Our epigenome-wide study jointly analyzed three inflammatory markers and found a total of 295 associations between blood-based DNAm and inflammation in more than 900 male PWH from VACS. Many of these associations were significant in more than one single-trait EWAS, particularly IL-6 and sCD14. By leveraging information from all three inflammatory markers, we identified several CpG sites that were not revealed by any of the single-trait EWAS. The beta-coefficients estimated in the three single-marker EWAS were highly correlated. They were also correlated with beta-coefficients estimated from an EWAS of CRP in the general populations36. This suggests concurrent increase or decrease of inflammatory markers might be associated with changes of DNAm at these identified CpG sites.
The leading CpG sites were mapped to genes implicated in immune and antiviral response, such as STAT1 and IFITM1. Many identified genes harbored multiple significant CpG sites, suggesting that CpG methylation in these regions may be regulated concomitantly in response to inflammation in PWH. Many of the identified CpG sites have been reported as associated with at least one autoimmune disease such as rheumatoid arthritis, Crohn's disease, or multiple sclerosis in the general population41–44. The gene-set enrichment analysis based on significant CpG sites also confirmed the role of these genes in inflammatory processes by highlighting the defense response to virus, type-1 interferon signaling, and innate immunity pathways.
Identified genes including STAT1, IFITM1, IFIT1, IFIT3, MX1, OAS2-3, PARP9, PSMB8, TMEM49, EPSTI1, IFI27, IFI44L, IRF7, ISG20 and ISG15 are all interferon stimulated genes (ISGs), namely genes whose expression can be robustly up-regulated by interferon signaling45. Beta-coefficients for all significant CpG sites located in these ISGs are negative, suggesting that hypomethylation at these CpG sites at ISGs, which usually leads to their higher gene expression, is associated with inflammation. CpG sites on PARP9, MX1, IFI44L, EPSTI1, and TAP1 have been previously reported to be hypomethylated in ART-naïve PWH compared to individuals without HIV, which could be related to inflammation in response to HIV infection46. In our study, over half of the participants are virally suppressed, and the hypomethylation of the CpG sites at ISGs may be due to oxidative stress triggering a non-specific response, herpesvirus reactivation, gut translocation of bacteria, or persistent viral replication47.
The transcription of ISGs is induced by interferon signaling through the Janus kinase (JAK)–signal transducer and activator of transcription (STAT) pathway48. The JAK-STAT pathway can both be triggered by and cause inflammation. In the JAK-STAT pathway, STAT1 molecules interact with other STAT proteins and MUC1 to bind to the promoter regions of other ISGs which leads to their increased expression49,50. The protein encoded by IFITM1, which harbored 10 significant CpG sites in our study, inhibits HIV replication partially through interfering with virus-cell membrane fusion, viral particle infectivity, and viral protein synthesis51–53. IFIT proteins can develop a functional complex to restrict viral translation and modulate innate immune signaling54. Specifically, IFIT1 molecules bind to an uncapped 5′-ppp on HIV RNA to sequester those RNA while IFIT3 proteins serve to enhance the binding function of IFIT1 molecules54,55. Proteins encoded by genes MX1, OAS2-3, IFI27, IRF7 and ISG20 have also been shown to inhibit HIV viral replication by diverse mechanisms56–59. On the other hand, proteins PARP9 and IFI44L may contribute to antiviral responses by mitigating STAT1 molecules in the JAK-STAT pathway60,61 while ISG15 functions as a negative regulator of IFN-1 signaling to prevent overt inflammation62. A recent study has shown that cells treated with IFITM1-targeted Small interfering RNA (siRNA) cells had attenuated levels of STAT1 and ISG15, suggesting a mediating effect of IFITM1 protein in IFN-γ stimulated protein synthesis63. The interferon signaling pathways could be sustaining the persistent inflammation in PWH. Targeting DNAm at the ISGs-mediated pathways may provide new therapeutic avenues for addressing HIV-associated chronic inflammation and its many downstream effects.
Genes outside the ISG group may also contribute the inflammatory process among PWH. For example, cg07839457 located within the promoter region of NLRC5 was previously shown to be hypomethylated in ART-naïve PWH compared to individuals without HIV, and to become significantly hypermethylated after two years of successful ART, although less so than in individuals without HIV46. This hypomethylation has also been observed in adults and children with HIV on ART22,64. In our study, we showed that hypomethylation of cg07839457 (NLRC5) was linked to IL-6, sCD14 and D-dimer levels. This finding supports the hypothesis that the effects of HIV and ART on DNAm at this specific site, are mediated through inflammation. NLRC5 encodes a protein that negatively regulates nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and type I interferon signaling pathways64, and may serve as a useful target for controlling inflammation in PWH. Further research, ideally with DNAm and inflammatory markers at multiple time points after ART, is needed to elucidate the relationship between DNAm at cg07839457 and persistent inflammation. Other promising genes identified by top CpG sites include FKBP5, which promotes NF-κB pathways,65 and KLHDC7B, which regulates ISGs such as IFITs and STAT1.
Using the EWAS atlas and EWAS catalog, we showed that many of the CpG sites we identified were previously reported as linked to HIV infection, chronic kidney disease, cancer and diverse autoimmune diseases in the general population (Supplementary Table 6). Because many of these disease outcomes have higher prevalence in PWH compared to the general population66,67, inflammation could increase HIV comorbidities through DNAm changes at the identified CpG sites. A longitudinal study design that establishes temporality between inflammation, DNAm and HIV comorbidities could help us test these hypotheses and better understand their relationships.
Compared to other methods, CPASSOC and OmniTest combine summary statistics from single-trait EWAS with fewer assumptions about relationships among the three inflammatory markers, which allowed us to treat the inflammatory markers as predictors as in our hypothesis. While we are most confident in the overlapping CpG sites identified by CPASSOC and OmniTest, both methods also provided unique CpG sites to explore due to their different ways of incorporating the correlation structure.
Owing to the unique multi-trait EWAS study design for PWH, we were not able to identify an independent dataset with the same phenotypic and DNAm data in which to replicate our findings. Additionally, our study included only men with HIV and the findings might not be generalizable to women or pediatric population. We used blood-based DNAm in this study and the observed associations might be different in other tissues. Lastly, this study had inflammatory markers measured at only one time point, which may not fully capture the complexity and dynamics of inflammation.
In conclusion, we present evidence that delineates molecular mechanisms associated with HIV-related inflammation through DNAm. We show that hypomethylation at specific CpG sites correlates with increased levels of three key inflammatory markers, making these sites promising targets for mitigating inflammatory responses. Our analyses underscore the importance of ISGs in combating inflammation, which may persist in PWH due to alterations in DNAm. Interestingly, some of these CpG sites have also been linked to diverse inflammation-related phenotypes in the broader population suggesting that a better understanding of the function of these CpG sites could enhance therapeutic strategies for a range of diseases with similar inflammatory pathophysiology.