Influence of polymorphisms within the CX3CR1 and MDR-1 genes on initial antiretroviral therapy response

Objective: Single nucleotide polymorphisms (SNP) in the genes encoding the human CX₃CR1 chemokine receptor and the P-glycoprotein multidrug transporter have been associated with accelerated disease progression in untreated individuals and implicated in therapeutic response, respectively. This retrospective study assessed the influence of SNP in the CX₃CR1 and MDR-1 genes on initial virological and immunological response in 461 HIV-infected, antiretroviral-naive individuals initiating antiretroviral therapy in British Columbia, Canada.

Methods: CX₃CR1 and MDR-1 SNP were determined by PCR amplification of human DNA from plasma, followed by DNA sequencing. Time to virological success [time to HIV plasma viral load (pVL) ≤ 500 copies/ml], virological failure (subsequent time to the second of two consecutive pVL ≥ 500) and immunological failure (time to the second consecutive CD4 cell count below baseline) were analyzed by Kaplan-Meier methods.

Results: Frequencies of CX₃CR1 amino acid haplotypes were 249V 280T (0.75), 249I 280M (0.15), and 249I 280T (0.1). Frequencies of MDR-1 nucleotide polymorphisms were 3435C (0.47) and 3435T (0.53). There was no effect detected for SNP in CX₃CR1 or MDR-1 on time to virological success, nor of CX₃CR1 and MDR-1 SNP on time to virological and immunological failure, respectively (P > 0.1). There was a trend to earlier virological failure in the MDR-1 3435C/C genotype group (P = 0.07), and a statistically significant trend to earlier immunological failure in individuals with the CX₃CR1 249I polymorphism (P = 0.02). These remained significant after correcting for baseline age, sex, pVL, CD4 cell count, type of therapy, and adherence (P ≤ 0.05).

Conclusion: Polymorphisms in MDR-1 and CX₃CR1 may be associated with accelerated virological and immunological therapy failure, respectively.

Introduction

Many recent studies have focused on the influence of naturally occurring polymorphisms within the human genome on HIV-1 infection and clinical disease progression. Probably the most notable polymorphism lies within the gene encoding the human chemokine receptor CCR5, a key cell-surface co-receptor for R5 strains of HIV-1 [^1-3]. A homozygous 32-base pair deletion in the coding region of CCR5 (CCR5Δ32/Δ32) has been associated with a protective effect with respect to HIV-1 infection [^4,5], while the heterozygous genotype (CCR5wt/Δ32) has been associated with slower HIV-1 disease progression [^4,6-8] in untreated individuals.

CX₃CR1, a leukocyte chemotactic and adhesion receptor for the human chemokine fractalkine, has also been defined as a minor co-receptor for HIV-1 [^9,10]. Two common single nucleotide polymorphisms (SNP) within the human CX₃CR1 gene have recently been identified [11]. HIV-1 infected patients homozygous for the CX₃CR1 249I and 280M amino acid substitutions (a G-to-A nucleotide substitution encoding a V-to-I amino acid substitution at position 249, and a C-to-T nucleotide substitution encoding a T-to-M amino acid substitution at position 280) were shown to progress to AIDS and death more rapidly than those with other haplotypes [¹¹]. These results were not confirmed in an independent study which examined the polymorphism at codon 280 only [¹²]. Although the influence of CCR5 polymorphisms on response to antiretroviral therapy has been examined [^13-16], to date, no published studies appear to have examined whether polymorphisms in the CX₃CR1 gene are associated with altered response to antiretroviral therapy.

SNP in other genes may also affect response to antiretroviral therapy more directly. The human multidrug-resistance (MDR)-1 gene encodes the ATP- dependent P-glycoprotein (P-gp) efflux membrane transporter which possesses a broad substrate specificity [^17,18] and therefore the ability to export a wide range of hydrophobic molecules from cells [^19,20]. The HIV-1 protease inhibitors (PI) are substrates for the P-gp drug efflux pump in vitro [^21-23], a finding which could potentially affect the bioavailability of HIV-1 PI and limit the therapeutic efficacy of PI in P-gp expressing cells [²⁴], including CD4 cells [²⁵]. A single C-to-T nucleotide polymorphism in exon 26 (C3435T) of the MDR-1 gene is significantly correlated with decreased intestinal expression levels and activity of P-gp [²⁶]. In other disease models, mainly cancer, the level of MDR-1 expression is correlated with the resistance of tumor cells against chemotherapeutic agents [^27,28]. It has also been found that over-expression of multidrug transporters significantly reduces the accumulation of PI in human lymphocytes [²⁴], and that individuals homozygous for the 3435C allele exhibited increased efflux of the P-gp substrate rhodamine 123 from CD56 natural killer cells [²⁹]. Based on these findings and evidence of racial differences in 3435C allele distribution, it was hypothesized that differential P-gp expression could affect the utility of PI in African populations [³⁰]. Finally, data have confirmed a correlation between initial CD4 response, nelfinavir and efavirenz levels, and MDR-1 polymorphisms in adherent patients [³¹].

To address whether previously identified SNP in CX₃CR1 or MDR-1 affect response to antiretroviral therapy in an unselected patient population, we retrospectively determined the prevalence of the CX₃CR1 249I and 280M amino acid substitutions, and the MDR-1 C3435T polymorphisms in a group of 461 antiretroviral-naive individuals initiating their first HIV therapy in British Columbia, Canada, between June 1996 and August 1998, and assessed the potential effects of these natural human polymorphisms on the initial virological and immunological response to antiretroviral therapy.

Materials and methods

Study design and patients

In the province of British Columbia, Canada, antiretroviral drugs are distributed free of charge to HIV-infected individuals through a centralized drug treatment program based at the British Columbia Centre for Excellence in HIV/AIDS (the B.C. Centre). Briefly, B.C. treatment guidelines recommend that plasma viral load testing be performed at baseline, again at 4 weeks after initiating or changing therapy, and approximately every 3 months thereafter (see [³²] for more details of policies and treatment guidelines). The median time between plasma viral load measurements for subjects included in this study was 3 months (interquartile range, 2-4 months). Less than 5% of sequential viral load measurements had an elapsed time of ≥ 8 months. All antiretroviral-naive individuals first seeking treatment at the B.C. Centre between 1 June 1 1996 and 31 August 1998, were eligible for analysis in this retrospective study (n = 479). Of these, 461 (96.2%) had a pre-therapy plasma sample available for testing. More than 80% of the antiretroviral-naive patients were prescribed triple therapy including two nucleoside analogues and one PI, consistent with treatment guidelines existing at the time. Adherence to medication was evaluated by analyzing patients' prescription records and determining the frequency of prescriptions filled on time during the first year of therapy [^33,34], with 'adherence' being defined here as filling 100% of prescriptions on schedule.

Ethical considerations

Ethical approval for this study was obtained from the local ethics board at St. Paul's Hospital, Vancouver. All patient samples were assigned new anonymous code identifiers prior to genotypic analyses. Laboratory analyses were performed using the coded identifiers only.

Plasma viral load and CD4 Cell Count

CD4 cell counts were measured by standard techniques. Plasma HIV-1 RNA levels (viral loads) were determined using the Roche Amplicor Monitor assay (Roche Diagnostics, Laval, Quebec, Canada) using either the standard method or the ultrasensitive adaptation. The ultrasensitive adaptation to this assay was not routinely available over the time period of this study; for this reason a constant cut-off of 500 HIV RNA copies/ml was used throughout the analysis.

Determination of CX₃CR1, MDR-1, CCR5 Δ32 polymorphisms, and HIV-RNA genotyping for drug resistance

Plasma nucleic acids were extracted and precipitated with guanidine thiocyanate and isopropanol followed by two ethanol washes. The region of CX₃CR1 containing the polymorphisms of interest was amplified in a single round of PCR, using the oligonucleotide primers CX₃CR1 F1 (5′-CTGAATCAGTGACA GAAAACTTT-3′) and CX₃CR1 R1 (5′-GTAGACA CAAGGCTTTGGGATTC-3′), to generate a 1.1-kb product. In the case of the MDR-1 polymorphism in exon 26, one round of PCR using oligonucleotide primers MDR-1 F1 (5′-CAATTATGACCTTGT TGGGTTAA-3′) and MDR-1 R1 (5′-AGATGCTT GTATACAGGTAAGG-3′) was performed, to generate a 0.5-kb product. PCR products were then sequenced directly in both the 5′ and 3′ directions using these primers and the BigDye dye terminator cycle sequencing kit (Applied Biosystems, Foster City, CA) and resolved on an ABI 3700 automated sequencer. CCR5Δ32 status [¹³] and HIV-1 drug-resistance genotyping [³⁵] were performed as previously described.

Statistical analysis

Time to virological success was defined as the time to achieve a plasma viral load (pVL) ≤ 500 copies HIV-1 RNA/ml from baseline. Time to virological failure was defined as the time elapsed between virological success and the second of two consecutive pVL measurements > 500 copies/ml. If a person never achieved virological success, they were excluded from the analysis of time to virological failure. If a patient reached therapeutic success and did not subsequently fail then they were censored at the time of their last viral load date (occurring prior to 1 December 2000). Time to immunological failure was defined as the time to the second of two successive CD4 cell counts below baseline. The effects of the CX₃CR1, MDR-1, and combined CX₃CR1 and CCR5 polymorphisms on time to virological and immunological success and/or failure were analyzed by Kaplan-Meier methods. Group differences were assessed using the log-rank test. Cox proportional hazard regression was used to model the effect of the CX₃CR1 and MDR-1 polymorphisms on time to virological and immunological failure while adjusting for age, sex, baseline pVL, baseline CD4 cell count, adherence, type of antiretroviral therapy at initiation (triple therapy with two nucleoside reverse transcriptase inhibitors and a PI versus dual therapy), and 'time to virological success' in the virological failure model only. Finally, Fisher's exact test was used to assess the potential correlation between MDR-1 genotype and the development of HIV-1 resistance to PI in the first 2 years of therapy (an analysis restricted to those patients initiating therapy including a PI).

All of the 461 patients in the study were included in the 'time-to-virological success' analyses. In order to be included in the 'time-to-virological failure' analysis, patients had to register at least one post-baseline HIV pVL ≤ 500 copies/ml; this included 374 of 461 individuals (81.1%). Of 461 individuals 382 (82.9%) had at least one post-baseline CD4 cell count, and were therefore included in the time-to-immunological- failure analysis.

Results

Baseline characteristics of the 461 HIV-1 infected individuals examined for CX₃CR1 249I and 280M amino acid substitutions and the MDR-1 C3435T polymorphisms are presented in Table 1. There were no significant differences with respect to age, sex, baseline plasma viral load, CD4 cell count, AIDS defining-illness or history of injecting drug use between the 479 patients eligible for the study and the 461 (96.2%) patients for whom plasma samples were available for testing. However, patients in the study group (n = 461) were more likely to have been prescribed triple therapy (P = 0.03). There were no statistically significant differences observed between any baseline parameter and presence or absence of SNP (P > 0.1, data not shown). The median follow up time after starting therapy was 40 months.

Table 1

PCR products and CX₃CR1 genotypes were obtained for a total of 451 of 461 individuals (97.8%). At CX₃CR1 codon 249, 282 individuals (62.5%) were homozygous G/G (encoding amino acid 249V), 54 individuals (12.0%) were homozygous A/A (encoding amino acid 249I), and the remaining 115 (25.5%) were heterozygous. At CX₃CR1 codon 280, 339 (75.2%) were homozygous C/C (encoding amino acid 280T), 27 (6.0%) were homozygous T/T (encoding amino acid 280M), and the remaining 85 (18.8%) were C/T heterozygous. The CX₃CR1 249G allele (encoding amino acid 249V) was in complete linkage disequilibrium with the 280C allele (encoding amino acid 280T) [¹¹]. PCR products and MDR-1 genotypes were obtained for 455 of 461 individuals (98.7%). A total of 137 (30.1%) individuals were 3435 C/C, 165 (36.3%) were 3435 T/T, and the remaining 153 (33.6%) were 3435 C/T. As described elsewhere, patients were also genotyped for the CCR5Δ32 [¹³]. The allele and haplotype frequencies for all polymorphisms in the CX₃CR1, MDR-1 and CCR5 genes are summarized in Table 2. There was no genetic linkage detected between any of the polymorphic groups (with the exception of the linkage observed within the CX₃CR1 gene).

Table 2

Kaplan-Meier analysis of the influence of MDR-1 genotype on time elapsed to virological failure (defined as the time elapsed to the second of two successive pVL > 500 copies/ml from initial success) is shown in Fig. 1. The proportion of individuals not achieving virological success was between 18-20% for all three groups. There was a trend to earlier virological failure in the homozygous 3435C/C group which did not attain statistical significance (P = 0.07). No effect of the 3435 C/T nucleotide polymorphism in the MDR-1 gene on immunological failure, defined as the time elapsed to the second of two consecutive CD4 cell counts below baseline, was detected (P = 0.46, data not shown). When only the homozygous groups (3435C/C and 3435T/T) were considered in the analysis, the correlation between 3435C/C and earlier virological failure reached statistical significance (P = 0.046), but no effect was detected on immunological failure (P = 0.97). In a secondary Kaplan-Meier analysis, there was no detectable influence of MDR-1 genotype on the time elapsed to virological success (time to HIV pVL ≤ 500 copies/ml) (P = 0.371, data not shown).

Fig. 1

In addition, we wished to determine whether the polymorphism in MDR-1 could be correlated with the development of resistance to HIV PI [²⁴]. A search of our HIV-1 genotype database identified a subset of 124 (of 461) individuals (29.5%) who had at least one HIV-1 antiretroviral resistance genotype requested by a physician performed during the course of study follow-up, at least 6 months after initiation of therapy. Thirteen of 124 patients (10.5%) displayed decreased susceptibility to one or more PI in at least one post-therapy genotype using a 'virtual' phenotype analysis [³⁵] where 'decreased susceptibility' was defined as a > 3-, 3.5-, 4-, 2.5-, 2-, or 2.5-fold decrease in susceptibility to indinavir, ritonavir, nelfinavir, saquinavir, amprenavir and lopinavir, respectively [³⁶]. Samples for which genotypes were available revealed no correlation between the MDR-1 polymorphism and development of resistance to PI using the Fisher's exact test, in an analysis restricted to those patients initiating therapy including a PI (n = 83; P = 0.99). PI resistance was evenly distributed between the MDR-1 3435 C/C, T/T, and C/T genotype groups.

For the Kaplan-Meier analyses of the influence of the CX₃CR1 genotype on time elapsed to virological and immunological failure, patients were classified into categories based on their amino acid polymorphisms at codons 249 and 280. The first analysis included homozygous amino acid combinations only (CX₃CR1 genotypes 249I 280T, 249I 280M and 249V 280T). The proportion of individuals censored in the time to virological failure analysis was similar for all three groups. No significant influence was observed between the three groups with respect to time elapsed to virological failure (P = 0.99; data not shown), however, a tendency to earlier immunological failure was observed in the two groups containing the 249I polymorphism which achieved statistical significance (P = 0.02; Fig. 2). When heterozygous allele combinations were included in the analysis, the effect on virological outcome remained not statistically significant (P = 0.92; data not shown), and a trend for earlier time to immunological failure for the two groups containing the homozygous 249I polymorphism remained (P = 0.07; data not shown). Results for the heterozygous allele groups fell near those of the homozygous 249V 280T group. Finally, when both homozygous and heterozygous allele combinations at codons 249 (I, I/V, V) and 280 (T, T/M, M) were analyzed independently (n = 451), no effects of variation at codons 249 or 280 on virological outcome could be demonstrated (P = 0.78 and P = 0.52, respectively). Similarly, no effect of variation at codon 280 alone [¹²] on immunological outcome could be demonstrated (P = 0.167). However, a statistically significant trend for earlier immunological failure in the homozygous 249I group was observed (P = 0.036; data not shown). In a secondary analysis, no influence of CX₃CR1 genotype on the time elapsed to virological success was observed (P = 0.89; data not shown).

Fig. 2

No significant association between the CX₃CR1 249I or 249V genotypes and CCR5Δ32 status was observed (n = 446; P > 0.1; data not shown). No effect of the combined CX₃CR1 and CCR5 genotypes on time to virological success or virological or immunological failure was observed (P > 0.6; data not shown).

Cox proportional hazard regression was used to model the effect of the CX₃CR1 and MDR-1 polymorphisms on time to virological and immunological failure while adjusting for age, sex, baseline pVL, baseline CD4 cell count, adherence, type of antiretroviral therapy at initiation, and 'time to virological success' in the virological failure model only (Table 3). The results were consistent with the previous analyses. The MDR-1 virological analysis reached statistical significance (P = 0.04) [risk ratios (95% confidence intervals) were 0.67 (0.45-0.99) for the MDR-1 3435 T/T and 0.71 (0.47-1.06) for the MDR-1 3435 C/T (Y) genotypes], and results of the CX₃CR1 immunological analysis also remained significant (P = 0.05) [risk ratios were 1.92 (1.00-3.69) for the CX₃CR1 249I 280T and 1.46 (0.80-2.70) for the 249I 240M genotypes]. Of interest, the only baseline variable that was consistently protective in the multivariate analyses was '100% adherence' (P = 0.0001; Table 3), which was evaluated using prescription refill records [^33,34], with 'adherence' being defined as filling 100% of prescriptions on schedule.

Table 3

In order to control for non-adherence, a secondary subset analysis was performed on only those individuals who were most likely to be highly adherent to therapy based on the above definition (240 of 461 subjects, or 52.1%). The adherence subset analyses were consistent with previous results (P > 0.05 for CX₃CR1 and MDR-1 virological failure and MDR-1 immunological failure, and P < 0.05 for CX₃CR1 immunological failure).

Discussion

This study investigated the effects of SNP within the human CX₃CR1 and MDR-1 genes on time to virological and immunological treatment failure in a large cohort of antiretroviral-naive patients initiating therapy. We report a small but statistically significant correlation between the CX₃CR1 249I polymorphism and slightly earlier immunological failure; and a trend to earlier virological failure in the MDR-1 3435C/C genotype group.

The role of MDR-1 gene polymorphisms and P-gp expression in HIV pathogenesis remains largely unresolved. It has been suggested that over-expression of multidrug transporters may accelerate the development of HIV-1 resistance to PI [²⁴], and that the MDR-1 3435C/C genotype may be a factor restricting access of HIV-1 PI to target cells expressing P-gp [³⁰]. The MDR-1 3435C/C genotype has a much higher frequency among individuals of African origin (between 60 and 80%) [³⁰] than in individuals of Caucasian or Asian origin (> 30%) [^26,30,37]. Although complete ethnic background data is not available for our cohort (individuals are not obligated to answer any questions pertaining to ethnic origin), partial data indicates that the cohort is mainly comprised of persons of Caucasian and Aboriginal descent. Based on this partial ethnic data, the frequency of the homozygous MDR-1 3435C/C genotype (30.1%) was consistent with the frequency reported previously for Caucasian and Asian populations [^26,30,37]. In support of previous reports suggesting a potential effect of the MDR-1 polymorphism on the effectiveness of HIV-1 PI [²⁴], a strong trend to earlier virological failure in the MDR-1 3435C/C genotype group was observed. However, there was no detectable effect of the MDR-1 polymorphism on initial immunological response to antiretroviral therapy using our a priori definition, or, in a limited subset, HIV-1 drug resistance development. These observations extend the findings from a recent study of 123 Caucasian, drug-naive patients initiating antiretroviral therapy, which reported a beneficial effect of the MDR-1 3435T/T genotype on short-term CD4 cell count increases [³¹]. Differences in virological response could not be detected over the 6-month follow-up period [³¹], compared to the 40 months of follow-up reported here.

The CX₃CR1 haplotype frequencies determined in this study are consistent with those previously reported [¹¹]: 249V 280T was the most common, with a haplotype frequency of 0.75 (compared to 0.678-0.730 in French cohorts), followed by 249I 280M, at 0.15 (French cohorts, 0.126-0.198), and finally 249I 280T, at 0.1 (French cohorts, 0.123-0.142). Similarly, we observed complete linkage disequilibrium within CX₃CR1 SNP, and no discernible genetic linkage between the CX₃CR1 and CCR5 polymorphisms [¹¹]. Thus, despite the fact that the CX₃CR1 and CCR5 genes are found in close proximity on chromosome 3 [¹¹], any effects of the CX₃CR1 polymorphisms on disease progression and therapy outcome are not due to linkage to the CCR5Δ32 allele. CX₃CR1 249I 280M was previously identified as a recessive HIV genetic risk factor in untreated individuals, with 280M (but not necessarily 249I) independently associated with faster progression to clinical AIDS and more rapid CD4 cell count decline [¹¹]. However, McDermott et al. examined CX₃CR1 280M alone and could not confirm these results [¹²]. Although not directly addressed here, the discrepancy between the two studies may be explained by our observation that the polymorphism at codon 249, and not codon 280, may be largely responsible for the differences observed in response to therapy. The homozygous 249I polymorphism was associated with reduced immunological response alone and in combination with variation at codon 280, while independent variation at codon 280 was not significant in our analyses. Variation in CX₃CR1 genotype at either codon 249 or 280 however, did not influence virological response in any of our analyses.

A limitation of this study relates to the frequency of viral load and CD4 cell counts performed on the study group. Although the majority of patients had tests performed at baseline, 1 month, and every 3 months thereafter (the median time between viral load tests in this study was 3 months), a minority of patients may have had viral load tests performed more or less often. It is possible therefore, that the therapy follow-up data may be slightly biased towards longer times to success and/or failure, due to the possibility that some study subjects may have large gaps between viral load or CD4 cell counts. Another possible limitation of this study relates to the potential confounding effect provided by incomplete adherence. We addressed this using a previously validated definition based on prescription refills [^33,34]. Although we cannot rule out the possibility of incomplete adherence, it is unlikely that this would be unbalanced between SNP groups. Other potential limitations of this study include its retrospective nature, the fact that the cohort, although large, represents only a geographically limited population, and the fact that the resistance subset for the MDR-1 analysis is non-random, and therefore potentially subject to bias. Finally, the biological mechanism of the effects of the CX₃CR1 and MDR-1 polymorphisms is unclear. In support of previous studies identifying the CX₃CR1 249I 280M genotype as a genetic risk factor in AIDS progression in untreated individuals [¹¹], and linking the MDR-1 3435 C/C genotype to reduced CD4 cell responses [³¹], we have observed a correlation between the CX₃CR1 249I polymorphism and the MDR-1 3435C/C genotype and slightly earlier immunological and virological failure, respectively. The magnitude of these effects is relatively minor despite the fact that they remain statistically significant after adjusting for multiple factors. The actual utility of determining these genotypes in routine clinical practice is therefore not clear.

Acknowledgements

Sponsorship: Z. Brumme is supported by a scholarship from the Natural Sciences and Engineering Research Council of Canada (NSERC).