Association of Viral Persistence and Atherosclerosis in Adults With Treated HIV Infection

This cohort study assesses whether higher levels of viral persistence are associated with atherosclerosis as assessed by changes in carotid artery intima-media thickness over time in patients living with HIV.


Introduction
The global burden of cardiovascular disease (CVD) in HIV has tripled during the past 2 decades, 1 and the risk of both CVD and myocardial infarction is increased 2-fold in persons living with HIV (PLWH). 2,3 This excess risk is owing in part to a higher prevalence of traditional risk factors in PLWH, but HIV-associated inflammation is likely also a key contributor. Other studies [3][4][5] have established that uncontrolled HIV disease is associated with increased cardiovascular risk. However, higher rates of myocardial infarction and atherosclerosis are still present in the setting of treated and suppressed HIV disease. Despite control of viremia with antiretroviral therapy (ART), inflammation persists at levels higher than in the uninfected person. 6 In turn, inflammatory and coagulation markers are strongly associated with cardiovascular events, mortality, and increased cardiovascular risk in a variety of HIV-infected populations. [7][8][9][10] Imaging studies suggest that the type of atherosclerosis in PLWH may be distinct from that in the general population. Namely, PLWH have higher arterial inflammation compared with uninfected controls, 11 and noncalcified plaque is more common in PLWH. 12,13 Carotid artery intima-media thickness (IMT) is a direct evaluation of the arterial wall 14 and accordingly has the potential to provide mechanistic insight into the pathogenesis of HIV-associated atherosclerosis. It is also a validated marker of atherosclerosis that has been shown to be associated with CVD events. 15,16 As a retrovirus, HIV integrates its genome into the host DNA. Any infected cells can consequently initiate new rounds of infection. Antiretroviral therapy prevents HIV from spreading and infecting new cells but does not eliminate those cells already infected. Because memory CD4 + T cells are maintained indefinitely via homeostatic and antigen-specific cell proliferation, 17-19 the virus persists even during long-term ART. The size of this viral reservoir can be estimated by measuring the frequency of HIV DNA in circulating memory CD4 + T cells. 20 The degree to which this reservoir is potentially "active" and produces new viral products or virions can be estimated by measuring the levels of HIV RNA in circulating cells. 21,22 Notably, much of the HIV DNA is defective and cannot support the production of replication-competent virions, but it can support the production of viral proteins. 23,24 The amount of cell-associated DNA and RNA in blood has been associated with levels of T-cell activation, T-cell proliferation, and soluble markers of inflammation in many [25][26][27][28][29] but not all 30 studies.
Given the potential role of HIV persistence in driving inflammation and given the role of inflammation in driving the development of coronary artery disease, we hypothesized that HIV DNA and RNA levels would be associated with overall atherosclerosis burden. We therefore examined the association between viral persistence and baseline carotid IMT, annual IMT progression, and incident plaque in a cohort of treated PLWH with viral suppression, controlling for traditional cardiovascular risk factors, HIV-related factors, and inflammatory markers. We also measured the association between inflammatory markers and HIV DNA and RNA levels.

Study Population
We conducted a longitudinal study of PLWH, who at our baseline visit were receiving ART and had an undetectable viral load. All participants were initially enrolled in the University of California, San Francisco-based SCOPE (Study of the Consequences of the Protease Inhibitor Era) cohort, 31 a longitudinal observational cohort of PLWH recruited from HIV/AIDS clinics at the San Francisco General Hospital and the San Francisco Veterans Affairs Medical Center from January 1, 2003, to December 31, 2012. From this cohort, we invited all participants to enroll in a substudy in which IMT was measured, as previously described. 4,[32][33][34] For the present analysis, we selected those individuals who were receiving ART and who at their first IMT visit had an undetectable viral load for at least the past 6 months. Participants were not recruited for any of these studies based on cardiovascular risk.
Candidate covariates were measured at the same time as the baseline IMT. Participants in the SCOPE cohort underwent blood tests and filled out questionnaires regarding medications, health-related behaviors, and symptoms every 4 months. Clinical data, including ART history, were obtained by extensive medical record review. The University of California, San Francisco, Committee on Human Research approved the study, and all participants provided written informed consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Carotid IMT Measurement
Carotid IMT was measured at baseline and the last visit, which occurred during a mean (SD) follow-up of 4.2 (2.7) years. High-resolution B mode ultrasonography was used to measure carotid IMT in the near and far walls of the bilateral common carotid, bifurcation, and internal carotid regions, according to the standardized protocol of the Atherosclerosis Risk in Communities Study, as previously described. 4,35 Within each segment (common, bifurcation, and internal), we calculated IMT as the mean of the near and far walls of the bilateral carotid arteries, blinded to other laboratory values.
Overall mean carotid IMT was calculated as the mean of the 12 segments. Plaque was defined as a focal region of carotid IMT greater than 1.5 mm. Incident plaque was defined as no plaque at baseline visit and plaque at last visit. Progression of IMT was defined as any positive change in carotid IMT from baseline to last visit. Annual IMT progression was defined as continuous absolute change in carotid IMT from baseline visit to last visit divided by the time between baseline and the last visit.

Measurement of Viral Persistence and Plasma Biomarkers
Viral persistence parameters were measured at baseline only. Using thawed, cryopreserved peripheral blood mononuclear cells, total cellular RNA and DNA from enriched CD4 + T cells were purified and measured using quantified polymerase chain reaction analysis. Detailed laboratory methods are described in the eMethods in the Supplement. Soluble markers of inflammation (levels of interleukin 6 [IL-6], tumor necrosis factor, C-reactive protein, soluble CD14 [sCD14], and sCD163) and coagulopathy (levels of D-dimer) were measured at baseline in cryopreserved plasma samples using a multiplex electrochemiluminescence assay (Meso Scale Discovery). multivariable approach to examine the association between each of the 3 primary independent variables (HIV DNA, HIV RNA, and RNA:DNA ratio) and the 2 outcomes (1) alone, (2) adjusted for demographic factors, and (3) adjusted for demographic factors plus traditional cardiovascular and HIV-related risk factors. In addition to the full multivariable models, we also examined models adjusted for demographic factors plus 1 additional cardiovascular or HIV-related risk factor. HIV DNA and RNA are known to have extremely right-skewed distribution. In these models, we normalized them by dividing by their interquartile ranges (IQRs) (on a log scale), so their coefficients reflect change from the 25th to the 75th percentile. In the models in which baseline carotid IMT and HIV DNA, HIV RNA, and RNA:DNA ratio are dependent variables, we log-transformed them to stabilize the residual variance to best fit the models and then exponentiated the coefficients to present estimated percentage differences.
We used Poisson regression models with log-transformed time of exposure as an offset to examine the bivariable associations between candidate covariates and the outcome of incident plaque among individuals who did not have plaque at baseline. 36,37 A robust variance estimator was applied to adjust for potential overdispersion in the models. We then used a staged multivariable approach, as described above, to examine the association of each of the 3 primary independent variables (HIV DNA, HIV RNA, and RNA:DNA ratio) with incident plaque. Incidence risk ratios (IRRs) were calculated for all of these models.
Because CD4:CD8 ratio and CD4 + T-cell counts are associated with the measures of viral persistence, 29,38 we performed sensitivity analyses in which we removed CD4:CD8 ratio or CD4 count from each of the models examining incident baseline carotid IMT, annual IMT progression, and incident plaque as outcomes. For our secondary analyses, we sought to identify factors associated with HIV RNA, HIV DNA, and RNA:DNA ratio. We created models that were adjusted for demographics and additionally included HIV-related factors and markers of inflammation (levels of IL-6, tumor necrosis factor, C-reactive protein, sCD14, and sCD163) or coagulopathy (level of D-dimer) that remained significant in the multivariable models.
In linear regression models with randomly missing inflammatory markers, we used the fullinformation maximum-likelihood approach in the setting of path analysis over multiple imputation approach for its efficiency. 39 In the Poisson analyses, however, multiple imputation with the Markov chain Monte Carlo method was used, with the number of repetitions similar to a percentage of missing values. 40,41 The variables used to create the imputation model included demographics and the cardiovascular and HIV-related risk factors specified above. Multiple imputation estimates of model parameters were computed by calculating the mean estimates from imputed models, and the variance and 95% CI of these estimates were computed using Rubin's combining formula. 42 All analyses were conducted using the SAS system, version 9.4 (SAS Institute Inc). Two-sided P < .05 indicated significance.

Baseline Characteristics
As shown in Table 1

Traditional Risk Factors Associated With Baseline IMT
We first evaluated factors associated with higher baseline carotid IMT. Cell-associated HIV RNA, HIV DNA, and the RNA:DNA ratio were not significantly associated with baseline carotid IMT in any of the models examined (

Assessment of Factors Associated With IMT Progression
We then determined which factors were associated with greater annual carotid IMT progression  Table 3). However, this association did not remain significant in the demographic-adjusted and multivariable models ( Table 3

Independent Association of HIV RNA and DNA With Incident Plaque
Among individuals who did not have plaque at baseline (

Discussion
Previous research has established that uncontrolled HIV disease is associated with increased cardiovascular risk. However, higher rates of myocardial infarction and atherosclerosis continue to be seen in PLWH in whom HIV is well controlled. Our study sought to explore this phenomenon further by evaluating cardiovascular risk in the setting of treated and suppressed HIV disease using more sensitive assessments of the viral load. In a cohort of PLWH with treated and suppressed disease, we found that levels of cell-associated HIV RNA and DNA were independently associated with incident  Abbreviations: CRP, C-reactive protein; NA, not applicable. a Includes 152 participants. A full-information maximum-likelihood approach was used for missing values.
b All models are adjusted for demographics and additionally included HIV-related factors and markers of inflammation or coagulation that remained significant in the multivariable models. carotid plaque. The association remained significant even after adjustment for traditional risk factors, HIV disease characteristics, and markers of inflammation, such as plasma IL-6 level. To our knowledge, our study is the first to demonstrate an association between the HIV reservoir and CVD in HIV. The independent association between cell-associated HIV RNA and DNA and the development of carotid plaque provides further evidence that HIV infection itself accelerates vascular disease. Although early initiation of ART reduced both AIDS-related and non-AIDS-related events in the START (Strategic Timing of Antiretroviral Treatment) study, the benefit of this strategy on cardiovascular events has not been established. 43 Our findings suggest that therapies that can reduce the size of the reservoir, which could include early ART initiation or HIV curative strategies, may have a beneficial effect on reducing cardiovascular risk-a hypothesis that will need to be evaluated in future clinical trials.

JAMA Network Open | Cardiology
Prior work has demonstrated that HIV infection is associated with higher carotid IMT 35,44,45 and that measures of HIV disease severity, such as lower CD4 count nadir, 35 were associated with higher carotid IMT. In addition to traditional risk factors, inflammatory markers-including high-sensitivity C-reactive protein, 32 expression of certain monocyte markers, 46 sCD163 and sCD14, 47,48 CCR5, 33,49 and plasma IL-6 33 -are associated with higher levels of vascular disease as assessed by carotid IMT.
Our study adds to this literature by demonstrating an association between cell-associated HIV RNA and DNA and incident carotid plaque development in the setting of treated HIV disease.
Interestingly, we found that the measures of viral persistence were associated with incident plaque formation but not annual IMT progression. It is plausible that the underlying pathophysiology of plaque formation may be most susceptible to the inflammatory effects of HIV. Carotid plaque reflects different pathophysiology as opposed to IMT and represents intimal thickening, including foam cells, smooth muscle cells, macrophages, lipid core, and fibrous cap. 50 Several studies 15,51,52 have suggested that carotid plaque may be a more clinically significant measure. A meta-analysis of 16 studies including more than 56 000 individuals 15 showed that mean IMT is significantly associated with cardiovascular events, but IMT progression over time was not. Although there is heterogeneity among studies regarding plaque definition and measurement, another meta-analysis of 11 population-based studies including more than 54 000 individuals 51,52 demonstrated that carotid plaque compared with IMT had significantly higher accuracy for prediction of future cardiovascular events. In the general population, most studies demonstrate that IMT does not provide additional prognostic information compared with traditional risk calculators. In addition, different imaging modalities demonstrate that atherosclerosis in HIV infection is distinct from that of the general population. Persons living with HIV tend to have increased formation of noncalcified plaques that have thin fibroatheroma caps, making them more prone to rupture and CVD-related events, 12,13 along with higher arterial inflammation that is related to inflammatory markers. 11,53 We also found no association between viral persistence measures and baseline carotid IMT. We suspect that the very strong associations between atherosclerosis and traditional risk factors, such

Limitations
Several limitations exist in this study. First, the short duration of follow-up may have limited our ability to demonstrate associations between the viral persistence markers and annual IMT progression. In addition, although IMT was measured at 2 time points, viral persistence parameters and inflammatory markers were only measured at baseline, so we cannot assess the effect of changes in these variables over time. The small sample size of our study also limited the number of candidate covariates that could be included in the multivariable models examining IMT outcomes.
For example, among the inflammatory markers, only IL-6 was included as a candidate covariate owing to its previously demonstrated strong association with carotid IMT. 33 The generalizability of this study among all PLWH, particularly in resource-limited settings, is unknown. This study assesses IMT and plaque in the carotid arteries and not a direct measurement of coronary artery plaque.
Inflammation could be driving both viral persistence and carotid IMT, and further research is needed to elucidate these associations. We also noted that certain traditional risk factors, such as diabetes and use of medications to lower cholesterol levels, were surprisingly not associated with IMT in our models. This outcome may be owing to lack of specific data about the degree of diabetes control and adherence to medications to lower cholesterol levels in the cohort. In addition, about 20% of the study population was missing data on inflammatory markers, although in our analyses of the missing data, these data appeared to be missing at random and were imputed for the multivariable analyses.
Finally, our measures of viral persistence have several important limitations. Most HIV DNA measured by quantitative polymerase chain reaction is defective and does not represent replicationcompetent virus, 23,64 although defective provirus may contribute to continued inflammation and pathology in individuals receiving ART. 65 In addition, HIV resides mainly in lymphoid tissue beyond the peripheral circulation, 66 which would require more invasive sampling to access. Nevertheless, quantitative measures of HIV DNA and RNA from peripheral blood CD4 + T cells, as used in our study, remain important and are commonly used to quantify viral persistence. Future studies ought to assess the effects of intact vs nonintact cell-associated DNA and the effect of the lymphoid viral reservoir on cardiovascular outcomes.