Targeting CD22 Reprograms B-Cells and Reverses Autoimmune Diabetes

A complete description of our methods is available in an online appendix at http://dx.doi.org/10.2337/db08-0420.

We first examined CD22 expression on B-cells in NOD mice. No differences were observed in terms of CD19 and CD22 expression on B220⁺ cells (B220⁺CD19⁺ cells, 82.0 ± 2.5 vs. B220⁺CD22⁺ cells, 83.1 ± 2.7%, NS) (Fig. 1A and B). Furthermore, CD22 is expressed on CD138⁺ cells (Fig. 1C).

The kinetics of CD45⁺CD19⁺ (B-cells) infiltration in the pancreata of NOD mice showed a significant increase at 8 weeks, when the mice began to exhibit islet peri-infiltration (from 22.8 ± 7.8% at 4 weeks up to 65.1 ± 5.0% at 8 weeks of age, P < 0.01) (Fig. 1D). This increase was not observed for CD45⁺CD3⁺ cells (T-cells) (Fig. 1E). The percentage of CD45⁺CD19⁺ cells (B-cells) was significantly higher than CD45⁺CD3⁺ cells (T-cells) in the pancreata of 8-week-old NOD mice (B-cells, 65.1 ± 5.0 vs. T-cells, 30.2 ± 3.2%, P = 0.004) (Fig. 1E).

We first treated NOD mice with anti-CD22/cal mAb to evaluate whether our antibody can successfully deplete B-cells in NOD mice. Two injections (160 μg/kg i.p. 5 days apart, D0–D5) of anti-CD22/cal mAb elicited a quick and profound depletion of B-cells in the peripheral blood of 10-week-old NOD mice; the effect appears at 1 week and lasts for 5–7 weeks (Fig. 1F and H). Control NOD mice did not appear to be depleted of B-cells (Fig. 1F and G), whereas the group treated with an equivalent dose of unconjugated anti-CD22 mAb (0.2 mg/injection i.p. 5 days apart, D0–D5) showed transient and partial B-cell depletion (Fig. 1F and I). B-cell recovery was complete by 8–10 weeks after therapy (Fig. 1F and H). No B-cell depletion was evident when using the control mAb, a mouse IgG1 anti-rat very-late antigen 4 mAb that does not bind to mouse cells and is conjugated to calicheamicin (GG5/cal) (data not shown).

The effect of anti-CD22/cal mAb treatment on diabetes onset was evaluated in female 10-week-old NOD mice. NOD mice were treated with two injections of 160 μg/kg anti-CD22/cal mAb 5 days apart and were monitored for diabetes development. As shown in Fig. 2A, we observed a significant delay in diabetes onset in the anti-CD22/cal mAb–treated mice (n = 20 mice, 50% protected in the long term) compared with untreated controls (n = 30 mice, P < 0.01, 10% protected in the long term) and with the group treated with calicheamicin alone (GG5/cal) (n = 10 mice, P < 0.01) (Fig. 2A).

We also treated 10-week-old female NOD mice with the unconjugated anti-CD22 mAb that only partially depletes B-cells (n = 20) (Fig. 1F and I). Diabetes onset was slightly delayed compared with controls (P = 0.06; Fig. 2A). This indicates that complete B-cell depletion is required to induce long-term protection from diabetes and stable tolerance toward autoantigens.

We also examined the effect of B-cell depletion on T-cell phenotype in NOD mice. A significant increase in the percentage of CD4⁺CD25⁺FoxP3⁺ cells is evident at 35 weeks of age (but not at 15 weeks) in the pancreatic lymph nodes of normoglycemic treated mice compared with both untreated hyperglycemic and normoglycemic untreated control NOD mice (anti-CD22/cal mAb treated, 20.3 ± 3.1 vs. normoglycemic control, 8.1 ± 0.6, P = 0.02, and vs. hyperglycemic control, 7.7 ± 3.1%, P = 0.009; Fig. 2B).

We sought to determine whether B-cell depletion can modify BDC2.5 peptide-driven interferon-γ (IFN-γ) production of T-cells, which can be considered an index of the T-cell anti-islet response (23). CD4⁺ T-cells extracted from splenocytes of normoglycemic anti-CD22/cal mAb–treated and normoglycemic control NOD mice were isolated at 15 and 35 weeks of age and were challenged with the BDC2.5 peptide and syngeneic dendritic cells in an ELISpot assay to evaluate IFN-γ production. Only at 35, but not at 15, weeks of age was the frequency of CD4⁺ T-cells extracted from normoglycemic treated animals responding to autoantigen significantly were reduced compared with responding CD4⁺ T-cells of normoglycemic 10-week-old and hyperglycemic untreated NOD mice (Fig. 2C). Interestingly, CD4⁺ T-cells extracted from splenocytes obtained from normoglycemic anti-CD22/cal mAb–treated NOD mice at 15 and 35 weeks of age are capable of mounting an immune response to alloantigen similar to the response by CD4⁺ T-cells extracted from normo- or hyperglycemic untreated control NOD mice (at 35 weeks: anti-CD22/cal mAb–treated, 15,210 ± 5,524 vs. normoglycemic control, 11,863 ± 2,470, NS, and vs. hyperglycemic control, 12,389 ± 897 [³H]thymidine incorporation counts/min, NS), indicating that the CD4⁺ T-cells are immunocompetent.

We then tracked the effect of B-cell depletion on survival and proliferation of autoreactive CD4⁺ T-cells in vivo (24–26). NOD.SCID mice were reconstituted with splenocytes from normoglycemic 10-week-old NOD mice. After 7 days (thereby allowing reconstitution of the immune system), mice were either treated with anti-CD22/cal mAb or were left untreated. After another 7 days, to allow ample time for B-cell depletion, isolated BDC2.5 T-cell receptor (TCR) Tg⁺ CD4⁺ cells extracted from splenocytes were transferred into B-cell–depleted or untreated NOD.SCID mice. After 72 h, mice were killed and examined for autoreactive CD4⁺ cell frequency in the spleen of recipients (easily tracked using the anti-ideotypic antibody against the Vβ4 chain of the TCR) (26). Interestingly, when B-cells are absent, fewer BDC2.5 TCR Tg⁺ CD4⁺ cells can be recovered from the host (reduction of 50%) (Fig. 2D and E, bottom quadrant) compared with the anti-CD22/cal mAb–treated NOD mice (Fig. 2D and E, top quadrant).

In the anti-CD22/cal mAb–treated NOD mice at 15 weeks of age, infiltrates were reduced compared with untreated control NOD mice and baseline untreated 10-week-old normoglycemic NOD mice (Fig. 3A1, B1, and C1). As expected, B220⁺ cells were very few in the anti-CD22/cal mAb–treated NOD mice but not in the control and in the baseline group (Fig. 3A2, B2, and C2). Surprisingly, very few CD3⁺ cells were apparent in the anti-CD22/cal mAb–treated NOD mice but not in the control and in the baseline group (Fig. 3A3, B3, and C3). Insulin (Fig. 3A5, B5, and C5) and glucagon (Fig. 3A6, B6, and C6) staining showed well-preserved islets in all the three groups. An increase in FoxP3⁺ cells was evident within the islets of the anti-CD22/cal mAb–treated group compared with the baseline and the untreated control group (Fig. 3A4, B4, and C4).

At 35 weeks of age, despite the complete recovery of the B-cell pool in the originally anti-CD22/cal mAb–treated group, pancreatic islets appeared to contain much less infiltrate than untreated control hyperglycemic NOD mice (Fig. 3D1 and E1). Again, neither the B220⁺ nor the CD3⁺ cells infiltrated the islets but instead remained at the islet border in the anti-CD22/cal mAb–treated group but not in the control group (Fig. 3D2, E2, D3, and E3). Insulin and glucagon staining confirmed the presence of many well-preserved islets in the anti-CD22/cal mAb–treated but not in the control group (Fig. 3D5, E5, D6, and E6). FoxP3 staining of islet infiltrate revealed reduced FoxP3 expression, particularly when compared with the massive presence of T-cells, in the untreated control but not in the anti-CD22/cal mAb–treated NOD mice (Fig. 3D4 and E4).

Finally, insulitis score revealed more well-preserved islets (0–50% of infiltration) in the anti-CD22/cal mAb–treated compared with untreated control NOD mice (Fig. 2G).

Newly hyperglycemic female NOD mice (defined on the basis of glucose levels >250 mg/dl for 3 consecutive days) were treated with a protocol identical to what is outlined above using the anti-CD22/cal mAb. A rapid reversal of hyperglycemia (within 2 days) was observed in all the B-cell–depleted NOD mice (10 of 10; Fig. 4A). Six of 10 remained normoglycemic in the long term (for >100 days). Three mice remained normoglycemic for 20–40 days and then reverted to hyperglycemia (Fig. 4A). One mouse remained normoglycemic for >50 days and then reverted to hyperglycemia (Fig. 4A). No consistent correlation was evident between glucose levels at baseline and the ability to restore or maintain normoglycemia after treatment (data not shown). None of the control NOD mice (n = 10) ever reverted from hyperglycemia spontaneously after 3 consecutive days of hyperglycemia (Fig. 4B). When hyperglycemic NOD mice were treated with anti-CD22/cal mAb after 5 days (n = 6) from hyperglycemia onset, a transient return to normoglycemia was evident in five of six NOD mice, which lasted for 20 days (Fig. 4B); subsequently, all mice then reverted to hyperglycemia. These data suggest that for treatment to be most effective, it must be initiated early after onset of hyperglycemia. We then treated newly hyperglycemic NOD mice with GC5/cal (n = 5), and no effect was observed in three of five mice treated, whereas in two NOD mice treated, some glycemic oscillations were observed before a return to stable hyperglycemia (Fig. 4B). Finally, B-cell depletion appeared to be mandatory for the restoration of normoglycemia. In fact, when B-cells composed >3–5% of the blood, normoglycemia was not restored. Only 2 of 10 partially depleted (with unconjugated CD22 treatment) hyperglycemic NOD mice showed a transient return to normoglycemia with a quick reappearance of hyperglycemia (Fig. 4B). Taken together, these data clearly indicate that optimal therapeutic efficacy for reversal of diabetes requires early initiation of therapy that effectively depletes B-cells in hyperglycemic animals.

Remarkably, a change in peripheral cytokine levels was observed during the restoration of normoglycemia (Fig. 4C). Most proinflammatory cytokines were reduced 10 days after treatment, when normoglycemia was restored (Fig. 4C). Particularly, 10 days after injection in the normoglycemic anti-CD22/cal mAb–treated mice, interleukin (IL)-17 levels were significantly reduced compared with hyperglycemic untreated controls (P = 0.01). Interestingly, although in the long-term hyperglycemic untreated control NOD mice, peripheral levels of tumor necrosis factor-α (TNF-α) (P = 0.06), IL-17 (P = 0.03), and IFN-γ (P = 0.03) were higher than in baseline hyperglycemic NOD mice, those normoglycemic NOD mice treated with anti-CD22/cal mAb showed peripheral levels of proinflammatory cytokines similar to newly hyperglycemic NOD mice (Fig. 4C).

The percentage of CD4⁺CD25⁺FoxP3⁺ cells was significantly increased in normoglycemic anti-CD22/cal mAb–treated long-term tolerant mice compared with hyperglycemic untreated control NOD mice, in both the pancreatic lymph nodes (anti-CD22/cal mAb–treated, 14.3 ± 1.9 vs. normoglycemic 10-week-old mice, 6.4 ± 0.7, P = 0.007, and vs. hyperglycemic mice, 8.7 ± 1.1%, P = 0.03; Fig. 4D) and in the spleen (anti-CD22/cal mAb–treated, 12.8 ± 1.7 vs. normoglycemic 10-week-old mice, 7.9 ± 0.4, P = 0.02, and vs. hyperglycemic mice, 8.4 ± 0.2%, P = 0.01; Fig. 4E).

Between normoglycemic anti-CD22/cal mAb–treated and hyperglycemic untreated control NOD mice, 100 days after injection, the effector-to-regulatory T-cell (Treg) ratio in the pancreatic lymph nodes and spleen was similar (data not shown).

After 3 days of hyperglycemia, islets are extensively infiltrated by lymphoid cells (Fig. 5A1) with disrupted structure, marked reduction in insulin staining, and a smaller reduction in glucagon staining (Fig. 5A5 and A6). The lymphoid infiltrate is composed predominately of B220⁺ cells with a smaller population of CD3⁺ cells (Fig. 5A2 and A3) and very few FoxP3⁺ cells (Fig. 5A4).

After 10 days, pancreas histology and immunohistochemistry show in untreated control NOD mice that all of the above features worsen (Fig. 5B1–B3, B5, and B6). Surprisingly, in the anti-CD22/cal mAb–treated NOD mice 10 days after treatment, islets showed very mild infiltrates confined to the borders of β-cells (Fig. 5C1) with an almost complete absence of B220⁺ and CD3⁺ cells (Fig. 5C2 and C3) and well-maintained and preserved insulin and glucagon staining (Fig. 5C5 and C6). In the anti-CD22/cal mAb–treated NOD mice 10 days after treatment but not in the controls, more infiltrating cells appeared to be FoxP3⁺ cells (Fig. 5B4 and C4).

In the anti-CD22/cal mAb–treated NOD mice, 100 days after treatment, two histological patterns were observed; islets still appeared almost completely free of infiltrates (Fig. 5D1) with very few B220⁺ and CD3⁺ cells inside islets (Fig. 5D2 and D3). Many small but well-preserved islets were present in the pancreas (Fig. 5D5 and D6) without a clear increase in FoxP3⁺ cells (Fig. 5D4). A smaller subset of islets showed an abundant infiltrate of B220⁺ and CD3⁺ B-cells (Fig. 5E1, E2, and E3), which remained, however, largely confined to the periphery of the islets. A high percentage of FoxP3⁺ Tregs (Fig. 5E4) and well-preserved islets (Fig. 5E5 and E6) were evident. Insulitis score confirmed that anti-CD22/cal mAb–treated NOD mice showed better preserved and less infiltrated islets compared with untreated control NOD mice (Fig. 4F).

We examined the gene expression profile of reemerging B-cells (obtained from normoglycemic NOD mice treated with anti-CD22/cal mAb 100 days after B-cell depletion) and compared it with that of B-cells obtained from naïve normoglycemic 10-week-old or hyperglycemic untreated control NOD mice. CD19⁺ cells were extracted from splenocytes with microbeads. Interestingly, a significant downregulation of inducible gene transcription was observed within the reemerging B-cell pool. Almost 200 genes were downregulated in reemerging B-cells compared with B-cells extracted from normoglycemic 10-week-old NOD mice (Fig. 6A), and 38 genes were downregulated in reemerging B-cells compared with B-cells extracted from hyperglycemic NOD mice (Fig. 6B).

When all three groups of B-cells were compared (naïve normoglycemic 10-week-old NOD mice, naïve hyperglycemic NOD mice, and reemerging), 21 genes appeared to be downregulated in the reemerging B-cell population (Table 1; Fig. 6C). It should be noted that the downregulation of many extracellular lytic enzyme products (elastase 1 and 2, lysozime, chymotrypsinogen B1, and amylase) may be associated with directed islet damage or a sustained proinflammatory effect (Table 1; Fig. 6C). Genes of the complement cascade (Fcna and C1qb) and proinflammatory (heme binding protein 1 [Hebp1], paired immunoglobin-like type 2 receptor β1 [Pilrb1], peroxisome proliferator–activated receptor-γ [PPARγ], and heme oxygenase-1 [Hmox-1]) genes are downregulated in reemerging B-cells as well. Hebp1 has been recently shown to be involved in monocyte chemotaxis (27); Pilrb1 is a receptor that can activate natural killer cells, dendritic cells, and monocytes (28); PPARγ has been related to dendritic cell/platelet activation/function (29); and Hmox-1 is an antioxidant gene (Table 1; Fig. 6C).

We then analyzed by fluorescence-activated cell sorting (FACS) the proportion of different B-cell subpopulations before B-cell depletion and after B-cell reconstitution with respect to expression of CD80, CD86, CD40, IgM, and major histocompatibility complex (MHC) class II and with respect to the presence of anergic B-cells; the latter can be identified as a small population of B220⁺CD93⁺CD23⁺IgM^lo cells (30).

No differences were detected in CD80, CD86, CD40, IgM, and MHC class II expression between naïve and reemerging B-cells obtained from splenocytes (too few B-cells can be recovered from pancreatic lymph nodes) of normoglycemic treated NOD mice (Fig. 7A). No differences were detected in the frequency of marginal zone B-cells or B-cell subpopulations as well (Fig. 7A). A small percentage of anergic B-cells is evident in naïve B-cells from either normo- or hyperglycemic untreated control NOD mice (Fig. 7B), whereas in the reemerging B-cell population obtained from normoglycemic treated NOD mice, an increase of B220⁺CD93⁺CD23⁺IgM^lo cells was detected, with a restoration of the original frequency found in naïve NOD mice, which was reduced in hyperglycemic NOD mice (Fig. 7B).

We also evaluated the functional ability of reemerging and naïve B-cells to present autoantigen to T-cells in vitro. We designed and optimized an in vitro assay in which B-cells are used as APCs and autoreactive BDC2.5 TCR Tg⁺ CD4⁺ cells are used as responders in the presence of the BDC2.5 peptide. Although B-cells from hyperglycemic untreated control NOD mice were capable of presenting autoantigen and stimulating IFN-γ production by CD4⁺ T-cells (Fig. 7C), reemerging B-cells obtained from normoglycemic treated NOD mice were defective in this capacity (Fig. 7C). In the same experiment, we evaluated the ability of reemerging B-cells to modulate cytokines profile of autoreactive BDC2.5 TCR Tg⁺ CD4⁺ cells; supernatant was therefore collected, and cytokines were evaluated with a Luminex assay. Interestingly, when reemerging B-cells but not naïve B-cells were used, BDC2.5 TCR Tg⁺ CD4⁺ cells downregulated their production of proinflammatory cytokines (Fig. 7F–I). Particularly, when reemerging B-cells obtained from normoglycemic treated NOD mice were used, BDC2.5 TCR Tg⁺ CD4⁺ cells produced less TNF-α compared with naïve B-cells extracted from normoglycemic and hyperglycemic untreated control NOD mice (P < 0.05 vs. both), less IL-17 compared with hyperglycemic untreated control NOD mice (P = 0.004), and less IFN-γ compared with hyperglycemic untreated control NOD mice (P = 0.02) (Fig. 7F–I).

To compare the regulatory functions of reemerging and naïve B-cells in vivo, diabetogenic CD4⁺ T-cells extracted with magnetic beads from splenocytes obtained from hyperglycemic NOD mice were adoptively transferred into NOD.SCID hosts. We then coadoptively transferred CD19⁺ cells extracted with magnetic beads either from splenocytes of normoglycemic anti-CD22/cal mAb–treated mice at 100 days (i.e., reemerging B-cells) or from untreated control NOD mice from our prevention studies at 35 weeks of age. Interestingly, when B-cells from untreated controls were transferred, NOD.SCID developed diabetes (particularly when naïve B-cells were extracted from hyperglycemic NOD mice) (Fig. 7D). Conversely, when reemerging B-cells were used, the onset of diabetes mediated by the transfer of CD4⁺ T-cells from hyperglycemic NOD mice was completely abrogated (Fig. 7D). To determine whether this protection is related to induction/expansion of Tregs in vivo, we analyzed the percentage of CD4⁺CD25⁺FoxP3⁺ cells (Tregs) in spleen of the NOD.SCID recipients of the diabetogenic CD4⁺ T-cells and reemerging B-cells or controls (B-cells from hyperglycemic animals or no cells) at day 30 after adoptive transfer. As seen in Fig. 7E, no differences were detected among the three groups. These data suggest that the reemerging B-cells may function to inhibit autoreactivity by a mechanism distinct from induction/expansion of Tregs, although further studies are required to define the exact mechanisms in vivo.

B-cell–depleting strategies are rapidly growing in popularity as a treatment approach for autoimmune diseases, thanks to the availability of an anti-CD20 mAb in humans. A recent paper from Yale University (8) showed that the human anti-CD20 mAb is capable of preventing autoimmune diabetes and reverses established diabetes in transgenic NOD mice expressing the human CD20 receptor on B-cells. Another recent paper showed that a murine anti-CD20 protects from diabetes onset when given at an early time point (4 weeks) and delays diabetes onset when given later on (15 weeks) (31). In this study, no data on reversal of diabetes were reported, and B-cell depletion was not complete (5% of B-cells were still found in NOD mice).

Our approach is novel for the following reasons: 1) We established a B-cell–depleting protocol in naïve NOD mice and not in transgenic NOD mice; 2) we targeted a new pathway, CD22, because CD22 expression is found on more mature B-cells and plasma cells and this may also have a positive effect on autoantibody production; 3) for the first time, the complete gene profiling of naïve NOD B-cells and reemerging B-cells has been identified; 4) our therapy not only depletes B-cells but also reprograms the entire pool of reemerging B-cells, generating a novel functionally impaired and regulatory B-cell population; and 5) anti-CD22 antibodies are available for human use as well, and this therapy is potentially different from an anti-CD20 approach in humans. Although anti-CD20 therapy can efficiently deplete B-cells in the blood, a number of reports suggest a potential incomplete depletion in lymphoid organs (32–35). This issue raises the question of whether pancreatic B-cells are depleted, and it is likely that the current ongoing trial will not clarify this issue due to the impossibility of performing pancreatic biopsies. On the contrary, anti-CD22 therapy seems to be more comprehensive regarding B-cell depletion (36).

We made use of a novel agent targeting CD22 (13) to develop and study a B-cell–depleting approach in naïve NOD mice as a model for human type 1 diabetes. Anti-CD22 treatment has been tested in humans, both for the immunoregulatory properties of CD22 engagement and for the possibility of depleting mature B-cells, with promising results in the fields of autoimmune disease (37,38) and B-cell malignancies (39). Surprisingly, the effect of targeting CD22 has never been tested in diabetes.

Our data show that anti-CD22/cal mAb treatment is capable of delaying diabetes onset in pre-diabetic NOD mice and, more importantly, of restoring normoglycemia in new-onset hyperglycemic NOD mice. In our experiments, complete B-cell depletion is required for restoration of normoglycemia; this is relevant from a clinical point of view, given that inefficient B-cell depletion after anti-CD20 mAb therapy (Rituximab) is a well-recognized phenomenon that can result in poor clinical outcome (32). Our data indicate that the absence of B-cells increases Tregs and reduces autoreactive T-cell proliferation, highlighting the importance of a persistent interaction of B-cells and autoreactive T-cells in maintaining the autoimmune response.

More than 150 treatments are capable of delaying diabetes in NOD mice (40–42); however, only anti-CD3 (43) and a few other strategies were found to be capable of restoring normoglycemia in NOD mice (44–46). Thus far, only the anti-CD3 regimen has been translated into clinical use in humans (47). Regarding our reversal studies, it is surprising how rapidly B-cell depletion rids islets of cellular infiltrates, because even treatment with anti-CD3 mAb appeared to be slower than B-cell depletion in restoring normoglycemia, and 20% of anti-CD3 treated animals did not revert from hyperglycemia (48). In our studies, 100% of hyperglycemic NOD mice reverted to normoglycemia within 2–3 days.

Our studies also show for the very first time that reemerging B-cells in NOD mice display a different phenotype confirmed by our transcriptome analysis, are functionally impaired in their ability to present antigen, and can regulate the autoimmune response, resulting in long-term tolerance to autoantigens in vivo. Our adoptive transfer studies show that reemerging B-cells can abrogate the transfer of diabetes in NOD.SCID by diabetogenic CD4⁺ T-cells.

In conclusion, we have shown for the very first time that anti-CD22 immunotherapy can deplete and reprogram B-cells, thereby reversing autoimmune diabetes in naïve NOD mice. Our study provides valuable knowledge to develop an approach using anti-CD22 in patients affected by type 1 diabetes.

P.F. is the recipient of an American Society of Transplantation (AST)–Juvenile Diabetes Research Foundation (JDRF) Faculty Grant and a JDRF Career Development Award. A.V. is the recipient of an AST-JDRF fellowship grant. R.A. is the recipient of a JDRF Regular Grant. M.H.S. is the recipient National Institutes of Health (NIH) Grant PO1-AI-41521, JDRF Grant 4-2007-1065, JDRF R&D Grant 4-2007-1065, and support from the Scott and Heidi Schuster Foundation. Microarray studies were performed by the Molecular Genetics Core Facility at Children's Hospital Boston supported by NIH Grants P50-NS-40828 and P30-HD-18655.