In vivo virulence of MHC-adapted AIDS virus serially-passaged through MHC-mismatched hosts

CD8+ T-cell responses exert strong suppressive pressure on HIV replication and select for viral escape mutations. Some of these major histocompatibility complex class I (MHC-I)-associated mutations result in reduction of in vitro viral replicative capacity. While these mutations can revert after viral transmission to MHC-I-disparate hosts, recent studies have suggested that these MHC-I-associated mutations accumulate in populations and make viruses less pathogenic in vitro. Here, we directly show an increase in the in vivo virulence of an MHC-I-adapted virus serially-passaged through MHC-I-mismatched hosts in a macaque AIDS model despite a reduction in in vitro viral fitness. The first passage simian immunodeficiency virus (1pSIV) obtained 1 year after SIVmac239 infection in a macaque possessing a protective MHC-I haplotype 90-120-Ia was transmitted into 90-120-Ia- macaques, whose plasma 1 year post-infection was transmitted into other 90-120-Ia- macaques to obtain the third passage SIV (3pSIV). Most of the 90-120-Ia-associated mutations selected in 1pSIV did not revert even in 3pSIV. 3pSIV showed lower in vitro viral fitness but induced persistent viremia in 90-120-Ia- macaques. Remarkably, 3pSIV infection in 90-120-Ia+ macaques resulted in significantly higher viral loads and reduced survival compared to wild-type SIVmac239. These results indicate that MHC-I-adapted SIVs serially-transmitted through MHC-I-mismatched hosts can have higher virulence in MHC-I-matched hosts despite their lower in vitro viral fitness. This study suggests that multiply-passaged HIVs could result in loss of HIV-specific CD8+ T cell responses in human populations and the in vivo pathogenic potential of these escaped viruses may be enhanced.

Potent HIV-specific CD8 + T cells select for MHC-I-associated mutations resulting in viral escape from CD8 + T-cell recognition often with reduced in vitro viral fitness [15][16][17][18]. Virus transmission to MHC-I-mismatched individuals could result in reversion of these mutations to recover viral fitness [6,17,[19][20][21]. Thus, it has been speculated that HIV may evolve by selection of individual MHC-I-associated mutations and their reversion after multiple transmissions among individuals with highly-diversified MHC-I genotypes.
Recent studies have suggested that HIV evolves to have lower in vitro replication capacity through accumulation of MHC-I-associated mutations in human populations [17,22]. These studies in HIV-infected humans, however, have had difficulties in addressing the following issues. First, it is difficult to precisely trace serial HIV transmission. Second, it is difficult to compare in vitro viral fitness among highly-diversified HIV variants. Finally, it is difficult to evaluate the in vivo replication capacity of transmitted viruses. A macaque AIDS model of simian immunodeficiency virus (SIV) infection could be helpful to address these issues.
In the present study, we performed serial transmissions of SIV adapted to the protective MHC-I haplotype 90-120-Ia through MHC-I-mismatched rhesus macaques. To determine how viruses with 90-120-Ia-associated mutations can change after multiple transmissions, we first infected 90-120-Iamacaques using a plasma sample from a 90-120-Ia + macaque at 1 year post-infection with the SIVmac239 clone, and performed further plasma transmission through 90-120-Iamacaques. Our analysis revealed that the viruses passaged through 90-120-Iamacaques maintained 90-120-Ia-associated mutations and induced persistent viremia in 90-120-Iamacaques despite their lower in vitro viral fitness. Notably, this passaged viral isolate showed rapid disease progression in 90-120-Ia + macaques when compared to wild-type SIV. These results suggest that passaged viruses can maintain escape mutations and are thus less sensitive to CD8 + T cells restricted by protective MHC-I alleles.
Of the twenty-nine mutations selected in 1pSIV, nine reverted in macaque #21 followed by one additional reversion and one re-selection in #31. Six reverted in macaque #22 but two were selected again in #32. Thus, twenty and twenty-five of twenty-nine mutations selected in 1pSIV remained in 3pSIV1 and 3pSIV2, respectively (Fig 3A and 3B and S1 Fig). Regarding the seven 90-120-Ia-associated CD8 + T-cell escape mutations described above, six remained without reversion even in 3pSIV1 and 3pSIV2. The GagL216S mutation reverted in macaque #21 but was maintained in macaques #22 and #32, while both 3pSIV1 and 3pSIV2 still had the GagD244E. Thus, the majority of 90-120-Ia-associated mutations remained without reversion even in 3pSIV through two passages. Macaques #21, #22, #31, and #32 elicited CD8 + T-cell responses targeting multiple SIV antigens ( Fig 3C). All of these four animals exhibited high frequency Nef-specific CD8 + T-cell responses. Several mutations in addition to the twentynine selected in 1pSIV were selected in macaques #21-#31 and #22-#32 ( Fig 3A).
Next generation sequencing (NGS) confirmed viral diversification in our transmitted plasma samples (S2 and S3 Figs). Phylogenetic distances of viral Gag CA-coding region from wild-type SIVmac239 decreased in macaque #21 but increased in macaques #31, #22, and #32, which may reflect the limited pressure exerted by Gag-specific CD8 + T-cell responses in #21 ( Fig 3C). Phylogenetic distances of Vif-coding and Nef-coding regions from wild-type SIV-mac239 increased in individual animals. Changes in viral genome sequences were the largest in the Nef-coding region, possibly reflecting larger CD8 + T-cell responses targeting Nef ( Fig 3C).

3pSIV exhibits lower in vitro viral fitness
We then attempted to compare the in vitro replication capacity of wild-type SIVmac239 and the passaged viruses 1pSIV, 2pSIVs, and 3pSIVs. It is not easy to compare in vitro replication capacity of plasma HIVs directly and previous studies mostly used recombinant viruses derived from molecular clones such as NL4-3 where gag is replaced by the predominant plasma HIV sequences for comparison of in vitro viral fitness [17,22]. In the present study, we examined the in vitro replication capacity of viruses recovered from peripheral blood mononuclear cells (PBMCs) and plasma as well as recombinant SIVmac239-derived viruses whose gag was replaced by the predominant plasma SIV sequences.
First, PBMCs from macaques #11, #21, #22, #31, and #32 at 1 year post-infection were cultured to obtain PBMC-derived virus stocks, referred to as c-1pSIV, c-2pSIV1, c-2pSIV2, c-3pSIV1, and c-3pSIV2, respectively. These viruses had the same nonsynonymous gag mutations with those in gag cDNAs amplified from plasma RNAs at 1 year post-infection ( Fig 4A). The culture supernatants of HSC-F cells (a macaque T cell line) on day 4 after infection with Plasma viral loads (SIV gag RNA copies/ml plasma) in macaques #11, #21, #22, #31, and #32 were determined as described previously [26]. The lower limit of detection is approximately 4 x 10 2 copies/ml. Viral loads in macaque #11 were reported previously [25].   We recovered viruses, c-1pSIV, c-2pSIV1, c-2pSIV2, c-3pSIV1, and c-3pSIV2, from PBMCs obtained at 1 year post-infection from macaques #11, #21, #22, #31, and #32, respectively. We also recovered viruses, p-2pSIV1, p-2pSIV2, p-3pSIV1, and p-3pSIV2, from 2pSIVs or 3pSIVs plasma but failed to recover viruses from 1pSIV plasma. In addition, we constructed recombinant SIVs, SIV3p1gag and SIV3p2gag, carrying 3pSIV1-derived and 3pSIV2-derived gag, respectively. B. C. D. Relative RT activity of the HSC-F cell culture supernatants on day 4 after infection with PBMC-derived (B), plasma-derived (C), and recombinant SIVs (D). The "E" in the graph C indicates the data on infection with the virus derived from plasma obtained at 1 year after SIVmac239 infection from a 90-120-Iabut 90-010-Ie + macaque (R08-014) that was used in our previous experiment [42]. Similar RT levels were detected in the supernatants of cells infected with the wild-type SIV and the E plasma-derived virus. The E plasma-derived virus had two dominant nonsynonymous mutations in gag at the positions encoding the residues 129 (the mutation/wild-type ratio: more than 4/1) and 468 (more than 1/1 but less than 4/1), but these mutations did not result in loss of in vitro viral fitness. RT activities relative to that of the wild type (set at 100) are shown. Representative results from two sets of experiments are shown. these PBMC-derived viruses showed lower reverse transcription (RT) activity compared to the wild-type SIVmac239 (Fig 4B). These results indicate that all the PBMC-derived viruses, c-1pSIV, c-2pSIV1, c-2pSIV2, c-3pSIV1, and c-3pSIV2, have lower in vitro replicative capacities compared to wild-type SIV.
Second, HSC-F cells infected with concentrated plasma samples obtained from macaques #21, #22, #31, and #32 at 1 year post-infection were cultured to obtain the culture supernatants as passaged plasma-derived virus stocks, referred to as p-2pSIV1, p-2pSIV2, p-3pSIV1, and p-3pSIV2, respectively. We failed to recover a plasma-derived virus stock from macaque #11. There were a few differences between p-3pSIV1-derived and plasma RNA-derived gag sequences, but p-2pSIV1, p-2pSIV2, and p-3pSIV2 had the same nonsynonymous gag mutations to those in gag cDNAs amplified from plasma RNAs at 1 year post-infection ( Fig 4A). Again, all of these virus-infected HSC-F cultures showed lower RT activity compared to the wild-type SIVmac239 in the culture supernatants on day 4 after infection ( Fig 4C).
Furthermore, we compared in vitro viral fitness of these viruses with the wild-type SIV by competition assay. For comparison of wild-type and passaged viruses, HSC-F cells infected with individual virus stocks were cocultured to determine which viral genome sequences become dominant in the culture supernatants. In competition assay of wild-type SIVmac239 with any of the PBMC-derived virus stocks, the wild-type sequences became dominant (S4 Finally, we constructed SIVmac239-derived recombinant viruses, SIV3p1gag and SIV3p2gag, where gag was replaced by the predominant 3pSIV1 and 3pSIV2 sequences, respectively ( Fig 4A). RT assay of the culture supernatants of HSC-F cells on day 4 after infection revealed lower in vitro viral fitness of both of these recombinant viruses compared to the wild-type SIV-mac239 ( Fig 4D).  (Fig 1).
All the 3pSIV-infected animals showed persistent viremia (Fig 5A and 5B), despite the lower in vitro viral fitness of 3pSIV1/3pSIV2. No clear difference was observed in viral loads between 3pSIV1 and 3pSIV2 infection in either macaques. Furthermore, no significant difference was observed in viral loads between 3pSIV-infected 90-120-Iamacaques and the SIVmac239-infected 90-120-Iacontrol group (n = 10) consisting of 90-010-Ie + (n = 6) and 90-120-Ib + (n = 4) animals that were previously reported to show typical levels of viremia [25]. Information on Mamu-A/B alleles in the control group was described in the previous report [25].

Discussion
HIV induces persistent infection and accumulates viral mutations largely due to selection by CD8 + T cells. These mutations often have viral fitness costs and some of them can revert after   [6,17,[19][20][21]. Recent studies in HIVinfected individuals have suggested that these MHC-I-associated mutations can accumulate in the population [22,31,32]. Analysis of HIV-infected transmission pairs has indicated that transmission of HIV mutations associated with the recipients' MHC-I alleles can result in higher viral loads [33,34]. Our present study in a macaque AIDS model demonstrated direct evidence indicating that MHC-I-adapted viruses that have been serially-passaged through MHC-I-mismatched hosts, even with lower in vitro viral fitness, can induce higher viral loads and more rapid disease progression in MHC-I-matched hosts.
The majority of the mutations selected for in 90-120-Ia + animal #11 were present in the 3pSIV virus isolate. Analysis of viral genome sequences in 3pSIV-infected 90-120-Iaanimals showed that the majority of the protective MHC-I haplotype 90-120-Ia-associated escape mutations were maintained after three serial passages. These mutations were preserved after three transmissions in 90-120-Iaanimals, supporting the notion that MHC-I-associated mutations can be maintained in circulating viruses in populations. The 90-120-Ia-associated mutations include GagL216S and GagD244E resulting in reduction of in vitro viral fitness [26,27]. Rapid reversion of these mutations was consistently observed after infection with SIV containing a single mutation in our previous study [35]. However, reversion occurred only rarely in infection with SIV carrying multiple mutations in the present study. No evidence of compensatory mutations which might have rescued viral fitness was found. After transmission, it may be more difficult for a virus with multiple prior CD8 + T-cell escape mutations to revert to wild-type when compared to a virus with a single escape mutation. The new host's CD8 + Tcell response may exert the most important selective pressure on the multiply previously escaped virus (selected for by the prior host's CD8 + T cells) and selection for the new host's MHC-I-associated escape variants may occur first. These new CD8 + T-cell escape variants may have a greater selective advantage in vivo than any reversion of the previous host's MHC-I-associated mutations and thus these changes will occur first perhaps delaying the reversion of the prior escape mutations.
Previous studies examined the effect of viral genome mutations on in vitro viral fitness in the context of molecular HIV clones such as NL4-3 [17,22]. In the present study, we constructed recombinant viruses carrying 3pSIV-derived gag in the context of wild-type SIV-mac239 and showed that these new recombinant viruses had lower in vitro replicative capacities when compared to the wild-type SIVmac239. Mutations in gag appeared to have greater suppressive impact on in vitro viral fitness than those in other regions, consistent with previous reports [15][16][17][18]. Furthermore, we confirmed lower in vitro replication capacity of viruses recovered from plasma and PBMCs. Our results indicate that viruses carrying multiple MHC-I-associated mutations with lower in vitro viral fitness can be serially transmitted through MHC-I-mismatched hosts with maintaining the potential for higher viral loads in MHC-I-matched hosts. This suggests that MHC-I-adapted viruses can circulate in the population.
3pSIV obtained by serial passage through 90-120-Iamacaques maintained several of the mutations selected for in the 90-120-Ia + macaque #11 and induced higher viral loads and more rapid disease progression in 90-120-Ia + hosts. These results demonstrate that MHC-I-adapted viruses can maintain the potential for higher virulence in MHC-I-matched hosts after serial transmissions through MHC-I-mismatched hosts. Although HIV transmissions from individuals with protective MHC-I alleles may be less efficient compared to those without protective MHC-I alleles, this study suggests that HIV isolates that are less sensitive to protective MHC-I alleles can be maintained and circulate in human populations.
3pSIV-infected 90-120-Ia + macaques appeared to generate fewer mutations than wild-type SIVmac239 post-infection, and it is speculated that there were only a limited number of CD8 + T-cell targets in 3pSIV-infected 90-120-Ia + macaques. This may be analogous to the situation in HIV-infected individuals where MHC-I homozygotes exhibit a more rapid course of disease progression [36]. Indeed, all of the 3pSIV-infected 90-120-Ia + animals developed AIDS in 28 months post-infection, whereas 30-40% of 90-120-Iaanimals were alive without AIDS onset at 28 months after wild-type SIVmac239 or 3pSIV infection.
In summary, we directly showed the impact of viral adaptation to MHC-I alleles on viral replication capacity in vivo. Protective MHC-I-adapted SIVs serially-passaged through MHC-I-mismatched hosts exhibited higher virulence in MHC-I-matched hosts despite their lower in vitro viral fitness. Our results indicate that MHC-I-adapted HIVs can circulate in populations, possibly resulting in loss of virus-sensitive MHC-I alleles in these populations. jp/ja/info/kohyo/pdf/kohyo-20-k16-2e.pdf). The experiments were in accordance with the "Weatherall report for the use of non-human primates in research" recommendations (https:// royalsociety.org/topics-policy/publications/2006/weatherall-report/). Animals were housed in adjoining individual primate cages allowing them to make sight and sound contact with one another for social interactions, where the temperature was kept at 25˚C with light for 12 hours per day. Animals were fed with apples and commercial monkey diet (Type CMK-2, Clea Japan, Inc.). Blood collection and virus inoculation were performed under ketamine anesthesia. Animals were euthanized at the end of experiments or at the endpoint determined by typical signs of AIDS including reduction in peripheral CD4 + T-cell counts (less than 200 cells/μl), 10% loss of body weight, diarrhea, and general weakness. At euthanasia, animals were deeply anesthetized with pentobarbital under ketamine anesthesia, and then, whole blood was collected from left ventricle.
The determination of macaque MHC-I haplotypes was based on the family study in combination with the reference strand-mediated conformation analysis of Mamu-A and Mamu-B genes and detection of major Mamu-A and Mamu-B alleles by cloning the RT-PCR products as described before [24]. Confirmed MHC-I alleles consisting of MHC-I haplotypes 90-120-Ia, 90-010-Ie, and 89-002-Ip were described before [24,25].

Analysis of viral genome sequences
Viral RNAs were extracted from plasma using the High Pure Viral RNA kit (Roche). Fragments of cDNAs encoding SIVmac239 (GenBank accession number M33262) Gag, Pol, Vif, Vpx, Vpr, Tat, Rev, and Nef were amplified from plasma RNAs by nested RT-PCR and subjected to direct sequencing by using dye terminator chemistry and an automated DNA sequencer (Applied Biosystems) as described before [28]. Predominant nonsynonymous mutations were determined. The Env-coding region known to have multiple antibody-related mutations was not included in the analysis.
For pyrosequencing, cDNA fragments corresponding to nucleotides (nt) 1760-2463 (containing entire Gag capsid [CA]-coding region), nt 5460-6340 (containing entire Vif-coding region), and nt 9257-10167 (containing entire Nef-coding region) were used for making fragmentation libraries using GS FLX Titanium Rapid Library Preparation Kit (Roche). The products were cleaned with Agencourt AMPure XP magnetic beads (Beckman Coulter) followed by quality control using Agilent 2100 Bioanalyzer (Agilent Technologies). Emulsion PCR was performed with GS junior Titanium emPCR Kit Lib-L (Roche). The emPCR products were deposited onto a GS Junior Titanium Pico Titer Plate and sequenced on the GS Junior System (Roche). Sequencing reads were analyzed by the GS Amplicon Variant Analyzer Software (Roche). After alignment of the FASTA files, populations of <1% were excluded. Molecular phylogenetic analyses were conducted by the Maximum Likelihood method using the MEGA6 software (http://www.megasoftware.net/).

Analysis of in vitro viral fitness
We recovered virus stocks from PBMCs and concentrated plasma samples obtained from macaques #11, #21, #22, #31, and #32 at 1 year post-infection. First, 1-5 x10 5 CD8 -T cells negatively-selected from PBMCs were cultured in RPMI with 10% fetal bovine serum and 10 ng/ml human interleukin-2 (hIL-2) (Roche) with stimulation by 2 μg/ml Phytohemagglutinin-L (Sigma) on the first 2 days. The culture supernatants on day 6 were added into HSC-F cells (a cynomolgus macaque T-cell line) [37], which were cultured for 5-7 days to obtain the culture supernatants as PBMC-derived virus stocks. Second, plasma samples were concentrated by 6-fold using Lenti-X Concentrator (Clontech) and cocultured with HSC-F cells for 6-14 days to obtain the culture supernatants as plasma-derived virus stocks. To prepare the wild-type virus stock, we first obtained culture supernatants from MT4 cells (a human T-cell line) expressing CCR5 after transfection with the wild-type SIVmac239 molecular clone DNA (pBRmac239) [38]. Then, these supernatants were added into HSC-F cells and the culture supernatants were obtained as the wild-type SIVmac239 stock, which was used for comparison of in vitro viral fitness with PBMC-derived and plasma-derived viruses. In addition, we constructed recombinant SIV clones by replacing the gag region (nt 1056-3408) in the wild-type pBRmac239 molecular clone with that amplified from plasma RNAs of macaques #31 and #32 at 1 year post-infection. We then obtained recombinant molecular clones whose gag had the predominant 3pSIV1 and 3pSIV2 sequences, respectively. COS-1 cells were transfected with these molecular clones to obtain recombinant SIV3p1gag and SIV3p2gag virus stocks. COS-1 cells were transfected with pBRmac239 to obtain the wild-type SIV-mac239 stock used for comparison of in vitro viral fitness with these recombinant viruses, SIV3p1gag and SIV3p2gag. Titers of these virus stocks were measured by RT assay as described previously [39,40]. For analysis of in vitro replication capacity, HSC-F cells were infected with these viruses (5 x 10 4 HSC-F cells were infected with viruses having the same RT activity with the wild-type SIVmac239 corresponding to 0.2 ng of p27), and RT activity of the culture supernatants on day 4 post-infection was measured. In the competition assay for comparison of in vitro replication capacity of two kinds of viruses, HSC-F cells infected with individual virus stocks were cocultured to determine which viral genome sequences become dominant in the culture supernatants. HSC-F cells were infected with individual virus stocks (normalized by RT activity) and their coculture started next day. Coculture was continued by transferring the culture supernatant into fresh HSC-F cells every 4 days. RNA was extracted from the coculture supernatant and the Gag CA (capsid)-coding region was sequenced. When only one viral sequence became dominant on day 2 after the coculture initiation, we confirmed that the virus became dominant even in the coculture of the virus-infected cells with larger numbers of the other virus-infected cells in which both viruses were equivalently detected on day 2.

Statistical analysis
All statistical analyses were performed using Prism software (GraphPad Software, Inc.) with significance set at p values of < 0.05. Comparisons were performed by Mann-Whitney U-test or log-rank test.