https://orcid.org/0000-0002-8611-4900
Abstract
Unbiased plasma proteomics in a matched case-control study of treated people with human immunodeficiency virus (PWH) revealed the complement cascade as being among the top pathways enriched in PWH. Specific complement components, namely C5, associated significantly with non-AIDS comorbidity prevalence, and did so more strongly than previously established predictive biomarkers.
While effective antiretroviral therapy (ART) has profoundly extended the lifespan of people with human immunodeficiency virus (PWH) by suppressing viral replication, PWH still experience a higher incidence of various non-AIDS and noninfectious comorbidities that contribute to morbidity and early mortality. Occurrence of these noncommunicable diseases (NCDs) has been linked with elevated levels of systemic immune activation, and numerous biomarkers, including C-reactive protein (CRP), interleukin 6 (IL-6), and soluble CD14 (sCD14) [1, 2], which have been identified to be independently associated both with comorbidity prevalence and early mortality. The precise immune pathways involved in the pathogenesis of these inflammation-mediated noninfectious complications remain poorly understood. Elucidation of such pathways is needed to provide novel targets for interventions, clinical trial endpoints that are more closely linked with pathogenesis, and improved predictors of morbidity/mortality that could be used in the clinical management of PWH.
The complement activation pathway represents a unique junction of innate and adaptive immunity directed against pathogens, which comprises a cascade of molecular switches that ultimately converge on pathogen opsonization and the assembly of a membrane attack complex that forms a lytic pore in the pathogen cellular membrane. Another function of this cascade is to recruit and activate innate immune cells including macrophages and granulocytes. Complement-mediated activation of these cells causes secretion of factors that overlap strikingly with immune-related biomarkers that are predictive of comorbidities in PWH, including tumor necrosis factor [3], soluble tumor necrosis factor receptors [4], IL-6 [5], sCD14 [6] via IL-6, and coagulation pathway components including D-dimer [7] and fibrinogen [8], with evidence of complex bidirectional relationships between these factors. More recently, the role of complement in adaptive immunity and inflammation, including the concept of the complosome, has been described [9]. Despite the links between complement pathway components and previously identified predictive biomarkers of non-AIDS comorbidity, the role of the complement cascade in the context of successfully treated PWH remains relatively unexplored. Herein, we report that several complement factors are significantly enriched in PWH on ART as compared to seronegative controls matched for sex/sexual behavior, age, and body mass index (BMI). The specific complement component C5 was significantly associated with comorbidity prevalence and exhibited a stronger association than previously established biomarkers of non-AIDS comorbidity including sCD14, IL-6, D-dimer, soluble urokinase plasminogen activating receptor (suPAR), and CRP.
METHODS
Single-timepoint plasma samples from a subset of participants of the Comorbidity and Aging With HIV (AGEhIV) Cohort Study [10] (NCT01466582) were subjected to aptamer-based proteomic screens (SOMAscan) [11], yielding quantification of 1317 proteins. All study participants signed informed consent agreements. Participants were selected in 2 groupings (N = 78 total). First, 27 PWH were selected at random, evenly divided by females (n = 9), men who have sex with men (MSM; n = 9), and men who have sex with women (MSW; n = 9). One seronegative control matched for sex/sexual behavior, age, BMI, and smoking status was chosen for each PWH participant (analysis group 1). Second, for examining associations with human immunodeficiency virus (HIV)–associated NCDs (analysis group 2), 24 additional PWH participants were randomly selected to expand sample size (n = 5 female, n = 12 MSM, n = 7 MSW), and all PWH from analysis group 1 were included except for 2 analysis group 1 PWH who were excluded to balance ages between the groups. Baseline characteristics of analysis groups 1 and 2 are presented in Supplementary Table 1. NCDs captured in analysis group 2 included chronic obstructive pulmonary disease (5 events), type 2 diabetes (5), angina pectoris (3), non-AIDS-defining cancers (4), peripheral arterial disease (3), ischemic stroke/transient ischemic attack (3), heart failure (1), myocardial infarction (3), advanced liver fibrosis (3), osteoporotic fracture (2), and renal insufficiency (4). Pathway analyses were performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) pipeline [12] on all plasma proteins enriched in PWH with all proteins measured as a background gene list. The same pathway analysis was performed using all proteins positively associated with comorbidities among PWH (linear mixed effects t > 0). Wilcoxon rank-sum tests were performed using the R package “exactRankTests,” and linear mixed-effects models as well as mixed-effects logistic regression were performed using R packages “lme4” and “lmerTest.” All mixed-effects models included age and BMI as fixed effects and sex/sexual behavior as random effects. Effect sizes were estimated by calculating odds ratios on scaled and centered analyte values, dividing by the interquartile range. Adjusted R2 values were calculated using the “rsq” R package. Outlier values >3 standard deviations from the mean were removed from analyses, resulting in exclusion of 1.8% of analyte values. Benjamini–Hochberg multiple comparisons corrections were performed using the “p.adjust” function in R. Visualizations were produced using R packages “ggplot2,” “viridis,” and “RColorBrewer.”
RESULTS
Using our plasma proteomics to group analytes by host pathways, we assessed the relative enrichment of pathways in PWH (n = 27) as compared to matched seronegative participants (n = 27). In line with expectation, pathways related to antiviral responses were most enriched in PWH (Supplementary Table 2). Surprisingly, we found that complement activation was among the top most enriched pathways in PWH when querying host pathways via the commonly used GO Biological Process pathway database (Figure 1A), while it was the top most enriched pathway when classifying proteins by the UniProt Keywords and among the top 3 for BioCarta Pathways databases (Supplementary Figure 1A and 1B). Furthermore, when specifically examining complement-related proteins that differed in their abundance between PWH and seronegative controls, we found that 7 of 8 complement components with unadjusted Wilcoxon P < .1 exhibited enrichment in PWH (Supplementary Figure 1C).
Complement pathway components are enriched in people with human immunodeficiency virus (PWH) and correlate significantly with comorbidity prevalence. A, Pathway analysis of plasma proteins enriched in PWH compared to seronegative controls reveals complement pathway upregulation in PWH (P = .013, using Gene Ontology Biological Process pathways). B and C, 27 PWH were compared to 27 paired seronegative controls matched for sex/sexual behavior, age, body mass index, and smoking status. Serum complement-related proteins enriched in PWH were complement C5 (panel B, Benjamini–Hochberg Q = .011, performed by Wilcoxon signed-rank tests for all complement proteins), and complement factor H–related protein 5 (CFHR5, C; Q = .001). The y-axes represent SOMAscan relative fluorescence units. D, Complement C5 was associated with comorbidity prevalence when examining PWH alone (N = 49 PWH, P = .012 by logistic regression adjusted for sex/sexual behavior). E, C5 correlates with C-reactive protein (CRP), adjusted for sex/sexual behavior via linear mixed effects (P = .0003). F, CRP shows a trend toward enrichment in PWH participants compared to seronegative (logistic regression P = .09 adjusted for sex/sexual behavior). Abbreviations: CRP, C-reactive protein; GO, Gene Ontology; MSM, men who have sex with men; MSW, men who have sex with women; PWH, people with human immunodeficiency virus.
When examining all complement proteins, complement component C5 and complement factor H–related protein 5 (CFHR5) were most enriched in PWH compared to matched seronegative participants (Benjamini–Hochberg Q = .011 and Q = .001, respectively, performed on paired Wilcoxon signed-rank tests; Figure 1B and 1C, Supplementary Figure 2A–C). Of these 2 complement pathway proteins, C5 correlated significantly with having experienced a clinical NCD event in PWH (n = 49 in total, logistic regression adjusted for age, BMI, and sex/sexual behavior; P = .012; odds ratio [OR] per 1 interquartile range [IQR], 1.63; Figure 1D and Supplementary Figure 2E), whereas CFHR5 exhibited a weak association with comorbidity prevalence (P = .201; OR per IQR, 0.97; Supplementary Figure 2C and 2D). From a pathway perspective, complement pathway activation trended toward enrichment in PWH having experienced a clinical NCD event as compared to PWH without NCDs (GO Biological Process: 0006957, P = .0645).
Plasma levels of the complement component C5 did not correlate strongly with previously established predictors of non-AIDS events IL-6, D-dimer, or sCD14, though they did correlate significantly with CRP and inversely with suPAR (Supplementary Figure 3A–E). Adjustment for IL-6, D-dimer, sCD14, and CRP individually yielded singificant correlations between C5 and comorbidity prevalence (P = .021, P = .014, P = .017, and P = .019, respectively). CRP is itself linked to the complement pathway and activates C1q [13], a factor upstream of the generation of C5. Accordingly, we found a strong correlation between CRP and C5 (P = .0003 by linear mixed effects; Figure 1E). Although CRP has been found to be enriched in PWH and to associate with comorbidities in large studies in PWH [1,2], its enrichment in PWH did not reach statistical significance in contrast to C5 (P = .090 vs P = .012; Figure 1E) and had no statistically significant association with comorbidities in our study (P = .340), suggesting that C5 may be more proximal in the pathologic pathway toward HIV-associated noninfectious comorbidities. Intriguingly, adjustment for suPAR, a major component of the coagulation pathway, increased the strength of association between C5 and comorbidity prevalence (P = .002 with suPAR vs P = .012 without). A substantially greater proportion of variance was explained from logistic models predicting comorbidity prevalence that included both C5 and suPAR (adjusted R2 = 0.207) as compared to either analyte alone (C5-adjusted R2 = .135, suPAR-adjusted R2 = 0.016) or the combinations of C5 and other previously established predictors (Supplementary Figure 3F). Examining other established comorbidity risk factors, we found that associations between comorbidity prevalence and IL-6 (P = .572; OR per IQR, 1.40), D-dimer (P = .762; OR per IQR, 0.72), suPAR (P = .398; OR per IQR, 0.89), and sCD14 (P = .533; OR per IQR, 0.97; Supplementary Figures 3G–J) were weaker and exhibited no statistical significance as compared to C5. When adjusting for D-dimer, IL-6, suPAR, CRP, sCD14, and the confounding variables of age, BMI, and sex/sexual preference simultaneously, C5 retained a significant association with NCD prevalence (P = .018; OR per IQR, 2.77).
Discussion
In our study, we find that a suite of complement pathway components are significantly enriched in PWH with suppressed viremia on ART as compared to seronegative controls matched for sex/sexual behavior, age, and BMI. Complement component C5 exhibited a significant association with comorbidity prevalence, which was stronger than previously established disease biomarkers including sCD14, IL-6, D-dimer, suPAR, and CRP.
Limitations of our study include its small sample size, case-control design, and the slightly older age of people with NCDs. However, inclusion in a linear mixed model framework of common predictive biomarkers (D-dimer, IL-6, suPAR, and sCD14) as well as age, BMI, and sex/sexual preference yielded a significant correlation between C5 and NCD prevalence, indicating a robustness to age for this association. In addition, all participants were older than 45 years of age, and thus generalizability of these findings in younger PWH will need to be determined. Due to the case-control study design, direction of causality between novel biomarkers and NCD prevalence cannot be directly assessed. Well-powered longitudinal studies are warranted to assess whether C5 and the combination of C5 and suPAR (which augmented strength of association between C5 and NCDs when included in the logistic regression) may robustly predict non-AIDS event incidence, and to what extent predictive capacity may differ for different comorbid events.
While links exist in literature between complement components such as C5 and inflammatory markers, we observed only a correlation with CRP and suPAR and no correlation between C5 levels and IL-6, D-dimer, and sCD14. This is not unusual as other biomarkers such as IL-6 and D-dimer also do not co-correlate, but can have additive discriminatory value as independent predictors of events. It remains possible that C5 modulates and is modulated by the noncorrelating markers, but that serum levels of these messenger molecules are uncoupled to C5 due to differential uptake or recycling. Regardless, the observation that C5 and suPAR (components of the complement and coagulation pathways, respectively) more strongly associate with NCDs when included together in a linear mixed model framework than when considered separately may indicate that activation of those 2 pathways contribute additively to NCDs. Both pathways can amplify production of proinflammatory cytokines [9], and thus may play a role in inflammation-associated NCDs. A putative link between C5 and NCDs in PWH may be rooted in its link to specific NCDs such as cardiovascular disease, in which it is predictive of onset [14].
We propose that levels of complement pathway components like C5 should be further evaluated and may serve as predictive biomarkers for HIV-associated noninfectious comorbidities and more informative surrogate endpoints in clinical trials, as well as novel targets for new therapeutic approaches. For example, the C5a-blocking monoclonal antibody eculizumab has been shown to lower levels of D-dimer and IL-6 [15] in severe acute diseases of complement activation. This raises the possibility that targeting this pathway with similar action novel drugs with lower toxicity and cost may reduce inflammation in PWH and mitigate the risk of inflammatory comorbidities.
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 acknowledge and thank the Center for Human Immunology of the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (NIAID/NIH). We also thank C. Cortes for helpful discussion.
Disclaimer. None of the funding bodies had a role in the design or conduct of the study, the analysis or interpretation of the results, the writing of the report, or the decision to publish.
Financial support. This work was supported in part by the intramural research program of NIAID/NIH and by The Netherlands Organization for Health Research and Development (ZonMW) (grant number 300020007) and AIDS Fonds (grant number 2009063). Additional unrestricted scientific grants were received from Gilead Sciences, ViiV Healthcare, Janssen Pharmaceuticals N.V., and Merck & Co.
Potential conflicts of interest. B. S. is a former SomaLogic, Inc (Boulder, Colorado) employee and a company shareholder. P. R. through his institution has received independent scientific grant support from Gilead Sciences, ViiV Healthcare, Merck & Co, and Janssen Pharmaceuticals Inc, and has served on scientific advisory boards for Gilead Sciences, ViiV Healthcare, and Merck & Co, for which his institution has received remuneration. All other authors report no potential conflicts of interest.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
