| ==== Front | |
| Cancer Med | |
| Cancer Med | |
| 10.1002/(ISSN)2045-7634 | |
| CAM4 | |
| Cancer Medicine | |
| 2045-7634 | |
| John Wiley and Sons Inc. Hoboken | |
| 36394080 | |
| 10.1002/cam4.5409 | |
| CAM45409 | |
| CAM4-2022-08-3480.R1 | |
| Research Article | |
| RESEARCH ARTICLES | |
| Clinical Cancer Research | |
| Immune checkpoint gene VSIR predicts patient prognosis in acute myeloid leukemia and myelodysplastic syndromes | |
| Yao et al. | |
| Yao Kevin https://orcid.org/0000-0001-6369-655X | |
| 1 | |
| Zhou Emily https://orcid.org/0000-0002-9194-7042 | |
| 2 | |
| Schaafsma Evelien https://orcid.org/0000-0002-3821-3568 | |
| 3 4 | |
| Zhang Baoyi 5 | |
| Cheng Chao https://orcid.org/0000-0002-5002-3417 | |
| 6 7 8 chao.cheng@bcm.edu | |
| 1 Department of Electrical and Computer Engineering Texas A&M University College Station Texas USA | |
| 2 Department of Biosciences Rice University Houston Texas USA | |
| 3 Department of Molecular and Systems Biology Dartmouth College Lebanon New Hampshire USA | |
| 4 Department of Biomedical Data Science The Geisel School of Medicine at Dartmouth College Lebanon New Hampshire USA | |
| 5 Department of Chemical and Biomolecular Engineering Rice University Houston Texas USA | |
| 6 Department of Medicine Baylor College of Medicine Houston Texas USA | |
| 7 Dan L Duncan Comprehensive Cancer Center, Baylor College of Medicine Houston Texas USA | |
| 8 Institute for Clinical and Transcriptional Research, Baylor College of Medicine Houston Texas USA | |
| * Correspondence | |
| Chao Cheng, Department of Medicine, Baylor College of Medicine, Houston, TX 77030, USA. | |
| Email: chao.cheng@bcm.edu | |
| 16 11 2022 | |
| 3 2023 | |
| 12 5 10.1002/cam4.v12.5 55905602 | |
| 08 10 2022 | |
| 11 8 2022 | |
| 24 10 2022 | |
| © 2022 The Authors. Cancer Medicine published by John Wiley & Sons Ltd. | |
| https://creativecommons.org/licenses/by/4.0/ This is an open access article under the terms of the http://creativecommons.org/licenses/by/4.0/ License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. | |
| Abstract | |
| Background | |
| Immune checkpoint proteins play critical functions during the immune response to cancer and have been targeted by immune checkpoint blockade therapy. V‐domain Ig suppressor of T cell activation (VSIR) is one of these immune checkpoint genes and has been investigated extensively in recent years due to its conflicting roles in cancer immunity. Specifically, in acute myeloid leukemia (AML), the prognostic value of VSIR is debated. | |
| Results | |
| In both patient tumor samples and cancer cell lines we find that VSIR has the highest expression in AML out of all cancer types and, in AML, has the highest expression out of all other immune checkpoint genes. Survival analysis indicated that AML patients with higher VSIR expression have significantly shorter survival than those patients with lower expression, even within established AML subgroups (e.g., FAB subtypes). Importantly, VSIR expression is predictive of progression from myelodysplastic syndromes (MDS) patients into AML, suggesting its potential role during the very early stage of AML development and progression. In addition to AML, VSIR also demonstrates prognostic values in other cancer types, including multiple myeloma and mesothelioma. | |
| Conclusion | |
| In summary, our analyses revealed the prognostic value of VSIR and its potential as a target for immunotherapy, especially in AML. | |
| The Kaplan‐Meier plot shows that patients with higher VSIR expression have a significantly worse prognosis. In addition, the forest plot showing that VSIR expression remains significant in predicting prognosis after adjusting for molecular and genetic variables like FAB score and cytogenetic risk. | |
| AML | |
| MDS | |
| prognosis | |
| VSIR | |
| Cancer Prevention and Research Institute of Texas 10.13039/100004917 RR180061 source-schema-version-number2.0 | |
| cover-dateMarch 2023 | |
| details-of-publishers-convertorConverter:WILEY_ML3GV2_TO_JATSPMC version:6.2.6 mode:remove_FC converted:21.03.2023 | |
| Yao K , Zhou E , Schaafsma E , Zhang B , Cheng C . Immune checkpoint gene VSIR predicts patient prognosis in acute myeloid leukemia and myelodysplastic syndromes. Cancer Med. 2023;12 :5590‐5602. doi: 10.1002/cam4.5409 | |
| Kevin Yao and Emily Zhou contributed equally to this work. | |
| ==== Body | |
| pmc1 INTRODUCTION | |
| Acute myeloid leukemia (AML) is a blood cancer of myeloid cells and is projected to have about 20,500 new cases and 11,540 deaths in 2022. 1 AML is characterized by the presence of abnormal or poorly differentiated, proliferative, and clonal myeloid cells, leading to higher concentrations of myeloblasts in bone marrow, blood, and other tissues. These blasts consume body resources and prevent the genesis of normal healthy cells. AML often progresses quickly, and the 5‐year survival for AML patients is as low as about 30%. 1 AML is more likely to develop in patients with underlying myelodysplastic syndromes (MDS), a heterogeneous class of diseases that result in ineffective hematopoiesis. In fact, about a third of patients with MDS will develop AML. 2 The distinguishing criteria for MDS or AML diagnosis are the percentage of blasts in the bone marrow and peripheral blood, which has been arbitrarily set as 30% by the French‐American‐British (FAB) classification system 3 and 20% by the World Health Organization (WHO). 4 | |
| The main treatment for AML has traditionally been chemotherapy and targeted therapy drugs. 5 In recent years, cancer treatment has been greatly advanced by the development of immunotherapy, including immune checkpoint inhibitors, adoptive cell therapies, and monoclonal antibodies. 6 Such an advancement is largely attributed to the discovery of immune checkpoint targets, especially PD‐L1 and CTLA‐4. Therapy via blockade of these targets has found success for solid cancers such as melanoma 7 and non‐small lung cancer. 8 For AML, there have also been pilot studies on the use of PD‐L1 checkpoint blockade. Ravandi et al. found that the median event‐free survival of patients treated with nivolumab in combination with idarubicin and cytarabine was not reached in the study, but a considerable number of patients experienced immune‐related adverse events. 9 In addition, Daver et al. found that the overall response rate of patients treated with nivolumab in combination with Azacitidine was 33%, with 11% of patients experiencing grade 3–4 immune‐related adverse effects. 10 Such immunotherapies have yet to be widely recommended for AML. This may be due to adverse effects or because AML has a lower mutation burden and a more suppressed immune system, 11 , 12 which tends to decrease the efficacy of immune checkpoint blockade therapy. | |
| In addition to PD‐L1 and CTLA‐4, other immune checkpoint targets like V‐domain Ig suppressor of T cell activation (VSIR, also called VISTA, C10orf54, B7‐H5, PD‐H1) have also been discovered and investigated. 13 VSIR is an integral membrane protein with an extracellular immunoglobulin domain, a stalk, a transmembrane domain, and a cytoplasmic tail. 14 Under physiological conditions, VSIR is highly expressed in hematopoietic lineages and tissues rich in infiltrating leukocytes and has lower expression in non‐hematopoietic tissue. In hematopoietic lineages, VSIR is expressed CD14+ monocytes, neutrophils, myeloid CD11c+ DCs, and CD4+ and CD8+ T cells. VSIR is not expressed on CD19+ B cells or CD56Hi NK cells. In mice, VSIR is highly expressed in tumor‐infiltrating leukocytes. Unlike PD‐1/PD‐L1, VSIR expression is restricted to cells from hematopoietic lineages. 15 VSIR not only serves as an inhibitor of T‐cell response, it also serves as a regulator for signaling and activation of innate immune cells. For example, in cancer, autoimmune, and inflammatory diseases, VSIR inhibits the production of inflammatory cytokines and chemokines in myeloid dendritic cells and macrophages. 15 It is mainly expressed in hematopoietic cells in humans. VSIR is encoded by a gene located at 10q22.1, which is entirely confined inside an intron of the CDH23 gene. 16 VSIR has attracted additional research interest due to its conflicting roles as both an inhibitory and a stimulatory immune checkpoint protein. 17 In AML, VSIR has been found to be highly expressed. 18 Knockout studies in mice found that VSIR induced immune evasion and caused the observed proliferation of AML cells. 18 This supports the candidacy of VSIR as a novel checkpoint target in AML treatment. Wang et al. found a positive correlation between the expression of VSIR in peripheral myeloid cells and the expression of PD‐1 in T cells, even though there is no evidence that these genes are directly regulated. 19 Since high PD‐1 expression has been implicated in a worse prognosis for AML patients, 20 we expect that VSIR expression might also be associated with patient prognosis. However, previous studies have reported inconsistent results on the prognostic association of VSIR expression in AML. 19 , 21 , 22 For instance, Wang et al. 19 found that VSIR expression did not correlate with prognosis, but Zhang et al. and Chen et al. 21 , 22 found that higher VSIR expression is related to poorer prognosis. In addition, there have not been reports on the prognostic value of VSIR in MDS, the pre‐AML disease. | |
| In this study, we investigated the prognostic value of VSIR gene expression in human AML and MDS. Using the TCGA (The Cancer Genome Atlas) and CCLE (Cancer Cell Line Encyclopedia) datasets, we first investigate the expression levels of VSIR in AML compared to other cancers and other immune checkpoint targets. We then establish the prognostic value of VSIR in AML based on survival analysis and show its potential for improving patient stratification in conjunction with established clinical factors, such as FAB subtype and cytogenetic risk. We also investigate the expression and prognostic value of VSIR in patients with myelodysplastic syndromes (MDS). Lastly, we perform a comprehensive analysis to examine the prognostic association of VSIR across seven different blood cancers and 33 other cancer types. | |
| 2 MATERIALS & METHODS | |
| 2.1 Datasets used in this study | |
| The Cancer Genome Atlas (TCGA) datasets for the various cancer types studied in this study were downloaded from Firehose (https://gdac.broadinstitute.org/). Specifically, for LAML, the dataset contains RSEM‐normalized gene expression data for 20,501 genes and 173 patient samples. Genomic mutation and copy number variation (CNV) data were downloaded from Firehose. 23 The genomic mutation data were downloaded in mutation annotation format (MAF) files and contain mutation information for 24,058 genes and 197 samples. Patient mutation burden is calculated as the sum of all non‐synonymous mutations. The CNV data were downloaded as a segmented copy number variation (sCNA) file and contain the CNV information for 23,311 genes and 380 patient samples. CNV burden is calculated as the fraction of genomic regions with abnormal chromosome copy numbers. 24 | |
| Other than the transcriptomic and genomic data from TCGA, the study has investigated the following gene expression datasets. The cell line data for 33 cancers were downloaded from the Cancer Cell Line Encyclopedia (CCLE) database (https://portals.broadinstitute.org/). The de novo AML dataset contains 526 total samples and was downloaded from the Gene Expression Omnibus (GEO) under the accession ID GSE14468. The MDS dataset contains 159 MDS and 17 wild‐type samples and was downloaded from GEO under the accession ID GSE58831. The processed single‐cell RNA‐seq dataset was downloaded from GEO under accession ID GSE116256. It contained scRNA data for 38,410 cells from 40 bone marrow biopsies. Those biopsies were obtained from 16 patients with AML and five healthy donors. In addition, we have analyzed a collection of cancer gene expression datasets from the PREdiction of Clinical Outcomes from Genomic (PRECOG) (http://precog.stanford.edu) database, including a total of 166 datasets with matched gene expression profiles and patient survival information for seven different cancer types. More detailed information for these datasets can be found in Table S1. | |
| 2.2 Survival analysis | |
| Overall survival (OS) event status and time were provided in the TCGA dataset. Using this data, we constructed univariate and multivariate Cox regression models using the “coxph” function with default parameters from the R library “survival.” The univariate Cox regression models were used to determine the association between overall survival and VSIR expression. The multivariate regression models were used to determine the association between OS and the following covariates: VSIR expression, FAB score, and cytogenetic risk. The results of the multivariate Cox regression models were visualized in forest plots, which was constructed using “forest_model” function from the R package “forestmodel.” Since VSIR expression is a continuous variable, we used the median VSIR expression to divide the samples into two equal‐sized groups (High and Low VSIR expression). The Kaplan–Meier method was used to plot the survival curves, and the log‐rank test was used to determine the difference between the two curves and calculate their p‐value. Survival analysis was performed using the R library “survival” with otherwise default parameters. The Kaplan Meier method was done using the function survfit from R library “survival” with default parameters, and the function “ggsurvplot” from the R library “ggplot” was used to plot the survival curves. | |
| 2.3 Comparing VSIR expression in subgroups | |
| To test if VSIR expression is different between three or more subgroups, such as the eight FAB subtypes or three cytogenetic risk categories, we performed one‐way ANOVA using the R function “aov” with default parameters. To determine if VSIR expression is higher in one subgroup compared to the others, such as if VSIR expression is higher in NPM1 mut patients compared to NPM1 WT patients, we performed the one‐sided Wilcoxon test using the “wilcox.test” function in R. | |
| 2.4 scRNA‐seq analysis of VSIR expression in healthy cells and AML cells | |
| Single‐cell RNA seq data of healthy cells were analyzed using data downloaded from the Bloodspot database, which provides mRNA expression profiles for a comprehensive list of hematopoietic cells. 25 We selected the HemaExp v1 dataset and filtered the genes to select the VSIR gene. Bloodspot then generated a diagram for the expression of VSIR depending on the cell type. | |
| For analysis of VSIR expression in AML patients, we utilized the GSE116256 dataset, which contains scRNA‐seq data on 38,410 cells from 40 bone marrow biopsies from 16 AML patients and five healthy patients. We chose to exclude the five healthy patients from this analysis, which left us with 30,712 cells. The cells were categorized as one of 14 cell types and determined to be cancerous or healthy. Of all the cell types, the only cancerous cells detected in this dataset were granulocyte monocyte progenitors (GMP), hematopoietic stem cells (HSC), progenitor cells (Prog), promonocytes (ProMono), classical dendritic cells (cDC), and monocytes (mono). | |
| 2.5 VSIR expression in other blood cancers | |
| The PRECOG dataset was used to determine the expression of VSIR in other blood cancers. This dataset contains curated data from 166 datasets, including gene expression data and survival data for around 18,000 patients. Specifically, PRECOG contained gene expression data on the following hematopoietic cancers: two datasets for chronic lymphocytic leukemia (CLL), two datasets for multiple myeloma, one dataset for Burkitt lymphoma, seven datasets for AML, two datasets for B‐cell acute lymphoblastic leukemia (B‐ALL), three datasets for diffuse large B‐cell lymphoma (DLBCL), and one dataset for follicular lymphoma. To consolidate the survival analysis results from all datasets within a cancer type, after calculating the p‐values using the log‐rank test for each dataset individually, we used Fisher's method to calculate the meta p‐value for each cancer type. | |
| 3 RESULTS | |
| 3.1 VSIR is the immune checkpoint gene with the highest expression and association with prognosis in AML | |
| To investigate the implication of immune checkpoint genes in leukemia, we compared their expression in AML patient samples from TCGA and leukemia cell lines from CCLE. The TCGA data reflect the mixed expression of genes in both tumor cells and non‐tumor cells in the tumor microenvironment, while CCLE data provide a proxy of gene expression in tumor cells only. As shown in Figure 1A, VSIR has the highest expression out of all the immuno‐checkpoint genes in both TCGA and CCLE datasets for AML. This indicates that VSIR expression is prominent in both cancer cells (CCLE) and the cancer microenvironment (TCGA), including immune cells that have infiltrated into the bone marrow. Furthermore, we examined the association of these immune checkpoint genes with the prognosis of patients with AML using the TCGA data. VSIR was determined to be the most prognostic immune checkpoint gene in AML (Figure 1B). | |
| FIGURE 1 VSIR has highest expression and is prognostic in AML. (A) VSIR has the highest expression out of all cancer types in both the TCGA and CCLE datasets. (B) VSIR has the highest expression out of other known immune checkpoint genes in AML samples for both the TCGA and CCLE datasets. (C) Within TCGA AML samples, VSIR expression has the highest statistical significance and hazard ratio when predicting survival compared to other immune checkpoint genes. (D) Kaplan–Meier plot showing that TCGA AML patients with higher‐than‐median VSIR expression have significantly worser prognosis. (E) Forest plot validating the prognostic value of VSIR expression after considering the contributions of other molecular and genetic variables such as cytogenetic risk and FAB subtype in a multivariate Cox proportional hazards model. | |
| Previous studies have suggested that VSIR is highly expressed in hematopoietic tissues. 14 To confirm this, we examined its expression pattern in different cancer types. Indeed, VISA shows the highest expression in AML out of all other cancer types in both TCGA and CCLE datasets (Figure 1C). We then differentiated TCGA AML samples into High and Low VSIR expression using the median as the threshold. The survival curves of the two groups are shown in Figure 1D. As shown, the VSIR‐high expression group has significantly shorter overall survival than the VSIR‐low expression group (p = 2 e‐3), with a median survival of 273 days compared with 471 days, respectively. Multivariate regression analysis indicated that VSIR expression remains significant after considering well‐known clinical factors, including the FAB category and cytogenetic risk group (Figure 1E), suggesting that it provides additional prognostic value to these established clinical features. | |
| 3.2 Validation of the prognostic value of VSIR in an independent dataset | |
| We further validate the prognostic association of VSIR in an independent AML dataset (GSE14468) with a larger number of samples (n = 526). Similar to previous results, after differentiating patients into two groups (High and Low VSIR expression) using the median as the threshold, patients with high VSIR expression exhibited significantly worse prognosis (p = 0.004) (Figure 2A), consistent with our previous observation. To account for possible confounding effects between VSIR and clinical variables, we performed a multivariate Cox regression analysis and found that high VSIR expression remained a significant prognostic indicator when clinical variables like FAB score and cytogenetic risk were considered (Figure 2B). | |
| FIGURE 2 Validation of VSIR prognostic value in GSE14468. (A) Kaplan–Meier plot shows that patients with higher VSIR expression have a significantly worse prognosis. (B) Forest plot showing that VSIR expression remains significant in predicting prognosis after adjusting for molecular and genetic variables like FAB score and cytogenetic risk. (C) VSIR expression is different depending on the FAB subtype (ANOVA p‐value = 3 e‐39). (D–F) AML patients with the FAB M1, M2, and M4 subtypes, respectively, have significantly poorer prognoses if they have high VSIR expression. (G) VSIR expression is different depending on cytogenetic risk (ANOVA p‐value = 4 e‐17). (H) AML patients with an intermediate cytogenetic risk have a significantly poorer prognosis if they have high VSIR expression. | |
| Biomarkers for cancer are often only applicable within specific molecular or clinical subsets of patients. For example, the Oncotype DX assay is only applicable to ER‐positive breast cancer patients. 26 Considering this, we aimed to determine the prognostic value of VSIR within certain subsets of AML patients. One such classification system of AML is the French‐American‐British (FAB) system, which distinguishes roughly eight categories of AML based on cell type and maturity: M0 (myeloblastic without differentiation), M1 (myeloblastic with minimal maturation), M2 (myeloblastic with maturation), M3 (promyelocytic), M4 (myelomonocytic), M5 (monocytic), M6 (erythroid), and M7 (megakaryoblastic). VSIR expression varied considerably depending on the FAB subtype (ANOVA p‐value = 3 e‐39), suggesting that VSIR may play different roles within different subtypes. We thus systematically evaluated the prognostic value of VSIR within each subtype by utilizing Kaplan–Meier analysis on patients dichotomized by the median VSIR expression within that subtype. As demonstrated in Figure 2D–F, high VSIR expression in M1 (p = 0.05), M2 (p = 0.001), and M4 (p = 0.01) are associated with poor prognosis. In the other FAB subtypes, the prognostic association was not significant. | |
| The next clinical subtype we considered was a cytogenetic risk (Figure 2G), which is a classification based on whether a patient's cytogenetics confer relatively favorable, intermediate, or poor prognosis. For example, patients with the t(8;21) translocation have a favorable prognosis, whereas the del(7q) abnormality confers poorer prognosis. 27 Intermediate cytogenetic risk, which is exhibited in the majority of AML patients, 28 appeared to have the highest VSIR expression, and higher VSIR expression is associated with significantly poorer prognosis (p = 0.01) (Figure 2H). Thus, we have demonstrated that high VSIR expression is associated with poor survival in both an external AML dataset and clinical variable categories commonly associated with AML. | |
| 3.3 VSIR is overexpressed in MDS patients and predicts overall and AML‐free survival | |
| MDS has been previously referred to as a preleukemia disorder, and about 30% of MDS patients eventually develop AML. 29 We, therefore, investigated the expression of VSIR in MDS samples. MDS patients exhibited significantly higher VSIR expression (p = 0.009) compared to their wild‐type or normal counterparts, which supports results from the previous work. 30 Additionally, VSIR expression remains prognostic in predicting overall survival in both a univariate (p = 0.008) and multivariate cox proportional hazards model (p = 0.02) when considering established clinical variables like age, gender, and cytogenetic risk (Table 1). Because MDS has a high frequency of transition to AML, we next investigated if VSIR is able to predict AML‐free survival, where events are defined as a patient with MDS progressing to AML. We find that higher VSIR expression puts patients at higher risk for progression to AML in a univariate model (p = 0.05) and after adjusting for possible confounding effects of age, gender, and cytogenetic risk (p = 0.02, Table 1). These results support the theory that VSIR acts as an immune checkpoint inhibitor, which may allow proliferating blasts to escape immune detection and progress to AML. | |
| TABLE 1 VSIR is prognostic for MDS patients | |
| PMID # Samples/Cancer types Tissue type | |
| 25,079,552 173 AML Hematopoietic Tissue | |
| 22,460,905 33 Cell Line | |
| 26,193,342 7 Various | |
| 25,574,665 176 Bone Marrow | |
| 30,827,681 40 Bone Marrow | |
| 19,171,880 526 Bone Marrow | |
| Note: VSIR expression significantly predicts overall survival and AML‐free survival for MDS patients. This is validated in both a univariate cox proportional hazards model and in a multivariate cox proportional hazards model considering clinical variables like age, gender, and cytogenetic risk. | |
| 3.4 Association of VSIR with the genomic landscape and biological pathways of AML patients | |
| To further investigate whether the expression of VSIR is affected by upstream genomic aberrations, we determined the correlation between the patients' VSIR expression and their mutation burden or their CNV burden in the TCGA dataset. Using Spearman's correlation test, we found that VSIR expression is significantly associated with mutation burden (p = 0.006, ρ = −0.218) but not CNV burden (p = 0.5, ρ = −0.0545). To explore specific mutations that may be associated with VSIR expression, we chose to only study NRAS, DNMT3A, IDH1, TET2, NPM1, WT1, FLT3, IDH2, TP53, CEBPA, and RUNX1 mutations, which are all present in greater than 5% of AML samples in the TCGA dataset. We then tested whether VSIR expression is different between patients with the mutation and those without the mutation. The p‐values were calculated by the Wilcoxon test and were adjusted using the Bonferroni multiple testing correction to limit the false discovery rate. We plotted that against the VSIR fold change between mutated and wild‐type samples (Figure 3A). NPM1 was the only mutation that remained significant, where patients with the mutation had higher VSIR expression (Figure 3B, p = 3 e‐5). This trend was also validated in the GSE14468 dataset (Figure 3C, p = 1 e‐23). | |
| FIGURE 3 VSIR association with the genomic landscape. (A) Volcano plot showing that NPM1 mutation is the only common genomic mutation in AML that has significantly different VSIR expressions between the group of patients with the mutation versus those without it. (B, C) Patients with the NPM1 mutation have significantly higher VSIR expression versus those without the mutation in both the TCGA (Wilcoxon p‐value = 1 e‐4) and GSE14468 (Wilcoxon p‐value = 1 e‐23) datasets. (D, E) Kaplan–Meier plots showing that patients are significantly dichotomized by their VSIR expression if they are NPM1 WT, but not if they have the NPM1 mutation, in both the TCGA and GSE14468 datasets. | |
| A previous study has reported that NPM1 mutation was associated with a favorable prognosis, 31 , 32 but another study observed no significant prognostic association. 33 Having shown that VSIR expression correlates with NPM1 mutation status, we further investigated if the prognostic value of VSIR can be explained by the prognostic value of the NPM1 mutation. We did not observe a significant survival difference between NPM1 mutant and wild‐type samples in both TCGA and GSE14468 (Figure S1). In addition, we investigated the prognostic association of VSIR expression within the NPM1 mutation and wild‐type subgroups. In the TCGA dataset, VSIR expression was negatively associated with patient prognosis (p = 0.01, HR = 1.837) in the NPM1 wild‐type subgroup, but not in the mutant subgroup (p = 0.8, HR = 1.119) (Figure 3D). In the GSE14468 dataset, similar results were observed (Figure 3E). As shown, higher VSIR expression was a poor prognostic factor for patients in the NPM1 wild‐type subgroup (p = 0.004, HR = 1.487), but high VSIR levels were not significantly associated with prognosis for NPM1 mutant patients (p = 0.2, HR = 1.470). These results seem to indicate that VSIR expression may only be prognostic for patients without the NPM1 mutation, but validation in a larger dataset is necessary to draw any definite conclusions. | |
| Although FLT3‐ITD mutations were not significantly associated with VSIR expression, since they occur in roughly 20%–25% of AML cases, 34 it is of importance to study the prognostic value of VSIR within populations with and without the FLT3‐ITD mutation. We found that VSIR expression significantly stratified patients without the mutation (p = 0.05, Figure S2A), but was unable to stratify patients with the FLT3‐ITD mutation (p = 0.4, Figure S2A), presumably due to the relatively small sample size (N = 63 for each group). | |
| Additionally, we note that NPM1 mutant AML is a clinically heterogeneous subtype due to its frequent coexistence with other mutations. Specifically, the FLT3‐ITD mutation is a about twice as frequent in patients with NPM1 mutations. 35 , 36 Thus, we next investigated the effect of FLT3‐ITD mutations on VSIR expression and prognostic value within NPM1 mutant patients. We found that patients with the FLT3‐ITD mutation in tandem with the NPM1 mutation had lower expression of VSIR (p = 0.05, Figure S2B). We then explored whether VSIR was prognostic within these subgroups and found that VSIR was not significantly prognostic within either (NPM1 mutants with FLT3‐ITD mutations p = 0.53, NPM1 mutants without FLT3‐ITD mutations p = 0.45, Figure S2C,D). The heterogeneity of NPM1 mutants may be confounded by other factors outside FLT3 mutations, such as stem cell signatures, 32 so further research in a larger dataset is needed to confirm the prognostic value of VSIR within further subtypes of NPM1 mutant samples. | |
| 3.5 The expression pattern of VSIR in different cell types | |
| To understand the potential mechanisms underlying the prognostic value of VSIR at the cellular level, we examined its expression pattern using scRNA‐seq data. First, we mapped the landscape of single‐cell VSIR expression for healthy patients using the HemaExplorer dataset based on their lineage in hematopoiesis (Figure 4A). As expected, cells of the myeloid lineage tend to have higher VSIR expression, especially monocytes and polymorphonuclear cells (PMN) from the bone marrow (BM) or peripheral blood (PB). | |
| FIGURE 4 VSIR expression in single cells associated with leukemia. (A) VSIR expression in various hematopoietic cells of different cell lineages. Myeloid cells appear to have the highest expression of VSIR. (B) Single‐cell RNA sequencing data show that monocytes have the highest nonzero percent expression of VSIR. (C) Boxplots showing the distribution of VSIR expression for cDC, GMP, HSC, monocytes, progenitor cells, and pro‐monocyte cells based on whether the cell is healthy or cancerous. Malignant cells have consistently higher VSIR expression than normal cells. (D) VSIR expression correlation with monocyte marker genes such as ITGAM and CCR2. Any point above the horizontal red line has a significant nonzero Pearson correlation coefficient (p‐value <0.05). | |
| We next aimed to compare this landscape of normal VSIR expression to the VSIR expression levels of patients with AML by utilizing the GSE116256 dataset, which contains scRNA‐seq data on 30,712 cells from bone marrow biopsies from AML patients. We first aimed to identify the percentage of cells which express VSIR in these AML patients (Figure 4B). Cancerous cells generally have a higher nonzero expression of VSIR, and monocytes have the highest expression of VSIR for both normal and cancerous cells. We further explored the distribution of VSIR expression for normal vs cancerous cells with nonzero VSIR expression (Figure 4C) and found that monocytes are the only cell type where the VSIR expression in cancerous cells is significantly higher than that in normal cells. Based on this evidence, VSIR expression in monocytes is the most prominent and appears to be most affected when a person develops AML. Hypothesizing that VSIR may be co‐expressed on monocytes, we correlated the expression of VSIR with monocyte marker genes (Figure 4D) and found that every monocyte marker gene besides CCR7 is significantly correlated with VSIR expression. This further suggests that monocytes may be the major cell type that VSIR affects in AML patients. | |
| 3.6 VSIR expression is prognostic in other cancer types | |
| Finally, we explored the association of VSIR expression with prognosis in other blood cancers and cancer types. Initially, we examined other types of blood cancers. As shown in Figure 5A, VSIR expression is significantly prognostic in chronic lymphocytic leukemia (CLL), AML, and multiple myeloma. We further divided the samples into two groups by using the median VSIR expression and constructed a survival curve. We revealed that low VSIR expression is significantly associated with poor prognosis in multiple myeloma (p = 2 e‐6) (Figure 5B). Similarly, low VSIR expression is significantly associated with poor prognosis in CLL (p = 8 e‐4) (Figure 5C). After examining other types of blood cancers, we expanded our scope to include all the cancer types available on TCGA. After performing a univariate Cox regression model using overall survival and VSIR expression in each cancer type, it was revealed that VSIR expression has the highest statistical significance and hazard ratio in AML (Figure 5D). As shown in Figures 5E,F similarly to previous findings, low VSIR expression is associated with poor prognosis in mesothelioma (MESO) (p = 2 e‐4) and cervical squamous cell carcinoma (CESC) (p = 0.03). These results are consistent with the previous studies. 17 , 37 , 38 | |
| FIGURE 5 Prognostic value of VSIR in other blood cancer types. (A) Barplot showing the meta‐p values when predicting survival using VSIR in multiple myeloma, AML, CLL, and other blood cancers using the PRECOG dataset. (B, C) Kaplan–Meier plot showing that patients with higher VSIR expression have significantly better prognosis for both multiple myeloma and CLL, respectively. (D) Volcano plot showing that VSIR expression has high statistical significance when predicting survival in AML compared to other cancer types. (E, F) Kaplan–Meier plot showing that patients with higher VSIR expression have significantly better prognosis for both mesothelioma and cervical squamous cell carcinoma, respectively. | |
| 4 DISCUSSION | |
| In this study, we demonstrated that VSIR is prognostic and has the highest expression out of all other immune checkpoint genes in AML. We validated this claim in the TCGA and GSE14468 datasets and accounted for possible confounding clinical variables like FAB category and cytogenetic risk. Favorable cytogenetic risk predicted better survival, which aligns with current guidelines. VSIR score was also found to be prognostic within the FAB M1, M2, and M4 categories, as well as patients with intermediate cytogenetic risk. We also found that VSIR is significant in predicting both overall survival and AML‐free survival for MDS patients. Specifically, MDS patients with higher VSIR expression have a higher risk of regressing into AML. Next, we considered genomic driver events that correlate with VSIR expression and found that NPM1 is the only common genomic mutation in AML that correlates significantly with VSIR expression. We then investigated the expression of VSIR in individual cells and found that monocytes have the highest expression of VSIR for both healthy cells and AML cells. Lastly, we determined that VSIR is also prognostic in other blood cancers like multiple myeloma and CLL and other cancers like mesothelioma and cervical squamous cell carcinoma. | |
| Our study supports the theory that VSIR acts as an inhibitory immune checkpoint in AML. 17 Higher VSIR expression suppresses immune responses at the cancer site and allows immune escape, leading to poorer prognosis. This should further support the ongoing efforts in determining the efficacy of VSIR blockade therapy in AML patients. 39 Interestingly, we observe the same trends in MDS patients, suggesting that VSIR also suppresses immune response for MDS patients, which would cause poorer prognosis and a higher risk of progressing to AML. These results can have implications in MDS treatment. Currently, immunotherapy is not recommended for MDS treatment, but our results suggest that VSIR blockade therapy may yield benefits. Clinical trials of administering immunotherapy to MDS patients should be evaluated more carefully. | |
| Currently, immune checkpoint inhibitors have modest efficacy in AML and MDS. The ORR rate of PD1 inhibitor combined with azacitidine in RR‐AML is about 30%, and the CR rate is about 20%. 40 Recently, a few clinical trials have been performed that evaluated the efficiency of anti‐VSIR‐based immunotherapy in multiple cancer types as monotherapy or combination therapies. Several VSIR inhibitors have undergone phase I and II clinical trials in solid tumors. 41 , 42 , 43 The VSIR inhibitors include CA‐170 (NCT02812875), CI‐8993 (NCT04475523), and W0180 (NCT04564417). CA‐170, an anti‐VSIR and anti‐PD‐L1 inhibitor, has shown very promising results. For example, in a phase II study for nonsquamous non‐small cell lung cancer, it was demonstrated that CA‐170 had a clinical benefit rate of 75% and patients had 19.5 weeks of progression‐free survival. 41 However, none of these trials were tested for treating AML or MDS. Our analysis in this study indicates that VSIR has the highest expression levels in AML compared to other cancer types; and in AML VSIR has the highest expression than the other immune checkpoint genes. These results suggest that VSIR might be a good target for delivering effective immune checkpoint blockade therapy. | |
| Importantly, for all cell types, VSIR expression increases when the cell develops AML, with monocytes having the highest VSIR expression with or without cancer. VSIR may directly bind to T cell receptors as a ligand, suppressing its activation. 44 , 45 Another possible pathway is that VSIR binds to and causes conformational changes in galectin‐9 that serve to enhance its effect. 46 Galectin‐9 has been reported to limit the activity of natural killer cells and induce apoptosis of T cells, thus limiting their ability to respond against cancer. 47 , 48 Our results suggest that the mechanism behind why higher VSIR expression leads to poorer prognosis may be that cancerous cells produce more VSIR protein, which through binding to T cells and galectin‐9, inhibits immune response. Monocytes appear to be the biggest player in the expression of VSIR, which may be understood since monocytes were the only cells (out of naive T cells, CD4 central memory T cells, CD4 effector memory T cells, CD8 EM T cells, NK cells, B cells, and basophils) that experienced significant transcriptional signature changes when VSIR was bound by anti‐VSIR antibodies. 49 Functionally, the transcriptional changes serve to enhance monocyte activity, such as increased secretion of IFNγ in a mixed lymphocyte reaction. | |
| Lastly, we note that higher VSIR expression leads to a better prognosis in multiple myeloma, CLL, MESO, and CESC. This is contrasted by the observation that higher VSIR expression leads to poorer prognosis in AML. The survival advantage of higher VSIR expression in multiple myeloma, MESO, and CESC agrees with the previous work, 17 , 37 , 38 but the prognostic value of VSIR in CLL has not been studied, to our knowledge. Both CLL and multiple myeloma are lymphoid cancers, which suggests that VSIR may play opposing roles depending on the hematopoietic lineage of cancer. Overall, this supports the consensus that VSIR plays a multifaceted role in cancer immunity across different cancers and may have co‐inhibitory or stimulatory roles in the immune response. 17 | |
| Our study is currently limited by the lack of large datasets with VSIR expression data for AML patients. Our study may also be further improved through the incorporation of VSIR proteomics data, since gene expression may have a low correlation with protein expression. 50 , 51 A direct correlation between VSIR protein expression and poorer prognosis may further support the idea that VSIR should be targeted in checkpoint blockade therapy. | |
| In conclusion, we have reported that VSIR is a poor prognostic factor for overall survival in AML and MDS. These associations hold after accounting for possible confounding clinical variables. VSIR is also prognostic within specific subgroups of patients. Importantly, MDS patients with higher VSIR expression are also more likely to develop AML. Not only do these results further support efforts in administering VSIR blockade therapy for AML patients, we also suggest that VSIR blockade therapy may be a promising treatment for MDS patients, which is a relatively unexplored prospect. | |
| AUTHOR CONTRIBUTIONS | |
| Kevin Yao: Data curation (equal); formal analysis (equal); methodology (equal); writing – original draft (equal). Emily Zhou: Formal analysis (equal); visualization (equal); writing – review and editing (equal). Evelien Schaafsma: Writing – review and editing (supporting). Baoyi Zhang: Writing – review and editing (supporting). Chao Cheng: Conceptualization (lead); data curation (lead); formal analysis (supporting); funding acquisition (lead); methodology (lead); project administration (lead); supervision (lead); writing – review and editing (lead). | |
| FUNDING INFORMATION | |
| This work is supported by the Cancer Prevention Research Institute of Texas (CPRIT) (RR180061 to CC). CC is a CPRIT Scholar in Cancer Research. | |
| CONFLICT OF INTEREST | |
| The authors have declared that no conflict of interest exists. | |
| Supporting information | |
| Figure S1 | |
| Click here for additional data file. | |
| Figure S2 | |
| Click here for additional data file. | |
| Table S1 | |
| Click here for additional data file. | |
| DATA AVAILABILITY STATEMENT | |
| All data generated or analyzed during this study are included in this published article and its supplementary information files. | |
| ==== Refs | |
| References | |
| 1 Siegel RL , Miller KD , Fuchs HE , Jemal A . Cancer statistics, 2022. CA Cancer J Clin. 2022;72 (1 ):7‐33.35020204 | |
| 2 DeVita VT , Lawrence TS , Rosenberg SA . DeVita, Hellman, and Rosenberg's Cancer: Principles & Practice of Oncology. Vol 2 . Lippincott Williams & Wilkins; 2008. | |
| 3 Varela B , Chuang C , Woll J , Bennett J . Modifications in the classification of primary myelodysplastic syndromes: the addition of a scoring system. Hematol Oncol. 1985;3 :55‐63.3857211 | |
| 4 Harris NL , Jaffe ES , Diebold J , et al. The World Health Organization classification of hematological malignancies report of the clinical advisory committee meeting, Airlie house, Virginia, November 1997. Mod Pathol. 2000;13 :193‐207.10697278 | |
| 5 Burnett A . The treatment of AML: current status and novel approaches. Hematology. 2005;10 :50‐53.16188635 | |
| 6 Kruger S , Ilmer M , Kobold S , et al. Advances in cancer immunotherapy 2019–latest trends. J Exp Clin Cancer Res. 2019;38 :1‐11.30606223 | |
| 7 Hodi FS , O'Day SJ , McDermott DF , et al. Improved survival with ipilimumab in patients with metastatic melanoma. N Engl J Med. 2010;363 :711‐723. doi:10.1056/NEJMoa1003466 20525992 | |
| 8 Brahmer J , Reckamp KL , Baas P , et al. Nivolumab versus docetaxel in advanced squamous‐cell non‐small‐cell lung cancer. N Engl J Med. 2015;373 :123‐135. doi:10.1056/NEJMoa1504627 26028407 | |
| 9 Ravandi F , Assi R , Daver N , et al. Idarubicin, cytarabine, and nivolumab in patients with newly diagnosed acute myeloid leukaemia or high‐risk myelodysplastic syndrome: a single‐arm, phase 2 study. Lancet Haematol. 2019;6 :e480‐e488. doi:10.1016/S2352-3026(19)30114-0 31400961 | |
| 10 Daver N , Garcia‐Manero G , Basu S , et al. Efficacy, safety, and biomarkers of response to Azacitidine and nivolumab in relapsed/refractory acute myeloid leukemia: a nonrandomized, open‐label, Phase II Study. Cancer Discov. 2019;9 :370‐383. doi:10.1158/2159-8290.CD-18-0774 30409776 | |
| 11 Lawrence MS , Stojanov P , Polak P , et al. Mutational heterogeneity in cancer and the search for new cancer‐associated genes. Nature. 2013;499 :214‐218. doi:10.1038/nature12213 23770567 | |
| 12 Han Y , Dong Y , Yang Q , et al. Acute myeloid leukemia cells express ICOS ligand to promote the expansion of regulatory T cells. Front Immunol. 2018;9 :2227. doi:10.3389/fimmu.2018.02227 30319662 | |
| 13 Qin S , Xu L , Yi M , Yu S , Wu K , Luo S . Novel immune checkpoint targets: moving beyond PD‐1 and CTLA‐4. Mol Cancer. 2019;18 :155. doi:10.1186/s12943-019-1091-2 31690319 | |
| 14 Flies DB , Wang S , Xu H , Chen L . Cutting edge: a monoclonal antibody specific for the programmed death‐1 homolog prevents graft‐versus‐host disease in mouse models. J Immunol. 2011;187 :1537‐1541. doi:10.4049/jimmunol.1100660 21768399 | |
| 15 Xu W , Hieu T , Malarkannan S , Wang L . The structure, expression, and multifaceted role of immune‐checkpoint protein VISTA as a critical regulator of anti‐tumor immunity, autoimmunity, and inflammation. Cell Mol Immunol. 2018;15 :438‐446. doi:10.1038/cmi.2017.148 29375120 | |
| 16 Nowak EC , Lines JL , Varn FS , et al. Immunoregulatory functions of VISTA. Immunol Rev. 2017;276 :66‐79. doi:10.1111/imr.12525 28258694 | |
| 17 Huang X , Zhang X , Li E , et al. VISTA: an immune regulatory protein checking tumor and immune cells in cancer immunotherapy. J Hematol Oncol. 2020;13 :83. doi:10.1186/s13045-020-00917-y 32600443 | |
| 18 Kim TK , Han X , Wang J , et al. PD‐1H (VISTA) induces immune evasion in acute myeloid leukemia. Blood. 2017;130 :2658. | |
| 19 Wang L , Jia B , Claxton DF , et al. VISTA is highly expressed on MDSCs and mediates an inhibition of T cell response in patients with AML. Onco Targets Ther. 2018;7 :e1469594. doi:10.1080/2162402X.2018.1469594 | |
| 20 Brodska B , Otevřelová P , Šálek C , et al. High PD‐L1 expression predicts for worse outcome of leukemia patients with concomitant NPM1 and FLT3 mutations. Int J Mol Sci. 2019;20 :2823. doi:10.3390/ijms20112823 31185600 | |
| 21 Zhang W , Zhang W , Gui L , et al. Expression and prognosis of the B7 family in acute myeloid leukemia. Ann Transl Med. 2021;9 :1530. doi:10.21037/atm-21-4255 34790736 | |
| 22 Chen CT , Wang PP , Mo WJ , et al. Expression profile analysis of prognostic long non‐coding RNA in adult acute myeloid leukemia by weighted gene co‐expression network analysis (WGCNA). J Cancer. 2019;10 :4707‐4718. doi:10.7150/jca.31234 31528236 | |
| 23 Weinstein JN , Collisson EA , Mills GB , et al. The cancer genome atlas pan‐cancer analysis project. Nat Genet. 2013;45 :1113‐1120.24071849 | |
| 24 Qin C , He X , Zhao Y , et al. Systematic computational identification of prognostic cytogenetic markers in neuroblastoma. BMC Med Genomics. 2019;12 :192. doi:10.1186/s12920-019-0620-6 31831008 | |
| 25 Bagger FO , Kinalis S , Rapin N . BloodSpot: a database of healthy and malignant haematopoiesis updated with purified and single cell mRNA sequencing profiles. Nucleic Acids Res. 2019;47 :D881‐D885. doi:10.1093/nar/gky1076 30395307 | |
| 26 Paik S , Shak S , Tang G , et al. A multigene assay to predict recurrence of tamoxifen‐treated, node‐negative breast cancer. N Engl J Med. 2004;351 :2817‐2826. doi:10.1056/NEJMoa041588 15591335 | |
| 27 Gupta M , Mahapatra M , Saxena R . Cytogenetics' impact on the prognosis of acute myeloid leukemia. J Lab Physicians. 2019;11 :133‐137. doi:10.4103/JLP.JLP_164_18 31160852 | |
| 28 Dohner K , Paschka P . Intermediate‐risk acute myeloid leukemia therapy: current and future. Hematology Am Soc Hematol Educ Program. 2014;2014 :34‐43. doi:10.1182/asheducation-2014.1.34 25696832 | |
| 29 Menssen AJ , Walter MJ . Genetics of progression from MDS to secondary leukemia. Blood. 2020;136 :50‐60. doi:10.1182/blood.2019000942 32430504 | |
| 30 Dufva O , Pölönen P , Brück O , et al. Immunogenomic landscape of hematological malignancies. Cancer Cell. 2020;38 :380‐399 e313. doi:10.1016/j.ccell.2020.06.002 32649887 | |
| 31 Rau R , Brown P . Nucleophosmin (NPM1) mutations in adult and childhood acute myeloid leukaemia: towards definition of a new leukaemia entity. Hematol Oncol. 2009;27 :171‐181. doi:10.1002/hon.904 19569254 | |
| 32 Mer AS , Heath EM , Madani Tonekaboni SA , et al. Biological and therapeutic implications of a unique subtype of NPM1 mutated AML. Nat Commun. 2021;12 :1‐13.33397941 | |
| 33 Sakaguchi M , Yamaguchi H , Najima Y , et al. Prognostic impact of low allelic ratio FLT3‐ITD and NPM1 mutation in acute myeloid leukemia. Blood Adv. 2018;2 :2744‐2754. doi:10.1182/bloodadvances.2018020305 30341082 | |
| 34 Blau O , Berenstein R , Sindram A , Blau IW . Molecular analysis of different FLT3‐ITD mutations in acute myeloid leukemia. Leuk Lymphoma. 2013;54 :145‐152. doi:10.3109/10428194.2012.704999 22721497 | |
| 35 Schnittger S , Schoch C , Kern W , et al. Nucleophosmin gene mutations are predictors of favorable prognosis in acute myelogenous leukemia with a normal karyotype. Blood. 2005;106 :3733‐3739. doi:10.1182/blood-2005-06-2248 16076867 | |
| 36 Suzuki T , Kiyoi H , Ozeki K , et al. Clinical characteristics and prognostic implications of NPM1 mutations in acute myeloid leukemia. Blood. 2005;106 :2854‐2861. doi:10.1182/blood-2005-04-1733 15994285 | |
| 37 Mutsaers P , Balcioglu HE , Kuiper R , et al. V‐domain Ig suppressor of T cell activation (VISTA) expression is an independent prognostic factor in multiple myeloma. Cancers (Basel). 2021;13 (9 ):2219. doi:10.3390/cancers13092219 34066382 | |
| 38 Alcala N , Mangiante L , le‐Stang N , et al. Redefining malignant pleural mesothelioma types as a continuum uncovers immune‐vascular interactions. EBioMedicine. 2019;48 :191‐202. doi:10.1016/j.ebiom.2019.09.003 31648983 | |
| 39 Ghosh A , Barba P , Perales MA . Checkpoint inhibitors in AML: are we there yet? Br J Haematol. 2020;188 :159‐167. doi:10.1111/bjh.16358 31808941 | |
| 40 Lee E , Koh Y , Hong J , Eom HS , Yoon SS . Recent clinical update of acute myeloid leukemia: focus on epigenetic therapies. J Korean Med Sci. 2021;36 :e85. doi:10.3346/jkms.2021.36.e85 33821592 | |
| 41 Radhakrishnan V , Banavali S , Gupta S , et al. Excellent CBR and prolonged PFS in non‐squamous NSCLC with oral CA‐170, an inhibitor of VISTA and PD‐L1. Ann Oncol. 2019;30 :v494. | |
| 42 DiMascio, L. et al. (Wolters Kluwer Health, 2021). | |
| 43 Tagliamento M , Agostinetto E , Borea R , et al. VISTA: a promising target for cancer immunotherapy? Immunotargets Ther. 2021;10 :185‐200. doi:10.2147/ITT.S260429 34189130 | |
| 44 Wang L , Rubinstein R , Lines JL , et al. VISTA, a novel mouse Ig superfamily ligand that negatively regulates T cell responses. J Exp Med. 2011;208 :577‐592. doi:10.1084/jem.20100619 21383057 | |
| 45 Flies DB , Han X , Higuchi T , et al. Coinhibitory receptor PD‐1H preferentially suppresses CD4(+) T cell‐mediated immunity. J Clin Invest. 2014;124 :1966‐1975. doi:10.1172/JCI74589 24743150 | |
| 46 Yasinska IM , Meyer NH , Schlichtner S , et al. Ligand‐receptor interactions of Galectin‐9 and VISTA suppress human T lymphocyte cytotoxic activity. Front Immunol. 2020;11 :580557. doi:10.3389/fimmu.2020.580557 33329552 | |
| 47 Golden‐Mason L , McMahan RH , Strong M , et al. Galectin‐9 functionally impairs natural killer cells in humans and mice. J Virol. 2013;87 :4835‐4845. doi:10.1128/JVI.01085-12 23408620 | |
| 48 Kang CW , Dutta A , Chang LY , et al. Apoptosis of tumor infiltrating effector TIM‐3+CD8+ T cells in colon cancer. Sci Rep. 2015;5 :15659. doi:10.1038/srep15659 26493689 | |
| 49 Rogers BM , Smith L , Dezso Z , et al. VISTA is an activating receptor in human monocytes. J Exp Med. 2021;218 :e20201601. doi:10.1084/jem.20201601 34106206 | |
| 50 de Sousa Abreu R , Penalva LO , Marcotte EM , Vogel C . Global signatures of protein and mRNA expression levels. Mol Biosyst. 2009;5 :1512‐1526. doi:10.1039/b908315d 20023718 | |
| 51 Vogel C , Marcotte EM . Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat Rev Genet. 2012;13 :227‐232.22411467 | |