Sequence Variations in Human Immunodeficiency Virus Type 1 Nef Are Associated with Different Stages of Disease

Study populations.

Most blood samples were derived from 165 HIV-1-infected individuals based at the Chelsea & Westminster Hospital, London, United Kingdom (referred to hereafter as the C&W cohort) who had been enrolled between 1994 and 1996 in a case-control study of the biological and behavioral correlates of nonprogression in HIV-1 infection. Full details of the recruitment strategy have been described recently (9). The patients were categorized into three main groups according to their clinical status and CD4⁺-cell count. Nonprogressors (NPs) (n = 47) were defined as individuals who had been HIV infected for at least 9 years but remained asymptomatic with an absolute CD4⁺-lymphocyte count of >500/mm³. Slow progressors (SPs) (n = 90) were defined as individuals who had also been infected for at least 9 years but whose CD4⁺-cell counts had declined to below 500/mm³. Forty-one of the SPs remained free of AIDS and Centers for Disease Control and Prevention stage IV disease; the remaining 49 SPs developed stage 4 disease or AIDS during follow-up but not within 5 years. Rapid progressors (RPs) (n = 28) were those who developed AIDS within 5 years of documented HIV-1 infection. The majority of the study participants had a first positive HIV antibody test in 1985 and 1986. Since HIV infection was introduced into the United Kingdom in the early 1980s, with a maximum incidence in 1984 (33a), it can be assumed that most patients were infected shortly before their first HIV test result. Of the 47 NPs, 90 SPs, and 28 RPs, samples from 24, 11, and 16 patients, respectively, were used for sequence analysis in this study.

Twelve patients with severe hemophilia A monitored by the New England Area Hemophilia Center at the University of Massachusetts Memorial Hospital, Worcester (named the Worcester cohort) were also investigated. Four of these patients were LTNPs, four were SPs, and four were RPs. LTNPs are those who were infected for more than 10 years and remained free of signs of HIV-1 disease in the absence of antiretroviral therapy. Further criteria were either an absolute CD4⁺-T-cell count maintained at >400/mm³ and a CD4 percentage of >30% or an absolute CD4⁺-T-cell count of >600/mm³ regardless of CD4 percentage. RPs are individuals who by 1992 progressed to death due to HIV-1 or had either an absolute CD4⁺-T-cell count of <200/mm³ with a CD4 percentage of <10% or an absolute CD4⁺-T-cell count of <100/mm³ regardless of CD4 percentage. SPs fit the definition of neither LTNPs nor RPs. The status of nef genes in one LTNP from the Worcester cohort has been described previously (12, 25). nef alleles from another LTNP in the Worcester cohort carried large deletions (20) and were not included in the present study. The 12 participants were infected with HIV-1 before 1984 through contaminated blood products.

All study participants gave informed consent, and the studies were approved by the research ethics committees of the Chelsea & Westminster Hospital and the Worcester hospital. In addition to those from these two cohorts, published nef sequences derived from patients with documented stages of disease progression and CD4⁺-T-cell counts (15, 27, 33, 38) were also included in the analysis.

CD4⁺-T-cell counts.

CD4 cell counts were determined by flow cytometry. For the C&W cohort, the same flow cytometer, monoclonal antibodies, analytical methods, and sample preparation were used throughout the 10-year period of analysis. All assays were performed with blinding as to the participant’s clinical status and CD4 measurements.

Viral load.

Proviral DNA copy numbers in patients in Worcester cohort were estimated by using a modification of the Amplicor HIV-1 test system (Roche Diagnostics Systems, Inc., Branchburg, N.J.) as described previously (12, 13, 25). Quantitative analysis of viral RNA loads was performed with the Amplicor HIV-1 Monitor assay (Roche Diagnostics Systems) according to the manufacturer’s instructions.

DNA preparation and PCR amplification.

Peripheral blood mononuclear cells (PBMCs) were isolated by centrifugation through Ficoll-Histopaque density gradients. Genomic DNA was extracted by standard methods from freshly purified PBMCs or from samples continuously frozen at −70°C since collection. Because HIV-1 proviral DNA is often present at very low copy numbers and shows substantial sequence variation, we employed nested PCR methods with degenerate sets of oligonucleotides to amplify the nef long terminal repeat (LTR) region from PBMC samples from the C&W cohort (31). The outer primers were VPRF (5′-ATGGAACAAGCMCCRGMAGACCA-3′; positions 5559 to 5581) or POLU (5′-CCCTAYAAYCCMCARAGYCARGG-3′; positions 4653 to 4675) paired with LTRR (5′-GACTACGGCCGTCTGAGGGATCTCTAGYTACCA-3′; positions 9689 to 9657); the inner primer pairs were either NOFP (5′-ATACCTASAMGAATMAGACACA RGG-3′; positions 8741 to 8763) and NORP (5′-CTGCTTATATGCAGCATCTGAGGG-3′; positions 9510 to 9487) or NEFFP (5′-TAAMATGGGKRGCAAMTGGTC-3′; positions 8783 to 8803) and NEFRP (5′-AGCAASYTCKRTGCAGCAGT-3′; positions 9420 to 9400). Standard abbreviations are used for positions of base ambiguity (18), and numbers in parentheses refer to the primer positions in the HIV-1 NL4-3 genome. For each sample 0.5 μg of template DNA was used for amplification. PCR amplification of nef genes from patients monitored by the New England Area Hemophilia Center at the University of Massachusetts Memorial Hospital was carried out as described previously (20).

Cloning and sequencing.

PCR fragments were purified from agarose gels with a GeneClean kit (Qiagen Inc., Chatsworth, Calif.) and cloned with a TA cloning kit (Invitrogen Corp., San Diego, Calif.) as recommended by the manufacturers. To minimize the possibility of sample cross-contamination, collection of clinical samples, DNA isolation, and cloning of PCR products were performed in separate laboratory spaces and frequently in different institutions. Moreover, negative controls were included in each PCR amplification experiment. Sequencing was performed with the PRISM sequencing kit (Perkin-Elmer, Foster City, Calif.) on an Applied Biosystems 373 DNA sequencer. An overview of the number of nef sequences analyzed is given in Table 1.

Sequence and statistical analyses.

Sequences were analyzed by using the Genetics Computer Group sequence analysis software package. Pairwise alignments of nucleotide and amino acid sequences were performed with the programs Gap and Distances. Multiple alignments were performed with the PileUp and Pretty programs and optimized manually. Statistical analysis of sequence variations in Nef protein sequences was performed with the chi-square test with the Yates correction. The relationships between Nef progression scores (NefProg scores) (see Results), viral titers, and absolute CD4⁺-T-cell counts were examined by using Pearson’s correlation coefficient. The significance of the correlation coefficient was assessed with Fisher’s transformation. Comparison of mean NefProg scores for the subgroups of the Worcester cohort defined by disease progression was by Student’s t test for independent samples. The software package StatView version 4.0 (Abacus Concepts, Inc., Berkely, Calif.) was used for statistical calculations.

Nucleotide sequence accession numbers.

HIV-1 nef sequences derived from the patients in the C&W and Worcester cohorts have been submitted to the GenBank sequence database and have been assigned accession numbers AF129333 to AF129395.

Defective nef genes are rare in HIV-1 infection.

To investigate the presence of gross deletions in nef, PCR amplification of nef LTR sequences was performed by using DNA extracted from PBMCs prepared from 165 patients (47 NPs, 90 SPs, and 28 RPs) monitored at the Chelsea & Westminster Hospital, London, United Kingdom. None of the 161 positive PCRs from these 165 HIV-1-infected individuals yielded fragments substantially shorter than that expected for a full-length nef gene (data not shown). To explore the frequency of inactivating point mutations or small deletions among the three progression groups, nef alleles derived from selected subgroups of 28 NPs, 15 SPs, and 20 RPs were cloned and sequenced. Four patients of each group were from the Worcester cohort; the remaining 24 NPs, 11 SPs, and 16 RPs were from the C&W cohort (Table 1). Consistent with previous studies (15, 27, 33), phylogenetic analysis revealed that nef sequences from patients with different rates of progression do not form distinct clusters (data not shown). Intrapatient length polymorphism of the nef gene in a previously defined variable region (38) was observed in 11 of the 28 NPs (39%) and in 14 of the 35 progressors (40%) (SPs and RPs combined). This frequency is higher than that reported by Huang et al. (15), who found no intrapatient length polymorphism in nef alleles derived from 10 LTNPs. The length of the nef open reading frames ranged from 585 to 657 bp in the NPs and from 615 to 648 bp in the progressors. The deduced amino acid sequences are shown in Fig. 1. nef open reading frames longer than 627 bp were amplified from 18 of 35 progressors (51%) and from 8 of 28 NPs (29%) (P = 0.12). This is due to a slightly higher frequency of N-terminal duplications in deduced Nef sequences derived from progressing individuals.

The majority of analyzed nef alleles from NPs, SPs, and RPs predicted intact reading frames (Fig. 1; Table 1). All nef alleles derived from one particular NP in the C&W cohort (patient 005) contained an in-frame deletion of 36 bp downstream of the variable-length region in nef (Fig. 1 and data not shown). Surprisingly, relatively high frequencies of defective nef alleles were found in two RPs (patients 161 and 165) with low CD4⁺-T-lymphocyte counts (66 and 4 CD4⁺ cells/mm³). In both patients, 4 of 10 analyzed nef alleles from one time point contained premature in-frame stop codons and/or mutations in the initiation codon (data not shown).

Features of Nef sequences associated with different stages of HIV disease.

To investigate whether specific amino acid substitutions in Nef are associated with different stages of infection, Nef protein consensus sequences from 41 NPs and 50 patients with progressive HIV-1 infection (SPs and RPs combined) were aligned (Fig. 1). Table 1 summarizes the research cohorts that contributed to the analysis. In addition to the nef alleles analyzed in the present study, published Nef sequences from 9 AIDS patients (38), 10 LTNPs (15), and 9 patients with different rates of disease progression (27) were included (Table 1). The three SPs SP1 to SP3 (Fig. 1 and Table 1) described by Michael et al. (27) were included in the NP group in the present study, because they showed a positive CD4 slope and final CD4⁺-cell counts of >500/mm³ over 8 years of follow-up without antiviral therapy.

Selected laboratory characteristics of the three progression groups are summarized in Table 1. On average, the NPs had 9-fold-higher mean CD4⁺-cell numbers and 20-fold-lower viral RNA loads than the RPs at the time of sampling (Table 1 and data not shown). All 24 subjects from the C&W cohort who were classified as NPs had maintained CD4⁺-T-cell counts of >500 per mm³ for 3 years following PBMC isolation for genomic DNA preparation. The average cell number, however, declined from 875 to 732 CD4⁺ cells/mm³ during this time period, indicating that many of these patients are slowly progressing to immunodeficiency (Table 1). The slight increase in the RPs from the C&W cohort from an average of 119 CD4⁺ cells/mm³ after 6.3 years to 147 CD4⁺ cells/mm³ after 7.8 years of documented HIV-1 infection is due to the use of intensive combination drug therapy in these patients starting after the time of sampling for genomic DNA preparation. The average number of nef alleles analyzed per patient varied from 5.3 for the RPs to 8.6 for the NPs. For clarity, only consensus Nef protein sequences representative for each patient were aligned (Fig. 1). The amino acid variation at each position is shown in Fig. 2.

With some exceptions, most previously defined conserved domains in Nef (30, 38) with putative functional relevance were conserved (Fig. 1 and 2). In Nef proteins obtained from five NPs and four SPs, changes in the N-terminal myristylation signal (MGGKWSK) were observed (Fig. 1). All deduced Nef protein sequences from one NP (patient 044) and one RP (patient 104) contained a change of a highly conserved cysteine at amino acid position 63 (044, C63→L; 104, C63→S/N). For patient 044, the C63→L substitution was present in 15 of 15 PCR clones obtained at three independent time points (Fig. 1 and data not shown). Nef sequences derived from several NPs (001, 039, 044, and RR) and a comparable number of progressors (104, 160, and S1) contained an additional G or K residue in the middle of the acidic charged region (amino acids 71 to 75). At other positions in the acidic stretch of amino acids, polar (Q [046, 104, 117, 130, and Pt164]), positively charged (K [165]) or uncharged (S [RP3]) amino acids were observed (Fig. 1 and 2). A (PxxP)₃ motif, resembling an SH3 binding site and shown to be involved in Hck kinase binding (21, 22, 34), and an ExxxLL motif that may be important for Nef-mediated endocytosis of CD4 (6) were universally conserved (Fig. 1 and 2). A second PxxP motif, located close to the C terminus of Nef, was changed to PxxQ, PxxT, or PxxK in five nonprogressing and in four progressing individuals (Fig. 1 and 2). A previously identified putative protein kinase C (PKC) recognition site was almost universally conserved (Fig. 1 and 2). For most Nef proteins the sequence was RPMTYK, although some variations were observed: RPMTYR (patients 016, 032, and 037), RPMSYR (LSS1), RPISYR (003), PPMTWK (127), PPMTFK (167), PPVTWK (MB), and PPMNWK (157) (Fig. 2). All of these except the RP157 sequence still predict a PKC recognition site (R/KX_0-2S/TX_0-2R/K). A predicted β-turn motif (GPG) was universally conserved (Fig. 1). A previously proposed preference for a threonine (GPGT) in nonprogressors, versus isoleucine or valine (GPGV/I) in progressors (15), could not be confirmed (Fig. 1 and 2).

The high degree of conservation of these putative functional motifs suggests that Nef proteins from all progression groups are functionally active. However, the analysis of a large number of samples allowed the identification of some amino acid variations in Nef that appear to be associated either with a well-preserved immune system as among the nonprogressors or with progression to more advanced immunodeficiency. The Nef consensus sequences obtained from the 41 NPs and the 50 progressing individuals differed in only 8 of the 220 amino acid positions aligned: 11V↔G, 15T↔A, 22 · ↔R, 63N↔T, 96A↔G, 98L↔V, 183L↔Q, and 195E↔M (Fig. 2) (numbering corresponds to the alignment; · specifies a gap in the NP Nef). Some of these differences are probably coincidental and/or represent conservative substitutions that should not alter Nef function. However, other variations are highly significant (Table 2) and do affect amino acids that may be important for Nef structure and function. As mentioned above, a (PxxP)₃ motif in the conserved central part of Nef and a second PxxP motif close to the C terminus showed similar degrees of conservation in nonprogressing and progressing individuals (Fig. 1 and 2). However, an additional PxxP motif close to the N terminus of Nef was present in only 1 of 41 NPs (2%) but in 8 of 32 RPs (25%) (P < 0.01) (Fig. 1; Table 2). Other variations include amino acids that may potentially be phosphorylated. At amino acid positions 51 and 157 (numbering corresponds to the NL4-3 Nef), nef alleles from RPs often predicted a threonine, whereas those from NPs usually predicted an asparagine (Fig. 2). Also, in the variable-length region, a threonine or serine was more frequently present in progressors (10 of 50; 20%) than in NPs (2 of 41; 5%) (P = 0.02) (Fig. 2). In contrast, Nef proteins from NPs more frequently contained a threonine at amino acid 15. A tyrosine at amino acid 102 in Nef was relatively conserved in progressors, whereas 34% of the NPs contained a histidine at this position. A serine at amino acid 169 was highly conserved in Nef sequences obtained from NPs (98%). In comparison, 28% of Nef sequences amplified from the RPs contained an asparagine at this position (Fig. 2; Table 2). Several of the sequence variations that seem to be associated with different stages of disease progression were located close to the carboxy terminus of the protein. Nef from progressors more frequently contained a cysteine at position 163 (Fig. 2; Table 2). At amino acid position 170, 58% of Nef proteins from RPs contained a polar amino acid (Q), whereas 68% of the NPs contained an aliphatic hydrophobic amino acid (L). This proportion was reversed at position 182, where Nef from NPs tended to have an acidic (E) or polar (Q) residue and Nef from progressors preferentially contained hydrophobic residues (M or V). These variations in the deduced amino acid sequences from NPs and rapidly progressing individuals suggest that some functional differences in Nef may be associated with different stages of disease. Some of the sequence variations close to the C terminus of Nef may be linked. For example, 79% (23 of 29) of the Nef sequences predicting a cysteine at position 163 contained a glutamine at position 170 (Fig. 1). In contrast, Nef sequences containing a serine residue at position 163 usually had a leucine at position 170 (33 of 51; 65%). These differences were highly significant (P < 0.001). Nef sequences containing the C163-Q170 combination had an asparagine at position 169 relatively frequently (8 of 23; 25%) compared to the remaining sequences (3 of 68; 4%) (P < 0.001).

Relationship between the number of Nef sequence variations and CD4⁺-cell-count and viral load.

Next, we investigated whether individuals representing the extreme ends of the clinical spectrum of nonprogression and rapid progression were likely to harbor proviruses expressing Nef proteins showing a greater number of the features typically observed in the different progression groups. Based on the Nef protein alignment shown in Fig. 1, we identified five features that were more frequently observed in NPs (T15, N51, H102, L170, and E182) and nine features that were more typically observed in RPs at late stages (an additional N-terminal PxxP motif and the presence of amino acids A15, R39, T51, T157, C163, N169, Q170, and M182 in Nef) (Fig. 2). Most of these differences are statistically significant (Table 2). The presence or absence of these Nef features was determined for each patient analyzed, and the number of features more typical for nonprogression was subtracted from the number of features more frequently observed in rapid progression. The corresponding number was termed the Nef progression score (NefProg score). Thus, a Nef sequence that contains all eight amino acids typical for RPs and SPs and an additional PxxP would be assigned a NefProg score of +9. Conversely, a Nef sequence showing all five properties typical for NPs and none of those typical for RPs would be assigned a score of −5. For each HIV-1-infected individual the number was calculated for a representative Nef consensus sequence, obtained from PCR clones derived from a single time point. Where substantial intrapatient variation was observed, the corresponding amino acid positions were not included in the analysis.

In the 41 NPs analyzed, NefProg scores ranged from −5 to +4 with an average of −1.4 (Table 1). For the 18 SPs, scores between −3 and +5 (average of +0.9) were observed, and for the 32 RPs, scores between −2 and +8 (average of +2.8) were observed. It was noteworthy that although the NefProg score varied considerably within each group, a tendency towards more positive numbers in immunodeficient individuals, irrespective of their origins, was observed (Table 1). The greatest difference (NPs, −3.3; RPs, +4.8) was found for the patients in the Worcester cohort.

Each of the three progression groups analyzed in this study is, to some extent, heterogeneous with regard to duration of infection, CD4⁺-T-cell count, and viral load. In order to further investigate how the stage of HIV disease relates to the Nef features identified, we compared the CD4⁺-cell counts and the viral load data with the NefProg scores. First, the 12 individuals in the Worcester cohort were analyzed, because very stringent criteria were used for nonprogressive infection and concurrent or nearly concurrent (within half a year of sampling for PCR analysis) data on CD4⁺-T-cell counts and plasma viral RNA load were available for most of these patients. Average viral DNA copy frequencies were determined with samples from 1994, the same calendar year as the NefProg score determination for most individuals. There was a strong correlation between the NefProg score and the absolute CD4⁺-cell counts (correlation coefficient, −0.854; P value, <0.001) (Fig. 3A), the plasma viral RNA load (correlation coefficient, +0.907; P value, <0.001) (Fig. 3B), and the proviral DNA copy number (correlation coefficient, +0.974; P value, <0.001) (Fig. 3C). The CD4⁺-T-cell count decreased and the viral load increased for patients harboring Nef protein sequences with high NefProg scores (Fig. 3). Next, we expanded the analysis to Nef consensus sequences derived from all 91 HIV-1-infected individuals listed in Table 1 and from 9 additional patients described by Ratner et al. (33). Concurrent data on viral load were not available for some of these patients and/or were obtained by using different methods. Therefore, only the relation between the CD4⁺-cell count and the NefProg score was investigated (Fig. 4). Although there were some exceptions, there was a strong correlation between the absolute CD4⁺-cell count and the NefProg score (correlation coefficient, −0.6; P value, <0.001). The majority of Nef proteins with a score below 0 were derived from individuals with >500 CD4⁺ cells/mm³ (34 of 42; 81%). In contrast, Nef proteins with a score greater than 0 were usually obtained from patients with <500 CD4⁺ cells/mm³ (38 of 45; 84%). These differences are even more significant for the extreme ends of the spectrum of the Nef sequence variations identified. Only 2 of 18 individuals (11%) (patients 490 and 191) from whom deduced Nef protein sequences with a score of ≤−3 were obtained had <500 CD4⁺ cells/mm³ at the time of sampling. In comparison, 1 of 18 patients (6%) from whom Nef proteins with a score ≥+4 were derived had >500 CD4⁺ cells/mm³. Our data suggest that certain sequence variations in Nef predominate in NPs with relatively normal CD4⁺-cell counts, while others are found in severely immunodeficient RPs with low CD4⁺-cell counts.

Sequence variations in Nef during progression from the asymptomatic stage to AIDS.

To investigate whether the sequence variations in Nef that appear to be associated with progressive disease are already present early during the asymptomatic stage or emerge during or after AIDS progression, sequential samples derived from two individuals from the Worcester cohort with progressive HIV-1 disease, FA and MB, were analyzed. Both patients have severe type A hemophilia and are seropositive for both hepatitis B and C viruses.

From the RP FA, blood samples were obtained between 1987 and 1994. In 1987, the CD4⁺-lymphocyte counts were just >500/mm³; by 1990, they had declined to 280/mm³; and in 1992 and 1994, they were <50 cells/mm³ (Fig. 5). FA initiated therapy in 1990 with zidovudine (AZT), changing to dideoxyinosine (ddI) in 1992, to AZT with Nevirapine in 1993, and to AZT alternating with ddI in 1994 until 1996. He has been treated for cytomegalovirus disease since 1995, for thrush (1994), and for presumed Pneumocystis carinii pneumonia (1993). Amino acid dominance in Nef changed at 12 positions between 1987 and 1994: 31D→A/V, 36V→A, 42K→R, 48N→S, 54N→T, 70 · →E, 138I→V/L, 154E→D, 162N→T, 174S→N, 187E→Q, and 197R→H (numbering corresponds to the alignment shown in Fig. 6; · specifies a gap). Three of these changes (48N→S, 70 · →E, and 197R→H) were already apparent in 1990. Seven additional alterations (31D→A/V, 36V · →A, 42K→R, 54N→T, 138I→V, 162N→T, and 174S→N) were observed in a subset of Nef protein sequences obtained from the 1992 blood sample, while the remaining two substitutions (154E→D and 187E→Q) were observed in 1994 (Fig. 6). Several of these changes in Nef, which coincided with progression to AIDS in FA, correspond to those that were more frequently observed in RPs. In addition, Nef proteins detected in FA in 1987 contained several features that were typically associated with a nonprogressive status (T15, K39, N51, N157, S169, and E182). In 1994, when the patient was severely immunodeficient, four of these positions (39K→R, 51N→T, 157N→T, and 169S→N) had changed to amino acids that were more frequently found in progressing individuals (Fig. 6; Table 3). Also, two other substitutions (154E→D and 187E→Q) observed during the late stages of infection in FA were more frequently observed in progressing infection (Fig. 2). Based on our criteria described above, the NefProg score changed from −2 in 1987 to +4 in 1994 (Table 3).

Four sequential samples were also analyzed from the individual MB (Fig. 5). The first sample was drawn in 1988, when the patient was asymptomatic with a CD4⁺-cell count of >1,000/mm³. This declined subsequently to <500/mm³ in 1991 and <50/mm³ after 1993 (Fig. 5). MB initiated therapy in 1991 with AZT, changing to ddI in 1993 and then to AZT alternating with ddI in 1993. In 1996, he changed therapy again to stavudine and Indinavir, with a good clinical response. Three of the 12 nef alleles derived from the 1988 PBMC sample predicted premature stop codons or frameshift mutations (Fig. 6). The nine intact Nef protein sequences showed a relatively high degree of divergence (up to 10%). Five of these Nef sequences showed features corresponding to an NefProg score of −1. The remaining four Nef sequences predicted an additional N-terminal PxxP and Q170, changing the score to +1. At the remaining three time points, all deduced Nef sequences predicted a PxxP close to the N terminus. Furthermore, in comparison with the majority of Nef sequences obtained in 1988, amino acid dominance changed at four positions after 1991: K39→R, N157→T, S169→N, L170→Q, and E182→Q. Similar to the results obtained with sequential samples from patient FA, these alterations represented changes to amino acids that were more typical for RPs (Fig. 6; Table 3).

Thus, several Nef features that were more frequently observed in RPs were not detected in early samples when the CD4⁺-T-cell numbers were high but came to predominate during (MB) or after (FA) progression to AIDS (Table 3). At four positions, the same changes of K39→R, N157→T, S169→N, and E182→Q were observed in both patients. In FA these changes occurred very late in infection when the CD4⁺-cell numbers were already below 50/mm³. In comparison, in patient MB the predominance at these amino acid positions changed when the CD4⁺-T-cell counts declined from the normal range in 1988 to just below 500/mm³ in 1991 (Table 3).

Localization of amino acid variations on the three-dimensional structure of Nef.

Finally, we investigated the localization of the amino acid variations observed between NPs and RPs on the published three-dimensional nuclear magnetic resonance solution structure of HIV-1 Nef (14). Six of the nine amino acid substitutions that may be associated with different stages of infection (Fig. 2) are located within the N-terminal domain (residues 15, 39, and 51) or in the unstructured loop between β-sheets c and d (residues 163, 169, and 170). These regions were deleted in the structural analysis of Nef to circumvent problems associated with aggregation (14). The localization of the three remaining positions (102, 157, 182) is shown in Fig. 7. Interestingly, amino acids 102 and 182, although well separated on the linear sequence, are in close proximity on the tertiary Nef structure. The presence of a His or Tyr at position 102, however, was not significantly associated with the presence of a certain amino acid at position 182. The remaining amino acid position, 157, is located at the tip of a loop structure and well exposed on the Nef surface (Fig. 7).