Loss of Intra-Islet CD20 Expression May Complicate Efficacy of B-Cell–Directed Type 1 Diabetes Therapies

NOD/LtDvs mice are maintained by sibling matings at The Jackson Laboratory. B-cell–deficient NOD.Igµ^null and totally lymphocyte-deficient NOD-scid mice have been previously described (14,29,30).

IgG1 and IgG2a isotypes of the B-cell–depleting mouse CD20-specific 18B12 monoclonal antibody have been described (31,32). Both antibodies were injected intraperitoneally at a 10 mg/kg body wt dose at indicated time points. Control mice received equivalent doses of the isotype-matched non–B-cell depleting 2B8 antibody. In one experiment, at the initiation of anti-CD20 or control antibody treatment, NOD mice also received 250 μg i.p. injections at 2-week intervals of the Treg-depleting PC61 CD25-specific antibody.

Splenic and/or pancreatic lymph node (PLN)-derived leukocytes were assessed for levels of various B-cell subsets by flow cytometry using FACSCalibur instrumentation (BD Biosciences, San Jose, CA) and FlowJo data analysis software (Tree Star Inc., Palo Alto, CA). Fluorochrome conjugated monoclonal antibodies specific for the CD45R/B220 (RA36B2), CD21/CD35 (7G6), CD23 (B3B4), and CD138 (281–2) cell surface molecules were obtained from BD Biosciences. Total B cells were identified based on B220 expression. Among total B cells, expression patterns of CD21 and CD23 defined the T1 (CD21⁻ CD23⁻), T2/pre-MZ (CD21hi CD23int), MZ (CD21hi CD23⁻), and FO (CD21int CD23hi) subsets. Plasma cells were identified by CD138 expression. Some studies used the IgG2a CD20-specific 18B12 antibody conjugated to allophycocyanin using the Alexa Fluor Protein Kit (Invitrogen, Carlsbad, CA). Other studies evaluated pancreatic islet–associated B-cell populations in NOD mice. Islets were isolated as previously described (33) and cultured overnight on an individual donor basis in previously described tissue culture medium (34), allowing for egress of associated leukocytes that were harvested for flow cytometric analyses of B-cell content and CD20 expression. Tregs were enumerated based on a CD4⁺CD25⁺FoxP3⁺ phenotype using the eBioscience (San Diego, CA) FoxP3 intracellular staining kit. Functional activity of magnetic bead–purified Tregs (CD4⁺ CD25⁺ phenotype) was evaluated by the previously described flow cytometric approach (35), assessing their ability to suppress anti-CD3–stimulated proliferation of carboxyfluorescein succinimidyl ester–labeled CD4⁺ CD25⁻ responder T cells. A previously described phycoerythrin-conjugated H2-A^g7 MHC class II tetramer supplied by Dr. Luc Teyton (Scripps Research Institute, La Jolla, CA) was used to identify CD4 T cells sharing antigenic specificity with the diabetogenic BDC2.5 clonotype (36).

Mice were monitored for glycosuria with Ames Diastix (Bayer Diagnostics Division, Elkhart, IN), with diabetes onset diagnosed as two consecutive values of ≥3.

Generation and measurement by enzyme-linked immunosorbent assay of antibodies produced in response to priming with hen egg lysozyme (HEL) were carried out by previously described methods (7–9).

IAAs were detected using the previously described radioimmunoassay methodology (37).

The IgG1 isotype of the 18B12 murine CD20-specific antibody rapidly depleted peripheral blood B cells in NOD mice, with minimal rebounding observed at day 21 posttreatment (Fig. 1). NOD mice could be maintained in a state where peripheral blood B cells remained virtually absent by continuous treatments with the IgG1 anti-CD20 reagent at 21-day intervals (Fig. 1). NOD mice receiving a second treatment at 14 days after the first with the IgG2a isotype of 18B12 all succumbed to an anaphylaxis response within 24 h. Hence, the IgG1 anti-murine CD20 reagent was used for all subsequent experiments.

There is only a short time frame after onset of overt hyperglycemia in which treatment with a Rituximab-like antibody (2H7) reportedly exerts diabetes reversal effects in a subset of NOD mice transgenically expressing human CD20 (18). Thus, we tested whether anti-CD20 treatment more effectively induces late-stage diabetes prevention effects when initiated in NOD mice that have already developed a significant level of pancreatic β-cell autoimmunity marked by the presence of IAAs. Starting at 10 weeks of age, NOD female mice were treated at 21-day intervals with the 18B12 mouse CD20-specific antibody. Other NOD females were treated at the same intervals with the isotype-matched 2B8 control antibody. Serum samples were obtained from all mice prior to the initiation of anti-CD20 or control antibody treatment and tested for IAAs. This allowed us to retrospectively determine if the efficacy of anti-CD20 treatment in blocking diabetes development potentially differed when initiated before or after IAA onset.

As expected, when initiated in pre-IAA onset NOD mice, anti-CD20 treatment significantly inhibited diabetes development (Fig. 2A). Among NOD mice treated with the irrelevant control antibody, the presence of IAAs marked a significantly increased propensity for progression to overt diabetes (100 vs. 64%) (Fig. 2A). Conversely, while anti-CD20 treatment may have a marginal capacity to inhibit diabetes development when initiated in already IAA-positive NOD mice, this effect did not achieve statistical significance (Fig. 2A). We cannot exclude the possibility that the minority subset of NOD mice typed as IAA negative at 10 weeks of age, and subsequently not protected from diabetes development by anti-CD20 treatment, was transiently positive at an earlier time point. However, overall, these results indicate B cells targetable by the 18B12 CD20-specific antibody appear to more prominently contribute to earlier rather than later stages of autoimmune diabetes development in NOD mice.

NOD B cells rebounding after transient depletion by anti-CD22 treatment are reportedly characterized by a reprogrammed gene expression profile inhibiting their diabetogenic capacity (17). Thus, we assessed if B cells rebounding in NOD mice receiving only one 18B12 anti-CD20 treatment also had suppressed diabetogenic activity. However, transiently depleting B cells by a single anti-CD20 treatment at 10 weeks of age in originally IAA-negative NOD females did not prevent subsequent diabetes development (Fig. 2B). Hence, B cells rebounding after anti-CD20–mediated transient depletion in NOD mice are not reprogrammed to a diabetes protective state.

A strain-specific and age-dependent expansion of the MZ B-cell subset has been hypothesized to be a diabetogenic component in NOD mice (22). MZ B cells are absent from peripheral blood that was monitored in the experiments described above. Thus, it was possible the poor ability of IAA-positive NOD mice to be protected from diabetes development by anti-CD20 treatment resulted from MZ B cells undergoing an age-dependent expansion, becoming refractive to deletion. To initially assess this possibility, we determined whether compared with the nonautoimmune-prone C57BL/6 (B6) strain, NOD mice housed at The Jackson Laboratory were also characterized by an age-associated expansion of splenic MZ B cells. A greater age- associated expansion of splenic MZ B cells was indeed observed in NOD than B6 mice (Fig. 3A). Conversely, mature FO subset B cells exhibited a greater age-associated expansion in B6 than NOD mice (Fig. 3B).

We next examined the effect of anti-CD20 treatment on MZ B cells in NOD mice at various ages. Starting at 5 weeks of age, NOD females were treated at 21-day intervals with the 18B12 CD20-specific or irrelevant control antibody. Splenic MZ B-cell numbers were assessed in a subset of mice in each group at various time points. At all ages analyzed, MZ B cells in NOD mice were resistant to anti-CD20–mediated deletion (Fig. 3C). Such deletion resistance was not due to an absence of CD20 expression by NOD MZ B cells (Supplementary Fig. 1). Thus, anatomical sequestration likely explains the resistance of MZ B cells to anti-CD20–mediated deletion. These collective results indicate the ability of the 18B12 CD20-specific antibody to inhibit early stages of diabetes development in NOD mice does not result from elimination of MZ B cells.

Unlike what was observed for the MZ subset, at all ages studied, splenic FO B-cell numbers were significantly lower in anti-CD20–treated than control NOD mice (Fig. 3D). Anti-CD20 treatment also significantly depleted the early developmental T1 and T2 subsets of immature B cells in NOD mice (Supplementary Fig. 2). FO B cells were also significantly deleted in the IAA-positive NOD mice depicted in Fig. 2A that were not protected from diabetes by anti-CD20 treatment (Supplementary Fig. 3). Hence, anti-CD20–mediated FO B-cell depletion strongly suppresses diabetes initiation mechanisms, but their absence does not significantly impair the progression of pancreatic β-cell autoimmunity established in IAA-positive NOD mice. As previously noted, Rituximab treatment also selectively depletes FO but not MZ subset B cells in humans {reviewed in Lund and Randall (20)}. Thus, use of the IgG1 18B12 anti-CD20 reagent in NOD mice robustly mimics the pattern of B-cell subset depletion elicited by Rituximab treatment in humans.

An IgG2c isotype of the MB20–11 CD20-specific antibody that inhibits diabetes development in NOD mice depletes both MZ and FO B cells (19). Thus, use of the IgG1 isotype of the 18B12 CD20-specific antibody that depletes FO, but not MZ, B cells now implicates the former subset as preferentially contributing to the initiation of diabetogenic autoimmunity in NOD mice. Furthermore, albeit at lower levels than in controls, NOD mice selectively depleted of FO B cells still generated a strong antibody response after priming with the exogenous antigen HEL (Fig. 3E). Thus, provided treatment could be initiated early enough, a further potentially important advantage of inhibiting diabetes development by selective depletion of the FO subset rather than total B cells would be induction of a less severe state of generalized immunosuppression.

The ability of anti-CD20 treatment to block early initiating but not late stages of diabetes development in NOD mice raised the question of what effects this agent may have on B cells present in pancreatic islet leukocytic infiltrates at various points of disease pathogenesis. We initially tested whether the proportion of B cells differed within islet-associated leukocytes from 7- or 13-week-old NOD female mice at, respectively, early and late stages of diabetes development. The level of B cells was significantly greater in insulitic infiltrates of 13- than 7-week-old untreated NOD female mice (Fig. 4A). We then assessed what effect anti-CD20 treatment initiated at early or late stages of diabetes development may have on proportions of B cells within insulitic infiltrates. NOD female mice were treated with anti-CD20 at 5 or 11 weeks of age, and levels of B cells in islet-associated leukocytes evaluated 2 weeks later. While both less than in age-matched controls, levels of B cells were significantly lower in islet-associated leukocytes from 7- than 13-week-old NOD female mice treated with anti-CD20 2 weeks earlier (Fig. 4B). It is interesting that islet-associated B cells in untreated NOD mice at both 13 (Fig. 4C) and 7 weeks of age (data not shown) developed a CD20-negative phenotype, in contrast to those from PLNs. The CD20-negative phenotype was associated with islet-derived B cells in untreated NOD mice converting to a CD138-positive plasma cell–like phenotype (Fig. 4D).

These collective results indicate that regardless of the stage of disease development when it is initiated, anti-CD20 treatment of NOD mice eliminates diabetogenic FO subset B cells only prior to their entry into pancreatic islets since subsequent to doing so, they downregulate CD20 expression while converting to a plasma cell–like phenotype. Plasma cells that have also been previously shown by others to convert to a CD20-negative phenotype (38) would by definition be the source of IAAs. Hence, the presence of IAAs would appear to mark the accumulation within pancreatic islets of a highly pathogenic level of diabetogenic B cells that have become refractive to anti-CD20–mediated deletion. However, while not assessed in the current study, anti-CD20 treatment could ultimately diminish IAA levels in NOD mice or humans by eliminating diabetogenic B cells before their conversion to plasma cells.

Diabetes resistance elicited in NOD mice treated with a global B-cell–depleting BAFF-blocking reagent reportedly results from an enhanced ability of residual remaining myeloid-type APCs to support a Treg expansion (28). Hence, we assessed whether preferential depletion of FO B cells in NOD mice also supported a Treg expansion and/or altered numbers of BDC2.5-like diabetogenic effector CD4 T cells. These possibilities were assessed in PLNs, given this is reportedly the site for pathogenic expansion of diabetogenic T cells (39). Totally B-cell–deficient NOD.Igµ^null mice served as an additional control.

Compared with controls, total CD4 T cells were significantly increased in the PLNs of NOD mice selectively depleted of FO B cells by anti-CD20 treatment (Fig. 5A). There was an even greater significant increase of total CD4 T cells within PLNs of totally B-cell–deficient NOD.Igµ^null mice (Fig. 5A). In contrast, numbers of BDC2.5-like diabetogenic CD4 T cells detected by tetramer staining did not differ in PLNs from NOD control or NOD.Igµ^null mice (Fig. 5B). Compared with controls, numbers of BDC2.5-like CD4 T cells were somewhat higher in PLNs of anti-CD20–treated NOD mice, but this difference achieved only marginal statistical significance (Fig. 5B). However, the ability to inhibit diabetes development in NOD mice by genetically ablating all B cells or selectively eliminating the FO subset was also associated with significantly increased numbers of phenotypic Tregs (CD4⁺ CD25⁺ FoxP3⁺) within PLNs (Fig. 5C). Splenic Treg numbers were not increased in either anti-CD20–treated NOD mice or the NOD.Igµ^null stock (data not shown). Thus, the intra-PLN expansion of Tregs in anti-CD20–treated NOD mice and the NOD.Igµ^null stock is most likely an antigen-dependent process.

While numerically increased, on a per cell basis, Treg functional activity in anti-CD20–treated NOD mice was equivalent to that in controls (Supplementary Fig. 4). We then tested if the elicited numerical increase in Tregs might contribute to the ability of anti-CD20 treatment to inhibit diabetes development when initiated in pre-IAA onset NOD mice. As expected, compared with controls, diabetes development was significantly reduced in NOD mice in which anti-CD20 treatment was initiated before IAA onset (Fig. 6). Unlike other T-cell types, Tregs constitutively express CD25 (40). It is significant that the ability of anti-CD20 treatment to inhibit diabetes development in pre-IAA onset NOD mice was abrogated by coinfusions with a depleting CD25-specific antibody to a disease rate indistinguishable from controls (Fig. 6). Hence, anti-CD25 coinfusions did not abrogate the diabetes protective effects of anti-CD20 treatment by eliciting alterations in pathogenic effector T-cell activity differing from in NOD control mice. Further supporting this conclusion is the finding that while Tregs were reduced, there was no loss in conventional CD4 and CD8 T cells in coanti-CD20/CD25–treated NOD mice (Supplementary Fig. 5). Instead, these results support a possible Treg-dependent component for the mechanism by which selective depletion of FO subset B cells inhibits diabetes development in pre-IAA onset NOD mice.

Our results indicate treatment with the murine CD20-specific 18B12 antibody that like the clinically used Rituximab reagent, selectively depletes the FO but not the MZ subset of B cells, strongly inhibits the early initiating events of autoimmune diabetes development in NOD mice. Preventing the early stages of diabetes development with an anti-CD20 reagent selectively depleting the FO subset, rather than total B cells, could have the important advantage of inducing a less severe state of generalized immunosuppression. However, individuals included in possible diabetes prevention trials are currently selected on the basis of markers of already established high levels of ongoing pancreatic β-cell autoimmunity, such as the presence of autoantibodies. Of potential importance in this regard, our current results indicate that when first initiated in NOD mice that are already IAA positive, the ability of selective FO B-cell depletion by anti-CD20 treatment to block progression to overt diabetes is largely lost. This likely results from B cells entering pancreatic islets of NOD mice rapidly converting to a CD20-negative plasma cell–like phenotype. If this is also true in humans, our results could have potential negative implications for use of the anti-CD20 Rituximab agent as a monotherapeutic clinical approach for either the late-stage prevention of diabetes or to reverse recent-onset disease.

Our current findings differ from a previous report that anti-CD20 treatment could reverse recent-onset diabetes in a subset of NOD mice (18). This previous study used a Rituximab-like antibody to target NOD B cells transgenically expressing human CD20 molecules. Thus, one potential explanation for the possible discrepancy between this previous study and our currently reported results is differential expression regulation of the transgene-encoded human versus endogenous murine CD20 molecules on NOD B cells entering pancreatic islets.

A strain-specific, age-dependent expansion of MZ B cells has been proposed to contribute to diabetes development in NOD mice (22). Our results suggest that while an age-dependent expansion of MZ B cells is an NOD strain characteristic, in the absence of the FO subset, this phenotype does not contribute to diabetes development. However, it remains possible that with increasing age, NOD mice shift from reliance on the FO to the MZ subset of B cells to support diabetogenic T-cell responses. Furthermore, agents such as BAFF inhibitors that eliminate MZ as well as FO B cells might have a stronger capacity than anti-CD20 treatment to prevent progression to overt diabetes after significant levels of pancreatic β-cell autoimmunity have already developed, but this prevention could come at the price of greater generalized immunosuppression. Such possibilities will be assessed in future studies.

It has been reported that in NOD mice, the ability of B cells to mediate the expansion of diabetogenic T cells may normally outpace the capacity of other APC subtypes to support Treg responses that functionally suppress such autoreactive effectors (28). This conclusion was based on a finding that diabetes protection elicited in NOD mice treated with a global B-cell–depleting BAFF-blocking agent was abrogated by anti-CD25–targeting of Tregs. The current study found selective depletion of FO B cells is sufficient to elicit a Treg expansion specifically within the PLNs of NOD mice. Furthermore, anti-CD25–treatment studies support the possibility that at least in part, the capacity of selective FO B-cell depletion to inhibit diabetes development when initiated in pre-IAA onset NOD mice is a Treg-dependent process. By possibly preventing the egress of pathogenic clonotypes into islets, the expansion of Tregs may account for why total CD4 T cells were also increased in the PLNs of FO B-cell–depleted NOD mice. This possibility is supported by a previous finding that pathogenic T cells were also increased in the PLNs of NOD mice protected from diabetes by the activation of NKT cells (41), reportedly to also involve a Treg induction component (42).

The current results also suggest a previously unconsidered, and potentially clinically relevant, means of diabetes intervention, this being a potential combination therapy that both limits the ability of B cells to mediate pathogenic T-cell responses and enhances the capacity of myeloid APC to support Treg activity. Such a combination therapy could entail using Rituximab to block the further entry of diabetogenic B cells into pancreatic islets in conjunction with granulocyte-colony stimulating factor (GCSF; Neupogen), another U.S. Food and Drug Administration–approved reagent that can increase recruitment of dendritic cells with a capacity to enhance Treg activity (43). Further support for this possibility includes the previous finding that while ineffective when used alone, cotreatment with GCSF significantly synergized the ability of antithymocyte globulin to reverse recent-onset diabetes in NOD mice (44). One potential concern regarding this previously described combination therapy is the generalized immunosuppression effects resulting from global antithymocyte globulin–mediated T-cell depletion. Thus, we propose a more benign yet still effective diabetes intervention may be achieved through use of GCSF in combination with Rituximab that selectively depletes FO B cells. However, regardless of their ultimate nature, B-cell targeting approaches continue to represent a potentially attractive means of diabetes intervention. Because of their potential significance as disease intervention targets, it will be important to continue gaining an increased understanding of the pathogenic contributions of various B-cell subsets to differing stages of autoimmune diabetes development. Our current results also indicate it will be important to determine if upon entry into pancreatic islets, diabetogenic B cells change expression levels of molecules that may be targets of agents hoped to attenuate such pathogenic effectors.

This work was supported by National Institutes of Health grants DK-46266 and DK-51090, Cancer Center Support Grant CA-34196, the Brehm Coalition, and by grants from the Juvenile Diabetes Research Foundation International. R.D. and M.R.K. were former employees of Biogen Idec, which supplied the CD20 antibody used in these studies. No other potential conflicts of interest relevant to this article were reported.

D.V.S. directed research and wrote the manuscript. H.D.C. and M.N. researched data. R.D. produced the CD20 antibody and contributed to discussion. M.R.K. and M.H. contributed to discussion. J.P.D. contributed to discussion. C.W. researched data. M.A.A. contributed to discussion and reviewed and edited the manuscript.