We recently described HyPR-MS (Hybridization Purification of RNA-Protein Complexes Followed by Mass Spectrometry), a strategy to identify the in vivo protein interactomes of specific viral RNAs, lncRNAs, and mRNAs (Knoener et al., 2017; Spiniello et al., 2018; Spiniello et al., 2019). Here, we present HyPR-MS_sv, a strategy that expands the capabilities of HyPR-MS to differentiate in vivo protein interactomes for multiple splice variants (SV) derived from a single primary transcript and isolated from a single-cell population. Applied here, we purified the three major classes of HIV-1 splice variants (US, PS, and CS) from a single population of infected Jurkat CD4+ T-cells, then identified and characterized their protein interactomes using mass spectrometry.

To preserve in vivo viral RNA-protein complexes prior to cell lysis, Jurkat cultures were treated with formaldehyde at 48 hr post-infection (h.p.i.) (multiplicity of infection [MOI] of ~1 infectious unit per cell). The Jurkat cell line was chosen because it is a well-characterized CD4+ T-cell line previously confirmed to support replication of the HIV-1_NL4-3 reporter virus used for this study (Knoener et al., 2017). To isolate the US, PS, and CS RNA pools, three biotinylated capture oligonucleotides (COs) were designed complementary to three distinct regions of the HIV RNA genome: intron-1 (unique to US), intron-2 (present in both US and PS), and the 3’-exon (present in US, PS, and CS) (Figure 1A, Supplementary file 1). Cell lysates were first depleted of the US HIV RNA through hybridization to the intron-1 CO, followed by its capture with streptavidin-coated magnetic beads, and subsequent release using toehold-mediated oligonucleotide displacement. Additional hybridization, capture and release steps were subsequently repeated iteratively using first the intron-2 CO and then the 3’-exon CO for isolation of the PS and CS RNA pools, respectively. Once purified, proteins cross-linked to each isolated HIV RNA class were identified by mass spectrometry (overview in Figure 1B).

RT-qPCR assays specific to intron 1, intron 2, and the 3’-exon (Supplementary file 1) were used to determine RNA capture specificity and efficiency. For three biological replicates of the US, PS, and CS captures, the magnitude of amplification using each qPCR assay confirmed strong capture specificity for the desired splice variant class relative to the other two classes (Figure 1C, Figure 1—source data 1). Enrichment of each HIV transcript over a cellular control transcript (GAPDH) was >100-fold (Figure 1—figure supplement 1A, Figure 1—source data 1). Capture efficiency (the amount of each transcript depleted from the lysate after capture) was >70% for each variant (Figure 1—figure supplement 1B, Figure 1—source data 1).

To identify host proteins differentially interacting with the US, PS, and CS RNA pools, we isolated the in vivo cross-linked HIV RNA variants from three biological replicate experiments of 5 × 10⁷ infected Jurkat cells; with each replicate generated from a separate set of cultured cells and virus preparation. Interacting proteins from the US, PS, and CS capture samples were purified, analyzed by bottom-up mass spectrometry, then identified and quantified using search and label-free quantitation algorithms (Cox and Mann, 2008; Tyanova et al., 2016). In all, 926 proteins were identified in at least two biological replicates of all three HIV splice variant captures (Supplementary file 2). More than two-thirds (633) of these common interactors were previously identified in at least one prior study identifying proteins associated with cellular polyadenylated mRNA (Hentze et al., 2018; Queiroz et al., 2019; Supplementary file 2), and thus likely represent RNA-associated proteins with general roles in mRNA processing, transport, and translation. These common interactors also featured several proteins previously implicated in HIV-1 replication including host factors regulating RNA transport and translational initation (e.g., NCBP1, DHX9, DDX3, EIF4G, and PABP) (Boeras et al., 2016; Bolinger et al., 2010; Soto-Rifo et al., 2013; Stake et al., 2015; Yedavalli et al., 2004), known HIV splicing factors (e.g., HMGA1, HNRNPA1, HNRNPAB, HNRNPH, HNRNPF, SRSF1, SRSF2, SRSF3, SRSF6, SRSF7, TRA2B, and U2AF2) (Dlamini and Hull, 2017; Mahiet and Swanson, 2016; Sertznig et al., 2018; Stoltzfus and Madsen, 2006), RNA nuclear export and transport proteins (e.g., ABCE1, RAB11A, RANBP2, and XPO1) (Friedrich et al., 2011), and proteins implicated in HIV-1 virus particle assembly (e.g., AP-2, PDCD6IP [ALIX], STAU2, UPF1, and VPS4) (Friedrich et al., 2011; Meng and Lever, 2013). Furthermore, several common interactors matched proteins previously identified in RNA capture screens that had used partial segments of HIV-1 RNA as bait, including 27 of 32 proteins identified by Kula et al., 32 of the 41 identified by Marchand et al., and 93 of the 121 identified by Stake et al. (Supplementary file 2; Kula et al., 2011; Marchand et al., 2011; Stake et al., 2015).

The primary goal of this study, however, was to use HyPR-MS_SV to differentiate subsets of proteins preferentially associated with one or more of the individual splice variant classes. Therefore, we conducted three pairwise comparisons: US vs PS, US vs CS, and PS vs CS. Using the Student’s t-test and a permutation-based false discovery rate (FDR) of 5%, we identified 212 proteins that differentially interacted with one or more of the HIV splice variant classes: 101, 93, and 68 proteins in the US, PS, and CS captures, respectively (Figure 2—source data 1). Hierarchical clustering was used to organize the 212 proteins into a heatmap for visualization (Figure 2A). The associated gene tree indicates the extent of similarity in ‘interaction profiles’ for the proteins shown; the length of the branches directly correlates with the degree of similarity. This analysis revealed clusters of proteins elevated for each individual class as well as proteins common to members of two HIV splice variant classes. The most abundant of these were proteins preferentially associated with both the US and PS HIV transcripts but not the CS pool (45 proteins) (Figure 2A).

We used these interactome data to infer biological pathways potentially relevant to the regulation of each splice variant class. Using gene ontology (GO) term enrichment algorithms (Mi et al., 2017), we evaluated each interactome for enrichment of proteins involved in specific biological processes. This analysis revealed over-representation of several GO terms in the interactome of each splice variant class; some common to more than one class (Figure 2B, Figure 2—source data 2). Notable among these were nine proteins, known to regulate RNA stability, associated with both the US and PS, but not CS, transcripts (FXR1, YBX1, YBX3, IGF2BP1, IGF2BP3, SYNCRIP, HNRNPR, YTHDF2, and UPF1). Proteins involved in mRNA splicing, and the regulation thereof, were also elevated in the US and PS relative to the CS capture samples (YBX1, SYNCRIP, HNRNPR, HNRNPL, BCAS2, RBMX, and MBNL1) but with less congruence. A subset of these proteins were elevated only in the US capture samples (HNRNPA2B1 and HNRNPK; negative regulators of splicing) or PS capture samples (DAZAP1 and TRA2A; positive regulators of splicing), but not in both. PS captures were also exclusively enriched for proteins found in nuclear bodies. For cytoplasmic activities, proteins involved in translation were highly enriched in both the US (12 proteins) and CS (18 proteins) interactomes. Interestingly, however, while the CS interactome included translation initiation proteins (as may be expected), the US interactome was enriched for proteins linked to mRNA translation in the mitochondria. Cellular component GO term enrichment analysis further revealed 45 mitochondrion proteins enriched in the US RNA interactome; eight are mitochondrial nucleoid proteins and several have known roles in the carboxylic acid metabolic process, a GO term also over-represented in the US RNA interactome (Figure 2C, Figure 2—source data 2).

To determine their potential relevance to HIV-1 gene expression, 121 host proteins identified by HyPR-MS_SV were targeted for siRNA KD (Supplementary file 3) and passed cell viability criteria in HEK293T cells (Figure 3—source data 1). The HEK293T cell line was selected here due to its extensive use in studies of HIV-1 expression and its compatibility with siRNA transfection experiments (König et al., 2008). Following KD, cells were infected with a two-color HIV-1 virus engineered to report single-cell levels of viral US (Gag-cyan fluorescent protein [CFP]) and CS (mCherry) gene expression (Figure 3A and B; Knoener et al., 2017). Relative to a scrambled siRNA control, statistically significant changes (p-value<0.05) to early (CS) and/or late (US) gene expression were observed for a remarkable 69% (84 of 121) of the targeted host genes (Figure 3C; Supplementary file 4). The KD of 33 host proteins affected the expression of US and CS protein products in the same direction (either both increased or both decreased) and with approximately the same magnitude. By comparing mCherry:CFP fluorescence ratios for each protein KD to the negative control, we determined that CS and US protein expression were differentially affected by KD of 51 host proteins; for 26 of the proteins the expression changes were in the same direction but with different magnitudes; for 18 only the expression of the US RNA protein product was affected; and for seven only the expression of the CS RNA protein product was affected (Figure 3C, Supplementary file 4). Based on the direction of the changes in HIV-1 gene expression (increased or decreased), we categorized 71 host genes as putative ‘positive’ effectors and 16 as putative ‘negative’ effectors (Figure 3C, Supplementary file 4). Interestingly, of the 16 negative effectors, 10 were implicated in mitochondria-associated pathways based on GO analysis; of those ten, nine were identified by HyPR-MS_sv to preferentially interact with the US HIV RNA (Figure 2—source data 2, Supplementary file 4).

We selected a subset of 20 HyPR-MS_SV identified host proteins for further validation studies. This subset was, in part, chosen based on the commercial availability of antibodies that allowed for immunoblot- and/or immunofluorescence (IF)-based detection of the host proteins (Supplementary file 5), and included five proteins linked to mitochondria (LRPPRC, DHX30, MBOAT7, GSDMA, and DLD; all negative effectors of US RNA gene expression), ten genes encoding proteins with functions related to mRNA processing, localization, and stability (FAM120A, HNRNPR, IGF2BP3, G3BP1, RBMX, CSDE1, SRSF6, SRRM2, RBM4, and RPL15, the majority of which were positive effectors of either US or CS gene expression), and five additional proteins that had not previously been linked to RNA regulation (TRIM56, BUB3, DNM2, DYNC1H1, and NCLN) (Figure 3D, Supplementary file 4). Fifteen of the 20 proteins were detected by immunoblot and siRNA KD was confirmed (31–95% relative to negative control siRNA) (Figure 3—figure supplement 1). IGF2BP3, SRRM2, DNM2, RPL15, and DYNC1H1 KDs were not confirmed by immunoblot (Supplementary file 6).

US HIV-1 RNA-protein interactions may commence as early as production of the nascent HIV transcript in the nucleus or as late as virus particle formation at the plasma membrane. To determine potential sites of interaction, we used three-color combined fluorescence in situ hybridization (FISH)/IF single-cell imaging to show host factor subcellular localization relative to US RNA and viral Gag proteins (Figure 4A and Figure 4—figure supplements 1–3). HeLa cells were chosen based on their microscopy-conducive size and shape (large and flat) and their common use to study HIV-1 gene expression (Jouvenet et al., 2008; Pocock et al., 2016). Cells were infected with an HIV-1 reporter virus (HIV-1 E-R-CFP) allowing for identification of infected cells and confirmation of specificity of the US RNA FISH probes (Stellaris FISH probe set specific to intron-1; Supplementary file 7) and Gag antibody (anti-p24Gag; Supplementary file 5). Host proteins were detected using the primary antibodies employed for our immunoblot analysis (Supplementary file 5). Seventeen of the 18 host proteins (all but DLD) were detected by IF and showed >40% decreases in IF signal after host protein siRNA KD. This imaging-based analysis also allowed verification of the efficacy of siRNA KD for three of the host proteins (IGF2BP3, SRRM2, and DNM2) that we had been unable to detect using immunoblot (Supplementary file 8).

FISH/IF was performed on HeLa cells 48 hr post-infection to localize US RNA, Gag, and each host protein. Analysis by single-cell fluorescence microscopy showed consistent co-localization of US HIV-1 RNA with 11 of the proteins (Figure 4A and Figure 4—figure supplements 1–3), with four (HNRNPR, RBMX, RBM4, MBOAT7) predominantly localized to the nucleus or near the nuclear membrane and seven (FAM120A, IGF2BP3, MOV10, TRIM56, DNM2, LRPPRC, CSDE1) predominantly localized to the cytoplasm in uninfected cells (Figure 4—figure supplement 4). In infected cells, we observed five recurrent US RNA-host protein co-localization phenotypes: (1) at nuclear puncta, (2) at puncta proximal to the nuclear membrane, (3) at cytoplasmic puncta, (4) at large, cytoplasmic complexes reminiscent of stress granules, and (5) at the plasma membrane (Figure 4A and Figure 4—figure supplements 1–3). In the nucleus, we typically observed one or two bright US RNA puncta per cell, consistent with prior reports describing sites of active HIV-1 transcription (Puray-Chavez et al., 2017). Puncta proximal to the nuclear membrane and cytoplasmic puncta were smaller, more numerous, and of lower intensity. Cytoplasmic granules were large with moderate intensity accumulations of US HIV RNA surrounded by or spotted with host protein. Plasma membrane puncta were variable in size and intensity and often co-localized with Gag, thus likely represent virion assembly sites (Figure 4A and Figure 4—figure supplements 1–3).

We quantified the frequency of each co-localization phenotype for 17–52 cells per antibody (Figure 4B, Figure 4—source data 1), excluding cytoplasmic granules that were only rarely observed. The data revealed that proteins that predominantly localize to the nucleus or proximal to the nuclear membrane (HNRNPR, RBMX, RBM4, MBOAT7) had a higher frequency of co-localization with HIV RNA at nuclear puncta (57–96%) relative to proteins that were predominantly localized to the cytoplasm (FAM120A, IGF2BP3, MOV10, DNM2, TRIM56, LRPPRC, CSDE1; 0–8%). Two proteins (HNRNPR and RBMX) co-localized frequently with HIV-1 US RNA at all four quantified sites (41–90%). All 11 host proteins co-localized with US RNA at small cytoplasmic puncta in a high percentage of cells (71–100%) with most (all but LRPPRC and MBOAT7) co-localizing with US RNA at the plasma membrane (31–100% of cells), generally with Gag also present (Figure 4B, Figure 4—source data 1).

Several HyPR-MS candidates (HNRNPR, FAM120A, IGF2BP3, RBMX, RBM4, CSDE1, DNM2, LRPPRC, MBOAT7) were observed to accumulate at bright US RNA nuclear puncta, suggesting that they associate with US RNA at or near sites of de novo transcription (Figure 4—figure supplements 1–3). Of these, HNRNPR, FAM120A, and IGF2BP3 were of particular interest because all three exhibited four US HIV RNA co-localization phenotypes (nuclear puncta, cytoplasmic granules, cytoplasmic puncta, and plasma membrane puncta) (Figure 4B); preferentially interacted with US and PS, but not CS, HIV RNA as determined by HyPR-MS_SV (Figure 4C, Supplementary file 4), and showed similar trends in gene expression upon siRNA KD (both US and CS expression decreased, with a marginally greater decrease in US expression) (Figure 4D, Supplementary file 4). By contrast, LRPPRC, a protein shown to localize to the mitochondria as well as the nucleus (Mili and Piñol-Roma, 2003; Ruzzenente et al., 2012), and MBOAT7, a protein shown to localize to mitochondria-associated membranes (Hirata et al., 2013), localized to US HIV RNA nuclear puncta, at perinuclear puncta, and at cytoplasmic puncta but were not observed to co-localize with US RNA and Gag at the plasma membrane. Similar to HNRNPR/FAM120A/IGF2BP3, both LRPPRC and MBOAT7 were preferentially associated with US and PS, relative to CS, transcripts based on HyPR-MS_SV analysis. However, unlike HNRNPR/FAM120A/IGF2BP3, each of these proteins were negative effectors of both US and CS HIV-1 gene expression (Figure 4D, Supplementary file 4). Interestingly, LRPPRC was detected not only near transcription sites but also in a trail-like pattern that extended to the periphery of the nucleus (Figure 4A and Figure 4—figure supplements 1–3) and MBOAT7 was observed at transcription sites, at smaller subnuclear US HIV RNA puncta, and with high frequency and abundance at US RNA puncta at or near the nuclear membrane (Figure 4A and Figure 4—figure supplements 1–3).

The FISH/IF single-cell analyses of US HIV RNA, Gag, and host proteins also allowed for tracking of host factor responses to infection (Figure 5). For example, HNRNPR, generally a nuclear protein, was primarily localized to the nucleus of cells expressing no, or low amounts of, Gag and US RNA, but exhibited marked shifts from the nucleus to the cytoplasm in cells with high levels of Gag and US RNA expression (Figure 5A). Changes to MBOAT7 were also striking, with much higher levels of expression in cells with abundant Gag and US RNA than in uninfected or early infected cells (Figure 5B).

To further track HyPR-MS_SV host factor changes, we plotted single-cell measurements of total Gag and total US HIV RNA and used the resulting inflection point to discriminate cells in ‘early’ and ‘late’ stages of HIV gene expression (Figure 5C and D; Figure 5—source data 1). We measured relative host protein abundances for 12 of these factors at these stages (HNRNPR, FAM120A, IGRF2BP3, LRPPRC, MBOAT7, CSDE1, DNM2, MOV10, RBM4, RBMX, SRRM2, TRIM56; Figure 5E and F, Figure 5—source data 1). In general, each host protein exhibited non-random, bimodal expression changes from ‘early’ and ‘late’ HIV gene expression (Figure 5—figure supplements 1–4, Figure 5—source data 1). For example, in early/uninfected cells, we observed linear increases in HNRNPR and MBOAT7 expression, positively correlating with the subtle increases in US HIV RNA expression (Figure 5E and F; slope m = 1.7, 0.9, respectively). However, in late cells, HNRNPR expression rose then fell again as per-cell US RNA increased, fitting a polynomial rather than linear trendline (R² = 0.690) (Figure 5E), while MBOAT7 expression levels plateaued (Figure 5F, slope m = −0.007).

A similar analysis was performed after image-based segmentation of cells into nuclear and cytoplasmic compartments to better discriminate the subcellular location in which host protein changes occurred (Figure 5—figure supplements 1–4, Figure 5—source data 1). For HNRNPR, the same trends were observed in the nucleus and cytoplasm as were seen for the total cell (Figure 5—figure supplements 1–4). For MBOAT7, nuclear expression plateaued as it did for total cell expression, but the cytoplasmic expression increased slightly as US RNA and Gag abundance increased (Figure 5—figure supplements 1–4). In all, the expression of each of the 12 host proteins showed significant correlation with the expression of US HIV RNA in at least one of the following sub-groups: early-nuclear, late-nuclear, early-cytoplasmic, late-cytoplasmic (Figure 5—figure supplement 5, Figure 5—source data 1). For HNRNPR and MBOAT7, the most evident differences were in cytoplasmic expression (cyto HNRNPR, median increase = 21%, p=0.055; cyto MBOAT7, median increase = 41%, p=3×10⁻⁵) (Figure 5G–J, Figure 5—figure supplements 1–4, Figure 5—source data 1). In all, changes to nuclear or cytoplasmic abundance were observed for five host proteins (p-values<0.05; MBOAT7, TRIM56, RBMX, MOV10, and IGF2BP3) (Figure 5K, Figure 5—figure supplements 1–4, Figure 5—source data 1). Three of these proteins showed differences to total cellular expression (MBOAT7, RBMX, and TRIM56), with RBMX and TRIM56 only increasing in the cytoplasm. Two proteins did not show net differences in overall expression but exhibited statistically significant differences (p-value<0.05) in expression in the nucleus (MOV10) or the cytoplasm (IGF2BP3).

To identify potential host protein translocation events, we evaluated single-cell nuclear-to-cytoplasmic (nuc/cyto) ratios relative to US RNA abundance and looked for statistically significant differences in early and late cells (Figure 5—figure supplements 1–4, Figure 5—source data 1). HNRNPR nuc/cyto ratios ranged from 1 to 3 in early/uninfected cells but only ranged from 0.6 to 0.9 in late infected cells, exhibiting a negative correlation with US HIV RNA expression (Figure 5L). For MBOAT7, the nuc/cyto ratio ranged from 1.7 to 3.9 in early cells and 1.5 to 3.9 in late cells, with no significant correlation with US RNA expression for either phase (Figure 5M). However, overall nuc/cyto ratios were significantly lower for late cells relative to early cells for both proteins (median decrease = -52%; p=5×10⁻¹¹ and median decrease=-27%; p=7×10⁻⁶, respectively) (Figure 5N and O). In all, the nuc/cyto ratios of six HyPR-MS candidate proteins showed notable changes to nuc/cyto ratio (HNRNPR, MBOAT7, TRIM56, SRRM2, RBMX, and RBM4); all with the exception of SRRM2, exhibiting relative increases to cytoplasmic abundance (Figure 5P, Figure 5—source data 1).

Taken together, these analyses demonstrated that many of the host factors identified by HyPR-MS_SV not only modulate HIV-1 gene expression (Figure 3) but also co-localize with HIV-1 US RNA (Figure 4) and respond during infection by increasing in abundance and/or undergoing alterations to subcellular distribution (Figure 5).

Jurkat E6 cells are T lymphocytes established from the peripheral blood of a 14-year-old male with acute T-cell leukemia and were obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH: Jurkat Clone E6-1 from Dr. Arthur Weiss (cat# 177) (Weiss et al., 1984). The cells were cultured in RPMI media supplemented with 10% fetal bovine serum and 1% L-glutamine-penicillin-streptomycin in roller bottles rotated at three rotations per minute (rpm) at 37°C in 5% CO₂. A cell density of 1 × 10⁶ cells/mL of media was maintained by regular quantification.

HEK293T cells are human embryonic kidney cells and were obtained from ATCC (CRL-11268). The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and 1% L-glutamine-penicillin-streptomycin at 37°C in 5% CO₂. This cell line was authenticated by morphology and are G418 resistant.

Human 293 T-cells obtained from ATCC were modified to stably express YFP-tagged F-actin (ACT-YFP) and were cultured in DMEM supplemented with 10% fetal bovine serum, 1% L-glutamine, and 1% penicillin-streptomycin.

HeLa cells obtained from ATCC were cultured in DMEM media supplemented with 10% fetal bovine serum and 1% L-glutamine-penicillin-streptomycin at 37°C in 5% CO₂.

All cell lines were validated as mycoplasma-negative using a PCR assay developed by Uphoff and Drexler, 2002; Uphoff and Drexler, 2004.

2.5 × 10⁶ HEK293T cells were plated in 10 cm tissue-culture-treated dishes in 10 mL media, then transfected using polyethylenimine with 1 µg of DNA plasmid expressing the G envelope glycoprotein from vesicular stomatitis virus (VSV-G) and 9 µg of plasmid DNA encoding the full-length NL4-3 molecular clone of HIV-1 bearing inactivating mutations in env, vpr, and either (1) expressing a CFP reporter from the nef reading frame (HIV-1 E-R-CFP) (Adachi et al., 1986; Becker and Sherer, 2017) or (2) expressing mCherry in the nef ORF and three copies of CFP, in tandem, between the matrix and capsid ORFs of Gag (E-R- Gag-CFP mCherry/nef) (Knoener et al., 2017). At 24 hr post-transfection, media was replaced with 4 mL fresh media. At 48 hr post-transfection, culture supernatants were harvested, filtered through a sterile 0.45 µm syringe filter, and frozen at −80°C. Dose of HIV-1 E-R-CFP viral inoculum required for effective infection was determined in small-scale infection titrations in Jurkat cells.

Images were processed and analyzed using FIJI/ImageJ2 (Rueden et al., 2017). For each well, only cell monolayers were used for quantitation of fluorescence. Cell viability for each gene-specific siRNA KD was assessed using the ACT-YFP marker expression in the 293 T-cell line. Wells that had YFP fluorescence detected within +/- 1.5 standard deviations of the plate mean were considered acceptable for further analysis. CFP and mCherry fluorescence for each well was normalized to the YFP fluorescence for the gene-specific siRNA and negative control siRNA (included in each 96-well plate). The Student’s t-test calculation was performed to determine if a statistically significant change in CFP or mCherry expression was detected between each gene-specific siRNA KD and the negative control siRNA KD (Figure 3—source data 1; p-value<0.05).

HEK293T cells, with and without gene-specific siRNA KD, were lysed in 1x radioimmunoprecipitation assay buffer (10 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1 mM EDTA, 0.1% sodium dodecyl sulfate (SDS), 1% Triton X-100, 1% sodium deoxycholate) and sonicated. Samples were then boiled for 10 min in 2x dissociation buffer (62.5 mM Tris-HCl [pH 6.8], 10% glycerol, 2% SDS, 10% β-mercaptoethanol), run on SDS-PAGE 10% polyacrylamide gels, and transfered to nitrocellulose membranes (0.2 μM pore size). Immunoblotting was performed as previously described (Becker and Sherer, 2017; Behrens et al., 2017; Garcia-Miranda et al., 2016) using the primary and secondary antibodies detailed in Supplementary file 5.

Fluorescence for each channel (HIV-RNA, Gag protein, and each host protein) was quantified using FIJI/ImageJ2 (Rueden et al., 2017). Nuclear and total cellular fluorescence were measured by drawing perimeters around the nucleus (defined by DAPI staining) and the entire cell (defined by Gag protein fluorescence in late stage cells or autofluorescence in uninfected/early stage cells), then using FIJI quantitation tools to measure the fluorescence within each drawn perimeter. Cytoplasmic fluorescence was calculated by subtracting nuclear fluorescence from total cell fluorescence and the nuc/cyto ratio was calculated by dividing the nuclear fluorescence by the cytoplasmic fluorescence (Figure 5—source data 1). For determining correlation of host protein expression with HIV gRNA expression, cells with outlier values in total HIV gRNA fluorescence were excluded from the dataset. An outlier here is defined as a value that is more than 1.5 interquartile ranges (IQRs) below the 1st quartile (Q1) or above the 3rd quartile (Q3). IQR is defined as (Q3–Q1), with Q3 and Q1 calculated using the quartile function in Excel (Figure 5—source data 1). The Pearson’s R value was calculated, excluding outliers, to determine correlation of US HIV RNA and host protein fluorescence expression in the nucleus, cytoplasm, total cell, and for the nuc/cyto ratios using the CORREL function in Excel. R² values were calculated using the chart tools in Excel (Figure 5—source data 1).

For determining host protein expression and distribution changes, outliers were determined, as described above, for host protein expression values in early and late cells in four categories: nuclear, cytoplasmic, total cellular, and nuc/cyto ratio. To determine if ‘early’ and ‘late’ cells showed statistically significant differences in host protein expression, a Student’s t-test, excluding outliers, was used to determine a p-value. The percent change in each category was calculated using the mean values for each category within early and late cells.

Statistical methods are described in the appropriate ‘Materials and Methods’ section or figure captions for all data analyses.