Proteomic Analysis of Hepatic Fibrosis in Human Immunodeficiency Virus–Associated Nonalcoholic Fatty Liver Disease Demonstrates Up-regulation of Immune Response and Tissue Repair Pathways Free

Abstract

Background

Human immunodeficiency virus (HIV)–associated nonalcoholic fatty liver disease (NAFLD) is characterized by a high prevalence of hepatic fibrosis as a strong clinical predictor of all-cause and liver-specific mortality risk.

Methods

We leveraged data from an earlier clinical trial to define the circulating proteomic signature of hepatic fibrosis in HIV-associated NAFLD. A total of 183 plasma proteins within 2 high-multiplex panels were quantified at baseline and at 12 months (Olink Cardiovascular III; Immuno-Oncology).

Results

Twenty proteins were up-regulated at baseline among participants with fibrosis stages 2–3 versus 0–1. Proteins most differentially expressed included matrix metalloproteinase 2 (P < .001), insulin-like growth factor–binding protein 7 (P = .001), and collagen α1(I) chain (P = .001). Proteins were enriched within pathways including response to tumor necrosis factor and aminopeptidase activity. Key proteins correlated directly with visceral adiposity and glucose intolerance and inversely with CD4⁺ T-cell count. Within the placebo-treated arm, 11 proteins differentially increased among individuals with hepatic fibrosis progression over a 12-month period (P < .05).

Conclusions

Among individuals with HIV-associated NAFLD, hepatic fibrosis was associated with a distinct proteomic signature involving up-regulation of tissue repair and immune response pathways. These findings enhance our understanding of potential mechanisms and biomarkers of hepatic fibrosis in HIV.

Nonalcoholic fatty liver disease (NAFLD) has risen in prevalence to affect more than one-third of people living with human immunodeficiency virus (HIV; PLWH) [1]. While NAFLD encompasses a broad histologic spectrum, hepatic fibrosis has been shown to be a major predictor of all-cause and liver-specific mortality risk [2–4]. Importantly, NAFLD among PLWH is characterized by high rates of fibrosis presence and progression [5], suggesting a poor prognosis in this population. We have further found that visceral fat accumulation, to which PLWH are prone [6], is a key clinical correlate of fibrosis in HIV-associated NAFLD [5]. Nonetheless, biologic mechanisms underlying fibrosis initiation and progression in PLWH remain poorly understood. Moreover, there are no Food and Drug Administration–approved therapies to prevent or to reverse fibrosis progression among any group of patients. These gaps in knowledge underscore the critical need for dedicated studies to characterize the biologic alterations that define NAFLD-related fibrosis among PLWH.

High-throughput proteomic technologies in recent years have allowed for the unbiased identification of novel protein candidates that may mediate or signify pathologic changes in various disease contexts [7]. Notably, to our knowledge, such approaches have never before been used to characterize hepatic fibrosis in the context of HIV-associated NAFLD. In the current study, we leveraged data from an earlier clinical trial to define for the first time the proteomic signature of fibrosis presence and progression in NAFLD among a modern cohort of PLWH. This work uncovers key proteins and protein networks characteristic of hepatic fibrosis in HIV-associated NAFLD that may serve as potential biomarkers and therapeutic targets of disease.

METHODS

Study Design

The current analysis used data from a 12-month randomized trial of tesamorelin versus placebo among participants with HIV-associated NAFLD (NCT02196831) [8]. Previously, we leveraged proteomic data from this trial to examine effects of tesamorelin versus placebo on plasma proteins within a single high-multiplex panel (Olink Immuno-Oncology) [9, 10]. The current report describes novel work involving an expanded array of proteins (Olink Cardiovascular III; Immuno-Oncology) to develop a proteomic signature of hepatic fibrosis, explore relationships of key proteins with clinical parameters, and identify proteins that change in association with fibrosis progression in a longitudinal assessment of the natural history of HIV-associated NAFLD in those not receiving tesamorelin.

In brief, the study enrolled 61 men and women aged 18–70 years who had documented HIV infection and liver steatosis, as defined by hepatic fat fraction ≥5% on proton magnetic resonance spectroscopy (¹H-MRS). Participants were required to have been on stable antiretroviral therapy for ≥3 months, with a CD4⁺ T-cell count >100/µL and an HIV viral load <400 copies/mL. Exclusion criteria included excess alcohol use, active hepatitis B or C, cirrhosis, and hemoglobin A1c >7% [8]. Participants were enrolled at Massachusetts General Hospital in Boston or the National Institutes of Health in Bethesda, Maryland, between 20 August 2015 and 16 January 2019. Written informed consent was obtained from each participant, and the study was approved by the institutional review boards at both institutions.

Study Procedures

Study procedures for the parent clinical trial have been described in detail elsewhere [8]. At screening, participants underwent detailed history, physical examination, and laboratory investigations to determine eligibility. Hepatic ¹H-MRS was performed with patients in a fasting state to measure liver fat content. Abdominal magnetic resonance imaging was performed to quantify visceral adipose tissue (VAT) and subcutaneous adipose tissue cross-sectional area at the L4 vertebral level.

Individuals meeting eligibility criteria subsequently completed baseline assessments, which included ultrasound-guided percutaneous liver biopsy and blood sample collection. Liver specimens were reviewed by a single expert pathologist (D. E. K., National Institutes of Health ) who performed histologic scoring using the Nonalcoholic Steatohepatitis Clinical Research Network scoring system [11]. Hepatic fibrosis was staged from 0 to 4. A standard 75-g oral glucose tolerance test was administered with the plasma glucose area under the curve calculated to reflect total glucose excursion after glucose loading. Study procedures were repeated in all participants 12 months after randomization to tesamorelin versus placebo.

Plasma Proteomic Assessment

Using blood samples obtained at baseline and 12 months, we measured plasma levels of 183 proteins from 2 curated high-multiplex panels (Olink Cardiovascular III; Immuno-Oncology). Proteins in these panels reflected biologic processes including inflammation, wound healing, vascular and tissue remodeling, and apoptosis. Protein abundance levels were reported as normalized protein expression on a log₂ scale. Assay characteristics are available from the manufacturer (http://www.olink.com).

Statistical Analysis

In this analysis, our primary objective was to delineate a circulating proteomic signature associated with hepatic fibrosis among individuals with HIV-associated NAFLD. We also aimed to identify clinical parameters that were correlated with key plasma protein levels. At baseline, participants were categorized by the presence or absence of significant hepatic fibrosis, defined as hepatic fibrosis stage 2–3 versus 0–1 by histologic assessment. Demographic and clinical characteristics were compared between fibrosis groups, using Student t (continuous variables) or χ² (categorical variables) tests, with P < .05 as the threshold for statistical significance.

For our proteomic analysis, we first ascertained differences in plasma protein levels between fibrosis groups at baseline, using Student t test. Statistical significance was defined with the combined criteria of Benjamini-Hochberg false discovery rate q ≤ .10 and P ≤ .01. Relationships between plasma proteins and hepatic fibrosis were further expressed using odds ratios with 95% confidence intervals. When we identified a subset of plasma proteins that were differentially expressed between groups, we used principal component analysis to reduce the dimensionality of our data pertaining to these key proteins, with eigenvectors derived from their covariance matrix. We then assessed for differences in principal components by hepatic fibrosis status, with significance defined at P < .05. Principal components found to differ between groups were further tested in multivariable logistic regression models to ascertain whether the association of hepatic proteomic signature with more advanced fibrosis persisted on adjustment for known clinical correlates of disease, including visceral adiposity.

To further contextualize differentially expressed plasma proteins at baseline within biologic pathways, we used the GeneMANIA database to delineate interrelated networks to which these proteins corresponded. Specifically, starting with the list of differentially expressed proteins, GeneMANIA searched publicly available data sets to find related proteins, based on parameters including coexpression, protein-protein interaction, shared protein domains, and shared biologic pathways, and then generated a weighted network among these proteins. Functional enrichment analysis in GeneMANIA was used to identify Gene Ontology pathways that were enriched among this protein network, based on false discovery rate q ≤ .01. In addition, we used Pearson correlation coefficients to assess relationships between plasma levels of differentially expressed proteins and key clinical parameters. For proteins found to be associated with visceral fat content and significant hepatic fibrosis and known to be expressed in visceral adipose tissue, we performed mediation analysis, using the Baron and Kenny method [12] to determine the proportion of the relationship between visceral fat and hepatic fibrosis that was mediated by each protein.

Finally, we identified plasma proteins found to change in association with fibrosis progression over 1 year among placebo-treated individuals, using Student t test with significance defined at P < .05. Fibrosis progression was defined as any increase in fibrosis stage between baseline and 12 months. These analyses were performed for exploratory purposes to extend the findings of our baseline assessment.

RESULTS

Baseline Characteristics of Study Participants

Of 61 participants enrolled in our clinical trial, 58 individuals had baseline liver biopsy specimens and plasma proteomics available for analysis (Table 1). Of these individuals, 16 (28%) had evidence of significant hepatic fibrosis on biopsy defined as fibrosis stage 2 (n = 10) or stage 3 (n = 6). Conversely, 42 individuals lacked significant fibrosis, with stage 0 (n = 33) or stage 1 disease (n = 9). Demographic and HIV-related characteristics were similar across fibrosis groups. Statin use was higher among individuals with versus without more advanced hepatic fibrosis (63% vs 31%; P = .03). While body mass index was also comparable between groups, individuals with significant hepatic fibrosis demonstrated higher visceral fat content than those with absent or mild disease (mean [standard deviation (SD)], 284 [81] vs 227 [102] cm², respectively; P = .03). As expected, individuals with versus without significant fibrosis exhibited higher NAFLD activity score (mean [SD], 4.3 [2.0] vs 2.0 [0.9], respectively; P < .001), alanine aminotransferase (45 [31] vs 25 [15] U/L; P = .03), aspartate aminotransferase (50 [26] vs 26 [16] U/L; P = .003), and Fibrosis-4 index (2.16 [1.09] vs 1.19 [0.47]; P = .004). CD4⁺ T-cell counts tended to be lower among those with more severe hepatic disease, although this difference did not reach statistical significance (mean [SD], 670 [272] vs 798 [276] cells/µL, respectively; P = .12).

Plasma Proteins Differentially Expressed by Hepatic Fibrosis Status

Proteomic analysis of 183 proteins in the combined immuno-oncology and cardiovascular protein sets revealed differences in 20 plasma proteins according to baseline hepatic fibrosis status (Table 2 and Supplementary Table 1). Among this subset, levels of each protein were higher among individuals with significant hepatic fibrosis than among those with absent or mild disease (Figure 1). Moreover, the odds ratios (95% confidence intervals) for hepatic fibrosis stages 2–3 (vs 0–1) per 1–standard deviation change in plasma protein level were robust, ranging from 2.10 (1.05–4.23) for tumor necrosis factor (TNF) receptor superfamily member 9 (TNFRSF9) to 3.68 (1.64–8.26) for matrix metalloproteinase 2 (MMP-2). The top 3 proteins that most significantly differed between groups were MMP-2, insulin-like growth factor–binding protein 7 (IGFBP-7), and collagen α1(I) chain (COL1A1). In the overall sample, there was no association of statin use with plasma levels of the 20 differentially expressed proteins.

Using principal component analysis, principal component 1 (PC1) was found to capture >50% of the variance in the 20 differentially expressed proteins (Supplementary Figure 1). Moreover, PC1 was higher in participants with than in those without significant hepatic fibrosis (mean [SD] 1.33 [1.37] vs −0.46 [1.28]; P < .001). This difference in PC1 between fibrosis groups persisted in models adjusting for Fibrosis-4 score (P = .02) or VAT (P = .002) as key clinical predictors of hepatic fibrosis. Notably, the relationship of VAT with significant fibrosis was no longer significant after adjustment for PC1 (P = .79). These findings suggested the potential for select plasma proteins to mediate the relationship between visceral adiposity and advanced hepatic disease, which we later explored in formal mediation analyses.

Network Analysis of Plasma Proteins Differentially Expressed by Hepatic Fibrosis Status

Network analysis of the 20 differentially expressed proteins and related proteins imputed by GeneMANIA revealed significant interactions between molecules, particularly based on coexpression and shared protein domains (Figure 2). The constructed network of proteins primarily converged on 2 nodes, one related to the TNF response and the other related to peptidase activity. The node pertaining to TNF response included differentially expressed proteins lymphotoxin-β receptor (LTBR; also known as TNF receptor superfamily member 3 [TNFRSF3]), TNFRSF9, osteoprotegerin (OPG; also known as TNF receptor superfamily member 11B [TNFRSF11B]), and TNF receptor superfamily member 14 (TNFRSF14; also known as herpesvirus entry mediator [HVEM]) in addition to imputed proteins TNF receptor superfamily member 8, CD40L receptor, and ectodysplasin A2 receptor. Moreover, the peptidase activity node contained differentially expressed protein aminopeptidase N (AP-N; also known as ANPEP) as well as imputed proteins endoplasmic reticulum aminopeptidase 2 and glutamyl aminopeptidase. Functional enrichment analysis of the collective set of differentially regulated and imputed proteins revealed enrichment within pathways pertaining to the immune response (eg, response to TNF; positive regulation of leukocyte activation, regulation of lymphocyte proliferation), tissue repair (eg, aminopeptidase activity, exopeptidase activity, response to mechanical stimulus), and cell death (eg, positive regulation of apoptotic process, neuron death, extrinsic apoptotic signaling pathway) (Table 3).

Associations of Key Plasma Proteins With Clinical Parameters

We next examined baseline relationships of the 20 proteins differentially expressed by hepatic fibrosis status with select clinical parameters among all study participants (Figure 3). Multiple plasma protein levels were found to be tightly correlated with NAFLD activity score (eg, IGFBP-7, r = 0.48 and P < .001; LTBR [TNFRSF3], r = 0.44 and P < .001). Furthermore, strong positive associations were observed between plasma protein levels and VAT (IGFBP-7, r = 0.50 and P < .001; TNFRSF14 [HVEM], r = 0.47 and P < .001) but not subcutaneous adipose tissue. As an index of metabolic dysfunction, differentially expressed proteins directly related to glucose area under the curve during an oral glucose tolerance test (eg, IGFBP-7, r = 0.40 and P = .002; TNFRSF9, r = 0.34 and P = .01). Finally, lower CD4⁺ T-cell count was associated with higher levels of key fibrosis-associated proteins (eg, IGFBP-7, r = -0.38 and P = .003; MMP-2, r = −0.38 and P = .003).

Mediation of Relationship Between Visceral Adiposity and Hepatic Fibrosis by Key Plasma Proteins

Given correlations between visceral fat content and key plasma proteins, we next sought to assess whether select proteins that were differentially expressed by hepatic fibrosis status may mechanistically underlie the relationship of visceral adiposity with hepatic fibrosis in HIV that we previously observed (Supplementary Table 2) [5]. Among the 13 proteins correlated with visceral adiposity, we hypothesized that 10 may be potential mediators based on (1) biologic plausibility and (2) known expression by visceral fat tissue according to the Human Protein Atlas [26]: IGFBP-7, galectin 4 (Gal-4; also known as LGALS4), osteopontin (OPN; also known as secreted phosphoprotein 1 [SPP1]), OPG (TNFRSF11B), granulins, interleukin 12 (IL12), cathepsin Z (CTSZ), LTBR (TNFRSF3), TNFRSF9, and TNFRSF14 (HVEM). In a mediation analysis, we found that each of these proteins mediated a significant proportion (>30%) of the relationship between VAT and hepatic fibrosis, indicating a potential biologic role in this regard.

Differential Changes in Plasma Protein Levels With Hepatic Fibrosis Progression

In an exploratory analysis, we examined differential changes in plasma protein levels over a 12-month period in placebo-treated individuals with versus without hepatic fibrosis progression. Among 23 individuals in the placebo arm with paired liver biopsy specimens and proteomic data available, 9 participants (39%) demonstrated fibrosis progression, whereas 14 participants (61%) exhibited no worsening of fibrosis. Baseline clinical characteristics within the placebo-treated arm were comparable to the overall study group [5]. Of the 183 plasma proteins examined, 11 proteins were found to increase over 12 months among individuals with versus without hepatic fibrosis progression (P < .05) (Supplementary Table 3). As in our baseline analysis, these proteins predominantly pertained to tissue repair or immune response pathways. Notably, TNFRSF9 and Gal-4 (LGALS4), which were up-regulated among individuals with significant hepatic fibrosis at baseline, were also found to increase in association with hepatic fibrosis progression in this independent analysis.

DISCUSSION

In an unbiased analysis leveraging 2 high-multiplex panels, we identified 20 plasma proteins up-regulated in the context of significant hepatic fibrosis among PLWH and NAFLD. These differentially expressed proteins broadly pertained to immune response and tissue repair pathways. Moreover, select proteins were shown to correlate with adiposity, glucose intolerance, and HIV-associated immune dysregulation, which may offer novel insights into their regulation and function. In an exploratory analysis, we further identified 11 plasma proteins that differentially increased over time among placebo-treated individuals with hepatic fibrosis progression, 2 of which overlapped with proteins up-regulated in our baseline analysis. Taken together, this study uncovers potential mechanisms and biomarkers of hepatic fibrosis in HIV-associated NAFLD, which have implications for disease detection and drug discovery.

Inflammation is well known to contribute to the pathogenesis of hepatic fibrosis in NAFLD owing to cross-talk between immune and hepatic stellate cells [16]. In the current study, at baseline, we found that hepatic fibrosis stages 2–3 were associated with up-regulation of key plasma proteins involved in immune modulation, compared with fibrosis stages 0–1. Four of these differentially expressed proteins corresponded to the TNF family: LTBR (TNFRSF3), TNFRSF9, OPG (TNFRSF11B), and TNFRSF14 (HVEM). In addition, network analysis revealed a node of TNF-associated proteins, whereas response to TNF was identified as the topmost enriched Gene Ontology pathway identified by functional enrichment analysis. Beyond members of the TNF family, we found elevations of other immunomodulatory proteins in the setting of significant hepatic fibrosis. In this regard, proinflammatory OPN (SPP1) [27], CC motif chemokine 2 (also known as monocyte chemoattractant protein 1) [17, 18], IL12 [17, 28], and integrin β2 [29] have been previously shown to be up-regulated in other contexts of hepatic fibrosis, and may contribute to the pathogenesis of disease in PLWH and NAFLD. Conversely, the soluble decoy receptors interleukin 1 receptor type 2 and interleukin 18–binding protein antagonize proinflammatory cytokines, and thus their up-regulation in our participants with hepatic fibrosis may signify a counterregulatory immunosuppressive response.

Beyond proteins involved in immune function, we have found that more advanced hepatic fibrosis at baseline was associated with higher circulating levels of proteins related to tissue repair and extracellular matrix remodeling. Notably, proteins with roles in both tissue formation (COL1A1, IGFBP-7, and decorin [DCN]) and tissue degradation (MMP-2, CTSZ, and AP-N [ANPEP]) were shown to be up-regulated among individuals with more severe hepatic disease. Relatedly, network analysis showed that differentially expressed and imputed proteins converged on a proteolytic node, whereas functional enrichment analysis revealed overrepresentation within Gene Ontology gene sets pertaining to aminopeptidase and metallopeptidase activity. Hepatic fibrosis is a dynamic condition characterized by accelerated tissue synthesis and breakdown [30]. These processes lead to alterations in composition, which in turn disrupt normal hepatic function and thereby contribute to morbidity and mortality risk [30]. In support of our findings in HIV-associated NAFLD, hepatic or circulating levels of COL1A1 [31, 32], IGFBP-7 [23, 33], DCN [34, 35], MMP-2 [36, 37], and AP-N (ANPEP) [19, 38] have been shown to be elevated in various contexts of hepatic fibrosis, whereas CTSZ [39] was found to be amplified and up-regulated in hepatocellular carcinoma.

While HIV-associated NAFLD is characterized by high rates of fibrosis presence and progression [5], HIV-specific mechanisms that underlie this exaggerated phenotype remain poorly understood. In the current analysis, we found inverse associations between CD4⁺ T-cell count and 10 of the 20 plasma proteins up-regulated with significant baseline hepatic fibrosis, which may implicate HIV-related immune dysfunction as an important driver of the fibrogenic response. Indeed, in prior studies of HIV/hepatitis C virus coinfection, lower CD4⁺ T-cell counts have been linked to higher rates of fibrosis prevalence and progression [40], end-stage liver disease [4], and hepatic decompensation [25]. Similarly, in our current study of HIV-associated NAFLD, CD4⁺ T-cell counts tended to be lower among individuals with versus without advanced fibrosis, though this relationship did not achieve statistical significance. In contrast to the inverse correlation with hepatic fibrosis, a 2017 meta-analysis demonstrated a direct association of CD4⁺ T-cell counts with NAFLD among PLWH [1]. The opposing relationships of CD4⁺ T-cell counts with hepatic steatosis and fibrosis suggest differential regulation of these processes, underscoring the critical need for dedicated studies in individuals with HIV and NAFLD-related fibrosis, who are at highest risk of adverse health outcomes.

As for additional HIV-specific mechanisms of hepatic fibrosis, visceral adiposity and fat redistribution are hallmark features of HIV [6]. Furthermore, visceral fat accumulation has been identified as a key clinical predictor of fibrosis in HIV-associated NAFLD, which may be due to profibrogenic mediators secreted by adipose tissue that target the liver by way of the portal circulation [5]. In the current study, we performed a mediation analysis to determine whether select plasma proteins might mechanistically underlie the association between visceral fat and hepatic fibrosis. Indeed, we showed for the first time that various proteins highly expressed in visceral fat including IGFBP-7, CTSZ, and TNFRSF14 (HVEM) mediated a substantial proportion of the relationship between visceral fat accumulation and liver fibrosis. These findings shed light on the biologic underpinnings whereby visceral fat may modulate the severity of NAFLD in HIV. Further studies are needed to examine these candidate proteins in the causal pathway between visceral fat and liver disease.

To extend our analysis of baseline data, we performed an exploratory investigation of longitudinal changes in plasma protein signatures within the placebo-treated arm. We identified 11 proteins whose circulating levels differentially increased over 12 months among individuals with versus without hepatic fibrosis progression. Notably, 2 of these proteins—TNFRSF9 and Gal-4 (LGALS4)—were also found to be up-regulated in those with more advanced hepatic fibrosis at baseline. TNFRSF9 is a proinflammatory molecule responsible for the activation and proliferation of various immune cell types [41]. Moreover, Gal-4 (LGALS4) has been shown to be aberrantly expressed in the context of hepatocellular carcinoma [15]. The association of TNFRSF9 and Gal-4 (LGALS4) with fibrosis presence and progression in separate unbiased analyses elevates the candidacy of these proteins as potential biomarkers or therapeutic targets of disease and provides a compelling rationale for their further examination in future work.

To our knowledge, this is the first study to leverage unbiased proteomics approaches to identify protein signatures of hepatic fibrosis among individuals with HIV-associated NAFLD. Moreover, our data link perturbations in these pathways to HIV-associated immune dysfunction, suggesting novel regulatory mechanisms in this population. Importantly, our cohort had HIV infection well controlled by modern antiretroviral therapy, which supports the generalizability of our findings with respect to contemporary PLWH. As a limitation of the current analysis, our sample size was relatively small, and thus some comparisons may have been underpowered. Nonetheless, with approximately 60 biopsy samples analyzed, the size was adequate to show robust associations of novel proteins with hepatic fibrosis at baseline. Moreover, we cannot determine from this observational study the causal directionality of relationships between protein levels and hepatic disease. However, we did perform analyses to suggest that these proteins may be in the pathway mediating effects of visceral fat on hepatic fibrosis. Finally, while we found relationships between key plasma proteins and CD4⁺ T-cell count, future studies should examine relationships of these proteins with other measures of HIV-associated immune dysfunction, including the magnitude of the HIV viral reservoir.

Taken together, in this unbiased proteomics analysis of individuals with HIV-associated NAFLD, we identified 20 plasma proteins that were differentially up-regulated in the context of significant hepatic fibrosis. These proteins primarily corresponded to immune activation and tissue repair pathways, including TNF response and aminopeptidase activity. Importantly, correlations of plasma peptides with visceral fat content and CD4⁺ T-cell count shed light on the biologic underpinnings linking adiposity and immune dysregulation to hepatic fibrogenesis in HIV. Overall, findings from this study suggest biomarkers and therapeutic targets for advanced hepatic disease in HIV-associated NAFLD to be further validated and explored in future work.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank the research volunteers who participated in this study as well as the Clinical Research Center staff at the Massachusetts General Hospital and the National Institutes of Health.

Financial support. This work was supported by the National Institutes of Health (grants K23HD100266 to L. T. F.; R01DK114114 to T. L. S. and K. E. C.; K23HL147799-01 to M. Toribio; R01DK108370, R01AI136715, and U19 AI082630 to R. T. C.; U01AI115711 to C. M. H. and S. K. G.; and P30DK040561 to S. K. G.); the American Heart Association–Harold Amos Medical Research Faculty Development Program, supported by the Robert Wood Johnson Foundation (M. Toribio); the MGH Research Scholars Program (R. T. C.); the National Institute of Allergy and Infectious Disease Intramural Research Program; and the Intramural Research Program of the NIH, National Cancer Institute.

References

Author notes