Plasma samples were collected anonymously and voluntarily from ART-naïve individuals living with HIV and residing in two cosmopolitan cities and 40 remote villages located in the equatorial rain forest in Cameroon from 2000 to 2013. Samples in remote villages were collected in 2000, 2012 and 2013, while those in the cities were sampled in 2007, 2008 and 2009. This study was approved by the Cameroonian National Ethics Committee (ethical clearance number: 2019/04/1156/CE/CNERSH/SP). Viral loads were measured at one time point before treatment initiation. The characterization of the genetic diversity of HIV-1M strains circulating in these cohorts has been described in (33, 35).

HIV RNA was extracted from stored plasma using the QIAamp^® Viral RNA Extraction Kit (Qiagen, Hilden, Germany) according to manufacturer protocols. Complementary DNA (cDNA) was generated using the ImProm-II™ Reverse Transcription System kit (Promega, Madison, Wisconsin, USA), also according to the manufacturer’s protocol. Following this, the entire nef region was amplified by nested PCR (ROCHE Expand High Fidelity kit; Roche-Mannheim, Germany) using non-subtype-specific HIV-1 group M primers. The PCR cycling conditions were the same for both the first and second rounds, consisting of 94°C for 2 min followed by 35 cycles of 94°C for 15 s, 55°C for 15 s, 72°C for 50 s and then 72°C for 7 min. The first round primers were: Nef outer5-le (HXB2: 8513 – 8533; 5’-GTGCCTCTTCAGCTACCACCG-3’ and Nef outer3-3e (HXB2:

9488 – 9508; Reverse primer 5’-AGCATCTGAGGGTTAGCCACT-3’). The nef second round primers were: NEF8746_Sgrl_Ascl_F (HXB2: 8736-8772; 5’- AGAGCACCGGCGCGCCTCCACATACCTASAAGAATMAGACARG-3’) and Nef 9474_Sacll_Clal_R 3-7e (HXB2: 9449-9491; 5’-GCCTCCGCGGATCGATCAGGCCACRCCTCCCTGGAAASKCCC-3’) (35). The second-round primers contain restriction sites AscI (forward) and SacII (reverse), shown in bold, to facilitate downstream cloning experiments. Each donor-derived HIV-1 nef amplicon was purified using the QIAQuick PCR purification kit (Qiagen, Germany), digested with AscI and SacII restriction enzymes at 37 ⁰C for 2 hours and cloned into a pSELECT-GFPzeo expression plasmid (Invitrogen, Canada) that was modified to contain AscI and SacII restriction sites. The pSELECT-GFPzeo plasmid contains two transcription units. The first expression cassete with hEF1/ HTLV promoter drives the expression of the inserted nef gene while, the second unit with the CMV/HTLV promoter drives the expression of the GFP-zeo gene (36, 37). The digested plasmid and nef amplicon were ligated and ligation mixture was transformed into OneShot TOP10 competent cells (Invitrogen, Canada) according to the manufacturer’s instructions. The transformed cells were plated onto Luria-Bertani agar plates containing zeocin and incubated at 37 ⁰C overnight for 16 hours. Colonies were picked, boiled at 95 ⁰C for 10 minutes in 10μl of nuclease-free water, and used as templates in colony PCR to confirm the presence of the nef gene. The bulk plasma-derived nef PCR products and the nef clones amplified by colony PCR were sequenced using the ABI Prism Big Dye Terminator v3.1 Sequencing Kit (Applied Biosystems, USA). The resulting bulk and clonal nef sequences were aligned using MEGA version 10 (38) and a maximum likelihood tree was constructed using IQ Tree using the best fit GTR+I+G nucleotide substitution model (http://iqtree.cibiv.univie.ac.at). Branch support values were computed through IQ Tree’s ultrafast bootstrap analysis (1000 resampling iterations) with Shimodaira-Hasegawa approximate likelihood ratios ([SH]-aLRT) ≥ 0.99] (39). The phylogenetic trees were rooted using an HIV-1 group P sequence (35) and visualized using Figtree v1.4.4 (40). The nef clone sequences were submitted to GenBank under the accession numbers OR979846 - OR980068.

Nef-mediated HLA-I and CD4 downregulation ability was measured as previously described (15). Briefly, an immortalized CD4+ T cell line previously modified to express high levels of HLA-A*02 (CEM-A*02) (15) was used to measure the HLA-I and CD4 downregulation activity of each Nef clone. The cells were maintained in R10 medium (RPMI-1640 supplemented with 10% FBS, 1% L-glutamine, 1% HEPES buffer, and 0.5% penicillin-streptomycin). A total of 6 x 10⁵ CEM-A*02 cells in MegaCell medium (Sigma, USA) were transfected with 8000 ng of Nef clone by electroporation (BioRad Gene Pulser Xcell Electroporation system). The transfected cells were supplemented with R10 and incubated for 20 hours overnight. The cells were stained with APC-labeled anti-CD4 and PE-labeled anti-HLA-A*02 antibodies (BD Biosciences, USA). The surface expression of CD4 and HLA-I were measured by flow cytometry. To determine the relative HLA-I or CD4 downregulation ability of each donor-derived Nef clone, the median fluorescence intensity (MFI) of CD4 or HLA-I expression in GFP-positive (i.e. Nef-expressing) cells was normalized to the MFI of HLA-I or CD4 expression in the negative control (empty pSELECT-GFPZeo plasmid) and the positive control (Nef from the HIV subtype B SF2 reference strain cloned into pSELECT-GFPZeo) using the following formula = [(Negative control – Donor Nef)/(Negative control – Positive control)] x 100%. A normalized value of 0% indicates no downregulation activity, while a value of 100% indicates a downregulation capacity equivalent to that of the positive control. We performed all assays in duplicate, and the results were presented as the mean of these two measurements.

The Nef mutation I43V was introduced into a participant-derived CRF02_AG nef sequence (BS21, GenBank accession JX244966) chosen for its high amino acid similarity to the consensus CRF02_AG sequence. Briefly, the BS21 sequence was cloned into a TOPO TA 3.1 plasmid (Invitrogen, San Diego, USA) and modified by site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis kit (Agilent Technologies, Texas, USA). Following sequence confirmation, the mutant nef was cloned into pSELECT-GFPzeo to assess CD4 and HLA-I down-regulation analysis as described above. Three independent HLA-I and CD4 down-regulation assays were performed.

We used Kruskal-Wallis tests to compare Nef-mediated HLA-I and CD4 downregulation abilities across all HIV-1M lineages and across different years of sampling, and Mann-Whitney U tests for two-group comparisons. We used Spearman’s correlation to assess the relationships between Nef-mediated activities and log₁₀ plasma viral load and between HLA-I and CD4 downregulation capacities. For the set of CRF02_AG Nef clones, we performed a codon-by-codon sequence-function analysis which is a technique that independently looks at each coordinate and determines whether the presence or absence of a particular amino acid is significantly associated with a change in function, to identify amino acid variants associated with significantly increased or decreased Nef-mediated HLA-I and CD4 downregulation ability. In this analysis, multiple comparisons were addressed using q-values (41). Functional comparisons between parental and mutant Nefs were performed using Unpaired t tests. Overall, p-values <0.05 were considered significant.

Two hundred and thirty-four (234) participants were enrolled between 2000 and 2013. Of these, 125 individuals were recruited in the cosmopolitan cities of Yaoundé (the country’s capital) and Douala (the country’s economic hub), 109 were enrolled from communities in remote locations within the equatorial rain forest. Biological gender data were available for only 103 of the participants; 55 in the cities (amongst whom 41 male and 14 female) sampled in 2007 and 48 in the remote villages (17 males and 41 female) collected in 2000; 101 participants reported their age and the median was 31 years (inter-quartile range [IQR] 27-38) for the cities cohort and 32 [IQR 26-41] for the second cohort. The median plasma HIV-1 RNA viral load was 4.41 log₁₀ copies/ml [IQR 3.7-4.9], available for 187 participants; 4.4 [IQR 3.8-4.9] in the cities and 4.5 [IQR 4.0-5.0] in remote villages ( Table 1 ). The statistically differences in gender, age and the plasma HIV-1 RNA viral loads of individuals living in cosmopolitan cities versus remote communities are summarized in Table 1 . All individuals were ART-naïve, and HIV infection dates were not known.

We have previously characterized HIV-1 group M (HIV-1M) genetic diversity in this cohort (35, 42). Overall, CRF02_AG accounted for 59% of all nef clones (n=137), with the remaining 41% (n=97) comprising various subtypes, sub-subtypes and circulating recombinant forms (CRFs) observed at frequencies of 1% to 9%. These included subtypes A (n=3), G and D (n=12 each), and H (n=3); sub- subtypes A1 (n=5), A2 (n=1), and F2 (n=9); and CRFs 01_AE/22_01A1 (n=22), 11_cpx (n=11), 13_cpx (n=2), 25_cpx (n=1), 36_cpx (n=2), 37_cpx (n=2), 45_cpx (n=1). Finally, variants that could not be classified into any known subtypes or CRFs were classified as either divergent or recombinants (n=11) ( Figure 1 ). It should be noted that CRF22_01A1 sequences are embedded within the CRF01_AE cluster in the nef gene. In addition, all these isolated Nef clones clustered with their respective bulk plasma sequences ( Supplementary Figure 1 ) and were free of gross genetic defects. Each Nef clone was transfected into the CEM-A*02 T cell line (15), after which, Nef-mediated HLA-A*02 and CD4 downregulation was measured by flow cytometry and normalized to that of the control subtype B strain SF2 (representative data in Supplementary Figures 2A–F ). Replicate measurements were consistent for HLA-I and CD4 downregulation (Spearman’s, r=0.85 and p < 0.0001; not shown). Overall, the HLA-I downregulation functions of the 234 isolated Nef clones ranged from 5% to 106% relative to SF2 (median 63%; IQR 45% - 83%) ( Supplementary Figure 2G ) while their CD4 downregulation function ranged from 9% to 108% relative to SF2 (median 94%; IQR 82% - 99%) ( Supplementary Figure 2H ). HLA-I downregulation significantly correlated with CD4 downregulation (Spearman’s, r=0.48 and p<0.0001; Supplementary Figure 3 ).

Previous data have shown that Nef function differ across HIV-1M lineages (43). To detect these differences in our cohort, we stratified our Nef-mediated HLA-I and CD4 downregulation data by HIV-1M lineage (where we only considered lineages with at least three representatives). Though HLA-I downregulation activities varied markedly among the isolates tested, we observed no overall significant differences in this function across the HIV-1M lineages represented in our cohort (p=0.52; Kruskal-Wallis test, Figure 2A ). Nevertheless, it was interesting that subtype H lineages exhibited the highest average Nef-mediated HLA-I downregulation ability (median 83%; IQR 26% - 87%) and subtype G the lowest (median 60% and IQR 48% - 90%). Somewhat in contrast, Nef-mediated CD4 downregulation showed a much narrower range of activity between isolates, with the majority displaying normalized function close to 100%. Similar to our observations for HLA-I downregulation, we observed no overall significant differences in CD4 downregulation function across the HIV-1M lineages in our cohort (p=0.40; Kruskal-Wallis’s test, Figure 2B ).

Categorizing our samples based on whether or not they belonged to the main circulating CRF02_AG lineage revealed no significant differences in the Nef-mediated receptor downregulation between the two groups (Kruskal-Wallis, p=0.12 for HLA-I and p=0.19 for CD4 downregulation; Supplementary Figures 4A, B ). Plasma viral loads (VL) also did not significantly correlate with either Nef-mediated HLA-I or CD4 downregulation activities (Spearman’s correlation, r=0.07 and p=0.36 for HLA-I and r=-0.10 and p=0.17 for CD4 downregulation; Supplementary Figures 5A, B ). Furthermore, analyses of the Nef functions according to the sex revealed no significant differences (data not shown).

It has been previously hypothesized that the evolutionary trajectory of the HIV-1 population is shifting towards attenuation (44). We therefore examined temporal trends in Nef function over time. Overall, there were 52 viruses sampled in 2000, 81 in 2007, 14 in 2008, 21 in 2009, 34 in 2012 and 32 in 2013. We therefore examined temporal trends in Nef function over time. A maximum-likehood (ML) tree revealed no strong evidence of temporal clustering ( Figure 3A ).

Our Nef functions analyses revealed an overall steady decline in HLA-I downregulation activities from 2000 (median 74% and IQR 49% - 88%) to 2013 (median 54% and IQR 37% - 70%), that reached statistical significance (p=0.04; Kruskal-Wallis test, Figure 3B ). In contrast, the CD4 downregulation activity of the viruses remained similar across the studied years ( Figure 3C ). Of note, the viral loads of these samples displayed a possibly decreasing trend over time ( Supplementary Figure 6 ).

We next compared the HLA-I and CD4 downregulation ability of Nef clones derived from viruses circulating in the cities with those derived from viruses circulating in remote settings. The rationale supporting this is that the two cohorts might have followed different evolutionary trajectories, possibly driven by sexual partner networks likely to be much broader in the highly connected cosmopolitan cities of Yaoundé and Douala than in the much less connected remote villages. For this analysis, we considered samples collected over approximately the same period: 2009 and 2012 (n=27) for the cities and 2012 (n=28) for the remote locations. We began by inferring a maximum-likelihood phylogeny from the 55 clone sequences involved in this comparison, which revealed some evidence of geographical compartmentalization (see round brackets in Figure 4A ). Of note, the Nef function analyses revealed a trend towards a higher HLA-I downregulation activity of viruses sampled in the cosmopolitan cities compared to viruses originating from the remote villages, though this was not statistically significant (Mann-Whitney U test, p=0.06, Figure 4B ). The CD4 downregulation ability of Nef clones derived from the two environments was similar (Mann-Whitney U test, p=0.78, Figure 4C).

We next explored the sequence determinants of Nef HLA-I and CD4 function within CRF02_AG, the dominant circulating variant in Cameroon, by undertaking a codon-by-codon sequence-function analysis. Here, we compared the receptor downregulation function of CRF02_AG sequences harboring (versus not harboring) each amino acid observed at each Nef codon in the dataset (we required a given amino acid to be observed at least five times at that codon to be included in the analysis). In total, 18 amino acids at 15 different Nef codons were associated with differential HLA-I downregulation activity at the exploratory threshold of p<0.05 ( Table 2 ), whereas 13 amino acids at 11 Nef codons were associated with differential CD4 downregulation activity at this threshold ( Table 3 ). Only one of these associations, between Nef codon 43 and HLA-I downregulation, met the q<0.2 threshold for multiple comparisons correction ( Table 2 ). Of note, the Nef I43V substitution was also associated with decreased CD4 downregulation with p<0.03 ( Table 3 ). We therefore selected this mutation for confirmatory testing. To do this, we engineered the I43V substitution into a participant-derived sequence of high similarity to the consensus CRF02_AG sequence by site-directed mutagenesis (participant BS21). As expected, the I43V mutant displayed significantly poorer Nef-mediated HLA-I and CD4 downregulation activity compared to the parental BS21 sequence (p=0.01 and p=0.002, respectively) ( Figures 5A, B ), consistent with the sequence-function analysis.