To establish a methodology that could simultaneously characterize protein and glycan features at the single-cell level, we developed a panel of lanthanide metal-conjugated antibodies and lectins compatible with CyTOF, which we refer to as CyTOF-Lec (Supplementary file 1 A). As tonsils provide an abundant source of both T and B cells, we used these cells for our initial validation of the panel. First, we confirmed that the staining patterns of the lanthanide-conjugated antibodies were consistent with the known differential expression of their target antigens on tonsillar T vs. B cells (Figure 1—figure supplement 1), and with results previously reported using CyTOF without lectin staining (Ma et al., 2020). To ensure that the lectin staining did not displace or alter the antibodies bound to their protein targets, we developed a protocol whereby surface antibody staining was completed prior to lectin staining (see Materials and methods). We confirmed that all five lectins (AOL, MAL-1, WGA, UEA-1, ABA, see Supplementary file 1) conjugated to lanthanides stained both T and B cells (Figure 1—figure supplement 1). Furthermore, we confirmed that antibody binding to CD3, CD4, and CD8 was the same whether or not the specimens were subsequently stained with lectins (Figure 1A). To establish the specificity of lectin binding, we assessed the effect of sialidase, which degrades cell-surface sialic acid. As expected, binding by WGA and MAL-1, which detect different forms of sialic acid (Supplementary file 1), was decreased after sialidase treatment of the cells (Figure 1B). By contrast, binding by AOL and UEA-1, which detect different forms of fucose, and ABA, which detects T antigen (Supplementary file 1), were all increased (Figure 1C–D). This was expected as removal of sialic acid should enable better detection of these other glycan structures, and is consistent with prior reports (Giron et al., 2020b).

To determine whether the conjugated lectins could detect differences in cell-surface glycans between subsets of immune cells, we compared binding by each of the five lectins to B cells, memory CD8+ T cells (CD8+ Tm cells), naïve CD8+ T cells (CD8+ Tn cells), memory CD4+ T cells (CD4+ Tm cells), and naïve CD4+ T cells (CD4+ Tn cells). We assessed the expression of glycans on these subsets not only among tonsillar cells, but also among PBMCs and endometrial T cells for comparison. For all three sites, the fucose-specific lectins AOL and UEA1 bound more to CD4+ Tm and CD8+ Tm cells than to their respective naïve counterparts, although the difference only reached significance for AOL binding in tonsils (Figure 2A). These results are consistent with the known role of core surface fucosylation for T cell activation, which is more prominent within the memory compartment (Fujii et al., 2016; Liang et al., 2018). The sialic acid-binding lectins WGA and MAL1 also consistently bound to memory T cells more than to their naïve counterparts, although this only reached statistical significance for MAL1 binding of endometrial CD8+ T cells (Figure 2B). Binding by ABA did not show a consistent pattern between memory vs. naïve T cells (Figure 2C). Binding by all five lectins was low on B cells, particularly as compared to memory CD8+ T cells (Figure 2A–C).

These results establish CyTOF-Lec as a panel that can quantitate glycan and protein expression at the single-cell level, and detect differential glycan expression between different subsets of immune cells.

We next applied CyTOF-Lec to determine the extent to which glycans are remodeled on the surface of HIV-infected cells. Fresh endometrial biopsies (n=6 donors) and whole tonsils from tonsillectomies (n=4 donors) from HIV seronegative individuals were processed into single-cell suspensions, and then immediately exposed to the HIV-F4.HSA, a replication-competent and Nef-sufficient virus that harbors the CCR5-tropic 109FPB4 transmitted/founder (T/F) Env (Cavrois et al., 2017). In addition, PBMCs were isolated from whole blood of HIV seronegative individuals (n=4 donors) and exposed to HIV-F4.HSA. To limit potential confounding effects of ex vivo-induced T cell activation, infection was carried on unstimulated cells. Infection was allowed to proceed for 3 days, after which the cells, as well as cells from paired uninfected control cultures, were harvested for CyTOF-Lec analysis. CD4+ T cells were identified as intact, live singlet CD3+ CD8- cells, while infected cells were identified as intact, live singlet CD3+ CD8- CD4^Low cells (Figure 3—figure supplement 1), to account for the downregulation of cell-surface CD4 by HIV (Doms and Trono, 2000; Garcia and Miller, 1991; Lama, 2003; Piguet et al., 1999). Consistent with our prior studies (Ma et al., 2020), endometrial T cells were the most susceptible to HIV-F4.HSA infection (Figure 3—figure supplement 2A). HIV-infected cells from all three sites were remodeled, as established qualitatively by assessing their locations on a t-SNE (Figure 3—figure supplement 2B) as well as quantitatively using SLIDE (Sen et al., 2014; Figure 3—figure supplement 2C). The remodeling of the infected cells is consistent with prior studies (Cavrois et al., 2017; Ma et al., 2020; Xie et al., 2021), and confirms that CyTOF-Lec is a valid panel for the analysis of remodeling.

To identify specific glycans that were remodeled, we implemented PP-SLIDE to identify the predicted precursor (PRE) cells (Cavrois et al., 2017; Ma et al., 2020; Neidleman et al., 2020b; Xie et al., 2021). PRE cells harbor the predicted original (pre-remodeling) features of T cells infected by HIV and are identified using k-nearest neighbor approaches by matching, in the high-parameter CyTOF space, the T cells in the uninfected culture most similar in phenotype to every HIV-infected cell (Figure 3—figure supplement 3A). As expected (Ma et al., 2020), the PRE cells from all three sites were preferentially memory CD4+ T cells (Figure 3—figure supplement 3B, C). Having identified the PRE cells, we then determined which glycans, if any, were remodeled by HIV infection. This was accomplished by assessing for lectins that differentially bound the PRE as compared to the infected cells. Glycans were quantitated by reporting the median signal intensity (MSI) of their corresponding lectins among each population of cells from each donor. Interestingly, we found that both fucose and sialic acid were upregulated during HIV infection. Infected cells from all three sites potently upregulated total fucose as assessed by AOL binding (Figure 3A, Figure 3—figure supplement 4A), although this upregulation did not reach statistical significance (after correcting for multiple comparisons) for the endometrium. Binding by UEA1, however, was not significantly different between infected and PRE cells, and tended to be downregulated in the endometrium and tonsils, and upregulated in PBMCs (Figure 3A). As UEA1 binds α1–2 branched fucose (Supplementary file 1), these results suggest that fucosylation is globally upregulated upon infection of tissue CD4+ T cells with HIV, although not the type of fucosylation that creates α1–2 branched structure. In contrast to the fucose-binding lectins, both sialic acid-binding lectins were increased on infected compared to PRE cells (Figure 3B, Figure 3—figure supplement 4B), and this was observed for all three sites although results did not reach statistical significance for the endometrium. These results are consistent with an upregulation of both total sialic acid (recognized by WGA, which also binds to N-acetylglucosamine [GlcNAc]) (Schwarz et al., 1999) and α2–3 linked sialic acid (recognized by MAL-1) by HIV during infection. In contrast to fucose and sialic acid, we did not observe any marked upregulation of T antigen on infected cells (Figure 3C).

In addition to revealing antigens that have been remodeled by infection, the PP-SLIDE approach can also identify antigens that are differentially expressed on cells before infection. In particular, antigens more abundant on PRE than uninfected cells correspond to antigens preferentially expressed on HIV-susceptible cells, while those less abundant on PRE cells correspond to those preferentially expressed on HIV-resistant cells. We used as our uninfected population CD4+ Tm cells, excluding CD4+ Tn, CD8+ Tm, and CD8+ Tn cells because these latter three populations harbored negligible numbers of HIV-susceptible cells (Figure 3—figure supplement 3B). This exclusion was important because otherwise antigens differentially expressed between uninfected and PRE cells could just reflect phenotypic differences between these major subsets. Significant differences in lectin binding between PRE and uninfected (CD4+ Tm) cells were only observed in PBMCs, with AOL, UEA1, WGA, and MAL-1 all binding at significantly higher levels on PRE cells (Figure 3, Figure 3—figure supplement 4C). Tonsillar PRE cells also bound these lectins more than their uninfected counterparts did, but these results did not reach statistical significance. Endometrial PRE cells did not show significant differences in lectin binding relative to their uninfected counterparts.

These results together with the remodeling analysis suggest that in blood and tonsils (but not endometrium), HIV preferentially infects memory CD4+ T cells with higher levels of fucose and sialic acid, and then further upregulates these cell-surface glycans through viral remodeling.

Remodeling of cells can occur in a cell-intrinsic manner as a result of direct infection, but may also result from bystander effects. For example, the inflammatory environment elicited by HIV replication may elicit phenotypic changes in bystander (uninfected) cells in the infected culture. We therefore assessed whether HIV infection elicits any glycosylation alterations in bystander cells. To identify bystander memory CD4+ T cells, we gated the infected culture on CD4+ Tm cells that were HSA-negative. Increased binding by all five lectins was observed among bystander tonsillar CD4+ Tm cells relative to their counterparts from uninfected cultures (Figure 3—figure supplement 5), suggesting that in at least some tissue sites, remodeling of glycans on bystander CD4+ T cells occurs. Interestingly, however, relative to the bystander CD4+ Tm cells, the infected cells still exhibited higher levels of total fucose and sialic acid (as assessed by AOL and WGA binding, respectively), suggesting possible additional cell-intrinsic glycan remodeling by replicating virus (Figure 3—figure supplement 5).

To examine whether HIV alters glycan expression in other bystander cellular subsets, we compared glycan levels on multiple subsets of B and T cells from the uninfected vs. infected cultures (Figure 3—figure supplement 6). Only tonsils exhibited significant differences between uninfected vs. bystander cells, and these differences were observed among all subsets. For example, sialic acid levels (as assessed by both WGA and MAL-1 binding) were significantly higher in all the analyzed subsets of bystander B, CD8+, and CD4+ T cells, relative to their counterparts from uninfected cultures. Fucose expression was also uniformly higher among bystander cells, although the difference among CD8+ Tm cells did not reach statistical significance. Differences in levels of T antigen were also observed among B cells, CD8+ Tm cells, and CD4+ Tm cells (Figure 3—figure supplement 6). These glycan changes may be elicited by HIV infection-induced inflammatory cytokines (Breen et al., 1990; Contreras et al., 2003; Sugawara et al., 2019), which can alter cell-surface glycosylation patterns (Dewald et al., 2016; Giron et al., 2020a).

The data presented thus far suggest that although there are differences between blood vs. the tissue sites examined, fucose and sialic acid are upregulated on HIV productively infected cells, and CD4+ T cells expressing high levels of fucose or sialic are preferentially targeted for infection. We next conducted manual gating to assess whether the HIV-susceptible cells express high levels of both fucose or sialic acid, or whether they belong to distinct subsets of fucose+ vs. sialic acid+ cells. We focused on the AOL and WGA datasets, as they cover total fucose and different forms of sialic acid, respectively. We first examined, within the HIV-infected cultures, the infection rates among CD4+ Tm cells expressing high vs. low levels of AOL or WGA. In both blood and both tissue compartments, AOL^High and WGA^High cells exhibited significantly higher HIV infection rates than did AOL^Low and WGA^Low cells, respectively (Figure 4A). To assess the extent to which this high level of infection was due to preferential infection of the AOL^High and WGA^High CD4+ Tm cells, we next compared, among the uninfected CD4+ Tm cells and PRE cells, the percentages of cells that were AOL^High, WGA^High, or AOL^High WGA^High. Consistent with the MSI data, the percentages of cells expressing high levels of AOL or WGA were higher among PRE cells in both tonsils and PBMCs (Figure 4B), suggesting preferential infection of fucose- and sialic acid-expressing cells by HIV at these sites. Interestingly, AOL^High WGA^High cells were also significantly over-represented among PRE cells at these sites (Figure 4B), suggesting that the HIV-susceptible Tm cells co-express fucose and sialic acid. Indeed, visualization of the PRE cells by t-SNE revealed cells binding high levels of both AOL and WGA among the tonsillar and blood compartments (Figure 4C). Although PRE cells from the endometrium did not preferentially express fucose or sialic acid (Figure 4B), PRE cells co-expressing AOL and WGA could be detected from this site (Figure 4C), suggesting the endometrium, like the other two sites, harbors HIV-susceptible cells co-expressing fucose and sialic acid. To better understand why the HIV-susceptible endometrial CD4+ Tm cells, unlike their tonsillar and blood counterparts, did not preferentially express high levels sialic acid or fucose, we compared the levels of WGA and AOL binding on uninfected CD4+ Tm cells from the three sites. This analysis revealed WGA, but not AOL, to be expressed at the highest levels on the endometrial cells (Figure 4D). These results suggest that the reason WGA^High cells are not preferentially targeted for infection in the endometrium may be that endometrial CD4+ Tm cells all express high levels of sialic acid. Fucose levels, however, are higher in PBMCs than endometrium, suggesting that the fact that AOL^High endometrial CD4+ Tm cells aren’t preferentially targeted for infection cannot be explained by exceptionally high levels of fucose expression. All together, these results suggest that among blood and tonsillar memory CD4+ T cells, those with the highest levels of fucose and sialic acid are preferentially targeted for infection by HIV; this phenomenon was however not observed among endometrial cells.

We then implemented a more global method of subset identification, using FlowSOM (Van Gassen et al., 2015). We combined the uninfected CD4+ Tm and PRE cells from all the donors, and identified 20 clusters for each of the three sites (Figure 5A). Endometrial T cells, which had the most PRE cells, were represented among most of the 20 endometrial cell clusters. To determine the extent of enrichment of each cluster among PRE cells, we calculated the ratio of the size of each cluster in the PRE vs. total uninfected CD4+ Tm cells. Enriched clusters identified in this manner (corresponding to those preferentially harboring HIV-susceptible cells, see Materials and methods) were detected among all three sites, with eight clusters from the endometrium, nine from tonsils, and six from PBMCs (Figure 5B). Interestingly, the fold-enrichment was highest among the tonsils (reaching almost 20-fold in one donor), suggesting that of all three sites, this one exhibits the most preferential selection of subsets for infection. Compared to the other two sites, the tonsils also harbored more enriched clusters with significantly elevated levels of fucose, sialic acid, and T antigens relative to their expression levels on uninfected CD4+ Tm cells (Figure 5—figure supplement 1). To assess whether the enriched clusters co-express fucose and sialic acid, we assessed by t-SNE heatmaps the levels of AOL and WGA binding on concatenated files of all the enriched clusters from each site. This analysis revealed regions of the t-SNEs co-expressing high levels of fucose and sialic acid (Figure 5C), confirming the manual gating data that HIV-susceptible cells co-express these two classes of glycans.

The results presented thus far suggest that CD4+ Tm cells from endometrium, tonsils, and blood are preferentially susceptible to HIV infection compared to their naïve counterparts, but only in tonsils and blood can high levels of fucose and sialic acid further distinguish HIV-susceptible CD4+ Tm cells from non-susceptible CD4+ Tm cells. To experimentally validate these findings, we conducted sorting experiments. As endometrial and tonsillar T cells do not maintain good viability after sorting, we limited these studies to blood specimens. CD4+ Tm cells from blood expressing low (WGA^Low), medium (WGA^Medium), or high (WGA^High) levels of sialic acid were isolated through sorting (Figure 6A). These sorted populations (along with total CD4+ Tm cells as a comparison control) were then exposed to HIV-F4.HSA for 3 days and then assessed by FACS for infection rates. Infection rates directly correlated with the expression levels of sialic acid, with the WGA^Low cells being the least susceptible and the WGA^High the most (Figure 6B and C). These results provide experimental confirmation that in PBMCs, differentially susceptible CD4+ Tm cells can be isolated based solely on sialic acid expression levels.

To better understand the mechanism behind the differential susceptibility of cells expressing high vs. low levels of sialic acid, we returned to our CyTOF datasets and manually gated on CD4+ Tm cells expressing high vs. low levels of total sialic acid as assessed by WGA binding, to assess what was differentially expressed among these two populations. The WGA^High cells preferentially expressed fucose (as assessed by AOL binding) (Figure 6D), consistent with earlier observations of co-expression of fucose and sialic acid on HIV-susceptible CD4+ T cells. Relative to their WGA^Low counterparts, the WGA^High cells also preferentially expressed higher levels of CD4 and the HIV co-receptor CCR5 (Figure 6E), potentially explaining the increased susceptibility of these cells to infection. As activated T cells are known to be preferentially susceptible to infection (Stevenson et al., 1990), we also compared expression levels of activation markers on the WGA^Low and WGA^High cells, and found that six markers of T cell activation (HLADR, CD69, CD38, CD25, CD28, and ICOS) were all elevated on the latter population (Figure 6F). To further validate the notion that the high susceptibility of WGA^High CD4+ Tm cells is closely associated with the activation status of these cells, we phenotyped CD4+ Tm cells from resting vs. PHA-stimulated PBMCs. As expected, the stimulated CD4+ Tm cells expressed higher levels of multiple activation markers (Figure 6—figure supplement 1a). Importantly, the stimulated CD4+ Tm cells also bound higher levels of WGA, consistent with upregulation of sialic acid upon T cell activation (Figure 6—figure supplement 1). Furthermore, activated cells, as defined as those expressing high levels HLADR, CD69, CD38, CD25, CD28, or ICOS, all expressed higher levels of sialic acid as compared to cells with low levels of these activation markers (Figure 6—figure supplement 1). These data together strongly support the notion that high sialic expression identifies the most activated subsets of CD4+ Tm cells.

We also considered the possibility that WGA^High cells may support higher levels of productive infection because these cells better survive the cytopathic effects of HIV replication. Consistent with this hypothesis, we found that WGA^High cells expressed higher levels of CD127, a marker of long-lived self-renewing cells, and Ox40 and BIRC5, which are involved in protecting HIV-infected cells from apoptosis (Kuo et al., 2018; Figure 6G). In comparison, markers preferentially expressed on the WGA^Low cells were those of central memory T cells, including CD27, CCR7, and CD62L (Figure 6H). Additional markers in our CyTOF panel differentially expressed between the WGA^Low and WGA^High cells included other glycans, transcription factors, homing receptors, and exhaustion markers (Figure 6—figure supplement 2).

Although these results suggest that high levels of sialic acid may simply be a biomarker of the most HIV-susceptible CD4+ Tm cells, it did not rule out the possibility that sialic acid plays a direct role in HIV susceptibility. To test this, we tried two approaches to diminish cell-surface sialic acid expression: transient treatment with either sialic acid synthase inhibitor P-3F_AX-Neu5Ac or sialidase. Inhibitor treatment did not decrease cell-surface sialic acid levels (Figure 6—figure supplement 1), but sialidase did (Figure 6I). We therefore exposed mock- or sialidase-treated cells to HIV-F4.HSA for 3 days and then assessed infection rates by FACS. We found that infection rates were decreased in a sialidase dose-dependent manner (Figure 6J and K). These results suggest that a single glycan – sialic acid – may not only be capable of distinguishing memory CD4+ T cells with vastly different phenotypic features and HIV susceptibility but may also play a direct role in promoting infection by the virus.

Endometrial tissues were obtained from the Women’s Health Clinic of Naval Medical Center Portsmouth (NMCP) in Virginia (CIP # NMCP.2016.0068) under standard operating procedures (Fassbender et al., 2014). The biopsies were transported to San Francisco in MCDB-105 (Sigma-Aldrich M6395) containing 10% heat-inactivated fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S). The collection of endometrial T cells was performed similar to recently described methods (Ma et al., 2020). Briefly, endometrial tissue was washed once with SCM media, which consisted of 75% phenol red-free Dulbecco’s Modified Eagle’s Medium (DMEM, Life Technologies), 25% MCDB-105, 10% FBS, 1% L-glutamine with P/S (Gemini), 1 mM sodium pyruvate (Sigma-Aldrich), and 5 mg/ml insulin (Sigma-Aldrich). The tissues were then digested at 37°C for 2 hr under rotation in SCM media mixed at a 1:1 dilution with Digestion Media, which consisted of HBSS containing Ca²⁺ and Mg²⁺ supplemented with 3.4 mg/ml collagenase type 1 (Worthington Biochemical Corporation LS004196) and 100 U/ml hyaluronidase (Sigma-Aldrich H3631). Cells were then filtered through a Falcon 40 μm cell strainer. The filtrate was then centrifuged, washed once with R10 media (RPMI 1640 containing 10% FBS and 1% P/S), and the cells were cultured in 96-well U-bottom polystyrene plates at a concentration of 10⁶cells/well in 200 μl R10 media.

Human tonsils obtained from the Cooperative Human Tissue Network (CHTN) were processed similarly to methods recently described (Ma et al., 2020). Briefly, tonsils were rinsed in tonsil media (RPMI supplemented with 15% FBS, 100 μg/ml gentamicin, 200 μg/ml ampicillin, 1 mM sodium pyruvate, 1% non-essential amino acids [Mediatech], 1% Glutamax [Thermo Fisher], and 1% Fungizone [Invitrogen]), dissected into small pieces, and then pressed through a 40 μm cell strainer using a syringe plunger. The cells were then filtered through a second 40 μm cell strainer, centrifuged, and resuspended with 200 μl tonsil media per 10⁶ cells. For sialidase treatment, 10⁶ tonsil cells were resuspended in 1 ml of tonsil media, and then treated with 20 μg of the sialidase for 1 hr at 37°C. The cells were washed once with PBS and processed for CyTOF analysis as described further below.

PBMCs were isolated from reduction chambers obtained from Vitalant Research Institute and Stanford Blood Bank using Ficoll-Hypaque density gradients, and then cultured in R10. For sorting experiments, CD4+ T cells were purified by negative selection using the EasySep CD4 enrichment kit (Stem Cell Technologies), and further enriched for memory cells by depletion of naïve T cells using CD45RA beads (Miltenyi Biotec), prior to lectin staining and sorting as described further below. Where indicated, PBMCs were first stimulated for 3 days with 5 µg/ml PHA in the presence of 10 IU/ml IL-2 prior to CyTOF-Lec analysis.

Viral stocks of the previously described HIV-1 reporter virus F4.HSA (Cavrois et al., 2017) were prepared similarly to recently described methods (Ma et al., 2020). Briefly, 293T cells (purchased directly from ATCC and therefore assumed to be authenticated, tested negative for mycoplasma contamination) were seeded in T175 flasks and transfected using polyethylenimine (Polysciences) with F4.HSA proviral DNA (70 μg/flask) (Longo et al., 2013). Two days after transfection, supernatants from 293T cultures were harvested, filtered through a 0.22 μm filter, and concentrated by ultracentrifugation at 20,000 rpm (Beckman Coulter Optima XE-90) for 2 hr at 4°C. p24^Gag concentrations were quantitated using the Lenti-X p24^Gag Rapid Titer Kit (Clontech). For infection, 10–20 ng/ml p24^Gag of F4.HSA was incubated with 10⁶ cells in 200 μl R10 media in 96-well U-bottom polystyrene plates. After 2 hr, cells were fed with fresh R10 media and cultured for another 3 days. Where indicated, PBMCs were first treated with the sialic acid inhibitor P-3F_AX-Neu5Ac (Tocris) for 24 hr, or sialidase for 1 hr, prior to infection with F4.HSA.

For sorting experiments, purified memory CD4+ cells (see above) were washed once with FACS buffer in 96-well V-bottom polystyrene plates, and then stained at room temperature for 15 min with a 1:200 dilution of the LIVE/DEAD Zombie Aqua Fixable Viability Kit (Biolegend) reagent to exclude dead cells. The cells were washed and stained for 30 min at 4°C with 5 μg/ml FITC-WGA (Vector Laboratories). After the wash, the purified memory CD4 + T cells were resuspended at a concentration of 10⁶ cells/ml, and sorted on an Aria II flow cytometer (BD Biosciences) into WGA^Low, WGA^Medium, and WGA^High populations. Total CD4+ Tm cells were also sorted as a control. The purity of the sorted cells was confirmed by analysis on Aria II immediately after sorting. All sorted populations were infected with F4.HSA for 3 days. For FACS analysis of the samples, 0.1–1 million cells of each sample were transferred into 96-well V-bottom polystyrene plates, washed once with FACS buffer (PBS containing 2% FBS and 2 mM EDTA), and stained for 30 min on ice with an antibody cocktail consisting of APC/Cy7-CD3 (SK7, Biolegend), PE/Cy7-CD4 (A161A1, Biolegend), APC-CD8 (SK1, Biolegend), FITC-CD24 (HSA, M1/69, BD Biosciences), and the LIVE/DEAD Zombie Aqua Fixable Viability Kit (Biolegend). The cells were then washed twice, fixed with 1% PFA (Electron Microscopy Sciences) in PBS, and analyzed by flow cytometry on an LSRFortessa (BD Biosciences).

A 39-parameter CyTOF panel was designed for this study, which included antibodies against markers of T cell differentiation states, activation markers, transcription factors, and homing receptors, and an antibody against HSA to identify productively infected cells. The panel also included numerous lectins enabling the characterization of glycan features (Supplementary file 1). X8 antibody-labeling kits (Fluidigm) were used to label antibodies that required in-house conjugation. The conjugated antibodies were quantitated for protein content by Nanodrop (Thermo Fisher). Prior to storage at 4°C, specimens were diluted 1:1 using a PBS-based Antibody Stabilizer (Boca Scientific) supplemented with 0.05% sodium azide.

Preparation of samples for CyTOF staining was conducted as previously described (Ma et al., 2020; Neidleman et al., 2020a; Neidleman et al., 2020b), Briefly, 1–6 million cells were washed once with CyFACS (metal contaminant-free PBS [Rockland] supplemented with 0.1% bovine serum albumin and 0.1% sodium azide). Where indicated, cells were first treated with 20 µg/ml of sialidase or PBS as control and incubated for 1 hr at 37°C. Sialidase was prepared in-house using the Vibrio cholerae nanH gene cloned into the pCVD364 vector, which was provided by Dr Eric R Vimr from the University of Illinois Urbana (Taylor et al., 1992). After centrifugation, the cells were resuspended with contaminant-free PBS (Rockland) supplemented with 2 mM EDTA (PBS/EDTA), and then treated with 25 μM cisplatin (Sigma-Aldrich) in 4 ml PBS/EDTA for 60 s at room temperature. The samples were then immediately quenched with CyFACS, centrifuged, resuspended in 2% PFA in CyFACS, and incubated for 10 min at room temperature. The cells were then washed three times with CyFACS, resuspended in 100 μl of CyFACS containing 10% DMSO, and frozen at –80°C until CyTOF staining.

To stain multiple specimens in the same reaction, cells were barcoded using the Cell-ID 20-Plex Pd Barcoding Kit according to the manufacturer’s instructions (Fluidigm). Briefly, 1–3 million cisplatin-treated cells were thawed and transferred into Nunc 96 DeepWell polystyrene plates (Thermo Fisher). After two washes with Barcode Perm buffer (Fluidigm), the cells were incubated with selected barcodes for 30 min. Cells were then washed with 0.8 ml Maxpar Cell Staining buffer (Fluidigm) followed by 0.8 ml CyFACS. Barcoded samples were combined and pelleted, and then blocked on ice for 15 min with sera from mouse (Thermo Fisher), rat (Thermo Fisher), and human (AB serum, Sigma-Aldrich). Cells were then washed twice with CyFACS, and stained on ice for 45 min with a cocktail of CyTOF surface-staining antibodies (Supplementary file 1) in a final volume of 100 μl/well. Cells were then washed three times with CyFACS buffer, and stained on ice for 45 min with the cocktail of lanthanide-conjugated lectins (Supplementary file 1) in a final volume of 100 μl/well. Cells were then washed three times with CyFACS buffer and fixed overnight at 4°C with 2% PFA in metal contaminant-free PBS. The next day, cells were incubated at 4°C for 30 min with fix/perm buffer (eBioscience), and then washed twice with Permeabilization Buffer (eBioscience). After another round of Fc blocking on ice for 15 min with sera from mouse (Thermo Fisher) and rat (Thermo Fisher), cells were washed twice with Permeabilization Buffer (eBioscience), and stained on ice for 45 min with a cocktail of CyTOF intracellular-staining antibodies (Supplementary file 1) in a final volume of 100 μl/well. Cells were then washed with CyFACS and incubated for 20 min at room temperature with 250 nM Cell-ID DNA Intercalator-Ir (Fluidigm) in PBS containing 2% PFA. After two more washes with CyFACS, cells were washed once with Maxpar Cell Staining Buffer (Fluidigm), once with Maxpar PBS (Fluidigm), and once with Maxpar Cell Acquisition Solution (Fluidigm). Immediately prior to sample loading, cells were resuspended to a concentration of 7 × 10⁵/ml in EQ calibration beads (Fluidigm) diluted 1:9 in Maxpar Cell Acquisition Solution. Cells were acquired on a Helios-upgraded CyTOF2 instrument (Fluidigm) at a rate of 250–350 events/s, at the UCSF Flow Core Facility.

Data were normalized to EQ calibration beads and then exported as FCS files. The data were then de-barcoded with CyTOF software (Fluidigm) and imported into FlowJo (BD) for gating. This study’s raw datasets, pre-gated on live, singlet events, are available for download via the following link in the Dryad public repository: https://doi.org/10.7272/Q6FT8J92.

Total T cells were identified by sequential gating on intact, live, singlet CD3+ CD19- cells (Figure 3—figure supplement 1). Total T cells were then re-exported as FCS files and imported into Cytobank for calculations of MSI, and high-dimensional analyses by t-SNE and FlowSOM. t-SNE and FlowSOM plots were generated with default settings except for a modification of total metaclusters from 10 to 20 for FlowSOM analysis. t-SNE and FlowSOM plots were generated excluding all parameters used upstream in the gating strategy (CD19 and HSA) and all glycan characterization parameters. To map defined populations onto t-SNE plots, subsets were defined by manual gating, and then pseudo-colored on the t-SNE plots using FlowJo software. Box plot graphs were generated using ggplot2 in R.

Identification of PRE cells by PP-SLIDE was implemented using recently described methods (Ma et al., 2020; Neidleman et al., 2020b) to match each infected cell against every CD4+ T cell in the uninfected sample and using k-nearest neighbor calculations to identify the phenotypically most similar. The degree of enrichment of each FlowSOM cluster in PRE cells was calculated by dividing each cluster’s relative size within the PRE cells by its relative size within total uninfected CD4+ Tm cells:

Enrichment ratio (Cluster X) = Number of Cluster X cells relative to PRE cells/Number of Cluster X cells relative to uninfected Tm cells.

Clusters with ratios > 1 were designated as enriched and those with ratios >0 and <1 as non-enriched, while clusters with undetectable PRE cells were not shown.

Expression levels were reported as MSI for each parameter (protein or glycan) within each cell population analyzed. Student’s two-sided paired t-tests were used to test for differences in MSI among phenotypic subsets (B cell, CD8+ Tm, CD8+ Tn, CD4+ Tm, and CD4+ Tn cells): among uninfected and bystander cells; or among uninfected cells, PRE cells, and infected cells. p-Values were adjusted for multiple testing using false discovery rate (FDR) via the Benjamini-Hochberg or Holm method as indicated in figure legends. FDR adjusted p-values that were <0.05 were considered as significant.

SLIDE analysis was conducted using the R package SLIDE (Mukherjee et al., 2018) as recently described (Ma et al., 2020). SLIDE was developed as a nearest-neighbor approach to identify and quantify viral-induced remodeling (Sen et al., 2014). The ratios between two distance measures in SLIDE (the remodeling score) provides a relative measure of remodeling, and is compared to a background remodeling score generated from SLIDE analysis of non-infected cells, as recently described (Mukherjee et al., 2018).