Real-time immune cell interactions in target tissue during autoimmune-induced damage and graft tolerance

To study CD4⁺ T cell responses in target tissue, we used CD4⁺ T_eff and T_reg cells from the NOD.BDC2.5 TCR transgenic mice (Katz et al., 1993), with a specificity against a natural antigen in the pancreatic islet β cells, chromogranin A (Stadinski et al., 2010). ACE offers the technical advantage of noninvasive access and high resolution in vivo imaging, but studies using ACE could be complicated by a status of immune privilege attributed to this compartment of the eye (Benhar et al., 2012). To test whether this impacts on the immune responses of antigen-specific T cells in the islet grafts in ACE, we compared the frequency of immune damage by BDC2.5 CD4⁺ T_eff cells against β cells in the ACE graft and that in the native pancreas. Donor pancreatic islets were injected into ACE through the cornea (Speier et al., 2008a,b; Abdulreda et al., 2011) at least 2 wk before T cell transfer, to ensure stable engraftment of the islets and complete healing of the minor injection wound.

Although previous studies in other settings showed that immune cells in ACE could be impacted by the eye-associated immunoprivilege (Benhar et al., 2012), in our model we found that the β cell–specific CD4⁺ BDC2.5 T_eff cells destroyed the islets in the pancreas and the islet grafts in ACE at a similar tempo. Importantly, the protective T_reg cells acted with a similar efficacy (∼50%) in controlling T_eff cell damage in ACE and in the endogenous pancreas (Fig. 1, A–D). Antigen-specific T_reg cells have been shown to be potent suppressors in several autoimmune settings, including models of type 1 diabetes (Tisch and Wang, 2008; Shevach, 2011). The main reason that only 50% of mice were protected in our experiments was likely because of the potency of the antigen-specific T_eff cells, which were purified BDC2.5 CD4⁺CD25⁻Foxp3⁻ cells. Nonetheless, these results are consistent with a previous study that found rejection of fully MHC-mismatched islets in ACE occurred similarly to that in a conventional extra-pancreatic implant site, the kidney sub-capsular space (Abdulreda et al., 2011). Therefore, the overall kinetics of immune destruction and protection of engrafted islet tissue in ACE was comparable to that in the native pancreas, and thus the islet grafts in ACE could serve as a surrogate in noninvasive and longitudinal imaging studies of basic T cell biology at the effector phase in the nonlymphoid target tissue.

CD4⁺ T cells are categorized into several helper and regulatory subsets. Their function as killers has also been shown (Hahn et al., 1995). The in vivo capacity of CD4⁺ T_eff cells killing target β cells was shown in Fig. 1 (A–D). How CD4⁺ T cells kill remains to be fully examined. Most target tissues do not express MHC class II molecules, which are necessary for antigen-specific, cognate interaction with CD4⁺ T cells. To study tissue destruction by antigen-specific CD4⁺ T_eff cells, pancreatic islets tagged with cyan fluorescence proteins (CFP; Hara et al., 2006) were grafted in ACE. CD4⁺Foxp3⁻ BDC2.5 T_eff cells marked with green fluorescence proteins (GFP) were injected into the intraocular graft-bearing animals. The GFP⁺CD4⁺ T_eff cells appeared in the islet grafts and engaged in direct contact with their target β cells. We then used Annexin V in situ cytolabeling to visualize apoptosis of β cells, by injecting allophycocyanin (APC)-conjugated Annexin V into ACE. The use of this in vivo assay for β cell apoptosis was described in detail in previous works (Speier et al., 2008a,b). Apoptosis signals were present in either the contact zone between T_eff cells and target islet cells, or on the target islet cells with T_eff cells in the vicinity but not in direct contact (Fig. 1 E and Video 1). We also examined the involvement of myeloid cells at the inflammatory site, by in situ immunocytolabeling (Abdulreda et al., 2011) with anti-CD11b antibodies. We could not detect CD11b⁺ cells in most of the areas wherein T_eff cells interacted with target β cells. A low frequency of CD11b⁺ cells were found but usually in the periphery of damaged grafts. Importantly, T_reg cells colocalized in the protected clusters of β cells that persisted amid areas of immune damage (Fig. 2, A–C). Overall, the majority of Annexin V signals associated with β cells rather than T_eff cells and the amount of the apoptotic signals on β cells positively correlated with the number of T_eff cells at the inflammatory site (Fig. 2, D and E). The exact molecular cause of the immunopathology by the CD4⁺ T_eff cells remains unclear. IFN-γ and IL-17A could be detected by flow cytometry in substantial proportions of the CD4⁺ T_eff cells in the draining cervical lymph nodes of the eyes (14 ± 2 and 6 ± 1%, respectively; mean ± SEM; n = 8 mice). However, further studies are needed to determine whether these or other cytokines have a pathogenic role. The imaging data suggest that, although not exclusive, direct contact may be involved in CD4⁺ T_eff cell killing of target cells, even in the absence of CD8⁺ T_eff cells. The contact-dependent mode is a hallmark of cytotoxicity by CD8⁺ T_eff cells.

We next studied noninvasive CD8⁺ T_eff cell activity in target tissue, by using CD8⁺ OT1 transgenic T cells (Hogquist et al., 1994), which are specific toward a model antigen, ovalbumin. Donor islets from both the RIP-mOVA transgenic mice (Kurts et al., 1996), which express ovalbumin in β cells, and the MIP-CFP mice, which lack ovalbumin but have CFP-labeled β cells (Hara et al., 2006), were grafted together in ACE of C57BL/6 (B6) animals. We selected transplanted animals that carried at least one pair of conjoined mOVA⁺CFP⁻ and mOVA⁻CFP⁺ grafts like Siamese twins, and transferred them with antigen-activated CD8⁺GFP⁺ OT1 T_eff cells. This was done to examine antigen-specific killing versus bystander tissue destruction (Tite and Janeway, 1984). The CD8⁺ OT1 T_eff cells selectively destroyed β cells carrying the specific antigen (Fig. 3, A–C).

Strikingly, the bystander RIP-mOVA⁻CFP⁺ islets conjoined to damaged mOVA⁺CFP⁻ islet grafts grew at the interface of immune destruction within days, rather than being subjected to killing (Fig. 3 A). Increase of the islet mass lacking the specific antigen required close juxtaposition with the site of antigen-specific immune responses. RIP-mOVA⁻CFP⁺ graft mass remained constant if it was not immediately adjacent (i.e., isolated) to a graft harboring the specific antigen (Fig. 3, A–C). Importantly, the increased mass of conjoined RIP-mOVA⁻CFP⁺ grafts was not a result of hypertrophy of the β cells, as the nuclear density of the conjoined and isolated islets was unchanged (Fig. S1). Moreover, imaging analyses showed preservation of the three-dimensional (3D) structure of these islets (Videos 2 and 3), precluding the possibility of imaging artifacts associated with islet flattening over time. Because the CFP expression in these islets is driven by the insulin promoter (Hara et al., 2006) and therefore labels differentiated β cells, this direct observation suggests that β cells can regenerate by replication at the site of immune damage, with an extraordinary potential of doubling in days.

To examine β cell replication under inflammatory conditions in a site other than ACE, we performed our experiments using a conventional islet transplantation model. Recapitulating the settings in ACE using islets with or without specific antigens for CD8⁺ OT1 T_eff cells, B6 recipient mice were transplanted under the kidney capsule with islets from either wild-type B6 or RIP-mOVA⁺ transgenic donors, or a mixture of the two. The premixed islets from RIP-mOVA⁺ and wild-type B6 donors, or single group controls, were pelleted by centrifugation before transplantation under kidney capsule. The kidney subscapular space, although does not allow noninvasive longitudinal analyses of the same graft tissue at a cellular level, enabled the retrieval of a relatively large number of islet grafts for histological analyses.

After engraftment of islets transplanted under the kidney capsule, the recipient animals were injected with activated CD8⁺ OT1 T_eff cells, as in the animals bearing ACE grafts (Fig. 3). We then administered BrdU to the animals to label proliferating cells. The CD8⁺ T_eff cells destroyed the RIP-mOVA⁺ islet grafts. This was confirmed by microscopic examination and insulin-staining of kidney tissue sections at the site of the islet engraftment in animals transplanted with RIP-mOVA⁺ islet alone (unpublished data). Using tissue sections of the islet grafts from the mice receiving B6 islets alone or a mixture of B6 and RIP-mOVA⁺ islets, we conducted immunofluorescence staining to detect BrdU incorporation in the nuclei of proliferating β cells. We found that in the group with the mixed engraftment (RIP-mOVA⁺ and B6 islets) and subsequent destruction of RIP-mOVA⁺ islets by OT1 CD8⁺ T_eff cells, there was a substantial increase in the proportion of BrdU⁺ β cells in the remaining B6 islets, compared with the group that was transplanted with B6 islets alone (Fig. 4).

Although it cannot be determined with absolute certainty in the histological analyses whether one particular B6 islet was once in direct contact with one particular RIP-mOVA⁺ islet after the latter was destroyed, the islets from RIP-mOVA⁺ and wild-type B6 donors were premixed and pelleted before being implanted in the renal subcapsular space. In addition, we counted 34–36 islets in each group. Thus, increased BrdU staining in the B6 islets in the mixed transplant group, compared with that in the B6 islets alone transplant group, represents at a group level the effect of close physical proximity of the bystander B6 islets with the RIP-mOVA⁺ islets before the latter was destroyed by OT1 T_eff cells. These results, obtained with a platform distinct from the ACE model, corroborated the notion that β cells lacking specific antigens are not subject to bystander killing or damage but replicate in an inflammatory setting, which is consistent with the observations from the imaging analyses of islet grafts in ACE.

Immune effector function at the target tissue is controlled by various mechanisms coordinated by T_reg cells (Josefowicz et al., 2012). How exactly T_reg cells suppress immune response in vivo is still debated. Initial studies with in vitro transwell culture systems showed that T_reg cell suppression was effective only if T_reg cells were placed in the same culture chamber with T_eff cells and antigen-presenting cells (Takahashi et al., 1998; Thornton and Shevach, 1998). Although subsequent studies showed that T_reg cells could inhibit T_eff cell activation by modulating antigen-presenting cells (Tadokoro et al., 2006; Onishi et al., 2008; Wing et al., 2008), several in vitro studies also demonstrated that both human and murine T_reg cells could directly suppress T_eff cells independent of antigen-presenting cells (Ermann et al., 2001; Nakamura et al., 2001; Piccirillo and Shevach, 2001; Baecher-Allan et al., 2006; Hagness et al., 2012; Huang et al., 2012). However, ex vivo and in vivo imaging studies in lymph nodes did not detect stable contact between T_reg and T_eff cells (Mempel et al., 2006; Tang et al., 2006). This contradiction between in vitro and in vivo studies has left a doubt about the in vivo relevance of contact-dependent T_reg suppression. The in vitro trademark activity of T_reg cells remains to be reconciled in vivo.

In this study, we examined the pathophysiological relevance of T_reg–T_eff contact at the effector phase in the target tissue in vivo. We used the NOD.SCID reconstitution model with antigen-specific T_reg and T_eff cells that are genetically tagged with different fluorescence markers for stable labeling and longitudinal study. Adoptive transfer of T cells to lymphopenic animals is followed by homeostatic proliferation and activation of the transferred T cells to fill empty niches in the lymphoid organs (Surh and Sprent, 2008), which could complicate studies of T cell activation and differentiation. However, lymphopenia-associated activation is likely to have minimal impact on our study, as we focused on T cell biology at the final effector phase in the nonlymphoid target tissue, i.e., during the effector phase after activation and differentiation. The lymphopenic reconstitution model is also necessary to avoid undercounting invisible interactions (see Materials and methods) and to generate meaningful measurement of the interactive behavior among T cell subsets.

Indeed, in the T_reg cell–protected grafts, a majority of T_eff cells were in direct contact with T_reg cells (Fig. 5 A); they displayed a dynamic and contact-featured choreography. The interaction between the T_reg and T_eff cells usually persisted for the entire length of the imaging sessions (≥30 min) and was characterized by reduced motility (Fig. 5, A–E; and Video 4). This direct contact between T_reg and T_eff cells was not due to mere crowdedness; in tissue areas that were only sparsely infiltrated, long-lasting contact between T_reg and T_eff cells was still evident (Fig. S2).

T_reg cells can dampen the expression of the co-stimulatory molecules CD80 and CD86 on the surface of DCs, and thus inhibit T cell activation (Shevach, 2008; Wing et al., 2008). Whether the function of T_reg cells in the target tissue depends on DCs during the effector phase remains unclear. To examine T_reg–T_eff interaction in the context of DCs in protected target tissues, we injected fluorescence-conjugated anti-CD11c antibodies to visualize DCs, in addition to GFP- and CFP-labeled T_reg and T_eff cells, respectively (Fig. 6, A and B). CD11c⁺ DCs could be detected in the islet graft, mostly at the periphery (Fig. 6 B). Consequently, the majority of T_reg–T_eff cell interactions within the graft occurred in the absence of DCs, and it was also the most abundant among the various types of interactions of T_reg, T_eff, and/or CD11c⁺ DCs. Clusters of the three types of cells, T_reg–T_eff–DC, could be detected but at a much lower frequency than that of T_reg–T_eff cell interaction without DCs. T_eff–DC or T_reg–DC interactions could be found at minor frequencies (Fig. 6 C). The interactions between T_reg–T_eff cells were also stable, with or without CD11c⁺ DCs (Fig. 6, D and E). The interactions between CD11c⁺ cells and T_eff or T_reg cells, although occurring in only a minor proportion of the T cells, were also mainly long lasting, with T_reg–CD11c⁺ cells interactions being somewhat less stable (Fig. 6, D and E). Overall, these results show that direct contact–based interactions between T_reg and T_eff cells persisted with or without CD11c⁺ DCs, which could reflect distinct subsets of T cells or distinct stages of the T cell function in the target tissues. The functional relevance of the different types of interactions has already been documented in vitro (Takahashi et al., 1998; Thornton and Shevach, 1998; Ermann et al., 2001; Nakamura et al., 2001; Piccirillo and Shevach, 2001; Baecher-Allan et al., 2006; Tadokoro et al., 2006; Onishi et al., 2008; Wing et al., 2008; Hagness et al., 2012; Huang et al., 2012). Our noninvasive in vivo imaging studies shows that those direct interactions do exist in vivo in target tissue. Further studies are needed to determine which interactions are most relevant in what settings for which types of functions.

Next, we examined the behavior of T_reg cells in a setting of failed immune regulation. We found that most T_reg cells at the site of extensive tissue damage were still persistently interacting with T_eff cells, with durations (interaction time) comparable to those in the protected tissues (Fig. 7, A and B). However, T_reg cells were largely outnumbered by T_eff cells; as a result, most T_eff cells were without T_reg cell interactions (Fig. 7, C and D). Thus, regardless of success or failure in protecting the target tissue, T_reg cells persistently interacted with T_eff cells, but an imbalance in the numbers of T_reg versus T_eff cells characterized the outcome, i.e., immune damage versus protection.

The imbalance of T_eff versus T_reg cells in the target tissue developed in some animals but not others even though they were injected with the same type of T_reg and T_eff cell mixture in the same batch of experiments. We studied the kinetics of the imbalance, taking advantage of our noninvasive platform to image both T_reg and T_eff cell populations in the same islet grafts longitudinally. The grafts were analyzed at two time points: days 10–12, when all animals were free of diabetes but had an onset of infiltration of both T_reg and T_eff cells (without substantial damage of the grafts); and days 17–20 when some animals suffered from new-onset diabetes (the islet damaged group) but other remains diabetes-free (the islet protected group). Of note, if an animal was still free of diabetes at this time point, it would remain protected for a long term, as shown in Fig. 1.

At the time point of days 10–12, the islets that would be damaged later and the islets that remain protected for a long term did not have significant difference in T_eff cell infiltrates (Fig. 7 E). The difference lays in the T_reg cells. The first group, at this predamage stage, had significantly lower numbers of T_reg cells compared with the protected group (Fig. 7 F). As a result, the T_reg/T_eff cell ratio was significantly reduced at this predamage time point in the would-be damaged group versus the would-be protected group (Fig. 7 G). It was only at the time point of day 17–20, when the damaged group had extensive destruction of the tissue, that the grafts had significantly more T_eff cells than the protected group (Fig. 7, E and G). Overall, these data show that the grafts with successful or failed T_reg cell protection did not have significant difference in the number of T_eff cells at the onset of T cell infiltration in the target tissue. These results suggest that in this experimental setting, the two different immune outcomes (graft protection or damage) are probably not caused by a difference in the numbers of T_eff cells expanded during the priming phase. Alternatively, the results highlighted the critical requirement of sufficient recruitment of T_reg cells to the inflammatory site and the continuous presence of T_reg cells for optimal suppression, as suggested in previous studies with models of autoimmune oophoritis (Samy et al., 2005), islet transplantation (Zhang et al., 2009), and other models of immune-mediated (Tisch and Wang, 2008; Shevach, 2011).

To further examine the behavior of T_reg cells in a setting of severe imbalance with T_eff cells, we examined the impact of acute removal of T_reg cells after establishment of target tissue protection. This approach also tested whether persistent contact with T_reg cells might impinge a lasting change of migratory behavior of T_eff cells. We used a Foxp3^DTR transgenic model, which carries a diphtheria toxin (DT) receptor transgene under the control of Foxp3 promoter, enabling acute depletion of 80–90% of T_reg cells with a low dose of DT (Feuerer et al., 2009). In ACE islet grafts, the T_eff cell–mediated islet damage was suppressed by using either Foxp3^DTR+ T_reg cells or Foxp3^DTR− T_reg controls. After stable protection of the islet grafts by the T_reg cells was established, the animals were treated with DT. This treatment led to an acute removal of the T_reg cells and a precipitous reduction in the T_reg/T_eff cell ratio in the tissue (Fig. 8 A). The efficacy of GFP⁺ cell removal in the islet graft indicated that the adoptively transferred GFP⁺ cells maintained Foxp3 expression even at the effector phase in the target tissues. Depletion of T_reg cells led to extensive tissue damage (Fig. 8 B), and increased motility of T_eff cells (Fig. 8, C–E). While the residual T_reg cells remained in stable contact with T_eff cells, the T_reg/T_eff cell disproportion caused by the T_reg-cell depletion treatment resulted in most of the T_eff cells in the target tissue no longer having T_reg-cell partners. Thus, intimate T_reg–T_eff interaction did not irreversibly alter the aggressiveness of the T_eff cells.

The function of T_reg cells depends on CTLA4 (Wing et al., 2008), which also regulates T_eff cell function (Teft et al., 2006). We tested here the role of CTLA4 in maintaining T_reg–T_eff cell interaction by administering anti-CTLA4-antibody blockade after stable T_reg–T_eff cell interactions and T_reg-cell protection were established. The anti-CTLA4 treatment under this condition did not cause diabetes (data not shown). It increased T_eff cell numbers in the target tissue, more so than in T_reg cell numbers (Fig. 9, A and B), and resulted in decreased T_reg/T_eff ratios (Fig. 9, C). The treatment did not immediately disrupt the interaction between the CD4⁺ T_reg–T_eff pairs. However, the proportion of interacting T_reg–T_eff pairs declined over time after CTLA4 blockade (Fig. 9, D), and their interaction time was shortened (Fig. 9, E and F). Although CTLA4 blockade led to increased motility of CD4⁺ BDC2.5 T_eff and T_reg cells in T_reg cell–protected grafts, it decreased the motility of CD8⁺ OT1 T_eff cells. Moreover, in a model wherein CTLA4 in CD8⁺ OT1 T_eff cells were modulated with RNAi (Chen et al., 2006; Miska et al., 2012), CTLA4 reduction decreased motility of T_eff cells, suggesting an intrinsic effect of CTLA4 in T_eff cell motility (Fig. 10 and Videos 5 and 6). Collectively, these results suggest that CTLA4 might influence T_reg–T_eff interaction through motility regulation, but the exact effect depends on the nature of the immune settings and cell types.

Control of immune damage at the effector phase is a crucial and perhaps the most realistic therapeutic target in clinical intervention of immune-mediated diseases (Chatenoud, 2011). Improvement of therapeutic interventions will require in-depth understanding of the immune cell behavior in target tissues and of the reaction of target tissue cells in response to insult. The current study suggests that the contact-dependent mode of immune cell interaction in the target tissue is a critical part of pathophysiology at the effector phase of immune responses, and immune tolerance induction may be facilitated by promoting intimacy between pathogenic and protective immune cells. In this regard, it is highly relevant that tissue antigen-specificity, as opposed to bystander killing (Tite and Janeway, 1984), shapes tissue fate in the effector phase.

With the tools currently available for longitudinal imaging of antigen-specific T cells in target tissues, we uncovered some basic behaviors of different lineages of T cells during the effector phase. CD8⁺ T cells are well known for contact-dependent killing. CD4⁺ T cells, on the other hand, are better known as various helper subsets, although increasing attention has been put on their potential cytotoxicity function. Although our current models do not allow us to compare the biology of CD4⁺ and CD8⁺ T cells in an ideally analogous setting, our studies provide in vivo evidence for contact-based killing of target cells by both. The observation adds to efforts to understand the behaviors of these two distinct lineages of T cells at various stages of their activation, differentiation and functioning (Mandl et al., 2012). Our results, however, do not exclude indirect mechanisms of target killing by CD4⁺ T_eff cells, and development of new tools should enable further studies to investigate such mechanisms in vivo.

It is important to make the distinction between our findings of stable T_reg–T_eff contact interaction in target tissue and those in previous reports on lack of direct T_reg–T_eff contact in lymph nodes (Mempel et al., 2006; Tang et al., 2006). T_reg cells play a major role in peripheral immune tolerance, likely through a variety of mechanisms (Tisch and Wang, 2008; Shevach, 2011; Josefowicz et al., 2012). However, key features of T_reg cell biology in vivo remain to be clarified, including whether T_reg cells interact with T_eff cells through direct contact. Two groups independently reported that T_reg cells could not suppress in vitro proliferation responses of T_eff cells if they were placed in a different chamber in a trans-well culture system (Takahashi et al., 1998; Thornton and Shevach, 1998). Although this could be attributed to an effect of T_reg cells on antigen-presenting cells (Tadokoro et al., 2006; Onishi et al., 2008; Wing et al., 2008), robust evidence has also been presented for direct suppression T_eff cells by T_reg cells independent of antigen-presenting cells (Ermann et al., 2001; Nakamura et al., 2001; Piccirillo and Shevach, 2001; Baecher-Allan et al., 2006; Hagness et al., 2012; Huang et al., 2012), such that contact-dependent suppression has been regarded as an in vitro trademark of T_reg cell activity (Shevach, 2006). On the other hand, imaging analyses of explanted lymph nodes in an autoimmune diabetes animal model did not detect stable interaction between CD4⁺ BDC2.5 T_eff cells and T_reg cells, but detected interaction between T_reg cell and DCs, suggesting a role of T_reg cells in the priming phase of T_eff cells (Tang et al., 2006). The suppressive effect of T_reg cells in the priming phase has also been documented with the findings from other studies (Tisch and Wang, 2008; Shevach, 2011). Absence of stable T_eff–T_reg cell contact in draining lymph nodes was also reported independently by another group using a different model system (Mempel et al., 2006).

In contrast, our in vivo studies focused on the effector phase in the nonlymphoid target tissue. We found that T_reg cells persistently interacted with T_eff cells through direct cell–cell contact. Importantly, the contact-based interactions between T_reg and T_eff cells in the target tissues were observed both in the presence and in the absence CD11c⁺ DCs, although more often for the latter. In this regard, it should be noted that a previous study (Tang et al., 2006) showed that, in draining lymph nodes, although both T_eff cells and T_reg cells had stable interactions with DCs, such interactions did not lead to stable T_reg–T_eff contact. Therefore, a platform of antigen-presenting cells does not obligate direct T_reg–T_eff cell interaction. In addition, recent evidence also suggested that niche-filling homeostasis of T_reg cells may occur independently of DCs (Pierson et al., 2013). Nevertheless, our observations do not deemphasize the role of DCs in initiation and progression of immune damage. Homeostasis of DCs was shown to play a critical role in autoimmune damage of pancreatic islets (Dissanayake et al., 2011). In the standard NOD model of autoimmune diabetes, as well as in the BDC2.5 TCR-transgenic model, the role of DCs in initiating autoimmune diseases and their potential in tolerogenic therapies have been clearly demonstrated (Turley et al., 2003; Tarbell et al., 2007; Mukhopadhaya et al., 2008; Driver et al., 2010; Tsai et al., 2013). We believe that the visual evidence of direct T_reg–T_eff interaction in target tissue in vivo, with or without the involvement of CD11c⁺ DCs, reconciles the discrepancy between in vitro and in vivo observations on a basic aspect of T_reg cell biology.

The molecular basis of T_reg–T_eff cell direct engagement remains to be elucidated. A previous study (Paust et al., 2004) reported that CD80/CD86-deficient T_eff cells were resistant to T_reg cell suppression. We explored along this line but did not detect expression of CD80 or CD86 on T_reg cells or T_eff cells in the target tissues by cytolableing in situ with specific antibodies, nor did we find a disruptive effect on T_reg–T_eff cell interaction after injecting anti-CD80 or anti-CD86 antibodies directly into ACE (unpublished data). However, our observation in this regard is preliminary in scope and limited to the target tissue, and thus does not invalidate the hypothesis that CTLA4 expressed by T_reg cells may bind to CD80/CD86 on T_eff cells to facilitate direct cellular interactions (Paust et al., 2004). On the other hand, the data from our studies suggest that CTLA4 might affect T_reg–T_eff cell interaction through other mechanisms. A study with a conditional knockout of CTLA4 showed that T_reg cells require CTLA4 for functioning in vivo (Wing et al., 2008). We showed here that CTLA4 played a role in regulating T_reg–T_eff cell interaction in target tissue. This role may be related to motility regulation of both T_reg and T_eff cells. Of note, the role of CTLA4 in T cell motility control has been debated. Schneider et al. (2006) reported that CTLA4 enhanced the motility of T cells and thus reversed the stop signal originated from engaging the TCR, an effect that may preferentially impact T_eff cells over T_reg cells (Lu et al., 2012). This evidence suggested that manipulating CTLA4-based motility control could lead to therapeutic advance. Indeed, a recent study found that anti-CTLA4 antibody treatment inhibited CD8⁺ T cell motility and promoted antitumor immunity (Ruocco et al., 2012). However, another study by Fife et al. (2009) showed that anti-CTLA4 treatment did not alter the motility of autoimmune CD4⁺ BDC2.5 T cells in draining lymph nodes. Our imaging analyses of CD4⁺ and CD8⁺ T cells during the effector phase within target tissue indicate that the exact effect of CTLA4 on T cell motility may vary in different T cell subsets and may be influenced by distinct circumstance of cellular interactions.

In our experiments, CTLA4 blockade caused only modest changes on cellular interaction. It did not substantially break tolerance under the conditions we tested. The small effect on T_reg–T_eff cell interaction could be contributed by altered motility controls, although it is still a challenge to determine the cause and effect relationship in such in vivo settings. It remains to be determined how CTLA4 blockade led to an increase in T_eff cells, rather than T_reg cells, in the target tissue. The resulting imbalance of T_reg/T_eff ratios, however, did not seem likely to account for the changes in durations of T_reg–T_eff cell interactions, as reduced T_reg/T_eff ratios did not lead to reduced T_reg–T_eff interaction time in the other settings of our studies. Although these results suggest novel facets of CTLA4 function beyond the scope of this study (Han et al., 2012), they may also reconcile the debate on whether CTLA4 controls T cell motility (Schneider et al., 2006; Fife et al., 2009). Importantly, in addition to its impact on the TCR stop signals, dysregulation of CTLA4-based motility control may lead to disruption of the T_reg–T_eff cell interaction time in target tissues, which may in turn lead to exacerbated tissue damage.

Our longitudinal and noninvasive observation of live tissue in animals documented growth of healthy tissue that avoided immune cell recognition. For most target tissue cells, unlike immune cells, motility is not typically in their nature. However, they may not be mere sitting ducks in a setting of immune destruction. Rather, they may be capable of a resilient response at the inflammatory front, as illustrated by the dramatic growth of the β cells in bystander islets that were not recognized by antigen-specific T_eff cells. The insulin-producing β cells can regenerate through various mechanisms, which is another topic of debate. Notably, an association between microenvironment inflammation and increased β cell proliferation was recognized during insulitis and pancreatitis (Sherry et al., 2006; Cano et al., 2008; Faleo et al., 2012). A mode of β cell regeneration through replication was demonstrated in animal models, but it was evident only after months of follow up (Dor et al., 2004). A recent study (Yi et al., 2013) showed that fast replication of pancreatic β cells could be induced by treatment with an antagonist compound of insulin receptors that stimulate production of betatrophin. The intravital evidence from our study is consistent with the notion that β cells can regenerate in vivo by replication from differentiated cells, yet with a surprising rate of doubling in mass within days. The fast regeneration could be contributed to by inflammatory signals released at the interface of immune damage. Future studies are needed to uncover the molecular signals and contextual cues that led to this surprising potential of β-cell growth under antigenic incognito. These studies may ultimately aid tissue regenerative therapies in type 1 diabetes and other disorders caused by immune damage.

Lines of transgenic mouse models were crossed or backcrossed to generate the necessary combinations of specific T cells and target tissue with distinct fluorescence reporters. Detailed descriptions of their genetic makeup, antigen-specificity, and fluorescence properties were provided in Table S1. NOD.BDC2.5 (Katz et al., 1993), NOD.Foxp3^DTR (Feuerer et al., 2009), NOD.CTLA4 shRNA (CTLA4KD7; Chen et al., 2006), NOD.PL4 (Chen et al., 2006), OT1 (Hogquist et al., 1994), RIP-mOVA (Kurts et al., 1996), MIP-CFP (Hara et al., 2006) and CAG-CFP (Hadjantonakis et al., 2002) transgenic lines were described previously. The CTLA4shRNA and PL4 lines were backcrossed onto B6 background for >10 generations (Miska et al., 2012). The CTLA4shRNA transgene caused 2–3-fold reduction in CTLA4 expression. The stability of the RNAi effect in the transgenic lines on different genetic background has been established (Chen et al., 2006; Miska et al., 2012). FIR (Foxp3-IRIS-RFP knock-in) mice (Wan and Flavell, 2005) were backcrossed onto the NOD background for 10 generation to create the NOD.Foxp3^FIR line. CAG-CFP, obtained on the C57BL/6 background, were backcrossed onto the NOD genetic background for 10 generations, and crossed with NOD.BDC2.5 and NOD.Foxp3^FIR. All animals were maintained in a specific pathogen–free barrier facility at the University of Miami, and the studies are approved by the Institutional Animal Care and Use Committee at the University of Miami.

Antigen-specific CD4⁺ BDC2.5 T_eff and T_reg cells were purified from transgenic mice on the NOD genetic background that carries the BDC2.5 transgene and the Foxp3^FIR knock-in alleles, in combination with either the PL4 transgene for a GFP reporter or the CAG.CFP transgene for a CFP reporter. Flow cytometry staining of spleen and lymph node cells was conducted according to a standard procedure (Lu et al., 2011). T_eff and T_reg cells were sorted by using a FACSAria II flow cytometer (BD), using the following parameters: CD4⁺CD25⁻CFP (or GFP) ⁺BDC2.5⁺CD62L⁺Foxp3^FIR− for T_eff cells, and CD4⁺CD25⁺GFP⁺BDC2.5⁺Foxp3^FIR+ for T_reg cells. The RFP expressed by the Foxp3^FIR allele is used as a specific lineage marker for flow cytometry purification of T_reg cells, but RFP signal is not strong enough for a reliable tracking of live cells in animals. Therefore we used an additional, nonlineage-specific GFP marker to track the cells in animals after the purified GFP⁺Foxp3^RFP+ cells were transferred. To enable punctuated, acute removal/depletion of Foxp3⁺ T_reg cells after T_reg-cell protection is established, an additional transgene, Foxp3^DTR, the diphtheria toxin receptor (DTR) driven by a Foxp3 promoter, is crossed with NOD.BDC2.5.Foxp3^FIR.PL4 transgenic mice, to generate NOD.BDC2.5.Foxp3^FIR.PL4.Foxp3^DTR+ mice or NOD.BDC2.5.Foxp3^FIR.PL4.Foxp3^DTR− controls. From these mice, Foxp3^DTR+ and Foxp3^DTR− T_reg cells, respectively, were purified with CD4⁺GFP⁺BDC2.5⁺Foxp3^FIR+CD25⁺ markers, for adoptive transfer. Purified T_eff cells, or a mixture of purified T_eff and T_reg cells were injected intravenously into NOD.SCID mice, at doses of 5–10 × 10⁴ cells with T_eff–T_reg ratios at 1:1. Some of the recipients had stable islet grafts established in ACE. In animals bearing islet grafts in ACE, the cervical lymph nodes draining the eyes likely play an important role in activation and differentiation of the antigen-specific T cells. The lymphopenia reconstitution model was necessary for our study, due to current limitations in tools for long-term, simultaneous tracking of different lineages of T cell in vivo. If one uses an immunocompetent animal as a recipient, rather than reconstitute a lymphopenic animal with highly purified, fluorescence-tagged T cell players, an invisible T_reg cell from host could be interacting with a fluorescence-tagged T_eff cell, but we would have to count that T_eff cell as if it had no T_reg cell interaction. That would prevent us from making meaningful conclusions on the extent of T_reg and T_eff cell interaction. Flow cytometry analyses of the recipient mice were performed to verify the stability of Foxp3 expression of the adoptive transferred T_reg cells ∼3 wk later, using the Foxp3-IRES-RFP reporter (Wan and Flavell, 2005).

Antigen-specific CD8⁺ T_eff cells were purified from transgenic mice on the C57BL/6 background carrying the OT1 transgene and the Foxp3^FIR knock-in allele, in combination with the PL4 transgene for a GFP reporter, or CTLA4shRNA/PL4 transgene for a GFP reporter and CTLA4shRNA for CTLA4 RNAi knockdown. Splenocytes from the mice were stimulated with the OT1 specific ovalbumin peptide (SIINFEKL) for 24 h. CD8⁺ T_eff cells were purified (>95% purity) with magnetic-bead-based cell sorting, and injected intravenously into B6 mice bearing islet grafts in ACE or under the kidney capsules.

To examine the molecular role of CTLA4 in maintaining T_reg–T_eff cell interaction, anti-CTLA4 antibody (clone UC10-4F10-11) were injected intraperitoneally into mice, at 40 µg/g body weight, two consecutive doses at 3 d apart, after T_reg cell established graft protection and stable T_reg–T_eff cell interaction in the graft were detectable (typically >15 d after T cell transfer). This monoclonal antibody blocks CTLA4 function without depleting T_reg cells (Read et al., 2000). To test the effect of CTLA4 blockade on CD8⁺ T_eff cells, one dose of anti-CTLA4 antibody was used at the time of OT1 T cell transfer into the animals. For punctual removal of diphtheria-toxin-receptor–tagged T_reg cells, mice carrying DTR⁺ T_reg cells or control DTR⁻ T_reg cells were injected (50 ng/g body weight) with DT at a schedule of day 0, 1, and 3, similar to a previously described regimen (Feuerer et al., 2009), after graft protection was established and stable T_reg–T_eff interactions were detected.

In some instances, fluorescently conjugated antibodies were injected directly into ACE for in vivo cytolabeling in situ. With our intravital imaging platform, the injected antibodies can effectively label cells up to 50 µm deep within the graft tissue, which is at the similar depth capacity for accurately tracking cells marked with genetic tags of fluorescence markers (Abdulreda et al., 2011). After mice were adoptively transferred with T_reg and T_eff cells, mice were injected with fluorochrome-conjugated antibodies directly into the anterior chamber. All antibodies were tested with isotype controls to determine specificity of the in situ labeling. The following is a list of the antibodies and their isotype controls. BV605 conjugated anti-CD11b monoclonal antibody or rat Ig2a isotype control; Alexa Fluor 647–conjugated anti-CD11c, anti-CD80, and anti-CD86 monoclonal antibodies and hamster IgG isotype control (BioLegend). APC-conjugated Annexin V (eBioscience) was used to assess apoptotic cells in ACE grafts. This in vivo assay for β cell apoptosis was described previously (Speier et al., 2008a,b).

The pancreatic islets were transplanted into ACE by injection through the cornea, as previously described (Abdulreda et al., 2011; Speier et al., 2008a,b). The cornea serves as a natural window for noninvasive visualization of transplants. This injection procedure does not entail substantial injury to the cornea or the eye chamber. At least 2 wk were allowed for complete engraftment of the islets before any experimentation. Intravital imaging of the transplants was conducted by confocal microscopy, with a Leica upright TCS SP-5broadband confocal microscope (using Leica 20X/0.5NA HCX APO L U-V-I 12 lens for PBS immersion), as previously described (2008a,b; Abdulreda et al., 2011). T_reg and T_eff cells were visualized by GFP or CFP fluorescence. Target islet cells were visualized by laser backscatter (reflection) or CFP fluorescence 3D (xyz) or time-lapse (xyzt; 4D) noninvasive in vivo imaging was acquired longitudinally. In time-lapse recordings, z stacks were acquired every 2 min for 30–75 min and the z-spacing ranged from 5–7 µm. A key strength of this intravital imaging platform is noninvasiveness that few other intravital cellular imaging platforms have been afforded. This strength not only avoids inadvertent inflammatory signals caused by surgical exposure for imaging needs, but enables us to monitor the same tissue in the same animal longitudinally at cellular resolution. A baseline measurement serves as a rigorous, internal control for all postintervention measurements of the same live tissue in the same live animal.

Animals were routinely monitored by urine and blood glucose levels. Animals transplanted with islet grafts in ACE were examined 2 wk after transplantation for engraftment. After T cell transfer, the islet grafts in ACE were examined every 2–3 d with the imaging microscope. Animals with two consecutive readings of BG > 250 mg/deciliter were considered diabetic. With regard to individual islet grafts in ACE in setting of immune responses, a graft maintaining >80% of its original mass was considered protected, and a graft with <20% of its original mass was considered damaged or failure in immune regulation. In experiments in which a relatively small number of islets were transplanted in ACE, the islet grafts served as indicators of immune responses and the endogenous pancreatic islets maintained blood glucose homeostasis of the animals. The immune damage of islet grafts in ACE always correlated with incidence of diabetes that was caused by immune destruction of endogenous pancreatic islets.

The responses of bystander islet cells in an inflammatory setting were also examined at a transplantation site that is different from ACE, the renal subcapsular space, which is the standard transplantation site for experimental studies of islet grafts in rodent recipients. It has been an invaluable research tool for decades (Ricordi et al., 1987). It allows transplantation of islets in a well-confined location that can then be retrieved for histopathological or molecular analyses of a relatively large number of islets. Islets were transplanted under the kidney capsule of B6 mice by the Diabetes Research Institute Preclinical Cell Processing and Translational Models Core facility following standard procedures (Berney et al., 2001; Faleo et al., 2012). In brief, pancreatic islets were isolated from either wild-type B6 or RIP-mOVA⁺ transgenic donors. B6 islets, RIP-mOVA⁺ islets, or a mixture of B6 islets and RIP-mOVA⁺ islets were prepared into individual aliquots for each transplant recipient. They were then handpicked with a Hamilton syringe and transferred into a polyethylene tube (PE50; BD; inside diameter 0.58 mm; outside diameter, 0.965 mm) that was kinked at one end, and then pelleted in the kinked tubing by centrifugation at 1,000 rpm for 2 min, to pack them together. The pelleting step was done before the transplantation procedure.

After induction of general anesthesia (isoflurane 2%/oxygen mix, to effect), a left flank incision was performed and the left kidney exteriorized and exposed. Under a dissection microscope and using microsurgical forceps, a small breach was performed on the capsule at the caudal pole of the kidney through which the tubing containing the pelleted islets was gently inserted and pushed toward the opposite (cranial) pole. Islets were gently released in the renal subcapsular space under visual microscopic inspection. Next, the tubing was gently removed and the breach on the capsule cauterized to prevent back flow. The kidney was repositioned in the abdominal cavity, and the muscular and cutaneous layer sutured. The graft-bearing mice were rested for ∼7 wk after transplantation to ensure engraftment, and then adoptively transferred with activated OT1 CD8⁺ T cells by intravenous injection. 2 d later, BrdU was injected at a concentration of 10mg/kg every 12 h. On day 10, kidneys were removed, fixed in 4% PFA overnight, followed by immersing in 30% sucrose overnight, and then embedded in OCT. Sections were cut and stained for BrdU incorporation with biotinylated antibodies against BrdU, using an in situ BrdU detection kit (BD) designed for histology use. We modified the secondary staining procedure for immunofluorescence staining. Although isotype controls were not included in the kit, we established the specificity of the BrdU staining procedure with spleen tissue sections from animals with or without BrdU injection treatment. To expose BrdU epitopes in the nuclear DNA, a covered plastic coplin (filled with 50 ml of diluted BD Retrievagen buffer) was preheated in a water bath at 95–97°C. After 30 min, the tissue slides were quickly placed into a preheated jar, and incubated for another 30 min. The tissue slides in the closed jar were then removed and left at ambient temperature for one hour. Primary anti-insulin antibodies (polyclonal guinea pig anti-insulin; DAKO; titration, 1:1,000) and the biotinylated anti-BrdU antibodies were incubated overnight with the tissue sections at 4°C, and then washed according to manufacturer’s instructions. Secondary staining was done with Alexa Fluor 488–conjugated streptavidin (Life Technologies; titration, 1:500) and Alexa Fluor 647–conjugated donkey anti–guinea pig F(ab)₂ (Jackson ImmunoResearch Laboratories; titration, 1:500). Sections were counterstained with DAPI and mounted for fluorescence microscopy. Images were acquired with a Leica inverted TCS SP-5 broadband confocal microscope (using Leica 40×/1.25–0.75NA HCX PL APO lens for oil immersion).

Image analyses were performed using the Volocity software (version 6; Perkin Elmer) as previously described (Abdulreda et al., 2011). Images were denoised and contrast-enhanced equally for consistent analyses. Quantitative analyses of cellular movement and T_reg–T_eff cell interaction dynamics were performed automatically in the Volocity with user feedback on drift-corrected 3D time-lapse recordings. Drift correction was performed in Volocity based on proprietary algorithms. T cell counting and movement tracking were performed automatically by the software and dynamic parameters (e.g., velocity, displacement) were derived from time-lapse recordings. T_reg–T_eff cell interaction time and interaction index were calculated manually. The interaction time between the T_reg–T_eff cell interacting pairs was calculated manually using the time stamps embedded in each image frame in a time-lapse series. The interaction index is calculated by dividing the number of T_eff cells interacting with T_reg cells with the number of T_eff cells not interacting with T_reg cells. β cell (islet) mass was measured by the software, as previously described (Abdulreda et al., 2011), based on volume detected by either laser backscatter or CFP fluorescence.

Annexin V labeling in islet grafts in ACE was quantified with z-stack images that were acquired ∼10–15 min after injection of APC-labeled Annexin V directly into ACE. Using proprietary algorithms in Volocity, we then measured in the 3D images the amount of overlap in the volume the Annexin V–positive stain with either that of β cells (visualized by CFP or backscatter) or CFP/GFP-labeled T_eff cells (Rodriguez-Diaz et al., 2011). Automatic selection, optimized with user feedback, based on fluorescence of Annexin V–positive cells, β cells, and T cells and volume measurements were performed automatically by the software. The overlap (volume) in Annexin V stain with either that of β or T cells was also derived by the software, and was expressed as a fraction of the total volume of the β and T cells in each islet. Similarly, the number of graft-infiltrating T_eff cells was automatically measured based on fluorescence intensity (Abdulreda et al., 2011). Islet cell mass was measured using Volocity software as previously described (Abdulreda et al., 2011), based on islet volume detected either by laser backscatter (reflection) or CFP fluorescence. For example, in Fig. 3, the volume of CFP-labeled β cells was measured based on CFP fluorescence which is in this case restricted to the bystander islets in this case. In brief, a proprietary detection algorithm built into the Volocity software was used to detect CFP signal based on fluorescence intensity. The detection threshold was set with user feedback to restrict the selection to the CFP-labeled β cells. Once the selection was made, the volume was derived automatically by the software. Longitudinal analyses on the same individual islets were performed using the same approach, and numerical values of islet volumes were expressed as means ± SEM at the different time points under the different conditions.

Unpaired Student’s t test was used to compare two samples. For multiple group comparisons, one-way ANOVA tests were performed followed up by Tukey’s post-hoc multiple comparisons test. P ≤ 0.05 was considered significant. Asterisks indicate significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Fig. S1 relates to Fig. 3 and shows digital estimation of cell nuclear density in islet grafts imaged over time in the living animal. Fig. S2 relates to Fig. 5 and shows that the long-lasting contact between T_reg–T_eff in the target tissue occurred even in areas with sparse infiltration of T cells. Table S1 lists the transgenic mouse models genetically tagged with antigen-specific T cell receptor, fluorescence reporters, and lineage markers for the imaging studies. Video 1 (corresponds to Fig. 1) shows direct interaction between antigen-specific CD4⁺ T_eff cells with target β-cells in pancreatic islet grafts in ACE. Videos 2 and 3 (corresponds to Fig. 3) shows antigen-specific CD8⁺ OT1 T_eff cell–mediated destruction of OVA⁺ islets and concomitant growth of juxtaposed bystander OVA⁻ islets. Video 4 (corresponds to Fig. 5) shows stable long-lasting interaction between T_reg and T_eff cells in the target tissue. Video 5 and 6 (corresponds to Fig. 10) shows CTLA4 blockade reduces CD8⁺ T_eff cell motility in target tissue.

This work was supported by grants from the National Institutes of Health (DP3DK085696 to Z. Chen; F32DK083226 to M.H. Abdulreda; 5U19AI050864-10 to A. Pileggi; and NIH-U-01DK089538 to A. Pileggi and P.-O. Berggren), and funds from the Diabetes Research Institute Foundation (to M.H. Abdulreda, A. Pileggi, P.-O. Berggren, and Z. Chen). Additional research support was provided through funds from the Swedish Research Council and the Family Erling-Persson Foundation (to P.-O. Berggren). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding institutions. We thank Mrs. E. Zahr-Akrawi, Mr. J. Enten, and Mr. K. Johnson, and Drs. O. Umland, G. McNamara, R.D. Molano, A. Bayer, A. Caicedo, S. Jacobs, and S. Opiela for their expert assistance and advice.

J. Miska, M.H. Abdulreda, P. Devarajan, J.B. Lui, J. Suzuki, A. Pileggi, and Z. Chen declare no financial or commercial conflicts of interest. P.-O. Berggren is one of the founders of the biotech company Biocrine, and he is also on the board of this company. Biocrine is going to use the anterior chamber of the eye as a commercial servicing platform.