Regulatory functions of CD8+CD28– T cells in an autoimmune disease model (original) (raw)
Role of CD8+ T cells in EAE. We investigated the role of CD8+ lymphocytes in the MOG-induced chronic EAE model with particular emphasis on CD28–/– mice that are naturally resistant to EAE (1–4). CD8+ lymphocytes constitute 9–14% of splenocytes in naive WT mice of which 60% are CD8+CD28+, while the remaining 40% are CD8+CD28– cells. Thus, overall, CD8+CD28+ cells make up 5.5–8.5% and CD8+CD28– cells 3.5–5.5% of all splenocytes (one representative example is shown in Figure 1). In WT mice, depletion of CD8+ T cells (less than 1% CD8+ T cells in peripheral blood by flow cytometry) prior to disease induction worsens clinical disease significantly as compared with non–CD8-depleted controls (mean maximal grade 3.8 ± 0.9 versus 2.8 ± 1.4, P < 0.03). The clinical disease observed was similar to that observed in CD8–/– mice (Table 1, Figure 2a). The mortality of animals lacking CD8+ T cells was higher than that of WT animals (Table 1), although it did not reach statistical significance.
Effect of lack of CD8+ T cells on MOG p35–55–induced EAE in WT C57BL/6 and CD28–/– mice. (a) Mice were immunized with MOG p35–55 and graded for disease daily. The mean daily grade for each group is shown. This is a representative experiment showing the disease course in C57BL/6 mice treated with rat IgG (filled squares), WT mice treated with anti-CD8 mAb (open circles), and CD8–/– mice (open squares). (b) A representative experiment showing disease induction in C57BL/6 WT mice (filled squares), CD28–/– mice treated with control rat IgG (filled triangles), and CD28–/– mice treated by anti-CD8 mAb before immunization (open circles). The mean daily grade for each group (n = 5–7) is shown.
Effects of absence of CD8 T lymphocytes on actively induced EAE in wild-type and CD28–/– mice
CD28–/– mice develop EAE after CD8+ T cell depletion. We and others have previously reported that unlike WT C57BL/6 mice, CD28–/– mice immunized with MOG 35-55 peptide are resistant to EAE (1, 3, 4). Administration of a depleting anti-CD8 mAb before immunization brings out clinical disease in CD28–/– animals (Table 1, Figure 2b), characterized immunohistologically by intraparenchymal cellular infiltrates similar to those observed in WT mice (data not shown), whereas in nondepleted CD28–/– mice the infiltrates were restricted to the meninges and within blood vessels, similar to our previous report (4).
In CD8-depleted animals, 12 of 13 mice developed disease compared with only 1 in 15 in the non-depleted CD28–/– animals. The disease onset in CD8-depleted CD28–/–mice was delayed and significantly less severe, however, compared with WT mice (Table 1). To further investigate the role of CD8+ T cells during the effector phase of EAE, we administered the depleting anti-CD8 mAb 2 weeks after immunization. The majority of these animals (8 of 11) also developed mild EAE (Table 1), with the onset being about 11 days after depletion. These data indicate a critical role for CD8+CD28– T cells in disease resistance in CD28–/– mice during the priming and effector phases of the immune response in vivo.
CD28–/– CD8–/– mice develop active EAE. To confirm the role of CD8+CD28– T cells in natural disease resistance in CD28–/– mice, we generated a new mouse colony that is deficient at both CD28 and CD8 loci by intercrossing CD28–/– with CD8–/– mice (see Methods). All the homozygous double-knockout mice developed EAE when immunized with MOG peptide (Table 1, Figure 3a). Interestingly, these mice developed two different forms of clinical EAE: approximately 50% of animals developed a mild form of “typical EAE” with the classic ascending flaccid paralysis and a significantly later disease onset as compared with WT controls. The remaining animals developed an “atypical” form of EAE, with the onset of symptoms not significantly altered compared with WT controls (Table 1). The clinical symptoms of those with atypical disease were primarily associated with ataxia, spastic reflexes, loss of coordinated movements, spinning, and head tilt. The disease led to premature death in all animals within 24 hours of onset of clinical symptoms.
Induction of EAE in CD28–/–CD8–/– and CD28–/–CD8+/– mice. (a) Complete lack or decreased expression of CD8 molecule leads to EAE. The mean daily grade for each group (n = 8–15 mice) is on the y axis. All CD28–/–CD8–/– mice immunized with MOG developed EAE (open circles): 7 of 15 animals developed a mild form of typical EAE, while 8 of 15 animals developed an atypical form of EAE that led to death in these animals within 24 hours. CD28–/–CD8+/– mice developed mild typical EAE that was half-way between that in CD28–/– (filled triangles) and CD28–/–CD8–/– mice (open circles), indicating a dose-response effect of CD8 expression (open squares). (b) Comparison of CD8 surface expression on peripheral blood lymphocytes derived from CD28–/– mice (solid histogram) and CD28–/–CD8+/– mice (line histogram).
CD28-deficient mice heterozygous for the CD8 locus demonstrate decreased CD8 surface density compared with CD28–/–CD8+/+ mice by flow cytometry (Figure 3b). We asked whether a decrease in CD8 expression would make CD28–/– animals susceptible to disease. Figure 3a shows that these mice develop mild typical EAE with a late onset of disease as compared with WT mice (Table 1). The disease course was halfway between that in CD28–/– and CD28–/–CD8–/– mice, indicating a dose-response effect of CD8 expression. It is interesting that decreased expression of surface CD8 in the heterozygous mice is linked to partial loss of regulatory function in vivo and suggests that surface CD8 is involved in the mechanism of CD8+CD28– T cell regulation.
To confirm the regulatory function of CD8+CD28– T cells in vivo, we developed an adoptive transfer system into CD8–/– animals that naturally develop EAE, which is more severe than WT animals. First, one million 100% purified CD8+CD28– T cells isolated from naive CD28–/– spleens were transferred into each CD8–/– recipient via tail vein, and animals were then immediately immunized with MOG peptide. Adoptive transfer of CD8+CD28– T cells led to a significant decrease in disease severity as compared with untreated CD8–/– mice (P = 0.02, n = 10) (Figure 4a). Interestingly, the disease severity was comparable to that in control WT animals (P = NS, n = 10). The onset of disease was significantly delayed as compared with CD8–/– mice (16.1 ± 4.25 versus 10.7 ± 1.6, P = 0.003). There was no significant difference in the disease incidence and the mortality between the two groups (incidence in CD8–/–, 10 of 10 versus 9 of 10 in adoptive transfer group; mortality in CD8–/– animals, 7 of 10 versus 3 of 10 in the transfer group, P = NS). Interestingly, adoptive transfer of 5 million purified CD8+CD28– cells did not result in further protection from disease (Figure 4b). Similar experiments with adoptive transfer of 0.5 × 106 purified CD8+CD28– T cells did not lead to significant suppression of EAE (data not shown), demonstrating the need for transfer of a minimum of one million cells. Next, the same experiment was repeated using one million CD8+CD28– and CD8+CD28+ T cells purified from naive WT mice. CD8+CD28+ cells did not modulate the course of EAE as compared with control animals (P = 0.14, n = 6) while CD8+CD28– T cells led to a significant decrease in disease severity (P = 0.0019, n = 6) (Figure 4b). These data confirm that CD8+CD28– T cells represent a distinct subset of CD8+ T cells with regulatory function in vivo in EAE.
Suppression of EAE by adoptive transfer of CD8+CD28– cells into CD8–/– mice. (a) 100% purified CD8+CD28– cells were generated from naive CD28–/– splenocytes and injected into CD8–/– recipients via the tail vein, as described in Methods. The recipient mice were then immunized with MOG peptide on the same day. The mean daily score for each group (n = 10) is shown on the y axis. The course of CD8–/– mice (open squares), CD8–/– recipients of CD8+CD28– cells (open circles), and WT mice (filled squares) is shown. (b) One hundred percent purified CD8+CD28+ and CD8+CD28– T cells were isolated from spleens of naive WT mice. Approximately one million cells were then injected into each CD8–/– recipient as described above. The mean daily score for each group (n = 6) is shown on the y axis. CD8+CD28– T cells (open circles) significantly suppress the EAE as compared with the control group (open squares), while CD8+CD28+ T cells do not show any significant effect on disease course (filled circles). Adoptive transfer of approximately five million CD8+CD28– T cells did not lead to any further suppression of disease (filled squares).
Finally, histological analysis of CD8–/–CD28–/– mice with typical EAE revealed the presence of inflammatory infiltrates and demyelination in the lumbar spinal cord starting at day 9 (Figure 5c), with no infiltrates or demyelination in the brainstem (Figure 5, a and b) or cervico-thoracic spinal cord (not shown). In contrast, CD28–/–CD8–/– mice with atypical EAE displayed inflammatory infiltrates in the brainstem meninges and parenchyma, including the anterior pons depicted in Figure 5d as well as demyelination in the anterior pons (Figure 5e). In some mice from this group, inflammation and demyelination was seen in the lateral pons and vestibular nerve also. In addition, CD28–/–CD8–/– mice with atypical EAE had large inflammatory infiltrates in the cervical and thoracic spinal cords with demyelination (Figure 5f), but little to no infiltrate in the lumbar spinal cord.
H&E and LFB stains of sections of the brainstem and spinal cords from CD28–/–CD8–/– mice with typical EAE (a–c) and CD28–/–CD8–/– mice with atypical EAE (d–f). (a) H&E-stained section from the anterior pons showing no inflammatory infiltrates. (b) LFB staining in the pons showing no demyelination. (c) LFB staining showing small infiltrate and demyelination in the lateral lumbar spinal cord (arrow). (d) H&E-stained section from pons showing inflammatory infiltrate in the anterior pons (arrow). (e) LFB staining showing demyelination in the anterior pons (arrow). (f) LFB staining showing demyelination and large inflammatory infiltrates in the lateral thoracic spinal cord (arrow). Photomicrographs a–f are taken at ×400 magnification. Enlargements are all at ×1,000 magnification.
Expansion of MOG-specific IFN-γ–producing CD4+ T cells in the absence of CD8+ T cells. We measured the frequency of MOG-specific Th1-producing (IFN-γ) and Th2-producing (IL-4, IL-5 and IL-10) cells on day 14 after immunization by ELISPOT in CD28–/–CD8–/– mice, CD28–/–, and CD28–/–CD8-depleted (by mAb therapy) mice. As shown in Figure 6a, the frequency of antigen-specific Th1 cells expanded significantly in both CD28–/– mice depleted of CD8+ T cells as well as CD28–/–CD8–/– mice, as compared with CD28–/– control mice (P < 0.02 for all antigen concentrations). In contrast, the frequency of Th2-producing cells was not significantly altered in any of the studied groups (IL-4 spots of 56.5 ± 4.9 for CD28–/– versus 46.5 ± 4.9 for anti-CD8–treated group versus 53 ± 11.3 in CD28–/–CD8–/– the mouse). Similar results were found in the WT CD8-depleted and CD8–/– mice compared with WT controls (Figure 6b). These results indicate that CD8+CD28– T cells play a role in limiting expansion of MOG-specific Th1 cells after immunization, whereas Th2 clone size is unaltered.
IFN-γ production of splenocytes in response to MOG peptide in vitro. MOG p35-55–specific IFN-γ–producing cells were measured by ELISPOT in cultures of splenocytes harvested on day 14 from: (a) CD28–/– mice (gray bars), CD28–/– mice depleted from CD8+ T cells in vivo by mAb (white bars) and CD28–/–CD8–/– mice (black bars). (b) WT mice (gray bars), WT mice depleted from CD8+ T cells in vivo by mAb (white bars), and CD8–/– mice (black bars). The y axis represents the number of positive cells per one million cells. Purified protein derivative (PPD) is used at a concentration of 100 μg/ml. (c) WT mice (gray bars), STAT4–/– mice (white bars), and STAT4–/–mice depleted from CD8+ T cells in vivo by mAb (black bars). (d) Expansion of IFN-γ–producing T cells after ex vivo depletion of CD8+ T cells: MOG p35-55–specific IFN-γ–producing cells were measured by ELISPOT in cultures of splenocytes harvested on day 14 from CD28–/– mice before and after CD8+ T cell depletion ex vivo by magnetic beads. The frequency of IFN-γ–producing cells is significantly higher after removal of CD8+ T cells at all concentrations of MOG or PPD (P < 0.02). s/p, status post.
Mice deficient in STAT4 lack IL-12 induced IFN-γ–production and Th1 differentiation and have been recently reported to be resistant to the induction of EAE (24). If CD8+ T cells exert their regulatory function through limiting the expansion of Th1 clone size, one would hypothesize that removal of CD8+ T cells would not overcome the disease resistance in STAT4–/– mice. As expected, depletion of CD8+ T cells prior to disease induction in STAT4–/– mice did not worsen the clinical disease as compared with control non–CD8-depleted STAT4–/– mice (mean maximal grade of 0.6 ± 0.65 in depleted group versus 0.5 ± 0.7 in control group, n = 5 in each group). As shown in Figure 6c, CD8+ T cell depletion in STAT4–/–mice did not lead to expansion of MOG-specific Th1 cells.
Enhancement of an anti–MOG 35-55 peptide Ab response in the absence of CD8+CD28– T cells. MOG-induced EAE is characterized by significant demyelination in the CNS, which is thought to be mediated by anti-MOG Ab’s (25, 26). We have shown previously that relative Ab titers from immunized CD28–/– mice sera were significantly lower than those from sera of immunized WT animals (4). These results are consistent with published data that CD28 signaling functions to augment T cell–dependent B cell growth and Ig secretion (27). To evaluate the effect of CD8+CD28– T cells on humoral immune response, we compared the Ab titers of the different groups of mice 14 days after immunization. We found that the relative Ab titers in sera from immunized CD28–/– mice were significantly lower than those from CD8-depleted CD28–/– animals (CD28–/– mean titer = 1/50; WT mean titer = 1/2,400; CD8-depleted CD28–/– mean titer = 1/3,200; and mean titer of CD28–/–CD8–/– mice = 1/3,200, P = 0.0001 for the latter two groups versus CD28–/–). Similarly, CD8-depleted WT mice demonstrate higher Ab titers in sera (WT mean titer = 1/2,400; CD8-depleted WT animals mean titer = 1/4,800; and mean titer of CD8–/– mouse = 1/5,600). These data establish that lack of CD8+CD28– T cells can cause expansion of CD4+ Th1 cells that then provide help to B cells to promote Ab production.
Depletion of CD8+ T cells ex vivo leads to expansion of Th1 CD4+ T cells. Primed splenocytes obtained from MOG peptide–immunized CD28–/– mice were cultured with MOG 35-55 peptide for 24 hours in ELISPOT plates. CD8 depleted splenocytes were obtained after magnetic bead separation and cultured similarly. Figure 6d shows increased frequency of IFN-γ–producing Th1 cells in CD8-depleted cultures consistent with the in vivo data described above. These data establish the inhibitory effect of CD8+CD28– T cells on Th1 clone size in vitro. The ex vivo CD8+ T cell depletion did not result in any significant change in the frequency of Th2 cells (not shown), again consistent with the in vivo data described above. Furthermore, since TGF-β is a potential effector of immune regulation by CD8+ T cells (13, 28), we examined the supernatants of in vitro cultures by ELISA for TGF-β. There was no statistically significant difference in the TGF-β content from splenocyte cultures incubated with MOG before and after in vivo or ex vivo CD8+ T cell depletion (data not shown). Thus, regulation by CD8+ T cells appears to be independent of secretion of TGF-β or Th2 cytokines.
The natural resistance to EAE in CD28–/–is in part mediated by a negative regulatory signal provided by B7-1 through CTLA4 (4). There is also growing evidence that signals mediated by cell-surface molecules such as CTLA4 are also involved in the effector function of some regulatory T cells (29–31). We investigated whether CD8+CD28– T cells express CTLA4 in naive or immunized CD28–/– animals. We found that naive CD8+CD28– cells lack both intracellular CTLA-4 and surface CD25 expression (data not shown). CD8+ T cells from immunized CD28–/– mice on day 14 lack CTLA4 expression, while CD4+ T cells show some CTLA-4 expression (3%). This low level of CTLA4 expression in CD4+ T cells is consistent with published data indicating that CD28 costimulation is required for optimal CTLA-4 expression (32). Thus, regulatory CD8+CD28– T cells do not express CTLA4.
Expression of a panel of markers including memory markers (CD45RB, CD44, CD62L), cytokine receptors (CD25, CD122, CD132, and CD210), and NK marker (NK1.1) was compared between the CD8+CD28– T cells with regulatory function and CD8+CD28+ T cells that lack such function. The two subpopulations were only distinguished by the expression of CD122 (IL-2 receptor β chain): CD122 was expressed almost entirely by CD8+CD28+ T cells. CD45RB and CD62L were both expressed by all CD8+CD28+ and CD8+CD28– T cells, while none of the subpopulations expressed any NK1.1, CD25, or CD210. Expression of both CD132 and CD44 was comparable in the two populations (data not shown).
In vitro suppression induced by CD8+CD28– T cells requires cell-cell contact and is APC dependent. To study the ability of CD8+CD28– T cells to suppress in vitro, a unique coculture system was set up using the ELISPOT assay. Using splenocytes from MOG-immunized CD8–/– mice as responder cells, we first demonstrated that 100% purified CD8+CD28– T cells derived from CD28–/– mice can suppress IFN-γ production only if in direct contact with responder cells (Figure 7a). Titration of the same number of CD28+/+ splenocytes as CD8+CD28– T cells into cultures did not lead to a decrease of IFN-γ, thereby excluding the possibility that an increase in total responder cell number was responsible for the suppressive effect (Figure 7a). Furthermore, CD8+CD28– T cells were able to induce suppression in a dose-dependent manner (responder cells–to–regulatory cells ratios of 2:1, 4:1, 16:1, with loss of suppression at a ratio of 32:1). This suppression was not a peculiarity of cells from CD28–/– animals, since CD8+CD28– but not CD8+CD28+ cells from WT mice demonstrated similar suppressive activity in vitro (Figure 7b). These in vitro findings are consistent with our adoptive transfer experiments demonstrating that regulatory CD8+ T cells do occur in WT mice and are confined to a subpopulation of CD8+ T cells lacking CD28 expression. Using the apoptosis kit, we have also determined that there was no increase in the number of apoptotic cells in either CD3+ or CD3– populations whether they were incubated with CD8+CD28– or CD8+CD28+ T cells (data not shown). Furthermore, blockade of Fas-FasL interaction in the suppression assay cultures did not result in reversal of suppression by the CD8+CD28– cells in vitro (data not shown). These results exclude apoptosis of T cells or APCs as a mechanism of regulation by these cells.
CD8+CD28– T cell–induced suppression in vitro requires cell-cell contact and is APC dependent. MOG p35-55–specific IFN-γ–producing cells were measured by ELISPOT in cultures of CD8–/– on day 14 after immunization. (a) Addition of 100% purified CD8+CD28– T cells in a 2:1 ratio leads to complete suppression of IFN-γ spots only if in direct contact with responder cells (white bar), but not if separated by a Transwell membrane (black bar). Titration of the same number of CD28+/+ splenocytes as CD8+CD28– T cells only led to an increase of IFN-γ spots (dark gray bar). (b) CD8+CD28–, but not CD8+CD28+ cells originating from WT mice demonstrate similar suppressive activity in vitro because CD8+CD28– cells generated from CD28–/– mice as demonstrated. (c) Purified CD8+CD28– T cells are not able to suppress IFN-γ production by 100% purified CD4 T cells stimulated by PMA (10 ng/ml) and ionomycin (400 ng/ml). The coculture of CD8+CD28– and CD4+ cells results in accumulation of spots produced by each individual group of cells (black bar) after stimulation with PMA plus ionomycin. Con A is unable to stimulate purified CD4 cells in the absence of accessory cells. (d) Purified CD8+CD28– T cells added to cultures in 2:1 contact induce complete suppression of IFN-γ production by naive CD8–/– splenocytes stimulated by Con A at 5 μg/ml.
CD8+CD28– T cells do not suppress IFN-γ production by purified CD4+ T cells stimulated by PMA and ionomycin, while they suppress CD8–/– splenocytes stimulated with Con A by ELISPOT (Figure 7, b–d). Since T cell proliferation to Con A is dependent on the presence of APCs as accessory cells (33, 34) (confirmed in Figure 7c), these data, taken together with the Transwell culture system data, clearly demonstrate that CD8+CD28– T cells require cell-cell contact and APCs for their regulatory function.
To study whether CD8+ CD28– T cell regulation is MHC class I restricted, we repeated the in vitro suppression assays in the presence of anti-MHC class I–blocking Ab’s or appropriate isotype control Ab and measured the frequency of IFN-γ–producing cells by ELISPOT. We found that 288 ± 45 IFN-γ spots were generated by responder cells incubated with MOG alone and 13 ± 5.7 spots by responder cells incubated with MOG in the presence of CD8+CD28– T cells at a ratio of 4:1, while 375.2 ± 100 spots were generated by responder cells after addition of MHC class I–blocking Ab in the presence of CD8+CD28– T cells but not when isotype control (9.2 ± 4.5 spots) was added to the latter group (P = 0.005 between the last two groups). In contrast, addition of anti–Qa-1b to the in vitro suppression assays (11.75 ± 3.5 spots) did not result in reversal of suppression by regulatory CD8+CD28– T cells. These data indicate that the regulatory functions of the CD8+CD28– T cells are MHC class I and not Qa-1b restricted.
CD8+CD28– T cells inhibit the upregulation of costimulatory molecules on APCs leading to less efficient antigen presentation. To evaluate the effect of regulatory cells on APC function, we first set up a functional MLR assay. CD8–/– splenocytes were cultured with purified CD8+CD28– T cells, purified CD8+CD28+ T cells, or alone and stimulated with Con A for 24 hours. Plastic adherent cells were isolated from these cultures, irradiated, and used as stimulators in a MLR. APCs previously cultured with CD8+CD28– T cells have markedly decreased ability to stimulate BALB/c responder cells in allogeneic mixed-leukocyte reactions compared with APCs previously cultured with CD8+CD28+ T cells (Figure 8a). Using flow cytometry, we then studied the expression of costimulatory molecules by DCs (CD11c+) from the APC cultures described above. Figure 8b shows the mean decrease in the percentage of CD11c+ cells that express costimulatory molecules after coculture with CD8+CD28– T cells relative to CD11c+ from cocultures with CD8+CD28+ cells. Thus, DCs precultured with CD8+CD28– cells have significantly downregulated CD80, CD86, and CD40 costimulatory molecules and are less-efficient APCs. To confirm these findings, we stimulated CD8–/– splenocytes from MOG-immunized mice in vitro with MOG peptide, in the presence of purified CD8+CD28– or CD8+CD28+ T cells similar to the experiment above. CD11+ cells isolated from these two culture conditions were then irradiated and used as APCs for 100% purified CD4+ T cells from MOG-immunized WT mice (1:1 ratio) in ELISPOT plates. As seen in Figure 8c, CD11c+ cells conditioned by interaction with CD8+CD28– T cells in vitro are significantly less efficient in presenting antigen to primed CD4+ T cells.
CD8+CD28– T cells induce suppression by modification of APCs. (a) APCs incubated with CD8+CD28– T cells and Con A for at least 24 hours have significantly decreased capacity to stimulate BALB/c splenocytes (white bar) than APCs exposed to Con A and CD8+CD28+ T cells (black bar). (b) Mean decrease in percentage of APCs expressing CD40, B7-1, and B7-2 (y axis). CD11c+ cells conditioned by preculture with CD8+CD28– T cells for 24 hours in the presence of Con A were stained for expression of CD40, B7-1, and B7-2, and the percentage of positive cells was compared with CD11c+ cells precultured with CD8+CD28+ T cells. (c) Antigen-presenting capacity of APCs cocultured with CD8+CD28– T cells and MOG (white bar) as compared with that of APCs cocultured with CD8+CD28+ T cells and MOG for 24 hours (black bar). As expected, 100% purified primed CD4+ T cells are unable to respond to MOG in the absence of APCs (gray bar).