Risk of All-Cause Mortality in HIV Infected Patients Is Associated with Clinical, Immunologic Predictors and the CCR5 Δ32 Deletion

Objective Investigation of the interplay between the CCR5 Δ32/wt genotype and demographic, epidemiological, clinical and immunological factors associated with mortality in the cART era. Design Longitudinal data from 507 HIV-infected patients following the Δ32 allele detection were analyzed. Methods Cumulative 15 years mortality was calculated using Kaplan-Meyer methodology. Hazard ratios were estimated using univariate Cox models. Basing on Akakie information criteria and statistical significance multivariate Cox model was constructed and effect plots presenting adjusted hazard ratio time-dependency were drawn. Analysis of the association of all-cause mortality and CCR5 Δ32/wt genotype prior to the antiretroviral treatment (cART) initiation (n = 507) and on the therapy (n = 422) was also performed. Results A mortality rate of 2.66 (CI 2.57–3.19) per 100 person-years was observed. Univariate analysis factors modifying the risk of death included the CCR5 genotype, gender, history of cART, AIDS diagnosis and also CD4 lymphocyte nadir, zenith, the latest CD4 count and stable levels >500 cells/µl. For multivariate analysis the following predictors were selected: CCR5 genotype (HR for wt/wt 2.53, CI 1.16–5.53, p = 0.02), gender (HR for males 1.91, 95%CI 1.1–3.36, p = 0.023), introduction of combined antiretroviral treatment (HR 4.85, CI 3.0–7.89, if untreated or treated <1 month, p<0.0001) CD4 count of 500 cells/µl for six months or more (HR 4.16, CI 1.95–8.88 if not achieved, p = 0.028), the latest CD4 count (HR 5.44, CI 3.39–8.74 for <100 cells/µl, p<0.0001) and history of AIDS (HR 1.69, CI 1.03–2.79, p = 0.039). Among untreated individuals the Δ32/wt genotype was associated with notably better survival (p = 0.026), while among cART treated individuals the Δ32 mutation did not correlate significantly with higher survival rates (p = 0.23). Conclusions The Δ32 CCR5 allele is associated with a reduction of the risk of all-cause mortality in HIV (+) patients alongside clinical and immunologic predictors such as AIDS, history of cART, lymphocyte CD4 cell count and gender.


Introduction
Infection with the human immunodeficiency virus type 1 (HIV-1) requires attachment to one of the principal chemokine coreceptors, namely CCR5 and CXCR4 for effective entry into CD4+ T-cells [1]. Use of CCR5 is frequently associated with the early stages of infection while progression of human immunodeficiency virus infection to AIDS and death is related to a viral tropism switch to CXCR4 [2]. Genetic variants of the chemokine receptors and their ligands modify susceptibility to HIV infection and the course of the disease, with a 32 base pair deletion (D32) in the reading frame of the CCR5 gene associated with reduced susceptibility to infection and delayed disease progression [3,4]. D32/D32 homozygotes remain resistant to infection with CCR5 tropic variants of HIV, while among heterozygous D32/wild-type (wt) subjects beneficial, AIDS protective effects, have been noted in the early years of infection [5,6]. Studies of the long-term nonprogression among HIV-infected patients have indicated higher frequency of this protective allele among individuals spontaneously controlling the infection [7]. This variant has also been associated with better virologic response to combined antiretroviral therapy (cART) as well as less frequent virologic failure, decreased risk of early death among perinatally infected children and protection from AIDS among antiretroviral-treated patients [8][9][10][11].
In the era of combined antiretroviral treatment (cART) and expected long term survival an array of interplaying factors influence mortality among people living with HIV. Lymphocyte CD4 metrics, early introduction of antiretroviral treatment and preservation of immunological function are the key contributors to the risk of mortality in this group [12,13]. As the interaction between clinical, immunological, virologic factors and genetic variants influencing survival have not been studied so far in detail this study aimed to evaluate the impact of CCR5 D32 mutation on the long-term mortality of patients living with HIV. Characteristically in Europe, as well as in Poland, most HIV cases are diagnosed late necessitating immediate initiation of combined antiretroviral treatment [14,15]. Therefore the rationale for this analysis was that beneficial effects of the mutation are likely to be disproportionately seen in this population due to a protective effect prior to diagnosis, enhanced by a favorable effect following initiation of the antiretroviral treatment.

Study population
For the study longitudinal data of 507 patients followed-up from January 1996 to June 2010 at the Department of Infectious Diseases and Hepatology, Pomeranian Medical University, Szczecin, Poland and Out Patients' Clinic for Acquired Immunodeficiency, Regional Hospital, Szczecin, Poland were analyzed. The study protocol was approved by the bioethical committee of Pomeranian Medical University, Szczecin, Poland (approval number BN-001/34/04). Written informed consent was obtained from all subjects participating in the study. Time zero was defined as a date of positive screening HIV test if later confirmed by Western-blot, immunoblotting or positive serum HIV-RNA or a confirmation test itself. The following data were collected: age, gender, date of HIV diagnosis, route of transmission, hepatitis C co-infection, clinical category at diagnosis according to CDC case definition [16], date and diagnosis of AIDS events, date and reason of death, baseline HIV viral load, history of cART as well as baseline, nadir, zenith and the latest CD4 counts, duration of CD4 count .500 cells/ml. Baseline CD4 counts are defined as the first documented result after diagnosis of HIV. Data on nadir and zenith (the lowest and the highest lymphocyte CD4 count throughout the period of observation) lymphocyte CD4 counts were collected as well. The latest lymphocyte CD4 count was taken as the last recorded value prior to the end of observation or death. CDC category at diagnosis was assumed based on the review of the clinical record of the patient, in cases of late care entry with documented prior HIV test category A (asymptomatic) was assumed if no apparent immunodeficiency was reported or available from medical records. Data on chronic hepatitis B were not included into the analysis due to small number of confirmed HIV/HBV co-infection cases while the parameter of the latest HIV-RNA levels was removed as it was performed only in 60% of patients, often being collected up to one year from the date of final observation, which was related to the poor availability of the assay.

Study endpoint
Study endpoint was defined as all-cause mortality excluding cases with documented accidental death. Determination of the reason of death was based on the following data: autopsy report (42 cases), medical record of in-hospital treatment with cause of death defined by the treating physician (32 cases), other medical report or letter (19 cases); in four cases the cause of death remained undetermined. Validation of the underlying cause of death was performed by an independent clinician not involved in the patient care with discrepancies in the data thoroughly discussed. AIDS related death was assumed if the criteria (either presumptive or definitive) outlined by the CDC case definition were met with at least one AIDS defining condition observed at the time of death. End of observation date was defined as either death date, last recorded date of visit (cases lost to follow-up) or 1 st of June 2010 for the patients remaining under care (termination of data collection). When it was impossible to determine the exact date of death (6 cases) the median date between the last recorded visit and information on death was assumed as the death date.

DNA extraction and CCR5 genotyping
For genomic DNA extraction from whole blood samples QIAamp DNA Blood Mini Kit (QIAgen, Hilden, Germany) was used with re-suspended DNA stored at 4uC for further analyses. If whole blood samples were unavailable frozen serum samples were used with extraction performed using Sherlock AX kits (A&A Biotechnology, Gdynia, Poland). For the CCR5 D32/wt genotyping a previously described methodology was used [17] with reaction products electrophoresed on a 3% agarose gel (SIGMA, Saint Louis, USA) stained with DNA-star dye (Lonza Inc, Rockland, USA) and visualized under UV light.

Statistics
Statistical analyses for nominal variables were performed with chi-square test using EPI6 Statcalc software (Department of Mathematics, University of Louisiana-Lafayette, Lafayette, LA, USA), for continuous variables, the Mann-Whitney U-test was used (Statistica software, Statsoft, Tulsa, OK, USA). Cox models for survival were used to assess the effect of analyzed parameter on the risk of death. For all analyzed parameters unadjusted (univariate) Cox models were used to calculate the hazard ratio (HR) and included CCR5 genotype, transmission route, hepatitis C co-infection, HIV infection stage at diagnosis, history of antiretroviral treatment, baseline viral load (log HIV-RNA copies/ml), baseline CD4 count (cells/ml), nadir CD4 count (cells/ml), zenith CD4 count (cells/ml), time with CD4 count .500 cells/ml, the most recent CD4 count (cells/ml) and history of AIDS. Moreover, for every parameter Kaplan-Meyer cumulative mortality was calculated and plotted with statistical significance testing using log-rank test. To select the most informative parameters for the multivariate model Akaike information criteria (AIC) were used to assess the fit of the each model (lower AIC indicates better fit) [18]. A final multivariate model was calculated by Cox regression for the 15-year observation and effect plots presenting adjusted hazard ratio changes over time were drawn for this model. To test the proportional hazards assumption for the multivariate Cox regression model coxph function of the R package was used. P-values of ,0.05 were considered significant with values in the range of 0.1-0.05 regarded as borderline statistically significant.
Notable differences in the frequency of the CCR5 D32 genotype were observed for baseline CDC category (CCR5 D32/wt genotype frequency among individuals diagnosed at asymptomatic stage of infection was 22.0% compared to 10.2% among symptomatic and AIDS patients, p,0.003) and overall number of deaths (the CCR5 D32/wt genotype was observed in 7.5% cases compared to 15.6% of patients surviving throughout the period of observation, p = 0.046). Similar CCR5 D32/wt genotype frequency was observed among patients with AIDS related deaths (7.6%) while in the group with non-AIDS related death the frequency of the CCR5 D32/wt genotype was lower (5.0%, p = 0.073 if compared to the entire surviving group). Other differences with borderline significance in the CCR5 D32/wt genotype distribution included transmission route (16.1%, 11.8%, and 14.2% for homosexual, heterosexual and injection drug use (IDU) related transmissions, respectively, p = 0.07), female vs. male gender (9.7% and 15.7%, respectively, p = 0.08) and initiation of antiretroviral treatment (20.0% among treatment naive individuals vs. 13.0% among cART treated patients, p = 0.09). Majority of women with the D32 allele were infected sexually (n = 9, 64.2%), while among men predominated IDU (n = 32, 56.1%) over sexual transmissions (homosexual route was observed in 18 (31.6%) while heterosexual in 7 cases (12.3%) (p = 0.17). No significant difference between the groups with and without the D32 allele in the overall number of AIDS diagnoses was observed (the CCR5 D32/wt genotype noted in     1 and 2). In seven patients who died (7.5%) the CCR5 wt/D32 genotype was noted, in majority of these patients (n = 5) AIDS-defining conditions were observed at the time of death while two deaths were non-AIDS related. There were no significant differences for the reasons of death between the groups with and without the CCR5 D32 allele. Table 2 shows univariate hazard ratios calculated with Cox regression and Kaplan-Meyer cumulative mortality for all parameters investigated in the study. Additionally, for every parameter included, Akaike information criterion (AIC) is presented to indicate the fitness of association with mortality. Identified predictors associated with a significantly greater mortality included the CCR5 genotype, male gender, symptomatic HIV infection at diagnosis, antiretroviral treatment history and AIDS diagnosis (figure 3a-e). The analysis also included a range of lymphocyte CD4 metrics, namely baseline, nadir, zenith and the most recent CD4 count and time of observation with CD4 above 500 cells/ml. Baseline, nadir and the most recent CD4 counts were further categorized using 100, 50 and 25 cells/ml thresholds; for zenith and time of observation the 500 cells/ml threshold was used. This category was also added in the calculations for the latest CD4 count. The strongest association between the baseline and nadir CD4 counts and mortality was noted for the 50 cells/ml category for baseline and nadir CD4 count (figure 3f-g) as well as CD4 zenith ,500 cells/ ml (figure 3h) and for patients with the most recent CD4 counts .500 cells/ml (figure 3i). It must be noted that Akaike information criterion for the most recent CD4 was the lowest for 100 cells/ml threshold therefore this value was selected for the multivariate analysis. Two categories were analyzed for time with CD4 count .500 cells/ml vs. patients who have never maintained stably high CD4 counts, namely six and twelve months, respectively (figures 3k,3l). No statistically significant differences were obtained for AIDS and non-AIDS related deaths analyzed separately.

Multivariate survival model with selected parameters
Basing on the AIC and statistical significance for each parameter a multivariate Cox model was constructed. In the initial model all parameters presented in  figure 4). Time-dependent HR plots adjusted for each of the six parameters were drawn to show the effect on survival (figure 5a-f).

Association between CCR5 D32 genotype and survival in treated and untreated individuals
To assess the influence of the CCR5 D32 genotype prior to the cART introduction two additional analyses were performed. The first one for the untreated individuals (n = 507 including 71 individuals with the CCR5 D32/wt genotype), starting from the HIV diagnosis and censored at the time of the cART initiation, loss to follow-up or death (figure 6a). The second one was based on cART treated individuals (n = 422, 55 individuals with the CCR5  All-Cause Mortality and the CCR5 D32 Deletion D32/wt, starting from the date of treatment initiation with censoring at the time of termination of data collection, loss to follow-up or death (figure 6b). Among untreated individuals the CCR5 D32/wt genotype was associated with better survival (p = 0.026), however among cART treated individuals the D32 mutation did not correlate significantly with higher survival rates (p = 0.23).

Discussion
This study investigates the combined effect of selected parameters and the CCR5 D32 allele on long term survival among HIV-1 infected patients in the era of cART. An array of factors influencing survival were analyzed in this study including demographics as well as clinical and immunological parameters. The frequency of the CCR5 D32/wt heterozygotes in this study was 14.0% and is consistent to the previous reports for Caucasian populations, while use of archived samples allowed to avoid the bias related to the survival benefit of the mutation in HIV infected patients who may have died [8,19,20]. The near-significant difference between the male and female gender frequency might be related to the transmission route -higher number of women was infected sexually. It is possible that the risk of HIV transmission is differently modified by the D32 allele for every gender, depending on the type of exposure.
An association between the D32 allele and delayed progression to AIDS has previously been reported including a suggestion that this deletion is the strongest protective factor for both viral and immununologic control of the HIV-associated disease and improved survival outside the HLA system [10,21]. In the study by Hendrickson et al. [11], both time to viral suppression and AIDS free survival following cART initiation was significantly longer in the group with the D32 allele. In the meta-analysis by Ioannidis et al., the relative hazard for both AIDS and death was notably lower among the CCR5 D32/wt heterozygotes if compared to the individuals homozygous for the wild-type variant [22]. The  beneficial effect of the CCR5 D32/wt genotype was also found by Brumme et al. [23], who observed improved survival in the median .5 year period after starting the antiretroviral therapy in the univariate analysis adjusted for baseline age, lymphocyte CD4 count, viral load and AIDS defining conditions. In our study the number of AIDS-free individuals with the CCR5 D32 variant was significantly higher at the time of HIV diagnosis, however for the entire follow-up period the number of AIDS cases was not notably different between the groups with and without this allele. This could be related to the fact that the cohort used for this analysis consisted of a large number of late diagnosed individuals, in 372 (73.4%) cases infection was diagnosed in either the symptomatic stage or with baseline CD4 count,350 cells/ml, which is characteristic for HIV infections diagnosed in Poland   [24]. The protective effect of this deletion may have been lost due to the time-dependency of the beneficial effect, as reported before [25,26]. Similarly, the slightly higher prevalence of the D32 allele among antiretroviral naive patients is probably related to a higher number of patients asymptomatic at baseline who would not require cART. Other differences of borderline statistical significance for the CCR5 D32 allele frequency include a lower frequency among women, confirming the finding by Philpott et al., of a more extensive protective effect of the CCR5 D32/wt heterozygous genotype in women [27] while a lower prevalence of the heterozygotes among heterosexually infected patients might be related to the fact that high CCR5 levels are required for these primarily macrophage-tropic sexual infections [28,29].
In the univariate analysis of all the factors significantly modifying all-cause mortality the strongest protective effect was observed for cART as well as CD4 metrics, namely nadir, zenith lymphocyte CD4, time with CD4 count above 500 cells/ml and AIDS-free observation. Effects of these parameters were extensively studied before and remain well documented [30][31][32][33]. These influences might be additionally modified by the D32 allele, found also to be a statistically significant protective influence on mortality, as it has been reported that in individuals bearing this variant time of progression to AIDS after initiation of cART as well as time from AIDS diagnosis to death is increased [25,11]. In our study the D32 allele was associated with decreased mortality risk observed for as long as 15 years of follow-up. This association remained significant after adjustment for other factors strongly influencing mortality: gender, nadir and zenith CD4 count, time with CD4 .500 cells/ml, most recent CD4 levels and AIDS diagnosis.
It must be noted, that the CCR5 D32/wt genotype was associated with better survival in untreated individuals only, which is probably associated to the slower development of the symptomatic disease as suggested by Brumme et al., [34] while among cART treated individuals the correlation with better survival was insignificant, however this analysis was performed on smaller number of patients (55 with the CCR5 D32/wt genotype with 23 on cART for 5 years and 6 for more than 10 years). No significant effect of the D32 allele, survival and AIDS free survival was also found by Laurichesse et al. [8]. Moreover, Brumme et al., did not find the association between the CCR5 D32/wt genotype and time to virological failure in the study of 436 individuals observed for the median follow-up of 22 months [22]. Similarly, Bratt et al, in the study of 147 patients, which included similar (37%) percentage of AIDS diagnosed patients as our analysis, observed little influence of the CCR5 D32/wt genotype on virologic treatment efficacy [35]. Lack of beneficial effect of the D32 allele among antiretroviral treated patients observed in our study might be associated not only to the size of the group, but also differences in the time from the infection to treatment initiation, which is related to late diagnosis of HIV and high prevalence of AIDS at the time of diagnosis. However, it must be noted that the D32 allele has been associated with the decrease in the likelihood of viral suppression failures, defined as HIV-RNA $200 copies/ml at 16-28 weeks of therapy, accelerated viral suppression to ,200 HIV-RNA copies/ml, as well as significantly better virologic response to antiretrovirals both at 6 and 12 months following treatment initiation [8,11,36]. Moreover, a higher rate of the sustained virologic suppression during the observation period of up to 5 years of antiretroviral treatment in the CCR5 D32/wt heterozygous patients was noted [9].
In adjusted multivariate analyses both AIDS diagnosis and antiretroviral treatment remained significant predictors of mortality. Moreover, it was confirmed that of all the CD4 metrics studied, the latest CD4 count is the strongest predictor of mortality, as has been suggested before [37]. When the timedependency for the six factors included into the multivariate model was studied, a continuous significant positive effect, regardless the period of observation both for this variable and history of cART, was observed. This confirms the value of these parameters and emphasizes the need for early introduction of the antiretroviral treatment. Stable CD4 count of .500 cells/ml for at least 180 days became a significant parameter after 30 months of observation, which is probably related to the time dependent immune recovery. Interestingly, the continuous protective effect of the CCR5 D32/wt genotype was observed over 36 months of follow-up with stable HR of 2.5-2.6 noted up to 96 months, the end of observation period. This suggests that in the cART era the protective effect of this variant on survival is not lost over time as was observed for the D32 allele and AIDS diagnosis. Hazard ratio effect plots for gender and diagnosis of AIDS proved time-dependent, with female gender associated with significant protection up to 108 month of observation while for AIDS two periods associated with favorable prognosis in AIDS free patients were noted, namely from months 36 to 54 and 108 to 180 months. The reason for this time dependent relationship remains to be elucidated.
Studies analyzing the impact of the CCR5 D32 deletion on long term survival among HIV (+) patients in the cART era remain sparse, especially with implementation of Cox regression models adjusted for clinical variables significantly contributing to the mortality risk.
A potential limitation of the study was use of interval censoring for six cases when the exact date was unknown, with the death date fixed as the median date between the last observation time and information that the patients actually died. Another limitation is related to the lack of possibility to observe the patients from the time of seroconversion, however in the setting of the late testing and referral of majority of cases with symptomatic immunodeficiency such a study was not feasible. Another limitation is related to the lack of information on adherence, therefore the results could not be controlled for this factor.
Host genetics strongly interacts with HIV leading to the modification of the course of infection. Our finding of the beneficial effect of the CCR5 D32 on long term survival in conjunction with well established factors influencing mortality provides an insight into the interplay between genetic and virus/ host-related factors. Such an effect is of the greatest importance in the late-tested populations, such as the one in Poland, allowing for the longer period of delay from infection to disease detection and treatment initiation, whereas it is of the lesser importance in the cART treated individuals. Testing for the CCR5 D32 mutation in clinical practice might be useful for elucidation of the reason for delayed progression of HIV infection, however it should probably not influence decision on antiretroviral treatment initiation or modification.