Reciprocal Interactions between Commensal Bacteria and γδ Intraepithelial Lymphocytes during Mucosal Injury¹

The intestinal immune system has coevolved with a vast nonpathogenic luminal microflora. These indigenous bacteria do not pose a significant threat to host health as long as they remain confined within the intestinal lumen. However, the epithelium can be injured by environmental factors such as toxins, rendering the host susceptible to opportunistic invasion by commensals. Thus, it is essential that the intestine be able to defend against opportunistic penetration of commensal bacteria across injured mucosal surfaces.

Intraepithelial lymphocytes (IEL)³ that bear γδ TCRs (γδ IEL) promote repair of injured gut epithelia (1). γδ IEL are intercalated between intestinal epithelial cells, residing on the basolateral side of epithelial tight junctions. Although rare in the circulation, γδ T cells are prominent at intestinal surfaces, where they are endowed with a number of properties that distinguish them from conventional T cells. These include the ability to secrete epithelial growth factors and to produce innate cytokines and chemokines that recruit inflammatory cells (1, 2). Analysis of mice lacking γδ T cells has revealed that γδ IEL play an essential role in promoting epithelial restitution following mucosal injury (1, 3). This function has been linked to up-regulated expression of keratinocyte growth factor (KGF), which stimulates proliferation of colonic epithelial progenitors (1). Consistent with their unique role in tissue repair, KGF expression is a distinctive feature of γδ IEL and does not occur in other mucosal T cell populations, including αβ IEL (1).

Despite the unique contributions of γδ IEL to mucosal healing, the molecular details of the γδ IEL response to intestinal injury remain poorly defined. Furthermore, little is known about the factors that regulate this response. This is due in large part to inherent experimental challenges posed by these cells, including the fact that they are refractory to experimental manipulation outside of their intestinal niche. In this study, we uncover new insights into the role of γδ IEL in maintaining intestinal homeostasis following mucosal injury. Using genome-wide analysis, we elucidate a dextran sulfate sodium (DSS)-induced transcriptional program in colonic γδ IEL that includes orchestrated expression of factors involved in epithelial protection, antibacterial defense, and inflammatory cell recruitment. We further show that commensal microbes direct key elements of the γδ IEL injury response, revealing a dialogue between commensal bacteria and γδ IEL. Finally, we find that γδ T cells are essential for controlling bacterial penetration across injured mucosal surfaces. Our results suggest that intestinal γδ IEL play a multifaceted role in maintaining mucosal homeostasis following injury, and reveal the existence of a dynamic and reciprocal cross-talk between the intestinal microbiota and γδ T cells.

Conventionally raised wild-type and MyD88^−/− C57BL/6 mice (from The Jackson Laboratory) and TCRδ^−/− mice were maintained in the barrier facility at the University of Texas Southwestern Medical Center. Germfree C57BL/6 mice were maintained in plastic gnotobiotic isolators, as previously described (4). All mice were maintained under a 12-h light cycle and were fed the same autoclaved chow diet. Six- to 10-wk-old mice were used for all experiments. For conventionalization studies, germfree mice were colonized with microflora from conventional mice 72 h before treatment with DSS. Dilution plating of luminal bacteria confirmed that these mice were reconstituted to conventional levels. All experiments were performed using protocols approved by the Institutional Animal Care and Use Committees of University of Texas Southwestern Medical Center.

Groups of five to eight C57BL/6 mice received 2% DSS (m.w. 40,000; ICN Biomedicals) in drinking water ad libitum for the indicated periods of time. For examination of mucosal healing, mice were treated with 2% DSS for 5 days, and then were returned to regular drinking water for an additional 3 days. The amount of DSS water consumed per animal was noted, and there were no marked differences between experimental groups. Control mice received water alone.

Total IEL were isolated as previously described (5). Total IEL from five to eight mice were pooled together and stained with PE-labeled anti-TCRγδ (GL3; BD Pharmingen), and γδ IEL were purified on a BD MoFlo cell sorter in the University of Texas Southwestern Flow Cytometry Core Facility. The purity of isolated γδ IEL was assayed postsorting and was always ≥98%.

Total RNAs were isolated from purified γδ IEL using the PicoPure RNA isolation kit (Arcturus). For each experimental condition, RNA was isolated from γδ IEL recovered from two independent groups of five to eight mice. Yields of total RNA were typically 10 ng/group. Total RNA (5 ng) was amplified using the Arcturus RiboAmp HS kit. Biotinylated cRNAs were generated by substituting the Enzo T7 BioArray Transcript kit during the last step and hybridized to Affymetrix Mouse Genome 430 2.0 GeneChips in the University of Texas Southwestern Microarray Core.

To identify genes whose expression was altered by DSS treatment, we performed two-way comparisons between untreated and DSS-treated groups, with untreated samples designated as baseline. Raw data were imported into Affymetrix GeneChip software for analysis, and previously established criteria were used to identify differentially expressed genes (6). Briefly, a ≥2-fold difference was considered significant if three criteria were met, as follows: 1) the GeneChip software returned a difference call of increased or decreased; 2) the mRNA was called present by GeneChip software in either untreated or DSS-treated samples; and 3) the difference was observed in duplicate microarray experiments. GeneChip quality and amplification linearity were assessed using polyadenylated spike-in control transcripts and oligo-B2 hybridization control (Affymetrix). Heatmaps to visualize signal intensities were generated using GeneTraffic software (Iobion). Table S3⁴ lists all gene expression changes identified by this analysis.

To identify DSS-induced genes whose expression is governed by intestinal microflora, we performed a microarray analysis on γδ IEL isolated from untreated and DSS-treated germfree mice. The list of 272 genes altered by DSS treatment in conventional mice was used to recover the corresponding signal intensities from the germfree dataset. Signal intensity data were converted to Z-scores (z = (x – μ)/σ, where x = signal intensity, μ = mean signal intensity for all samples, and σ = SD across all samples) and subjected to unsupervised hierarchical clustering using GeneTraffic software. The cluster analysis was used to identify the subset of probe sets in which the signal intensity in at least three of four germfree samples fell at or below the mean signal intensity averaged across all eight arrays. This subset of genes is depicted in Fig. 3 and in Fig. S1.⁴

Total RNA was isolated from purified γδ IEL using the Arcturus PicoPure RNA isolation kit and was subjected to mRNA amplification with the Arcturus RiboAmp kit. cDNAs were generated from the amplified cRNAs using random primers and were used as a template for Q-PCR with gene-specific primers and SYBR Green Master Mix (Invitrogen). Expression levels were calculated relative to GAPDH. The sequences of Q-PCR primers are as follows: βig-h3, forward 5′-CGAAACCGACATCATGGCCACAAA, reverse 5′-TGGAATACGCTGACGCCTGTTTGA; RegIIIγ, forward 5′-TTCCTGTCCTCCATGATCAAAA, reverse 5′-CATCCACCTCTGTTGGGTTCA; lysozyme, forward 5′-ATGCCTGTGGGATCAATTGCAGTG, reverse 5′-TCTCTCACCACCCTCTTTGCACAT; IL-1β, forward 5′-TGGTACATCAGCACCTCACAAGCA, reverse 5′-AGGCATTAGAAACAGTCCAGCCCA; CXCL-9, forward 5′-TCAGATCTGGGCAAGTGTCCCTTT, reverse 5′-TGAGGTCTATCTAGCTCACCAGCA; MIP2α, forward 5′-GCAGTATTCCTTGGCTGGCCATTT, reverse 5′-ATTCTTCCTACACCGGCATGACCT; KC, forward 5′-TGTGTGGGAGGCTGTGTTTGTATG, reverse 5′-AATGTCCAAGGGAAGCGTCAACAC; and GAPDH, forward 5′-TGGCAAAGTGGAGATTGTTGCC, reverse 5′-AAGATGGTGATGGGCTTCCCG.

For surface staining, isolated IEL were suspended in FACS buffer (PBS and 0.5% BSA), stained for 20 min with PE-conjugated anti-TCRγδ (BD Pharmingen), and washed twice. For intracellular staining, cells were then fixed in 5% formaldehyde in PBS, and permeabilized with PBS containing 0.5% BSA and 0.1% saponin. Staining was done with rabbit anti-lysozyme (Chemicon International) or rabbit anti-RegIIIγ (6), followed by FITC-labeled goat anti-rabbit secondary Ab (Jackson ImmunoResearch Laboratories). Cells were analyzed on a FACSCalibur flow cytometer (BD Biosciences) in the University of Texas Southwestern Flow Cytometry Core Facility.

Conventional C57BL/6 wild-type and TCRδ^−/− mice were cohoused 5 days before initiation of DSS treatment to control for microflora differences between mouse strains. Groups of three to six C57BL/6 wild-type and TCRδ^−/− mice received 2% DSS (m.w. 40,000; ICN Biomedicals) in drinking water ad libitum for up to 5 days. In regeneration studies, mice were returned to regular drinking water for an additional 3 days after 5 days of DSS treatment. Control mice received water alone. Mice were sacrificed, and bacterial translocation to mesenteric lymph nodes (MLN) was determined by dilution plating of homogenized tissue. Intestinal colonization levels were measured by dilution plating of luminal contents.

Intestinal γδ IEL present a number of unique experimental challenges that do not exist for other T cell populations. A key experimental roadblock is the fact that these cells readily undergo spontaneous apoptosis when cultured outside of the gut (7), and therefore cannot be studied extensively in vitro. Insights into the characteristics of γδ IEL relative to other T cell populations have been obtained through in vivo functional genomic studies of γδ IEL isolated from the small intestine (8, 9). However, the relatively low numbers of colonic γδ IEL have precluded genome-wide analysis of this population. This has posed a serious problem for analyzing how γδ IEL respond to injury, because key mucosal damage models, such as the DSS model, specifically damage colonic epithelia (10). As we detail below, we have used mRNA amplification techniques and genome-wide microarray analysis, previously used for analysis of rare epithelial cell populations (6), to gain new insight into the biological functions of colon γδ IEL.

Prior studies have shown that γδ IEL play a unique role in tissue repair through localized epithelial growth factor expression following colonic epithelial injury (1). We hypothesized that this is representative of a more complex response to tissue damage. To test this idea, we used DNA microarrays to gain a genome-wide view of the γδ IEL transcriptional response to epithelial damage. We chose the DSS model of colonic epithelial injury for our studies because it has been used previously to elucidate the contributions of γδ IEL to mucosal repair in mice (1). DSS treatment results in colon-specific epithelial damage that is characterized by focal lesions (10). Because previous studies of γδ IEL function were performed on cells isolated during recovery after DSS-induced damage (1), we conducted our analyses on colonic γδ IEL purified from mice treated orally for 5 days with 2% DSS, followed by 3 days’ recovery. Epithelial cell proliferation was quantitated using BrdU, which is incorporated into the DNA of actively dividing cells. Consistent with prior studies (10), DSS-treated mice showed expanded zones of proliferation in crypts bordering ulcerated areas, indicating active epithelial regeneration (data not shown).

To study the transcriptional program expressed by γδ IEL during mucosal repair, we used flow cytometry to isolate pure γδ IELs from untreated and DSS-treated C57BL/6 mice. Consistent with prior results (11, 12), we obtained ∼10,000 cells per mouse colon, and the numbers of cells isolated from untreated and DSS-treated colons were virtually identical (data not shown). The small numbers of γδ IEL present in the mouse colon initially presented a major obstacle to our study, because our total RNA yields (∼10 ng from pooled cells isolated from five mice) were insufficient for direct functional genomic analyses. We addressed this limitation by performing linear amplification of mRNAs to generate sufficient cRNA for hybridization to Affymetrix Mouse Genome 430 2.0 arrays. To confirm the purity of our cells, we assessed the resulting transcriptional profiles for transcripts representative of other colon intraepithelial and subepithelial cell populations. Although the TCRγ transcript was abundantly present in all samples, the TCRα transcript was undetectable, indicating an absence of αβ IEL. Moreover, there was no detectable expression of the macrophage-specific transcript NO synthase 2, or of RELMβ, an epithelial cell-specific mRNA, confirming the absence of macrophages and epithelial cells in our isolated cell populations.

Comparison of the DSS-treated and -untreated transcriptional profiles revealed that DSS treatment elicits a complex γδ IEL transcriptional response. Expression of 272 transcripts was altered 2-fold or more in γδ IEL isolated from DSS-treated mice (Fig. 1^,A and Table S1).⁴ A total of 217 of these transcripts was enriched, whereas 55 showed reduced abundance. Prior studies have established that DSS treatment leads to increased expression of KGF, which is critical for promoting repair of epithelial lesions (1). Although we were unable to detect KGF transcripts in γδ IEL from either DSS-treated or untreated mice, we observed enhanced expression of transcripts encoding other factors with established cytoprotective properties, including heat shock proteins (13) and the chemokine KC (14) (Fig. 1,B). The microarray also revealed enhanced expression of βig-h3 (Fig. 1,B), a secreted TGF-β-induced factor that supports keratinocyte proliferation and wound healing (15). These findings were substantiated by Q-PCR analysis of amplified mRNAs from independently isolated cell populations (Fig. 2 A), suggesting that γδ IEL regulate epithelial regeneration through multiple factors.

A large proportion of DSS-induced transcriptional changes occurred in genes involved in immunoregulation and inflammatory cell recruitment (Fig. 1^,B and Table S1).⁴ This substantiates the fact that γδ T cells recruit inflammatory mediators into injured tissues (16, 17), and is consistent with prior observations of skin γδ T cells (18). Q-PCR analyses of independently generated samples verified that DSS treatment elicited enhanced expression of the cytokines KC, IL-1β, MIP2α, Cxcl9, and Cxcl16 (Fig. 2 B), which are chemotactic for neutrophils, macrophages, and CD4⁺ T cells (19, 20, 21).

A second prominent category of DSS-regulated transcripts encompassed factors involved in innate immune responses to bacteria. These included the microbial pattern recognition receptors TLR1, which dimerizes with TLR2 to recognize bacterial lipopeptides (22), and CD14, which participates in LPS recognition in complex with TLR4. DSS treatment also enhanced expression of transcripts encoding directly bactericidal proteins, including complement components 1qa and 1qb and lysozyme (Fig. 1,B). Increased lysozyme expression was verified by Q-PCR analysis of amplified mRNA (Fig. 2,C). We also observed increased expression of pancreatitis-associated protein (PAP), a member of the RegIII family of C-type lectins that includes RegIIIγ. Although the biological function of PAP remains undefined, we have previously shown that RegIIIγ is directly bactericidal for Gram-positive organisms (6). Because PAP and RegIIIγ are coexpressed in intestinal epithelial cells (6), we tested whether RegIIIγ expression is also enhanced by DSS treatment. Q-PCR analysis of amplified mRNA revealed a 4.5-fold increase in RegIIIγ expression following DSS treatment (Fig. 2 C).

Because production of directly bactericidal proteins is a previously unappreciated function of γδ IEL, we further analyzed the expression of both RegIIIγ and lysozyme by flow cytometry. The number of γδ IEL-expressing RegIIIγ increased 6.6-fold after DSS treatment, and the number of cells expressing lysozyme underwent a 5-fold increase, in agreement with the Q-PCR analysis (Fig. 2 D). The fact that antimicrobial protein expression is induced in a limited subset of the cell population is consistent with the prior observation that γδ IEL are activated to express KGF in a localized manner, specifically at focal sites of injury (1).

These findings reveal that DSS treatment elicits a complex transcriptional program in intestinal γδ IEL that includes coordinate expression of cytoprotective, immunomodulatory, and antibacterial factors. This suggests that γδ IEL orchestrate multiple responses that restore epithelial integrity, recruit inflammatory cells, and maintain host-microbial homeostasis following intestinal damage.

Several studies have revealed that commensal microbiota play an essential role in maintaining homeostasis during acute DSS-induced colonic injury. As a result, mice lacking intestinal microbes exhibit increased susceptibility to colonic epithelial damage (10, 23). Although bacteria are known to govern many key functions of epithelial cells (4, 6), there has been a lack of experimental support for the idea that intestinal microbes significantly alter the properties of γδ IEL. Challenge experiments suggest that pathogenic bacteria do not trigger significant changes in γδ IEL gene expression (8). However, there have been no studies addressing the role of indigenous intestinal bacteria in modulating the global properties of γδ IEL. Therefore, we sought to test the hypothesis that indigenous intestinal bacteria govern elements of the γδ IEL injury response.

We tested this idea by assessing expression of the complex DSS-induced transcriptional program in germfree mice. BrdU incorporation studies disclosed that germfree C57BL/6 mice exhibit markedly reduced epithelial proliferation during the repair phase following injury (data not shown), in agreement with prior studies (10). We isolated γδ IEL from DSS-treated germfree C57BL/6 mice and untreated germfree controls and determined that the numbers of γδ IEL isolated from both groups were similar to the numbers obtained from conventional mice (data not shown). Duplicate Affymetrix DNA microarray analyses were performed on pooled cells from DSS-treated and untreated germfree mice using the same methods used for the analysis of conventional mice. We then used the list of 272 DSS-regulated transcripts identified by the conventional analysis to recover the signal intensities for these transcripts from the germfree microarray dataset. Signal intensity data were subjected to unsupervised hierarchical clustering and heatmap analysis.

The cluster analysis revealed a subset of genes that was regulated by damage only in conventional mice (68 genes; Fig. 3^,A and Fig. S1).⁴ Commensal microbes were required for DSS-induced expression of a subset of transcripts involved in cytoprotection (e.g., heat shock protein transcripts). These results agree with studies showing that intestinal bacteria enhance heat shock protein and cytokine expression in intact colonic tissue during DSS-induced injury (23). Although one probe set corresponding to βig-h3 clustered with the microbe-regulated subset of genes (Fig. 3^,B), three other βig-h3 probe sets were excluded from this cluster (Fig. S1 and Table S1).⁴ Q-PCR analysis of independently isolated γδ IEL populations revealed that βig-h3 transcripts were more abundant in γδ IEL from DSS-treated germfree mice than from DSS-treated conventional mice (Fig. 4 A), correlating with the increased severity of epithelial wounding in germfree mice (10). Thus, DSS-induced expression of βig-h3 does not require microbial input, suggesting that a component of the γδ IEL-cytoprotective response is elicited independently of bacterial signals.

Remarkably, the microbiota were required for DSS-induced expression of the majority of the proinflammatory cytokine/chemokine transcripts up-regulated by mucosal injury, including KC, IL-1β, MIP2, MIP2α, and CXCL9 (Fig. 3,B). Q-PCR analysis of independently isolated cells verified that the abundance of transcripts encoding the cytokines KC, CXCL9, IL-1β, and MIP2α was enhanced between 6- and 270-fold by DSS treatment of conventional as compared with germfree mice (Fig. 4 B). The fact that commensal microbes are required for DSS-induced expression of these genes makes it unlikely that this transcriptional program is activated by the cell isolation procedure or as a nonspecific response to DSS exposure.

The microarray analysis of germfree mice further identified PAP mRNA as requiring bacterial signals for DSS-induced expression (Fig. 3^,B). Because bacterial signals induce coordinate expression of PAP and RegIIIγ in epithelial cells (6), we assessed whether DSS-induced γδ IEL expression of the bactericidal lectin RegIIIγ also requires microbial input. Q-PCR analysis of amplified γδ IEL mRNA revealed that RegIIIγ transcripts were enriched 38-fold in cells isolated from DSS-treated conventional mice as compared with DSS-treated germfree mice (Fig. 4 C), establishing that a component of the DSS-induced antimicrobial response is governed by intestinal bacteria. In contrast, DSS-induced expression of lysozyme, another component of the γδ IEL antibacterial response, occurred independently of the intestinal microbiota (Table S1).⁴

We next assessed whether the defective injury response exhibited by γδ IEL in germfree mice is reversible. Adult germfree mice were colonized with a complete ileal/cecal microflora harvested from conventionally raised mice 72 h before DSS exposure. Dilution plating of luminal bacteria established that colonization levels of the conventionalized mice were similar to those of conventionally raised mice (∼10⁸ CFU/ml). Q-PCR analysis of amplified γδ IEL mRNA revealed that the conventionalized mice exhibited DSS-induced expression of KC, CXCL-9, IL-1β, MIP2α, and RegIIIγ (Fig. 4, B and C). The defective injury response exhibited by γδ IEL in germfree mice is therefore reversible and is unlikely to be due to an inherent developmental defect in γδ IEL from germfree mice. Thus, commensal bacteria provide critical regulatory input into the γδ IEL response to mucosal injury, eliciting orchestrated expression of directly antibacterial factors and chemotactic cytokines that function in inflammatory cell recruitment. These findings reveal for the first time that indigenous gut microbiota govern a complex transcriptional program in a mucosal T cell population.

We next sought to gain insight into the mechanisms by which bacteria direct γδ IEL responses to DSS-mediated epithelial injury. Several elements of the regenerative response to colonic damage, including production of cytoprotective factors (23) and recruitment of activated macrophages to areas of damage (10), require the TLR signaling adaptor MyD88. To determine whether MyD88-dependent pathways govern microbe-dependent responses to mucosal injury in γδ IEL, we analyzed cells isolated from the colons of untreated and DSS-treated conventional MyD88-deficient mice. Numbers of γδ IEL recovered from MyD88-deficient mice were similar to numbers recovered from wild-type mice (data not shown), in agreement with published data (24). Analysis of amplified cellular mRNAs revealed that DSS treatment failed to trigger CXCL9 and KC expression in conventionally raised MyD88-deficient mice, in contrast to conventional wild-type mice (Fig. 4,B). Thus, CXCL9 and KC expression is induced in γδ IEL through a MyD88-dependent pathway. In contrast, DSS treatment of MyD88-deficient mice elicited enhanced γδ IEL expression of IL-1β, MIP2α, and RegIIIγ transcripts (Fig. 4, B and C), revealing activation through a MyD88-independent mechanism. Intestinal microbes therefore regulate the γδ IEL response to mucosal damage through both MyD88-dependent and MyD88-independent pathways.

Our finding that commensal bacteria orchestrate antimicrobial and proinflammatory responses in γδ IEL following mucosal injury suggested that γδ IEL may play a role in antibacterial defense of damaged surfaces. These cells inhabit the intraepithelial spaces on the basolateral side of epithelial tight junctions that inhibit paracellular crossing of luminal contents, including bacteria (25). However, the integrity of epithelial tight junctions is compromised by DSS-induced injury (26), leaving the host vulnerable to opportunistic invasion by members of the commensal microbiota. Thus, γδ IEL are ideally situated to sense epithelial damage and/or bacterial penetration immediately following injury, and to orchestrate direct antibacterial defenses with recruitment of additional immune cell populations. Based on these observations, we hypothesized γδ IEL may play a role in limiting opportunistic penetration of commensals across damaged mucosal surfaces. To test this idea, we compared numbers of mucosa-penetrant commensal bacteria in wild-type and TCRδ^−/− mice, which lack γδ T cells (27).

Commensal bacteria that penetrate the intestinal barrier spread to the MLN, which confine them to the mucosal immune compartment and prevent their further penetration to the systemic immune system (28, 29). We therefore monitored bacterial penetration of the intestinal mucosa by quantitating MLN bacterial loads. We initially examined mice treated with the 5-day DSS/3-day recovery regimen that was used for our transcriptional analyses. Consistent with prior results (1), we observed increased mucosal damage and delayed epithelial repair in TCRδ^−/− mice (data not shown). However, the numbers of MLN bacteria recovered from TCRδ^−/− mice were not significantly different from wild-type mice using this treatment regimen (data not shown). We reasoned that this might be due to the fact that colonic injury elicits a complex cellular response (10), and that by 5 days of DSS treatment and 3 days of recovery, other recruited immune cell populations (e.g., neutrophils) might be present in sufficient numbers to limit opportunistic penetration of commensals even in the absence of γδ IEL. Thus, we postulated that γδ T cells may function to control bacterial penetration specifically during (as opposed to prolonged) exposure to a damaging agent.

To test this idea, we quantitated bacterial numbers at earlier time points following DSS-induced injury. In agreement with prior studies (1, 10), a time course of DSS treatment resulted in detectable mucosal damage within 3 days of treatment, and this damage was augmented in TCRδ^−/− mice compared with wild-type mice (data not shown). Numbers of MLN bacteria recovered from TCRδ^−/− mice were significantly higher than in wild-type mice after 3 days of DSS treatment (Fig. 5,A), paralleling the development of mucosal damage. We found no significant differences in the numbers of MLN bacteria before 3 days of DSS treatment, suggesting that overt damage to the intestinal mucosa is necessary for increased bacterial penetration (Fig. 5,A). Numbers of MLN bacteria were further elevated after 4 days of DSS treatment, with significantly more bacteria recovered from MLN of TCRδ^−/− mice than wild-type mice. However, by 5 days of DSS treatment, numbers of MLN bacteria were reduced in both TCRδ^−/− and wild-type mice, and were not statistically different between the two groups (Fig. 5,A). We detected no differences in the overall numbers of colonizing anaerobic bacteria between wild-type and TCRδ^−/− mice at any time during the DSS treatment time course (Fig. 5 B), indicating that differences in numbers of penetrating bacteria were not due to intestinal bacterial overgrowth. These results indicate that γδ T cells help to limit opportunistic penetration of commensals across injured mucosal surfaces, and that they carry out this function specifically at early time points following exposure to an epithelium-damaging agent.

Although γδ IEL constitute a major intestinal T cell population, their exact biological functions have remained unclear. In this study, we have demonstrated that γδ IEL mount a complex response to mucosal injury, and that commensal bacteria direct key elements of this response, including expression of immunomodulatory and antibacterial factors. Furthermore, we have shown that γδ T cells are essential for controlling opportunistic penetration of commensal bacteria immediately following damage. In combination with prior studies showing that γδ IEL promote epithelial repair (1), our findings suggest that γδ IEL play a multifaceted role in restoring homeostasis after epithelial damage.

Because low absolute numbers of γδ IEL are present in mouse colon, and because these cells are difficult to manipulate ex vivo, there has been little information regarding the molecular details of how intestinal γδ IEL respond to mucosal injury. Through the application of mRNA amplification techniques and functional genomic approaches, we were able to gain a comprehensive view of the colonic γδ IEL injury response that revealed orchestrated expression of cytoprotective factors, immunomodulatory factors, and directly bactericidal proteins. The regulated production of microbicidal proteins such as RegIIIγ indicates a previously unappreciated function for γδ IEL. Furthermore, this finding suggests that γδ IEL may make multifaceted contributions to antibacterial defense of damaged mucosal surfaces by producing proteins that directly target invading bacteria, and simultaneously initiating a secondary line of defense through recruitment of additional immune cells (Fig. 6).

By analyzing the γδ IEL injury response in germfree mice, we have discovered that commensal bacteria provide critical regulatory input to γδ IEL. Strikingly, intestinal microbes are required for enhanced expression of both the antibacterial factor RegIIIγ and several chemotactic cytokines that function in inflammatory cell recruitment. We further found that γδ IEL injury responses can be restored in adult germfree mice reconstituted with a conventional microbiota. Thus, γδ IEL from germfree mice are not irreversibly defective in their ability to respond to mucosal injury, arguing against an inherent developmental defect in γδ IEL from germfree mice. Rather, they appear to lack the appropriate acute bacterial signals required to drive this program.

Bacterial signaling through the TLR adaptor protein MyD88 is critical for maintaining mucosal homeostasis during DSS-induced epithelial damage (10, 23). Bacterial molecules are detected through MyD88-dependent TLR signaling on subepithelial macrophages, which position themselves next to colonic epithelial progenitors and drive proliferation (10). Similarly, we have shown that a component of the γδ IEL response to DSS-induced injury is governed by bacterial signaling through a MyD88-dependent pathway. This indicates that MyD88-dependent repair of mucosal damage is a complex process involving multiple cell types, including macrophages and γδ IEL. More importantly, it suggests a role for innate pattern recognition in activating γδ IEL responses to tissue damage.

A critical remaining question is whether γδ IEL responses to epithelial injury are elicited by direct interactions between bacteria (or bacterial products) and γδ IEL, or whether other cells, such as epithelial cells or macrophages, detect bacteria and then alter γδ IEL responses through indirect mechanisms of cell-cell communication. Prior studies provide evidence consistent with both models. In vitro experiments suggest that circulating γδ T cells can respond directly to bacterial products (30), supporting the concept of direct bacterial detection. The idea that γδ IEL could directly sense the presence of invading bacteria also makes sense given that these cells inhabit the intraepithelial spaces on the basolateral side of epithelial tight junctions that prevent paracellular crossing of luminal contents (31). Although γδ IEL in healthy, intact epithelia should have limited contact with commensal populations that are confined to the gut lumen, they are strategically situated to detect bacteria that penetrate through damaged epithelia.

There is also experimental support for the idea of indirect bacterial detection by γδ IEL. Studies in skin have revealed that epidermal γδ T cells respond to Ag expressed by stressed keratinocytes (2), indicating the possibility of an indirect detection mechanism. Furthermore, analyses of intestinal IEL have shown that MyD88-dependent IL-15 production by epithelial cells is crucial for γδ IEL development (32). Future studies will be required to distinguish between the direct and indirect bacterial detection models with respect to intestinal γδ IEL mucosal injury responses, and will also require development of mouse models that allow specific genetic manipulation of γδ IEL.

The results of our functional genomic studies prompted us to investigate whether γδ T cells might defend against opportunistic penetration of commensals across damaged mucosal surfaces. We found that γδ T cells were essential for limiting the spread of commensal bacteria to MLN immediately following the appearance of mucosal damage, suggesting that γδ T cells may be protective specifically following acute epithelial insult. The cellular response to prolonged exposure to damaging agents such as DSS is complex (10), and thus, other cell populations that are recruited to damaged areas under the control of γδ T cell-independent signals (33) could restrict the spread of commensals to MLN during chronic injury. A protective role for γδ IEL during acute damage furthermore makes sense given that these cells are well situated to mount an immediate response designed to limit opportunistic penetration of bacteria through epithelial tears. This is likely to be especially important in otherwise healthy individuals in which there may be frequent transient (as opposed to chronic) exposures to damage-inducing environmental factors such as toxins. However, the fact that we detected bacterial regulation of γδ IEL transcriptional responses after 5-day DSS/3-day recovery suggests that commensals impact γδ IEL function even after bacterial spread to MLN has been contained. This could result from continued exposure of γδ IEL to bacterial products and/or damaged epithelial cells, and is consistent with the fact that intestinal mucosa still show signs of damage at this stage (1).

The responses of γδ IEL uncovered by our study provide a plausible molecular explanation for the protective function of these cells. Through coordinated expression of antibacterial and immunomodulatory factors, γδ IEL could limit bacterial penetration through acutely injured mucosa and prevent bacterial spread to MLN. However, it remains to be determined experimentally whether the regulated production of antibacterial and chemotactic factors accounts for this protective function. Given the difficulties in experimentally manipulating γδ IEL in vitro, such studies will require development of animal models that allow cell-specific genetic manipulations of γδ IEL.

In this study, we have shown that γδ IEL engage in a dynamic and reciprocal cross-talk with commensal bacteria. Commensal bacteria provide critical regulatory input to γδ IEL by directing the expression of key immunomodulatory and antibacterial responses following mucosal injury. Although commensals were previously known to elicit complex gene transcription programs in epithelial cells (6), we have shown for the first time that intestinal bacteria also extensively regulate gene transcription in a mucosal T cell population. At the same time, we have found that γδ T cells defend against opportunistic penetration of commensal bacteria immediately following mucosal injury. Thus, γδ IEL make multifaceted contributions to restoring homeostasis after epithelial damage by both promoting epithelial repair and limiting opportunistic invasion of commensals through damaged mucosal surfaces. Together, these findings provide new insights into the role of γδ T cells in maintaining tissue homeostasis, and indicate that bacteria-lymphocyte cross-talk plays a critical role in mucosal immunity.

We thank Dr. David Farrar and Angie Mobley for advice on flow cytometry and for helpful discussions throughout the course of this work. Array data are deposited with Gene Expression Omnibus under accession number GSM343263 (http://www.ncbi.nlm.nih.gov/projects/geo/).

The authors have no financial conflict of interest.