Rapid HIV-1 Disease Progression in Individuals Infected with a Virus Adapted to Its Host Population

HIV-1 escape from CTL is predictable based on the Human Leukocyte Antigen (HLA) class I alleles expressed by the host. As such, HIV-1 sequences circulating in a population of hosts will harbor escape mutations specific to the HLA alleles of that population. In theory, this should increase the frequency of escape mutation transmission to persons expressing the restricting HLA allele, thereby compromising host immunity to the incoming HIV-1 strain. However, the clinical impact of infection with HIV-1 containing immune escape mutations has not conclusively been demonstrated. Japan’s population features limited HLA diversity which is driving population-level HIV adaptation: for example, >60% of Japanese express HLA-A*24:02 and its associated Nef-Y135F escape mutation represents the population consensus. As such, Japan is an ideal population in which to examine this phenomenon. Here, we combine genetic and immunological analyses to identify A*24:02-positive individuals likely to have been infected with Y135F-containing HIV-1. Over a ~5 year follow-up, these individuals exhibited significantly lower CD4 counts compared to individuals inferred to have been infected with wild-type HIV-1. Our results support a significant negative clinical impact of pathogen adaptation to host pressures at the population level.


Introduction
The highly polymorphic human leukocyte antigen class I (HLA-I) molecules present HIV-1-derived peptide epitopes on the surface of infected cells, targeting these for elimination by cytotoxic T lymphocytes (CTL). HLA-Is are thus critical to HIV-1 immune control [1,2]. They also represent strong evolutionary pressures that drive the selection of CTL escape mutations in the HIV-1 genome [3][4][5][6], which act by disrupting intracellular epitope processing [7,8], abrogating viral peptide-HLA binding [9], or altering interactions between the HLA-bound peptide and the T-cell receptor (TCR) [10]. Critically, CTL escape mutations are highly predictable based on the HLA-I alleles expressed by the host [11,12]. By extension, HIV-1 sequences circulating in a given host population exhibit adaptations that reflect the HLA-I allele distributions of that population. This in turn means that the frequency of an HLA allele in a population will tend to correlate positively with the frequency of its associated escape mutations in circulation [13], though exceptions apply [14]. Intuitively, higher circulating frequencies of HIV-1 CTL escape mutations should result in their increased transmission to persons expressing the relevant HLA, thereby compromising host cellular immunity to the incoming viral strain. Indeed, observations from high-seroprevalence settings support a role for HLA-driven HIV-1 adaptation in undermining the protective effects of certain HLA-I alleles at the population level [15]. However, the impact of population-level adaptation on HIV-1 disease progression at the individual level has yet to be directly demonstrated.
Japan's HIV epidemic, largely concentrated in men who have sex with men (MSM), is unique in terms of the limited HLA-I diversity of the host population. For example, >60% of Japanese individuals express HLA-A Ã 24:02 (A Ã 24:02) (http://www.hla.or.jp/). It has been hypothesized that population-level HIV adaptation, and its consequences, may occur more rapidly in populations with limited HLA-I diversity [16]. Indeed, one of the most well-characterized HLA class I-associated escape mutations in HIV, a tyrosine to phenylalanine mutation at Nef codon 135 (Y135F) that is reproducibly selected in A Ã 24:02-expressing individuals [17][18][19][20][21][22] represents the population consensus sequence in Japan, whereas the global HIV subtype B consensus is the wild-type tyrosine (Y135) (http://www.hiv.lanl.gov/) [17].
Nef codon 135 serves as the N-terminal anchor residue for two overlapping A Ã 24:02-restricted epitopes starting at Nef's 134 th residue: an 8mer (RW8; RYPLTFGW, Nef134-8) [21,23,24] and a 10mer (RF10; RYPLTFGWCF; Nef134-10) [17-19, 22, 25]. In theory, CTL responses to these epitopes in A Ã 24:02-expressing persons may differ depending on whether the individual is infected with HIV-1 harboring wild-type (Y135) versus escaped (Y135F) form at this residue: in particular, initial CTL responses to these epitopes in the latter case might be reduced or absent. Furthermore, the presence of Y135F in the infecting HIV-1 strain may signal the presence of other unknown A Ã 24:02-associated escape mutations in other viral epitopes that might also compromise CTL responses to the incoming HIV strain. If this were the case, infection of A Ã 24:02-expressing persons (representing the majority of Japanese individuals) by Y135F-containing HIV-1 (representing the majority of Japanese HIV strains) could bring about graver prognosis than infection by Y135-containing virus. To investigate this, we identified A Ã 24:02-expressing individuals likely to have been infected with escaped (Y135F) versus wild-type (Y135) HIV-1, from within our wellcharacterized chronic HIV-1 infection cohort, and compared these two groups with respect to their longitudinal HIV-1 clinical outcomes. Our analysis suggested more rapid CD4 decline in A Ã 24:02-expressing individuals inferred to have been infected with Y135F-containing HIV-1.

Materials and Methods Participants
This study comprised 31 A Ã 24:02-positive adults (29 male, 2 female) who were randomly retrospectively selected from participants of a longstanding HIV cohort study based at the Institute of Medical Science at the University of Tokyo (Table 1). All subjects reported unprotected sexual intercourse and were infected with HIV-1 subtype B. Date of infection is not known for study participants. All study patients were untreated with untiretroviral drugs for the duration of followup. In addition to routine viral load and CD4 + cell count monitoring, blood samples were separated into peripheral blood mononuclear cell (PBMC) and plasma and preserved in liquid nitrogen or at -80°C, respectively.

Ethics Statement
This study was approved by the internal review board of the Institute of Medical Science at the University of Tokyo (certificates 20-47-210521, 20-31-1120). All study participants were adults and provided written informed consent to participate in the cohort study.

Proviral DNA pyrosequencing
One microgram of genomic DNA (representing DNA from 1.5 x 10 5 cells) was subjected to nested-PCR using Platinum Taq DNA polymerase High Fidelity (Invitrogen) using first round primers Nef-2F and Nef-2R and second round Nef-specific primers (underlined). The forward primer additionally contained an A-adaptor (italics) and Multiplex Identifier (MID; X 10 ): (5'-CCATCTCATCCCTGCGTGTCTCCGACTCAGXXXXXXXXXXCTCAGGTACCTTTAAGA CCAATG (nt 9028-9050) [26] and the reverse primer contained a B-adaptor (5'-CCTA TCCCCTGTGTGCCTTGGCAGTCTCAGGAAATGCTAGTTTGCTGTCAAAC, nt 9387-9365). After gel purification, pyrosequencing was performed using the GS-FLX (Titanium kits; Roche Applied Science). GS-FLX sequences were processed by removing primer sequences and extracting sequences corresponding to SF2 nucleotides 9051-9364. Identical reads were collapsed into a single sequence and aligned using MUSCLE [27]. Possible instances of cross-contamination or tag-switching (a known problem during multiplex pyrosequencing [28]) were detected using a genetic distance approach. Nucleotide differences between patient plasma and provirus sequences were identified using MEGA6 [29]. Briefly, if the maximum homology between a patient's plasma and provirus was <95%, we then examined homology to other patient sequences processed in the same run; if any were >95% homologous these were removed as possible cross-contaminants (four such cases were identified, with putative contaminants occurring at <0.4% in all cases). The remaining sequences were aligned to SF2 nucleotides 9213-9242 (Nef codons 134-143) using a Smith-Waterman algorithm [30,31] modified to stringently detect insertions/deletions. Insertion/deletion-containing sequences were removed and Y135 frequencies calculated from the resulting alignment.

HLA typing and haplotype prediction
where f AiBj is the frequency of HLA-Ai and HLA-Bj) was > 1, the haplotype was predicted as A1B1 and A2B2; otherwise as A1B2 and A2B1.

Statistical analysis
Statistical analyses were performed in GraphPad Prism 6 (GraphPad Software). P-values <0.05 were considered statistically significant.
We then evaluated cell surface expression of endogenously-derived epitope using a minigene transfection approach. Luciferase readouts from the fusion constructs indicated that Y135F did not affect minigene transcription and translation (Fig 1B), however this mutation dramatically reduced epitope recognition by CTL clones known to be dually-specific for Y135 and Y135F-containing epitopes ( Fig 1C). Specifically, the introduction of Y135F into the minigene abolished the response of CTL clone T26-102, dually specific to Nef134-8(Y135) and Nef134-8(Y135F) (Fig 1C, left), to the background level. Interestingly, weak but significantly higher response than background level (about 20% of the Y135-containing wild type) of CTL clone C1-28, dually specific for Nef134-10(Y135) and Nef134-10(Y135F) (Fig 1C, right), was observed. Together, these results suggest that infection with Y135F-containing HIV-1 may be able to stimulate weak CTL responses to Nef134-10 but not to Nef134-8.
Follow-up of the four individuals in subgroups 1A and 1B with plasma Y135 at baseline revealed that all subsequently developed escape mutations in this region (Fig 3). The two subgroup 1A individuals subsequently escaped within Nef134-10 (IMS0671 developed Y135F while IMS0738 developed F139L, an escape mutation we identified previously [20]) while the two subgroup 1B patients (IMS0525 and IMS0735) also selected Y135F later in their course. The observation of subsequent escape in these individuals suggest that they may have been recruited into our study at a relatively early infection stage, though it is important to emphasize that infection date is unknown for study participants.

Sequence-based inference of transmitted HIV-1
Though infection dates (and transmitted HIV-1 sequences) were unknown for study patients, we hypothesized that proviral HIV-1 sequences would contain archived viral sequences, thus  yielding clues as to the sequence of the founder virus. Specifically, we hypothesized that the presence of wild-type (Y135) HIV-1 in proviral sequences would indicate that this sequence had previously been present in the patient (i.e. presumably acquired at transmission). Similar approaches have been employed in the context of identifying transmitted drug resistance mutations that may have subsequently reverted in vivo [34][35][36].
We thus performed Nef proviral deep-sequencing for all patients except IMS0617 (who harbored Y135W) from PBMCs collected a median of 0 (IQR 0-19.5) weeks after the first visit (Fig 3). IMS0543 and IMS0638 were subsequently removed from the analysis due to very low sequence reads. This left 28 patients with a median of 1611 (IQR 1204-2509) reads each. The potential ability of proviral DNA analysis to infer the sequence of the transmitting strain was tested in patient IMS0509 whose initial Y135 plasma viruses were replaced by Y135F variants by week 58: in this patient, Y135-containing proviral sequences could still be detected at 2.75% frequency >12 years later (data not shown).
Consistent with Y135 transmission and subsequent persistence, Y135 provirus dominated (>93%) in all group 1 individuals analyzed (Fig 3). In contrast, 7 of 22 (32%) individuals in group 2 (plasma Y135F) harbored Y135 provirus at frequencies ranging from 0.06-45.27%, identifying these as possible Y135 transmission cases that had subsequently escaped to Y135F in vivo. The remaining 15 of 22 (68%) individuals with plasma Y135F harbored no detectable Y135 provirus. This finding suggests possible Y135F transmission (though rapid replacement of Y135 sequences by replicating Y135F virus after selection in the infected host cannot be ruled out).

Immunological inference of transmitted HIV-1
We next wished to explore whether CTL reactivity to Nef134-8(Y135) and Nef134-10(Y135) might be used to further infer the sequence of the infecting HIV-1 strain within this epitope. To this end, we analysed CD8 + T cell responses to Nef134 As expected given their wild-type epitope sequences in plasma HIV-1 RNA, the subgroup 1A patients were both Tet-10-and Tet-8-positive (Figs 3 and 4). Also as expected given the presence of C-terminal mutations within Nef134-10 in plasma HIV-1 RNA (that are beyond the boundaries of the Nef134-8 epitope), the two subgroup 1B patients were Tet-10-negative but Tet-8-positive. Similarly, the only patient within subgroup 2B (IMS0871), who harbored plasma Y135F with a C-terminal mutation in Nef134-10, was also Tet-10-negative and weakly Tet-8-positive. Interestingly, of the 21 subgroup 2A patients who harbored Y135F with no additional substitutions in plasma HIV-1 RNA, 20 (95%) were Tet-10-positive while only 11/21 (52%) were Tet-8-positive (p = 0.0036, Fisher's exact test). Taken together with the observations described above, we inferred that Tet-8-negative responses in this group could reflect two possibilities: either loss of immune memory following Y135-to-Y135F selection in vivo (in which case archived Y135 proviruses should be detectable in these patients) or lack of responsiveness to this epitope as a result of infection with HIV-1 containing Y135F.
We then combined the immunological and proviral deep sequence data to infer the transmitted viral sequence. Considering the dramatic negative impact of Y135F on recognition of Nef134-8, we reasoned that the 7 patients who were Y135-provirus-negative and Tet-8-negative represented the most plausible candidates for having been infected with HIV-1 carrying Nef-Y135F. In contrast, the presence of Y135 provirus and/or detectable Nef134-8 CTL responses precludes our ability to rule out infection with wild-type Y135 HIV-1 for the remaining 13 individuals in subgroup 2A. Finally, we stratified our A Ã 24:02-positive individuals into those plausibly infected with HIV-1 containing the Nef-Y135F escape mutation (N = 7) and those for whom infection with global wild-type Y135 was probable (or could not be ruled out) (N = 18). Since clinical markers such as plasma viral loads (pVL) and CD4 + cell counts fluctuate, we calculated intra-patient median values at 50 week intervals, and analyzed differences between groups over a 250 week period. No significant differences in pVL were observed longitudinally between the two groups ( Fig  5A). Notably however, the individuals inferred to have been infected with HIV-1 containing Nef-Y135F exhibited significantly lower CD4 + cell counts at all time intervals except the first visit ( Fig 5B). Clinical data at all time points are shown in Supporting Information (S1 Fig). Taken together, this suggests that infection with HIV harboring a key A Ã 24:02-restricted Nef escape mutation is associated with significantly poorer long-term CD4 outcomes in A Ã 24:02-expressing individuals.

Discussion
Japan features low population HLA-I diversity and an HIV-1 epidemic that is highly focused in a single risk group (MSM). The high prevalence (>60%) of A Ã 24:02 in Japan allows the opportunity to examine the impact of population-level viral adaptation to host HLA on HIV progression at the individual level. To this end, we used the A Ã 24:02-associated Nef-Y135F substitution, located at position 2 of the overlapping Nef134-8 and Nef134-10 epitopes, as a marker of viral adaptation to the Japanese population and we integrated immunologic and viral genetic information to identify A Ã 24:02-expressing individuals who likely acquired Rapid Disease Progression with HIV-1 Adapted to Its Host Population Y135F-containing HIV-1 at transmission. Based on the major impact of Y135F on Nef134-8 binding and presentation, and the archival nature of proviral HIV sequences, we reasoned that such persons would harbor Y135F in plasma HIV, would have no detectable wild-type Y135 proviruses by deep-sequencing, and would have Nef134-8 tetramer-negative PBMCs. We thus identified 7 individuals likely to have acquired Y135F at transmission and 18 individuals for whom acquisition of wild-type Y135 could not be ruled out (a conservative classification, as the prevalence of Y135F in Japan suggests that it is transmitted in >50% of cases). Notably, the former group exhibited significantly lower CD4 counts compared to the latter over nearly 5 years of follow-up. These observations support the idea that acquisition of HIV-1 "pre-adapted" to one's cellular immune responses can accelerate disease progression.
Our results also extend our understanding of the Nef-Y135F mutation [21]. Firstly, our assays revealed a stronger negative impact of Y135F on binding and presentation of Nef134-8 compared to Nef134-10. Moreover, patients unable to mount responses to Nef134-10 due to the presence of C-terminal epitope mutations selected Y135F during follow up, presumably as a result of CTL responses against Nef134-8. This suggests that CTL responses to Nef134-8 play Rapid Disease Progression with HIV-1 Adapted to Its Host Population a crucial role in selecting the mutation in vivo. Furthermore, all 14 patients in whom no wildtype Y135 proviruses could be detected by deep-sequencing were Nef134-10 tetramer-positive, while only 7 of them were Nef134-8 tetramer-positive (p = 0.0058). We showed previously that the tertiary structure of Nef 134-10 bound to A Ã 24:02 is not substantially influenced by the presence of Y135F [20]. Although Nef134-10 presentation is substantially reduced in the presence of Y135F, we speculate that a small amount of Y135F-containing Nef134-10 can still be presented and can still stimulate cognate CTLs. The observation that, in patients with plasma Y135F, TCRs of CD8-positive T cells are often dually-specific for Nef134-10(Y135) and Nef134-10(Y135F) and exhibit a restricted repertoire also supports this notion [19]. The CTL repertoire responding to Nef134-10 thus stimulated may also maintain pressure on Y135F, such that it does not revert to Y135 in A Ã 24:02 + persons who acquire it at transmission.
The lack of association between inferred infection with Nef-Y135F and pVL in A Ã 24:02-expressing individuals merits mention, though this is in line with our previous studies indicating no changes in pVL at the population level in Japan over time [37]. Plasma viral loads predict the rate of HIV progression (in terms of CD4 decline) [38] and are thus commonly used as a surrogate of virulence. However, a recent study suggested that populationlevel HIV-1 adaptation to host HLA may lower the average viral replication capacity [15], which could offset pVL consequences of acquiring escaped HIV. Indeed, we also reported reductions in Gag-protease-mediated viral replication capacity over time among A Ã 24:02-positive patients in our cohort, without discernible effects on viral load [39]. Furthermore, epitope switching from Nef134-10 to the upstream Nef126-10, triggered of Y135F, might also contribute to viral control following escape [22].
CD4 counts were lower for inferred Y135F transmission recipients at all time points except the first visit. Although we cannot formally rule out the possibility that these patients visited the hospital late in their clinical course, a review of their charts did not reveal any information suggestive of late presentation. Furthermore, intra-patient pair-wise genetic distances of group 2 patients were not significantly different between individuals inferred to have been infected with Y135F vs. wild-type Y135 HIV, (data not shown) suggesting that these groups were generally comparable with respect to infection stage. We hypothesize that factors other than viral replication, such as immune activation, could have contributed to the observed accelerated CD4 decline [40].
We interpret our data to suggest that population-level adaptation to the limited HLA diversity in Japan is causing two opposing phenomena. On one hand, these forces may be contributing to reducing viral replication capacity such that modulations in viral load are not apparent at the population level. At the same time, HIV-1 sequences carrying escape mutations may be undermining CTL-mediated viral control, thereby enhancing pathogenesis (in terms of persistently lower CD4 T-cell counts) in persons carrying the most common HLA allele in the population.
The limitations of the study are as follows. First, the sample size was small, which limited statistical power in our study. The modest sample size is attributable to a paucity of retrospective longitudinal specimens and data collected while patients remained antiretroviral-naive (as Japan, like many other global regions, has adopted earlier HIV treatment initiation guidelines), as well as Japan's low HIV incidence and prevalence.Second, infection dates are not known for our cohort patients. As such, we cannot formally exclude the possibility that Y135-provirusand Tet-8-negative patients were those who had come to the clinic late in their disease. Thirdly, we inferred the founder virus sequence by combining proviral deep-sequencing and immune response data. However, we did not estimate the decay of Y135-provirus and CTL response against Nef134-8(Y135). If these decay rates were random from patient to patient, our inference may have been inappropriate. Lastly, we cannot directly ascribe the observed differences in CD4 count to the Nef-Y135F mutation (though it is intriguing to note that Y135F has been associated with increased HIV-1 replication capacity [41], but also ablation of Nef's HLA class I downregulation function, at least in vitro [42]). Rather, we used Y135F as a marker of escaped viruses, and acknowledge that unknown escape mutations could be involved. Further studies in larger cohorts including acute infection cohorts (if available) are needed.
In conclusion, our results suggest that infection with HIV strains harboring escape mutations specific to one's HLA alleles, transmitted as a result of ongoing HIV adaptation to host populations, may contribute to accelerated clinical progression, presumably because such strains fail to stimulate CTL responses against key epitopes. Results suggest that HIV adaptation may represent challenge to both natural and vaccine-induced immunity in populations with limited HLA diversity and support the importance of prevention, early diagnosis and treatment efforts.