First, we wanted to know what ape CD4 was like before SIVs spilled over to apes and began to exert infection pressure on them. We used an alignment of CD4 from diverse simian primates, and the program PAML (Yang, 2007a), to infer ancestral CD4 sequences at the base of the hominin and hominid clades, at the evolutionary positions shown with red and blue nodes in Fig. 2A. These ancestral nodes yielded CD4 sequences that differ from human CD4 by only 2 (hominin) and 5 (hominid) nonsynonymous substitutions. Only one of these changes mapped to the D1 domain of CD4 (N52S; Fig. 2B). We then synthesized these ancient and extinct CD4 genes. We constructed stable cells lines where Cf2Th (canine) cells were transduced with retroviral vectors that stably integrated each of these CD4 genes (hominin ancestral CD4, hominoid ancestral CD4, human CD4, gorilla CD4, chimp CD4, or an empty vector). All cell lines were also engineered to express human CCR5, a critical co-receptor for SIV and HIV.

We then tested these extinct and modern CD4 proteins for their ability to support viral entry mediated by SIVcpz Env. Since we don’t know the actual genetic sequence of the first SIV(s) to infect chimpanzees, the best alternate strategy is to test a phylogenetically diverse set of extant SIVcpz strains (Fig. 2C). We also tested HIV-1 strains that are embedded within the SIVcpz clade, because these represent the virus spillovers from chimpanzees to humans. To generate pseudoviruses bearing SIVcpz and HIV Env, different SIVcpz and HIV-1 Env expression plasmids were co-transfected into 293T cells along with a plasmid encoding HIV-1ΔEnv-eGFP. The cell lines stably expressing various CD4 proteins and human CCR5 were then infected with each of these pseudoviruses. The percent of GFP+ (infected) cells was measured by flow cytometry and viral titers on the different cell lines were calculated and reported as transducing units per milliliter (TDU/mL). All tested pseudoviruses displayed similar levels of infection on cells bearing human or the ancestral CD4 proteins (Fig. 2D, 2E). This suggests that the ancestral versions of CD4 in apes were susceptible to primate lentivirus entry, just as human CD4 is known to be today. On the other hand, cells bearing chimpanzee and gorilla CD4s were generally less permissive to virus entry (Fig. 2D, 2E). We conclude that CD4 was originally permissive to primate lentiviruses, but that selective pressures exerted by SIVs in the chimpanzee and gorilla lineages have led to the retention of mutations that confer resistance to primate lentivirus infection. This has not happened in humans where selective pressure by HIV-1 is too new.

Natural selection operates on individuals within populations, and only over time can the effects of this selection be seen in the divergence of gene orthologs between species. We and others have already shown that SIVcpz has placed selective pressures on chimpanzees such that multiple CD4 alleles circulate in chimpanzees which resist SIVcpz entry better than human CD4 (Bibollet-Ruche et al., 2019; Warren et al., 2019b). We next wanted to know if the same is true in gorilla populations. To similarly analyze gorilla CD4, we used the SNP data from the Great Ape Genome Project (Prado-Martinez et al., 2013) to identify extant CD4 alleles. We analyzed genetic data from 32 gorilla (Gorilla gorilla gorilla [n = 28]; Gorilla gorilla diehli [n = 1]; Gorilla beringei graueri [n = 3]) individuals and found six non-synonymous and five synonymous SNPs separating the individual alleles encoded. Five out of six of the non-synonymous polymorphisms are located within domain 1 of CD4 (two are in the same codon, codon 27) and one in domain 2 (Fig. 3A). A study of over 100 fecal samples from gorillas at field sites in Africa recently identified the same set of SNPs (Russell et al., 2021). Within the gorilla samples, these amino acid differences result in eight distinct allelic CD4 protein haplotypes. The allele frequencies of these protein haplotypes are heterogeneous, where allele 5 is the most common (Fig. 3B). (This, allele 5, was also the gorilla CD4 that was tested in Figure 2D, E and shown in the alignment in Figure 2B.) From looking at the sequences of these different alleles, we noticed a predicted glycosylation site (N-glycosylation tripeptide NXT) at position 15 that is fixed in the gorilla population but absent in the other African apes (Fig. 3A). Interestingly, the gorilla CD4 allele 2 codes for a proline at position 18, immediately after the tripeptide NCT, which strongly reduces the likelihood of glycosylation (Gavel and Heijne, 1990). Since five of the six protein altering polymorphisms are located in domain 1 of gorilla CD4, which directly binds to the lentiviral Env (Fig. 3C) we next wanted to test their functional significance.

We made stable cell lines expressing each gorilla CD4 allele, along with human CCR5 (Fig. S1). We then infected each of these with GFP pseudoviruses displaying envelopes from different strains of SIVcpz, as described above. We quantified the number of GFP+ cells to measure for the differential usage of each CD4 allele. Again, we don’t know the exact strain of SIVcpz that initially infected gorillas, so instead we have tested a phylogenetic diversity of SIVcpz strains. We found substantial differences in susceptibility to pseudovirus entry between the alleles, varying by up to 2 orders of magnitude in some cases (Figure 4). All gorilla alleles were equal to, or more resistant to infection than, the human CD4. We also tested pseudoviruses displaying a diverse set of Envs from HIV-1 groups M and N, and found similar patterns (Figure S2). These data are consistent with SIV putting selective pressure on gorillas in favor of resistant alleles of CD4. However, as would be expected in a host-virus arms race, the viruses are evolving too. As such, we found considerable differences on the entry phenotype for each SIVcpz strain evaluated, where a single host CD4 allele can be highly restrictive to one strain, while being fully functional for entry of another. As an outlier, SIVcpz TAN2.69 showed a uniformly strong ability to use any of the gorilla CD4 alleles.

We next sought to evaluate if CD4 polymorphisms found in gorilla individuals are protective when engineered in the human version of CD4, a widely susceptible receptor for primate lentiviruses. First, we investigated gorilla CD4 allele 2, which encodes a proline at position 18 that is predicted to prevent an otherwise fixed N-glycosylation at position 15 (Fig. 3A). We noticed that gorilla allele 2 CD4 is highly susceptible to most of the SIVcpz strains tested in this study (Fig. 4). Allele 3, which differs from allele 2 only by this proline, supported less entry by SIVcpz TAN1.910 presumably due to this change in glycosylation status (Fig. 5A). To explore the effects of this gorilla specific glycan at residue 15, we generated cell lines stably expressing a mutated version of human CD4 that encodes for the gorilla specific glycosylation motif. We then challenged these cells with pseudoviruses displaying the envelope of different SIVcpz strains and consistently found a decrease in susceptibility to entry compared to wild type human CD4 (Fig. 5B). To confirm the glycosylation status of CD4, we performed CD4 western blotting on lysates from cells stably expressing each of the different versions of CD4. As expected, human T15N CD4, as well as gorilla allele 3, migrated at a higher molecular weight compared to human wild type CD4 and gorilla allele 2, corresponding to the predicted number of glycans on CD4 domain 1 (Fig. 5C). We then treated the lysates with PNGase F, an N-linked glycosidase, and found that all CD4 versions migrated at the same rate, confirming that the mobility shift is due to the glycosylation status of CD4. Thus, most gorilla CD4 alleles have gained a glycan at position 15 that reduces entry of SIV as compared to human CD4. It seems that allele 2, which doesn’t have this glycan, would be at a fitness disadvantage. In support of this, allele 2 is one of the least frequent alleles in the gorilla population that we surveyed (Figure 3B).

We proceeded to test the amino acid residues at the other 3 polymorphic positions in domain 1 of gorilla CD4. We infected cells individually expressing mutant forms of human CD4 that coded for the gorilla specific residues at positions 27, 31, and 34. We found that in all cases, the mutant form of human CD4 encoding the gorilla specific amino acid was significantly more restrictive to at least two of the four SIVcpz strains when compared to human wild type CD4 (Fig. 5D). In several cases, such as H27R CD4 expressing cells infected with SIVcpz MB897, we found drastic changes rendering a fully supportive receptor now highly refractory to infection by a single amino acid substitution. We found that the protective role of these gorilla specific substitutions was SIVcpz strain specific, demonstrating that collectively, the diversity found in gorilla individuals can confer relative protection to all the SIV strains we tested, but that SIVs are counter-evolving as well. These results suggest that single amino acid changes in domain 1 can drastically modify the interaction between CD4 and the lentivirus envelope, directly influencing virus entry.

To evaluate the reverse - if gorilla CD4 mutated to recapitulate the amino acids encoded in human CD4 may render the CD4 a better receptor for SIVcpz - we made and constructed cells expressing those CD4s and quantified the level of infection. We did not observe a full gain-of-function phenotype here and instead found only minimal increases in entry of SIVcpz through these receptors (Fig. 5E). These results imply that the resistance to SIVcpz found in gorilla individuals is not dependent on single amino acids, but rather the cumulative effect of multiple SNPs. Overall, our data suggest that population-level diversity of CD4 in SIV-endemic gorillas confers some level of protection against multiple SIVcpz strains.

Natural selection influences the frequency of SNPs within populations. SNPs with deleterious effects will be kept at low frequency by purifying selection. On the other hand, mutations that confer a selective advantage will reach higher frequencies and/or be maintained in a population longer than expected due to different forms of positive selection (i.e., selective sweeps or frequency dependent selection). We next tested how polymorphism in CD4 has been shaped in ape species, with comparison made between apes that have been endemically infected with SIV (chimpanzees and gorillas) and those that have not (bonobos and orangutans).

Formal methods to detect the influence of selection on population-level nucleotide variation exist (Fay and Wu, 2000; Fu and Li, 1993; Tajima, 1989), but their statistical power is decreased in non-human ape species due to their small sample sizes and lower levels of variation (Prado-Martinez et al., 2013). Thus, we use a comparative approach to detect the signature of natural selection. To do this, we compare patterns of population level diversity between CD4 verus its neighboring genes. We compared SIV endemic apes (chimpanzee and gorillas) to apes uninfected (bonobos and orangutans) or recently infected on an evolutionary timescale (humans). This was performed for both nonsynonymous (protein altering) and synonymous (not protein altering) polymorphisms. Polymorphism data for chimpanzee, gorilla, bonobo, and two orangutan species were obtained from the Great Ape Genome Project (Prado-Martinez et al., 2013) for CD4 and 11 neighboring genes spanning 250 kb of the X-chromosome. For these same genes, human variation was obtained from the 1000 Genomes project. We calculated nucleotide diversity either based on the number of single nucleotide polymorphisms (SNPs; Watterson’s 𝜭_ω(Watterson, 1975)) or mean pairwise difference between individuals (𝜭_ν (Tajima, 1983)). The mutation rate at CD4 does not appear to be elevated given similar levels of variation between CD4 and its neighboring loci when based on the number of SNPs (𝜭_ω) (Table S1 and Fig. S2).

However, within the endemically infected species, nonsynonymous SNPs in CD4 are at a significantly higher frequency compared to neighboring loci, represented by 𝜭ν, (Fig. 6A) (Tajima, 1983). This difference is not observed for synonymous variation. This discordance between nonsynonymous and synonymous variation suggests that the higher frequency of nonsynonymous variants at CD4 in the endemically infected species is not explained by neutral or demographic evolutionary forces. In addition, the higher frequency of segregating nonsynonymous variation is restricted to the endemically infected species. Taken together, these patterns are consistent with positive selection increasing the frequency of and/or maintaining nonsynonymous polymorphisms at CD4 within the endemically infected species only. Also, in support of this, we find that gorilla and chimpanzee nonsynonymous polymorphic sites are significantly concentrated on the domain 1 of CD4 when compared to the un/recently infected species (Fig. 6B). This difference is statistically significant (Fig. 6C). This data suggests that long-term endemic infection of SIV in ape populations may be driving nonsynonymous SNPs to higher frequency in CD4, particularly on the domain 1, the region that directly interacts with the primate lentivirus Env glycoprotein.

The ancestral state of CD4 was determined using the PAML software package as previously described (Yang, 2007a, 2007b; Yang et al., 1995). As input, we used an alignment of CD4 sequences from the following species: human (Homo sapiens; NM_000616.4), common chimpanzee (Pan troglodytes; NM_001009043.1), western lowland gorilla (Gorilla gorilla gorilla; XM_004052582.2), bonobo (Pan paniscus; XM_008973678.1), northern white-cheeked gibbon (Nomascus leucogenys; XM_004092147.1), Sumatran orangutan (Pongo abelii; XM_024256502.1), rhesus monkey (Macaca mulatta; NM_001042662.1), green monkey (Chlorocebus sabaeus; XM_007967413.1), sooty mangabey (Cercocebus atys; NM_001319342.1), pig-tailed macaque (NM_001305921.1), crab-eating macaque (Macaca nemestrena; XM_005569956.2), gelada (Theropithecus gelada; XM_025401282.1), black snub-nosed monkey (Rhinopithecus bieti; XM_017891844.1), drill (Mandrillus leucophaeus; XM_011982990.1), Angolan colobus (Colobus angolensis palliatus; XM_011952091.1), golden snub-nosed monkey (Rhinopithecus roxellana; XM_010385914.1), and olive baboon (Papio anubis; XM_003905871.3).

Human (Genbank ID# MK170450) and chimpanzee (Genbank ID# NM_001009043.1) CD4 expression plasmids were constructed in a previous study (Warren et al., 2019a). The chimpanzee CD4 allele tested here is “allele 6” as defined by us previously (Warren et al., 2019b), and has 2 glycans that impede virus binding to the receptor. Gorilla CD4 alleles and ancestral CD4s were commercially synthesized (IDT GeneBlocks) and gateway cloned into the pLPCX retroviral packaging vector (Clontech). Mutant versions of human and gorilla CD4 were constructed by standard site-directed mutagenesis methods using overlapping PCR primers encoding the modification. Both wild-type and mutant CD4 constructs were analyzed by Sanger sequencing prior to use.

HEK293T (ATCC CRL-11268) were cultured in DMEM (Invitrogen) with 10% FBS, 2 mM L-glutamine, and 1X penicillin-streptomycin (complete medium) at 37 °C and 5% CO2. Cf2Th (ATCC CRL-1430) cells stably expressing human CCR5 (from (Warren et al., 2019a)) were cultured in complete medium supplemented with 250 μg/mL hygromycin. To produce retroviruses for transduction, HEK293T cells plated in antibiotic free media (1x10⁶ cells per well in a six well plate) were transfected with 2 μg of pLPCX transfer vector containing the CD4 gene of interest (or empty), 1 μg of pCS2-mGP (MLV gag/pol), and 0.2 μg of pC-VSV-G (VSV-G envelope) using a 3:1 ratio of TransIT-293 (Mirus) transfection reagent to DNA according to the manufacturer’s instructions. Forty-eight hours post transfection, supernatant was collected, filtered through 0.22 μm cellulose acetate filters, and retrovirus stored at -80 °C in single-use aliquots. Cf2Th cells stably expressing human CCR5 were plated at 2x10⁴ cells per well of a 12-well dish (15% confluent) and 24-h later, transduced with 500 μL of retroviral supernatant by spinoculation at 1,200 xg for 75 min in the presence of 5 μg/mL polybrene. Forty-eight hours post transduction, the cells were placed in complete medium containing selection antibiotics (250 μg/mL hygromycin and 3 μg/mL puromycin) and cultured until stable outgrowth was noted (>1 week). Stable cell lines were maintained indefinitely in selection media. To confirm expression of CD4, cells were analyzed by flow cytometry (Fig. S3). Briefly, cells were harvested from culture plates, washed two times with PBS, fixed in 2% paraformaldehyde, and washed 2 times in flow buffer (1X PBS, 2% FBS, 1mM EDTA). Fixed cells were stained for 30 min at 4° C with PerCP-Cy5.5 mouse anti-human CD195 (CCR5, BD Biosciences 560635) and AlexaFluor647 mouse anti-human CD4 (BD Biosciences, 566681), and analyzed using a BD Accuri C6 Plus flow cytometer (BD Biosciences).

Envelope clones for HIV-1 and SIVcpz EK505 and MB897 were constructed in a previous study (Warren et al., 2019b). SIVcpz MT145, TAN1.910, and TAN2.69 molecular clones were a gift from Brandon Keele, Frederick National Laboratory for Cancer Research, Frederick, MD and used as template for PCR amplification. The RevEnv cassettes of SIVcpz were amplified by PCR using the following primer pairs, where the lowercase sequence corresponds to an added Kozak sequence for enhanced translation: MT145 (JN835462) forward 5’-tcgccaccATGGCAGGAAGAAGCGAGGGAGACG-3’, reverse 5’- TTAAAGCAAAGCTCTTTCTAAGCCTTGT-3’; TAN1.910 (AF447763.1) forward 5’- tcgccaccATGGCAGGAAGAGAAGAGGACGC-3’, reverse 5’- TTAATTTAAGGCTAGTTCCAGACCC-3’; TAN2.69 (DQ374657.1) forward 5’-tcgccaccATGGCAGGAAGAGAAGAGGACGC-3’, reverse 5’-TTAATTTAAGGCTATTTCTAGACCCTGT-3’. PCR products were cloned into the pCR8/GW/TOPO TA plasmid (Thermo Fisher) and then shuttled into a Gateway-converted pCDNA3.1 mammalian expression vector (Invitrogen).

To produce HIV-1Λ1Env-eGFP reporter viruses, 13x10⁶ HEK293T cells were seeded into a 15-cm dishes in antibiotic free media and 24 h later transfected with 13.25 μg of Q23Λ1Env-GFP (group M backbone; (Humes and Overbaugh, 2011)) and 6.75 μg of envelope plasmid. Forty-eight hours post transfection, the cell supernatant was harvested, concentrated (∼100-fold) using Amicon Ultracel 100K filters (Millipore), and stored at -80 °C in single use aliquots. Cf2Th cells stably expressing CD4 and CCR5 were plated at 3x10⁴ cells/well of a 48-well plate 24 h before infection. The cells (∼80% confluent) were then infected with HIV-1 pseudoviruses in three different volumes (Fig 2 and 4), or a volume corresponding to 10-20% infection of cells expressing human CD4 (Fig 5). Infections were carried by spinoculation at 1,200 xg for 75 min in the presence of 5 μg/mL of polybrene. Forty-eight hours post infection, the cells were harvested from the plate and fixed in 2% paraformaldehyde. Fixed cells were washed three times with PBS and resuspended in 50 μL flow buffer (1X PBX buffer containing 2% FBS and 1 mM EDTA) and stained for 30 min at 4 °C with the following antibody mixture: PerCP-Cy5.5 mouse anti-human CD195 (CCR5, BD Biosciences 560635) and AlexaFluor647 mouse anti-human CD4 (BD Biosciences, 566681), and analyzed using a BD Accuri C6 Plus flow cytometer (BD Biosciences). Following singlet cell discrimination, gates were drawn to capture double-positive cells expressing CD4 and CCR5, and then the percent GFP+ cells was enumerated within that population. The data from ∼2x10⁴ cells per technical replicate were analyzed using FlowJo v10. To calculate virus titers (Fig 2 and 4), the linear range of the infectivity curve was determined, and two points within the linear range were selected to calculate the mean virus titer in TDU/mL. The limit of detection for the titer calculation corresponds to a value of 0.2 % GFP positive cells. TDU/mL mean values were normalized to the titer of infection in cells expressing human CD4, and data used to construct a heat map using the Morpheus server (https://software.broadinstitute.org/morpheus); rows and columns were hierarchically clustered by Euclidian distance.

Statistical comparisons were performed between percentages of infected cells in some cases. Values of technical replicates of each biological replicate were compared between mutant and wild type CD4 versions by one-way ANOVA. If a statistically significant difference was found (p < 0.05) in both independent biological replicates, an asterisk was added to the mutant column in the dot plot.

Cf2Th cells stably expressing CD4 cells were lysed in Nonidet P-40 buffer [150 mM NaCl, 50 mM Tris·HCl pH 7.4, 1% Nonidet P-40 substitute, 1 mM DTT, 1 μL/mL Benzonase (Sigma-Aldrich #E1014), and protease inhibitor mixture (Sigma- Aldrich, #11873580001)] by resuspending the cell pellet and rocking at 4° C for 30 min. Cell lysate was cleared by centrifugation at maximum speed for 15 min. Whole-cell extracts were quantified using the BCA assay and 10 μg was subjected to PNGase F treatment according to the manufacturer’s protocol, including a paired sample with no glycosidase as control (New England Biolabs, #P0705S). Treated whole cell extracts (5 μg per lane) were resolved on a 12% TGX Stain-free polyacrylamide gel (Bio-Rad, #1610185) by applying 180V until loading dye ran off the gel. Protein was transferred to a PVDF membrane (Millipore Sigma, #IPVH07850) using a wet transfer apparatus set at 100V for 60 min. The membrane was incubated with blocking buffer (tris-buffered saline 1X, Tween-20 0.1%, 5% milk) for 60 min at room temperature. Primary antibodies were diluted in blocking buffer and incubated with the membrane overnight at 4° C (1:1,000 anti-CD4, Abcam #ab133616). After primary Ab incubation, membrane was washed 4 x 5 min in TBST (0.1% Tween-20). Secondary antibodies were diluted in blocking buffer and incubated with the membrane for 60 min at room temperature (1:10,000 anti-rabbit-HRP, Promega #W401B). After secondary Ab incubation, membrane was washed 4 x 5 min in TBST (0.1% Tween-20) and developed using ECL reagent (Sigma-Aldrich, #GERPN2232), and imaged on a Bio-Rad ChemiDoc Imaging System. As loading control, membranes were reblotted to detect β-acting using a primary (Cell Signal #3700) and secondary (Promega #W402B) antibodies and developed as described.

To compare the pattern of molecular evolution at CD4 relative to neighboring loci we pulled population level re-sequencing data for loci located within 100 kb downstream and upstream of CD4. Primate sequences were obtained from the Great Ape Genome project (Prado-Martinez et al., 2013) and size-matched human sequences were selected to represent diverse ethnic groups from Human 1000 Genomes project.

To identify the individual-specific SNPs within the selected loci, genotype data in variant call format (VCF) was directly downloaded from International Genome Sample Resources (internationalgenome.org/) and the Great Ape Genome Project (biologiaevolutiva.org/greatape/). For human variants, the variant calls were made based on human reference genome annotation hg38, and individual-specific haplotypes were extracted by altering the reference sequence with the alternative SNPs annotated in the VCF files via custom Perl script. For non-human primate variants, the primate genome short read sequences were mapped to human reference genome hg19 to generate the VCF files, as previously described(Prado-Martinez et al., 2013). The SNPs in the VCF files were further filtered by the variant call quality (GQ ζ 15). Like the human sequences, the individual-specific haplotype sequences are re-constructed by correcting the reference sequence with VCF annotations.

In total we obtained population level variation for CD4 plus 15 other loci (six upstream and nine downstream). Four loci were removed from analysis because they have previously been shown to directly interact with a viral protein (USP5 and SPSB2; (Jia et al., 2020; Rathore et al., 2020; Wang et al., 2019; Zhang et al., 2021)) or non-human primate sequencing reads did not map well with the human reference due to repetitive sequence (LAG3 and P3H3). Coding loci included in this study (in order 5’ to 3’) are: ZNF384, PIANP, COPS7A, MLF2, PTMS, CD4, GPR162, GNB3, CDCA3, TPI1, LRRC23 and ENO2. This was done for great ape species endemically infected with immunodeficiency viruses (chimpanzee and gorilla) and those newly or not infected (human, bonobo, Sumatran and Bornean orangutans).

Sequences were aligned for each species individually using the Muscle alignment program (Edgar, 2004). DnaSP (Rozas et al., 2017) was used to haplotype-phase the downloaded sequences and to calculate levels of nucleotide diversity for each locus. Rarely we would observe an internal stop codon within a locus’ reading frame. In these cases, both haplotypes for that individual were removed from analysis. We analyzed the subspecies of gorilla and chimpanzee together. While there is evidence of genetic differentiation between these subspecies (Prado-Martinez et al., 2013) this should not affect our comparisons as the differentiation is expected to be similar across all loci.