Differences in Baseline Lymphocyte Counts and Autoreactivity Are Associated With Differences in Outcome of Islet Cell Transplantation in Type 1 Diabetic Patients

Between September 2000 and January 2006, 35 nonuremic type 1 diabetic patients received an islet cell transplant under ATG induction therapy and maintenance immune suppression with mycophenolate mofetil (MMF) and tacrolimus. They were all C-peptide negative, had large within-subject variation of fasted glycemia (coefficient of variation of prebreakfast glycemia [CVfg] >25%), and one or more signs of diabetic lesions (hypoglycemic unawareness, microalbuminuria, or retinopathy). The first 24 patients had been included in a phase 1 graft-dose finding study and the last 11 patients in a protocol that aims to assess influence of tapering of tacrolimus after month 12. Graft survival with this immune-suppressive regimen was previously reported for the first 24 patients (17,18). Informed consent had been obtained from all candidate recipients before they were listed as such by the Eurotransplant Foundation. Selection for transplantation occurred on basis of listing date, bloodgroup compatibility with the available graft, and health status. At the time of transplantation, none presented symptoms of acute infectious disease or inflammation. Analysis for cytomegalovirus (PCR and serology) and hepatitis A, B, and C (serology) at baseline excluded active disease. Two patients tested positive for complement-binding HLA antibodies pretransplantation, two patients that discontinued immune suppression during the first 6 months and one patient that died from a cerebral hemoraghe at 18 weeks posttransplant. These five patients were excluded from the current analysis.

Islet cells were isolated and cultured according to standardized protocols (17,19,20). For all islet cell grafts used, a sample was taken before transplantation and analyzed for its insulin synthesis capacity in the absence and presence of glucose (10 mmol/l). They all fulfilled the set criteria for function (i.e., minimally 20 pmol insulin synthesized per 2 h per million β-cells). Preparations were analyzed for their cellular composition and combined to grafts before being infused into the portal vein using either a laparoscopic (n = 16) (21) or subcutaneous transhepatic approach (n = 14) (22). Donor and graft characteristics are listed in appendix 1 (available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-0160/DC1).

Induction therapy consisted of rabbit ATG (Fresenius HemoCare, Redmond, WA) initiated 1–4 days before transplantation. A first dose of 9 mg/kg body wt was followed by 3 mg/kg for 6 days; no injection was given on days with a T-cell count <50 cells/mm³. Maintenance immunesuppression consisted of MMF (2,000 mg/day Cellcept starting on the day of the first ATG injection; gift of Roche, Brussels, Belgium) and tacrolimus (Prograft, starting 1 day before the last ATG injection, Fujisawa). Tacrolimus troughlevels were maintained between 8 and 10 ng/ml during the first 3 months after an islet cell transplantation and 6–8 ng/ml thereafter. This MMF-tacrolimus treatment was continued when a second islet cell transplant was given, without additional antibody course, but with one injection of 500 mg methylprednisolone 3 h before transplantation. Anticoagulant therapy was started 1 day after infusion and consisted of acetylsalycilic acid 100 mg once daily in all patients and additional low–molecular weight heparin at preventive dosage in 18 subjects (0.3 ml Fraxiparin once daily). No heparin was administered during islet infusion.

Efficacy and safety criteria were examined weekly during the first 6 weeks posttransplant, every 2 weeks between posttransplant week 6 and 12, and monthly thereafter. The CVfg was assessed using home glucose monitoring. Plasma C-peptide (TRFIA; Perkin-Elmer, Turku, Finland) with corresponding glycemia, A1C levels (HPLC; Pharmacia Biotech, Upsala, Sweden), and autoantibodies [islet cell antibody, IA2A, GAD antibody, and I(A)A] were assayed in the central laboratory of the Belgian Diabetes Registry (23). There was also a central measurement of lymphocyte subsets CD3+, CD4+, CD8+, CD19+, and NK cells (CD3−CD16+CD56+) (EpicsXl flow cytometer; Beckman Coulter, Miami, FL). Total lymphocyte counts were determined by routine hematology in local centers. Baseline characteristics of graft recipients are shown in online appendix 1. The majority of patients (n = 24) were treated with a subcutaneous insulin pump for at least 2 months prior to transplantation. In one patient, insulin pump was started in the second week after first islet infusion. Insulin tapering was only considered in patients with plasma C-peptide values ≥1.0 ng/ml (at glycemia 120–220 mg/dl), CVfg <25%, and mean fasting glycemia <125 mg/dl and was started after month 2 at a rate of −2 IU every 3–5 days or faster when patients experienced hypoglycemic episodes (<70 mg/dl).

Baseline T-cell autoreactivity to islet cell antigens was assessed at the Leiden University Medical Center and data analyzed blinded from clinical outcome. Blood was drawn before the first ATG administration, and peripheral blood mononuclear cells were isolated and processed as described before (24). Briefly, 150,000 fresh peripheral blood mononuclear cells were cultured in triplicate in 96-well round-bottomed plates in Iscove's modified Dulbecco's medium with 2 mmol/l glutamine (Life Technologies, Paisley, Scotland) and 10% pooled human serum in presence of islet autoantigens IA-2 (10 μg/ml) or GAD65 (10 μg/ml), of interleukin-2 (35 units/ml), or of medium alone. After 5 days, ³H-thymidine (0.5 μCI/well) was added and its incorporation measured after 16 h. Data were expressed as a stimulation index (S_i) by comparison with the medium alone value. An S_i ≥3 for any of the two antigens was considered a sign of T-cell autoreactivity against an islet cell antigen. In three patients, cellular autoreactivitiy could not be assessed because autoantigens were not available for testing of baseline samples.

In one patient, a third islet cell preparation was injected at posttransplant week 60. During the laparoscopically guided infusion, a liver biopt was taken from a steatotic area on the liver surface. Immunohistochemistry was performed on semiconsecutive paraffin-embedded sections using a rabbit insulin antibody (raised by Dr C. Van Schravendijk, Vrije Universiteit Brussels, Brussels, Belgium) and monoclonal CD3 (NeoMarkers, Freemont, CA) and CD20 (DAKO, Glostrup, Denmark) antibodies. Human tonsils and pancreas was used as control. For epitope retrieval, sections were heated to 98°C with citrate buffer, pH 6.0.

All values are expressed as median and interquartile range (IQR), unless indicated otherwise. Baseline and postransplant characteristics were related with status of insulin independence at month 6 as well as β-cell graft function and glycemic variability during the first 6 months. To assess differences between subgroups, we used nonparametric Mann-Whitney U test for continuous data and Fisher's exact test for categorical data. Correlations between baseline characteristics and CVfg or mean C-peptide during the first 6 months after transplantation were assessed by calculating Pearson's correlation coefficient.

To determine independent predictor ability of the variables, we used forward stepwise binary logistic regression analysis for insulin independence and a stepwise linear regression model for both CVfg and mean C-peptide. The analysis was performed on 27 subjects. Three subjects could not be included because of missing data. In our multivariate model we included parameters with P value <0.05 in univariate analysis.

All analysis were performed using SPPS (version 16.0), and graphics were computed by GraphPad Prism (version 4.0). All reported P values are two sided, and P < 0.05 was considered significant.

All 30 recipients became C-peptide positive after transplantation; 1 of them returned to C-peptide negativity before posttransplant month 6. At posttransplant month 6, 15 patients were insulin independent while the other 15 were on lower-dose insulin therapy (Table 1). Both groups had similar A1C concentrations and fasting mean glucose levels (Table 1). However, insulin-independent recipients had significantly higher basal C-peptide levels and exhibited a lower variability of fasting glycemia (Table 1).

No differences between both patient groups were noticed in terms of graft characteristics (Table 2); respectively, 87 and 80% of their subjects had received at least 2 milion β-cells per kg body wt in the first graft.

Baseline recipient characteristics such as age, sex, body weight–BMI, duration of disease, autoantibody positivity, metabolic control, and insulin dose were also similar (Table 2); a tendency to higher A1C concentrations was noticed in patients who did not become insulin independent (P = 0.054). After first transplantation, 14 patients who would achieve insulin independence continued treatment with insulin pump for a median time of 15 weeks (IQR 10–18). Among patients not achieving insulin independence, seven were treated with insulin pump during the 6-month follow-up. In four patients, insulin pump was discontinued after a median of 20 weeks because of low daily insulin need and replaced by subcutanuous injections four times a day. Metabolic control during the first 2 months did not differ between groups with median A1C values of 5.0% (IQR 4.4–5.9) in insulin-independent subjects vs. 5.2% (IQR 4.7–5.9) in insulin-requiring patients (P = 0.48). The number of patients treated with low–molecular weight heparin after transplantation did not differ between groups.

On the other hand, a significant difference was observed in the baseline immune state (before ATG treatment), as expressed by the absolute number of lymphocytes, of CD3+ cells, and of CD19+ cells, with higher initial counts in the recipient group that would not become insulin independent (Table 2). Of four patients with a baseline lymphocyte count >3,000 cell/mm³ (range 3,005–3,455), none became insulin independent. The higher initial number of T-cells in the insulin-dependent group was associated with a higher number of CD8+ cells (547/mm³ [IQR 462–657] vs. 393/mm³ [305–503] in insulin-independent patients; P = 0.021) but not of CD4+ T-cells (850/mm³ [742–1,134] and 716/mm³ [570–936], respectively; P = 0.093) or of NK cells (184/mm³ [147–302] vs. 174/mm³ [144–271], respectively; P = 0.72). At the day of islet infusion, a similar tendency was noted (Table 2). B-cell and T-cell autoreactivity were not measured at the day of islet infusion.

At the time of a second infusion (Table 3) we observed slightly higher CD3+ counts in insulin-dependent patients. It is important to note that in these 17 patients we still observed a significant difference in lymphocyte subsets in samples taken before ATG administration. There were no differences in immune profile and reactivity between patients receiving one or two grafts.

Both groups also differed in their baseline T-cell autoreactivity. This in vitro test could be performed in 27 of 30 subjects listed in Table 2. In the group that did not become insulin independent, 12 of 14 patients scored positive for IA2 and/or GAD65, whereas this was only the case in 6 of 13 patients who would become insulin independent (P = 0.046; Table 2). The role of baseline T-cell autoreactivity remained confined to a subgroup of recipients receiving amounts of β-cells below the median, as we previously reported (18; online appendix 2). On the other hand, no difference was seen in baseline autoantibody status or in the presence of multiple autoantibodies prior to transplantation (Table 2). Of 19 patients positive for IA2 and/or GAD65 antibodies at baseline, 10 became insulin independent, whereas this was the case for 5 of 10 patients that were negative for these antibodies (P = 1.0 by Fisher's exact test). This was also the case for islet cell antibody positivity. The number of HLA mismatches did not differ between groups (data not shown).

The observed correlation between baseline immune status of the recipient and the achievement of insulin independence at posttransplant month 6 was further examined by multivariate analysis (Table 4). Baseline B-cell count and T-cell autoreactivity were found to be independently correlated with the ability to achieve insulin independence (odds ratio 0.989 [95% CI 0.979–0.999] and 0.101 [0.009–1.067], respectively). Of nine patients without T-cell autoreactivity at start, seven became insulin independent, whereas this was only the case for 6 of 18 patients that tested positively (Fig. 1 B). When these patients were further stratified according to their baseline B-cell count (i.e., under or above the 50th percentile [259 B-cells/mm³]), insulin independence was seen in seven of eight patients without baseline T-cell autoreactivity and a B-cell count <p50, while only 1 of 11 patients with T-cell autoreactiviy and a B-cell count >p50 became insulin independent (P = 0.001 by Fisher's exact test).

In univariate analysis, similar correlations were found with other clinical outcome parameters, such as the average coefficient of variation of prebreakfast glycemia (CVfg) and the mean plasma C-peptide levels during the first 6 months (Table 4). After multivariate analysis, baseline positivity for T-cell autoreactivity and total lymphocyte counts correlated positively with the values of CVfg and negatively with the mean C-peptide levels (Fig. 2,A). A correlation with leukocytes at the day of infusion was maintained after multivariate analysis for CVfg but not C-peptide. Recipients with baseline T-cell reactivity had significantly higher CVfg (24% [IQR 20–35]) and lower C-peptide (1.0 ng/ml [0.5–1.3]) than those without (18% [15–24], P = 0.014, and 1.9 ng/ml [1.3–2.2; P = 0.001], respectively) (Fig. 2,B). These differences became more pronounced when a further stratification was made according to presence or absence of a baseline total lymphocyte count >p50. Thus, a negative test for T-cell autoreactivity and a lymphocyte count <p50 was associated with a lower CVfg and a higher C-peptide, whereas a positive test and a count >p50 appeared to predispose for a high CVfg and a low C-peptide (Fig. 2 A and B). The number of β-cells transplanted per kilogram bodyweight in the first graft had a tendency to correlate with CVfg in univariate analysis (R = −0.34; P = 0.07) but not with C-peptide levels during the first 6 months posttransplant (R = 0.27; P = 0.15). This could not be retained when it was added to the multivariate model (P = 0.56 and P = 0.42, respectively). The total number of β-cells transplanted or the number of donors used did not affect CVfg nor C-peptide levels.

One week after start of the immune therapy, the number of CD3+ cells had markedly dropped in both groups, reaching similar low values that were maintained until posttransplant month 6 (Fig. 1,A). On the other hand, baseline CD19+ counts did not decrease during the first week, thus remaining higher in the patient group that did not become insulin independent (Fig. 1 A); at later time points, CD19+ counts had decreased to similar levels in both groups. Total ATG dose in insulin-independent patients was similar to that used in insulin-requiring patients (21.9 mg/kg body wt [IQR 20.0–24.7] and 23.6 mg/kg body wt [22.0–26.2], respectively, P = 0.11). Mean tacrolimus trough levels (9.2 ng/ml [8.6–9.5] and 9.2 ng/ml [7.4–9.7]) and MMF dose (2,000 mg/day [IQR 1,500–2,000] and 2,000 mg/day IQR [1,500–2,000]) during the first 6 months posttransplant were also comparable between both groups (P = 0.66 and P = 0.45, respectively). A similar observation was made for CVfg and mean C-peptide levels during the first 6 months (data not shown).

We were unable to demonstrate any correlations between posttransplant immune measures and clinical outcome in the first 6 months posttransplant (Table 2). Interestingly, we observed a clear difference between subjects showing T-cell reactivity posttransplant when these results were combined with baseline T-cell reactivity. In eight subjects where T-cell reactivity was present both at baseline and during the first 6 months, only one achieved insulin independence. C-peptide levels were low in these patients (mean 0.74 ng/ml) and CVfg high (mean 33%). However, among five recipients who only tested positive after transplantation, four achieved insulin independence (P = 0.03) with mean C-peptide levels of 1.9 ng/ml (P = 0.01) and low mean CVfg (20%; P = 0.02). Among 10 recipients with baseline T-cell autoreactivity who where negative during the first 6 months posttransplant, results were intermediate. Five reached insulin independence; C-peptide levels were 1.1 ng/ml with CVfg 24%. Highest C-peptide levels (1.7 ng/ml) and lowest CVfg (18%) were seen in four patients who never presented T-cell autoreativity. Three of them reached insulin independence.

A liver biopt was available from a patient who had received two intraportal islet cell injections >1 year earlier. Two islets were identified in a portal tract. Endocrine cell clusters were identified by their positivity for chromogranin, proinsulin, and insulin (Fig. 3). They were surrounded by CD20+ and CD3+ cells; there were only few CD8+ cells detected. The CD20+ cells were dominant in the examined sections. The patient had not become insulin independent, and his posttransplant month 6 mean plasma C-peptide was low (0.62 ng/ml) and CVfg high (32%). At baseline, he did present a cellular autoreactivity against IA2, a high total lymphocyte count (3,393/mm³) with a B-cell count (536/mm³) above the 50th percentile of the recipient population before treatment (259 B cells/mm³); despite >1 year of immune suppressive treatment, the mean B-cell count until the time of the biopsy (412/mm³) was still above this 50th percentile. Interestingly, we did not observe a dominant CD20+ infiltrate in a liverbiopsy done in the patient who died 18 weeks posttransplant of cerebral hemorrhage. This patient had lower numbers of total lymphocytes (1,776/mm³) and B-cells (188 cell/mm³) at baseline. Last circulating CD19+ cell count was 30 cells/mm³ and mean C-peptide level 2.2 ng/ml in the last 4 weeks before she died.

The outcome of islet cell transplantation in type 1 diabetic recipients is variable, ranging from rapid failure to return of insulin independence. This variation can be caused by differences in graft and in recipient characteristics, some of which can be anticipated such as an inadequate size and/or viability of the donor β-cell mass and some being more difficult to distinguish. We have reduced graft variability by standardizing the composition of cultured islet cell grafts (17,20). In a series of 30 C-peptide–negative recipients of such graft, we retrospectively searched for possible recipient characteristics that are associated with an inadequate outcome during the first 6 months posttransplantation. The present data show a correlation with the baseline immune status at the time of the implantation. A higher blood B-cell count was associated with an inability to induce an insulin-independent state, as was also the case for a baseline T-cell autoreactivity against islet antigens, both characteristics being independent variables after multivariate analysis. A higher total lymphocyte count and a baseline T-cell autoreactivity were correlated with lower plasma C-peptide levels and a higher variability in fasting blood glucose.

Of 11 patients with a baseline B-cell count >p50 and a T-cell autoreactivity, only 1 became insulin independent, while this occurred in 7 of 8 with a B-cell count <p50 and negativity for T-cell autoreactivty. In 1 of these 10 patients with elevated baseline immune status and inadequate graft outcome, a biopt of the implant during posttransplant year 2 showed massive B-cell accumulations in the vicinity of insulin-positive cell aggregates; CD3+ cells were also present but less abundant, at least in the examined sections. Throughout follow-up, this patient had maintained a mean B-cell count above the 50th percentile of the recipient population before treatment. The biologic or pathogenic significance of these observations is unclear. Interestingly, no such infiltrate was observed around the graft of a patient who died 18 weeks posttransplant with a well-functioning graft and low B-cell counts. In animal models of diabetes, B-cells appear implicated in the destruction of β-cells (25,,,–29). However, their role in the development of human disease is uncertain. That type 1 diabetes can occur in a patient with severe hereditary B-cell deficiency is seen as evidence that the disease can occur without participation of B-cells (30) but does not exclude their involvement in other patients. In islet cell transplantation, it is not yet clear whether B-cells play a role through production of antibodies and/or through antigen presentation. Islet cell autoantibodies have been correlated to recurrence of autoimmunity and clinical outcome (8,31,,–34), but this was not confirmed by others (18,20) or by the present study; we did not detect a correlation between the presence or titer of autoantibodies and the number of circulating B-cells. On the other hand, alloantibodies can mediate rejection of organ transplants and islet cell grafts have been found to induce alloantibody formation (35,–37). Alloantibodies after transplantation were detected in only two recipients in the current study, one of which became insulin independent. Besides being a source of antibodies, B-cells can also operate as antigen-presenting cells that activate auto- and/or alloimmunity (38,,–41). This mechanism has been proposed in rejecting kidneys with a B-cell infiltration (42,43). Whether B-cells can interfere with engraftment and/or early graft function of intraportal islet cell grafts by one of these mechanisms cannot be concluded from the present data. Intravenous injection of the donor cells can nevertheless be expected to immediately subject them to influences of circulating lymphocytes and more so to populations that are more abundant and/or rapidly activated by β-cell autoantigens. It is then also conceivable that an early local activation of B-cells may result in their sustained presence in the islet microenvironement as was noticed in the biopt. To replace speculation by evidence, an immune therapy protocol will be needed in which B-cell depletion is achieved just before and during the first posttansplant weeks. In a study in nonhuman primates, Liu et al. (44) demonstrated that B-cell–directed immunotherapy promotes long-term islet allograft survival in nonhuman primates that also received T-cell–depleting ATG antibodies and a maintenance dose of rapamycin. It would thus be interesting to examine whether addition of rituximab, which selectively targets CD20+ B-cells, to our ATG-based protocol increases the percent recipients that become insulin independent.

Islet culture is known to help survival in rat allograft by enriching islet isolates in endocrine cells (45), but this has not yet been demonstrated in humans. Furthermore, islet culture can be expected to reduce inflammatory and/or immune reactivity in the implant in rodents (46). We systematically used islet culture, and a comparative study between freshly isolated and cultured islet cells has not yet been possible. Therefore, we are unable to assess whether these factors may have influenced graft survival, but it cannot be excluded.

Our data confirm in a larger cohort the previously reported correlation between a baseline T-cell reactivity against islet cell antigens and the failure of the implant to induce an insulin-independent state (18). They extend this correlation to other signs of inadequate graft function such as low plasma C-peptide levels and a high variability in fasting glycemia. As we previously reported, this correlation is confined to recipients receiving lower numbers of β-cells. The role of autoreactive T-cells in β-cell destruction is well accepted (47,–49). It is conceivable that the ATG-mediated lymphopenia induces a proliferation of autoreactive memory T-cells (50), which appear present or more abundant at baseline in a number of patients and might therefore predispose to a more rapid or extensive immune destruction. Patients rejecting a pancreas transplant were also found to present memory CD4+ T-cells that were specific for the islet autoantigen GAD65 (51). Indeed, we were able to demonstrate in the current study that in those patients where baseline T-cell autoreactivity persisted in the first 6 months posttransplant, clinical outcome was very poor. Although recipients who lost baseline-detectable T-cell reactivity after transplantation had slightly better outcome, both are still in contrast with the well-functioning grafts in recipients where T-cell autoreactivity was only detectable after transplantation.

Our data demonstrate an important predictive role for immune measures taken at baseline, before initiation of immunosuppression, but not for samples from the day of islet infusion. However, we did not measure B-cell and T-cell autoreactivity at the day of infusion. We could therefore not fully assess the correlations between those parameters at the day of infusion and clinical outcome, although the available data point in the same direction. However, it is evident that B-cells are more abundant in insulin-dependent patients in the peritransplant period, which, we suggest, may impair graft function via mechanisms we mention above. In a severe lymphopenic environment such as during ATG treatment, assessment of T-cell reactivity is impossible because the frequency of circulating lymphocytes is too low (24). Yet, T-cell autoreactivity remains present at a later time point, demonstrating the insufficient irradication of autoreactive T-cells, which may be susceptible to proliferation by lymphopenia-induced homeostatic proliferation (50).

In conclusion, this study correlates circulating characteristics of the baseline immune state with the metabolic outcome of intraportal islet cell transplants in type 1 diabetic patients. A higher number of total and of B-cell and a T-cell reactivity against β-cell antigens are associated with a lower return to insulin independence, lower plasma C-peptide levels, and/or a higher variability in fasting glycemia; in the one case where a liver biopt was available (>1 year after the transplant) B-cells were predominantly present around the β-cell aggregates. These data support the rationale of targeting also B-cells during the induction phase of immune therapy.

This study was supported by the Juvenile Diabetes Research Foundation (grant 4/2005/1327). R.H. is a PhD Fellowship of the University Research Council (OZR) of the Brussels Free University-VUB. D.J.-T.-T. is a PhD Fellow and B.K. and C.M. are Senior Clinical Investigators of the Research Foundation Flanders (FWO-Vlaanderen). V.A.L.H. and B.O.R. are supported by the Dutch Diabetes Research Foundation.

No potential conflicts of interest relevant to this article were reported.

A part of this study was reported in abstract form at the 14th European Society for Organ Transplantation (ESOT) Congress, Paris, France, 30 August–2 September 2009

We thank Koen Verbeeck, Ursule Van de Velde, Hilde Morobe, and Marijke Carpentier for their excellent collaboration in organizing this project, and we acknowledge the loyal support of the Beta Cell Bank, the Belgian Diabetes Registry, the Diabetes Research Center of the Brussels Free University (VUB), the Clinical Biology Department of Universitair Ziekenhuis Brussels (Brussels Free university-VUB), the physicians and nurses of the patients included in this study, and the Eurotransplant Foundation and its transplant surgeons and coordinators.

Author contributions: D.G.P., B.K., C.M., and B.O.R. designed the study and wrote the study protocol. R.H., V.A.L.H., P.G., J.H.L.V., B.M. and D.J.-T.-T. implemented the study. R.H., V.A.L.H., L.K., and B.K. performed the statistical analyses. F.K.G. supervised the autoantibody and C-peptide studies. M.P.-M. performed histological analysis. C.M., P.G., R.H., and B.K. were responsible for patient recruitment and follow-up. All authors assisted with interpretation of results and manuscript preparation.