Increased prevalence of clonal hematopoiesis of indeterminate potential amongst people living with HIV

People living with human immunodeficiency virus (PLWH) have significantly increased risk for cardiovascular disease in part due to inflammation and immune dysregulation. Clonal hematopoiesis of indeterminate potential (CHIP), the age-related acquisition and expansion of hematopoietic stem cells due to leukemogenic driver mutations, increases risk for both hematologic malignancy and coronary artery disease (CAD). Since increased inflammation is hypothesized to be both a cause and consequence of CHIP, we hypothesized that PLWH have a greater prevalence of CHIP. We searched for CHIP in multi-ethnic cases from the Swiss HIV Cohort Study (SHCS, n = 600) and controls from the Atherosclerosis Risk in the Communities study (ARIC, n = 8111) from blood DNA-derived exome sequences. We observed that HIV is associated with a twofold increase in CHIP prevalence, both in the whole study population and in a subset of 230 cases and 1002 matched controls selected by propensity matching to control for demographic imbalances (SHCS 7%, ARIC 3%, p = 0.005). We also observed that ASXL1 is the most commonly mutated CHIP-associated gene in PLWH. Our results suggest that CHIP may contribute to the excess cardiovascular risk observed in PLWH.


Methods
We identified CHIP in a multi-ethnic sample of 600 PLWH from the Swiss HIV Cohort Study (SHCS), aged 21-83. The SHCS is a multicenter, prospective observational study for interdisciplinary HIV research 7 . Established in 1988, the SHCS currently comprises more than 20,000 PLWH with median 51 years of age. Samples of 600 patients, used for exome sequencing, were chosen randomly in terms of gender (genetic sex, gender at birth), age, category of transmission, as well as HIV management and control 8 .
We utilized a set of 8111 individuals with available exome sequences from the Atherosclerotic Risk in the Community study (ARIC), aged 45-84 years, as population controls 9 . The ARIC study is a prospective longitudinal investigation of the development of atherosclerosis and its clinical sequelae which enrolled 15,792 individuals aged 45 to 64 years at baseline 10 . At study enrollment (1987)(1988)(1989), the participants were selected by probability sampling from four United States communities: Forsyth County, North Carolina; Jackson, Mississippi; the northwestern suburbs of Minneapolis, Minnesota; and Washington County, Maryland.
Exome capture kits used in the compared cohorts were different (SHCS: xGen Exome Research Panel v 1.0, Sureselect All Exon V5 and TruSeq DNA Exome, ARIC: HGSC VCRome 2.1 design (42 Mb, NimbleGen)), which is the major limitation of the study. To minimise these differences we performed a set of statistical analysis to normalize the coverage among cohorts (see below the adjusting for the depth of sequencing and table S4).
CHIP was called in both exome sequenced cohorts using an identical and previously described pipeline 4,11 . Briefly, short read sequence data were aligned to the hg19 reference genome using the BWA-mem algorithm and processed with the Genome Analysis Toolkit MuTect2 tool to detect somatic variants 12 . To identify individuals with CHIP, we used a pre-specified list of variants in 74 genes known to be recurrent drivers of myeloid malignancies (Table S3) and variant filtration process (> = 20 reads in total, > = 3 Alt reads including one on a forward and one on a reverse strand, VAF limit > 2%) which filter in the biologically relevant cases 13 .
As CHIP prevalence depends strongly on age, we performed a 1:5 case/control propensity matching on age, sex and self-reported ethnicity using nearest neighbor matching 14 as implemented by the MatchIt package version 3.0.2 in R. Next, we used univariate Fisher's exact test and multivariable logistic regression to test the association between HIV status and CHIP prevalence. Multivariable models were adjusted for age, sex, self-reported ethnicity, and smoking status.
To take into account potential difference in the depth of sequencing of the CHIP-associated genes between the matched cohorts, we used a backward stepwise multiple logistic model, describing CHIP status (0/1) as a function of cohort (0/1) and coverage of the four most common CHIP-associated genes (DNMT3A, TET2, ASXL1 and JAK2). Analyses were performed in R version 3.6. A threshold of p < 0.05 was considered statistically significant.
All methods were performed in accordance with the relevant guidelines and regulations.

Results
First, we compared the prevalence of CHIP across the entire SHCS PLWH cohort (N = 600) and ARIC cohort (N = 8111) ( Fig. 1). SHCS PLWH and ARIC participants had mean (SD) age 44 (11) and 57 (6) years (p = 1.8 × 10 -167 ), were 25% and 56% female (p = 1.9 × 10 −46 ), and were 95% and 74% of European ancestry (p = 5.2 × 10 −36 ) respectively. With adjustment for age, age 2 , sex and self-reported ethnicity, we observed a significant association between HIV case status and CHIP (OR: 1.77, 95% CI: 1.33-2.21, p = 0.02). Second, given the overall demographic imbalances, to confirm an excess of CHIP amongst SHCS under all else equal, we used a propensity matching strategy to match the two cohorts by age, gender, self-reported ethnicity and smoking status (ever-smoker or not). Propensity matching analyses yielded a set of 230 (out of 600) PLWH cases and 1002 (out of 8111) ARIC population controls. Neither age nor sex differed significantly between the matched cohorts (Table 1) and the standardized mean difference across age, sex and self-reported ethnicity were all less than 0.1 indicative of adequate matching. In this subset, CHIP was detected in 7% of exomes from PLWH, but only 3% of the controls (   www.nature.com/scientificreports/ The limited sample size precluded inference on the association of HIV status with specific CHIP driver genes, however we observed differences in the genes most likely to carry CHIP mutations between PLWH (table S1) and population controls (table S2). The most common CHIP gene in the SHCS was ASXL1 (13 out of 27 CHIP mutations, 48%) followed by TET2 (8 out of 27 CHIP mutations, 30%) and DNMT3A (5 out of 27 CHIP mutations, 19%). Overall this distribution was inverted from the control cohort where CHIP mutations were more frequent in DNMT3A (14 out of 28 CHIP mutations, 50%), followed by TET2 (5 out of 28 CHIP mutations, 18%) and ASXL1 (5 out of 28 CHIP mutations, 18%). In total, 22 PLWH had a single CHIP mutation, while one individual had 2 mutations and one individual had 3 mutations, while in the ARIC cohort all CHIP carriers had a single CHIP mutation (tables S1 and S2). Additionally, we compared VAFs between matched ARIC and SHCS CHIP carriers (tables S1 and S2) and observed a trend of increased VAF in ARIC (28 CHIP mutations in ARIC and 14 CHIP mutations in SHCS: 12 patients among which one patient has 3 CHIP variants, p-value = 0.026, Mann-Whitney U test).
Within the full PLWH cohort (N = 600) we considered additional phenotypes, which might be a cause or consequence of CHIP. First, we observed a trend toward an increase in CAD among CHIP carriers (Fisher's exact test OR: 2.99, p = 0.068) and increased cases of diabetes among CHIP carriers (Fisher's exact test OR: 3.76, p = 0.037) (see the patient-specific information in Table S1). Second, we observed that duration of antiretroviral therapy (ART) was twice as long in CHIP carriers versus non-carriers (ART mean [st. dev.] 2675 [1850] days vs. 1322 [1454] days in carriers vs. non-carriers, respectively; p = 0.0004, Mann-Whitney U test). This association was directionally concordant after adjusting for patient age in multiple logistic regression (p = 0.066). It is important to note that although ART duration positively correlated with the total duration of HIV infection (Spearman's rho = 0.58, p = 2.0 × 10 −54 ), the total duration of HIV infection was not associated with CHIP (p = 0.452; paired Mann-Whitney U test on matched CHIP carriers and non-carriers, p = 0.22).

Discussion
Here, we report that HIV infection is associated with increased prevalence of CHIP. In the present samples, we identify a more than twofold enrichment of CHIP among PLWH versus controls after careful adjustment for known factors predisposing to CHIP (age, smoking status, ethnicity, gender). Although PLWH and controls originate from different cohorts, identical bioinformatics pipelines were used for the identification of CHIPassociated variants, and the statistical matching of cohorts and multiple logistic models controlling for the gene coverage (see Methods) assure robustness of our results. Of note, a very similar twofold excess of CHIP among PLWH has been described recently in an independent study from Australia 15 . Altogether, we demonstrate that HIV infection is the second strongest factor, after age, associated with increased CHIP prevalence among PLWH.
Our finding is based on cross-sectional design and thus we cannot infer causality. However, assuming that CHIP is highly unlikely to be a risk factor for HIV acquisition, we focus on potential mechanisms that could promote CHIP development among PLWH. HIV infection or, more generally, HIV-related factors may promote CHIP development either through an increased rate of occurrence of CHIP-associated mutations and/or increased rate of clonal expansion of these somatic variants.
An increased somatic mutational rate in PLWH up to date has been shown only for the mitochondrial genome, which is particularly sensitive to some antiretroviral therapies [16][17][18][19] . However, taking into account several potential factors such as: a mutagenic effect of the virus, DNA-replication errors associated with an increased turnover rate of hematopoietic stem cells, mutagenic effects of antiretroviral therapy and other HIV-specific confounders (such as tobacco smoke, the effect of which has been controlled in the current study) we can not rule out an increased rate of somatic mutagenesis in the nuclear genome of PLWH.
A potentially higher rate of occurrence of somatic mutations alone is unlikely to provide a comprehensive explanation of increased CHIP prevalence among PLWH. A recent study, performing accurate detection of rare (with variant allele frequency higher than or equals to 0.0003) CHIP-associated mutations, demonstrated that such CHIP variants are nearly universal in healthy individuals by the age of 50 and often stable longitudinally 20 , showing that for the majority of people, decades elapse between the acquisition of a CHIP-associated mutation and CHIP itself. Thus, an understanding of the rate of clonal expansion of initially rare CHIP-associated variants is of great importance to shed light on the excess of CHIP among PLWH. A recent model proposed that many of the CHIP-associated mutations increase cell fitness, ensuring their proliferation with age 21 . Thus, HIV infection may modify the fitness landscape of CHIP-associated mutations, accelerating their clonal expansion and thus providing a fertile substrate for CHIP development. Various HIV-related mechanisms may be responsible for this, including induced immunodeficiency, increased prevalence of tobacco smoking and other comorbid conditions, as well as chronic immune activation from antigenic stimulation. Indeed, it has been recently shown that mutations in both the most common CHIP-associated genes DNMT3A 22 and TET2 23 are getting selective advantage in case of chronic infection. According to our results, ART could additionally induce CHIP development, however an elucidation of the mechanisms as well as relative contribution of different HIV-specific factors to CHIP risk requires future studies.
The relationship we identify between HIV and CHIP may be a mechanistic basis of shared phenotypes. For example, recent study showed that HIV infection leads to a greater risk of myelodysplastic syndrome (MDS), a downstream consequence of CHIP and precursor to myeloid malignancy 24 . Furthermore, similar to the gene distribution in MDS, we find a greater relative prevalence of ASXL1 mutations among PLWH compared to controls. Of note, while cigarette smoking selects for ASXL1 clonal hematopoiesis 25 , our cohort of PLWH still had an increased prevalence of ASXL1 mutations compared to the control cohort despite being matched for smoking status. Another shared phenotype is an increased risk for cardiovascular disease. We propose that CHIP may be one mechanism that elevates risk for CAD among PLWH and further studies are required to evaluate this hypothesis.

Scientific Reports
| (2022) 12:577 | https://doi.org/10.1038/s41598-021-04308-2 www.nature.com/scientificreports/ PLWH have accelerated biologic aging. CHIP detection may represent a new opportunity for identification of at-risk patients with particular relevance for HIV medicine. Conversely, PLWH may provide a rich source of information to understand mechanisms of clonal expansion of different CHIP-associated variants under longterm low-grade inflammation.

Data availability
CHIP-associated genetic variant callsets and associated participant level phenotype data used in this study are available to qualified investigators by application to the SHCS and ARIC.