HIV-1 Vpr Induces Widespread Transcriptomic Changes in CD4+ T Cells Early Postinfection

ABSTRACT The interactions between a virus and its host are complex but can be broadly categorized as either viral manipulation of cellular functions or cellular responses to infection. These processes begin at the earliest point of contact between virus and cell and frequently result in changes to cellular gene expression, making genome-wide transcriptomics a useful tool to study them. Several previous studies have used transcriptomics to evaluate the cellular responses to human immunodeficiency virus type 1 (HIV-1) infection; however, none have examined events in primary CD4+ T cells during the first 24 h of infection. Here, we analyzed CD4+ T cells at 4.5, 8, 12, 24, and 48 h following infection. We describe global changes to host gene expression commencing at 4.5 h postinfection and evolving over the ensuing time points. We identify upregulation of genes related to innate immunity, cytokine production, and apoptosis and downregulation of those involved in transcription and translation. We further demonstrate that the viral accessory protein Vpr is necessary for almost all gene expression changes seen at 12 h postinfection and the majority of those seen at 48 h. Identifying this new role for Vpr not only provides fresh perspective on its possible function but also adds further insight into the interplay between HIV-1 and its host at the cellular level.

A s obligate intracellular parasites, viruses require many elements of cellular machinery to successfully complete their life cycles, often referred to as dependency factors. Beyond this, to replicate efficiently, it is often necessary for viruses to actively modify host cell processes from the earliest point of contact with the cell. Illustrative examples of such viral manipulation of early events include vaccinia virus initiation of macropinocytosis by mimicry of apoptotic bodies, use of multiple cellular kinases by simian virus 40 (SV40) to promote entry through caveolar endocytosis and transport to the smooth endoplasmic reticulum, and repression of cellular transcription by the matrix protein of vesicular stomatitis virus (1)(2)(3)(4)(5). Conversely, host cells have evolved myriad mechanisms to impede viral infection at the earliest possible opportunity and limit subsequent damage; the antiviral protein effectors of inhibition are often referred to as restriction factors (6). The wide and varied signaling cascades that ensue from these interactions inevitably change cellular transcription and RNA expression, making transcriptomics an important approach for acquiring a broad and unbiased overview of host-pathogen interactions.
A number of prior studies have assessed transcriptomic changes following human immunodeficiency virus type 1 (HIV-1) infection in different contexts. Studies can be grouped into those examining gene expression in vivo in infected individuals and those examining RNA levels following infection of cells in vitro. The in vivo studies of humans or following simian immunodeficiency virus infection of nonhuman primates have shown an upregulation of genes associated with innate immunity. Many of these studies have also noted differential expression of genes related to the cell cycle but have differed in their conclusions as to the consequences of these changes (7)(8)(9).
Transcriptomic studies of cells infected in vitro with HIV-1 can be further divided into those studies performed using immortalized cell lines and those performed on primary CD4 1 T cells ex vivo. There are many studies of the former genre, and these have been reviewed elsewhere (10,11); in sum, no unifying picture of genome-wide changes in RNA expression has emerged, perhaps due to the use of different cell lines, viral isolates, and experimental workflows. To our knowledge, three studies using primary CD4 1 T cells and full-length infectious virus have been performed. The first concluded that effector memory T cells and cells with high levels of active AP-1 (a transcription factor) were more permissive for infection (12). The second examined changes to transcription at 48 h postinfection and identified increases in the expression of genes related to innate immunity (13). Significant differential expression of genes related to apoptosis and cytokine production was also noted. A third, more recent, study sought to assess the impact of Vpu on cellular gene expression 72 h postinfection; however, following infection with wild-type virus, this study demonstrated upregulation of genes associated with antiviral immunity and downregulation of genes in ontology categories, including the spliceosome, RNA transport, and ribosome biogenesis (14).
Given the potential of transcriptomics to identify dependency and restriction factors, we employed this approach to investigate the very earliest events during infection. While one previous study documented transcriptional changes in the first 24 h of HIV-1 infection in SupT1 cells (15), none have investigated primary T cells. We therefore infected primary CD4 1 T cells with HIV-1 and assessed genome-wide RNA expression at 4.5, 8, 12, 24, and 48 h following infection. Specific changes were evident by 4.5 h, and the identities and functions of affected genes varied over the subsequent time points. Further infections using primary memory CD4 1 T cells and different HIV-1 strains, including transmitted founder viruses, confirmed these observations. Finally, we sought to identify the viral factors responsible for initial transcriptomic reprogramming and, for the first time, defined the accessory protein Vpr as necessary for almost all cellular gene expression changes occurring within the first 12 h of HIV-1 infection.

RESULTS
HIV-1 induces rapid genome-wide transcriptome changes following infection of primary CD4 + T cells. There has been limited analysis of the primary CD4 1 T-cell transcriptional response to HIV-1 infection in vitro, and prior work has focused on gene expression changes occurring between 24 and 72 h postinfection (12)(13)(14). Other studies have used viral antigen (e.g., gp120) or T-cell lines to examine transcriptional responses in the first 24 h postchallenge (15)(16)(17). We sought to build upon these observations by examining responses in primary CD4 1 T cells using replication-competent virus at multiple early time points across multiple donors. Primary CD4 1 T cells were isolated from the blood of healthy volunteers and activated for 40 h before challenge with either no virus (mock), the wild-type CXCR4 (X4)-tropic strain HIV-1 IIIB (herein called IIIB), or an engineered derivative lacking a functional env gene, HIV-1 IIIBDEnv (IIIBDEnv), that is unable to enter cells. Excess input virus was removed by extensive washes with phosphate-buffered saline (PBS) 3 h postinfection, and an aliquot of cells was used to determine the efficiency of infection by intracellular p24 Gag staining and flow cytometry. On average, 81% of cells exposed to wild-type IIIB were positive for intracellular p24 Gag , whereas less than 1% were stained following IIIBDEnv challenge (see Fig. S2 in the supplemental material). Subsequently, at 4.5, 8, 12, 24, and 48 h postinfection, cells were harvested, total RNA was isolated, and gene expression was determined using Illumina's BeadArray HT12v4 (Fig. 1A). Of the 47,324 probes on the array (representing 34,697 gene loci), 26,079 detected expression significantly (B) Line plots that illustrate the qualitative difference in the degrees of observed gene expression changes between IIIBDEnv-(left) and IIIB-infected (right) cells relative to mock-infected cells. Each line corresponds to the gene expression changes detected by a particular microarray probe at each of the five time points. The set of probes shown on the left and right is the same (n = 1,303) and corresponds to the 5% of microarray probes that detected the greatest gene expression changes overall. Specifically, all probes were ranked according to the amount of gene expression change (F statistic) detected across the 10 pairwise comparisons (two by five time points) between mock-infected and either IIIBDEnv-(left) or IIIB-infected (right) cells. Therefore, probe selection was independent of experimental condition. Nevertheless, the selected 5% of probes detected gene expression changes greater than 2-fold in IIIB-but not in IIIBDEnv-infected cells. Note that, while overall differential expression for a probe may have been high, differential expression at an individual time point may still be zero.
Vpr Modifies T-Cell Gene Expression ® above the background noise level, defined as the global median signal intensity, under at least one of our experimental conditions (false-discovery rate [FDR] , 20%). They represent the gene expression "universe" of our studies.
Across the entire time course, we found the mock infection and IIIBDEnv experimental conditions to yield equivalent results. Even at an FDR threshold of 80%, at most 2 probes detected large (absolute log 2 fold change [jLFCj] . 1) expression differences between mock-and IIIBDEnv-infected cells at any of the time points (Fig. S3). We therefore used both conditions to control for infection-unrelated, time-driven gene expression changes. A total of 5,280 probes detected significant (FDR , 5%) expression changes in cells infected with wild-type virus at some point during the time course relative to the expression levels in these controls (Data Set S1). Of these probes, 1,411 detected large changes (jLFCj . 1), starting with 176 probes at 4.5 h postinfection ( Fig. 1B and 2A and B). By 8 h postinfection, this number had increased to 455, reaching a peak of 817 probes at 24 h and decreasing to 583 probes by 48 h.
The identities of the genes affected by infection varied over time, sometimes within relatively short time spans. Considering all intersections between the time point-specific sets of probes detecting either up-or downregulation of gene expression ( Fig. 2A and B), we found that, for example, at 4.5 h, most differentially expressed genes (DEGs) fell into two categories: genes either up-or downregulated exclusively at 4.5 h (77 of 176 probes) and upregulated genes that remained so through to the 48-h time point (33 of 176 probes). The four largest intersections also pointed to dynamic changes in the set of DEGs over time; an exceptionally large group of genes (163 probes) were downregulated only at 24 h, accompanied by a large group of upregulated genes (124 probes) that also were specific to the 24-h time point (including a notable upregulation of interferon [IFN]-stimulated genes [ISGs]) ( Fig. 2A). Genes upregulated from 8 h onwards gave rise to the largest group of genes (109 probes), which remained upregulated for the rest of the time course; finally, many upregulated DEGs (205 probes) were observed only at 48 h, not earlier. There were small numbers of genes that appeared to be upregulated or downregulated at nonconsecutive time points. While this may be an artifact of the analysis generated by setting a 2-fold cutoff, the possibility that these genes truly have cyclical expression cannot be excluded. It should be noted that the frequency with which genes were observed to be differentially expressed at any combination of nonconsecutive time points was significantly less than that which would be observed by chance (P , 1 Â 10 266 ). Therefore, even if all of these observations represent true cycle expression, this phenomenon occurs significantly less frequently than differential expression at consecutive time points. In sum, HIV-1 infection and entry into CD4 1 T cells triggers rapid, dynamic, and genome-wide changes in the host transcriptome; as a general trend, the magnitude of expression changes tended to be greater for upregulated genes than for downregulated genes (Fig. 1B).
WGCNA. To identify and characterize infection-dependent temporal patterns of gene expression without fixed significance and fold change thresholds, we applied weighted gene coexpression network analysis (WGCNA) to group together probes whose gene expression measurements correlated across samples into so-called modules. WGCNA is an unsupervised method that does not take into account the time point or experimental condition represented by each sample. From our previously defined gene universe, WGCNA grouped 15,407 probes into 35 modules. Of these, 20 modules exhibited an average gene expression profile over time for IIIB-infected cells that significantly (FDR , 5% [see Materials and Methods]) differed from those for mock-and IIIBDEnv-infected cells (Fig. 3A). A large proportion (4,108/7,443; odds ratio [OR] = 18.35; P , 10 215 ) of the probes comprising these 20 modules were among the 5,280 probes that detected significant (FDR , 5%) differential expression between IIIBand mock-or IIIBDEnv-infected cells in our time point-specific tests, indicating that these 20 modules represent most of the temporal gene expression patterns that are characteristic of the ex vivo CD4 1 T-cell response to infection with HIV-1. Indeed, and as expected, substantial numbers of probes grouped into modules (e.g., modules 3, 4, We applied gene ontology (GO) enrichment analysis to the genes represented by the probes in each module to assess the degree to which each characteristic temporal gene expression pattern represents known functional categories (Fig. 3B). Immune cell activation and proliferation categories were enriched for genes upregulated from 4.5 h postinfection onwards (module 16). Innate immune response categories were enriched for genes upregulated later, from 24 h onwards (module 12). p53-mediated apoptosis was enriched for genes upregulated from 8 h, with upregulation increasing over time (module 4). Transport-related categories were enriched in a small group of upregulated genes with overall low absolute expression levels (module 9). Four modules with genes downregulated from 12 to 24 h postinfection substantially overlapped in terms of enriched GO categories that mostly related to translation but also related to viral transcription and nonsense-mediated decay (modules 1, 13, 14, and 18). For module 3, to which the highest number of genes grouped, only a single GO category, protein heterotetramerization, was enriched. This enrichment was due to differential expression of a number of histone genes.
Application of the WGCNA method to previously published primary T-cell data showed upregulation of similar groups of genes ( Fig. 3C). An upregulation of genes related to the type I IFN response has also been noted in multiple previous analyses of gene expression in CD4 1 T cells isolated from viremic patients, though it should be recognized that the majority of CD4 1 T cells are not, themselves, infected in this context (7,18). The downregulation of genes related to transcription and translation has been identified in previous studies of HIV-1 infection (14,15,19). The previously identified Nef-mediated upregulation of genes involved in cholesterol metabolism was also recapitulated in our data (20). Ontology categories enriched among genes in module 9 have not prominently featured in previous studies of HIV-1 infection in either primary cells or cell lines. Quantitative real-time PCR (qPCR) was then used to verify the differential expression of 3 protein-coding genes (PTK2, ZBP1, and STAP1) that were identified as consistently upregulated in infected cells by microarray analysis and an expressed sequence tag (NCBI accession no. AF086468), the corresponding microarray probe for which had detected the highest upregulation of all probes in the data set ( Fig. 4). These probes were not chosen based on the WCGNA; however, AF086468 and ZBP1 fall into module 9, PTK2 falls into module 12, and STAP1 falls into module 5.
Host gene expression modulation is a shared feature of CXCR4-and CCR5-tropic viruses. To ensure that the observed transcriptomic response was not idiosyncratic to IIIB, total CD4 1 T cells were isolated from three donors and challenged with either IIIB, BK132 (an X4-tropic primary isolate of HIV-1), or CH077.t (a dual tropic transmitted founder virus). Gene expression in infected cells was analyzed by qPCR at 12 and 48 h postinfection and compared with that after mock infection; a response to all three viral strains was demonstrated ( Fig. 5A and B).
Previous studies have shown that the principal target cells for HIV-1 infection in vivo are memory CD4 1 T cells and that transmitted HIV-1 strains are predominantly CCR5 (R5) tropic (21,22). For these reasons, CD4 1 CD45RO 1 memory T cells were isolated from three further donors and infected with either IIIB, Ba-L (a CCR5-tropic primary isolate of HIV-1), CH077.t, CH106.c, or REJO.c (the last two of which are CCR5-tropic transmitted founder viruses). RNA expression analyses confirmed a similar transcriptional response in memory CD4 1 T cells to all viral strains tested (Fig. 5C).

FIG 2 Legend (Continued)
all possible mutually exclusive and potentially nonempty subsets of probes that can be generated via the intersection of the five time point-specific sets of probes (n = 2 5 -1 = 31). By definition, the probes in a particular subset detected significant change in expression at a specific set of time points, indicated below the vertical bar by a column of filled black circles connected by lines, while not detecting significant change at the other time points, indicated by the unconnected light-gray circles in the same column. This is equivalent to a five-way Venn diagram with areas proportional in size to the numbers of probes in each subset. For example, 112 probes in total detected significant upregulation at 4.5 h after infection (first horizontal bar). Of these, 42 probes (1st column) did so exclusively at 4.5 h, while 3, 4, 6, and 33 probes (2nd to 5th columns) did so exclusively at 4.5 and 8 h, 4.5 and 8 and 12 h, 4.5 and 8 and 12 and 24 h, and all time points, respectively. The rapid gene expression changes that occur by 12 h are interferon independent. The analysis described above identified an upregulation of ISGs from 24 to 48 h postinfection (Fig. 3B). This observation agreed with previous studies of the transcriptomic response of primary CD4 1 T cells both in vivo and ex vivo (9,12,13,18). It was not, however, possible to identify increased expression of type I IFNs themselves by microarray analysis, qPCR, or bioassay (data not shown for the last two). Accordingly, to examine the possible contributions of IFN and IFN-induced signaling to ISG expression changes at 12 and 48 h postinfection, IIIB infections of CD4 1 T cells were carried out in the presence or absence of an antibody specific for the type I IFN receptor (MMHAR-2) that efficiently blocks IFN-induced signaling, and RNA levels of a subset of ISGs (ISG15, IFN-induced protein with tetratricopeptide repeats [IFIT1], and 29-59-oligoadenylate synthetase 1 [OAS1]) were then monitored by qPCR. The addition of MMHAR-2 had a negligible impact on HIV-1-induced gene expression at 12 h but  partially suppressed induction when measured at 48 h (Fig. 6). We interpret these data as revealing that the HIV-1-induced expression changes seen at 12 h occur independently of IFN but that IFN contributes to the activation of ISG expression at later time points.
The HIV-1 accessory protein Vpr initiates cellular RNA expression changes during the initial hours of infection. The observed early transcriptional response was likely to have been stimulated by the constituents of incoming virions. In particular, genomic RNA, the DNA products of reverse transcription, and the capsid lattice are known to be sensed by pattern recognition receptors and to stimulate innate immune signaling (23)(24)(25)(26)(27)(28). For this reason, a series of mutated viruses were used to challenge primary CD4 1 T cells, and differential RNA expression was monitored at 12 h by qPCR. Specifically, IIIBDEnv (Fig. 1B) served as the negative control, IIIBDW has a disrupted guide RNA (gRNA) packaging signal and encapsidates reduced levels of gRNA (29,30), IIIBDRT carries an inactivated reverse transcriptase (RT) gene and is unable to synthesize viral cDNA, IIIBDTat lacks the transcriptional transactivator tat and is severely Vpr Modifies T-Cell Gene Expression defective for viral transcription (31), and IIIBDVpr lacks the accessory protein Vpr, an ;14-kDa nonstructural protein that is selectively incorporated into nascent virions via an interaction with the p6 region of p55 Gag (32-34) but whose functions in the context of virus replication are not fully understood (35)(36)(37). Disruption of gRNA packaging, RT, or Tat had negligible effects on virus-induced alterations to gene expression, but the removal of Vpr abrogated the cellular response (Fig. 7A). It was formally possible that the loss of the inductive capability of IIIBDVpr was due to the mutation of viral sequence rather than the loss of the Vpr protein itself. To address this, IIIBDVpr was complemented in the cultures used to produce viral stocks by cotransfection with Vpr expression vectors corresponding to three different HIV-1 isolates. In each case, the activation of host gene expression at 12 h in CD4 1 T cells was efficiently restored to levels comparable to those of wild-type virus (Fig. 7B), Vpr Modifies T-Cell Gene Expression ® demonstrating that the initiating signal for the transcriptomic changes seen in primary CD4 1 T cells is Vpr dependent. Lastly, the efficient incorporation of these Vpr proteins in trans into IIIBDVpr particles was confirmed using purified virions and immunoblotting (Fig. 7C).
Transcriptomic analysis of Vpr-regulated genes. To more broadly determine the role of Vpr, repeat experiments with primary CD4 1 T cells were analyzed using Illumina's BeadArray HT12v4 (Fig. 8). At the 12-h time point, almost all differential gene expression seen with the wild-type virus was lost in the absence of Vpr but restored when Vpr was provided in trans. At 48 h postinfection, a low level of differential expression was observed with IIIBDVpr; however, the response remained far more limited than in the presence of Vpr (Fig. 8A and B). As expected, at 12 h postinfection, there was a large degree of overlap in the identities of cellular genes upregulated following challenge with IIIB and IIIBDVpr plus Vpr (in trans), additionally confirming that virionpackaged Vpr was responsible for inducing the observed transcriptome response (Fig. 9). At 48 h postinfection, a larger number of genes was seen upregulated following IIIB infection but not after infection with IIIBDVpr plus Vpr (in trans). Previous studies have suggested that Vpr may play a role in either eliciting or suppressing a type I IFN response following HIV-1 infection (38)(39)(40)(41).

DISCUSSION
Studying the transcriptomic response of cells to viral infection in terms of the genes that may promote or impede infection can provide new insights into host-pathogen interactions. While there are a number of prior studies for HIV-1, few have been conducted with primary CD4 1 T cells, and none of these have interrogated the first 24 h of infection. We therefore sought to expand on published data by examining transcriptional responses in primary CD4 1 T cells across a range of early time points using X4and R5-tropic viral strains. By doing so, we showed for the first time that changes to cellular gene expression in the early hours following infection are triggered by Vpr.
Our study identified patterns of differential gene expression that were broadly similar to those found in previously published data sets (11, 13-15, 19, 42, 43). In particular, we documented an upregulation of genes associated with the innate immune response, cytokine production, and apoptosis (13,42,43). The downregulation of genes related to transcription and translation that we observed has previously been reported both in primary CD4 1 T cells and T-cell lines (14,15,19,42). In addition, of particular interest, an overrepresentation of histone genes led to protein heterotetramerization being identified in our gene ontology enrichment analysis. The confirmation of transcriptomic responses to infection in memory as well as total primary CD4 1 T cells and with multiple X4-and R5-tropic viral strains further validated the generality of our findings.
One previous study has provided a detailed examination of the early (first 24 h) transcriptional response of SupT1 cells to HIV-1 infection (15). The study identified groups of genes with different temporal patterns of expression, many of which we similarly observed, for example, genes that were downregulated throughout infection, genes with rising levels of upregulation across successive time points, and genes that were specifically upregulated later in infection. Though direct comparison of the identities of genes in these similar groups was not possible owing to data inaccessibility, a minimum of overlap might be anticipated due to differences in cell types and the methods used for analysis (13). With regard to overall trends, both our study and this previous work noted initial downregulation of genes associated with transcription and translation and later upregulation of genes involved in immune signaling.
Our most striking finding was the dependence of initial gene expression changes on Vpr. Arguably the least well understood of the HIV-1 accessory proteins, Vpr has previously been associated with a number of activities, most prominently the arrest of the cell cycle in G 2 /M (44-47). It has also been linked to the modulation of innate immune signaling and enhanced gene expression from HIV-1 and cellular promoters, both activities that potentially impact cellular transcriptional profiles. With respect to signaling, multiple studies have examined the role of Vpr in NF-k B signaling pathways. However, they disagree as to whether Vpr promotes (48,49) or inhibits (50-52) NF-k B signaling and at which point in the pathway this occurs (48,50,52). Equally, in the context of primary CD4 1 T-cell infection with replication-competent virus, other viral proteins have been shown to play a role (14). In addition, a previous study of the transcriptional response of monocyte-derived macrophages (MDMs) to Vpr suggested an upregulation of genes associated with innate immunity (53). However, this is in apparent contrast with other studies showing that HIV-1 evades stimulation of the type 1 IFN response in MDMs and that Vpr restricts the translocation of IRF3 to the nucleus in monocytic cell lines (52, 54).  (55,56). Although the heterogeneity of prior studies makes direct comparisons with our results challenging, our findings, overall, demonstrate a central role for Vpr in the broad modulation of cellular mRNA expression early postinfection in a pattern that has not previously been described.
Vpr has been shown to co-opt the cellular E3 ubiquitin ligase consisting of cullin 4A (CUL4A), DNA damage binding protein 1 (DDB1), and CUL4A-and DDB1-associated factor 1 (DCAF1) to degrade cellular targets (57)(58)(59)(60)(61)(62). Recent evidence supports the hypothesis that many of these are recruited by interaction with their DNA binding domains (63)(64)(65). One recent extensive proteomic study used multiple mass spectrometry approaches to evaluate infection with viruses in the presence or absence of Vpr and identified many new primary protein targets of Vpr, a proportion of which have the potential to directly modify transcription (65). Other publications have shown DCAF1-dependent degradation of proteins capable of epigenetic modification, including class I histone deacetylases (66,67), ZGPAT (Zinc finger CCCH-type and Gpatch domain containing) (68), and TET2 (10-11 translocation) (69). Two groups have also recently shown Vpx (but not HIV-1 Vpr)-mediated degradation of members of the human silencing hub (HUSH) complex, which resulted in reactivation of HIV-1 in a latency model and more rapid replication kinetics in spreading infection (70,71). These studies may therefore provide clues as to how Vpr may promote the transcriptional reprogramming of genome-wide gene expression and will help direct future work examining molecular mechanisms. It is also not yet clear whether the Vpr-mediated modification of the cellular transcriptome is beneficial or detrimental to HIV-1 replication, and answering this question represents a further priority.
A notable (though experimentally necessary) limitation of our study was the relatively high viral inoculum used to generate a synchronously infected population of cells where transcriptomic responses could be monitored during the initial phases of infection. Whether cells are naturally exposed to such high levels of viral particles in vivo is not clear, though high multiplicities of HIV-1 infection have been reported in the context of lymphoid tissue (72). Equally, it was not possible to show how many cells identified as p24 Gag positive at 3 h postinfection were productively infected. However, experiments using DVpr virus complemented with Vpr during viral production (Fig. 8), together with a panel of replication-deficient viruses (Fig. 7), demonstrated that early changes to transcription were a consequence of virion-associated Vpr, regardless of whether infection was ultimately productive. Future work using methodologies at the single-cell resolution will seek to provide additional insights into these observations, though the general correspondence between our findings and the patterns of transcription described in previous studies provides important overall confirmation.
The in vivo importance of Vpr is clearly demonstrated by its conservation among lentiviral lineages, attenuated infection in its absence, and reversion of disrupted alleles over the course of infection in both humans and animals (73)(74)(75)(76)(77)(78)(79). However, the underlying basis for this importance remains elusive. Our demonstration that encapsidated Vpr has a profound effect on the CD4 1 T-cell transcriptome in the earliest phases of HIV-1 infection therefore casts new light on previous transcriptomic analyses of HIV-1 infection and highlights a new and important aspect of the interaction between HIV-1 and its host. Further studies to determine whether this property is conserved among the Vpr proteins from other lentiviral lineages would be of interest.

MATERIALS AND METHODS
HIV-1 molecular clones and plasmids. The wild-type HIV-1 proviral clone pIIIB and its env-deficient derivative pIIIBDEnv have been described previously (80)(81)(82). The pIIIBDW mutant was generated by deleting a portion of SL2 and the entire SL3 region (i.e., between nucleotides 750 and 788, in accordance with the work of Aldovini and Young [29]). Point mutations were introduced into the IIIB reverse transcriptase catalytic site (YMDD338-341YAAA) to generate the pIIIBDRT mutant. pIIIBDVpr and pIIIBDTat were generated by introducing 2 stop codons at the beginning of the vpr or tat open reading frames with point mutations (G5629T and G5632T for vpr and ATCATC5855-5860TAGTAG for tat). The transmitted founder HIV-1 molecular clones CH077.t, CH106.c, and REJO.c were kindly provided by Beatrice Hahn (83). The HIV-1 Ba-L and BK132 virus isolates were obtained from the NIH AIDS Reagent and Reference Program (generously provided by Suzanne Gartner, Mikulas Popovic, and Robert Gallo [84] and Nelson Michael [85], respectively).
Amino-terminally Flag-tagged vpr genes from NL4-3, CH077.t, and Ba-L were amplified by PCR from proviruses (NL4-3, CH077.t) or by reverse transcription-PCR from Ba-L-infected CCR5 Jurkat cells and cloned into pcDNA3.11. pcDNA3.11Tat was made by PCR amplification of IIIB Tat from pcTat (86) and cloning into pcDNA3.1 using HindIII and XbaI and was a kind gift from Chad Swanson.
Cells and cell culture. Human primary CD4 1 T cells were obtained from peripheral blood mononuclear cells (PBMCs) of healthy volunteer donors (approved by the Guy's research ethics committee [reference no. 03/02/06]). Briefly, PBMCs were isolated using Lymphoprep (Axis-Shield), and total or memory CD4 1 T cells were then isolated from the remaining cells using CD4 1 T-cell isolation kit II or the memory CD4 1 T-cell isolation kit, respectively (Miltenyi Biotec). The purity of isolated cell populations was always .90%, as judged by flow cytometric analysis of specific markers (CD4 and CD3 or CD4 and CD45RO for total or memory T cells, respectively) (see Fig. S1 in the supplemental material). CD4 1 T cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco) supplemented with 10% autologous human serum and stimulated for 40 h with anti-CD3 and anti-CD28 antibodies (BD Biosciences; final concentration, 1 to 2 mg/ml) prior to infection, and the antibodies were readded upon medium Vpr Modifies T-Cell Gene Expression ® changing after infection. Human 293T cells were maintained in complete Dulbecco's modified Eagle's medium (DMEM) (Gibco) plus 10% fetal bovine serum (FBS), and Jurkat T cells were grown in RPMI 1640 medium supplemented with 10% FBS.
Viral production. Wild-type or mutant HIV-1 IIIB , HIV-1 CH077.t , HIV-1 CH106.c , or HIV-1 REJO.c were produced by standard polyethylenimine (PEI) transfection of 293T-cell monolayers. IIIBDTat was produced by cotransfection of the corresponding provirus and pcDNA3.11Tat at a ratio of 1:0.2. To test the effect of Vpr proteins from other strains, pIIIBDVpr provirus was cotransfected with pcDNA3.11Vpr expression plasmids at a ratio of 1:0.25. The culture medium was changed at ;6 h, and the virus-containing supernatant was harvested at ;36 h. HIV-1 Ba-L and HIV-1 BK132 were grown in CCR5-expressing Jurkat cells as previously described (87).
Viruses were filtered and incubated with 20 U/ml RQ1 DNase (Promega) and 10 mM MgCl 2 for 3 h at 37°C. Viruses were then purified by ultracentrifugation through a sucrose cushion (20%, wt/vol; 75 min of 4°C at 145,370 Â g using a Sorvall Surespin630 rotor), resuspended in RPMI 1640 medium without serum, and stored in aliquots at 280°C. Viral particles were normalized according to the results of an HIV-1 p24 Gag enzyme-linked immunosorbent assay (ELISA) (Perkin Elmer).
HIV-1 infection. Infections were initiated by spin infection of the cells with HIV-1 particles (0.5 to 1mg/ ml p24 Gag ) for 2 h at 2,000 Â g and at 30°C (89). The cells were then incubated at 37°C for 1 h prior to extensive washes in PBS. Fresh and prewarmed medium (containing anti-CD3 and -CD28 antibodies) was added, and the cells were placed back at 37°C for the indicated incubation times, prior to RNA extraction.
The percentage of cells in which the virus had efficiently entered was routinely analyzed by p24 Gag intracellular staining of cells at 3 h postinfection (i.e., at the time of the PBS washes) and flow cytometry (FACSCalibur; BD Biosciences). Briefly, the cells were washed in PBS, incubated for 10 min in prewarmed trypsin to remove surface-associated virion particles, fixed, permeabilized (IntraStain kit; Dako), and stained with a p24 Gag -specific antibody (KC57-RD1; Beckman Coulter) (90). The percentage of p24 Gagpositive cells was typically .70% for IIIB (Fig. S2). This antibody does not recognize p24 Gag from HIV-1 CH106.c or HIV-1 REJO.c and was therefore not used on memory T-cell infections with these viruses. For experiments assessing the role of IFN in gene expression, a monoclonal antibody specific for the type 1 IFN receptor (MMHAR-2; Invitrogen) was added to the indicated wells at a final concentration of 10 mg/ ml 1 h before infection. For conditions treated with IFN-a (Intron A; Merck, Sharpe & Dohme Corp.), this was added to a final concentration of 100 U/ml at the start of the infection.
Microarray data generation and analysis. Primary CD4 1 T cells (2 Â 10 5 to 2 Â 10 6 ) were harvested at the indicated times postinfection. RNA was isolated using the miRNeasy kit with on-column DNase treatment (Qiagen). One hundred to 500 ng RNA was used for cRNA probe preparation using the Illumina TotalPrep RNA amplification kit (Ambion) according to the manufacturer's instructions. The probes were hybridized on Illumina HT12v4 bead arrays by following the manufacturer's standard hybridization and scanning protocols.
Raw signals, detection P values, bead numbers, and bead-level standard errors were exported for regular and control probes from GenomeStudio (Illumina). Data were imported into R using the Bioconductor bead array and illuminaHumanv4.db packages. For the purposes of preprocessing, normalization, and removal of probes that did not detect a signal above background, we used a larger data set than is presented here. All arrays used for this analysis have been deposited in GEO as described below. The raw signal intensity distributions and the signal intensities for control probes (labeling, hybridization, housekeeping gene, and negative controls) were nominal for all but one array; signal intensities for the negative control and regular probes alike were elevated for array 9031292066_A. Signal intensities were background corrected, quantile normalized, and log 2 transformed using the limma (91) implementation of the neqc method Shi (92). Unsupervised hierarchical clustering of the normalized data did not identify any outliers but rather a systematic array batch effect (November 2012 versus March 2013 batches) so that batch always was included as a covariate in subsequent linear models of data comprising arrays from both batches. Array 9031292066_A, in particular, belonged to an unremarkable cluster of other arrays from the same batch assaying the same cell type and, therefore, was retained for analysis. Multidimensional scaling analysis of the array measurements for XIST and for highly expressed genes on chromosome Y exhibiting a gender-corresponding, bimodal, expression-level distribution (KDM5D, TXLNG2P, EIF1AY, RPS4Y1, RPS4Y2, UTY, USP9Y, PRKY, ZFY) bipartitioned arrays into two distinct genderspecific clusters. Cluster membership always corresponded to the annotated proband (donor) gender of the sample; i.e., no sample swaps became apparent.
Probes that did not detect a signal significantly above the global (for all arrays and all probes) signal median (considered the background) under any of the distinguished experimental conditions were excluded from analysis. Specifically, we fitted three zero intercept linear models that included batch as a covariate and the following experimental factors: (i) cell type (total versus memory CD4 1 T cells), (ii) cell type and infection state (infected with HIV-1 versus mock infected with IIIBDEnv or no virus), and (iii) cell type, infection state, and time point after infection. Separately for each model and experimental condition, probes were tested for above-background signals (limma treat function [93] with LFC, which is the global median). Probes detecting an above-background signal in any of the conditions of any of the models at a significance threshold FDR of ,20% were retained for analysis.
Array-quality weights were inferred with the limma arrayWeightsSimple method (94) using a model with batch and proband as covariates and cell type, infection status, and time point as the experimental factors. The array weights were used in all subsequent applications of the limma lmfit model fitting function. All linear models addressing biological questions (Table 1) also included proband as a covariate to account for interindividual differences in gene expression and/or in probe hybridization efficiency due to genetic variants from the probe sequence. For each model, multiple-testing correction was carried out globally using the Benjamini-Hochberg method for controlling the FDR (limma decideTests with method set as "global" and adjust.method set as "BH"). The bar chart visualizations of the sizes of sets of differentially expressed genes and their intersections were generated with a customized version of UpSetR v1.0 (95).
WGCNA. The weighted gene coexpression network analysis (WGCNA) tool (96) was utilized. Since WGCNA employs an unsupervised algorithm, unwanted sources of variation were removed prior to its application. Specifically, the effects of individual proband (donor) were removed using the ComBat function of the sva package (97). The data set comprised more control than infected samples. To avoid a disproportionate contribution of control samples to gene module definition, the data were separately analyzed in two parts: (i) infected samples and uninfected controls and (ii) infected samples together with DEnv-challenged samples separately. The full WGCNA workflow was then applied for identification of gene expression modules. First, adjacency matrices were generated from probe-level expression data using the pickSoftThreshold and adjacency functions. These matrices were converted to distance matrices before construction of topological overlap matrices (TOM) with the TOMsimilarity function. The resultant TOMs (one generated using the uninfected controls and one using the DEnv controls) were scaled in order to equate the 95th percentiles, and a single consensus TOM was generated by taking the minimum topological overlap of the two input TOMs for all pairwise probe comparisons. This allowed an "average" gene expression network of the two control conditions (uninfected controls and DEnv controls) to be captured. The resultant TOM was used for hierarchical clustering prior to module definition via the cutTreeDynamic function/algorithm (96). Per-sample-per-module summarizations were generated via principal-component analysis (PCA), using the first eigenvector (also known as the eigengene). Modules with a pairwise Pearson correlation coefficient of .0.75 were combined.
Statistics. Differences in levels of expression of specific genes between conditions measured by qPCR were assessed for statistical significance using an unpaired Student t test, with individual variances computed for each comparison, and P values were adjusted using Holm-Šídák's multiple-comparison test. Calculations were performed in Prism version 9.0.0.
Data availability. The microarray data are available under GEO accession no. GSE166375.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.