Differential Expression of CD8⁺ T Cell Cytotoxic Effector Molecules in Blood and Gastrointestinal Mucosa in HIV-1 Infection

This article is featured in In This Issue, p.1531

Mucosal tissues are primary targets for HIV (HIV-1) infection, and the gastrointestinal mucosa is a major site of virus transmission and pathogenesis. Accordingly, understanding mucosal immune responses to HIV-1 is important for the development of strategies to fight infection. CD8⁺ T cells are thought to play a crucial role in containing HIV-1 infection, as depletion of these cells results in viral rebound (1, 2). However, the specific effector functions by which CD8⁺ T cells mediate viral containment remain unclear. In blood, both perforin-mediated cytotoxicity and polyfunctionality, defined as the simultaneous production of multiple cytokines and chemokines, have been identified as possible correlates of immune-mediated protection from disease progression (3–9). CD8⁺ T cell polyfunctionality, including robust production of macrophage inflammatory protein (MIP)-1β, has also been identified as a potential correlate of protection in mucosa (10). However, gastrointestinal CD8⁺ T cells exhibit strikingly low perforin expression and reduced cytotoxicity compared with blood T cells, suggesting CD8⁺ T cells housed in the gut are primed for cytokine release rather than cytotoxicity (11).

Despite significantly lower perforin levels in the gut compared with blood, HIV-1⁺ individuals not on antiretroviral therapy (ART) display greater proportions of perforin and granzyme (Gzm) B–positive CD8⁺ T cells in gastrointestinal mucosa compared with those on ART and HIV-1–negative adults (10–12). Whether this indicates expression of cytotoxic effector proteins by polyfunctional HIV-1–specific CD8⁺ T cells, the emergence of a cytotoxic HIV-1–specific effector T (T_EFF) cell subset, and/or bystander activation-induced perforin and GzmB expression in non–HIV-1–specific CD8⁺ T cells in the mucosa during chronic HIV-1 infection is unknown.

Interestingly, despite low perforin expression, gastrointestinal CD8⁺ T cells express GzmB, albeit at lower levels compared with blood, and display ample degranulation, measured by expression of CD107, suggesting that these mucosal CD8⁺ T cells may produce and release cytotoxic granules containing little or no perforin (11, 13). Numerous in vitro and mouse in vivo studies have demonstrated noncytolytic, extracellular functions of GzmA, B, and K, including extracellular matrix degradation and the direct and indirect processing and release of proinflammatory cytokines (14–17). The concept of Gzms functioning in a noncytolytic manner to elicit a proinflammatory state appears consistent with the strong propensity of gut CD8⁺ T cells to produce chemokines and cytokines in response to antigenic stimulation. Accordingly, the major goal of this study was to more fully explore the expression and coexpression patterns of GzmA, B, and K in parallel with perforin in HIV-1–specific and non–HIV-1–specific CD8⁺ T cells in rectosigmoid mucosa during chronic HIV-1 infection.

HIV-1 positive and seronegative (SN) participants were enrolled through the SCOPE study based at San Francisco General Hospital. Written informed consent for phlebotomy and rectosigmoid biopsy was obtained through protocols approved by the Committee on Human Subjects Research, University of California, San Francisco (protocols 10-01218, 10-00263, and 10-01330). Participants were classified in four groups defined by plasma viral load (VL) and ART as follows: controllers (C, n = 10) consistently maintained VL <2000 copies per ml without ART; viremic untreated (V, n = 15) subjects maintained VL ≥2000 copies per ml without ART; ART-treated (Tx, n = 13) participants had VL <40 copies per ml; and SN (n = 10) controls were negative for HIV-1. SN participants were healthy volunteers from whom samples were collected for research purposes only. Exclusion criteria included active sexually transmitted infections other than HIV-1; inflammatory bowel disease or other known inflammatory conditions affecting the gastrointestinal tract; colorectal cancer or other malignancies; and severe anemia or blood clotting disorders. Additional demographic information (gender, race/ethnicity, age, VL, CD4 count, and time since diagnosis) is supplied in Table I. In total, 20–40 ml of blood was collected by sterile venipuncture into tubes containing EDTA. Approximately 30 rectosigmoid biopsies were obtained by flexible sigmoidoscopy 10–30 cm from the anal verge (18). Biopsies were collected in 50 ml conical tubes containing 15 ml of R-15 medium [RPMI 1640 supplemented with FCS (15%), penicillin (100 U/ml), streptomycin (100 mg/ml), and glutamine (2 mM)]. Blood and biopsy samples were transported at room temperature to the laboratory and processed immediately upon arrival.

Ficoll-Paque (Pfizer-Pharmacia, New York, NY) was used to isolate PBMCs from blood. Rectosigmoid mucosal mononuclear cells were liberated from biopsies using enzymatic and mechanical disruption as previously described (19, 20). Briefly, biopsies underwent shaking incubation at 37°C for 30 min in 25 mg/ml Liberase DL (Roche, Indianapolis, IN), followed by passage through a 16-gauge blunt-end needle to disrupt tissue. Free cells were collected through a sterile 70 μm cell strainer. The disruption process was repeated until all biopsy tissue was digested. Free rectosigmoid mucosal mononuclear cells were washed three times in 20 ml of R-15. All isolated cells were rested overnight at 37°C, 5% CO₂ in R-15 supplemented with 200× Zosyn (Pfizer-Pharmacia).

The following fluorochrome-labeled monoclonal Abs were used in flow cytometry: CD107a (H4A3: PE-Cy5, PE-Cy7), CD8 (SK1: APC-H7, FITC), CCR7 (3D12: PE-Cy7), GzmB (GB11: V450, PE-CF594), MIP-1β (D21-1351: APC-H7, PerCP-Cy5.5), IFN-γ (B27: PE-Cy7), TNF-α (Mab11: Ax700), CD69 (L78: PE) from BD Biosciences (San Jose, CA); GzmK (GM6C3: FITC) from Santa Cruz Biotechnology (Dallas, TX); CD4 (T4D11: ECD) from Beckman Coulter (Brea, CA); CD4 (RPA-TA: BV785), CD45RO (UCHL1: BV785), CD27 (O323: BV650), CD103 (Ber-ACT8: Ax488), GzmA (CB9: Ax647, PacBlue), and CD3 (UCHT-1: Ax700) from BioLegend (San Diego, CA). CD3 (S4.1: Qdot655) was purchased from Invitrogen (Carlsbad, CA). Sphingosine-1-phosphate receptor (S1PR1) (SW4GYPP: eFluor660) was purchased from eBioscience (San Diego, CA); unlabeled CD28 (L293) and CD49d (L25) were from BD Pharmingen (San Diego, CA). Perforin (B-D48: PE) was from Cell Sciences (Canton, MA). This perforin Ab detects de novo as well as preformed perforin and is suitable for monitoring perforin expression following TCR stimulation (21). The HIV-1 Gag peptide pool (Consensus Clade B) consisted of 123 15-mers overlapping by 11 residues (JPT Innovative Peptide Solutions, Berlin, Germany). Staphylococcal enterotoxin B (SEB) was from Sigma-Aldrich.

Staining was performed as previously described with slight modifications (12) on ex vivo blood and mucosal mononuclear cells rested overnight at 37°C, 5% CO₂. For stimulation assays, cells were incubated at 2 × 10⁶ cells per 200 μl R-15 for 5.5 h in the presence of anti-CD28 (1 μg/ml) and anti-CD49d (1 μg/ml), anti-CD107a, 1 μM GolgiStop (BD Biosciences), brefeldin A (5 μg/ml; Sigma-Aldrich), and the appropriate antigenic stimulation: Gag-peptide pool (final concentration 3.5 μg/ml of each peptide), or medium containing an equivalent amount of DMSO (peptide solvent) as the negative control. SEB (5 μg/ml) was used as positive control. Following incubation, cells were stained for surface markers and viability (Aqua Live/Dead cell stain kit; Invitrogen) for 20 min at room temperature. Cells were then fixed in 4% paraformaldehyde and permeabilized using FACS Perm 2 (BD Biosciences) prior to intracellular staining for CD3, cytotoxic effectors, chemokines, and cytokines. Staining for cytotoxic effectors following antigenic stimulation used staining protocols and Ab clones previously demonstrated to reveal perforin and GzmB expression following TCR stimulation (21). Cells were resuspended in 1% paraformaldehyde and stored at 4°C in the dark no longer than 24 h until analysis.

An LSR II flow cytometer and FACSDiva software (Becton Dickinson Immuno-Cytometry Systems) were used to collect flow cytometry data. Flowjo software (Flowjo, Ashland, OR) was used to perform initial flow cytometric analysis and SPICE was used to visualize multifunctional T cell populations (22). CD8⁺ T cells were identified using the following gating strategy: lymphocytes [side scatter -A versus forward scatter (FSC)-A] → singlets (FSC-H versus FSC-A) → live, CD3⁺ (CD3 versus viability) → CD8⁺, CD4^Neg (CD8 versus CD4). Memory subsets within the CD8⁺ T cell population were identified by first establishing positive gates for each memory marker within the live, CD3⁺, CD4^Neg, CD8⁺ T cell population followed by Boolean gating to define memory subsets as follows: naive (CD45RO⁻CD27⁺CCR7⁺), central memory (CD45RO⁺CD27⁺CCR7⁺), transitional memory (T_TM CD45RO⁺CD27⁺CCR7⁻), effector memory (T_EM CD45RO⁺CD27⁻CCR7⁻), and T_EFF (CD45RO⁻CD27⁻CCR7⁻) (23). Polyfunctional cells and cells coexpressing cytotoxic effector molecules were identified by first establishing positive gates for each response analyte and/or cytotoxic effector molecules within live, CD3⁺, CD4^Neg, CD8⁺ T cell populations, followed by Boolean gating to define subsets coexpressing these effector molecules. We then used Boolean gating to define memory marker expression within each cytotoxic effector subset. An in-house statistical algorithm was used to determine whether Ag-specific responses differed significantly from the negative control (12). For responses found to be statistically significant, net responses were then calculated by subtracting the negative control values from Ag-specific responses. Statistical analyses were performed with GraphPad Prism V.6 (GraphPad Software, San Diego, CA) or SPICE using nonparametric statistical test as follows: Wilcoxon matched-pairs signed rank test for paired samples, Mann–Whitney U test for unpaired samples, and Spearman correlation and linear regression analysis for correlative analysis.

Expression of perforin, and to a lesser extent GzmB, is reduced in resting and stimulated CD8⁺ T cells in gastrointestinal mucosa compared with blood, but whether this trend extends to other Gzms, or to their coexpression, has not been fully explored (13). It is also unclear whether coexpression of these proteins varies with HIV-1 disease status. To address these questions, we compared intracellular expression of GzmA, B, K and perforin in resting CD8⁺ T cells freshly isolated from blood and rectosigmoid mucosal biopsies of participants in the following groups: HIV-1⁺ controllers (C); HIV-1⁺ viremic (V) individuals; HIV-1⁺ individuals on ART (Tx); and SN controls (Table I).

GzmB, GzmK, and perforin all displayed bimodal expression patterns with distinctive positive and negative populations (Fig. 1A). GzmA displayed a trimodal expression pattern, with a small fraction of CD8⁺ T cells identified as GzmA^Bright, expressing GzmA at high fluorescence intensity, and a large subset displaying intermediate intensity, designated as GzmA^Int (Fig. 1A). The GzmA^Bright population was particularly prominent in mucosa. Irrespective of HIV-1 disease status, GzmA was the most abundantly expressed cytotoxic effector protein in both blood and mucosa, whereas perforin was the rarest (Fig. 1A, 1B). In agreement with our previous work, the proportion of CD8⁺ T cells expressing GzmB and perforin was greater in blood compared with mucosa in both SN and HIV-1⁺ participants (11, 13). In contrast, the relative abundance of GzmA^Int cells in blood and gut differed between HIV-1⁺ and SN participants. In HIV-1⁺ individuals, the proportion of GzmA^Int CD8⁺ T cells was similar in blood and mucosa; in SNs, it was significantly greater in mucosa. No significant differences were detected between blood and gut in the proportion of GzmK⁺ CD8⁺ T cells in either HIV-1⁺ or SN individuals (Fig. 1B).

Expression of cytotoxic effectors varied across HIV-1 disease status. With the exception of GzmA^Bright cells, which were most abundant in SNs, percentages of CD8⁺ T cells expressing cytotoxic effectors were typically higher in HIV-1⁺ groups compared with SN in both tissues, with the highest percentages observed in HIV-1⁺ individuals not on ART (Fig. 1C, Supplemental Fig. 1A). Among HIV-1⁺ groups, C and V individuals not on ART had higher percentages of GzmA^Int and GzmK-positive rectosigmoid CD8⁺ T cells compared with ART-Tx participants. V individuals not on ART also had greater GzmB expression in the gut compared with ART-Tx subjects (Fig. 1C). Similar trends were observed in blood but not as pronounced (Supplemental Fig. 1A).

Taken together, these data demonstrate elevated expression of GzmA^Int, B, K, and perforin in rectosigmoid CD8⁺ T cells from HIV-1⁺ individuals, particularly those not on ART, compared with SNs, and to our knowledge highlight for the first time a novel subset of CD8⁺ T cells expressing GzmA at high fluorescence intensity in the mucosa.

Differences in the abundance of individual cytotoxic effectors led us to investigate the coexpression patterns of these molecules in mucosa. In blood, HIV-1–specific CD8⁺ T cells have been previously characterized as Perforin^−/+ GzmB⁺ GzmA⁺ GzmK⁺ (24). As expression patterns of cytotoxic effectors were elevated in blood CD8⁺ T cells of HIV-1⁺ individuals compared with SNs, we predicted that their coexpression would likewise be elevated in the mucosa of HIV-1–infected compared with healthy individuals. To test this prediction, we used flow cytometry and SPICE software to analyze coexpression of cytotoxic effectors in unstimulated CD8⁺ T cells from rectosigmoid mucosa and blood of chronically HIV-1–infected and SN adults (Fig. 2). This dataset captured information from all CD8⁺ T cells, regardless of specificity.

Irrespective of HIV-1 disease status, we observed significant differences in the percentages of CD8⁺ T cells coexpressing certain cytotoxic effectors between blood and mucosa, most notably in the proportions of Perforin⁺GzmA^IntGzmB⁺ and GzmA^Bright CD8⁺ T cells. Although in both tissues, perforin was typically coexpressed with GzmA^Int and GzmB, the percentage of Perforin⁺GzmA^IntGzmB⁺ CD8⁺ T cells was significantly higher in blood compared with mucosa (Fig. 2A, 2B). This putative cytotoxic subset was one of the most abundant in blood (Blood_HIV+ median 17.1%, Blood_SN median 6.17%) but one of the rarest in mucosa (Rectosigmoid_HIV+ median 1.18%, Rectosigmoid_SN median 0.508%). In contrast, the mucosa contained a significantly higher percentage of GzmA^Int single-positive CD8⁺ T cells compared with blood (Rectosigmoid_HIV+ median 17.6%, Rectosigmoid_SN median 28.9%; Blood_HIV+ median 2.69%, Blood_SN median 1.43%). A similar trend was observed with GzmA^Bright CD8⁺ T cells. Interestingly, in addition to differences in abundance, the coexpression pattern of GzmA^Bright also differed between blood and mucosa. In mucosa, GzmA^Bright CD8⁺ T cells typically lacked expression of any other cytotoxic effectors; in contrast, blood GzmA^Bright CD8⁺ T cells typically coexpressed perforin and GzmB (Fig. 2A, 2B).

Perhaps not surprisingly, HIV-1⁺ participants also had higher median frequencies of CD8⁺ T cells coexpressing cytotoxic effectors compared with SNs in both blood and gut, including a large proportion of CD8⁺ T cells coexpressing GzmA^Int, B, and K (Rectosigmoid_HIV+ median 21.7%, Rectosigmoid_SN median 6.89%; Blood_HIV+ median 15.9%, Blood_SN median 5.89%) (Fig. 2B). To elaborate on this observation, we compared the abundance of coexpressing subsets across HIV-1 disease status (Fig. 2C, Supplemental Fig. 1B). In mucosa, the proportions of GzmA^IntGzmB⁺GzmK⁺ and GzmA^IntGzmB⁺ CD8⁺ T cells were significantly greater in all HIV-1⁺ subgroups compared with SNs. The proportion of GzmA^IntGzmB⁺GzmK⁺ CD8⁺ T cells was also greater in V individuals compared with those on ART. Both C and V individuals not on ART had significantly higher percentages of phenotypically cytotoxic CD8⁺ T cells, defined as those coexpressing perforin and Gzms, compared with SNs. Specifically, C and V individuals not on ART had higher percentages of Perforin⁺GzmA^IntGzmB⁺GzmK⁺ CD8⁺ T cells compared with SNs; C had higher percentages of Perforin⁺GzmA^IntGzmB⁺ CD8⁺ T cells compared with SNs. In contrast, SNs displayed greater frequencies of GzmA^Int and GzmA^Bright single-positive CD8⁺ T cells compared with all HIV-1⁺ subgroups (Fig. 2C). Similar trends were observed in the blood (Supplemental Fig. 1B).

Taken together, these data reveal that, in general, the rectosigmoid mucosa contains a lower percentage of phenotypically cytotoxic CD8⁺ T cells compared with blood, but an abundance of GzmA single-positive CD8⁺ T cells. Furthermore, individuals with chronic HIV-1 infection, particularly those not on ART, have greater coexpression of cytotoxic effector proteins in mucosal CD8⁺ T cells compared with uninfected individuals.

We next turned our attention to Ag-specific CD8⁺ T cells. To explore the range of functionality exhibited by Ag-specific rectosigmoid CD8⁺ T cells, lymphocytes isolated from blood and mucosa of HIV-1⁺ individuals during early and chronic infection or SN controls (Table I) were stimulated with an HIV-1 Gag peptide pool or SEB in a 5.5 h ex vivo stimulation assay followed by intracellular staining for MIP-1β, IFN-γ, and TNF-α along with CD107a, GzmB, and perforin as surrogates of cytotoxicity. For these experiments we used reagents and protocols previously demonstrated to reveal perforin newly produced in response to TCR stimulation (21). Expression of perforin and GzmB was measured within Gag- and SEB-responding CD8⁺ T cells as defined by one or more of the following responses: degranulation (CD107a), MIP-1β, IFN-γ, or TNF-α production (Fig. 3A).

In both HIV-1–infected and SN participant groups, a large fraction of rectosigmoid CD8⁺ T cells responding to antigenic stimulation did not express detectable perforin or GzmB (median_Gag 49.6%, median_SEB 74.5%) (Fig. 3B). For both Ags, the proportion of total responding CD8⁺ T cells coexpressing perforin and GzmB was significantly greater in blood compared with mucosa (Fig. 3B); in contrast, the proportion of total responding CD8⁺ T cells expressing neither perforin nor GzmB was greater in mucosa compared with blood, reaching statistical significance for SEB (Fig. 3, Supplemental Fig. 2). As no statistically significant differences were observed between HIV-1⁺ subgroups, HIV-1⁺ participant data were consolidated into a single group. Interestingly, in mucosa but not blood, the proportion of responding CD8⁺ T cells expressing GzmB in the absence of perforin (Perforin^NegGzmB⁺) was significantly greater for Gag compared with SEB stimulation (medians 38.7% and 19.9% respectively) (Fig. 3B). The Gag-specific Perforin^NegGzmB⁺ CD8⁺ T cell response was also greater in mucosa compared with blood (Fig. 3B).

We next assessed the expression of perforin and GzmB within CD8⁺ T cell subsets producing multiple cytokines/chemokines and/or CD107, and again found differences between Gag- and SEB-responding CD8⁺ T cells. Similar to previous observations in blood (7), the proportion of Gag-responding CD8⁺ T cells expressing perforin and/or GzmB tended to increase as cytokine polyfunctionality decreased in both blood and gut (Fig. 3D). This trend was especially pronounced in mucosa, where the majority of CD8⁺ T cells responding to Gag with three or four soluble markers and/or CD107, did not express perforin or GzmB (medians 60.18 and 67.33% for three- and four-function cells, respectively) (Fig. 3D). In contrast, single-function mucosal Gag-responding CD8⁺ T cells displayed the greatest GzmB expression (Fig. 3D). This was not observed for blood SEB-responding CD8⁺ T cells, in which the proportion of perforin and/or GzmB-expressing cells appeared greater in polyfunctional compared with single-function CD8⁺ T cells (Supplemental Fig. 2C, 2D). No obvious trend emerged for SEB in the mucosa, as the majority of rectosigmoid SEB-responding CD8⁺ T cells did not express perforin or GzmB, irrespective of cytokine expression (Supplemental Fig. 2).

Taken together, these data demonstrate an apparent inverse relationship between polyfunctionality (defined as coexpression of multiple cytokines/chemokines and CD107a) and cytotoxic potential (defined as coexpression of perforin and GzmB) in rectosigmoid HIV-1 Gag-specific CD8⁺ T cells, and highlight discordance in the expression of perforin and GzmB between Gag- and SEB-responding CD8⁺ T cells in the mucosa.

The data presented in Fig. 3 suggest that most cytokine/chemokine-producing HIV-1–specific rectosigmoid CD8⁺ T cells do not express perforin and GzmB. The abundance of mucosal CD8⁺ T cells expressing Gzms independently of perforin led us to re-examine this result using a broader range of cytotoxic effectors. As before, lymphocytes isolated from blood and rectosigmoid mucosa of chronically HIV-1–infected and SN adults were stimulated with an HIV-1 Gag-peptide pool or SEB in a 5.5 h ex vivo stimulation assay. However, in this series of experiments, stimulation was followed by intracellular staining for perforin, GzmA, GzmB, and GzmK, along with MIP-1β and CD107 (Fig. 4). As CD107a and MIP-1β expression were the two strongest responses previously detected in rectosigmoid mucosa in response to HIV-1 Gag and SEB stimulation, they were used in this study as indicators of Ag responsiveness (10, 12).

For both HIV-1 Gag and SEB stimulation, the mucosa displayed a greater proportion of responding CD8⁺ T cells not expressing any cytotoxic effectors compared with blood. These Perforin⁻GzmA⁻GzmB⁻GzmK⁻ responses dominated in the mucosa and were particularly striking for SEB-responding CD8⁺ T cells (Fig. 4C). Rectosigmoid mucosa also exhibited greater proportions of GzmA single-positive Ag-responding CD8⁺ T cells compared with blood. In contrast, blood had greater proportions of Perforin^+/−GzmA⁺GzmB⁺–responding CD8⁺ T cells compared with mucosa (Fig. 4C). Differences between antigenic stimulations were also observed. In mucosa but not blood, Gag-responding CD8⁺ T cells exhibited a larger proportion of GzmA⁺GzmB⁺GzmK⁺ CD8⁺ T cells compared with SEB-responding CD8⁺ T cells (Fig. 4C, data not shown). In contrast, rectosigmoid SEB-responding CD8⁺ T cells included a greater proportion of cells not expressing any cytotoxic effectors compared with Gag-responding CD8⁺ T cells (Fig. 4C, data not shown).

These findings expand upon the data presented in Figs. 1 and 3, and further support the interpretation that rectosigmoid CD8⁺ T cells are typically primed for cytokine expression rather than cytotoxic effector production. However, this observation may be partly Ag dependent, as rectosigmoid Gag-specific CD8⁺ T cells displayed proportionately greater expression of GzmA, B, and K compared with SEB-responding cells.

Based on observations in blood, it is thought that CD8⁺ T cells acquire expression of cytotoxic effectors as they mature into effector subsets. For example, Perforin⁻GzmA⁻GzmB⁻, GzmA⁺GzmB⁺GzmK⁺, and Perforin⁺GzmA⁺GzmB⁺ subsets are associated with central memory, transitional, and effector maturation stages, respectively (24–26). However, despite housing a high proportion of T_EM cells (27), the mucosa displays an abundance of CD8⁺ T cells expressing either only GzmA or not expressing any cytotoxic effectors (Fig. 1). To address this apparent discrepancy, we assessed the intracellular expression of cytotoxic effectors in conjunction with T cell memory markers CD45RO, CCR7, and CD27 on unstimulated blood and rectosigmoid CD8⁺ T cells from chronically HIV-1–infected and SN adults.

Not surprisingly, blood CD8⁺ T cells not expressing any cytotoxic effectors primarily had a naive phenotype (Supplemental Fig. 3), whereas in rectosigmoid mucosa most quadruple-negative Perforin⁻GzmA⁻GzmB⁻GzmK⁻ CD8⁺ T cells were memory cells, with a large fraction displaying a T_TM phenotype (Fig. 5). Because the proportion of GzmA^IntGzmB⁺GzmK⁺ was elevated in the mucosa of HIV-1⁺ individuals not on ART (Fig. 2C), we were interested in the memory phenotype of these cells. Rectosigmoid subsets expressing GzmK, such as GzmA^IntGzmB⁺GzmK⁺ and GzmA^IntGzmK⁺ CD8⁺ T cells, exhibited a more pronounced T_TM phenotype than their counterparts in blood (Fig. 5, Supplemental Fig. 3). Mucosal single-positive GzmA^Int/Bright CD8⁺ T cells were predominantly T_EM cells (Fig. 5). In blood, GzmA^Bright cells were absent, and GzmA^Int cells were distributed among T_TM, T_EM, and T_EFF categories. The relatively small subset of rectosigmoid CD8⁺ T cells expressing perforin, primarily Perforin⁺GzmA^IntGzmB⁺ CD8⁺ T cells with or without GzmK, was comprised of T_EFF cells with some T_TM and T_EM (Fig. 5B), a distribution similar to that of Perforin⁺GzmA^IntGzmB⁺ CD8⁺ T cells in blood (Supplemental Fig. 3). Taken together, these data reveal differences between blood and gut regarding the memory phenotype of CD8⁺ T cells lacking expression of cytotoxic effectors, and identify the GzmA^Bright CD8⁺ T cell subset, limited to mucosa, as T_EM cells.

The human rectosigmoid mucosa houses a large fraction of CD8⁺ T cells with a tissue-resident phenotype, including in chronic HIV-1 infection (28). As GzmA^Bright CD8⁺ T cells were abundant in mucosa but not blood, and primarily displayed a T_EM phenotype, we hypothesized that these cells might also express markers of tissue residency. To test this, we analyzed intracellular expression of GzmA and GzmB in conjunction with the canonical tissue-resident markers CD69, integrin αE (CD103), and S1PR1 on unstimulated blood and rectosigmoid CD8⁺ T cells from chronically HIV-1–infected and SN adults.

Irrespective of HIV-1 disease status, the overwhelming majority of rectosigmoid GzmA^Bright CD8⁺ T cells were CD69⁺CD103⁺S1PR1⁻, indicative of tissue residency (Fig. 6). In contrast, the expression of tissue-resident markers on GzmB⁺ and GzmA^Int CD8⁺ T cells varied across HIV-1 disease status. In ART-Tx and healthy adults, GzmB⁺ and GzmA^Int CD8⁺ T cells were primarily CD69⁺CD103⁺S1PR1⁻. In contrast, in HIV-1⁺ individuals not on ART, GzmB⁺ and GzmA^Int CD8⁺ T cells were primarily single-positive for CD69 or lacked expression of all three markers (Fig. 6). Together, these data support the interpretation that GzmA^Bright CD8⁺ T cells are likely a subset of tissue-resident CD8⁺ T cells unique to the mucosa, whose functions and antigenic specificity remain to be determined.

The gastrointestinal mucosa is an important site of HIV-1 transmission and viral replication, and is generally considered as one of the front lines of defense against mucosal pathogens. Understanding how mucosal CD8⁺ T cells respond to microbial challenge is important for developing therapeutic and preventative strategies targeted to mucosal tissues. Previous reports have demonstrated that rectosigmoid CD8⁺ T cells are robust producers of chemokines and cytokines but have limited perforin expression, supporting the interpretation that gut CD8⁺ T cells are primed for cytokine and chemokine secretion rather than cytotoxicity (11–13, 29). However, missing from earlier studies of mucosal T cells were detailed analyses of cytolytic effector proteins GzmA, B, and K and their patterns of expression or coexpression with cytokines and chemokines. In the current study we provide a more comprehensive view of the effector molecules produced by rectosigmoid CD8⁺ T cells, which may be of particular interest in the light of recent literature demonstrating novel extracellular, proinflammatory functions of Gzms (14, 15, 17, 30, 31).

As previously reported, the percentages of CD8⁺ T cells expressing perforin and GzmB were significantly reduced in rectosigmoid mucosa compared with blood (11–13, 29). In contrast, the percentages of GzmK- and GzmA-positive CD8⁺ T cells were similar between the two tissues. These observed differences in expression levels of cytotoxic effectors suggest differential regulation at the transcriptional and/or posttranscriptional level. Interestingly, the gene for human GzmB is located on chromosome 14 whereas genes for GzmA and GzmK, whose expression was not deficient in the gut, are located in close proximity to one another on chromosome 5 (30, 32). Little is known about the transcriptional regulation of GzmA and GzmK but their spatial separation from GzmB might suggest differing transcriptional regulation (33). Differences in posttranscriptional regulatory mechanisms may also account for the variations in expression levels. For example, although resting murine NK cells express mRNA for perforin, GzmB, and GzmA, they must be activated to express perforin and GzmB protein; in contrast, both resting and activated NK cells express high levels of GzmA protein (34).

Although GzmA was the most abundantly expressed cytotoxic effector in both blood and gut, we observed discordance in its expression pattern between tissues. In mucosa, a large percentage of CD8⁺ T cells were single positive for GzmA; a subset of these expressed GzmA at high levels (GzmA^Bright) and displayed a T_EM, tissue-resident phenotype (CD103⁺CD69⁺S1PR1⁻) (35, 36). In blood, however, GzmA^Bright CD8⁺ T cells were scarce and GzmA was nearly always coexpressed with other cytotoxic effectors. These data suggest that similar to perforin and GzmB, GzmA expression may be regulated differently in rectosigmoid CD8⁺ T cells compared with blood.

In contrast to the abundance of GzmA, the percentage of CD8⁺ T cells coexpressing perforin and Gzms was sharply reduced in the gut compared with blood, further evidence of the diminished cytotoxic potential of rectosigmoid CD8⁺ T cells (11). We previously speculated that perforin and GzmB are tightly regulated in mucosa to avoid bystander damage to the fragile mucosal epithelium (11). As Gzms are believed to be unable to induce target-cell apoptosis without perforin (37), one interpretation of these data is that limited perforin expression enables rectosigmoid CD8⁺ T cells to use Gzms for noncytolytic functions. Consistent with this interpretation, although the cytotoxicity of GzmB and perforin are well established, controversy remains as to whether other Gzms contribute significantly to target-cell apoptosis in vivo (15, 38). Rather, GzmA and GzmK may function in extracellular, noncytotoxic capacities in the mucosa. For example, GzmA and GzmK have been shown to facilitate cytokine responses to the microbial cell-wall component LPS (16, 39). Additionally, GzmA, B, and K have demonstrated extracellular activities that promote cellular migration and inflammation (14, 15, 17, 30, 31). Proinflammatory functions of Gzms appear consistent with the polyfunctional nature of rectosigmoid CD8⁺ T cells, whose expression of cytokines and β-chemokines likely have similar proinflammatory effects (40, 41).

Disparities in the expression levels of cytotoxic effectors varied with HIV-1 disease status. The percentages of CD8⁺ T cells expressing GzmA^Int, GzmB, GzmK, and perforin were higher in HIV-1–infected compared with SN adults, with the highest percentages observed in HIV-1⁺ individuals not on ART. Interestingly, the opposite trend was observed for single-positive GzmA^Bright cells. One interpretation is that HIV-1 infection results in upregulation of other cytotoxic effectors in GzmA^Bright CD8⁺ T cells. The overwhelming majority of GzmA^Bright CD8⁺ T cells expressed αE integrin (CD103), which is associated with proximity to epithelial cells expressing the CD103 ligand, E-cadherin (42–44). A characteristic of HIV-1 infection is loss of epithelial barrier integrity (45, 46). Accordingly, an alternative explanation is that lower proportions of GzmA^Bright CD8⁺ T cells in HIV-1 infection reflects a loss of CD103⁺ CD8⁺ T cells, due in part to reduced availability of epithelial-derived E-cadherin.

Notably, the proportion of CD8⁺ T cells coexpressing GzmA, B, and K was significantly greater in HIV-1–infected compared with SN participants in both blood and gut. In blood, coexpression of GzmA, B, and K is associated both with an intermediate memory maturation status and HIV-1 specificity (24–26). Similarly, rectosigmoid GzmA⁺GzmB⁺GzmK⁺ CD8⁺ T cells primarily displayed a T_TM phenotype and a substantial fraction of rectosigmoid Gag-responding CD8⁺ T cells coexpressed GzmA, B, and K. Whether coexpression of GzmA, B, and K in Gag-specific CD8⁺ T cells is a reflection of maturation status or a consequence of HIV-1–specific priming of the immune response will require further exploration. However, previous work has demonstrated that blood HIV-1–specific CD8⁺ T cells are skewed toward an intermediate maturation stage, suggesting the former (47).

Consistent with the hypothesis that rectosigmoid CD8⁺ T cells are skewed toward cytokine production rather than cytotoxicity, the proportions of Gag- and SEB-responding CD8⁺ T cells expressing cytotoxic effectors were significantly lower in rectosigmoid mucosa compared with blood. Furthermore, we observed that perforin and GzmB expression increased in Gag-specific CD8⁺ T cells as polyfunctionality decreased, suggesting that cytotoxic and cytokine/chemokine-producing Gag-specific CD8⁺ T cells are to some extent separate subsets. In the mucosa, responding CD8⁺ T cells often lacked expression of perforin and Gzms. As postulated elsewhere (11), degranulation in the absence of cytotoxic effectors may reflect the release of other granule components like MIP-1α and MIP-1β (48). Alternatively, it may reflect differences in the kinetics of cytotoxic effector upregulation between blood and gut CD8⁺ T cells. In addition to differences between tissues, we also observed differences in expression of cytotoxic effectors between HIV-1 Gag- and SEB-responding CD8⁺ T cells in the mucosa. The proportion of rectosigmoid Gag-responding CD8⁺ T cells expressing cytotoxic effectors was greater compared with SEB-responding CD8⁺ T cells, indicating that expression of cytotoxic effectors is, to a certain degree, influenced by Ag specificity. It is currently unclear whether elevated cytotoxic effector expression in Gag-responding rectosigmoid CD8⁺ T cells reflects the upregulation of cytotoxic effectors by tissue resident CD8⁺ T cells or, alternatively, an influx of cytotoxic effector-expressing CD8⁺ T cells from the circulation. Addressing this question will require further work, likely involving experimental infection studies in nonhuman primate models.

Interestingly, unlike GzmA^Bright CD8⁺ T cells that were overwhelmingly tissue resident, a large fraction of GzmB⁺ and GzmA^Int rectosigmoid CD8⁺ T cells did not express the phenotypic markers of tissue residency, suggesting they might be recent immigrants. Notably, Gag-specific polyfunctional CD8⁺ T cells were not completely devoid of perforin and GzmB, suggesting a small fraction of Gag-specific CD8⁺ T cells may be capable of both cytotoxic and polyfunctional responses. Determining whether the coexpression of cytotoxic effectors with cytokines/chemokines confers a protective advantage against HIV-1 will require further analysis. No differences were observed between HIV-1 C and non-C in the proportion of Gag-specific CD8⁺ T cells expressing perforin and/or Gzms, although this analysis was limited by small sample size.

Taken together, these data demonstrate that rectosigmoid CD8⁺ T cells produce Gzms independently of perforin, likely complementing dominant cytokine and chemokine responses to promote an inflammatory response. The work described in this study supports a model in which rectosigmoid CD8⁺ T cells are skewed toward proinflammatory responses rather than direct cytotoxicity, suggesting new insights into the mucosal adaptive immune response that may inform the development of therapeutics to treat mucosal infections such as HIV-1.

We thank Rebecca Hoh, Montha Pao, Monika Deswal, and the clinical staff at San Francisco General Hospital for assistance with participant recruitment and sample collection. We thank the participants for willingness to contribute to this study. We thank the University of California Davis Flow Cytometry Shared Resource Laboratory staff members Bridget McLaughlin and Jonathan Van Dyke for assistance with flow cytometry.

The authors have no financial conflicts of interest.