Long-Lived Colitogenic CD4⁺ Memory T Cells Residing Outside the Intestine Participate in the Perpetuation of Chronic Colitis¹

Studies of Ag exposure in murine models of acute virus infection have provided much information about the dynamics of naive, effector, and memory CD8⁺ T cell responses (1, 2, 3). Acute virus infection elicits massive proliferation of viral Ag-specific CD8⁺ T cells, which acquire effector functions (effector phase). After the peak of T cell proliferation, most of the effector CD8⁺ T cells are eliminated (contraction phase). Following virus (Ag) clearance, a small proportion of the remaining cells differentiate into memory CD8⁺ T cells that are maintained by cytokine (IL-7 and/or IL-15)-dependent homeostatic proliferation and survive in the absence of their corresponding Ags (memory phase). Thus, Ag clearance is essential for the emergence of memory CD8⁺ T cells.

In contrast to the generation of CD8⁺ memory T cells, it is controversial whether Ag clearance is needed for the generation of CD4⁺ memory T cells. Indeed, there is evidence that persistence of the Ag itself is essential for the maintenance of CD4⁺ memory T cells (4, 5, 6). Given that Ag-specific effector T cells are thought to be terminally differentiated and short-lived cells, there must be cellular mechanisms by which memory T cells specific for persistent Ags are maintained in the host as “memory-stem cells.” In support of the idea that persistent Ag helps maintain CD4⁺ memory T cells, rats immunized with irradiated sporozoites are protected from malaria infection, but this protection is lost after drug treatment to remove the remaining parasites (5). A similar phenomenon is found in Leishmania major infection in mice where CD4⁺CD25⁺ regulatory T cells and IL-10 are thought to prevent a Th1 immune response from clearing the infection, thus allowing the host to retain CD4⁺ memory T cells specific to the organism (6). Furthermore, more recent data have shown that in the absence of MHC class II signals that are essential for Ag presentation to CD4⁺ T cells, surviving memory T cells are functionally impaired, suggesting that TCR signaling is involved in the maintenance of CD4⁺ memory T cells (7).

All complex metazoans, including humans and mice, are colonized with microbial organisms that comprise an indigenous microflora (8). Although the host evidently benefits from the resident microflora in the gut (9), the presence of commensal bacteria appears to be of crucial importance in the development of human inflammatory bowel disease (IBD)³ and in almost all animal models of IBD (10, 11). For instance, results from most IBD models showing that in germfree (GF) rodents intestinal inflammation is absent indicate that commensal bacteria are indispensable contributors to the pathogenesis of chronic immune-mediated intestinal inflammation.

IBD is caused by excessive and tissue-damaging chronic inflammatory responses, which are thought to be due to inappropriate activation of the immune system in many cases and which commonly take a persistent, disabling course (10, 11). In some patients, disease progresses steadily, whereas in others, it follows a relapsing-remitting course. According to present understanding, the disease is caused and controlled by colitogenic effector and memory CD4⁺ T cells presumably reacting to commensal bacterial Ags. Importantly, however, it is not known whether different effector CD4⁺ T cells are recruited at each relapse or whether sequential memory CD4⁺ T cells are derived from members of the initial attack cohort throughout the entire course of disease. In other words, the nature and regulation of colitogenic effector and memory CD4⁺ T cells in the host-commensal interaction over time and the correlation between colitogenic effector and memory CD4⁺ T cells in chronic colitis remain largely unknown.

Given the possible importance of the microflora in intestinal inflammation, we conducted a series of experiments to test the in vivo effect of commensal bacteria using a GF system for the establishment and maintenance of colitogenic CD4⁺ T cells. We also attempted to assess whether colitogenic CD4⁺ T cells are self-limited effector cells or long-lived memory cells in a host-commensal bacteria mutualism to understand the perpetuation of IBD and to develop a strategy for IBD treatment.

C3H/HeN mice (6–8 wk old) and C57BL/6-Ly5.2 mice were purchased from CLEA Japan. C57BL/6-Ly5.1 and C57BL/6-Ly5.2 RAG-2^−/− mice were obtained from Taconic Laboratory and the Central Experimental Animal Institute. Specific pathogen-free (SPF) and GF breeding colonies of C3H-SCID mice (C3Smnc Prkdc scid/J; The Jackson Laboratory) were maintained at the Animal Facilities (SPF) and the Gnotobiotic Facilities (GF), respectively, of our institute. Sterility in the Gnotobiotic Facilities was tested monthly by culturing of feces and bedding as well as Gram staining. The Institutional Committees on Animal Research of both Tokyo Medical and Dental University and Yakult Central Institute approved the experiments.

The following mAbs other than biotin-conjugated anti-mouse IL-7Rα (A7R34; eBioscience) and anti-CCR7 (EBI-1; eBioscience) were obtained from BD Pharmingen and used for purification of cell populations and flow cytometric analysis: Fcγ (CD16/CD32)-blocking mAb (2.4G2); PE-, PerCP-, and PECy5-conjugated anti-mouse CD4 (RM4-5); FITC-conjugated anti-mouse CD3 (145-2C11); PE- and allophycocyanin-conjugated anti-mouse CD44 (IM7); FITC- and PE-conjugated anti-mouse L-selectin (CD62L) (MEL-14); FITC-conjugated anti-mouse CD69 (H1.2F3); PE-conjugated anti-mouse α₄β₇ (DATK32); FITC-conjugated anti-mouse CD45RB (16A); PE-conjugated anti-mouse Ly5.1 (CD45.1); PE-conjugated streptavidin; biotin-conjugated rat IgG2; PE-conjugated mouse IgG; and PE-conjugated rat IgG.

To assess the role of commensal bacteria in the persistence of T cell-mediated chronic colitis, CD4⁺ T cells (5 × 10⁵ cells/mouse) isolated from spleen (SP) of colitic mice induced by an adoptive transfer of CD4⁺CD45RB^high T cells into C3H-SCID mice under SPF conditions were injected into new GF (n = 16) or SPF (n = 8) C3H-SCID mice. Six weeks after the transfer, one group of GF SCID recipients was maintained in GF conditions (GF→GF group, n = 8), and the other was moved into SPF conditions (GF→SPF group, n = 8). All groups (SPF, GF→GF, and GF→SPF) were kept for an additional 4 wk and sacrificed 10 wk after cell transfer. They were observed for clinical signs such as hunched posture, piloerection, diarrhea, and blood in the stool. After sacrifice, mice were given a clinical score defined as the sum of four parameters: hunching and wasting, 0 or 1; colon thickening, 0–3 (0, no colon thickening; 1, mild thickening; 2, moderate thickening; 3, extensive thickening); and stool consistency, 0–3 (0, normal beaded stool; 1, soft stool; 2, diarrhea; 3, bloody stool) (12). To assess the histological scores, two parts of the colon were evaluated. The proximal section was defined as 1 cm anal to the cecum, and the distal section was defined as 1 cm oral from the anus. Both tissue samples were cut 5 mm long and fixed in PBS containing 10% neutral-buffered formalin. Paraffin-embedded sections (5 μm) were stained with H&E. The sections were analyzed without prior knowledge of the type of T cell reconstitution or recipient. The mean degree of inflammation in the colon was calculated using a modification of a previously described scoring system as the sum of three parameters: crypt elongation, 0–3; mononuclear cell infiltration, 0–3; and frequency of crypt abscesses, 0–3 (13).

CD4⁺ T cells were isolated from lamina propria (LP) of colitic RAG2^−/− mice previously injected with Ly5.1⁺ CD4⁺CD45RB^high T cells and were labeled with CFSE (Invitrogen) at a concentration of 5 μM according to the manufacturer’s instructions; then, CFSE-labeled LP CD4⁺ T cells (1 × 10⁷/mouse) were injected into new Ly5.2⁺ wild-type (WT) C57BL/6J mice.

C57BL/6J mice in the antibiotic-treated group were given drinking water, including ampicillin (1 g/L), vancomycin (500 mg/L), neomycin sulfate (1 g/L), and metronidazole (1 g/L), 2 wk before beginning the adoptive transfer and during the course of the experiment. Control mice received drinking water without antibiotics.

Anti-IL-7Rα mAb (A7R34) was described previously (14). GF SCID mice were treated with rat anti-murine IL-7Rα mAb by i.p. injection of a 1-mg dose once per week for 2 wk (0, 1, and 2 wk after transfer). Control mice were treated with the same amounts of rat control IgG (Sigma-Aldrich). All mice were killed on the day after the last treatment.

To detect the surface expression of a variety of molecules, isolated bone marrow (BM), SP, mesenteric lymph nodes (MLN), or colonic LP mononuclear cells were preincubated with an FcR-blocking mAb (CD16/32 and 2.4G2; BD Pharmingen) for 20 min before incubation with specific FITC-, PE-, PerCP-, and APC-labeled Abs for 30 min on ice. When biotin-conjugated Abs were used, cells were incubated with Abs for 30 min, then with PE-labeled streptavidin for 30 min on ice. Standard four-color flow cytometric analyses were performed using a FACSCalibur (BD Biosciences) and analyzed by CellQuest software (BD Biosciences). Background fluorescence was assessed by staining with irrelevant isotype-matched mAbs.

To measure cytokine production, 1 × 10⁵ CD4⁺ T cells from LP were cultured in 200 μl of culture medium at 37°C in a humidified atmosphere containing 5% CO₂ in 96-well plates (Costar) precoated with 5 μg/ml hamster anti-mouse CD3ε mAb (145-2C11; BD Pharmingen) and 2 μg/ml hamster anti-mouse CD28 mAb (37.51; BD Pharmingen) in PBS overnight at 4°C. Culture supernatants were removed after 48 h and assayed for cytokine production. Cytokine concentrations were determined by specific ELISA as per the manufacturer’s recommendation (R&D Systems).

Consecutive cryostat BM and colon sections were used in all studies. Immunohistochemistry was performed using purified mAb against mouse CD4 (RM4-5; BD Pharmingen) or biotin-conjugated polyclonal IL-7 Ab (BAF407; R&D Systems). In brief, fresh frozen tissue samples were cut into serial sections 6 μm thick, placed on coated slides, and fixed with 4% paraformaldehyde phosphate buffer solution for 10 min. Slides were incubated with the primary Ab at 4°C overnight, then stained for 60 min at room temperature with AlexaFluor 488 goat anti-rat IgG for CD4 detection or AlexaFluor 488 streptavidin (Molecular Probes) for IL-7 detection. All slides were counterstained with 4,6-diamidino-2-phenylindole (Vector Laboratories) and observed under a BioZERO BZ8000 microscope (Keyence).

The results were expressed as the mean ± SEM. Groups of data were compared by the Mann-Whitney U test. Differences were considered to be statistically significant when p < 0.05.

In research into antiviral CD8⁺ memory T cells, memory T cells are classically defined as cells that persist in the organism once the viral Ag has been cleared and that mediate a much quicker and stronger response when the Ag is met again (1, 2, 3). However, current evidence suggests that maintenance of a memory CD4⁺ T cell population responding to chronic infection is dependent on the persistent presence of Ag (4, 5, 6), and it is controversial whether CD4⁺ memory T cells can be maintained in the absence of Ag. In particular, to understand the persistence of IBD, it is important to know whether the colitogenic CD4⁺ T cells are self-limited effector cells or are long-lived memory cells within the host-commensal bacteria mutualism. We attempted to clarify this issue using a well-known chronic colitis model induced by adoptive transfer of CD4⁺CD45RB^high T cells into SCID mice, which characteristically involves the differential activation of Th1/Th17 cells (15, 16). We first checked the phenotype of CD4⁺ T cells in colonic LP, MLN, SP, and BM of colitic C3H-SCID mice that had previously been injected with syngeneic CD4⁺CD45RB^high T cells. Consistent with our previous reports using BALB/c/C.B.17 or C57BL/6J strains (12, 13), LP CD4⁺ T cells, as well as colitic SP, MLN, and BM CD4⁺ T cells, are exclusively CD44^highCD62L⁻CCR7⁻IL-7Rα^high T_EM cells (Fig. 1). CD69 was also expressed by a high proportion of the CD4⁺ T cells in various site of colitic mice (Fig. 1). Consistent with this, we have previously shown that in sharp contrast to IL-7^+/+ × RAG-1^−/− recipients, IL-7^−/− × RAG-1^−/− recipients given these colitic CD4⁺ T cells do not develop colitis (13), indicating that these CD4⁺ T cells could be categorized as T_EM cells.

Although it is well established that the existence of commensal bacteria is required for the initial development of most animal models of chronic colitis (10, 11, 17), it is still unclear whether commensal bacteria are also needed for the persistence of colitis and/or the maintenance of colitogenic CD4⁺ T cells after the T cell-priming process. To this end, we conducted an adoptive retransfer experiment under SPF or GF conditions (Fig. 2,A). We used SP not LP CD4⁺ T cells isolated from colitic SCID mice as donor cells for two reasons: 1) we wanted to exclude possible contamination by commensal bacteria from colitic LP or MLN samples, and 2) we have previously shown that most of the SP CD4⁺ T cells in this mouse colitis model express the T_EM cell phenotype with a capacity to induce colitis similar to that of LP CD4⁺ T cells (12). Thus, we injected these SP cells into new SPF or GF C3H-SCID mice (hereafter called SPF or GF SCID mice, respectively) in a retransfer. Furthermore, to assess whether GF SCID recipients develop colitis when exposed to resident commensal bacteria, one group of the GF SCID recipients was moved into the SPF environment 6 wk after the retransfer and kept for an additional 4 wk (GF→SPF SCID mice), and the other group was kept in GF conditions for 10 wk (GF→GF SCID mice) (Fig. 2,A). Consistent with previous reports (12, 13) and in sharp contrast to GF SCID mice, SPF SCID mice showed an enlarged colon with a greatly thickened wall (Fig. 2,B) and manifested progressive weight loss from 3 wk after retransfer (data not shown). Consistent with this, the clinical score of SPF SCID mice at 10 wk after retransfer was significantly increased compared with that of GF SCID mice, which showed no clinical signs of colitis or weight loss throughout the observation period (Fig. 2,C). It was notable that GF→SPF SCID recipients also showed an enlarged and thickened colon, and their clinical scores were significantly increased compared with those of GF→GF SCID recipients, being comparable to those of SPF SCID recipients. As previously reported (18), a markedly enlarged cecum was also reproducibly observed in the GF→GF SCID-recipient mice (Fig. 2 B). The integrity of the GF conditions was confirmed by repeated stool culture tests in the GF groups (data not shown). SP and colon weights of SPF SCID mice and GF→SPF SCID mice were significantly greater than those of GF→GF SCID mice (data not shown).

Histological examination showed a prominent epithelial hyperplasia with glandular elongation and massive infiltration of mononuclear cells in the LP of proximal and distal colon from SPF SCID and GF→SPF SCID mice (Fig. 3,A). In contrast, the glandular elongation was mostly abrogated, and fewer mononuclear cells were observed in the LP of the colon from the recipient GF→GF SCID mice (Fig. 3,A). This difference was also confirmed by histological scoring of multiple proximal and distal colon sections: 3.80 ± 0.37 and 4.40 ± 0.43, respectively, in SPF SCID mice; 4.00 ± 0.30 and 5.17 ± 0.056 in GF→SPF SCID mice; and 0.125 ± 0.001 and 0.125 ± 0.001 in GF→GF SCID mice (SPF SCID mice or GF→SPF SCID mice vs GF→GF SCID mice; p < 0.01) (Fig. 3 B).

A further quantitative evaluation of CD4⁺ T cell infiltration was made by isolating LP, SP, and BM CD3⁺CD4⁺ T cells. As shown in Fig. 4, the number of CD4⁺ T cells recovered from LP and SP of SPF and GF→SPF SCID mice was significantly higher than that from GF→GF SCID mice. Surprisingly, however, a substantial number of CD4⁺ T cells remained resident in BM and SP in GF→GF SCID mice, and the number of CD4⁺ T cells recovered from BM of colitic SPF and GF→SPF SCID mice was not significantly higher that that recovered from noncolitic GF→GF SCID mice. This suggests that even in the absence of commensal bacteria, the colitogenic SP CD4⁺ T donor cells are not all short-lived effector cells but, in accordance with high expression of IL-7Rα, are long-lived T_EM cells or a mixture of effector cells and T_EM cells. Alternatively, it is possible that most persisting CD4⁺ T cells in GF SCID mice respond to environmental Ags derived from food and bedding.

We next examined the cytokine production by isolated LP CD4⁺ T cells from each group. As shown in Fig. 5, LP CD4⁺ T cells from the GF SCID recipients produced significantly smaller amounts of IFN-γ, TNF-α, and IL-17 than those from the SPF SCID recipients after in vitro stimulation by anti-CD3/CD28 mAbs.

Flow cytometry analysis revealed that the cell-surface phenotype of LP, SP, and BM CD4⁺ T cells isolated from SPF or GF→SPF SCID recipients was very similar to that from the primarily injected colitic mice, that is, T_EM cell type with CD44^highCD62L⁻CD69⁺IL-7Rα^high cells (Figs. 1 and 6). Again, the expression level of CD69 on SP CD4⁺ T cells was significantly lower than that on LP and BM cells in the SPF or GF→SPF SCID recipients (percentage of positive cells in SPF: SP 18.9 ± 3.40, LP 60.4 ± 1.79, and BM 61.8 ± 3.37; SP vs LP or BM, p < 0.05; in GF→SPF, SP 9.83 ± 0.95, LP 55.8 ± 4.21, and BM 53.8 ± 6.28; SP vs LP or BM p < 0.05), suggesting the presence of activation mechanisms in LP and BM. Importantly, the expression pattern of all examined markers in all sites was not significantly different between SPF and GF→SPF SCID recipients (Fig. 6, right panels).

In contrast, the expression pattern on cells from GF→GF SCID recipients was quite different. First, surprisingly, IL-7Rα expression on LP but not SP and BM CD4⁺ T cells was significantly down-modulated compared with the paired LP cells of SPF or GF→SPF SCID recipients (Fig. 6,A). Second, and as expected, CD69 expression on cells from all sites was also reduced compared with the matching cell source from SPF or GF→SPF SCID recipients (Fig. 6,B). Third, and quite interestingly, the proportion of CD3⁺CD4⁺CD44^highCD62L⁺ T_CM cells was significantly increased in SP and BM but not in LP of GF→GF SCID recipients (SP 2.41 ± 0.85%; BM 5.21 ± 0.95%; LP 3.09 ± 0.95%) compared with SPF SCID recipients (SP 1.14 ± 0.1%, p < 0.05; BM 2.80 ± 0.48%, p < 0.05; LP 0.35 ± 0.059%, p = 0.10) and GF→SPF SCID recipients (SP 1.31 ± 0.17%, p < 0.05; BM 2.23 ± 0.33%, p < 0.05; LP 0.18 ± 0.040%, p = 0.28), suggesting the conversion of cells from T_CM to T_EM after movement of mice from GF→SPF (Fig. 6 C).

The TCR Vβ repertoires of SP CD4⁺ T cells from SPF, GF→GF, and GF→SPF SCID mice differed only slightly and only in such Vβ families as Vβ7, Vβ9, and Vβ11, which are not major groups in this model (Fig. 7). This suggested that long-lived colitogenic memory CD4⁺ T cells are broadly polyclonal rather than specifically oligoclonal. Furthermore, this finding suggests that in GF conditions, colitogenic memory CD4⁺ T cells may be maintained by homeostatic cytokines that would lead to polyclonal proliferation, rather than by response to a specific Ag such as food or bedding that may lead to a specific oligoclonal expansion.

We have previously reported that colitogenic memory CD4⁺ T cells are retained in an IL-7-dependent manner in BM of SPF-conditioned CD4⁺CD45RB^high T cell-injected colitic mice (19). Thus, we hypothesized that BM IL-7 may maintain colitogenic memory CD4⁺ T cells in noncolitic GF→GF SCID mice even in the absence of commensal bacteria. To test this hypothesis, frozen sections of colon and BM from each group were stained with polyclonal anti-IL-7 Ab (green) and anti-CD4 mAb (red). First, Fig. 8 clearly demonstrates marked infiltration of CD4⁺ T cells in the colon of colitic SPF and GF→SPF SCID recipients, although the expression of IL-7 in epithelial cells was markedly decreased in these two groups (Fig. 8, left top and bottom). In sharp contrast, only a scattering of CD4⁺ T cells were found in the LP of noncolitic GF→GF SCID recipients despite the presence of epithelial IL-7 (Fig. 8, left middle). However, consistent with the constant expression of IL-7 in BM, substantial numbers of CD4⁺ T cells were resident in BM of all three groups, regardless of the presence or absence of commensal bacteria (noncolitic or colitic) (Fig. 8, right panel).

To confirm the importance of IL-7 for the maintenance of colitogenic memory CD4⁺ T cells even in GF conditions, we next blockaded IL-7. GF SCID mice injected with colitogenic SP CD4⁺ T cells were separated in two groups: one group was treated with anti-IL-7Rα mAb, and the other group was treated with control rat IgG (Fig. 9,A). Two weeks after transfer, all mice were sacrificed, and LP, MLN, SP, and BM cells were analyzed by flow cytometry to determine the number of CD3⁺CD4⁺ T cells recovered (Fig. 9,B). As expected, the numbers of CD4⁺ T cells recovered from sites other than LP of GF SCID mice treated with anti-IL-7Rα mAb were significantly reduced compared with the control IgG-treated group (Fig. 9 B). However, although the number of CD4⁺ T cells recovered from LP of mice treated with anti-IL-7Rα mAb tended to be lower than that from LP of mice treated with control IgG, the difference was not significant, suggesting the possibility of additional stimuli, such as food and/or bedding Ags and pathogen-associated molecular patterns included in sterile foods, maintaining CD4⁺ T cells in LP. Collectively, this result supports our hypothesis that IL-7 may maintain colitogenic memory CD4⁺ T cells outside the intestine of the injected GF SCID mice even in the absence of commensal bacteria.

Finally, we asked whether colitogenic memory CD4⁺ T cells can be retained in lymphocyte-sufficient normal mice, because it is very important to know whether our findings are applicable to the pathogenesis of human IBD where the patients are always immunosufficient. To this end, we injected CFSE-labeled colitogenic LP Ly5.1⁺CD4⁺ T cells into Ly5.2⁺ WT mice treated with either a mixture of antibiotics in distilled water or distilled water alone (Fig. 10,A). One week and 4 wk after transfer, cell numbers were recovered from LP, SP, and BM from injected WT mice, and their numbers of cell divisions were assessed by a flow cytometry (Fig. 10, B and C). As shown in Fig. 10 B, a substantial number of Ly5.1⁺CD4⁺ T cells were recovered from LP, SP, and BM of lymphocyte-replete mice at 1 and 4 wk after cell transfer, regardless of antibiotic treatment. Notably, there is almost no difference between antibiotic-treated and control mice in the number of CD4⁺ T cells in various sites. Consistent with this, a CFSE dilution assay revealed that Ly5.1⁺CD4⁺ T cells divided well in LP, SP, and BM at the indicated time points regardless of antibiotic treatment. Collectively, these results suggest that in the noncolitic immunosufficient condition, the maintenance of the colitogenic memory CD4⁺ T cells is entirely independent of the presence of commensal bacterial Ag but is presumably dependent on homeostatic cytokines such as IL-7.

The present study has demonstrated that latent long-lived colitogenic memory CD4⁺ T cells not only reside outside the intestine (for example in the IL-7-sufficient BM environment of noncolitic mice that had been previously injected with colitogenic CD4⁺ T cells under GF conditions) but also participate in the recurrence of colitis in the presence of commensal bacteria. In other words, these data clearly showed that colitogenic CD4⁺ memory T cells are established during the process of development of colitis even in the presence of commensal bacteria and can be maintained outside the intestine, possibly in an IL-7-dependent manner, in the remission stage of chronic colitis. Thus, our present results may explain why IBD is an intractable lifelong disease if it depends on the persistence of latent colitogenic CD4⁺ memory T cells.

Immunologically, “memory” is now generally believed to be a phenomenon that occurs after Ags have been eliminated (1, 2, 3). More specifically, Ags must be eliminated from the body in order for CD8⁺ memory T cells to form. For example, in the lymphocytic choriomeningitis virus model of acute viral infection, lymphocytes differentiate into CD8⁺ memory T cells for the first time after the virus has been eliminated (20, 21). In contrast, in a chronic viral infection model using a mutant lymphocytic choriomeningitis virus, it is thought that no CD8⁺ memory T cells are produced and that the CD8⁺ effector T cells are ultimately destroyed as a result of “exhaustion” (22, 23). The fact that the Ags responsible for IBD are derived from commensal bacteria that can never be eliminated suggests the possibility that effector T cells continue to emerge and become exhausted without memory T cells ever being produced. However, a requirement for constant production of colitogenic CD4⁺ effector cells from naive CD4⁺ T cells that are continuously supplied by the thymus may be difficult to explain in view of the involution of the thymus in both mice and humans. We therefore hypothesized that even though commensal bacteria may permanently reside in the intestine, colitogenic CD4⁺ T cells could be retained as memory cells. Consistent with this hypothesis, we previously demonstrated that IL-7^−/− × RAG-1^−/− mice injected with colitogenic CD4⁺ T cells never develop colitis, whereas evidence of disease was seen in IL-7^+/+ × RAG-1^−/− recipients (13). These findings demonstrated that IL-7 is essential for the development and maintenance of chronic colitis. Accordingly, it is now assumed that IL-7-dependent memory T cells make up at least part of the population of colitogenic CD4⁺ T cells and contribute to the overall maintenance of colitis even in the presence of enteric bacteria.

In contrast, our group has previously reported that IL-7 is produced by epithelial cells in the intestine, especially by intestinal goblet cells (24), and it is known that the pathology of IBD is characterized by a decrease in goblet cells when it becomes chronic (25). There was also a marked decrease in strongly staining Alcian blue-positive goblet cells at the sites of the chronic colitis lesions in the CD4⁺CD45RB^high T cell transfer model that we used, and we discovered that there was a concomitant decrease in IL-7 production by the epithelial cells (Ref. 26 ; Fig. 8). Why then does IL-7 derived from inflamed intestine decrease when IL-7 is essential to the development of chronic colitis? We hypothesized that intestinal IL-7 is not required to maintain chronic colitis and that colitogenic CD4⁺ memory T cells are maintained by IL-7 outside the intestine. We previously performed parabiosis surgery that connected the flanks of colitic RAG-2^−/− mice into which CD4⁺CD45RB^high T cells had been injected with the flanks of untreated IL-7^+/+ × RAG-1^−/− mice or IL-7^−/− × RAG-1^−/− mice. We allowed the hemodynamics of the two mice to be shared for several days after the parabiosis, and even though the intestinal epithelial cells of the IL-7^−/− × RAG-1^−/− host mice do not produce IL-7, they can be viewed as mice that produce IL-7 outside the intestine because of the shared hemodynamics. As expected, ∼4 wk after the parabiosis, the transfer of colitogenic CD4⁺ T cells from the colitic RAG-2^−/− donor mice was associated with the development of chronic colitis in the IL-7^+/+ × RAG-1^−/− host mice, which was accompanied by marked CD4⁺ T cell infiltration of the large intestine. Surprisingly, despite the deficiency of intestinal IL-7, frank colitis also developed in the IL-7^−/− × RAG-1^−/− host mice in the same way as in the control IL-7^+/+ × RAG-1^−/− host mice (26). These previous findings demonstrated that intestinal IL-7 is not essential for the maintenance of colitogenic CD4 memory T cells in mice with chronic colitis.

In this study, we further attempted to resolve the following two issues. 1) How are colitogenic CD4⁺ T cells maintained in inflamed mucosa despite the lack of epithelium-derived IL-7? 2) Can colitogenic CD4⁺ T cells be maintained for a long period in the absence of commensal bacteria in IL-7-sufficient conditions, as is the case for memory CD8⁺ T cells? If so, are latent colitogenic CD4⁺ memory T cells involved in the recurrence of colitis after moving from a GF to an SPF environment?

First, we showed that almost all CD4⁺ T cells in LP of noncolitic GF→GF SCID mice disappeared despite the presence of epithelium-derived IL-7 (Fig. 8). In contrast, a substantial number of CD4⁺ T cells still resided outside the intestine of GF→GF SCID mice, showing that the number of CD4⁺ T cells recovered from these noncolitic mice was comparable with that from colitic SPF or GF→SPF SCID mice, especially in BM (Fig. 4). This result suggests that colitogenic CD4⁺ T cells may be reciprocally regulated in intestine and in the BM so that colitogenic LP and BM CD4⁺ T cells are stimulated via commensal bacterial Ags and via IL-7 signaling, respectively. Consistent with this idea, we previously showed that: 1) synchronous stimulation of human LP CD4⁺ T cells by anti-CD3 mAb and IL-7 induces apoptosis of those cells (16), and 2) colitogenic CD4⁺ T cells actively circulate in colitic mice and the blockade of the circulation by FYT720 treatment ameliorates the colitis in this model (27). This suggests that epithelial IL-7 plays a physiologically important role in intestinal immune homeostasis that prevents the triggering of IBD, whereas the absence of epithelium-derived IL-7 itself may be needed for the persistence of chronic colitis. Further study using IL-7-deficient recipient mice under GF conditions will be needed to test this hypothesis.

Second, and surprisingly, we showed that colitogenic SP CD4⁺ T cells survived for 6 wk in GF conditions and are involved in the recurrence of colitis after their movement to SPF conditions. Because effector T cells are believed to be short lived in contrast to long-lived memory T cells, this result indicates that the colitogenic SP CD4⁺ T cells contained at least some fraction of colitogenic memory CD4⁺ T cells. As an additional explanation for the perpetuation of IBD, the ratio of T_CM/T_EM cells in GF→GF SCID mice may support the idea that colitogenic “memory stem”-like T_CM cells are generated in the process of development and/or persistence of chronic colitis.

The results of our current project show some differences from the findings published by Veltkamp et al. (17). Using their original model of chronic colitis induced by transplantation of BM from normal mice into CD3ε transgenic (BM→Tgε26) mice, they showed that BM→Tgε26 mice do not develop colitis in GF conditions whereas control BM→Tgε26 in SPF conditions do. However, they also showed that viable intestinal bacteria are not necessary for the survival of CD4⁺ T cells that retain their functional integrity in noncolitic BM→Tgε26 mice in GF conditions and induce colitis when mice are moved to SPF conditions and that Tgε26 mice develop colitis in SPF conditions when MLN cells from BM→Tgε26 mice maintained in GF conditions are injected. Thus, they concluded that continuous stimulation by commensal bacteria is essential for the development of colitis. Although the results of their study seem to be similar to ours, the interpretation of each study is distinct. First, it is very likely that peripheral MLN cells isolated from noncolitic BM→Tgε26 mice in GF conditions contain a substantial number of naive CD4⁺ T cells, and thus it is very possible that lymphopenic Tgε26 mice injected with these cells develop colitis when transferred to SPF conditions. In contrast, we first isolated colitogenic SP CD4⁺ memory T cells from colitic mice, and as shown in Fig. 1, these cells all have the phenotype of memory CD4⁺CD44^high T cells. Also, it is very important that naive CD4⁺ T cells are continuously generated in their BM transplantation system but not in our adoptive transfer system using T cell-deficient SCID mice that lack the ability to generate naive CD4⁺ T cells continuously from the thymus.

Finally, it should be noted that the expression of IL-7Rα was significantly reduced on CD4⁺ T cells from all sites of noncolitic GF→GF SCID mice compared with the paired sites in colitic SPF or GF→SPF SCID mice. This was very surprising, being inconsistent with the recent immunological dogma that memory CD4⁺ T cells express high levels of IL-7Rα. Although it makes sense that the high expression of IL-7Rα on CD4⁺ T cells of noncolitic GF→GF SCID mice would be maintained because the significant decrease in expression of the activation marker CD69 on those memory T cells suggests that they are in the resting state, we also obtained the same result using antibiotic-treated mice that had previously been injected with CD4⁺CD45RB^high T cells. Although further study will be needed to address this issue, colitogenic memory CD4⁺ T cells may be quite different from conventional memory CD4⁺ T cells.

Taken together, these findings suggest that long-lived colitogenic memory CD4⁺ cells can be established as CD4⁺ T_EM or T_CM cells during the process of development of colitis, even in the presence of commensal Ags, and that these cells participate in the perpetuation of colitis. Thus, blocking a stimulus of colitogenic memory CD4⁺ cells such as IL-7, or alternatively an immunological reset by BM transplantation to remove these colitogenic memory CD4⁺ T cells (28), may have therapeutic benefit for treatment of IBD.

The authors have no financial conflict of interest.