
2CHU Clermont-Ferrand, Emergency Medicine, F-63000 Clermont-Ferrand, France
3CHU Clermont-Ferrand, Occupational and Environmental Medicine, WittyFit,
F-63000 Clermont-Ferrand, France
*Correspondence: jbb.bouillon@gmail.com
Metabolic syndrome is a significant public health concern linked to the obesity pan-
demic. Even if glycemia, triglycerides, and HDL are mandatory for the assessment of
metabolic syndrome, other biomarkers have recently been proposed to be linked. Recent
studies indicate that both chronic and (repeated) acute stress are involved in developing
metabolic syndrome. Furthermore, both oxidative and psychosocial stress have been linked
to heart disease and metabolic syndrome. The main hypothesis is the disruption of the
hypothalamic–pituitary–adrenal axis (HPA axis). Indeed, a dysfunction in the HPA axis
increases cortisol levels in the blood, increasing both glucose and insulin levels, causing the
apparition of insulin resistance, the promotion of dyslipidemia, high blood pressure, and
visceral adiposity. Secondly, HPA axis dysfunction has an impact on bones, cardiovascular
diseases, and psychiatric disorders.
This Special Issue is composed of five articles, three original articles and two reviews,
including a systematic review with a meta-analysis.
The first study concerns 535 obese patients aged 0–18 years. Children born large for
gestational age predominated over those born small for gestational age. Birth weight had
an independent effect on triglycerides and insulin resistance (two well-known biomarkers
of cardiometabolic risk) during childhood, whereas obesity had a direct influence on
hypertension, an impaired glucose metabolism, and hypertriglyceridemia [1].
The second one also studied pediatric obesity, but this time, the impact of acute stress
was assessed for 137 obese youngsters with the Trier Social Stress Test. Those overweight
and with a high level of chronic stress seemed to have a higher stress vulnerability (stronger
relative salivary cortisol reactivity and weaker happiness recovery) and a higher fat/sweet
snack intake. Those patients would benefit from stress therapy to reduce the risk of
obesity [2].
The third one studied the link between dysfunction pancreatic ß-cells and nonalcoholic
fatty liver disease (NAFLD). This disease is associated with a decreased insulin sensitiv-
ity. Among 6168 participants, those with NAFLD had a much higher HOMA2-%B level.
However, when evaluating the -cell function in the context of insulin resistance by using
the disposition index, NAFLD subjects had a lower disposition index. Thus, it seems that
pancreatic -cell function might be a novel predictor for the presence of NAFLD, and an
insufficient compensatory -cell function is associated with NAFLD [3].
The fourth one explored the link between metabolic syndrome and sarcopenia—
two common ailments among elderly patients. Indeed, skeletal muscle is a major organ
in the glucose metabolism. The loss of muscle mass has been closely linked to insulin
resistance and metabolic syndrome through the accumulation of intramuscular fat using
a combination of factors (oxidative stress, inflammatory cytokines, mitochondrial dys-
function, insulin resistance, and inactivity). Persistent inflammation, fat deposition, and
insulin resistance are thought to play a complex role in the association between metabolic
Nutrients 2022 ,14, 2935. https://doi.org/10.3390/nu14142935 https://www.mdpi.com/journal/nutrients | e3c15d56530a8101c39767fb9252d7e3773286ac | page_0000 |
Nutrients 2022 ,14, 2935 2 of 2
syndrome and sarcopenia, both affecting quality of life and contributing to increased frailty,
weakness, dependence, morbidity, and mortality [4].
Lastly, leptin, the main satiety hormone presenting a circadian rhythm, was studied
in an acute way. Indeed, it seems that leptin can be considered a biomarker of acute
stress, with a 34% decrease following acute stress. Individuals with a normal weight and
women had a higher variation of leptin levels after stress, suggesting that leptin may have
implications in obesity development in response to stress in a sex-dependent manner [5].
In conclusion, it was an honor to be a guest editor to this very interesting Special Issue.
All the different findings provide a higher comprehension on the link between metabolic
syndrome and stress through the use of novel biomarkers. We hope that future research
is performed, aiming to find novel pathway to increase quality of life among obese and
stressed people.
Author Contributions: J.-B.B.-M. and F.D. wrote this editorial hand-to-hand. All authors have read
and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Bizerea-Moga, T.O.; Pitulice, L.; Pantea, C.L.; Olah, O.; Marginean, O.; Moga, T.V . Extreme Birth Weight and Metabolic Syndrome
in Children. Nutrients 2022 ,14, 204. [CrossRef] [PubMed]
2. Wijnant, K.; Klosowska, J.; Braet, C.; Verbeken, S.; De Henauw, S.; Vanhaecke, L.; Michels, N. Stress Responsiveness and Emotional
Eating Depend on Youngsters’ Chronic Stress Level and Overweight. Nutrients 2021 ,13, 3654. [CrossRef] [PubMed]
3. Chen, X.; Xiao, J.; Pang, J.; Chen, S.; Wang, Q.; Ling, W. Pancreatic -Cell Dysfunction Is Associated with Nonalcoholic Fatty
Liver Disease. Nutrients 2021 ,13, 3139. [CrossRef]
4. Nishikawa, H.; Asai, A.; Fukunishi, S.; Nishiguchi, S.; Higuchi, K. Metabolic Syndrome and Sarcopenia. Nutrients 2021 ,13, 3519.
[CrossRef] [PubMed]
5. Bouillon-Minois, J.-B.; Trousselard, M.; Thivel, D.; Benson, A.C.; Schmidt, J.; Moustafa, F.; Bouvier, D.; Dutheil, F. Leptin as a
Biomarker of Stress: A Systematic Review and Meta-Analysis. Nutrients 2021 ,13, 3350. [CrossRef] [PubMed] | e3c15d56530a8101c39767fb9252d7e3773286ac | page_0001 |
How the sequence of a gene can tune its translation
Kurt Fredrick and Michael Ibba
Department of Microbiology, Ohio State Biochemistry Program, and Center for RNA Biology, The
Ohio State University, Columbus, Ohio 43210, USA
Abstract
Sixty-one codons specify 20 amino acids offering cells many options for encoding a polypeptide
sequence. Two new studies (Cannarozzi et al, 2010, Tuller et al., 2010) now foster the idea that
patterns of codon usage can control ribosome speed, fine-tuning translation to increase the efficiency
of protein synthesis.
Just about every molecular biologist, intentionally or not, has conducted experiments on the
roles of codon usage during translation. For example, to prepare a protein of interest, a foreign
gene might be expressed in a heterologous host like the bacterium Escherichia coli, but
disappointingly either no protein is produced or the resulting product is inactive. There are
numerous possible reasons such experiments fail, but one that can usually be excluded from
these is that the genetic code differs between the host and the foreign organism. With very few
exceptions, the genetic code is universal – no matter how evolutionarily distant two organisms
are, they will always use the same combination of nucleotide triplets (codons) to encode the
same amino acids. That being said, the universality of the genetic code comes with a few
caveats. There are not twenty codons, but up to 61. As a consequence, some amino acids are
encoded multiple times within the genetic code whereas others are not. For example, there are
six different codons for leucine, but only two for lysine. These codons, with different
sequences, but coding for the same amino acid, are termed synonymous codons. The frequency
with which different synonymous codons are used for a particular amino acid varies greatly
between different organisms – the phenomenon is referred to as codon usage. Understanding
differences in codon usage, and making appropriate adjustments, can help to improve yields
when trying to express foreign proteins, but the varying success and unpredictability of these
approaches suggests that we don’t yet understand all the rules guiding translation. One long-
standing idea is that the order with which codons are used is far from random. Examining
individual mRNA sequences to try and identify patterns in codon choice has yielded relatively
little information. Recent studies have instead focused on searching for genome-wide trends
in codon choice, with increasingly striking results as clear patterns start to emerge. These
include roles for codon selection in protein folding (e.g. Zhang et al., 2009) and, as described
in this issue of Cell, potential functions in controlling translation elongation (Cannarrozzi et
al., 2010; Tuller et al., 2010).
© 2010 Elsevier Inc. All rights reserved.
Correspondence to: Dr. Michael Ibba, Dr. Kurt Fredrick, Department of Microbiology, The Ohio State University, 318 West 12th Avenue,
Columbus, Ohio 43210-1292, ibba.1@osu.edu, fredrick.5@osu.edu.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting
proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.
NIH Public Access
Author Manuscript
Cell. Author manuscript; available in PMC 2011 April 16.
Published in final edited form as:
Cell. 2010 April 16; 141(2): 227±229. doi:10.1016/j.cell.2010.03.033.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 2f2cb8883590e6338a1d35b307970a3176d6915b | page_0000 |
Translation speed varies over mRNA length
One of the first questions to arise when considering the possible effects of altering codon usage
is how it might impact the rate at which an mRNA is translated. The notion that translation
rates can change for different regions within a single mRNA, and in so doing facilitate processes
such as frame shifting or the folding of nascent polypeptides, has been discussed in the literature
for some time and has recently garnered additional experimental support (Siller et al., 2010;
Zhang et al., 2009). To get a more global picture of translation rates, Tuller and colleagues first
set out to estimate how efficiently each individual codon is translated in a particular gene, based
upon the predicted availability of the corresponding tRNA. Similar approaches have been
applied before and averaged over individual genes, but never codon-by-codon over an entire
transcriptome. To further refine the analysis, Tuller et al. also factored in whether or not
translation of a codon required a perfect codon-anticodon match (more efficient) or a codon-
anticodon wobble interaction (less efficient). The final adjustment compared to earlier studies
was to take advantage of recent technological advances to measure global cellular tRNA pools
(Zaborske et al., 2009) and to use these values when calculating tRNA availability. For yeast
grown on rich medium, there was a robust correlation between actual tRNA abundance and
tRNA gene copy number. Extrapolating from this finding allowed the authors to broaden their
studies to organisms other than yeast. One qualification here is that whereas this correlation
between copy number and tRNA abundance holds true under some conditions, there are other
cellular states where this may not reflect the charging of tRNAs with their appropriate amino
acids (Dittmar et al., 2005; Elf et al., 2003). When all of these factors are taken into
consideration, conserved patterns start to emerge both within a single organism’s transcriptome
as well as across species. The most obvious finding is that the speed of translation is predicted
to be slow during the first 30–50 codons (the “ramp”), and then to increase to a plateau level
for the remainder of the gene (Figure 1A). Support for this model comes from the striking
correlation between profiles of ribosome density along mRNAs predicted computationally and
those observed experimentally (Ingolia et al., 2009). The clear exception to this rule is for the
second codon, which is predicted to be translated at a higher rate than the surrounding codons,
thereby promoting rapid release and recycling of the initiator tRNA.
What purpose could this ramp play in translation? Slowing translation elongation immediately
after initiation would effectively generate more uniform spacing between ribosomes further
down the mRNA, which should prevent ribosome congestion and promote efficient protein
synthesis. The ramp is predominantly associated with highly-expressed genes, consistent with
a role for the ramp in increasing product yield amid heavy ribosome traffic. Traffic jams in
translation, and also in transcription (Ehrenberg et al., 2010), can decrease processivity and
increase stalling and termination, providing precedence for the idea that the ramp increases the
overall efficiency of protein synthesis. Although this hypothesis is attractive, it remains to be
tested experimentally. Another potential role for the ramp involves protein folding. The length
of the ramp corresponds remarkably well to the length of polypeptide needed to fill the exit
tunnel of the ribosome (Ban et al., 2000), so the nascent peptide chain should emerge from the
ribosome as it transitions from the slow (ramp) stage to the fast stage of elongation. This raises
the possibility that the ramp might somehow facilitate interactions between the emerging
peptide and the chaperone proteins, thereby increasing the fraction of correctly folded product.
This idea will also be worth investigating experimentally.
The importance of the ramp may explain why foreign gene expression sometimes fails – the
ramp’s function would diminish or disappear in heterologous systems due to changes in tRNA
availability ( Tuller et al., 2010). The need to compensate for the absence of ramps by slowing
down translation might contribute both to anecdotal “cures” for increasing foreign protein
production (e.g., lowering growth temperature) and to more rational recoding-based
approaches (Siller et al., 2010). It also may help to explain why the over-expression of rareFredrick and Ibba Page 2
Cell. Author manuscript; available in PMC 2011 April 16.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 2f2cb8883590e6338a1d35b307970a3176d6915b | page_0001 |
tRNAs, another common strategy to improve foreign protein production, does not always work
and can even be counterproductive. In some cases, rare codons can indeed be problematic as
the scarcity of the corresponding tRNA may lead to premature termination; in other examples
excessive over-expression of a low abundance tRNA could significantly change the translation
efficiency of certain codons thereby disrupting the ramp to the detriment of protein synthesis.
Although the effects of the ramp on the early rounds of elongation may indeed help to tune
translation, it is important to keep in mind that the rate of product yield for a given mRNA is
strictly governed by the rate of initiation. Kudla et al. (2009) recently examined the effects of
synonymous codon substitutions on the efficiency of translation of a gene encoding green
fluorescent protein in E. coli . They found that the sequence at the beginning of the gene strongly
influenced translation, and that the expression level was inversely correlated with predicted
mRNA secondary structure. These data are consistent with other studies pointing to the
importance of mRNA structure in controlling translation initiation (de Smit and van Duin,
2003 ; Studer and Joseph, 2006). As the regions defined by Kudla et al. and Tuller et al. overlap,
the nucleotide sequence in this region may control initiation and/or early elongation, depending
on the particular gene.
Finding order from degeneracy
The genetic code is degenerate, that is to say it can use many different combinations of codons
to make exactly the same protein. The 61 codons that encode the 20 standard amino acids are
not equally abundant in mRNA. Within synonymous sets, some codons are used far more
frequently than others, and the limited strategic use of rare codons can sometimes be exploited
for regulatory purposes (Chandra and Chater, 2008 ). The discovery of ramps emphasizes that
codon choice is not uniform, but is instead highly selected and broadly conserved (Tuller et
al., 2010). In a related vein, Cannarrozzi and colleagues describe a different pattern of codon
usage in yeast that appears to reflect differences in translation rate (Cannarrozzi et al., 2010).
They find that when an amino acid recurs in a protein (for instance, xLxxL) there is a strong
tendency to use the same codon the second time as for the first occurrence of the amino acid.
This predisposition towards selecting particular codons rather than arbitrarily choosing one
from the synonymous set has important implications for the dynamics of mRNA translation
and the protein synthesis machinery.
Cannarrozzi and colleagues focused on groups of codons that encode the same amino acid, and
asked whether these synonymous codons were randomly or non-randomly ordered along genes.
In so doing, they were able to analyze all consecutive synonymous codon pairs in the yeast
genome. What this revealed was that identical codons have a strong tendency to be used again
when an amino acid recurs, and if the same codon is not reused, there is instead a bias towards
the most closely related synonymous wobble codons. This observed reuse of codons, termed
“auto-correlation,” is not simply the result of the most frequent codons accumulating in genes,
as rare codons are just as likely to be reused as common ones. These highly conserved patterns
suggest that reusing codons may benefit translation in some way, a notion reinforced by the
frequent reuse of rare codons in highly expressed genes. This trend extends to larger groups
of genes. For example, regulons that are highly upregulated under certain conditions, such as
those involved in cell cycle progression and environmental stress responses, showed the
strongest correlations in sequential codon usage. The overall model that emerges from these
and other analyses performed by the authors is that, simply put, codon reuse may provide an
effective mechanism to speed up translation. To put this model to the test, the authors
engineered reporter genes in which the synonymous codons were either positionally auto-
correlated or anti-correlated (i.e., the two opposite extremes of their model) (Figure 1B). These
reporter genes encoded the identical protein and had an identical total codon composition–
only the arrangement of the synonymous codons differed. These genes were expressed in yeastFredrick and Ibba Page 3
Cell. Author manuscript; available in PMC 2011 April 16.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 2f2cb8883590e6338a1d35b307970a3176d6915b | page_0002 |
cells in the presence of radiolabeled amino acids, the cells were lysed, and the translation
products (partial and complete) were purified from the remainder of the lysate using an N-
terminal epitope tag. These translation products were separated using SDS-PAGE, and the size
distribution of the nascent chains allowed ribosome density along each mRNA to be compared.
It was found that the reporter mRNAs with auto-correlated codons had lower ribosome density
than their counterparts with anti-correlated codons. Because ribosome density is inversely
proportional to elongation rate, the authors could deduce that translation on auto-correlated
mRNA was substantially (~30%) faster than on anti-correlated mRNA.
This significant increase in the speed of translation achieved by the appropriate arrangement
of codons has important implications for how the translation machinery functions. To achieve
the higher translation speed that comes from correlating codon choices, the availability of the
corresponding tRNAs may need to be greater. To assess the availability of tRNAs during high-
speed translation, Cannarrozzi et al. modeled a number of different scenarios in several
eukaryotes, and also took into account the effect of distance between synonymous codons on
the degree of auto-correlation. They concluded that the advantage to codon reuse comes from
an ability to reuse the corresponding tRNA. This implies that the tRNA molecules exiting the
ribosome remain associated with the translational machinery so that they are readily available
when the next identical codon comes along. This model necessitates that the enzymes
responsible for attaching amino acids to tRNAs, the aminoacyl-tRNA synthetases, are also
associated with ribosomes. Extensive evidence exists that the aminoacyl-tRNA synthetases
form ordered complexes in eukaryotes ( Deutscher, 1984; Mirande et al., 1985), and that these
complexes associate with ribosomes (Kaminska et al., 2009). Moreover, it has recently been
shown that these complexes can increase translation rates by promoting the “channeling” of
charged tRNAs to the ribosome for protein synthesis (Kyriacou and Deutscher, 2008), which
is highly consistent with the model of Cannarrozzi and coworkers. One interesting issue that
remains to be explored is how these models will play out in bacteria, where translation rates
are substantially higher but evidence for channeling is scant at best.
Perspectives
The studies of Cannarrozzi, Tuller and their colleagues provide evidence that the pattern of
codon usage modulates the rate of protein synthesis, and suggest how this might be exploited
on the genome scale to fine-tune the efficiency with which certain sets of genes are translated.
These effects may also be accentuated by translation factors, thereby providing an additional
layer of regulation. For example, the translation factor eIF5A is known to increase the
efficiency of translation (Saini et al., 2009), and how this and other factors could potentially
change the gradient of a ramp or alter the impact of codon correlation is of significant interest.
Further modulation of ramps and codon correlation effects may also be provided by isodecoder
tRNAs, that is, tRNAs with identical anticodons but otherwise diverse primary sequences.
Recent studies have shown that huge numbers of isodecoders exist in eukaryotes. There are
approximately 270 isodecoders in the human genome alone (Geslain and Pan, 2010), and
variations in the expression of these tRNAs may further heighten the codon-dependent changes
in translation efficiency predicted by Tuller, Cannarrozzi and their colleagues ( Cannarrozzi et
al., 2010; Tuller et al., 2010).
Acknowledgments
This work was supported by grants from the National Institutes of Health (GM 072528 to KF; GM 065183 to MI).
REFERENCES
Ban N, Nissen P, Hansen J, Moore PB, Steitz TA. Science 2000;289:905–920. [PubMed: 10937989]Fredrick and Ibba Page 4
Cell. Author manuscript; available in PMC 2011 April 16.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 2f2cb8883590e6338a1d35b307970a3176d6915b | page_0003 |
Cannarrozzi G, Schraudolph NN, Faty M, von Rohr P, Friberg MT, Roth AC, Gonnet P, Gonnet G, Barral
Y. Cell. 2010 In Press.
Chandra G, Chater K. Antonie van Leeuwenhoek 2008;94:111–126. [PubMed: 18320344]
de Smit MH, van Duin J. J Mol Biol 2003;331:737–743. [PubMed: 12909006]
Deutscher MP. J Cell Biol 1984;99:373–377. [PubMed: 6746733]
Dittmar KA, Sorensen MA, Elf J, Ehrenberg M, Pan T. EMBO Rep 2005;6:151–157. [PubMed:
15678157]
Ehrenberg M, Dennis PP, Bremer H. Biochimie 2010;92:12–20. [PubMed: 19835927]
Elf J, Nilsson D, Tenson T, Ehrenberg M. Science 2003;300:1718–1722. [PubMed: 12805541]
Geslain R, Pan T. J Mol Biol 2010;396:821–831. [PubMed: 20026070]
Ingolia NT, Ghaemmaghami S, Newman JR, Weissman JS. Science 2009;324:218–223. [PubMed:
19213877]
Kaminska M, Havrylenko S, Decottignies P, Le Marechal P, Negrutskii B, Mirande M. J Biol Chem
2009;284:13746–13754. [PubMed: 19289464]
Kudla G, Murray AW, Tollervey D, Plotkin JB. Science 2009;324:255–258. [PubMed: 19359587]
Kyriacou SV, Deutscher MP. Mol Cell 2008;29:419–427. [PubMed: 18313381]
Mirande M, Le Corre D, Waller JP. Eur J Biochem 1985;147:281–289. [PubMed: 3971983]
Saini P, Eyler DE, Green R, Dever TE. Nature 2009;459:118–121. [PubMed: 19424157]
Siller E, Dezwaan DC, Anderson JF, Freeman BC, Barral JM. J Mol Biol. 2010 doi:10.1016/j.jmb.
2009.12.042.
Studer SM, Joseph S. Mol Cell 2006;22:105–115. [PubMed: 16600874]
Tuller T, Carmi A, Vestsigian K, Navon S, Dorfan Y, Zaborske J, Pan T, Dahan O, Furman I, Pilpel Y.
Cell. 2010 In Press.
Zaborske JM, Narasimhan J, Jiang L, Wek SA, Dittmar KA, Freimoser F, Pan T, Wek RC. J Biol Chem
2009;284:25254–25267. [PubMed: 19546227]
Zhang G, Hubalewska M, Ignatova Z. Nat Struct Mol Biol 2009;16:274–280. [PubMed: 19198590]Fredrick and Ibba Page 5
Cell. Author manuscript; available in PMC 2011 April 16.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 2f2cb8883590e6338a1d35b307970a3176d6915b | page_0004 |
Figure 1. Messenger RNA sequences set the speed limit
There are three times as many codons as amino acids, meaning that a given amino acid can be
encoded by several synonymous codons and identical proteins can be made from very different
mRNA sequences. Codon choice is not random, but is highly selected across a broad range of
organisms to optimize protein production. ( A) For many genes, codons recognized by low-
abundance tRNAs are overrepresented in the first part of the gene. This pattern suggests that
ribosomes translate more slowly over the initial 50 codons or so (ramp stage), and then translate
the remainder of the mRNA at full speed. The mRNA template itself controls the speed of
ribosomes, somewhat analogous to how poor road conditions limit the speed of cars in a
construction zone. ( B) The arrangement of synonymous codons along a gene influences
translation speed. Shown is a simple example in which two different codons (represented by
different shades of blue) encode the same amino acid. When the identical codons are
consecutively arranged along the mRNA (auto-correlated), translation is faster than when they
are alternatively arranged (anti-correlated).Fredrick and Ibba Page 6
Cell. Author manuscript; available in PMC 2011 April 16.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript | 2f2cb8883590e6338a1d35b307970a3176d6915b | page_0005 |
Machine Learning for Quantum Mechanical Properties of Atoms in Molecules
Matthias Rupp,1,Raghunathan Ramakrishnan,1and O. Anatole von Lilienfeld1,y
1Institute of Physical Chemistry and National Center for Computational Design and Discovery of Novel Materials (MARVEL),
Department of Chemistry, University of Basel, Klingelbergstr. 80, CH-4056 Basel, Switzerland
(Dated: August 26, 2015)
We introduce machine learning models of quantum mechanical observables of atoms in molecules.
Instant out-of-sample predictions for proton and carbon nuclear chemical shifts, atomic core level
excitations, and forces on atoms reach accuracies on par with density functional theory reference.
Locality is exploited within non-linear regression via local atom-centered coordinate systems. The
approach is validated on a diverse set of 9 k small organic molecules. Linear scaling of computational
cost in system size is demonstrated for saturated polymers with up to sub-mesoscale lengths.
PACS numbers: 03.65.-w,31.15.A-,31.15.E-,02.60.Ed
This work has subsequently been published in the
Journal of Physical Chemistry Letters 6(16): 3309{
3313, American Chemical Society, 2015. To ac-
cess the nal edited and published work see DOI
10.1021/acs.jpclett.5b01456.
Accurate solutions to the many-electron problem in
molecules have become possible due to progress in hard-
ware and methods. [1{4] Their prohibitive computational
cost, however, prevents both routine atomistic modeling
of large systems and high-throughput screening. [5] Ma-
chine learning (ML) models can be used to infer quantum
mechanical (QM) expectation values of molecules, based
on reference calculations across chemical space. [6, 7]
Such models can speed up predictions by several orders
of magnitude, demonstrated for relevant molecular prop-
erties such as enthalpies, entropies, polarizabilities, elec-
tron correlation, and, electronic excitations. [8{10]
A major drawback is their lack of transferability, e.g.,
ML models trained on bond dissociation energies of small
molecules will not be predictive for larger molecules. In
this work, we introduce ML models for properties of
atoms in molecules. These models exploit locality to
achieve transferability to larger systems andacross chem-
ical space, for systems that are locally similar to the ones
trained on (Fig. 1). These aspects have only been treated
in isolation before.[6, 11]
We model spectroscopically relevant observables,
namely13C and1H nuclear magnetic resonance (NMR)
chemical shifts [12] and 1 score level ionization ener-
gies (CIE), as well as atomic forces, crucial for struc-
tural relaxation and molecular dynamics. Nuclear shifts
and ionization energies are dominated by inherently lo-
cal core electron-nucleus interactions. Atomic forces are
expectation values of the di erential operator applied to
an atom's position in the Hamiltonian [13], and scale
quadratically with inverse distance.
Inductive modeling of QM properties of atoms in
molecules constitutes a high-dimensional interpolation
problem with spatial and compositional degrees of free-
dom. QM reference calculations provide training exam-
⟶z
QFIG. 1. Sketch illustrating local nature of atomic properties
for the example of force h j@RQ^Hj iacting on a query atom
in a molecule (mid), inferred from similar atoms in training
molecules (top, bottom). Shown are force vectors (arrows),
integrated electron densityR
dxdyn(r) (solid) and integrated
electronic term of Hellmann-Feynman force along z(dashed).
plesf(xi;yi)gn
i=1, where the xiencode atoms in their
molecular evironment and yiare atomic property values.
ML interpolation between training examples then pro-
vides predicted property values for new atoms.
The electronic Hamiltonian is determined by num-
ber of electrons, nuclear charges fZIgand posi-
tionsfRIg, which can be challenging for direct in-
terpolation. [14] Proposed requirements for representa-
tions include uniqueness, continuity, as well as invariance
to translation, rotation, and nuclear permutations. [15]
For scalar properties (NMR, CIE), we use the sorted
Coulomb matrix [6] to represent a query atom Qand
its environment: MII= 0:5Z2:4
IandMIJ=ZIZJ=jRI
RJj, where atom indices I;Jrun overQand all atoms
in its environment, sorted by distance to Q. Note that
all molecules in this study are neutral, and no explicit
encoding of charge is necessary.
Atomic forces are vector quantities requiring a basis,
which should depend only on the local environment; in
particular, it should be independent of the global framearXiv:1505.00350v2 [physics.chem-ph] 25 Aug 2015 | ca914bd3cfced5f10907d642b2d64b2f9260d67d | page_0000 |
2
of reference used to construct the Hamiltonian in the
QM calculation. We project force vectors into a local
coordinate system centered on atom Q, and predict each
component separately. Later, the predicted force vector
is reconstructed from these component-wise predictions.
We use principal component analysis (PCA) to obtain
an atom-centered orthogonal three-dimensional local co-
ordinate system. In analogy to the electronic term in
Hellmann-Feynman forces,R
dr(r RQ)ZQn(r)=kr
RQk3[13], we weight atoms by ZI=kRI RQk3, increas-
ing in
uence of heavy atoms and decreasing in
uence
of distant atoms. Non-degenerate PCA axes are unique
only up to sign; we address this by de ning the center of
charge to be in the positive quadrant. A matching matrix
representation is obtained via MI= (ZI;X0
I;Y0
I;Z0
I),
whereX0;Y0;Z0are projected atom coordinates, and
rows are ordered by distance to central atom Q, yield-
ing anm4 matrix, where mis number of atoms. In
both representations, we impose locality by constraining
Q's environment to neighboring atoms within a sphere of
radius .
For interpolation between atomic environments we use
kernel ridge regression (KRR) [16], a non-linear regular-
ized regression method e ectively carried out implicitly
in a high-dimensional Hilbert space (\kernel trick"). [17]
Predictions are linear combinations over all training ex-
amples in the basis of a symmetric positive de nite ker-
nelk:f(z) =Pn
i=1 ik(xi;z), where are regression
weights for each example, obtained from a closed-form
expression minimizing the regularized error on the train-
ing data. See Refs. [6, 18{20] for details. As kernel k, we
use the Laplacian kernel k(x;z) = exp
kx zk1=d
,
wherekk 1is theL1-norm,is a length scale, and
d= dim(x). This kernel has shown best performance
for prediction of molecular properties. [18]
Our models contain three free parameters: cut-o
radius , regularization strength , and kernel length
scale. Regularization strength , controlling the
smoothness of the model, was set to a small constant
(10 10), forcing the model to t the QM values closely.
As for length scales , note that for the Laplacian kernel,
non-trivial behavior requires jj;jj1. We setto four
times the median nearest neighbor L1-norm distance in
the training set. [21] Cut-o radii were then chosen to
minimize RMSE in initial experiments (Fig. 2). For the
comparatively insensitive FC, other statistics (maxAE,
R2) yielded an unambiguous choice.
We used three datasets for validation: For NMR chem-
ical shifts and CIEs, both scalar properties, we employed
a dataset of 9 k synthetically accessible organic molecules
containing 7{9 C, N, or O atoms, with open valencies
saturated by H, a subset of a larger dataset. [23, 24]
Relaxation and property calculations were done at the
DFT/PBE0/def2TZVP level of theory [25{30] using
Gaussian [31]. For forces, we distorted molecular equi-
librium geometries using normal mode analysis [32{34]
●
●
●●●●●●●●●●●●■
■■■■■■■■■■■■▲
▲
▲
▲
▲
▲▲▲▲▲▲▲○
○○○○○○○○○○○○○□
□
□□□□□□□□□□□□* ** ** *
* ** *
2 4 6 824682 4 6 8
2468
cut-off radius/ÅRMSE/ %●13Cδ■1Hδ▲1s Cδ ○FC□FHFIG. 2. Locality of properties, measured by model perfor-
mance as a function of cut-o radius . Root mean square
error (RMSE) shown as fraction of corresponding property's
range [22] for nuclear shifts (13C,1H), core level ionization
energy (1s C ), and atomic forces ( FC,FH). Asterisksmark
chosen values. Shaded areas indicate 1.6 standard deviations
over 15 repetitions.
by adding random perturbations in the range [ 0:2;0:2]
to each normal mode, sampling homogeneously within
an harmonic approximation. Adding spatial degrees of
freedom considerably increases the intrinsic dimensional-
ity of the learning problem. To accommodate this, we
reduced dataset variability to a subset of 168 constitu-
tional isomers of C 7H10O2, with 100 perturbed geome-
tries for each isomer. Computationally inexpensive semi-
empirical quantum chemistry approximations are readily
available for forces. We exploit this to improve accuracy
by modeling the di erence between baseline PM7 [35]
and DFT reference forces (-learning [10]). To demon-
strate linear scaling of computational cost with system
size, we used a third dataset of organic saturated poly-
mers, namely linear polyethylene, the most common plas-
tic, with random substitutions of some CH units with NH
or O for chemical variety. All prediction errors were mea-
sured on out-of-sample hold-out sets never used during
training.
Table I presents performance estimates for models
trained on 10 k randomly chosen atoms, measured on a
hold-out set of 1 k other atoms. Comparison with liter-
ature estimates of typical errors of the employed DFT
reference method suggests in all cases that the ML mod-
els achieve similar accuracy|at negligible computational
cost after training. Statistical learning theory shows that
under certain assumptions the accuracy of a ML model
asymptotically improves with increasing training set size
asO(1=pn). [36] Fig. 3 presents corresponding learning
curves for all properties. Errors are shown as percentage
of property ranges [22], enabling comparison of properties
with di erent units. All errors start o in the single digit | ca914bd3cfced5f10907d642b2d64b2f9260d67d | page_0001 |
3
●●●
●●●●●●●■■■■■■■■■■▲▲▲▲▲▲▲▲▲▲○○○○○○○○○○□□□□□□□□□□1 k 2 k 5 k 10 k12481 k 2 k 5 k 10 k
1248
nRMSE/ %●13Cδ■1Hδ▲1s Cδ ○FC□FH
FIG. 3. Systematic improvement in accuracy of atomic prop-
erty predictions with increasing training set size n. Root
mean square error (RMSE) shown as fraction of correspond-
ing property's range [22] for nuclear shifts (13C,1H), core
level ionization energy (1s C ), and atomic forces ( FC,FH).
Values from 15 repetitions; see Table I for ranges and stan-
dard deviations. Solid lines are ts to theoretical asymptotic
performance of O(1=pn).
percent range at 1 k training atoms, and decay system-
atically to roughly half their initial value at 10 k training
atoms.
ML predictions and DFT values for chemical shifts of
all 50 k carbon atoms in the dataset are featured in Fig. 4.
The shielding of the nuclear spin from the magnetic eld
is strongly dependent on the atom's local chemical envi-
ronment. In accordance with the diversity of the dataset,
we nd a broad distribution with four pronounced peaks,
characteristic of up- or downshifts of the resonant fre-
quencies of nuclear carbon spin. The peaks at 30, 70,
150, and 210 ppm typically correspond to saturated sp3-
hybridized carbon atoms, strained sp3-hybridized car-
bons, conjugated or sp2-hybridized carbon atoms, and
carbon atoms in carbonyl groups, respectively. A ML
model trained on only 500 atoms already reproduces all
major peaks; larger training sets yield systematically im-
proved distributions. For 10 k training examples predic-
tions are hardly distinguishable from the DFT reference,
except for a small deviation at 140 ppm. Using the same
model, we predicted shifts for 847 k carbon atoms in all
134 k molecules published in Ref. [24]. The resulting dis-
tribution is roughly similar, re
ecting similar chemical
composition of molecules in this much larger dataset,
which is beyond the current limits of DFT reference cal-
culations employed here.
The presented approach to model atomic properties
scales linearly: Since only a nite volume around an atom
is considered, its numerical representation is of constant
size; [45] in particular, it does not scale with the system's
overall size. Comparing atoms, and thus kernel evalua-
tions, therefore requires constant computational e ort,
0 50 100 150 20010102103104
13Cδ/ppm# DFT ML 0.5k ML 1k
ML 10k GDB9FIG. 4. Distribution of 50 k13C chemical shifts in 9 k or-
ganic molecules. ML predictions for increasing training set
sizes approach DFT reference values. Molecular structures
highlight chemical diversity and e ect of molecular environ-
ment on chemical shift of query atom (orange; see main text).
GDB9 corresponds to ML predictions for 847 k carbon atoms
in 134 k similar molecules published in Ref. [24].
rendering the overall computational cost of predictions
linear in system size, with small prefactor. Furthermore,
a form of chemical extrapolation can be achieved despite
the fact that ML models are interpolation models. As
long as local chemical environments of atoms are similar
to those in the training set, the model can interpolate.
Consequently, using similar local \building blocks", large
molecules can be constructed that are very di erent from
the ones used in the training set, but amenable to pre-
diction.
To verify this, we trained a ML model on atoms drawn
from the short polymers in the third dataset, then ap-
plied the same model to predict properties of atoms in
polymers of increasing length. Training set polymers had
a backbone length of 29 C,N,O atoms; for validation, we
used up to ten times longer backbones, reaching lengths
of 355 A and 696 atoms in total. Fig. 5 presents nu-
merical evidence for excellent near-constant accuracy of
model predictions, independent of system size, validated
by DFT. Although trained only on the smallest instances,
the model's accuracy varies negligibly with system size,
con rming both transferability and chemically extrapola-
tive predictive power of the ML model.
Individual ML predictions are 4{5 orders of magnitude
faster than reference DFT calculations. Overall speed-up
depends on dataset and reference method, and is domi-
nated by training set generation, i.e., the ratio between
number of predictions and training set size. DFT and
ML calculations were done on a high-performance com-
pute cluster and a laptop, respectively.
In conclusion, we have introduced ML models for QM
properties of atoms in molecules. Performance and ap- | ca914bd3cfced5f10907d642b2d64b2f9260d67d | page_0002 |
4
TABLE I. Prediction errors for ML models trained on 10 k atoms and predicting properties of 1 k other out-of-sample atoms.
Calculated properties are NMR chemical shifts (13C,1H), core level ionization energy (1s C ), and forces ( FC,FH).a
Property Ref. Range MAE RMSE maxAE R2 /A
13C/ppm 2.4 [30, 37, 38] 6 { 211 3.9 0.28 5.80.30 368.0 0.9880.001 203.4 3
1H/ppm 0.11 [38{40] 0 { 10 0.28 0.01 0.420.02 3.21.1 0.9540.005 0.531.2 3:5
1s C/mE h7.5 [41{43] -165 { -2 4.9 0.12 6.50.27 3417 0.9710.002 1810.0 7
FC/mE h/a01 [44] -99 { 96 3.6 0.10 4.70.15 295.5 0.9830.002 0.690.1 6
FH/mE h/a01 [44] -43 { 43 0.8 0.02 1.10.03 7.42.6 0.9960.003 0.350.0 3
aShown are MAE of DFT reference from literature (Ref.), property ranges [22], mean absolute error (MAE), root mean
squared error (RMSE), maximum absolute error (maxAE), squared correlation ( R2) and hyperparameters (kernel length
scale, cut-o radius ). Averagesstandard deviations over 15 randomly drawn training sets.
●●
●●●●●●●●■■
■■■■■■■■▲▲▲▲▲▲▲▲
▲▲
○○○○○○○○○○
□
□□□□□□□□□
4 14 25 35123234 906 1578 2250
1357
polymer length/nmRMSE/ %polymer size/ #electrons
compute time/days●13Cδ■1Hδ▲1s Cδ ○FC□FH
FIG. 5. Linear scaling and chemical extrapolation for ML
predictions of saturated polymers of increasing length. Shown
are root mean square error (RMSE), given as fraction of cor-
responding property's range [22], as well as indicative com-
pute times of cubically scaling DFT calculations (gray bars)
and ML predictions (black bars, enlarged for visibility), which
scale linearly with low prefactor. See Table I for property
ranges.
plicability have been demonstrated for chemical shifts,
core level ionization energies, and atomic forces of 9 k
chemically diverse organic molecules and 168 isomers of
C7H10O2, respectively. Accuracy of predictions is on par
with the QM reference method. We have used the ML
model to predict chemical shifts of all 847 k carbon atoms
in the 134 k molecules published in Ref. [24]. Locality of
modeled atomic properties is exploited through use of
atomic environments as building blocks. Consequently,
the model scales linearly in system size, which we have
demonstrated for saturated linear polymers over 30 nm in
length. Results suggest that the model could be useful
in mesoscale studies.
For the investigated molecules and properties the local-
ity assumption, implemented as a nite cut-o radius in
the representation, has proven sucient. This might not
necessarily be true in general. The Hellmann-Feynman
force, for example, depends directly on the electron den-sity, which can be altered substantially due to long-range
substituent e ects such as those in conjugated -bond
systems. For other systems and properties, larger cut-
o s or additional measures might be necessary.
The presented ML models could also be used for nu-
clear shift assignment in NMR structure determination,
for molecular dynamics of macro-molecules, or condensed
molecular phases. We consider ecient sampling, i.e.,
improving the ratio of performance to training set size
(\sample eciency"), and improving representations to
be primary challenges in further development of these
models.
We thank Tristan Bereau, Zhenwei Li, and Kuang-Yu
Samuel Chang for helpful discussions. OAvL acknowl-
edges the Swiss National Science Foundation for sup-
port (SNF grant PP00P2 138932). Calculations were
performed at sciCORE ( scicore.unibas.ch ) scienti c
computing core facility at University of Basel. This re-
search used resources of the Argonne Leadership Com-
puting Facility at Argonne National Laboratory, which
is supported by the Oce of Science of the U.S. DOE
under contract DE-AC02-06CH11357.
mrupp@mrupp.info; Current address: Fritz Haber Insti-
tute of the Max Planck Society, Faradayweg 4{6, 14195
Berlin, Germany.
yanatole.vonlilienfeld@unibas.ch
[1] C. Bekas and A. Curioni, Comput. Phys. Comm. 181,
1057 (2010).
[2] I. Duchemin and F. Gygi, Comput. Phys. Comm. 181,
855 (2010).
[3] G. H. Booth, A. J. Thom, and A. Alavi, J. Chem. Phys.
131, 054106 (2009).
[4] A. Benali, L. Shulenburger, N. A. Romero, J. Kim, and
O. A. von Lilienfeld, J. Chem. Theor. Comput. 10, 3417
(2014).
[5] T. Helgaker, P. J rgensen, and J. Olsen, Molecular
Electronic-Structure Theory (Wiley, Chichester, Eng-
land, 2000).
[6] M. Rupp, A. Tkatchenko, K.-R. M uller, and O. A. von
Lilienfeld, Phys. Rev. Lett. 108, 058301 (2012).
[7] O. A. von Lilienfeld, Int. J. Quant. Chem. 113, 1676
(2013). | ca914bd3cfced5f10907d642b2d64b2f9260d67d | page_0003 |
5
[8] G. Montavon, M. Rupp, V. Gobre, A. Vazquez-
Mayagoitia, K. Hansen, A. Tkatchenko, K.-R. M uller,
and O. A. von Lilienfeld, New J. Phys. 15, 095003 (2013).
[9] R. Ramakrishnan, M. Hartmann, E. Tapavicza, and
O. A. von Lilienfeld, arXiv , 1504.01966 (2015).
[10] R. Ramakrishnan, P. O. Dral, M. Rupp, and O. A. von
Lilienfeld, J. Chem. Theor. Comput. 11, 2087 (2015).
[11] Z. Li, J. R. Kermode, and A. D. Vita, Phys. Rev. Lett.
114, 096405 (2015).
[12] Calculated as = (ref sample )=(1 ref).
[13] R. P. Feynman, Phys. Rev. 56, 340 (1939).
[14] L. M. Ghiringhelli, J. Vybiral, S. V. Levchenko, C. Draxl,
and M. Scheer, Phys. Rev. Lett. 114, 105503 (2015).
[15] O. A. von Lilienfeld, R. Ramakrishnan, M. Rupp, and
A. Knoll, Int. J. Quant. Chem. 115, 1084 (2015).
[16] T. Hastie, R. Tibshirani, and J. Friedman, The Ele-
ments of Statistical Learning , 2nd ed. (Springer, New
York, 2009).
[17] B. Sch olkopf and A. Smola, Learning with Kernels (MIT
Press, Cambridge, 2002).
[18] K. Hansen, G. Montavon, F. Biegler, S. Fazli, M. Rupp,
M. Scheer, O. A. von Lilienfeld, A. Tkatchenko, and
K.-R. M uller, J. Chem. Theor. Comput. 9, 3543 (2013).
[19] K. Vu, J. Snyder, L. Li, M. Rupp, B. F. Chen, T. Khelif,
K.-R. M uller, and K. Burke, Int. J. Quant. Chem. 115,
1115 (2015).
[20] L. Li, J. C. Snyder, I. M. Pelaschier, J. Huang, U.-
N. Niranjan, P. Duncan, M. Rupp, K.-R. M uller, and
K. Burke, arXiv , 1404.1333 (2014).
[21] J. Moody and C. J. Darken, Neural. Comput. 1, 281
(1989).
[22] Instead of max( p) min(p), which are extrema that
strongly
uctuate across subsets of the data, we use the
more robust 99 % and 1 % quantiles.
[23] L. Ruddigkeit, R. van Deursen, L. C. Blum, and J.-L.
Reymond, J. Chem. Inf. Model. 52, 2864 (2012).
[24] R. Ramakrishnan, P. O. Dral, M. Rupp, and O. A. von
Lilienfeld, Scienti c Data 1, 140022 (2014).
[25] K. Burke, J. Chem. Phys. 136, 150901 (2012).
[26] A. D. Becke, J. Chem. Phys. 140, 18A301 (2014).
[27] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev.
Lett. 77, 3865 (1996).
[28] F. Weigend and R. Ahlrichs, Phys. Chem. Chem. Phys.
7, 3297 (2005).
[29] F. Weigend, Phys. Chem. Chem. Phys. 8, 1057 (2006).
[30] C. Adamo and V. Barone, Chem. Phys. Lett. 298, 113
(1998).
[31] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria,M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone,
B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Cari-
cato, X. Li, H. P. Hratchian, A. F. Izmaylov, J. Bloino,
G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara, K. Toy-
ota, R. Fukuda, J.-Y. Hasegawa, M. Ishida, T. Naka-
jima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, J. Mont-
gomery, John A., J. E. Peralta, F. Ogliaro, M. Bearpark,
J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov,
R. Kobayashi, J. Normand, K. Raghavachari, A. Ren-
dell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi,
N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B.
Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts,
R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi,
C. Pomelli, J. W. Ochterski, R. L. Martin, K. Morokuma,
V. G. Zakrzewski, G. A. Voth, P. Salvador, J. J. Dan-
nenberg, S. Dapprich, A. D. Daniels, O. Farkas, J. B.
Foresman, J. V. Ortiz, J. Cioslowski, and D. J. Fox,
\Gaussian 09 revision D.01," Gaussian Inc., Wallingford
CT 2009.
[32] J. W. Ochterski, Vibrational Analysis in Gaussian , Gaus-
sian Inc., Wallingford, Connecticut, USA (1999).
[33] W. Schneider and W. Thiel, Chem. Phys. Lett. 157, 367
(1989).
[34] M. J. Mills and P. L. Popelier, Theor. Chim. Acta 131,
1137 (2012).
[35] J. J. P. Stewart, J. Comput. Chem. 10, 209 (1989).
[36] S. Amari, N. Fujita, and S. Shinomoto, Neural. Comput.
4, 605 (1992).
[37] K. Dybiec and A. Gry -Keller, Magn. Reson. Chem. 47,
63 (2009).
[38] D. Flaig, M. Maurer, M. Hanni, K. Braunger, L. Kick,
M. Thubauville, and C. Ochsenfeld, J. Chem. Theor.
Comput. 10, 572 (2014).
[39] P. d'Antuono, E. Botek, B. Champagne, M. Spassova,
and P. Denkova, J. Chem. Phys. 125, 144309 (2006).
[40] D. E. Hill, N. Vasdev, and J. P. Holland, Comput. Theor.
Chem. 1051 , 161 (2015).
[41] V. Myrseth, J. D. Bozek, E. Kukk, L. J. Sthre, and
T. D. Thomas, J. Electron Spectros. Relat. Phenom. 122,
57 (2002).
[42] T. Karlsen, K. J. B rve, L. J. Sthre, K. Wiesner,
M. B assler, and S. Svensson, J. Am. Chem. Soc. 125,
7866 (2002).
[43] A. Holme, K. J. B rve, L. J. Sthre, and T. D. Thomas,
J. Chem. Theor. Comput. 7, 4104 (2011).
[44] Gaussian's default convergence threshold of 1 mE h/a0
was used as target accuracy.
[45] Although the size of the representation may vary, it is
bounded from above by a constant. | ca914bd3cfced5f10907d642b2d64b2f9260d67d | page_0004 |
PlantPhysiol.(1987)84,613-618
0032-0889/87/84/0613/06/$0 1.00/0
ProtonTransport inPlasmaMembrane andTonoplast Vesicles
fromRedBeet(BetavulgarisL.)StorageTissue1
ACOMPARATIVE STUDY OFIONEFFECTS ONApHANDA'I
Received forpublication November 17,1986andinrevisedformFebruary 25,1987
JOHNL.GIANNINI ANDDONALD P.BRISKIN*
Department ofAgronomy, University ofIllinois,
ABSTRACT
Theprotontransport properties ofplasmamembrane andtonoplast
vesiclesisolatedfromredbeet(BetavalgarisL.)storagetissuewere
examined andcompared. Membrane vesiclesisolatedwith250millimolar
KOinthehomogenization mediaandrecovered atlowdensityfollowing
sucrosedensitygradientcentrifugation displayed characteristics ofproton
transport (nitrateinhibition, noinhibition byorthovanadate, pHoptimum
of7.75,pyrophosphate-driven protontransport) whichwereconsistent
withatonoplastoripgn.WhentheKGinthehomogenization medium
wasreplaced by250milhimolar KI,sealedmembrane vesicles were
recovered athigherdensities insucrosegradients anddisplayed proper-
ties(orthovanadate sensitivity, noinhibition bynitrate,pHoptimum of
6.5)consistent withaplasmamembrane origin.Acomparison ofanion
effects(potassium salts)uponApHandAtrevealed adirectcorrespond-
encebetweentherelativeabilityofanionstostimulate protontransport
andreduceA+.Fortonoplast vesicles, therelativeorderforthiseffect
wasKI>KBr-K>KCO3>K2SO4whileforplasmamembrane
vesicles, adifferent orderKI>KNO32KBr2KClO3>KG>K2S04
wasobserved. Protontransport inplasmamembrane andtonoplast
vesicleswasinhibited byfluoride;however, plasmamembrane vesicles
appeared tobemoresensitive tothisanion.Inordertocorrelate anion
effectsinthetwovesiclefractions withaniontransport, thekineticsof
anionstimulation ofsteady-state pHgradients established intheabsence
ofmonovalent ionswasexamined. Anionswereaddedaspotassium salts
andthetotalpotassium concentration (100millimolar) wasmaintained
throughtheadditionofK/Mes. Forplasmamembrane vesicles,chlorate
andnitratedisplayed saturation kineticswhilechloridedisplayed stimu-
lationofprotontransport whichfollowed alinearprofile.Fortonoplast
vesicles,thekineticsofchloridestimulation ofprotontransport displayed
asaturable component. Theresultsofthisstudyindicatedifferences in
protontransport properties ofthesetwovesicletypesandprovideinfor-
mationonconditions whereprotontransport inthetwofractions canbe
optimized.
Theplasmamembrane andtonoplast constitute majormem-
branebarriersfornutrient uptakeandcompartmentation in
higherplantcells.Associated witheachofthesemembranes are
energy-dependent systemsfortheprimarytransport ofprotons
whichresultintheproduction ofaninwardly directed proton
electrochemical gradient acrosstheplasmamembrane andan
outwardly directed protonelectrochemical gradient acrossthe
'Supported byUnitedStatesDepartment ofAgriculture Competitive
GrantNo.86-CRCR-1-1977 andaMcKnight Foundation Individual
Research Grantawarded toD.P.B.UrbanaIllinois61801
tonoplast (27andreferences therein). Thereissubstantial evi-
dencethatthemechanism ofenergycoupling toprotontransport
involves theactionofprotontranslocating ATPases ateach
membrane (20andreferences therein). Inaddition, protontrans-
portacrossthetonoplast canalsobeenergized byaproton
translocating pyrophosphatase (24,25).Theprotonelectrochem-
icalgradients established ateachmembrane, throughprimary
protontransport, canthenservetodrivethesecondary transport
ofothersolutesbyadditional carriersassociated withthemem-
branes(15).
Isolatedpreparations ofmembrane vesicleshaveproventobe
ausefulsystemfortheinvitrocharacterization oftransport
systemsassociated withplantmembranes (27andreferences
therein). Thisworkhasgenerally emphasized thestudyoftrans-
portsystems associated withthetonoplast sinceithasproven
muchmoredifficulttoisolatetransport competent vesiclesfrom
theplasmamembrane. Thesestudieswithisolatedtonoplast
vesicleshaveallowedthecharacterization ofthenitratesensitive
ATPaseassociated withthismembrane (19,21,22),theproton
transporting pyrophosphatase (24,25)andsecondary transport
carriersfornitrate(2),sucrose(8),sodium(3),andcalcium(4,
26).Inrecentstudies,sealedplasmamembrane vesicleshave
beenisolatedfromcorncoleoptiles (9),radishseedings(23),and
zucchini fruit(17).However, anextensive characterization of
thetransport properties ofthesemembrane preparations empha-
sizingcomparisons betweentheplasmamembrane andtonoplast
hasnotbeencarriedout.
Inaprevious report(13),wedescribed amethodforthe
selectiveisolationofsealedplasmamembrane ortonoplast ves-
iclesfromthestoragetissueofredbeet(BetavulgarisL).This
methodislargelybasedupontheinclusion of0.25MKIorKCI
inthehomogenization mediawhichresultsintheproduction of
sealedmembrane vesiclesderivedfromeithertheplasmamem-
brane(KI)orthetonoplast (KCI).Thisapparent selectivity in
theproduction ofsealedmembrane vesiclestogetherwiththe
abilitytoproducemembranes inlargequantityfromthebulky
storagetissueofredbeetsuggestthatthissystemwouldbeuseful
intheinvitrocharacterization andcomparison oftransport
takingplaceattheplasmamembrane andtonoplast. Thisstudy
willfocusupontheseaspects.
MATERIALS ANDMETHODS
PlantMaterial. Redbeet(BetavulgarisL.)storagerootswere
purchased commercially. Thetopsoftheplantswereremoved
andthestoragerootswerestoredinmoistvermiculite at2to
4°Cuntiluse.Allroottissueusedwasstoredatleast10dto
ensureuniformity inmembrane isolation (21).
Membrane Isolation. Plasmamembrane andtonoplast vesicles
wereisolatedaccording tothemethodofGiannini etal.(13).
613 | 71a2d86e587ff2f69e923f36dafd3195c5acb3c0 | page_0000 |
GIANNINI ANDBRISKIN
Storagerootswerepeeled,cutintosmallsquares, andthen
rapidlyplacedintoahomogenization mediacontaining 250mm
sucrose,2mMEDTA,2mmNa2ATP, 1.0%(w/v)BSA(fraction
Vpowder), 0.5%(w/v)polyvinylpyrrolidone (40,000molwt),2
mMPMSF,215mm,B-mercaptoethanol, 4mMDTE,10%(v/v)glycerol, and70mMTris/HCl (pH8.0).DTE,PMSF,andf3-mercaptoethanol wereaddedtothemediumjustpriortouse.In
theisolation ofplasmamembrane vesicles250mmKIwas
included inthehomogenization mediawhileintheisolation of
tonoplast vesicles,theKIwasreplaced byanidentical concen-
trationofKCI.Thetissuewastreatedwithhomogenization
medium priortohomogenization inavegetable juiceextractor
byvacuuminfiltration ina1.5:1medium:tissue ratiofor5min
aticetemperature.
Thestoragetissuehomogenate wasfilteredthrough4layersof
cheesecloth andthencentrifuged at13,000g(8,500rpm)for15
mininaSorvallGSArotor.The13,000gpelletwasdiscarded
andthesupematant wascentrifuged at80,000g(32,000rpm)for
30mininaBeckman type35rotortoobtainamicrosomal
membrane pellet.Themicrosomal membrane pelletwassus-
pendedin4mlofasuspension buffercontaining 250mm
sucrose,2mMBTP/Mes (pH7.0),0.2%(w/v)BSA,1mmPMSF
(addedfresh),10%(v/v)glycerol, 1mmDTE(addedfresh),and
gentlyhomogenized inadouncetypehomogenizer. Thesuspen-
sionofmembranes enrichedforeithersealedplasmamembrane
ortonoplast vesicleswaslayeredovera25/38%(w/w)discontin-
uoussucrosedensitygradient andthencentrifuged at100,000g
(25,000rpm)for2hinaBeckman SW28rotor.Thegradient
solutions werebuffered with1mMBTP/Mes (pH7.0)and
contained 1mMDTE.Themembranes presentateitherthe8/
25%(tonoplast) or25/38%(plasmamembrane) interface were
removed usingaPasteurpipet.Themembranes wereeitherused
immediately intransport assaysorstoredat-80Cafterfreezing
inliquidN2.Inthelattercase,fulltransport activitywasretained
forupto3months.
OpticalMeasurement oftheVesiclepHGradient andMem-
branePotential. Protontransport inmembrane vesicleswas
measured bythequenching ofquinacrine fluorescence (1,7).
Thestandard assaycontained 250mMsorbitol,3.75mMATP
(BTPsalt,assaypH),3.75mMMgSO4,2.5gMquinacrine, 25
mMBTP/Mes (pH6.5or7.75),50mmmonovalent ions(when
present)and50to150Mugofmembrane protein.Theassaywas
carriedoutatpH6.5forplasmamembrane vesiclesandpH7.75
fortonoplast vesicles.Theionophore reversible quench(lono-
phoreReversible; Q/mgprotein)wasdetermined following the
additionof5gMgramicidin D(7).Fluorescence measurements
werecarriedoutat25CusingaPerkin-Elmer Model203spec-
trofluorimeter withtheexcitation monochronometer setat430
nmandtheemission monochronometer setat500nm.Opticalmeasurement ofthevesiclemembrane potential wascarriedout
underthesameconditions asthemeasurement ofthevesiclepHgradientexceptthat15MmoxonolVreplacedquinacrine inthe
assay,theexcitation monochronometer wassetat590nm,and
theemission monochronometer wassetat650nm.Anyvariation
inthesereactionconditions areindicated in"Results andDis-
cussion."
ProteinAssay.Proteinwasdetermined bythemethodofBradford (5)usingBSAasastandard. TheBradford assayreagent
wasfilteredjustpriortouse.
RESULTS ANDDISCUSSION
Inprevious studies,itwasapparent thattheinclusion of
potassium salts(KIorKCI)atarelatively highconcentration(250mM)inthehomogenization medium hadadramatic effect
2Abbreviatons: PMSF,phenylmethylsulfonyl fluoride;BTP,Bis-Trisupontherecovery ofredbeetplasmamembrane orputative
tonoplast vesiclesinamicrosomal membrane fractionthatwere
sealedandcompetent incarrying outATPdependent proton
transport (13).Inclusion of250mmKIinthehomogenization
mediaresultedintheproduction ofsealedvesiclesdisplaying
ATPdependent protontransport inhibited byorthovanadate
andmigrating tothedensityexpected forplasmamembrane on
linearsucrosedensitygradients. Furthercharacterization ofthe
vesiclesrevealedproperties consistent withaplasmamembrane
origin.Although inourpreviousstudy,theinclusion of250mm
KCIwasdemonstrated toproducesealedvesiclesthatdisplayed
nitrateinhibited protontransport andbandedatalowerdensity
insucrosedensitygradients, furthercharacterization ofthese
vesicleswasnotcarriedout.Therefore, itwasimportant to
confirmatonoplast originforthelowdensityvesiclesproduced
duringhomogenization ofredbeetstoragetissueinthepresence
of250mmKCIpriortobeginning comparative studies.
Whenthelowdensityvesiclesproduced inthepresence of250
mMKCIwerecharacterized, properties consistent withatono-
plastoriginforthevesicleswerefound.Thecorresponding
properties ofsealedplasmamembrane vesiclesareshown(for
TableI)fromourprevious work(13)forcomparison. Proton
transport inthelowdensityfraction wasinsensitive toortho-
vanadate butinhibited bynitrate(TableI).Inaddition, thelow
densityvesiclescouldusePPiasasubstate todriveproton
transport. Incontrast, theplasmamembrane vesiclesdisplayed
protontransport whichwasinhibited byorthovanadate, stimu-
latedbynitrate,andcouldnotbeenergized withPPiasa
substrate. WhentheeffectofassaypHuponprotontransport
wasexamined forthelowdensityvesicles, theoptimum for
transport wasbroadwithpeakactivityoccurring atpH7.75(Fig.
1).Incontrast, thepHoptimum forprotontransport withthe
highdensityvesicleswassharpwithapeakatpH6.5(Fig.1);
similartothepHoptimum forplasmamembrane ATPaseactiv-
ity(6).
Fromtheseresults,itisapparent thatthelowdensityvesicles
displayproperties consistent withatonoplast origin(21,27).In
addition todifferential sensitivity tonitrateandvanadate, other
distinguishing featuresincludetheassociation ofPPidriven
protontransport andabroader,alkalinepHoptimum. These
resultsaresimilartothosefoundfortonoplast preparationsisolatedfromavarietyofplantspecies(27andreferences
therein). Forthesubsequent comparative studies,transport as-
sayswerecarriedoutunderoptimized conditions withtheassay
pHforplasmamembrane vesicletransport assaysat6.5andthe
assaypHfortonoplast membrane vesicletransport assaysat
7.75.
TheeffectofvariousanionsonApHandA'formation in
sealedplasmamembrane andtonoplast vesicles wasexamined
TableI.Characteristics ofProtonTransport inTonoplast andPlasma
Membrane VesiclesIsolatedfrom RedBeetStorageTissue
Treatment LowDensityVesicles HighDensityVesicles
%Q/min*mgQ/mg%Q/min -mgQ/mg
Control' 62(IOO)b 152(100) 61(100) 181(100)
Nitrate(100mM) 19(31) 28(18)65(107)202(112)
Vanadate (100Mm)62(100) 166(109) 16(26) 36(20)
-ATP,PKiC 22(35) 55(36)0 0
aControlassaywascarriedoutinthepresenceof250mMsorbitol,
3.75mMATP(BTPsalt,assaypH),3.75mmMgSO4,2.5uMquinacrine,
25mmBTP/Mes (pH6.5)(plasmamembrane) or7.75(tonoplast), 50
mMKCI,and100ggofmembrane protein. Protontransport was
measured bythequenching ofquinacrine fluorescence overa4min
periodasdescribed in"Materials andMethods." bValuesinparen-
thesesrefertothepercentofthecontrolassay. cPPireplaced ATP
propane. inthecontrolassayandwaspresent at3.75mm.614 PlantPhysiol.Vol.84,1987 | 71a2d86e587ff2f69e923f36dafd3195c5acb3c0 | page_0001 |
H+TRANSPORT INREDBEETPLASMA MEMBRANE ANDTONOPLAST VESICLES
z
0
0x
E
z
wL
z
0
0xes
zI
E
z
a
5.7566.256.56.7577.257.57.7588.25
ASSAY pH
FIG.1.EffectofpHonprotontransport byplasmamembrane and
tonoplast vesicles fromredbeetstoragetissue.Protontransport was
measured attheindicated assaypHbythequenching ofquinacrine
fluorescence asdescribed in"Materials andMethods."
usingfluorescence quenching ofApHsensing(quinacrine) and
sensing(oxonolV)probes(1,7).Inordertofacilitate the
presentation ofdatafromalargenumber offluorescence quench-
ingtraces,thedataforprotontransport arequantitated bothin
termsoftheinitialrateoffluorescence quenching andthefinal,
steadystatepHgradient revealed aftertheaddition of5AM
gramicidin D(fordiscussion seeRef.7).Thedatafor A'P
production ispresented intermsofasteadystatepotential which
wasrapidlyestablished (lessthan 1min)inplasmamembrane
andtonoplast vesiclesfollowing theaddition ofMg:ATP. The
potential remained constant forupto4minandwasrelatedto
abaseline determined aftertheaddition of5uMgramicidin D.
Thisrapidestablishment ofasteadystatemembrane potential
intheredbeetvesiclesdiffered fromtheresultsofBennett and
Spanswick (1)wheremembrane potentials intonoplast vesicles
fromcornrootsweremeasured usingoxonolVI.Instead of
remaining constant overashortperiodoftimefollowing the
addition ofMg:ATP, theinitial Alrapidlydeclined asthepH
gradientinthevesiclesincreased toasteadystatelevel.Sincethe
measured A*remained constant overtimeinthisstudy,the
steadystate valueswouldrepresent areasonable estimate of
theinitialpotential whichwouldberequired incomparisons of
Al'andApHundervariousexperimental conditions.
Therelativeorderbywhichanions(50mmpotassium salts)
stimulate protontransport intonoplast vesicles wasfoundtobe:
KI>KBr2KCI>KCl03 >K2SO4(TableII).Thisrelative
orderwasobserved bothforquantitation intermsoftherateof
fluorescence quenching andthesteadystatepHgradient. This
relativeorderforanionsinstimulating protontransport was
identical totheorderinwhichtheseanions arecapable of
reducing thevesicle A*(TableII).Theseresults aresimilarto
whatwasobserved fortonoplast vesiclesisolatedfromsugarbeet
storagetissue(7)whichindicate thatasubstantial partofthe
effectofanionsinstimulating protontransport isrelatedtoa
reduction inthemembrane potential throughcharge compen-
sation.Inaddition, thereisanadditional effectofmonovalent
anions,especially chloride, inthedirectstimulation oftonoplast
ATPhydrolytic activity(7,21,27).WhenATPase assays were
carriedoutwiththeredbeetvesiclesinthepresence ofgramicidin
D,monovalent anionsstimulated activity.However, thelevelof
stimulation wasthesameforthemonovalent anionstested, so
thatdirectstimulation oftonoplast ATPaseactivity cannot ac-
countfortheobserved relativeeffectiveness ofanionsinstimu-
latingprotontransport (datanotshown). Thiswassimilar to
whatwasfoundinprevious studiesusingsugarbeet tonoplast
vesicles(6).
Although theplasmamembrane ATPaseisdirectlystimulated
bycations(16andreferences therein), anionshaveeffects uponTableII.AnionEffectsuponApHandA^inTonoplastVesiclesfrom
RedBeetStorageTissue
ApH AA
Treatment lonophore lonophore
Initialrate reversible Reversible
%Q/min*mg Q/mg Q/mg
Controla 12.9 4.3 268.9
KI 57.9 125.0 48.6
KBr 55.7 120.7 66.4
KC1 52.1 107.1 75.7
KC103 28.6 66.4 116.0
K2SO4 20.0 42.3 164.3
aControlassaywascarriedoutinthepresenceof250mmsorbitol,
3.75mmATP(BTPsalt,pH7.75),3.75mMMgSO4,25mMBTP/Mes
(pH7.75),and100to150jigofmembrane protein.AcidinteriorpH
gradients weremeasured inthepresenceof2.5gMquinacrine while
interiorpositivemembrane potentials weremeasured inthepresenceof
15juMoxonolV,overa4minperiodasdescribed in"Materials and
Methods." Whenpotassium saltsofanionsweretested,theanioncon-
centration was50mM.
TableIII.AnionEffectsuponApHandA4'inPlasmaMembrane
VesiclesfromRedBeetStorageTissue
ApH A4
Treatment Ionophore Ionophore
Initialrate reversible Reversible
%Q/min*mg Q/mg Q/mg
Control' 12 25 1120
KI 136 525 250
KNO3 102 475 312
KBr 94 438 462
KCQ03 75 425 436
KCI 56 212 888
K2SO4 36 120 920
aControlassaywascarriedoutinthepresence of250mmsorbitol,
3.75mmATP(BTPsalt,pH6.5),3.75mMMgSO4,25mmBTP/Mes
(pH6.5),and100Mgofmembrane protein.AcidinteriorpHgradients
weremeasured inthepresenceof2.5Mmquinacrine whileinteriorpositive
membrane potentials weremeasured inthepresenceof15juMoxonolV,
overa4minperiodasdescribed in"Materials andMethods." When
potassium saltsofanionsweretested,theanionconcentration was50
mM.
theprotonpumping aspectofthisenzymewhenpresentinsealed
membrane vesicles(TableIII).Anionswerefoundtostimulate
protontransport inthefollowing relativesequence: KI>KNO3
2KBr>KC103>KCI>K2SO4.Aswithtonoplast vesicles,the
samerelativeorderforstimulation ofprotontransport byanionswasfoundwithrespecttotherelativeabilityofanionstoreduce
thevesiclemembrane potential.
Fromacomparison ofresultsobtained inTablesIIandIIIa
numberofinteresting pointscanbemade.Forbothmembrane
systems, KIgivesthegreatestenhancement ofprotontransport
throughchargecompensation. Intermsoftheeffectiveness of
themonovalent halidestested,thesequence I->Br->Cl-was
observed forbothplasmamembrane andtonoplast vesicles.This
sequence issimilartothesequence ofincreasing heatsofhydra-
tionanddecreasing ionicradii(11)sothattheseparameters may
beinvolved indetermining theselectivity ofmonovalent anion
movement, possiblythroughmembrane associated channels (12
andreferences therein).However, thisinterpretation issomewhat
uncertain sincesomedegreeofpassiveionconductance maybe
introduced inthevesiclesthrough damagetothemembranes
duringisolation. Nitrate,whichactstoinhibitthetonoplast615 | 71a2d86e587ff2f69e923f36dafd3195c5acb3c0 | page_0002 |
GIANNINI ANDBRISKIN
protonpumping ATPase,givessubstantial stimulation ofproton
transport inplasmamembrane vesicles.Sincenitrateinhibited,
ATPdependent protontransport isoftenusedtoquantitate
tonoplast vesiclesinisolatedmembrane preparations, problems
mayariseifnitrateinsensitive protontransport isusedasa
parameter toquantitate residualprotontransport associated with
plasmamembrane vesicles(9).Undertheseconditions, thetend-
encywillbetooverestimate therelativeamount ofplasma
membrane vesiclessinceprotontransport inthesemembranes
willbestimulated whiletonoplast protontransportwillbeinhib-
ited.Forthisreason,orthovanadate inhibited, ATPdependent
protontransport wouldappeartobeabetterquantitative marker
forthepresence ofsealedplasmamembrane vesicles(16,27).
Chlorate, ananionchemically similartonitrate(10),stimulates
protontransport inplasmamembrane vesicles,buttoamuch
lesserextentintonoplast vesicles. Whilethisanionhasbeen
showntodirectlyinhibittonoplast ATPaseactivitytoalesser
extentthannitrateinredbeettonoplast vesicles(measured in
thepresence ofgramicidin D)(14),thisapparent inhibition is
overshadowed byitseffectonreducing AIinthesealedvesicles
examined inthisstudy.Incontrast, inhibition ofthetonoplast
ATPasebynitrateissufficiently strongsothatprotontransport
inthepresence ofthisanionisgenerally lowerthanthelevel
observed intheabsenceofmonovalent ions(7,21).However, it
hasbeenshownthatnitratecantransiently stimulate proton
transport priortoinhibition whenaddedtotonoplast vesiclesin
whichapreexisting pHgradient hasbeenestablished byATP-
dependent protonpumping (2).Aswithchlorate, thisstimulation
wasrelatedtoareduction inthevesiclemembrane potential
withpossibletransport ofnitrateintothevesicleinterior(2).
Alsoofinterestisthemuchlowerstimulation ofproton
transport inplasmamembrane byCl-thanintonoplast vesicles.
Whilemostprotontransport assaysfortonoplast vesiclesare
carriedoutinthepresence ofKCI,theseresultssuggestthatan
optimized protontransport assayforplasmamembrane vesicles
shouldbecarriedoutinthepresence ofeitherKIorKNO3tomaximize bothdirectcationeffectsupontheATPase(6,16)and
anioneffectsuponprotontranslocation.
Inordertofurthercharacterize theseeffectsforselectedions
andtorelatetheireffectsonchargecompensation toionmove-
mentsacrossthevesiclemembrane, protontransport assayswere
carriedoutasdescribed byLewandSpanswick (18).Inthese
assays,vesiclesareallowedtogenerateasteadystateAgH'across
themembrane whichisdominated byA,.Thisiscarriedoutby
initiating protontransport withMg:ATP intheabsenceofcharge
compensating monovalent ions.Whenthefluorescence ofthe
ApHsensingprobe(quinacrine) reachesasteadystatelevelof
quenching, themonovalent ionisaddedwhichcausesanincrease
intheextentoffluorescence quenching. Asdescribed byLew
andSpanswick (18),thisistheresultofelectrophoretic transport
oftheionintothevesicleinteriorwithaconcomitant reduction
inA'Iandacorresponding increaseinApH.Itbecameapparent
duringourstudiesthatinordertousethistechnique toobtain
kineticdata(i.e.protontransport versusionconcentration), a
methodtocontrolcationconcentration wasrequired. Thisis
important sincecationssuchasK+directlystimulate theplasma
membrane ATPase(16,20)sothatitwouldbedifficult to
differentiate directcationstimulation fromanionreduction of
A'Iwhenusingpotassium salts.Ourapproach wastouseK+saltsofselectedanions,holdingthepotassium concentration
constantusingK/Mestomaximize thestimulation oftheplasma
membrane ATPase. Inaddition, itwasdesirable touseBTP/C1-tomaintain aconstant Cl1concentration whenexamining K+effectsonprotontransport. However, priortotheseexperiments
itwasimportant todetermine ifBTPorMeshadanyeffect
themselves uponprotontransport inplasmamembrane and
tonoplast vesicles.Inordertoaddressthisquestion, experiments werecarriedout
todetermine ifhighconcentrations ofBTPorMes(100mM)affectedprotontransport inthesealedvesicles.Theseexperi-
mentsinvolved measurements ofeffectsuponbothproton
pumping andthevesiclesmembrane potential (datanotshown).
Transport assayswithplasmamembrane vesiclescarriedoutin
thepresence ofeither100mmKCIor100mMK/Mes+100
mMBTP/C1showednodifference, indicating thatthesebuffers
didnotaffectprotontransport orthemembrane potential inthis
system.However, intransport assaysusingtonoplast vesicles,
bothprotonpumping andthemembrane potential werereduced
whenmeasured with100mMK/Mes+100mMBTP/Clas
compared to100mMKC1.Inordertodetermine, inthissitua-
tion,whetherBTPorMesactedtoinhibittonoplast transport,
assayswherecarriedoutinthepresence of100mMKCIand
either100mMMes(pH7.75withBTP)orBTP(pH7.75with
Mes).Onlytheassaycarriedoutinthepresence ofhighlevelsof
BTPshowedaninhibitory effectuponprotontransport. Al-
thoughthemembrane potential wascompletely collapsed under
theseconditions, experiments inwhichanionstimulation of
protonpumping wasobserved revealed nosubstantial effectof
100mMMes.
Thekineticsofanionstimulation ofprotontransport in
plasmamembrane andtonoplast vesiclesisshowninFigure2.
Fortheseexperiments, theK+concentration wasmaintained at
100mmwithK/Mesandtheanionconcentration wasvaried
usingK+salts.Inthisway,exposure ofthetonoplast vesiclesto
highlevelsofBTPwasavoided. Chloride stimulation ofproton
transport inredbeettonoplast vesiclesdemonstrated asaturable
component asobserved insoybean tonoplast vesicles(18).In
theseprevious studies,Cl-wasaddedasKCIsothatthesimilarityinresultsindicates thatK+hasaminimal effectontonoplast
vesicles.Incontrasttothepresence ofasaturable phaseinCl-effectswithtonoplast vesicles,Cl-stimulation ofprotontrans-
portinplasmamembrane vesiclesincreased linearlywithin-
creasedCl-whentheK+wasmaintained at100mm.Thisresult
wassurprising andsuggested theabsenceofcarriermediated
transport forwhatwouldcorrespond toCl-effluxinplantcells.However, previous studiesbyRasi-Caldogno etal.(23)onthe
plasmamembrane ATPaseinvesiclesfromradishdemonstrated
KCIstimulation ofprotontransport whichdisplayed saturation
kinetics. Inordertofurtherclarifythisresult,theeffectofK+concentration onpost-steady stateApHwasexamined bymain-
tainingtheCl-concentration constant at100mmusingBTP/
Cl-andaddingvariousconcentrations ofK+/Mes. Underthese
conditions, saturation kineticswereobserved forthestimulation
ofprotonpumping. Therefore, theresultsofRasi-Caldogno et
al.(23)aremostlikelyrelatedtoK+effectsupontheATPase
ratherthanCl-effectsuponchargecompensation.
Incontrasttothelinearstimulation ofprotontransport byC1-inplasmamembrane vesicles,NO3-andC103stimulation of
protontransport demonstrated saturation kinetics.Thestimula-
toryeffectofK+wasseparated fromtheseanioneffectsby
maintaining allassaysat100mmK+throughtheaddition of
K+/Mes. Inaddition, acontrolassayinvolving theaddition of
100mmK+/Mes wassubstracted fromeachassaytogivethe
stimulation ofprotonpumping duetoanioncompensation of
themembrane potential. Theidentical kineticprofilesforthese
twoanionsareconsistent withtheirchemical similarity andoften
similarbiochemical effects(10).Thisprofileofsaturation kinetics
wouldsuggesttheinvolvement ofamembrane carrierfornitrate
movement corresponding tonitrateeffluxinintactplantcells.Attempts weremadetomeasure thestimulation ofpost-steady
stateprotonpumping inthepresence ofC103-withtonoplast
vesicles.Thiswasunsuccessful undertheconditions ofthese
experiments usingKC103withK+maintained at100mM.Previous studiesbyLewandSpanswick (18)haveshownthat616 PlantPhysiol.Vol.84,1987 | 71a2d86e587ff2f69e923f36dafd3195c5acb3c0 | page_0003 |
H+TRANSPORT INREDBEETPLASMA MEMBRANE ANDTONOPLAST VESICLES61
61
5
4
z
w3
0
0.
*L
zi
z-
wi
beI14-
12-
0-
a
6'-
4'-
2
oi0 20 40 60 s0
IONCONCENTRATION, mm
FIG. 2.Ionstimulation ofproton transport inmembrane vesicles
with apHgradient established intheabsence ofmonovalent ions. Proton
transport assays were carried out asdescribed in"Materials andMethods"
intheabsence ofmonovalent ions and allowed toreach asteady state
level ofquinacrine fluorescence quenching. Following theattainment of
asteady state pHgradient, 20glaliquots ofionstocks were added to
yield theindicated final concentration ofanions ofK'.TheK' concen-
tration wasmaintained at100 mminanion stimulation experiments by
theaddition ofK~/Mes and the rate ofquenching observed inthe
presence of100mmsK~/Mes wassubtracted from each value tospecifi-
cally indicate anion stimulatory effects. Potassium stimulation ofquin-
acmie quenching was observed byadding 20MAIaliquots ofK~/Mes
stocks andmaintaining theCl- at100mmswithBTP/Cr-.
100
0.~~~
0soP Tonoplast
0
0 10 20 30 40 50 60
FLUORIDE CONCENTRATION, mM
FIG. 3.Fluoride inhibition ofproton transport inplasma membrane
andtonoplast vesicles. Proton transport was measured asdescribed in
"Materials and Methods" inthepresence oftheindicated concentration
offluoride (asKF). TheKIconcentration was maintained at100mms
through theaddition ofKI/Mes.plasmamembrane andtonoplast ATPaseinmembrane vesicles
fromsoybeanhavedifferential sensitivity toP-withtheplasmamembrane ATPasebeingdirectlyinhibitedbythisanion.The
effectsofP-onprotontransport intonoplast andplasma mem-
branevesiclesofredbeetwastestedinordertodetermine
whetherdifferential sensitivity tothisanionoccursinthisplanttissue.InFigure3,theeffectofincreasing concentration ofP-
(K'maintained at100mm)ontheinitialrateofquinacrinefluorescence quenching wasexamined. Unliketheresultspre-
sentedbyLewandSpanswick (18),protonpumping wasinhib-
itedbyfluorideinbothplasmamembrane andtonoplast vesicles
preparations. However, theplasmamembrane vesiclesappearedtobemoresensitive tothisinhibitor. Sincethevesicleprepara-
tionmethod usedinthisstudyusesbothdifferential vesicle
sealing(KCIversusKIinhomogenization media)anddensitygradientcentrifugation toproducesealedplasmamembrane and
tonoplast vesiclesfreefromcross-contamination oftransportingvesicles,theseresultsraisequestion totheproposal thatP-can
beusedasaspecificinhibitor forplasmamembrane transport
activity.Inconclusion, theresultsofthisstudyindicate substantial
differences intheproperties ofplasmamembrane andtonoplastvesicleswithrespecttopHoptimum forprotontransport, sen-
sitivitytobuffers,response toionsandtheabilitytousePPias
asubstrate todriveprotontransport. Thesedifferences may
represent characteristic featureswhichcouldbeusedtodistin-
guishthesevesiclepopulations. However, amajorpointdem-
onstrated bythisstudyisthedifference intheoptimum condi-
tionsforprotontransport inthetwovesiclefractions. Therefore,pundertheconditions whereitisdesirable tomaximize the
measurement ofplasmamembrane versustonoplast transport
activity, separate assayconditions mayneedtobeusedfor
optimal measurement ofprotontransport.
LITERATURE CITED
1.BENNErAB,RMSPANSWICK 1983Opticalmeasurements ofApHandA*Iin
cornrootmembrane vesicles:kineticanalysisofCl1effectsonaproton
translocating ATPase.JMembrBiol71:95-107
2.BLUMWALD E,RJPOOLE1985NitratestorageandretrievalinBetavulgarisL:
effectsofnitrateandchloride onprotongradients intonoplast vesicles.Proc
NatiAcadSciUSA82:3683-3687
3.BLUMWALD E,RiPOOLE1985Na'/H'antiportinisolatedtonoplast vesicles
fromstoragetissuesofBetavulgaris. PlantPhysiol78:163-:167
4.BLUMWALD E,RJPOOLE1986Kinetics ofCa`~/H'antiport inisolated
tonoplast vesiclesfromstoragetissueofBetavulgarisL.PlantPhysiol80:
727-731
5.BRADFORD MM1976Arapidandsensitive methodforthequantification of
microgram quantities ofproteinutilizingtheprinciple ofprotein-dye bind-
ing.AnalBiochem 72:243-254
6.BRISKINDP,RIPOOLE1983Characterization ofaKI-stimulated adenosine
triphosphatase associated withtheplasmamembrane ofredbeet.Plant
Physiol71:350-355
7.BRISKINDP,WRTHORNLEY, REWYSE1985Membrane transport'in isolated
vesiclesfromsugarbeet taproot.I.Isolation andcharacterization ofenergy-
dependent transporting vesicles.PlantPhysiol78:865-870
8.BRISKINDP,WRTHORNLEY, REWYSE1985Membrane transport inisolated
vesiclesfromsugarbeet taproot.II.Evidence forasucrose/H' antiport. Plant
Physiol78:871-875
9.DEMICHELIS MI,RMSPANSWICK 1986H'-pumping drivenbythevanadate-
sensitive ATPaseinmembrane vesiclesfromcornroots.PlantPhysiol81:
542-547
10.DEANE-DRUMMON6 CE,ADMGLASS1983Short-term studiesofnitrateuptakeintobarleyplantsusingion-specific eletrodes andMC103.I.Controlofnet
uptakebyN03-efflux.PlantPhysiol73:100-104
11.DIAMOND JM,EMWRIGHT 1969Biological membranes: thephysicalbasisof
ionandnonelectrolyte selectivity. AnnuRevPhysiol31:581-646
12.EISENMAN G,RHORN1983Ionicselectivity revisted: theroleofkineticand
equilibrium processes inionpermeation throughchannels. JMembrBiol
76:197-225
13.GIANNINI JL,LHGILDENSOPH, DPBRISKIN 1987Selective production of
sealedplasmamembrane vesiclesfromredbeet(BetavulgarisL)storage
tissue.ArchBiochemBiophys. 254:621-630
14.GRIFFITH CI,PAREA,EB7LU'MWALD, RIPOOLE1986Mechanism ofstimu-
lationandinhibition oftonoplast H+-ATPase ofBetavulgarisbychloride
andnitrate.PlantPhysiol81:120-125'oP1asMa Membrane VesiCles
0 ___________*~~~ N3
0CIO03U o~~~K+
;o~~~~~
Po
Tonoplast VesCicle
cl-617 | 71a2d86e587ff2f69e923f36dafd3195c5acb3c0 | page_0004 |
618 GIANNINI ANDBRISKIN
15.GUNNRB1980Co-andcounter-transport mechanism incellmembranes.
AnnuRevPhysiol42:249-259
16.LEONARD RT1984Membrane associated ATPases andnutrientabsorption by
roots.In:PBTinker,ALauchli.eds,Advances inPlantNutrition, VolI.
PraegerScientific, NewYork,pp209-240
17.LEwRR,NBUCHUNOW, RMSPANSWICK 1985ATP-dependent proton-
pumping activities ofzucchini fruitmicrosomes. Astudyoftonoplast and
plasmamembrane activities. BiochimBiophys Acta821:341-347
18.LEwRR,RMSPANSWICK 1985Characterization ofanioneffectsonthenitrate-
sensitiveATP-dependent protonpumpingactivityofsoybean(Glycine max
L.)seedlingrootmicrosomes. PlantPhysiol77:352-357
19.MANOLSON MF,PAREA,RJPOOLE1985Identification of3-044-ben-zoyl)benzyoladenosine 5'-triphosphate andN'N-dicyclohexylcarbodiimide-
bindingsubunitsofahigherplantH+-translocating tonoplast ATPase. JBiol
Chem260:12273-12279
20.MARREE,ABALLARIN-DENTi 1985Theprotonpumpsoftheplasmalemma
andthetonoplast ofhigherplants.JBioenerg Biomembr 17:1-21
21.POOLERJ,DPBRISKIN, ZKRATKY, RMJOHNSTONE 1984DensitygradientPlantPhysiol.Vol.84,1987
localization ofplasmamembrane andtonoplast fromstoragetissueof
growinganddormant redbeet.Characterization ofprotontransport and
ATPaseintonoplast vesicles.PlantPhysiol74:549-556
22.RANDALL SK,HSZE1986Properties ofthepartially purified tonoplast
pumping ATPasefromoatroots.JBiolChem261:1364-1371
23.RASI-COLDOGNO F,MCPUGLIARELLO, MIDEMICHAELIS 1985Electrogenic
transport ofprotonsdrivenbytheplasmamembrane ATPaseinmembrane
vesiclesfromradish:biochemical characterization. PlantPhysiol77:200-
205
24.REAPA,RJPOOLE1985Proton-translocating inorganic pyrophosphatase in
redbeet(BetavulgarisL.)tonoplast vesicles.PlantPhysiol77:46-52
25.REAPA,RJPOOLE1986Chromatographic resolution ofH+-translocating
pyrophosphatase fromH+-translocating ATPaseofhigherplanttonoplast.
PlantPhysiol81:126-129
26.SCHUMAKER K.HSzE1986Calcium transport intothevacuoleofoatroots.
Characterization ofH+/Ca2 exchange activity.JBiolChem261:12172-
12178
27.SzEH1985H+-translocating ATPases: advances usingmembrane vesicles.
AnnuRevPlantPhysiol36:175-208 | 71a2d86e587ff2f69e923f36dafd3195c5acb3c0 | page_0005 |
Aromatic Small Molecules Remodel Toxic Soluble Oligomers
of Amyloid /H9252through Three Independent Pathways*□S
Received for publication, August 19, 2010, and in revised form, October 26, 2010 Published, JBC Papers in Press, November 23, 2010, DOI 10.1074/jbc.M110.173856
Ali Reza A. Ladiwala‡, Jonathan S. Dordick‡§, and Peter M. Tessier‡1
From the Departments of‡Chemical & Biological Engineering and§Biology, Center for Biotechnology and Interdisciplinary Studies,
Rensselaer Polytechnic Institute, Troy, New York 12180
In protein conformational disorders ranging from Alzhei-
mer to Parkinson disease, proteins of unrelated sequence mis-
fold into a similar array of aggregated conformers rangingfrom small oligomers to large amyloid fibrils. Substantial evi-dence suggests that small, prefibrillar oligomers are the mosttoxic species, yet to what extent they can be selectively tar-geted and remodeled into non-toxic conformers using smallmolecules is poorly understood. We have evaluated the con-formational specificity and remodeling pathways of a diversepanel of aromatic small molecules against mature solubleoligomers of the A
/H925242 peptide associated with Alzheimer dis-
ease. We find that small molecule antagonists can be groupedinto three classes, which we herein define as Class I, II, and IIImolecules, based on the distinct pathways they utilize to re-model soluble oligomers into multiple conformers with re-duced toxicity. Class I molecules remodel soluble oligomersinto large, off-pathway aggregates that are non-toxic. More-over, Class IA molecules also remodel amyloid fibrils into thesame off-pathway structures, whereas Class IB molecules failto remodel fibrils but accelerate aggregation of freshly disag-gregated A
/H9252. In contrast, a Class II molecule converts soluble
A/H9252oligomers into fibrils, but is inactive against disaggregated
and fibrillar A /H9252. Class III molecules disassemble soluble oligo-
mers (as well as fibrils) into low molecular weight species thatare non-toxic. Strikingly, A
/H9252non-toxic oligomers (which are
morphologically indistinguishable from toxic soluble oligo-mers) are significantly more resistant to being remodeled thanA
/H9252soluble oligomers or amyloid fibrils. Our findings reveal
that relatively subtle differences in small molecule structureencipher surprisingly large differences in the pathways theyemploy to remodel A
/H9252soluble oligomers and related aggre-
gated conformers.
A central tenet of protein folding is that a given amino acid
sequence encodes a single folded structure (1). By analogy,
one would expect that a given protein sequence would encodea single misfolded structure (e.g. a single amyloid fibril confor-
mation). Instead, each protein sequence encodes numerousaggregated isoforms that possess unique secondary and terti-ary structures (2–12). Previous work has firmly establishedthat small, prefibrillar conformers (herein referred to as solu-
ble oligomers) of diverse polypeptides are the most toxic ag-gregates both in vitro andin vivo (11, 13–17). However, eluci-
dating the structural attributes of such toxic conformers thatdifferentiate them from their non-toxic counterparts hasproven difficult (see Refs. 11 and 18–22 for recent progress).
Significant evidence linking protein misfolding to cellular
toxicity in numerous aggregation disorders has motivated thesearch for small molecules that prevent aggregation (see Refs.23–25, and references therein). A general conclusion of thesestudies is that many small molecules redirect the aggregationcascade rather than inhibiting it completely (26). In hindsight,this finding is logical based on the large amount of buried sur-face area within protein aggregates compared with the smallsize of inhibitor molecules (27, 28). Therefore, using smallmolecules to alter the nucleation pathway by disrupting spe-cific intermolecular contacts or promoting atypical ones ap-pears to be a more feasible approach to preventing formationof toxic aggregates than antagonizing all possible intermolec-ular contacts.
Much less is known about the capacity of small molecules
to remodel mature protein aggregates (see Refs. 12 and 29–31for recent progress), despite the therapeutic importance ofabrogating toxic aggregates. This is surprising because it ismore complex to understand how small molecules alter theaggregation of monomers where proteins necessarily undergoconformational change (unless prevented by small molecules)than it is in the reverse direction where mature aggregatedconformers can be isolated that do not change structurallyduring experimentally relevant time scales. Nevertheless, dif-ficulties in forming homogeneous populations of differentaggregated conformers and discriminating between themhave hampered mechanistic studies of protein disaggregation.The development of several conformation-specific antibodiescapable of selectively detecting aggregated conformers rang-ing from intermediates (e.g. soluble oligomers (32–34), fibril-
lar oligomers (21), and annular protofibrils (35)) to end prod-ucts (i.e. fibrils (36, 37)) of amyloid assembly have been
critical to overcoming such challenges.
Indeed, such conformation-specific antibodies and related
biochemical assays are beginning to illuminate pathways em-
ployed by aromatic small molecules to remodel mature solu-ble oligomers of A
/H9252and other disease-associated proteins
(29–31, 38). Multiple polyphenols have been found recentlyto convert mature soluble oligomers of A
/H9252and Tau into off-
pathway, SDS-resistant aggregates that are non-toxic (12, 31,39). In fact, these and related studies suggest that conversion*This work was supported by the Alzheimer’s Association Grant NIRG-08-
90967 (to P. M. T.).
□SThe on-line version of this article (available at http://www.jbc.org) con-
tains supplemental Figs. S1–S8.
1To whom correspondence should be addressed: 110 Eighth St., Troy,
NY 12180. Fax: 518-276-3405; E-mail: tessier@rpi.edu.THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 5, pp. 3209 –3218, February 4, 2011
© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.
FEBRUARY 4, 2011• VOLUME 286 • NUMBER 5 JOURNAL OF BIOLOGICAL CHEMISTRY 3209
This is an Open Access article under the CC BY license. | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0000 |
of soluble oligomers into high molecular weight aggregates
may be a common remodeling pathway employed by othersmall molecules. Nevertheless, small molecules may neutral-ize the toxicity of mature A
/H9252soluble oligomers via other
mechanisms as well (38, 40). Herein, we demonstrate thatdiverse aromatic small molecules utilize three independentpathways to remodel mature A
/H9252soluble oligomers into be-
nign conformers with highly dissimilar biochemicalproperties.
EXPERIMENTAL PROCEDURES
Preparation of A /H9252Conformers—A /H925242 (American Peptide)
was dissolved in an aqueous, 50% acetonitrile solution (1 mg/ml), aliquoted, dried under vacuum and lyophilized, and thenstored at /H1100220 °C. The preparation of A
/H9252soluble oligomers,
non-toxic oligomers, and fibrils is described elsewhere (12).Briefly, A
/H9252soluble oligomers and non-toxic oligomers were
prepared by dissolving the peptide in 100% hexafluoroisopro-panol (Fluka). After the hexafluoroisopropanol was evapo-rated, the dried peptide was reconstituted in 50 m
MNaOH (1
mg/ml A /H9252), sonicated (30 s), and diluted in PBS (25 /H9262MA/H9252).
The peptide was then centrifuged (22,000 /H11003gfor 30 min),
and the pelleted fraction (5% of starting volume) was dis-carded. The supernatant was incubated at 25 °C for 0–6 dayswithout agitation. For preparing amyloid fibrils, aliquoted A
/H9252
was solubilized as described above (12), diluted into PBS (25
/H9262M), and mixed with pre-existing fibrils (10–20 wt % seed)
without mixing for 24 h at 25 °C.
Thioflavin T (ThT)2Assay—A /H9252(25/H9262M) was diluted with
ThT (44 /H9262M) at a ratio of 1:19 by volume. The fluorescence
was measured in 384-well microtiter plates (Microfluor 1,Thermo Fisher Scientific) using a Tecan Safire
2plate reader
(450/482 nm excitation/emission, 15 nm bandwidth). Seedingexperiments were conducted using A
/H9252monomers (25 /H9262M)
and various aggregated A /H9252conformers (20 wt % seed) without
agitation.
AFM—A /H9252samples (25 /H9262M) were spotted on cut mica
mounted on glass slides. The samples were adsorbed (30 min),and then washed with water and dried overnight. Images weretaken using an Asylum Research MFP three-dimensionalAFM system with Olympus AC240TS cantilevers.
Cell Toxicity Assay—The procedure for assaying the toxic-
ity of A
/H9252conformers is reported elsewhere (12). Briefly, rat
adrenal medulla cells (PC12, ATCC) were cultured in Dulbec-co’s modified Eagle’s medium (5% fetal bovine serum, 10%horse serum, and 1% penicillin-streptomycin). The cell sus-pension was aliquoted (90
/H9262l) into 96-well microtiter plates
(CellBIND, Corning) and allowed to adhere for 24 h. After-ward, A
/H9252or control samples (10 /H9262l) were added to microtiter
plates, and the cells were further incubated for 48 h at 37 °C.The medium was then removed, and fresh medium (200
/H9262l)
and thiazolyl blue tetrazolium bromide (Sigma; 50 /H9262lo f2 . 5
mg/ml) were added to each well for3ha t3 7° C .Finally, thesesolutions were discarded, 250
/H9262l of DMSO was added, and theabsorbance was measured at 562 nm. The toxicity values werenormalized to the PBS control without A
/H9252.
Gel Electrophoresis and Silver Staining—A /H9252samples (25
/H9262M) were diluted into sample buffer (Novex LDS, Invitrogen),
sonicated, analyzed using 10% BisTris gels (Invitrogen), andsilver stained (SilverXpress kit, Invitrogen).
Antibody Dot Blot Analysis—Dot blot analysis of A
/H9252con-
formers (25 /H9262M) was conducted as reported elsewhere (12).
Briefly, each A /H9252sample was spotted (1 /H9262l) on nitrocellulose
membranes (Hybond ECL, GE Healthcare). After the blotswere dried, they were blocked overnight at 4 °C (10% nonfatdry milk in PBST) and washed in PBST. The blots were thenincubated with A11 (Invitrogen), OC or 6E10 (Millipore)antibodies. The blots were washed and incubated with the ap-propriate horseradish peroxidase-conjugated secondary anti-body. After washing, the blots were exposed to substrate (ECLWestern blotting Substrate, Thermo Fisher) and developed.
Circular Dichroism Spectroscopy—The secondary struc-
tures of A
/H9252conformers (25 /H9262Min water; 300 /H9262l) were evalu-
ated using a Jasco 815 Spectrometer (1 mm path length cu-vette) at 25 °C. Each sample spectra is the average of at least10 readings.
ANS Fluorescence Analysis—8-Anilino-1-naphthalene
sulfonate (ANS, Sigma) was used at 7.5–10
/H9262Mto assay the
conformation of A /H9252(2.5/H9262M). The ANS fluorescence spectra
(/H9261ex/H11005380 nm) was measured using a Tecan Safire2plate
reader.
RESULTS
Aromatic Small Molecules Rapidly Remodel Mature A /H9252
Soluble Prefibrillar Oligomers—The goal of this work is toelucidate pathways employed by aromatic small molecules toremodel A
/H925242 toxic soluble oligomers into alternative con-
formers with low toxicity. To accomplish this, we utilized in
vitro assembly conditions that we have reported elsewhere
(12) to form homogeneous preparations of A /H9252soluble olig-
omers (25 /H9262M). Such soluble oligomers possess a combination
of biochemical properties and toxicities distinct from freshlydisaggregated and fibrillar A
/H9252conformers, including immu-
noreactivity with the conformation-specific antibody A11 thatselectively recognizes soluble prefibrillar oligomers (32). Wefirst employed this antibody to identify small molecules thatrapidly eliminate the A11 epitope of mature soluble olig-omers. We excluded small molecules that were inactive dur-ing a short time interval (4 h) relative to the time required forsoluble oligomers to convert into alternative structures (2days; data not shown). Based on our preliminary studies andprevious work (38–41), we identified seven aromatic mole-cules that rapidly remodel A
/H9252prefibrillar oligomers (Fig. 1;
structures of small molecules are shown in supplemental Fig.
S1). Four of these molecules are polyphenols or derivativesthereof (myricetin, tannic acid, piceid, and nordihydroguaia-retic acid or NDGA), one is a phenothiazine (methylene blue),and two are benzothiazoles (riluzole and benzothiazole). Allsmall molecules except benzothiazole and riluzole neutralizethe oligomer-specific conformational epitope at substoichio-metric concentrations, with tannic acid being the most effec-tive (Fig. 1).
2The abbreviations used are: ThT, thioflavin; NDGA, nordihydroguaiaretic
acid; BisTris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-
1,3-diol; ANS, 8-anilino-1-naphthalene sulfonate.Unique Pathways for Remodeling A /H9252Soluble Oligomers
3210 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 5 • FEBRUARY 4, 2011 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0001 |
Small Molecule Antagonists Utilize Unique Remodeling
Pathways—To elucidate whether these molecular antagonists
employ common or unique remodeling mechanisms to neu-tralize toxic A
/H9252soluble oligomers, we first performed SDS-
PAGE analysis of A /H9252soluble oligomers dosed with either thevehicle (1% DMSO) or a range of concentrations of each an-
tagonist (Fig. 2). We have included four A /H9252conformational
controls for comparison. Two A /H9252controls, soluble oligomers
and freshly disaggregated A /H9252peptide that we herein refer to
as soluble low molecular weight species, migrate on an SDS-PAGE gel as monomers, trimers, and tetramers. The othertwo A
/H9252conformational controls, fibrils and resveratrol-re-
modeled soluble oligomers, display different amounts of highmolecular weight aggregate. We previously found that res-veratrol-remodeled soluble oligomers are large off-pathwayaggregates that are non-toxic (12).
We find that myricetin and NDGA remodel A
/H9252soluble
oligomers into large, SDS-resistant aggregates that are indis-tinguishable from resveratrol-remodeled soluble oligomers atconcentrations /H1135020
/H9262M(Fig. 2A). We confirmed that these
aggregates are non-covalently associated by dissociating them
with heat (Fig. 2A). Benzothiazole and riluzole (200 /H9262M) also
convert A /H9252soluble oligomers into large SDS-resistant aggre-
gates similar to resveratrol-remodeled soluble oligomers (Fig.2B). In contrast, methylene blue (/H1135020
/H9262M) remodels soluble
oligomers into SDS-resistant aggregates that are indistin-guishable from A
/H9252fibrils (Fig. 2C). For each aromatic mole-
cule in Fig. 2, A–C, the concentration required to generate
SDS-resistant aggregates (Fig. 2, A–C) is identical to that re-
quired for eliminating the A11 epitope (Fig. 1). Finally, tannicacid and piceid do not convert A
/H9252soluble oligomers into
SDS-resistant aggregates (Fig. 2D). Based on these prelimi-
FIGURE 1. Analysis of the remodeling activity of aromatic small mole-
cules for A /H9252soluble oligomers. A/H9252soluble oligomers (25 /H9262M) were assem-
bled for 1 day (without agitation), and then incubated with and without
compounds for 4 h (25 °C). Each sample was spotted on nitrocellulose, andprobed with the conformation-specific antibody A11 (specific for solubleprefibrillar oligomers) and the sequence-specific antibody 6E10 (specific forthe N terminus of A
/H9252).
FIGURE 2. Characterization of SDS resistance of remodeled A /H9252soluble oligomers. A/H9252soluble oligomers (25 /H9262M; formed after 1–2 days without agita-
tion) were dosed with (A) myricetin and NDGA, (B) benzothiazole and riluzole, (C) methylene blue, and (D) tannic acid and piceid, and analyzed using SDS-PAGE. The A
/H9252conformational controls are freshly disaggregated A /H9252(soluble low molecular weight (LMW)), fibrils, soluble oligomers remodeled by resvera-
trol (20 /H9262M, 4 h), and soluble oligomers.Unique Pathways for Remodeling A /H9252Soluble Oligomers
FEBRUARY 4, 2011• VOLUME 286 • NUMBER 5 JOURNAL OF BIOLOGICAL CHEMISTRY 3211 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0002 |
nary findings, herein we refer to these molecules as Class I
(myricetin, NDGA, benzothiazole, and riluzole), II (methyleneblue), and III (tannic acid and piceid) antagonists, and we in-vestigate the pathways they employ to eradicate mature A
/H9252
soluble oligomers.
Class I Molecules Convert A /H9252Soluble Oligomers into Large,
Non-toxic Conformers—Our SDS-PAGE analysis in Fig. 2 re-vealed that Class I molecules utilize a pathway in which solu-ble oligomers are converted into large, SDS-resistant aggre-gates that appear similar to resveratrol-remodeled solubleoligomers (12). Because SDS may alter the size of the remod-eled A
/H9252soluble oligomers, we first performed AFM imaging
analysis of A /H9252soluble oligomers before and after addition of
Class I molecules to evaluate if A /H9252oligomers are remodeled
into large aggregates (Fig. 3A). Indeed, we find that all fourClass I molecules convert A
/H9252soluble oligomers into large ag-
gregates of similar morphology, whereas the vehicle does not.Moreover, circular dichroism analysis revealed that eachClass I molecule remodels soluble oligomers into unstruc-tured conformers that are similar to the secondary structuresof soluble oligomers prior to remodeling and significantly dif-ferent from
/H9252-sheet-rich fibrils (supplemental Fig. S2). Finally,
soluble oligomers display negligible ThT fluorescence beforeand after being remodeled by Class I molecules (supplementalFig. S2). Collectively, our findings reveal that Class I mole-cules convert soluble oligomers into off-pathway aggregatesthat are unstructured.We posited that the remodeled soluble oligomers would
possess low toxicity given their large size and lack of the A11epitope. Indeed, relative to freshly disaggregated A
/H9252(supple-
mental Fig. S3), we find that all four Class I molecules elimi-nate the toxicity of soluble oligomers in a dose-dependentmanner (Fig. 3B) that is in exact concordance with loss of theA11 epitope (Fig. 1) and gain of SDS-resistant aggregates (Fig.2,AandB). Thus, Class I molecules convert toxic soluble
oligomers into large aggregates that are non-toxic.
The similar morphologies, biochemical properties, and tox-
icities of the remodeled soluble oligomers suggests that ClassI molecules utilize a similar remodeling pathway. However, itis possible that Class I molecules remodel soluble oligomers atdifferent rates (or generate unique intermediates) during theremodeling process. Thus, we sought additional insights intothe kinetics of loss of the A11 epitope and whether
/H9252-sheet-
rich intermediates are transiently populated during the re-modeling process. We performed kinetic dot blot analysisusing conformation-specific antibodies specific for prefibrillaroligomers (A11) and fibrillar conformers (OC; supplemental
Fig. S2). We find that the four aromatic molecules eliminatethe A11 epitope after 3–4 h without generating OC-positiveconformers for at least 24 h (longer times not evaluated). Thesimilarity in the kinetics of remodeling and the absence ofOC-positive intermediates further suggests that Class I mole-cules employ a similar remodeling pathway to convert solubleoligomers into unstructured, off-pathway aggregates.
Despite the similar remodeling activity of Class I molecules,
polyphenols (myricetin and NDGA) and benzothiazoles (ri-luzole and benzothiazole) differ significantly in structure anddisplay unique dose dependence for remodeling soluble olig-omers (Figs. 1 and 2). Because benzothiazoles are more hy-drophobic than polyphenols, we suspected the former mole-cules utilize hydrophobic interactions to antagonize toxicsoluble oligomers that could be attenuated by addition of sur-factant. Thus, we evaluated the remodeling activity of Class Imolecules in the presence of 0.1% Triton X-100 using A11 dotblot analysis (supplemental Fig. S2). Indeed, we find that bothbenzothiazoles are inactive in the presence of surfactant,whereas the remodeling activity of polyphenols is unchanged.Moreover, we confirmed that both benzothiazoles did notform colloidal aggregates that sequester polypeptides in anonspecific manner (supplemental Fig. S2) (42). Thus, poly-phenols (herein referred to as Class IA molecules) and benzo-thiazoles (herein referred to as Class IB molecules) employsimilar remodeling pathways, yet they utilize unique interac-tions to convert soluble oligomers into high molecular weightaggregates.
A Class II Molecule Converts Prefibrillar A
/H9252Oligomers into
Fibrils—The SDS-PAGE analysis in Fig. 2C suggests that the
Class II molecule methylene blue converts soluble oligomersinto fibrils in a rapid manner. Indeed, we confirmed such aconformational transition using AFM, and found that A
/H9252sol-
uble oligomers convert into clumped fibrils (supplementalFig. S4) that can be readily dissociated with surfactant (2%SDS; Fig. 4A). We also performed dot blot analysis to probethe conformers that form after methylene blue remodeling(Fig. 4B). Strikingly, concentrations of methylene blue suffi-
FIGURE 3. Analysis of the pathway utilized by Class I molecules to re-
model A /H9252soluble oligomers. A/H9252soluble oligomers (25 /H9262M; formed after 1
day without agitation) were incubated with each small molecule (4 h) and
analyzed using (A) AFM imaging, and (B) cell toxicity analysis (n /H110053). In A,
mature soluble oligomers were incubated with the vehicle (1% DMSO) or200
/H9262Mcompound, and each image is 3 /H110033/H9262m.Unique Pathways for Remodeling A /H9252Soluble Oligomers
3212 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 5 • FEBRUARY 4, 2011 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0003 |
cient to eliminate the A11 epitope (/H1135020 /H9262M) correspond to
those that generate OC-positive conformers (Fig. 4B). These
dose dependent findings are in precise accordance with theconcentration dependence of methylene blue-mediated con-version of soluble oligomers into SDS-resistant aggregates(Fig. 2C). Moreover, we find that the A11 epitope is elimi-nated (and the OC epitope is generated) after2h( supplemen-
tal Fig. S4). These results strongly suggest that methylene bluerapidly remodels A
/H9252soluble oligomers into fibrils.
Nevertheless, we further characterized the biochemical
properties of these remodeled oligomers. We first evaluatedthe ThT fluorescence of the methylene blue-remodeledoligomers, and found that the same concentrations of methyl-ene blue that promote OC immunoreactivity also promotesignificant ThT fluorescence (Fig. 4C). Moreover, the highestconcentration of methylene blue (200
/H9262M) generates A /H9252con-
formers that display ThT fluorescence indistinguishable tofibrils (Fig. 4C). Moreover, we find that the remodeled struc-tures possess
/H9252-sheet secondary structures similar to fibrils
(supplemental Fig. S4).
We also evaluated the templating activity of methylene
blue-remodeled soluble oligomers because A /H9252fibrils effi-
ciently seed A /H9252monomers into ThT-positive conformers,whereas many other fibrillar intermediates do not (7, 12, 21).Importantly, we find that methylene blue-remodeled olig-omers template A
/H9252monomers in a manner that is similar to
mature fibrils (Fig. 4D). In contrast, soluble oligomers (whichwere not treated with methylene blue) fail to seed A
/H9252mono-
mers (Fig. 4D).
These findings provide strong evidence that methylene blue
converts soluble oligomers into fibrils. Because fibrils aremore toxic than A
/H9252monomers and off-pathway aggregates
but less toxic than soluble oligomers (14, 15, 43), we assayedthe toxicity of the remodeled soluble oligomers (Fig. 4E). Wefind that methylene blue concentrations /H1135020
/H9262Mconvert A /H9252
soluble oligomers into conformers that possess toxicities in-distinguishable from mature A
/H9252fibrils and significantly lower
than soluble oligomers. In summary, the Class II moleculemethylene blue rapidly remodels A
/H9252toxic soluble oligomers
into/H9252-sheet-rich fibrils.
Class III Molecules Disaggregate A /H9252Soluble Oligomers—
Unlike Class I and II molecules that promote high molecularweight aggregation, the SDS-PAGE analysis in Fig. 2D sug-
gests that Class III molecules tannic acid and piceid remodelsoluble oligomers into either SDS-soluble aggregates or lowmolecular weight species. To differentiate between these pos-sibilities, we used AFM analysis (Fig. 5A). Strikingly, we findthat both small molecules disaggregate A
/H9252soluble oligomers.
Moreover, we find that the remodeled oligomers are largelyunstructured and ThT-negative (supplemental Fig. S5). Theseremodeled oligomers also possess toxicities indistinguishableto freshly disaggregated A
/H9252, and both Class III molecules
eliminate the toxicity of soluble oligomers in a dose-depen-dent manner (Fig. 5B) identical to that observed for eliminat-ing the A11 epitope (Fig. 1). Finally, the kinetics of elimina-tion of the A11 epitope are similar for both molecules (A11
FIGURE 4. Analysis of the pathway employed by the Class II molecule
methylene blue to remodel A /H9252soluble oligomers. A/H9252soluble oligomers
(25/H9262M; formed after 1 day without agitation) were incubated with methyl-
ene blue (4 h), and analyzed using (A) AFM, (B) dot blots probed with con-
formation-specific (A11, soluble oligomer and OC, fibrillar conformer) andsequence-specific (6E10) antibodies, (C) ThT fluorescence (n /H110055), (D) tem-
plating activity (20% seed), and (E) cell toxicity analysis (n /H110053). The soluble
oligomers remodeled with methylene blue in Awere incubated with 2%
SDS to dissociate fibril clumps.
FIGURE 5. Analysis of the pathway utilized by Class III molecules to dis-
aggregate A /H9252soluble oligomers. A/H9252soluble oligomers (25 /H9262M; formed
after 1 day without agitation) were incubated with each small molecule(4 h) and analyzed using ( A) AFM imaging, and (B ) cell toxicity analysis (n /H110053).
InA, mature soluble oligomers were incubated with 2
/H9262Mtannic acid or 200
/H9262Mpiceid, and each image is 3 /H110033/H9262m.Unique Pathways for Remodeling A /H9252Soluble Oligomers
FEBRUARY 4, 2011• VOLUME 286 • NUMBER 5 JOURNAL OF BIOLOGICAL CHEMISTRY 3213 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0004 |
epitope is lost after 3–4 h; supplemental Fig. S5), and this re-
modeling activity is insensitive to addition of non-ionic sur-
factant (supplemental Fig. S5). Collectively, these resultsstrongly argue that Class III molecules remodel solubleoligomers into low molecular weight species that possess acombination of biochemical properties and toxicities charac-teristic of disaggregated A
/H9252.
Class I, II, and III Molecules Display Unique Remodeling
Activities for Fibrillar and Disaggregated A /H9252—Our findings
that aromatic small molecules utilize three unique remodelingpathways led us to investigate if they employ similar pathwaysto remodel other A
/H9252conformers or if these pathways are spe-
cific to A /H9252soluble oligomers. Thus, we first evaluated
whether each small molecule could rapidly remodel maturefibrils (Fig. 6). We find that Class IA (myricetin and NDGA)and III (tannic acid and piceid) molecules eliminate the OCepitope specific for fibrillar conformers, whereas Class IB(benzothiazole and riluzole) and II (methylene blue) mole-cules do not (Fig. 6A). Kinetic analysis of the fibril remodelingprocess revealed that each active molecule required 3–4 h toeliminate the OC epitope, and A11-positive intermediateswere not generated (supplemental Fig. S6). In each case, thedose-dependent loss of the OC epitope is mirrored preciselyby the dose-dependent loss of ThT fluorescence (supplemen-tal Fig. S6). In contrast, the Class IB and III molecules that failto eliminate the OC epitope also fail to eliminate the ThTfluorescence of A
/H9252fibrils (supplemental Fig. S6).
Given that soluble oligomers and fibrils possess signifi-
cantly different biochemical properties, we posited that ClassIA and III molecules would utilize different pathways to re-model fibrils relative to the pathways they employ to remodelsoluble oligomers. To test this hypothesis, we first performedSDS-PAGE analysis of fibrils before and after incubation witharomatic small molecules (Fig. 6B). We find that Class IAmolecules convert fibrils into SDS-resistant aggregates,whereas Class III molecules disassemble fibrils into speciesthat are indistinguishable from freshly disaggregated A
/H9252.W e
confirmed these findings using AFM (supplemental Fig. S6).Class IA molecules remodel fibrils into large aggregates (sup-plemental Fig. S6) that are indistinguishable from solubleoligomers after being remodeled by the same molecules (Fig.3A). In contrast, Class III molecules disaggregate fibrils intolow molecular weight species not detectable by AFM (supple-mental Fig. S6), as observed for soluble oligomers disaggre-gated by the same molecules (Fig. 5A). Finally, we confirmedthat each active small molecule eliminated the toxicity of A
/H9252
fibrils in a dose-dependent manner identical to that observedfor elimination of the OC- and ThT-epitopes (Fig. 6C). Incontrast, the inactive molecules did not prevent fibril-medi-ated toxicity (supplemental Fig. S6). Taken collectively, ourfindings strongly argue that Class IA and III molecules re-model A
/H9252fibrils through the same pathways they employ to
remodel soluble oligomers, and Class IB and II molecules areselective for remodeling A
/H9252soluble oligomers.
Several previous reports suggest that aromatic small
molecules can alter the aggregation pathway of freshly dis-aggregated A
/H9252, yet many such experiments were conducted
over long times or at high concentrations where A /H9252ma-
tures structurally even in the absence of small molecules(30, 38, 44, 45). Thus, we evaluated if the small moleculesinvestigated in this work would rapidly accelerate aggrega-tion of freshly disaggregated A
/H9252over a time interval (4 h)
in which A /H9252otherwise fails to aggregate when not agitated
(12). Strikingly, we find that Class IB molecules (benzothia-zole and riluzole) rapidly convert soluble A
/H9252into large ag-
gregates that are similar to soluble oligomers remodeledwith the same molecules (Fig. 3), whereas the other aro-matic small molecules were inactive ( supplemental Fig. S7).
FIGURE 6. Impact of Class I, II, and III molecules on mature A /H9252fibrils. A/H9252fibrils (25 /H9262M; formed via seeding for 1 day without agitation) were incubated
with each small molecule (4 h) and analyzed using (A) dot blots probed with conformation-specific (OC, fibrillar conformer) and sequence-specific (6E10)
antibodies, (B) SDS-PAGE, and (C) cell toxicity analysis (n /H110053).Unique Pathways for Remodeling A /H9252Soluble Oligomers
3214 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 5 • FEBRUARY 4, 2011 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0005 |
Moreover, we find that Class IB molecules do not acceler-
ate soluble A /H9252aggregation in the presence of non-ionic
surfactant (supplemental Fig. S7). Thus, Class IB moleculesremodel freshly disaggregated and oligomeric A
/H9252through
a common mechanism that is distinct from the other aro-matic molecules investigated in this work.
A
/H9252Non-toxic Oligomers Are More Resistant to Remodeling
Than Soluble Oligomers or Fibrils—We recently identified anoligomeric form of A
/H9252that is morphologically indistinguish-
able from A /H9252soluble oligomers, yet these oligomers are non-
toxic and possess a collection of biochemical properties notfound in other A
/H9252conformers (12). Interestingly, these
non-toxic oligomers are negative for ThT and multiple con-formation-specific antibodies (A11 and OC), yet are SDS-resistant. Thus, we employed AFM and SDS-PAGE analysis toevaluate the impact of each small molecule antagonist onthese conformers (Fig. 7 and supplemental Fig. S8). Strikingly,
we find that all aromatic small molecules were inactiveagainst these oligomers, including those that are active against
A
/H9252fibrils. Nevertheless, over longer time intervals (8–24 h),
we find that Class I molecules remodel non-toxic oligomersinto large SDS-resistant aggregates, whereas Class II mole-cules are inactive and Class III molecules disaggregate non-toxic oligomers into low molecular weight species (supple-mental Fig. S8). Thus, non-toxic oligomers are more resistantto being remodeled by diverse aromatic small molecules,which confirms that they possess structural features absent inother A
/H9252conformers.DISCUSSION
Our findings illuminate multiple pathways employed by
small molecules to remodel soluble oligomers of A /H925242 (Fig.
8). Indeed, even in the absence of small molecule antagonists,
we have reported previously that A /H9252soluble oligomers are
not committed to a single nucleation pathway (12). Instead,soluble oligomers can mature into multiple A
/H9252conformers
(fibrillar intermediates and non-toxic oligomers) that possesshighly dissimilar biochemical properties, and each conformercan be formed in a specific manner via simple changes in agi-tation. This multifaceted capacity of A
/H9252soluble oligomers to
mature into different isoforms is consistent with our findingsthat small molecules can remodel soluble oligomers into con-formers with significantly different sizes and structures. Thatonly four of the small molecules investigated in this work re-model mature A
/H9252fibrils (whereas all seven remodel soluble
oligomers) highlights the conformational plasticity of solubleoligomers to evolve into diverse conformers.
Our results for Class I molecules confirm that converting
soluble oligomers into large, unstructured aggregates is acommon pathway for remodeling toxic oligomers. We reasonthat it is simpler to promote nonspecific intermolecular inter-actions between A
/H9252peptides within soluble oligomers than it
is to promote specific ones necessary for fibril formation or to
FIGURE 7. Impact of Class I, II, and III molecules on mature A /H9252non-
toxic oligomers. A/H9252non-toxic oligomers (25 /H9262M; formed after 5 days
without agitation) were incubated with each small molecule (200 /H9262M
except for 2 /H9262Mtannic acid; 4 h) and analyzed using AFM imaging (each
image is 3 /H110033/H9262m).
FIGURE 8. Summary of pathways utilized by aromatic small molecules
to remodel A /H9252soluble oligomers. Class I molecules remodel A /H9252soluble
oligomers (ThT- and OC-negative, unstructured as judged by circular di-
chroism, highly toxic, SDS-soluble) into large aggregates that are SDS-resis-tant, negative for multiple conformational probes (OC and A11 antibodies,and ThT), unstructured, incompetent for templating A
/H9252monomers, and
non-toxic relative to freshly disaggregated A /H9252. In contrast, a Class II mole-
cule selectively converts soluble oligomers into fibrils (which are OC- andThT-positive,
/H9252-sheet rich, SDS-insoluble, A11-negative and mildly toxic).
Class III molecules convert soluble oligomers into low molecular species(which are OC-, A11-, and ThT-negative, unstructured, disaggregated as
judged by AFM, SDS-soluble and non-toxic relative to freshly disaggre-gated A
/H9252). Class IA molecules also convert mature fibrils into large off-
pathway structures, and these molecules are inactive against freshlydisaggregated A
/H9252. Class IB molecules convert A /H9252monomers into large
off-pathway aggregates, and are inactive against fibrils. Class II mole-cules are inactive against A
/H9252monomers and fibrils. Class III molecules
remodel fibrils into low molecular weight species, and they are inactiveagainst freshly disaggregated A
/H9252.Unique Pathways for Remodeling A /H9252Soluble Oligomers
FEBRUARY 4, 2011• VOLUME 286 • NUMBER 5 JOURNAL OF BIOLOGICAL CHEMISTRY 3215 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0006 |
prevent all possible intermolecular interactions. Studies of
small molecule inhibitors of protein aggregation also supportour findings because many such antagonists promote alterna-tive, unstructured aggregates rather than preventing conden-sation of protein monomers (30, 44, 45).
Nevertheless, we found that a Class II molecule (methyl-
ene blue) selectively converts soluble oligomers into fibrils,which is consistent with previous work (38). Importantly,methylene blue fails to remodel amyloid fibrils within 24 hor accelerate aggregation of freshly disaggregated A
/H9252
within 4 h (longer times not tested). Thus, the primary ac-tivity of methylene blue is to accelerate the conversion ofA
/H9252oligomers into fibrillar conformers. Because a key
structural difference between A /H9252soluble oligomers and
fibrils is the presence of /H9252-sheets, it is possible that Class II
molecules facilitate mature self-stacking interactions be-tween residues necessary for
/H9252-sheet formation. Alterna-
tively, Class II molecules may accelerate conversion of sol-uble oligomers into fibrils by inhibiting nonspecificinteractions sampled within A
/H9252soluble oligomers that gov-
ern assembly of non-toxic oligomers.
The conformational specificity of methylene blue to re-
model A /H9252soluble oligomers into fibrils makes it and re-
lated variants potential drug candidates to evaluate the hy-pothesis that A
/H9252soluble oligomers are the primary cellular
insult in Alzheimer disease. The therapeutic activity ofmany small molecules for remodeling toxic A
/H9252oligomers is
limited because they do not cross the blood-brain barrier.
In contrast, methylene blue crosses the blood-brain barrier,and possesses many other desirable drug properties, in-cluding high solubility and low toxicity (46–48). Indeed,methylene blue and derivatives of it are currently beingevaluated as therapeutic candidates to treat Alzheimer dis-ease (49).
Given that many small molecule antagonists fail to disag-
gregate mature A
/H9252soluble oligomers and fibrils into low mo-
lecular weight species (including the Class I and II moleculesstudied here), the activity of Class III molecules is intriguing.Tannic acid was the most active small molecule we studied(which is consistent with previous work (41, 50, 51)), and itappears that its large size enciphers this activity. However, theremarkable disaggregation activity of piceid is more difficultto explain. Piceid was found previously to partially disaggre-gate oligomers and amyloids of a more soluble A
/H9252peptide
fragment (residues 25–35; neither A /H9252conformer was con-
firmed with A11 or OC antibodies) (40). However, its disag-gregation activity for A11-positive soluble oligomers and OC-positive fibrils of A
/H925242 was unexpected not only because the
full-length peptide is significantly more aggregation-prone,but also because piceid is closely related to another polyphe-nol (resveratrol) that converts A
/H9252oligomers and fibrils into
large off-pathway aggregates (12). Because piceid is identicalto resveratrol except for the 3-O-
/H9252-glucosidic moiety, it ap-
pears that the sugar moiety is an important determinant of itsdisaggregation activity. Future structure-activity analysis isnecessary to resolve whether the stilbenoid backbone is im-portant for the disaggregation activity of piceid and whetherother phenolic glycosides possess similar disaggregation
activity.
Are there additional pathways by which aromatic small
molecules remodel toxic soluble oligomers into non-toxicassemblies? Indeed, we expect that several other pathwaysare possible. For example, soluble oligomers could be con-verted into various types of fibrillar conformers other thanfibrils or precipitated into large aggregates without struc-tural changes (thereby reducing toxicity due to increasedsize). Moreover, some aromatic small molecules modifypolypeptides covalently (52, 53) and are likely to remodelmature soluble oligomers through unique pathways notconsidered here.
A puzzling finding of our work is that Class IA polyphenols
selectivity remodel A
/H9252soluble oligomers and fibrils into in-
distinguishable, unstructured aggregates with identical dosedependence. A simple explanation would be that solubleoligomers and fibrils possess common structural features rec-ognized by Class IA molecules. However, we were unable todemonstrate a single biochemical property shared by thesetwo conformers. We find that A
/H9252soluble oligomers are A11-
positive, OC-negative, ThT-negative, SDS-soluble, highlytoxic and largely unstructured, whereas fibrils possess the“opposite” properties (i.e. they are A11-negative, OC-positive,
ThT-positive, SDS-insoluble, mildly toxic and
/H9252-sheet rich).
Because polyphenols are suspected to antagonize /H9266-stacking
interactions between aromatic side chains in amyloids (54), itmay be that such side chains stack in soluble oligomers (de-spite that they lack mature
/H9252-sheets detectable by circular
dichroism) as they do in A /H9252fibrillar oligomers and fibrils (21,
55). Such stacking interactions between aromatic side chainsof A
/H9252may render soluble oligomers and fibrils sensitive to
being remodeled by polyphenols. It is also possible that ClassIA molecules recognize other structural motifs that may becommon to soluble oligomers and fibrils, such as sheet-likestructures in both prefibrillar oligomers (
/H9251-pleated (56), anti-
parallel (57), or disordered (58, 59) sheets) and fibrils (parallel(55) or anti-parallel (60)
/H9252-sheets), although additional analy-
sis is needed to test these speculative hypotheses.
A striking finding of our work is that non-toxic oligomers
are more resistant to being remodeled by diverse aromaticsmall molecules than all other A
/H9252conformers, including
highly stable A /H9252fibrils. A possible explanation is that non-
toxic oligomers lack aromatic stacking interactions, whichallows them to more effectively evade the remodeling activityof the small molecules studied in this work. This hypothesiswould also predict that substituting aromatic amino acids fornon-aromatic ones would desensitize soluble oligomers to theremodeling activity of polyphenols. However, it is also possi-ble that A
/H9252non-toxic oligomers are simply more stable than
other A /H9252conformers (61), which would also explain their
increased resistance to being remodeled by diverse aromaticmolecules.
It is notable that two oligomeric forms of a non-disease
associated protein (HypF) have been reported to also possessunique toxicities (11). Similar to A
/H9252soluble and non-toxic
oligomers reported here, the N-terminal domain of HypF (de-noted HypF-N) forms two oligomeric forms, one of which isUnique Pathways for Remodeling A /H9252Soluble Oligomers
3216 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 5 • FEBRUARY 4, 2011 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0007 |
toxic. Chiti and co-workers (11) demonstrate that the toxic
conformer is less structured, which is generally consistentwith our findings that soluble oligomers are less stable thannon-toxic oligomers (as judged by SDS resistance). However,both HypF-N oligomers bind ThT and possess extensive
/H9252-sheets, confirming that they are fibrillar intermediates
rather than prefibrillar oligomers. In contrast, we find that A /H9252
soluble and non-toxic oligomers possess random coil struc-tures (as judged by circular dichroism) and, thus, the rele-vance of these and related (62) observations remains to beevaluated for A
/H9252soluble and non-toxic oligomers. We expect
that future site-specific structural studies of A /H9252toxic and
non-toxic oligomers will illuminate the structural differencesbetween these oligomers that confer their unique toxicitiesand susceptibilities to being remodeling by aromatic smallmolecules.
Acknowledgments—We greatly appreciate the gift of the OC anti-
body from Dr. Charles Glabe (University of California, Irvine), andassistance with the figures from Dr. Juan Mauricio Mora Pale.
REFERENCES
1. Anfinsen,C.B.(1973) Science 181,223–230
2. Frost,B.,Ollesch,J.,Wille,H.,andDiamond,M.I.(2009) J. Biol. Chem.
284,3546–3551
3. Chien,P.,andWeissman,J.S.(2001) Nature 410,223–227
4. Chien,P.,DePace,A.H.,Collins,S.R.,andWeissman,J.S.(2003) Na-
ture424,948–951
5. Tanaka,M.,Chien,P.,Naber,N.,Cooke,R.,andWeissman,J.S.(2004)
Nature 428,323–328
6. Tessier,P.M.,andLindquist,S.(2007) Nature 447,556–561
7. Goldsbury,C.,Frey,P.,Olivieri,V.,Aebi,U.,andMu ¨ller,S.A.(2005) J.
Mol. Biol. 352,282–298
8. Yamaguchi,K.,Takahashi,S.,Kawai,T.,Naiki,H.,andGoto,Y.(2005) J.
Mol. Biol. 352,952–960
9. Petkova,A.T.,Leapman,R.D.,Guo,Z.,Yau,W.M.,Mattson,M.P.,and
Tycko,R.(2005) Science 307,262–265
10. Paravastu,A.K.,Leapman,R.D.,Yau,W.M.,andTycko,R.(2008) Proc.
Natl. Acad. Sci. U.S.A. 105,18349–18354
11. Campioni,S.,Mannini,B.,Zampagni,M.,Pensalfini,A.,Parrini,C.,
Evangelisti,E.,Relini,A.,Stefani,M.,Dobson,C.M.,Cecchi,C.,and
Chiti,F.(2010) Nat. Chem. Biol. 6,140–147
12. Ladiwala,A.R.,Lin,J.C.,Bale,S.S.,Marcelino-Cruz,A.M.,Bhatta-
charya,M.,Dordick,J.S.,andTessier,P.M.(2010) J. Biol. Chem. 285,
24228–24237
13. Ono,K.,Condron,M.M.,andTeplow,D.B.(2009) Proc. Natl. Acad.
Sci. U.S.A. 106,14745–14750
14. Glabe,C.G.(2008) J. Biol. Chem. 283,29639–29643
15. Chiti,F.,andDobson,C.M.(2006) Annu. Rev. Biochem. 75,333–366
16. Cleary,J.P.,Walsh,D.M.,Hofmeister,J.J.,Shankar,G.M.,Kuskowski,
M.A.,Selkoe,D.J.,andAshe,K.H.(2005) Nat. Neurosci. 8,79–84
17. Silveira,J.R.,Raymond,G.J.,Hughson,A.G.,Race,R.E.,Sim,V.L.,
Hayes,S.F.,andCaughey,B.(2005) Nature 437,257–261
18. Ahmed,M.,Davis,J.,Aucoin,D.,Sato,T.,Ahuja,S.,Aimoto,S.,Elliott,
J.I.,VanNostrand,W.E.,andSmith,S.O.(2010) Nat. Struct. Mol. Biol.
17,561–567
19. Yu,L.,Edalji,R.,Harlan,J.E.,Holzman,T.F.,Lopez,A.P.,Labkovsky,
B.,Hillen,H.,Barghorn,S.,Ebert,U.,Richardson,P.L.,Miesbauer,L.,Solomon,L.,Bartley,D.,Walter,K.,Johnson,R.W.,Hajduk,P.J.,andOlejniczak,E.T.(2009) Biochemistry 48,1870–1877
20. Chimon,S.,Shaibat,M.A.,Jones,C.R.,Calero,D.C.,Aizezi,B.,and
Ishii,Y.(2007) Nat. Struct. Mol. Biol. 14,1157–1164
21. Wu,J.W.,Breydo,L.,Isas,J.M.,Lee,J.,Kuznetsov,Y.G.,Langen,R.,andGlabe,C.(2010) J. Biol. Chem. 285,6071–6079
22. Zhang,A.,Qi,W.,Good,T.A.,andFernandez,E.J.(2009) Biophys. J.
96,1091–1104
23. Bulic, B., Pickhardt, M., Schmidt, B., Mandelkow, E. M., Waldmann,
H., and Mandelkow, E. (2009) Angew. Chem. Int. Ed. Engl. 48,
1740–1752
24. Amijee,H.,andScopes,D.I.(2009) J. Alzheimers Dis. 17,33–47
25. Stains,C.I.,Mondal,K.,andGhosh,I.(2007) Chem. Med. Chem. 2,
1674–1692
26. Kodali,R.,andWetzel,R.(2007) Curr. Opin. Struct. Biol. 17,48–57
27. Roberts,B.E.,andShorter,J.(2008) Nat. Struct. Mol. Biol. 15,544–546
28. Wells,J.A.,andMcClendon,C.L.(2007) Nature 450,1001–1009
29. Roberts,B.E.,Duennwald,M.L.,Wang,H.,Chung,C.,Lopreiato,N.P.,
Sweeny,E.A.,Knight,M.N.,andShorter,J.(2009) Nat. Chem. Biol. 5,
936–946
30. Wang,H.,Duennwald,M.L.,Roberts,B.E.,Rozeboom,L.M.,Zhang,
Y.L.,Steele,A.D.,Krishnan,R.,Su,L.J.,Griffin,D.,Mukhopadhyay,S.,
Hennessy,E.J.,Weigele,P.,Blanchard,B.J.,King,J.,Deniz,A.A.,Buch-wald,S.L.,Ingram,V.M.,Lindquist,S.,andShorter,J.(2008) Proc. Natl.
Acad. Sci. U.S.A. 105,7159–7164
31. Bieschke,J.,Russ,J.,Friedrich,R.P.,Ehrnhoefer,D.E.,Wobst,H.,
Neugebauer,K.,andWanker,E.E.(2010) Proc. Natl. Acad. Sci. U.S.A.
107,7710–7715
32. Kayed,R.,Head,E.,Thompson,J.L.,McIntire,T.M.,Milton,S.C.,Cot-
man,C.W.,andGlabe,C.G.(2003) Science 300,486–489
33. Lambert,M.P.,Viola,K.L.,Chromy,B.A.,Chang,L.,Morgan,T.E.,Yu,
J.,Venton,D.L.,Krafft,G.A.,Finch,C.E.,andKlein,W.L.(2001)
J. Neurochem. 79,595–605
34. Zameer,A.,Kasturirangan,S.,Emadi,S.,Nimmagadda,S.V.,andSierks,
M.R.(2008) J. Mol. Biol. 384,917–928
35. Kayed,R.,Pensalfini,A.,Margol,L.,Sokolov,Y.,Sarsoza,F.,Head,E.,
Hall,J.,andGlabe,C.(2009) J. Biol. Chem. 284,4230–4237
36. O’Nuallain,B.,andWetzel,R.(2002) Proc. Natl. Acad. Sci. U.S.A. 99,
1485–1490
37. Kvam,E.,Nannenga,B.L.,Wang,M.S.,Jia,Z.,Sierks,M.R.,and
Messer,A.(2009) PLoS One 4,e5727
38. Necula,M.,Breydo,L.,Milton,S.,Kayed,R.,vanderVeer,W.E.,Tone,
P.,andGlabe,C.G.(2007) Biochemistry 46,8850–8860
39. Taniguchi,S.,Suzuki,N.,Masuda,M.,Hisanaga,S.,Iwatsubo,T.,Goe-
dert,M.,andHasegawa,M.(2005) J. Biol. Chem. 280,7614–7623
40. Rivie `re,C.,Delaunay,J.C.,Immel,F.,Cullin,C.,andMonti,J.P.(2009)
Neurochem. Res. 34,1120–1128
41. Ono,K.,Hasegawa,K.,Naiki,H.,andYamada,M.(2004) Biochim. Bio-
phys. Acta 1690,193–202
42. Feng,B.Y.,Toyama,B.H.,Wille,H.,Colby,D.W.,Collins,S.R.,May,
B.C.,Prusiner,S.B.,Weissman,J.,andShoichet,B.K.(2008) Nat.
Chem. Biol. 4,197–199
43. Walsh,D.M.,andSelkoe,D.J.(2007) J. Neurochem. 101,1172–1184
44. Necula,M.,Kayed,R.,Milton,S.,andGlabe,C.G.(2007) J. Biol. Chem.
282,10311–10324
45. Ehrnhoefer,D.E.,Bieschke,J.,Boeddrich,A.,Herbst,M.,Masino,L.,
Lurz,R.,Engemann,S.,Pastore,A.,andWanker,E.E.(2008) Nat. Struct.
Mol. Biol. 15,558–566
46. Peter,C.,Hongwan,D.,Ku ¨pfer,A.,andLauterburg,B.H.(2000) Eur.
J. Clin. Pharmacol. 56,247–250
47. Ku ¨pfer,A.,Aeschlimann,C.,Wermuth,B.,andCerny,T.(1994) Lancet
343,763–764
48. DiSanto,A.R.,andWagner,J.G.(1972) J. Pharm. Sci. 61,1086–1090
49. Gura,T.(2008) Nat. Med. 14,894
50. Boye ´-Harnasch,M.,andCullin,C.(2006) J. Biotechnol. 125,222–230
51. Kocisko,D.A.,Baron,G.S.,Rubenstein,R.,Chen,J.,Kuizon,S.,and
Caughey,B.(2003) J. Virol. 77,10288–10294
52. Zhu,M.,Rajamani,S.,Kaylor,J.,Han,S.,Zhou,F.,andFink,A.L.(2004)
J. Biol. Chem. 279,26846–26857
53. Meng,X.,Munishkina,L.A.,Fink,A.L.,andUversky,V.N.(2009) Bio-
chemistry 48,8206–8224
54. Gazit,E.(2002) FASEB J. 16,77–83
55. Margittai,M.,andLangen,R.(2008) Q. Rev. Biophys. 41,265–297Unique Pathways for Remodeling A /H9252Soluble Oligomers
FEBRUARY 4, 2011• VOLUME 286 • NUMBER 5 JOURNAL OF BIOLOGICAL CHEMISTRY 3217 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0008 |
56. Armen,R.S.,DeMarco,M.L.,Alonso,D.O.,andDaggett,V.(2004)
Proc. Natl. Acad. Sci. U.S.A. 101,11622–11627
57. Cerf,E.,Sarroukh,R.,Tamamizu-Kato,S.,Breydo,L.,Derclaye,S.,Duf-
re ˆne,Y.F.,Narayanaswami,V.,Goormaghtigh,E.,Ruysschaert,J.M.,
andRaussens,V.(2009) Biochem. J. 421,415–423
58. Manavalan,P.,andJohnson,W.C.(1983) Nature 305,831–832
59. Gursky,O.,andAleshkov,S.(2000) Biochim. Biophys. Acta 1476,93–102
60. Tycko,R.,Sciarretta,K.L.,Orgel,J.P.,andMeredith,S.C.(2009) Bio-
chemistry 48,6072–6084
61. Lee, S., Fernandez, E. J., and Good, T. A. (2007) Protein Sci. 16,
723–732
62. Nekooki-Machida,Y.,Kurosawa,M.,Nukina,N.,Ito,K.,Oda,T.,and
Tanaka,M.(2009) Proc. Natl. Acad. Sci. U.S.A. 106,9679–9684Unique Pathways for Remodeling A /H9252Soluble Oligomers
3218 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 5 • FEBRUARY 4, 2011 | 1bb858e6182f64414b1370218ba794832fbfe662 | page_0009 |
\Wh6le Number ,
CCLXXIX. AUGUST, 1885.Number i.
Volume XXV.
TheBUFFALO
Medical Surgical Jurnal
EDITORS:
• r HdsNiLdxjEmoi ',. m. n. A. R. DAVIDSON, M. ».
Published by the Medical Journal Association .
No. 8 WEST CHIPPEWA ST.
Entered at the Post-Office at Buffalo, N. Y., as Second-class Mail Matter-
| b1a421a1f7c2787f0729682e2144aa57f5ca327e | page_0000 |
PDF hosted at the Radboud Repository of the Radboud University
Nijmegen
The following full text is a publisher's version.
For additional information about this publication click this link.
http://hdl.handle.net/2066/98925
Please be advised that this information was generated on 2023-04-03 and may be subject to
change. | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0000 |
Deceleration and trapping of ammonia using time-varying electric fields
Hendrick L. Bethlem, Floris M. H. Crompvoets, Rienk T. Jongma, Sebastiaan Y. T. van de Meerakker, and Gerard Meijer
FOM–Institute for Plasma Physics Rijnhuizen, P.O. Box 1207, NL-3430 BE Nieuwegein, The Netherlands
and Department of Molecular and Laser Physics, University of Nijmegen, Toernooiveld 1, NL-6525 ED Nijmegen, The Netherlands
~Received 1 February 2002; published 10 May 2002 !
Apolar molecule experiences a force in an inhomogeneous electric field. Using this force, neutral molecules
can be decelerated and trapped. It is shown here that this can in principle be done without loss in phase-spacedensity. Using a series of 64 pulsed inhomogeneous electric fields a supersonic beam of ammonia molecules(
14NH3,14ND3,15ND3) is decelerated. Subsequently, the decelerated molecules are loaded into an electro-
static quadrupole trap. Densities on the order of 107molecules/cm3at a temperature of 25 mK are obtained for
14ND3and15ND3separately and simultaneously. This corresponds to a phase-space density in the trap of
2310213, 50 times less than the initial phase-space density in the beam.
DOI: 10.1103/PhysRevA.65.053416 PACS number ~s!: 33.80.Ps, 33.55.Be, 39.10. 1j
I. INTRODUCTION
Currently there is great interest in the physics of cold
molecules @1–5#. There are now three techniques that have
been able to produce samples of trapped cold molecules. Inthe first technique, homonuclear diatomic molecules areformed by photoassociation of cold atoms @6,7#. In this way
ultracold H
2@8#,H e2*@9#,L i2@10#,N a2@11#,K2@12,13 #,C a2
@14#,R b2@15,16 #, and Cs 2@17#have been formed.The trans-
lational temperature of the molecules can be as low as that ofthe atoms from which they are formed. Recently, photoasso-ciation of atoms has been demonstrated in a Bose-Einsteincondensate @18,19 #. In most cases, the formation of mol-
ecules is deduced from an observed reduction in the numberof trapped atoms as a function of the wavelength of the pho-toassociation laser. In some cases, the produced moleculeshave also been detected directly @13,16,17 #. Formation of
mixed-alkali species is also being pursued @20,21 #, but has
turned out to be considerably more difficult. Once formed,the molecules can be trapped in the focus of an intense laserbeam @22#. In their triplet state, the molecules can also be
trapped in an inhomogeneous magnetic field. Recently, mag-
netic trapping of 2 310
5Cs2molecules at a temperature of
50mK has been reported @23#.
In the second technique molecules are cooled via colli-
sions with a helium buffer gas in a cryogenic cell. For3He,
temperatures down to 250 mK can be reached at buffer gasdensities that are still sufficiently high to allow for efficient
cooling. In this way, about 10
8CaH molecules have been
cooled and magnetically trapped at a temperature of 400 mK@24#. This method is applicable to any paramagnetic mol-
ecule ~or atom !provided that a method is available to bring
the molecule into the cryogenic cell. In the case of CaH,laser ablation of a solid precursor mounted in the cell is used.In order to obtain a truly isolated sample of cold molecules,the buffer gas needs to be rapidly pumped out. This hasalready been demonstrated to work for atoms @25#.
In this paper the third technique, in which time-varying
electricfieldsareusedtodecelerateandtrappolarmolecules,is described in detail. Static electric fields have been used todeflect and focus polar molecules since the 1920s.Already inthe 1950s it was realized that time-varying electric fields canbe used to change the longitudinal velocity of polar mol-
ecules @26–30 #. Molecules having their dipole-moment ori-
ented antiparallel to an electric field will gain Stark energyupon entering this field. The gain in Stark energy ~‘‘poten-
tial’’ energy !is compensated by a loss in kinetic energy. If
the electric field is greatly reduced before the molecule hasleft the electric field the molecule will not fully regain thelost kinetic energy. This process may be repeated by lettingthe molecules pass through multiple pulsed electric fields.Molecules can thus be slowed down and eventually broughtto a standstill. In Ref. @31#it was first experimentally dem-
onstrated that the longitudinal velocity of polar moleculescan be changed using time-varying electric fields.Abeam ofmetastable CO was decelerated from 225 m/s to 98 m/s usingan array of 63 pulsed electric fields. In Ref. @32#time-
varying electric fields were used to manipulate the velocityof Cs atoms in a fountain via the dipole polarizability. In Ref.@33#a theoretical model was presented to describe the mo-
tion of molecules in an array of time-varying electric fields.In particular, it was shown that the phase-space density isconstant during the deceleration process. This is nontrivialsince this is not generally true when a Hamiltonian is explic-
itly dependent on time @34#.I nR e f @35#
14ND3molecules
were decelerated and loaded in an electrostatic quadrupole
trap at a density of 106molecules/cm3and a temperature
below 350 mK. In Ref. @36#a prototype electrostatic storage
ring for neutral molecules has been demonstrated; bunchesof deuterated ammonia molecules with a forward velocity ofaround 100 m/s at a translational temperature of 10 mK weretrapped in a hexapole torus. Here we report trapping of
14ND3and15ND3molecules in a quadrupole trap at a den-
sity on the order of 107molecules/cm3. The temperature is
measured to be 25 mK. The phase-space density in the trap
corresponds to 2 310213, about 50 times less than the phase-
space density of the initial beam.
This paper is organized as follows. In Sec. II a model
describing the longitudinal and transverse motion in the de-celerator is outlined. From this model, the distribution ofmolecules in phase-space that is transmitted by the decelera-tor is derived. In Sec. III matching of the phase-space distri-bution throughout the apparatus is described. Experiments on
the deceleration of a beam of
14ND3molecules are presentedPHYSICAL REVIEW A, VOLUME 65, 053416
1050-2947/2002/65 ~5!/053416 ~20!/$20.00 ©2002 The American Physical Society 65053416-1 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0001 |
in Sec. IV, supporting the model. In Sec. V simultaneous
deceleration of different molecules is discussed and experi-
ments on simultaneous deceleration of the14NH3,14ND3,
and15ND3ammonia isotopomers are presented. The proper-
ties of the electrostatic trap and the method used to load thetrap, are discussed in Sec. VI. In Sec. VII experimental re-sults on the loading of the decelerated beam of ammoniamolecules into the trap are presented. The actual trapping ofthe molecules, including the experimental study of oscilla-tions in the trap, is described in Sec. VIII. Conclusions andfuture prospects are given in Sec. IX.
II. MOTION IN THE STARK DECELERATOR
A. Longitudinal motion
There is a strong similarity between deceleration of polar
molecules in a Stark decelerator and the acceleration ofcharged particles in a linear accelerator. In a charged particleaccelerator an electric field acts on the charge of a particle,whereas in a neutral particle accelerator an electric field gra-dient acts on the dipole moment of a particle. With a suitabletranslation all concepts that are used throughout for chargedparticle accelerators can be used for neutral particle accelera-tors as well. In particular, the principle of ‘‘phase-stability,’’i.e., the principle that is used to keep a bunch of particlestogether throughout the acceleration process, can be ex-ploited. Phase stability was independently discovered byVeksler @37#and McMillan @38#and forms the basis of all
modern charged particle accelerators. In this section it willbe shown how phase stability can be applied to the decelera-tion of polar molecules.
Consider an array of electric field stages separated by a
distanceL, as shown in Fig. 1. Each stage consists of two
parallel cylindrical metal rods with radius r, centered a dis-
tance 2r1dapart. One of the rods is connected to a positive
and the other to a negative switchable high-voltage powersupply.Alternating stages are connected to each other. Whena molecule in a quantum state with a positive Stark effect ~a
so-called low-field seeker !moves through the array of elec-
tric field stages as indicated in Fig. 1, it will gain Starkenergy. This gain in potential energy is compensated by aloss in kinetic energy. If the electric field is abruptlyswitched off, the molecule will keep its instantaneous veloc-ity. If, simultaneously, the electric field of the next stage isswitched on, the process will repeat itself. In Fig. 1 the po-
tential energy of the molecule, W(z), is depicted as a func-
tion of its position zalong the beam axis. The energy a mol-
ecule loses per stage depends on its position at the time thatthe fields are being switched. In analogy with concepts usedin charged particle accelerators, this position is expressed in
terms of a ‘‘phase-angle’’
fthat has a periodicity of 2 L.
Molecules that are in maximum electric field just prior to thetime at which the fields are being switched are assigned a
phase angle of
f590°.
First the situation where the electric fields are switched at
equal time intervals DT, is discussed. Let us consider a mol-
ecule at a phase f50° and with a velocity that matches the
frequency of the electric fields, i.e., a molecule that travels
exactly the distance Lin a time interval DT. This molecule
will be referred to as the ‘‘synchronous’’molecule. Its phase
and velocity are indicated as the equilibrium phase f0and
the equilibrium velocity v0, respectively. It is readily seen
that~i!the phase and velocity of the synchronous molecule
remains unchanged, and ~ii!that molecules with a slightly
different phase or velocity will experience an automatic cor-rection towards these equilibrium values. A molecule with a
phase slightly larger than
f0and a velocity equal to v0, for
instance, will lose more energy per stage than the synchro-nous molecule. It will thus be slowed down relative to thesynchronous molecule and consequently its phase will getsmaller, until it lags behind. At this point, the situation isreversed and it will lose less energy than the synchronousmolecule will, etc. This argument shows that molecules with
a slightly different phase from
f0and/or a slightly different
velocity from v0will oscillate with both phase and velocity
around the equilibrium values; the molecules are trapped in apotential well travelling at the velocity of the synchronousmolecule.
In order to decelerate the molecules one has to lower the
velocity of the potential well, by gradually increasing the
time intervals DTafter which the electric fields are being
switched. The synchronous molecule will still travel a dis-
tanceLin the interval DT, but
f0will now be different from
zero. By definition, the synchronous molecule is always at
the same position when the fields are being switched ( f0
remains constant !; it will achieve this by losing exactly the
required kinetic energy per stage. Again, the phase and ve-locity of a nonsynchronous molecule will oscillate aroundthose of the decelerated synchronous molecule.
The kinetic energy lost by the synchronous molecule per
stage DK(
f0), is given by W(f0)2W(f01p). It is conve-
nient to express W(f) as a Fourier series. In the expression
forDK(f0) all the even terms cancel, yielding
DK~f0!52a1sin~f0!12a3sin~3f0!1. ~1!
When adjacent electric field stages are not too far apart,
i.e.,L;2r1d,DK(f0) is predominantly determined by the
first term. As mentioned above, the phase is only defined atthe moment at which the fields are being switched. To beable to mathematically describe the motion of the moleculesthrough the Stark decelerator, a description in terms of con-tinuous variables is needed. For a description of the motion
FIG. 1. Scheme of the Stark decelerator, together with the Stark
energy of a molecule as a function of position zalong the molecular
beam axis.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416
053416-2 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0002 |
of a nonsynchronous molecule relative to the motion of the
synchronous molecule, the instantaneous difference in phase,
Df5f2f0, and velocity, Dv5v2v0, is introduced. One
can regard the lost kinetic energy per stage of the synchro-nous molecule to originate from a continuously acting aver-
age force F¯(f0)52DK(f0)/L. This approximation can be
made provided that the deceleration rate, i.e., the amount bywhich the velocity of the synchronous molecule is reduced in
a given stage relative to
v0, is small. When Dv!v0, the
average force on a nonsynchronous molecule can be written
asF¯(f01Df).2DK(f01Df)/L. The equation describ-
ing the motion of the nonsynchronous molecule relative tothe motion of the synchronous molecule is thus given by
mL
pd2Df
dt212a1
L@sin~f01Df!2sin~f0!#50, ~2!
withmthe mass of the molecule. This is analogous to the
equation for a pendulum driven by a constant torque. When
Dfis small, sin( f01Df).sinf01Dfcosf0and the equa-
tion of motion becomes
mL
pd2Df
dt212a1
L~Dfcosf0!50. ~3!
In Fig. 2 a numerical integration of Eq. ~2!is shown for
various equilibrium phases f0, with parameters as used in
the experiment on14ND3, to be described later. The solid
curves are lines of constant energy, and indicate trajectoriesthat molecules will follow. The inner curves are nearly ellip-tical, corresponding to a nearly linear restoring force on the
molecules, as can be seen from Eq. ~3!. For
14ND3the lon-
gitudinal oscillation frequency, given by
vz/2p5Aa1
2pmL2cosf0, ~4!
is approximately 1 kHz for f0570°. Further outward, the
restoring force is less than linear and the oscillation fre-quency is lowered, approaching zero on the separatrix ~thick
curve !. Beyond the separatrix the total energy of the mol-
ecules will be larger than the effective potential well and thetrajectories of the molecules are unbound.The area inside theseparatrix ~‘‘bucket’’ !is the longitudinal acceptance of the
decelerator. It is seen that this area is larger for smaller val-
ues of
f0. The deceleration per stage is given by Eq. ~1!and
increases when f0is increased from 0 to1
2p. Since both a
large acceptance and efficient deceleration are desirable thereis a trade-off between the two.
The phase-stability diagrams showing the longitudinal ac-
ceptance of the Stark decelerator are very similar to thoseused to describe charged particle accelerators ~see, for in-
stance, Fig. 2.20 of Ref. @39#!. There is an important differ-
ence, however. In a charged particle accelerator energy isadded to the particles at a certain position, while the amount
of energy that is added depends on the timeat which they
arrive at this position. In the Stark decelerator energy isadded to the molecules at a certain time, while the amount ofenergy that is added depends on their positionat that time.
Therefore, while in charged particle accelerators energy andtime are conjugate variables, in the Stark decelerator this roleis played by velocity ~energy divided by velocity !and posi-
tion~time multiplied by velocity !. As a consequence, while
theenergyspread in the laboratory frame remains constant in
a charged particle accelerator, the velocityspread in the labo-
ratory frame remains constant in the Stark decelerator.
Phase stability ensures that the phase-space density re-
mains constant during the deceleration process. In addition,it makes the deceleration insensitive to small deviations fromthe designed electric field in the decelerator, caused by, forinstance, misalignments of the electrodes or fluctuations inthe applied voltages. Small deviations from the anticipated
electric field will make the phase
f0, assumed to be constant
throughout the decelerator, differ per stage. This will lead toa slightly modified acceptance area, with some blurring inthe region of the separatrix. The final velocity, however, isdetermined by the time sequence only and will not bealtered.
In Ref. @33#this model was tested using metastable CO
molecules.Inthisexperimentthemoleculesweredeceleratedto 200 m/s, such that the deceleration rate and the relative
velocity spread ( D
vz/vz) was kept low. The model was
found to accurately describe the deceleration process. In thisarticle data are presented where a beam of ammonia mol-ecules is decelerated to much lower velocities, where theseapproximations no longer hold. In addition, these experi-
FIG. 2. Phase-stability diagrams for various values of f0, ob-
tained via numerical integration of Eq. ~2!with parameters as used
in the experiment for14ND3. In the experiment, a difference in the
phase angle of 2 pcorresponds to a distance of 11 mm.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416
053416-3 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0003 |
ments are sensitive to the effects of the transverse motion in
the decelerator.
B. Transverse motion
For the array of electrodes schematically depicted in Fig.
1 the electric field close to the electrodes is higher than thaton the molecular beam axis. Therefore, molecules in low-field seeking states will experience a force focusing themtowards the molecular beam axis. This focusing occurs onlyin the plane perpendicular to the electrodes, i.e., in the planeof the figure. In order to focus the molecules in both trans-verse directions, the electrode pairs that make up one decel-eration stage are alternately positioned horizontally and ver-tically.
In Fig. 3 the electric field in the decelerator is shown for
two mutually perpendicular planes. Both planes that areshown have the molecular beam axis in common. The elec-tric fields are calculated using a finite-element program @40#,
with parameters as in the experiment. The two opposing rodsthat make up one deceleration stage have a radius rof 1.5
mm and are spaced a distance dof 2 mm apart. Two adjacent
deceleration stages are spaced a distance Lof 5.5 mm apart.
With a maximum voltage of 110 kV and 210 kV on the
electrodes the maximum electric field on the molecular beamaxis is 90 kV/cm. Close to the electrodes at high voltage theelectric field is 120 kV/cm. The electric fields shown in Fig.3 are calculated for the situation with the vertical electrodes~front part !at ground potential and with the horizontal elec-
trodes ~partly hidden !at high voltage. Therefore, molecules
in low-field-seeking states are focused in the plane perpen-dicular to the electrodes at high voltage, as shown in theupper part of the figure. In the other direction the electricfield is ~nearly !constant and therefore there will be no fo-
cusing or defocusing in this direction. During the decelera-tion process the molecules pass through electric field stagesthat are alternately positioned horizontally and vertically, andare therefore focused in either direction.
As can be seen from Fig. 3, the force experienced by a
molecule depends on its position zin the decelerator. In ad-
dition, it depends on the time sequence of the high-voltagepulses that are applied. The molecule is strongly focusedwhen in between the electrodes at high voltage but much lessso further away from these electrodes. Therefore, the trajec-tories will generally be complicated. When the focusing isnot too strong the zdependence of this force can be elimi-
nated by introducing an averaged force. The transverse mo-tion throughout the decelerator can then be described usingthis averaged force, which will now depend on the phase
angle
fonly. The motion of the molecules in the transverse
potential well is characterized by a transverse angular oscil-
lation frequency vt(f). The condition that the focusing is
not too strong can be formulated as 2 p/vt@2L/vz, i.e., that
the time of transverse oscillation is much larger than the timeit takes for the molecules to traverse one period ~two stages !
of the decelerator. Molecules with a longitudinal velocityclose to the velocity of the synchronous molecules will
traverse a distance Lduring DT, the time interval after which
the voltages are being switched. The force needs to be aver-
aged over 2 LF
¯t~f!5E
fL/p(f12p)L/pFt~z!
2Ldz. ~5!
Close to the molecular beam axis, the average force is
linearly dependent on the displacement from this axis. Thesolid curve in Fig. 4 shows the resulting force constant for
14ND3(1 aN/m .0.05 cm21/mm2). The dashed curve
shows the maximum depth of the transverse potential well,i.e., the depth 1 mm away from the molecular beam axis.Since the linear dependence of the force on the displacementholds rather well throughout, the curves are almost identical.
For
f570° the transverse well is ;0.014 cm21,o r
;20 mK, deep. Molecules with a maximum transverse ve-
locity of 4 m/s will, therefore, still be accepted in the decel-eration process. The frequency of transverse oscillation,
vt/2p, is approximately 800 Hz.
Generally a molecule will oscillate between a minimum
and a maximum phase and it will, therefore, experience a
FIG. 3. The electric field in two mutually perpendicular planes
that have the molecular beam axis in common, together with aschematic view of two adjacent electric field stages. The two verti-cal rods are at ground potential, while the horizontal rods are at110 kV and 210 kV.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416
053416-4 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0004 |
different ~averaged !transverse force along its path. The cou-
pling between the longitudinal motion and the transversemotion may result in parametric amplification of the trans-verse oscillation. This effect might be expected to be strong,as the frequencies involved are rather similar. On the otherhand, as the deceleration process only lasts 2–3 oscillationperiods, parametric amplification is not expected to be im-portant, as confirmed by Monte Carlo simulations.The maxi-mum transverse velocity that will be accepted is determinedby the minimum depth of the transverse potential well thatthe molecules pass through. This implies that the maximumtransverse acceptance is obtained for molecules with a phasearound that of the synchronous molecule.
Since a maximum of the electric field strength cannot be
produced in free space @41#, molecules in high-field-seeking
states cannot be simultaneously focused in both transversedirections. It is possible, however, to produce a field that hasa maximum in one direction and a minimum in the otherdirection. By alternating the orientation of these fields it ispossible to obtain net focusing ~‘‘alternate gradient focus-
ing’’!for molecules in high-field-seeking states in either di-
rection @28#. When combined with the use of time varying
electric fields, an alternate gradient decelerator can be con-structed @42#.
III. PHASE-SPACE MATCHING
As shown in the preceding sections the area in phase
space remains constant throughout the deceleration process.This in itself is not enough. When the phase-space distribu-tion of the molecular beam ~‘‘emittance’’ !does not match the
phase-space acceptance of the decelerator ~where acceptance
is used to designate the shapeof the curves of constant en-
ergy rather than the maximum curve of constant energy !the
area in phase-space sampled by the molecules will be muchlarger. ~see for instance Fig. 3 in Ref. @33#!. Although the
area remains constant in this case, the original distribution isin practice irretrievable, and the effective area in phase spaceis increased, i.e., the effective ~time-averaged !phase-space
density is decreased. In order to avoid this, one needs toshape the emittance properly, a process known as ‘‘phase-space matching’’ @34#. The setup required to decelerate andtrap a molecular beam therefore consists of essentially five
parts; ~i!the molecular beam source, ~ii!a device to match
the beam to the input of the decelerator, ~iii!the decelerator,
~iv!a device to match the beam exiting the decelerator to the
trap, and, ~v!the trap itself. In this way a 6D area in phase
space is imaged from the source region of the molecularbeam onto the trap.
IV. DECELERATION EXPERIMENTS
The experimental setup, schematically depicted in Fig. 5,
consists of a compact molecular beam machine, with twodifferentially pumped vacuum chambers. The source cham-ber and decelerator chamber are pumped by a 300 l/s and a240 l/s turbo pump, respectively.Apulsed beam of ammoniais formed by expanding a mixture of less than 1% ammoniain xenon through a modified solenoid valve with a 0.8 mmdiameter opening into vacuum. The solenoid valve ~General
Valves series 9 !is modified such that it can be operated
down to liquid nitrogen temperatures. In these experimentsthe valve housing is cooled to 200 K, such that the vaporpressure of ammonia remains sufficiently high. The stagna-tion pressure is typically about 1.5 atm. The valve opens for
a duration of 80
ms. The experiments run at 10 Hz, and
under operating conditions the pressure in the source cham-
ber and the decelerator chamber is typically 4 31026and 2
31028Torr, respectively. The peak intensity of the beam is
limited by the pumping capacity; at higher pressures in thesource chamber the beam collapses.
The various ammonia isotopomers have different possible
symmetries for the nuclear spin wave functions of the iden-tical H/D nuclei. For each rovibrational level of the ammoniamolecule a combination with a nuclear spin wave function ofone specific symmetry is allowed. In the expansion, the mol-ecules are adiabatically cooled, and only the lowest rota-tional levels in the vibrational and electronic ground state arepopulated in the beam. In this cooling process the symmetryof the nuclear spin wave function is preserved.Therefore, theratio of the populations in the levels having different sym-metries of the nuclear spin wave function is the same in the
molecular beam as in the original sample. For NH
3and ND 3
this implies that roughly 60% of the molecules in the beamreside in the
uJ,K&5u1,1&level, the ground-state level for
molecules having a nuclear spin wave function of Esymme-
try. For historical reasons, levels of Esymmetry in NH 3and
FIG. 4. The average force constant ~close to the molecular beam
axis!for14ND3as a function of f~solid curve !together with the
maximum depth of the transverse potential well ~dashed curve !, i.e.,
the depth 1 mm away from the molecular beam axis.
FIG. 5. Schematic view of the experimental setup. A beam of
ammonia molecules is decelerated in a 35-cm-long decelerator andloaded into an electrostatic trap. In order to detect the ammoniamolecules a pulsed laser is focused inside the trap. The resultingions are extracted and counted using an ion detector.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416
053416-5 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0005 |
ND3are denoted as para levels.The only other possible sym-
metry of the nuclear spin wave function in NH 3isA1sym-
metry, and levels having this symmetry are denoted as ortho
levels. For ND 3bothA1andA2symmetry is possible; levels
having either one of this symmetry are denoted as ortholevels.
In zero electric field, the
uJ,K&5u1,1&level of NH 3and
ND3is split due to tunneling in a symmetric and in an anti-
symmetric component. Each of these components is furthersplit due to hyperfine interactions. The hyperfine splitting is
less than 2 MHz for
14ND3and less than 0.5 MHz for15ND3
@43#, and will be neglected in the following. Upon applying
an electric field, the symmetric and the antisymmetric com-ponents interact and repel each other, leading to four levels,as shown in Fig. 6. Throughout this work molecules in the
uJ,MK&5u1,21&state are used, which is the only low-field-
seeking state populated in the molecular beam; roughly 20%of the ammonia molecules in the beam are in this state. Atlow electric fields the Stark shift of this state is quadratic.When the Stark interaction becomes large relative to thezero-field splitting, the Stark shift is linear.
It is essential that the molecules remain in the low-field-
seeking state throughout the decelerator. If the moleculeswould come in zero electric field, they might project onto the
uJ,MK&5u1,0&state, and they would be lost for the decel-
eration process. These so-called Majorana transitions areavoided by assuring that the electric field never drops belowa certain minimum value @44#. Furthermore, the rapid
switching of the electric fields generates rf radiation that caninduce transitions between these states in nonzero fields. Toavoid these transitions, the minimum electric field needs tobe kept sufficiently high that the Stark splitting is larger thanthat in the highest-frequency component of the rf radiation.Molecules moving through the decelerator never experiencean electric field below 300V/cm, and no sign of ~loss due to !
these transitions has been observed. It should be noted, how-ever, that the rapid switching of electric fields is expected toscramble the various hyperfine levels of the low-field-seeking component.
The average velocity of the molecular beam is around 285
m/s (E
kin568 cm21). This is slightly higher than expected@45#, indicating that the translational velocity of the ammonia
molecules is not fully equilibrated with the carrier gas. Thevelocity spread of the beam @full width at half maximum
~FWHM !#is 20%, corresponding to a translational tempera-
ture of 1.6 K. The average velocity and the translational tem-
perature are averages over the 80
ms duration of the gas
pulse. The beam passes through a 1.0-mm-diameter skimmer
into a second vacuum chamber and then flies into a 5-cm-long pulsed hexapole that acts as a positive lens for mol-ecules in low-field-seeking states. After exiting the hexapolethe molecules enter the 35-cm-long decelerator. The decel-erator consists of an array of 64 deceleration stages. Eachdeceleration stage is formed by two parallel 3-mm-diametercylindrical rods, spaced 2 mm apart. The rods are made fromhardened steel and are highly polished. The two opposingrods are simultaneously switched by two independent high-
voltage switches to a maximum of 110 kV and 210 kV.
All horizontal and vertical stages are electronically con-nected requiring a total of four independent high-voltageswitches ~Behlke Electronic GmbH, HTS-151-03-GSM !.
The switches are triggered by a programmable delay genera-tor running at a clock frequency of 100 MHz. The time se-quences that are used are generated by the same program thatis used for the Monte Carlo simulations ~vide infra !.
It is advantageous to operate the decelerator at the highest
possible electric fields, since this will increase the acceptanceof the decelerator. The maximum obtainable electric fieldstrength is limited by field electron emission at the electrodesurfaces. It is assumed that this will take place at field
strengths over ;10
4kV/cm @46#, in principle allowing a volt-
age difference of 2000 kV over the 2 mm gap between theelectrodes. In practice, however, electrical breakdown willtake place at much lower voltage differences due to fieldenhancement at microfeatures associated with the intrinsicmicroscopic roughness of the electrode surfaces. Breakdownevents need not be disastrous as long as the energy that isdissipated in the gap is kept low. In fact, since dischargeswill locally melt the surface, these events will reduce theroughness of the surface, allowing the gap to withstand ahigher voltage difference after the event. Since in the decel-erator the electric fields are switched rapidly ~within 200 ns !,
the resistance between the switches and the electrodes needs
to be small (1 k V). Under these conditions it is likely that a
discharge will permanently damage the surface, instead ofimproving it. Therefore, before operating the deceleratorwith time varying electric fields, a constant electric field isused for high-voltage conditioning. The voltage difference isslowly increased while the current of the induced dischargesis being limited. This procedure allows for an increase of themaximum attainable voltage difference between the elec-trodes without sparking. Typically, the voltage difference is
increased by 5 kV/h while the current is limited to 1
mA.
This procedure is repeated every time the decelerator hasbeen exposed to air. After this treatment, the decelerator isoperated at a voltage difference 25% below the maximumvoltage difference sustainable during the high-voltage condi-tioning. In this way the decelerator could be operated rou-
tinely at 110 kV and 210 kV, corresponding to a maxi-
mum electric field of 120 kV/cm. It is believed that at
FIG. 6. Stark shift of the uJ,K&5u1,1&level of14NH3and
14ND3ammonia molecules in an electric field of up to 150 kV/cm.
On this scale the Stark shifts of14ND3and15ND3~not shown !are
identical.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416
053416-6 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0006 |
present the maximum electric field is limited by currents
running over the surface of the aluminum-oxide rings thatare used for suspension of the decelerator. By improving themounting of the decelerator, it is expected that the electric
fields can be increased to 200 kV/cm or more.
In all experiments presented here, the decelerator is oper-
ated at voltages of 110 kV and 210 kV, leading to a
maximum electric field on the molecular beam axis of 90
kV/cm. In this field, the
uJ,KM&5u1,21&state is shifted by
1.1 cm21relative to zero electric field for14ND3and
15ND3, and by 0.8 cm21for14NH3~see Fig. 6 !. In most of
the experiments, a phase angle f0570° is used. With 64
deceleration stages14ND3and15ND3are then slowed down
to 15 m/s, while14NH3is decelerated to ;100 m/s, when
starting from 255 m/s.With these settings, the 6D acceptance
area of the decelerator is approximately @1.5 mm 38 m/s #
3@2m m 38 m/s #2. The longitudinal emittance of the
beam, typically @25 mm 360 m/s #, is much larger than the
longitudinal acceptance of the decelerator, and there is noth-ing to be gained by changing the shape of the emittance inthis direction. The transverse emittance of the beam, typi-
cally
@1m m 320 m/s #2, is better matched to the transverse
acceptance of the decelerator when it is expanded spatiallywhile the velocity distribution is compressed. This isachieved by using a short hexapole to make a 1:2 image ofthe beam exiting the skimmer onto the entrance of the decel-erator.
The hexapole consists of 3-mm-diameter rods placed
equidistantly on the outside of a circle wit ha3m m radius.
The rods of the hexapole are alternatingly at ground potentialand at 10 kV, creating an electric field that is cylindricallysymmetric close to the molecular beam axis. Generally, the
field in an npole is proportional to r
(n/2)21@47#, yielding a
quadratic dependence of the electric field on rin our case.
Molecules with a linear positive Stark effect will thus expe-rience a linear force towards the molecular beam axis. In freeflight from the skimmer to the hexapole the beam spreads outin the transverse spatial direction, while the transverse veloc-ity remains unchanged.This creates an elongated distributionin the corresponding phase space, tilted with respect to thespatial coordinate. When switching on the field of the hexa-pole, this distribution will start to rotate in phase space.When the hexapole is switched off at the appropriate time,the beam will refocus in free flight from the hexapole to thedecelerator, forming an image of the beam ~at the skimmer !
onto ~the entrance of !the decelerator. By placing the hexa-
pole at the right position, or, alternatively, by using a longhexapole that is pulsed at the right time, a 1:2 image of thebeam can be formed. Due to the zero-field splitting, mol-ecules close to the molecular beam axis will be less wellfocused, and the image will not be perfect @48#. Almost a
factor of 4 increase in phase-space density is expected at theexit of the decelerator compared to the situation where thehexapole is not used. A clear increase has indeed been ob-served. In most of the data presented here the hexapole couldnot be used, however, due to problems with electrical dis-
charges. In Fig. 7 the density of
14ND3molecules 24.5 mm
behind the decelerator, i.e., at the center of the trap, is shownas a function of time after the start of the time sequence.Using a pulsed tunable UV laser,
14ND3molecules in the
uJ,K&5u1,1&upper component of the inversion doublet are
selectively ionized in a (2 11) resonance-enhanced multi-
photon ionization ~REMPI !scheme. The laser radiation,
about 10 mJ of energy around 317 nm in a 5-ns-durationpulse, is focused inside the trap using a lens with a focallength of 75 cm. Mass-selective detection of the parent ionsis performed using the trap electrodes as extraction elec-trodes in a Wiley-McLaren-type mass spectrometer setup.
For this, a voltage of 2300 V is applied to the exit endcap.
The field-free flight tube is kept at 21.2 kV and the ion
signal is recorded using a microchannel plate detector placedfurther downstream. The ion signal is proportional to thedensity of neutral ammonia molecules at the center of thetrap, and amounts to approximately 125 ion counts per laserpulse for the decelerated molecular beam.The lower curve inFig. 7 shows the time of flight ~TOF!profile of the original
beam. The beam has an average velocity of 285 m/s and avelocity spread of 60 m/s. The upper curve ~bold!shows the
TOF profile when the decelerator is used to slow down themolecules from 271.5 m/s to 91.8 m/s.At the entrance of thedecelerator the beam has an extension along zof around 25
mm~FWHM !. This is more than twice as long as the peri-
odicity 2Lof the decelerator, and thus more than one bucket
will be filled ~see Fig. 2 !.Three peaks are clearly observed in
the TOF profile around 2.5 ms. The central peak originatesfrom molecules that were at the right position near the en-trance of the decelerator at the start of the time sequence, and
FIG. 7. TOF profiles for14ND3molecules recorded 24.5 mm
behind the decelerator as a function of time after the start of thetime sequence. Using (2 11)-REMPI with a pulsed laser around
317 nm,
14ND3molecules in the uJ,K&5u1,1&upper inversion level
are selectively detected. The lower curve ~bold!shows the original
beam when the decelerator is off. The upper curve ~bold!shows the
time-of-flight ~TOF!profile when the beam is decelerated from
271.5 m/s to 91.8 m/s. Bunches of slow molecules exit the decel-erator at later times, arriving in the detection region around 2.5 msafter the start of the time sequence.The gray curves are the result ofa Monte Carlo simulation of the experiment. The signal intensitytypically increases with a factor of 14 due to transverse focusingwhen the decelerator is switched on; the signal of the slow mol-ecules is larger than that of the original beam.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416
053416-7 | 6a29bdbbc33cbdac22141ee1826fcd55640e89b8 | page_0007 |