This clinical trial has been registered at http://clinicaltrials.gov (identifier: NCT00315133) and conducted in accordance with good clinical practice guidelines and approved by the institutional review board of the Ribeirão Preto Medical School of the University of São Paulo, Brazil (Comitê de Ética em Pesquisa do HCFMRP-USP, No 10095/2002). Written informed consent was received from each participant, in accordance with the Declaration of Helsinki, prior to inclusion in the study. A data safety monitoring board and an independent medical monitor provided additional study oversight.

Twenty-five newly diagnosed type 1 diabetes patients fulfilled inclusion criteria and agreed to participate in this phase 2, single-arm, open-label AHSCT clinical trial and were prospectively followed-up for mean (SD) of 67.5 (26.9) months (19, 20) (Figure 1). Data from the first 23 patients have been previously described, as well as detailed inclusion and exclusion criteria (19, 20). Here, we report extended follow-up, including two additional patients.

Patients with less than 6 weeks of clinically diagnosed T1D and positive anti-glutamic acid decarboxylase (GAD) antibody levels had their HSC mobilized with 2 g/m² cyclophosphamide and G-CSF and harvested through leukapheresis. Subsequently, patients were treated with 200 mg/kg cyclophosphamide plus 4.5 mg/kg rabbit anti-thymocyte globulin (rATG, Genzyme, USA), followed by infusion of unmanipulated, cryopreserved autologous HSC. Insulin doses were gradually reduced according to glycemic levels, aiming for glycated hemoglobin A_1c (HbA_1c) of less than 7%, and fasting and 2 h after meal glucose levels of 120 and 140 mg/dL, respectively. Before and periodically after AHSCT, patients were evaluated, clinical data were registered, and blood samples were drawn for hematologic, metabolic, and immunological evaluations. Mean area under the curve (AUC) of C-peptide levels was measured during mixed-meal tolerance test, at baseline, 6 months and yearly after transplantation. From 25 patients, four were treated off-protocol and 21 were retrospectively analyzed for immunological parameters. The cohort of 21 patients was divided into upper and lower 50th percentile, based on duration of insulin freedom (Figure 1; Table 1).

Peripheral whole blood samples (150 µL) from 21 patients were immunophenotyped by flow cytometry, with previously titrated monoclonal antibodies (Becton-Dickinson, San Jose, CA, USA, EUA; Figures S1–S3 in Supplementary Material). Cells were analyzed with BD Biosciences FACSCalibur flow cytometer (Becton-Dickinson, San Jose, CA, USA, EUA). Gating strategies are shown (Table S1 in Supplementary Material). Thirty-thousand events per sample were acquired for each subset and 100,000 events for immunoregulatory T cells. Results were expressed as absolute cell numbers (cells per microliter).

Peripheral blood mononuclear cells (PBMCs) from 18 patients were isolated by Ficoll-Hypaque™ density gradient centrifugation (Amersham-Pharmacia, Uppsala, Sweden). DNA was purified using DNeasy Blood & Tissue kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s recommendations. DNA was diluted to concentration of 100 ng/μL and used for quantification of TREC molecules (25). Absolute quantification of TRECs (target gene) and albumin (endogenous gene control) in PBMCs was performed by real-time PCR, using TaqMan^® reagents and 7500 Real-Time PCR system (Applied Biosystems, Foster City, CA, USA). For each real-time PCR reaction, standard curves of TREC and endogenous genes were prepared. Plasmids containing TREC or albumin genes were kindly provided by Dr. Ngai Ka-Leung (Northwestern University, EUA). All experiments were performed in duplicates. The number of TREC molecules from each sample was divided by the number of endogenous gene copies (for normalization). Results were expressed as TREC molecules/100 ng DNA.

Peripheral blood mononuclear cells from 21 patients were isolated by Ficoll-Hypaque™ density gradient centrifugation (Amersham-Pharmacia, Uppsala, Sweden). RNA was purified using Pure Link Mini Kit (Invitrogen Life Technologies, Carlsbad, CA, USA). Complementary DNA (cDNA) was synthesized from RNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA), according to the manufacturer’s recommendations. For further TCRBV CDR3 Length Spectratyping, TCR βV-βC segments were first PCR amplified with specific primers for the 24 variable TCR regions (24 βV families) and with one specific primer for the constant region of the TCR (βC primer), as described elsewhere (26, 27). The product from the first reaction underwent another PCR amplification (elongation or run-off reaction) that included an internal fluorescent βC primer. The segment containing the CDR3 region of the TCR was amplified (26, 27). Run-off reaction products were submitted to electrophorectic run in an automated DNA sequencer (ABI 3500xL Genetic Analyzer, Applied Biosystems, Foster City, CA, USA). Data were analyzed by Gene Mapper™ software (Applied Biosystems, Foster City, CA, USA), and distribution graphs of CDR3 segments were generated (26, 27). Graphs represent fluorescence intensity in arbitrary units according to the size of the CDR3 region in base pairs. Considering that the positions of the βV and βC primers are fixed, the size of PCR product labeled βV-βC depends on the size of the VDJ junction of each specific TCR. The size of CDR3 segment with 10 amino acid residues for each Vβ has been previously estimated (26, 27).

We then calculated the diversity of the TCR repertoire by the complexity score (CS) (28). Each Vβ family was classified from 0 to 8, based on the degree of TCR complexity/diversity. A score of 0 was given if the Vβ family was absent, of 1 if the Vβ family presented only one CDR3 peak, of 2 if the family presented two peaks, and subsequently. The score 8 was given to a spectratype with 8 or more peaks. The overall CS was calculated by the sum of the CSs of each Vβ family (28).

Multimeric peptide–MHC (pMHC) complexes were prepared as previously described (29). Briefly, recombinant HLA-A2 and human β2M were solubilized and injected together with each synthetic peptide into a refolding buffer. Multimeric pMHC complexes were produced by addition of streptavidin-conjugated Qdots (Invitrogen, Breda, The Netherlands) to achieve a 1:20 streptavidin-Qdot/biotinylated pMHC ratio. Qdot-585, -605, -655, -705, and -800 were used. PBMCs from HLA-A2 positive subjects were stained with a mixture containing six HLA-A2-associated diabetes epitopes (Table 2). Cells (2 × 10⁶) were stained simultaneously with all Qdot-labeled pMHC multimers (0.1 µg of each). Subsequently, 10 µL allophycocyanin-labeled anti-CD8 and 10 µL fluorescein isothiocyanate-labeled anti-CD4, -CD14, -CD16, -CD20, and -CD40 antibodies (Becton Dickinson, Franklin Lakes, NJ, USA) were added. Cells were resuspended in PBS/0.5% BSA containing 7-aminoactinomycin-D (7-AAD; eBioscience, San Diego, CA, USA) to exclude dead cells and analyzed using the LSRII (Becton Dickinson).

Patients were retrospectively divided into two groups, based on duration of insulin independence. The median was chosen as cut point because it dissected the patients into two groups of equal size who presented sustained vs short-lived response to therapy. Patients above the median duration of insulin independence were named “long-remission group” and those below, “short-remission group.” Further data analyses were performed in each of the subgroups, as well as in the undivided cohort of patients. For each patient, in the respective periods, C-peptide levels were calculated by mean AUC, which was estimated by the trapezium rule. The linear regression mixed model was used to analyze data from AUC, immune reconstitution, and frequency of autoreactive CTL (29). This method allowed for multiple longitudinal observations per individual across a baseline period, and subsequent time points after transplantation. The linear mixed models are composed of random and fixed effects. This model applies for analysis of data on which responses are grouped (more than one measure to the same individual), and the assumption of independence between observations in the same group is not adequate. The fixed effects are groups and periods. The random effects are associated with patients, since it is necessary to control correlations between repeated measures. For variable frequency, we used a logarithmic transformation to fit the data to the proposed model. Kaplan–Meier method was used to calculate insulin-free survival curve. Data analysis was performed using SAS^®9.0 statistical software (SAS Institute Inc., Cary, NC, USA). Statistical significance was set at p < 0.05.

Twenty-five patients were enrolled between November 2003 and April 2010, and followed until 2014. Patient characteristics have been reported previously (19, 20) and are summarized (Table 1; Table S2 in Supplementary Material; Figure 1). Two patients had presented with ketoacidosis at diagnosis, and two others inadvertently received steroids during transplant procedure (Figure 1; Table S3 in Supplementary Material). These off-protocol patients were excluded from immunological analyses, allowing us to study a more homogeneous group of patients, with similar clinical and metabolic characteristics and also treated under the same transplant protocol. In fact, these four patients did not experience any period of insulin freedom after AHSCT. Details on this subpopulation of patients have been published (19, 20).

In the remaining patients, insulin doses were progressively reduced after AHSCT, based on glycemic levels, until complete suspension. Twenty-one patients became insulin free for median of 43 months (6–100 months) after AHSCT, and three of these patients are still continuously free from insulin (Table 1; Table S4 in Supplementary Material). Ten patients remained insulin free for less than 3.5 years after AHSCT and thus were named “short-remission group.” Eleven patients persisted insulin free for at least 3.5 years, therefore labeled “prolonged-remission group.” From 30 to 42 months of follow-up, there was no change in insulin use status, whether patients being were insulin free or insulin-dependent (Table S4 in Supplementary Material). This observation indicates the different profiles of each group, short-term and long-term responders, and reinforces the importance of investigating possible mechanistic differences between groups.

Mean (SD) AUC of C-peptide levels significantly increased from baseline compared with 6, 12, 24, 36, and 48 months post-AHSCT (Figures 2A,B; Table S5 in Supplementary Material). Furthermore, although at 60 and 72 months most patients had already resumed insulin, C-peptide levels were similar than those from before transplant. Moreover, when compared with baseline, the prolonged-remission group presented significant increase of C-peptide levels from 6 to 60 months after AHSCT, while the short-remission group showed stabilized C-peptide levels along same follow-up (Figure 2B; Figure S4 in Supplementary Material).

The profile of pancreas-infiltrating lymphocytes is partly mirrored in the peripheral blood of patients (29). Islet-autoreactive CTL have been detected in insulitic lesions from patients diagnosed with TID (30). These data indicate that measuring islet-specific CTL in peripheral blood may represent an excellent biomarker for treatment intervention. Thus, we first evaluated whether autoreactive islet-specific CD8⁺ T cell (aCTL) frequencies were associated with clinical outcomes of transplanted patients.

Cumulative frequencies of islet-specific CTL recognizing six different islet epitopes were determined over time (Figure 3; Table S6 in Supplementary Material). We did not detect significant changes in aCTL frequencies following transplantation (Figure 3A). Subsequently, cumulative frequencies of aCTLs were divided into upper and lower 50th percentile, clustering patients according to the CD8⁺ T-cell autoreactivity at baseline (Figure 3B; Table S7 in Supplementary Material). Initial frequencies only minimally changed after AHSCT, indicating that the conditioning regimen was not able to eliminate aCTLs (Figure 3C). Patients with higher CTL autoreactivity before AHSCT maintained significantly elevated cumulative frequencies of aCTLs up to 24 months after transplantation. Moreover, patients with low CTL autoreactivity demonstrated higher C-peptide levels after AHSCT (Figure 3D; Table S8 in Supplementary Material).

Further analyses of insulin-free patients indicated that those with higher cumulative aCTL frequencies at baseline relapsed earlier post-AHSCT than patients with lower autoreactivity (p = 0.007; Figure 3E). Indeed, all patients with lower islet autoreactivity at baseline remained free from insulin for at least 24 months after AHSCT, while 83% of patients with higher baseline levels of islet autoreactivity had already relapsed at this time point (Figure 3E). In summary, our data show that baseline cumulative frequency of aCTLs predicts duration of insulin independency after AHSCT in T1D patients.

We next evaluated the dynamics of aCTL frequencies in the short- and prolonged-remission groups after AHSCT. Those remaining insulin free for longer periods had persistently lower frequencies of aCTLs that differed significantly at 6 and 12 months posttransplant (Figure 3F).

Once AHSCT did not affect specific islet autoreactivity, we searched for other immunological mechanisms that might be involved in the control of β-cell destruction. Thymic production of naive T-cell populations (recent-thymic emigrants) was determined through analysis of signal-joint T-cell receptor excision circles. In our cohort of 21 patients, TREC levels were lower than baseline at 3 months, normalized at 12 months, and increased from 18 to 60 months after AHSCT (Figure 4A; Figure S5A and Table S9 in Supplementary Material). Concurrently, the T-cell repertoire diversity as defined by the CS was sustained until 24 months (Figure 4B; Figure S5B and Table S10 in Supplementary Material), progressively declining after 30 months post-AHSCT.

Lymphopenia was observed following transplantation in both groups, reflecting the immunosuppressive effect of the procedure (Figures S6A,B in Supplementary Material). We examined whether T- and B-cell subset reconstitution was associated with metabolic control of patients (Figure S6 in Supplementary Material). For the entire follow-up, CD3⁺CD4⁺ T-cell numbers remained lower than baseline in both groups (Figure S6C in Supplementary Material), whereas CD3⁺CD8⁺ T-cell levels did not change, resulting in a CD4/CD8 ratio inversion (Figures S6D,E in Supplementary Material). B cells reconstituted to baseline levels approximately 2–3 months post-AHSCT in both patient groups (Figure S6F in Supplementary Material).

We also investigated whether clinical response to AHSCT was associated with imbalanced distribution of memory T-cell subsets. In both patient groups, reconstitution to baseline numbers of central-memory CD4⁺ (CD4⁺T_CM) cells was not detected throughout follow-up (Figure 5A), while overall central-memory CD8⁺ (CD8⁺T_CM) cell counts increased at 2 and 3 months post-AHSCT, decreasing after 54 and 60 months (Figure 5B). The short-remission group had higher effector-memory CD4⁺ (CD4⁺T_EM) cell counts at 2–9 months posttransplantation when compared with the prolonged-remission group (Figure 5C), while the prolonged-remission group presented higher CD8⁺T_CM values at 30, 36, and 60 months posttransplantation than the short-remission group. In both groups, effector-memory CD8⁺ (CD8⁺T_EM) cell counts raised early after AHSCT (Figure 5D). In summary, memory CTL comprehended most of T cells detected on long-term follow-up of patients after AHSCT, indicating that the immunosuppressive regimen may not sufficiently target potentially autoreactive and pathogenic memory T cells.

Considering that total CD8 T cells were elevated after transplant, especially in those patients with longer clinical remission, we investigated the dynamics of CD8⁺CD28⁻CD57⁺ T cells post-AHSCT, due to their potential to decrease activity of autoreactive T cells (31). The prolonged-remission group showed increased numbers of CD8⁺CD28⁻CD57⁺ T cells at 6, 12, 30, and 42 months posttransplantation, compared with baseline levels, while the short-remission group showed elevated CD8⁺CD28⁻CD57⁺ cell counts only at 18 months post-AHSCT (Figure 6A). The percentage of CD8⁺CD28⁻CD57⁺ cells was increased from 6 to 42 months after transplantation (Figure 6C). Thus, during early years post-AHSCT, CD8⁺CD28⁻CD57⁺ and CD8⁺T_CM cell subsets constituted most of the T cells, especially in patients that presented better response to transplantation.

Next, we evaluated the reconstitution of CD4⁺CD25^hiFoxP3⁺ regulatory T cells (Tregs) after AHSCT (32). At baseline, the short-remission group had higher absolute Treg numbers than the prolonged-remission group. However, after transplantation, the percentage and absolute numbers of Tregs significantly increased in the prolonged-remission group only (Figures 6B,D). Altogether, these data show that expansion of immunoregulatory T cells is associated with higher metabolic responsiveness to AHSCT, possibly by temporary reestablishment of self-tolerance.