• unlimited access with print and download
    $ 37 00
  • read full document, no print or download, expires after 72 hours
    $ 4 99
More info
Unlimited access including download and printing, plus availability for reading and annotating in your in your Udini library.
  • Access to this article in your Udini library for 72 hours from purchase.
  • The article will not be available for download or print.
  • Upgrade to the full version of this document at a reduced price.
  • Your trial access payment is credited when purchasing the full version.
Buy
Continue searching

Human leukocyte antigen supertypes in relation to human immunodeficiency virus infection among populations of African ancestry

Dissertation
Author: Aleksandr Lazaryan
Abstract:
The role of human leukocyte antigen (HLA)-A and HLA-B supertypes in controlling human immunodeficiency virus type 1 (HIV-1) infection in African-Americans remains poorly understood. We examined the effects of HLA class I supertypes on the outcomes of HIV-1 clade B infection among 338 African-American adolescents and women from Reaching for Excellence in Adolescent Care and Health and HIV Epidemiology Research Study cohorts. In addition, we investigated HLA-B58 supertype (B58s) and its constituent alleles in relation to control of HIV-1 clade A (202 Rwandans) and C infection (423 Zambians). Our analysis of clade B-infected African-Americans demonstrated associations of certain previously or newly implicated HLA class I alleles (A*32, A*36, A*74, B*14, B*3501, B*45, B*53, B*57, Cw*08, Cw*12, and Cw*18) with one or more of the HIV-1 outcomes. B58s ( P < 0.0001) and B62s ( P = 0.01) were associated with more favorable outcomes, whereas B7s ( P < 0.001) and B44s ( P = 0.02) were associated with less favorable ones. The unfavorable effect of B*1510 ( P = 0.04) masked an otherwise protective effect of other alleles in B27s ( P = 0.03). HLA-A supertypes exert no appreciable influence. With notable exceptions, we observed broad homogeneity of allelic effects within specific HLA-B supertypes in the context of clade B HIV-1 infection. The associations of B supertypes (B58s, B7s) were largely explained either by well known constituent alleles or by previously unimplicated B alleles grouped together within a supertype (B44s, B62s). The frequencies of HLA-B genotypic supertypes correlated positively with viral load (VL) ( r = 0.63, P = 0.027) and negatively with CD4+ T-cells ( r = -0.62, P = 0.03). Multilevel linear analysis of VL heterogeneity between the genotypic supertypes suggested that B58s and B*57 in particular contributed disproportionally to the explainable VL variation. We detected no significant advantage for B58s as a whole in native Africans with clade A or C, due to contrasting influences of favorable B*57 and unfavorable B*5802. Taken as a whole, we demonstrated the dominant role of HLA-B alleles and supertypes in HIV-1 clade B-infected African-Americans and further dissected the contributions of individual class I alleles and their population frequencies to the supertype effects across the spectrum of viral clades in North America and Sub-Saharan Africa.

TABLE OF CONTENTS Page

ABSTRACT....................................................................................................................... iii DEDICATION.....................................................................................................................v ACKNOWLEDGMENTS.................................................................................................vi LIST OF FIGURES...........................................................................................................ix LIST OF TABLES.............................................................................................................xi INTRODUCTION...............................................................................................................1 Background.................................................................................................................1

HLA class I supertypes...............................................................................................3

Within-supertype effects of HLA class I alleles.........................................................4 HLA class I supertypes and HV-1 outcomes..............................................................6 Driving forces of supertype effects.............................................................................7 PRIMARY HYPOTHESIS AND STUDY AIMS...............................................................8 MATERIALS AND METHODS.........................................................................................9 Subjects.......................................................................................................................9 Study data..................................................................................................................10 Supertype assignment...............................................................................................11 Human subjects.........................................................................................................12 Statistical methods....................................................................................................12

HUMAN LEUKOCYTE ANTIGEN CLASS I SUPERTYPES AND HIV-1 CONTROL IN AFRICAN-AMERICANS............................................................................................23

THE INFLUENCE OF HLA CLASS I A LLELES AND THEIR POPULATION FRE- QUENCIES ON HIV-1 CONTROL IN AFRICAN-AMERICANS.................................53

HUMAN LEUKOCYTE ANTIGEN B58 SU PERTYPE AND HUMAN IMMUNODE- FICIENCY VIRUS TYPE 1 INFECTION .......................................................................86

vii

TABLE OF CONTENTS (Continued) Page

CONCLUSIONS..............................................................................................................106 GENERAL LIST OF REFERENCES.............................................................................109 APPENDIX A B58 SUPERTYPE AND HIV-1 TRANSMISSION..........................................114

B INSTITUTIONAL REVIEW BOARD FOR HUMAN USE APPROVAL FORMS..............................................................................................................118

viii

LIST OF FIGURES Figures

Page

INTRODUCTION

1.

Correlation in VLs between 128 epidemiologically linke d seroprevalent partners of Zambian couples........................................................................................16

2.

Power curves for HIV-1 viral load analysis.................................................................17

HUMAN LEUKOCYTE ANTIGEN CLASS I SUPERTYPES AND HIV-1 CONTROL IN AFRICAN-AMERICANS

1 Frequencies of HLA class I genotypic supertypes and their correlation with HIV-1 outcomes......................................................................................................................49

Supplemental Figures 1

Carriage of HLA-A and -B supertypes in HIV-1-infected African-Americans from REACH and HERS......................................................................................................50

2

HLA-B genotypic supertype frequencies a nd HIV-1 viral load in both sub-cohorts..51

3

HLA-C population frequenc ies and HIV-1 outcomes am ong African-Americans.....52

THE INFLUENCE OF HLA CLASS I A LLELES AND THEIR POPULATION FREQUENCIES ON HIV-1 OUTCOME S IN AFRICAN-AMERICANS

1

Distribution of continuou s clinical outcomes among African-Americans from REACH and HERS cohorts.........................................................................................74

2

Distribution of HLA-B alleles accordi ng to their population frequencies (infrequent vs. frequent ) and HIV-1 control................................................................75

Supplemental Figures 1

a. Population frequencies of HLA- A alleles among 338 African-Americans.............76 b. Population frequencies of HLA-B alleles among 338 African-Americans.............77 c. Population frequencies of HLA- C alleles among 338 African-Americans.............78

ix

LIST OF FIGURES (Continued) Page

2

Distribution of population frequenc ies of HLA-A and -B alleles................................79

HUMAN LEUKOCYTE ANTIGEN B58 SUPERTYPE AND HUMAN IMMUNODEFICIENCY VIRUS TYPE 1 INFECTION IN NATIVE AFRICANS

1

Viral load distribution of HLA-B 58 supertype alleles in Zambians..........................102

2

a. Distribution of HLA-B58 supertype al leles within three VL categories in Zambians....................................................................................................................103 b. Distribution of HLA-B58 supertype alleles within three HIV-1 disease progression groups of Rwandans...............................................................................103

B58 SUPERTYPE AND HIV-1 TRANSMISSION 1

HIV-1 transmission in Zambian couples base d on carriage of B*5802 among seroprevalent index partners......................................................................................115

2

HIV-1 transmission in association with B*57 carriag e among susceptible seronegative partners of serodiscordant Zambian couples........................................116

3

HIV-1 transmission based on B58s carriage among susceptible seronegative partners of Zambian couples......................................................................................117

x

LIST OF TABLES Tables

Page INTRODUCTION

1

Overview of the studies on HLA cla ss I supertypes and HIV-1 outcomes.................18

2

Description of the study cohorts..................................................................................20

3

Description of the study data.......................................................................................21

4

HLA class I supertype framework...............................................................................22

HUMAN LEUKOCYTE ANTIGEN CLASS I SUPERTYPES AND HIV-1 CONTROL IN AFRICAN-AMERICANS

1

Demographic, immunologic, virologic, and genetic features of the individual and combined African-American cohorts....................................................................45

2

HLA class I supertypes and HIV-1 control in African-Americans..............................46

3

Multivariable analysis of HLA class I s upertypes in relation to HIV-1 outcomes among African-Americans...........................................................................................47

4

Factors associated with HIV-1 vira l load heterogeneity across genotypi c supertypes among African-Americans.........................................................................48

THE INFLUENCE OF HLA CLASS I A LLELES AND THEIR POPULATION FREQUENCIES ON HIV-1 OUTCOME S IN AFRICAN-AMERICANS

1

HLA class I markers in association with HIV-1 control among African-Americans.......................................................................................................80

2

HLA class I markers in association with conti nuous HIV-1 outcomes among African-Americans.......................................................................................................81

xi

LIST OF TABLES (Continued)

Tables

Page 3

HLA B*35-Px and -Py subtypes in association with HIV-1 outcomes among African-Americans.......................................................................................................82

Supplemental Tables 1

The associations between selected HLA class I haplotypes and HIV-1 viral load among African-Americans...........................................................................................83

2

False discovery rate analysis of HLA cl ass I associations with HIV-1 outcomes among African-Americans...........................................................................................85

HUMAN LEUKOCYTE ANTIGEN B58 SUPERTYPE AND HUMAN IMMUNODEFICIENCY VIRUS TYPE 1 INFECTION IN NATIVE AFRICANS

1

Effects of HLA-B58s alleles on HIV-1 VL, stratified according to heterosexual HIV-1 transmission categories in Zambians..............................................................104

2

Linear association of HLA-B58s and HLA-C alleles with HIV-1 viral load in Zambians................................................................................................................105

xii

1 INTRODUCTION Background. In less than 30 years hum an i mmunodeficiency virus 1 (HIV-1), which is responsible for acquired immunodefi ciency syndrome (AIDS), has claimed over 30 million lives in different parts of the wo rld. By the end of 2007 the estimated world- wide number of HIV-1-infected adults and children exceeded 33 million, with over 22 million of those living in Sub-Saharan Africa (4 0). Individuals of African ancestry in the United States have been also disproportionately affected by HIV-1/AIDS epidemic, de- spite accounting for the minority (13%) of the total US population. The number of newly diagnosed AIDS cases among African-American s has continued to exceed the cumulative incidence of AIDS cases among all other US racial and ethnic groups together (9). The development of a salutary vaccine will radi cally diminish the ongoing global HIV-1 pan- demic. With numerous social, economic, and scientific obstacles , the understanding of intricate host-virus interaction remains one of the most significant challenges faced by vaccine designers. Their research efforts have been repeatedly impeded in different ways, including: (i) destruction of host immunity by HIV-1; (ii) inadequacy of natural human immune responses to HIV-1; (iii) enormous mutation propensity of HIV-1 leading to vi- ral escape and growing diversity of the virus across geographic regions and even within a single human host; and (iv) unknown correlates of host immunity to HIV-1. The latter is particularly important in de signing appropriate immunogens fo r an effective vaccine. De- spite the progress in developing the broadl y neutralizing monoclonal antibodies against the HIV-1 envelope glycoprotein, the human tr ials of vaccines aimed at generating neu-

2 tralizing antibodies have not been successful. More intense attention has been directed towards the cytotoxic T lymphocytes (CTL) within the cellular arm of immunity against HIV-1. Although results of trial of a vaccine designed to stimulate T cell immunity has also recently proved disappointing (6, 21), the essential role played by CTLs in recogniz- ing and destroying HIV-infected cells (25, 26, 29) will remain a focus of attention in vac- cine development. As compared to antibody-based vaccines, the induction of the CTL responses has an advantage of ensuring immunity against both internal and external foreign antigens (aka epitopes). CTLs are triggered by com plex interaction between the human leukocyte antigen (HLA) molecules, small antigenic peptides, and the cytotoxic T-cell receptors (TCR) on the CTL surface. Once activated, CTLs destroy the HIV-infected cells via mul- tiple pathways, including release of cytolytic granules, induction of apoptosis, and secre- tion of cytokines. It has been demonstrated that the specificity of the CTL response de- pends on the type of antigenic peptide as well as the structural and functional features of the HLA class I molecules, which engage and display epitope(s) on cellular surface. HLA genes, located on chromosome 6, are one of the most polymorphic in the whole hum an genome, encoding over 2100 HLA class I alleles (http://www.ebi.ac.uk/- imgt/hla/stats.html). Because of this extreme polymorphism, HLA molecules are capable to bind a large variety of epitopes and display them on the cell surface for TCR recogni- tion. As the number of discovered HIV-1 epitopes for different HLA class I molecules has multiplied, the prospect of developing a multiepitope vaccine that would include every epitope targeted by an HLA specificity has become less feasible.

3 HLA class I supertypes. The identification of the HLA supertypes in the late 1990s has simplified the complex HLA alleli c interrelationships by consolidating the huge spectrum of individual HLA-A and HLA-B alleles into relatively few groups (32, 33). The rationale for this grouping has emer ged from experimental HLA-peptide binding assays demonstrating that different HLA cla ss I molecules can actually present similar epitopes. Individual alle les within each supertype have ei ther proven or predicted ability to present antigenic peptides with similar anchoring residue s at second and terminal posi- tions. The majority of the pe ptides presented by HLA class I molecules contain 8 to 11 amino acids (31). As initially proposed by Sette and Sidney (32-34), four HLA-A (A1s, A2s, A24s, A3s) and five HLA-B (B7s, B27s, B44s, B58s , B62s) supertype categ o ries incorporated a number of known HLA class I alleles. With rapidly growing number of both HLA class I alleles and their peptide ligands, this experi mental approach to assigning supertype be- came more tedious. Moreover, most of the analyzed samples belonged to Caucasians, hence, many of non-Caucasian cl ass I alleles could not be su ccessfully categorized into supertypes. More recent efforts to improve supertype catego rization have employed computa- tional methods in predicting peptide binding re pertoires. Lund and colleagues proposed a predicted peptide-based approach to supert ype assignment. In th eir proposed algorithm, the HLA ligands were first extracted from SYFPEITHI and MHCPEP databases (1, 22), and the weight matrices for each individual HLA class I molecules were constructed by using a Gibbs sampling algorithm capturing th e highest relative en tropy of binding motifs (16). Clustering of HLA molecules was subseque ntly based on the measures of similarity

4 between these motifs. Reche and colleagues us ed computationally predicted binders and estimated their similarity based on the numbe r of the overlapping peptides (23). Another study of supertype categorizati on, led by Doytchinova (4), used bioinformatics strategy to established the similarities of the cross-re acting HLA proteins inst ead of their binding epitopes. Finally, Hertz and Yanover combined pe ptide- and protein-based similarities in their proposed supertype categorization (11). One important observation with the supert ypes is that, indepe ndent of their as- signm ent methods, the combinations of just few of the major supertypes have been esti- mated to provide substantial cross-population coverage ( 33). As opposed to individual alleles, which were shown to vary across hum an ethnicities, supertypes are invariably ba- lanced across all human populations. The selective pressures that maintain this invariable supertype composition might indi cate the importance of the mi x of supertypes in protect- ing humans from various pathogens in wide ly differing populations. If conserved HIV-1 epitopes that cross-react with representative HLA class I supertypes can be identified, supertype-guided approach could open door to the development of effective HIV-1 sub- unit vaccine containing relatively few epitope s. However, the uniformity in the magni- tude of the CTL responses to viral epitopes is necessary to ensure the success of this ap- proach to the supertype-based HIV-1 vaccine (14). Within-supertype effects of HLA class I alleles. Because HLA class I alleles within each supertype sh are their epitope-binding specificiti es, it is possible that they might also demonstrate similar associations with HIV-1 outcomes. Frahm and colleagues (7) demonstrated that subjects carryi ng HLA-B*1516 and B*1517 – both of which be- long to B63 serologic subgroup within B58s, were associat ed with lower HIV-1 viral

5 load. HLA-B*5801, another member of B58s, was found to be protective against HIV-1 among Sub-Saharan Africans with clade C infection (13). The study of commercial sex workers in Kenya demonstrated that HLA-A*0202 and A*6802 from A2s conferred resistance to HIV-1 infection (17, 18). In the same population both of these alleles were also highly represented among the uninfected chil- dren of HIV-positive Kenyan mothers, thereby extending the protection of the A2s alleles to vertical HIV-1 transmission from mother to child (18). In their study of HIV-infected males from the Multicenter AIDS Cohort Study, Trachtenberg et al. reported a residual significant association for the alleles within B27s, B7s, and a trend for those within the B58s after excluding the individual HLA-B alleles known to strongly influence HIV-1 outcomes (39). The associations between individual alleles within each supertype and th e HIV-1 outcomes have to be reviewed with careful attention to previously detected trends in as- sociations. Those associations might be interpreted differently when viewed from the su- pertype perspective. This might be particularly important for alleles whose lower popula- tion frequencies have provided insufficient power to make statistically meaningful infer- ences in the previous studies. Moreover, a systematic assessment of population-specific effects of alleles within the same supertype might reveal either important differences in their epitope binding characteristics or epitope-independent mechanisms for immune con- trol of HIV-1 infection (e.g., regulation via natural killer cell activity). If functional dif- ferences among alleles within a single HLA supertype are substantial, then the knowl- edge of those differences would be crucial for predicting responses to vaccines designed on the basis of supertype. Moreover, detection of the functional allelic heterogeneity

6 within the supertypes might point either toward the important features of the virus, HLA class I alleles, TCR polymorphism or other as yet undefined factors. Viral features may potentially include clade-depende nt epitope specificities or fitness cost of viral escape. Conversely, the avidity of the epitope or its tolerance to viral escape mutations may sig- nificantly vary from one HLA cl ass I molecule to another. HLA class I supertypes and HIV outcomes. P rior research demonstrated that HLA class I supertypes vary in their associ ations with CTL res ponses and control of HIV-1 viral load. Key findings from the prev ious studies are summa rized in Table 1. In particular, association of B58s with more vigorous CTL-response and lower viral load has been contrasted with the opposing effects of B7s in Caucasians with HIV-1 clade B and B44s in native Africans with clade C (20, 28, 39). Such population-specific supert ype associations with HI V-1 control m ight facili- tate prediction of host immune responses to specific vaccine constr ucts. Similar predic- tion has been already illustrated at the le vel of the individual HLA alleles among AL- VAC-HIV canarypox vaccine trial participants (12). In that trial, the carriers of HLA- B*27 and HLA-B*57 responded significantly be tter to both Gag and Env HIV-1 proteins. These findings emphasized the importance of i ndividual alleles in th e population genetic background for the efficacy assessment of th e candidate HIV-1 vaccine. However, if there is functional sim ilarity among the alleles within a single supertype, predictions at the level of HLA supertypes would become possible. Shared epitope recognition by al- leles within a single supertype would be e xpected to produce relatively uniform associa- tions of immunologic or clini cally relevant CTL responses.

7 Driving forces of supertype effects. It has been thought that the observed effects of supertypes are predominantly driven by cer tain individual HLA class I alleles, inde- pendent of their population frequencies. Howe ver, it has been also suggested that HLA class I population frequency can al so be responsible for some of the differences in HIV-1 outcomes. Rare HLA class I alleles and supe rtypes might provide selective advantage against HIV-1 infection because the virus is less capable of circumventing class I mole- cules to which it has not been sufficiently expo sed in the past. These two explanations for the supertype effects on HIV-1 outcomes might no t be mutually exclusive. In fact, Trach- tenberg et al. acknowledged th at the contribution of the s upertype frequency to HIV-1 progression in their study (39) was separate from the possi ble effects of the individual HLA alleles. However, these two driving fo rces behind the supert ype effect have not been examined within the same study, and the relative magnitude of each separate contri- bution is unknown. Finally, if ra re alleles are indeed advant ageous against HIV-1 infec- tion, then it would be important to know how much is contribu ted by alleles at each of the HLA class I loci, HLA-A, -B, and -C. As a follow-up of the earlier studies on HLA class I supertypes, our research aim s to examine the associations of HLA class I supertypes with HIV-1-related outcomes in the populations of African ances try since they continue to account for a disproportionate burden of HIV-1/AIDS worldwide and in the US.

8 PRIMARY HYPOTHESIS AND STUDY AIMS Primary hypothesis. HLA class I supertypes are diffe rentia lly associated with HIV-1 outcomes among African-Americans from REACH and HERS cohorts of patients infected with HIV-1 clade B. Aim 1.

To determine and contrast the frequencies of the HLA class I supertypes and alleles in African-Americans. Aim 2.

To examine the associations of th e HLA class I supertypes and alleles with HIV-1 outcomes in African-Americans. Aim 3. To explore the degree of uniform ity in the control of the disease exerted by the separate alleles of the supertypes a ssociated with HIV-1 progression in African- Americans. Aim 4.

To assess the contribution of the indivi dual HLA class I alleles and allele frequencies to the supertype effects in African-Americans. Aim 5. To exam ine the effect of B58 supert ype on control of HIV-1 infection and its heterosexual transmission among native Africans.

9 MATERIALS AND METHODS Subjects. The description of the study cohorts and subjects is summarized in Ta- ble 2. W e analyzed the HLA class I data with in the combined cohort of seroprevalent Af- rican-American adolescents and women from the Reaching for Excellence in Adolescent Care and Health (REACH) and from HIV Epid emiologic Research Study (HERS). In ad- dition, we used the data from native African cohorts of HIV-1-infected Zambian sex partners and sexually active HIV-1-infected Rwandan women to investigate B58 super- type. The design and data structure of these cohorts were previously described in detail elsewhere (5, 19, 36, 38, 41). In short, REACH was a prospective observational study which enrolled over 500 teenagers from 15 locations in 13 US cities (4 in New York/New Jersey area, 3 in the Mid-Atlantic, 6 in the Southeast, 1 in Chi cago, and 1 in Los-Angeles). By the beginning of 2001, 355 seropositive partic ipants (71% African-Americans, 75% females, median age of 18 years, 91% with CD4 ≥ 200 cells/ml) were successf ully accrued into the study (41). Out of those, 227 subjects (66% African -Americans, 75% females, median age of 18, 93% with CD4 ≥ 200 cells/ml) had both virologic a nd clinical data available during their treatment-free follow-up in tervals. We have restricted this study to 161 African- American REACH participants. Our subset refl ected closely the base population of Afri- can-Americans by their demographic (71% female s with median age of 18) and clinical (93% with CD4 ≥ 200 cells/ml) characteristics.

10 HERS is a larger multicenter prospe ctive cohort which enrolled 871 HIV-1- infected women (60.3% African-Americans, median age of 35 years, 83% with CD4 ≥

200 cells/ml) in 4 US cities (Baltimore, Ne w York, Providence, and Detroit) from 1993 to 1995 (36). Genetic analyses were perf ormed on 286 eligible women (62% African- Americans, median age of 34, 86% with CD4 ≥ 200 cells/ml) selected for their outcomes during the treatment-free follow-up intervals. Out of those, 177 African-American wom- en, selected for this study, continued to re present the global HERS cohort in their age (median age of 35) and immunologic profile (86% with CD4 ≥ 200 cells/ml). Our subset did no differ from the base cohort by the number of subjects with CD4 ≥ 200 cells/ml ( P > 0.2). The subset of 423 HIV-1 clade C-infected Zam bian partners was selected based on the availability of biologi cal samples for HLA class I de finition and RNA viral load measurements (38). 202 seropositive Rwanda n women with mostly clade A of HIV-1 were selected for this study based on the ava ilability of clinical and hematologic indica- tors of disease progression (3). Study data. Data on patient dem ographics, HL A class I genotyping, virologic, and clinical profile were r eadily available for analyses. The major outcomes among Afri- can-Americans in this study included VL, CD4, and their combinations defining the dis- tinct disease control categories (i.e., ‘controllers,’ ‘intermedi ate,’ and ‘non-controllers’). Rwandan participants were observed longitudina lly for over 8 years and they were classi- fied into three ordinal dis ease progression catego ries (i.e., ‘rapid,’ ‘intermediate,’ and ‘slow’) (3).

11 Viral RNA levels were quantified by Nu cleic Acid Sequence-Based Amplifica- tion (NASBA; NucliSens, Organon, Teknika, Durham, NC) in REACH as compared to Quantiplex branched chain DNA (b-DNA) a ssay (Chiron Corp., Emeryville, CA) in HERS, and RT-PCR (Roche Amplicor 1.0, Roch e Diagnostic Systems Inc., Branchburg, NJ) in Zambian and Rwandan samples. As a ru le, the target viral load assays (e.g., RT- PCR and NASBA) tend to yield somewhat highe r values as compared to those based on signal amplification (e.g., b-DNA). Hence, the viral load values in REACH and Zambian participants were predictably higher than those in HERS. We accounted for this inherent difference by adjusting all analyses among African-Americans for the dummy indicator variable of the cohort membership (REACH served as a referent cohort). The quantifica- tion of CD4 was based AIDS Clinical Tria ls Group standard flow cytometry method across all study cohorts. The highest definition for HLA class I al leles was achieved by polym erase chain reaction with sequence-specific primers as a part of a commercial kit (Pel- Freez/Invitrogen, Brown Deer, WI). HLA class I alleles with residual resolution ambigui- ties were further analyzed by the reference- strand conformation an alysis, followed by the sequencing-based typing, as necessary. Additional data were available from one or m ore of the study cohorts on the fol- lowing variables relevant to this study: chemokine receptor genotypes, age, gender, his- tory of antiretroviral treatment, and number of study visits. The summary of the key study variables is presented in Table 3. Supertype assignment. To enable direct comparison of our findings on supertype association s with previous studies, we essent ially adhered to the recently updated original

12 supertype classification by Sette and Sidney (32, 33, 35) (Table 4). For those class I al- leles that were not successfully captured by this categorization, we reserved alternative supertype assignment strategies (4, 16, 23). Human subjects. All original studies were revi ewed and approved by the Univer- sity of Alabama at Birmingham Institutio nal Review Board (Appendix B). The proce- dures for data safety included encoding the st udy subjects with c linical and laboratory identification numbers so that no pe rsonal information can be revealed. Statistical methods. We used PopGen ( Alberta, Canada) software to assess for Hardy-W einberg population equilibrium and detect the patterns of th e linkage disequilib- rium (LD) between the closely positioned HL A class I loci. We continued searching for LD with SAS/Genetics (Cary, NC), which, in addition to validating the findings from PopGen, revealed a number of additional LD associations. Global association Cochran- Mantel-Haenszel χ 2 test was used to compare and cont rast the differences in distribution of class I alleles and supertypes amo ng the sub-cohorts (REACH vs. HERS). We analyzed ordinal HIV-1 control and progression categories by Cochran- Mante l-Haenszel (CMH) χ 2 and by proportional-odds logistic regression.

The normality assumptions for both VL and CD4 were validated by Kolmogorov-Smirnov test. The nor- mality in VL distribution was attained after log 10 transformation. Random coefficients (RC) mixed linear mode ls were subsequently used to assess the effects of genetic m arkers on the conti nuous repeated HIV-1 outcomes. These models were assumed to be a random deviation from some population regression model and they contained a random intercept ( β 0i ) and a slope ( β i ) for each subject (15, 27). Using this approach to analyze continuous repeated VL s and CD4 was particularly important be-

13 cause the study subjects were independent of each other. In the meantime, both VL and CD4 measurements were clustered within each subject across multiple visits for the same individual, and for that particular reason they were assumed to be correlated. This is why the multilevel analysis is superior in approaching clustered data as compared to the tradi- tional methods in general linear modeling. First and foremost, the RC linear model tests the null hypothesis that the population variances for the longitudinal outcomes are equal to zero. Once this hypothesis is rejected, the model next examines the variation in inter- cepts and slopes in relation to the supertypes and other covariates included into the mod- el. In the absence of differential association between HLA alleles and HIV-1 outcomes throughout patient visits, the estimates of HLA class I effects were set by the models to reflect the average differences within the study follow up. For the alleles which were dif- ferentially associated with CD4 based on patient visit, the differences at the first visit along with the average changes per visit were reported. These estimates were further ad- justed for the effect of an individual cohort (REACH served as a referent cohort) and pa- tient age. We used the unstructured correlation matrix to achieve the best model fit as as- sessed by the lowest Akaike Information Criterion. To avoid an inevitable loss of subjects and to ens ure stability of the estimates from repeated measures analysis, we used a Markov Chain Monte Carlo technique (30) (PROC MI, PROC MIANALYZE, SAS 9.1 Cary, NC) to impute a small proportion of missing consecutive VL (2.3% of all values) and CD4 (3.7%) measurements. In the effort to account for multiple hypothesis testing of the HLA class I associa- tions with HIV-1 outcome s, we estimated q-value for each individual association with

Full document contains 133 pages
Abstract: The role of human leukocyte antigen (HLA)-A and HLA-B supertypes in controlling human immunodeficiency virus type 1 (HIV-1) infection in African-Americans remains poorly understood. We examined the effects of HLA class I supertypes on the outcomes of HIV-1 clade B infection among 338 African-American adolescents and women from Reaching for Excellence in Adolescent Care and Health and HIV Epidemiology Research Study cohorts. In addition, we investigated HLA-B58 supertype (B58s) and its constituent alleles in relation to control of HIV-1 clade A (202 Rwandans) and C infection (423 Zambians). Our analysis of clade B-infected African-Americans demonstrated associations of certain previously or newly implicated HLA class I alleles (A*32, A*36, A*74, B*14, B*3501, B*45, B*53, B*57, Cw*08, Cw*12, and Cw*18) with one or more of the HIV-1 outcomes. B58s ( P < 0.0001) and B62s ( P = 0.01) were associated with more favorable outcomes, whereas B7s ( P < 0.001) and B44s ( P = 0.02) were associated with less favorable ones. The unfavorable effect of B*1510 ( P = 0.04) masked an otherwise protective effect of other alleles in B27s ( P = 0.03). HLA-A supertypes exert no appreciable influence. With notable exceptions, we observed broad homogeneity of allelic effects within specific HLA-B supertypes in the context of clade B HIV-1 infection. The associations of B supertypes (B58s, B7s) were largely explained either by well known constituent alleles or by previously unimplicated B alleles grouped together within a supertype (B44s, B62s). The frequencies of HLA-B genotypic supertypes correlated positively with viral load (VL) ( r = 0.63, P = 0.027) and negatively with CD4+ T-cells ( r = -0.62, P = 0.03). Multilevel linear analysis of VL heterogeneity between the genotypic supertypes suggested that B58s and B*57 in particular contributed disproportionally to the explainable VL variation. We detected no significant advantage for B58s as a whole in native Africans with clade A or C, due to contrasting influences of favorable B*57 and unfavorable B*5802. Taken as a whole, we demonstrated the dominant role of HLA-B alleles and supertypes in HIV-1 clade B-infected African-Americans and further dissected the contributions of individual class I alleles and their population frequencies to the supertype effects across the spectrum of viral clades in North America and Sub-Saharan Africa.