Peripheral blood mononuclear cells (PBMCs) were collected from pancreas transplant recipients admitted for pancreas transplantation and at the time of pancreas graft biopsies. Biopsies were performed either for-cause or for surveillance. Blood collection was performed prior to the transplant or the biopsy procedure. For this study we conducted a retrospective analysis using the stored patient samples. Collection and use of patient blood samples for the current study was approved by local ethical IRB board (HCB_2016_0479) and was conducted in full adherence to the Declaration of Helsinki. All patients provided written informed consent to participate in the study.

Between January 2017 and December 2019 a total of 108 pancreas graft biopsies were performed in our center. We excluded all samples in which a biopsy-matched blood sample was not obtained (n = 17), and those in which graft biopsy could not be performed or sample was not suitable for histological diagnosis (n = 24). During the same period a total of 15 simultaneous pancreas-kidney transplants were performed, from which PBMCs were available both at 1) the day of transplantation; and 2) pancreas graft biopsy performed within the first 3 months (either for surveillance or for cause, whichever was first).

For cause biopsies were indicated according to hospital´s protocol–1) >3xs increase in serum amylase or lipase; 2) hyperglycemia (fasting blood glucose >120 mg/dl); 3) de novo donor-specific antibodies (DSA); or 4) de novo anti-glutamic acid decarboxylase antibodies (GAD). Surveillance biopsies were performed according to center protocol at 3 weeks and at 12 months after transplantation, or as surveillance 4 weeks following the completion of the treatment for an acute rejection episode. Samples were obtained by ultrasound-guided percutaneous needle punch. Histological and immunohistochemical evaluation of pancreas graft biopsies was performed according to the 2011 Banff criteria (36, 37).

Blood samples were obtained contemporaneously to pancreas graft biopsy and used to measure glucose (mg/dl), amylase (U/L), lipase (U/L), creatinine (mg/dl), C-Peptide (ng/ml), HbA1C (%), and anti-GAD (U/ml). Serum samples at time of biopsy were screened for HLA class I and II antibodies using the Lifecodes LifeScreen Deluxe flow bead assay (Immucor, Stamford, CT, United States). Antibody specificities, including the presence of DSA, were determined using the Lifecodes Single Antigen bead assay (Immucor, Stamford, CT, United States) in patients with positive screening for HLA antibodies.

Blood samples were collected in two separate EDTA tubes and processed separately. In one red blood cells (RBCs) were removed from whole blood samples with RBC lysis buffer (Invitrogen™) and cells were resuspended at a concentration of 10⁶/ml in complete MACS buffer to determine a broad spectrum of leukocytes. In the other, PBMCs were isolated from whole blood samples by standard Ficoll density gradient (Ficoll-Paque premium, GE healthcare Bio-Science AB) and were resuspended at a concentration of 10⁶/ml in complete MACS buffer to determine T and B cell subsets. Six different panels were designed aiming at interrogating the immune cells for markers of cell activity, memory, and differentiation, with focus on T and B cells. The gating of T cell subsets and B cell subsets are defined in Supplementary Figure S1. Cell surface markers were stained with antibodies indicated in Supplementary Table S1, and used according to the instructions of the manufacturer. Except for the leukocyte panel, Aqua Live/Dead fixable dead cell kit (Thermo Fisher Scientific, Waltham, MA, United States) was used unambiguously to remove dead cells. Flow cytometry analysis was performed on a FACS Canto II (BD Biosciences, Heidelberg, Germany). Data were analyzed using FlowJo software (Tree Star, Ashland, OR, United States).

Induction therapy was used in all patients with rabbit anti-human lymphocytes polyclonal antibodies (Thymoglobulin 1.25 mg/kg/d for 4 consecutive days), and maintenance immunosuppression protocol with tacrolimus, mycophenolate, and prednisone. Prednisone withdrawal was attempted between months 3–6 in all patients with low immunological risk, absence of acute rejection episodes during the first 90 days, and good tolerance to mycophenolate treatment doses.

Comparisons of median measurements were performed using Mann-Whitney U test and p value < 0.05 was considered statistically significant. Kaplan–Meier was used to estimate unadjusted patient, graft, and rejection-free survivals and compared using log-rank test. Binominal logistic regression was used to calculate odds ratio, and Cox proportional regression performed to estimate grafts’ hazards. Statistical analysis was performed using SPSS (version 22, IBM, United States) software, with all tests 2-tailed and significance considered if p < 0.05.

A total of 67 biopsy-matched (biopsy-proven acute rejection [BPAR] n = 28; No rejection n = 39) PBMCs samples were performed during the period analyzed. Most biopsies were performed per protocol (52%) or surveillance following treatment of an acute rejection (AR) episode (23%), and performed for-cause in 18 cases (25%), with a median time from transplant to biopsy of 11.9 months [IQR 0.9–13.8]. At time of biopsy 18 patients (28%) presented DSA. Patients with a BPAR had higher lipase (p < 0.001) and glucose levels (p = 0.02), without differences in serum creatinine nor amylase, compared to those without AR. The demographics and immunological parameters at time of biopsy are described in detail in Table 1.

Phenotypic characterization of immune cells by flow cytometry showed a significant increase of T cells (CD3⁺CD19⁻CD56⁻) (p = 0.0187) in BPAR compared to those without AR (Figures 1A,B), including CD8⁺ (CD3⁺CD8⁺CD4⁻; p = 0.007) and CD4⁺ (CD3⁺CD4⁺CD8⁻; p = 0.0459) T cell lineages (Figures 1C,D). The only T cell subsets significantly increased in patients with BPAR were CD8⁺ naïve and central memory (Figure 2A), and CD4⁺ naïve (Figure 2B). No other major differences within the CD8⁺ or CD4⁺ lineages were observed between groups, neither TCRαβ⁺ nor TCRγδ⁺ (p > 0.05). The percentage of B cells (CD19⁺CD3⁻) was higher in BPAR group compared to biopsies without signs of rejection (p = 0.005) (Figure 1E). A deeper analysis of B cell subsets using anti-CD27 and anti-IgD antibodies revealed that both naive and classical memory B cells were increased in patients with BPAR (Figure 2C; Supplementary Figure S1B). In addition, the percentages of NK (CD3⁻CD56⁺), of NKT (CD3⁺CD56⁺) and monocytes (CD14⁺) were not different between those with or without BPAR (Figures 1F,G).

Having identified that both CD8⁺ and CD4⁺ T cells, as well as B cells, were increased at the time of acute rejection, we explored their ability to classify between those with and without BPAR. In a receiver operating curve analysis, both CD4⁺ (AUC 0.66 [95% CI 0.53–0.80]; p = 0.027), CD8⁺ (AUC 0.68 [95% CI 0.54–0.82]; p = 0.017), and B cells (AUC 0.69 [95% CI 0.55–0.83]; p = 0.012) presented a poor discriminative capacity. Most relevant, lipase outperformed any of the cell markers (AUC 0.72 [95% CI 0.58–0.85]; p = 0.004). We then evaluated whether timing post-transplant could influence the correlation between immune cell subtypes and acute rejection episodes. We observed that in biopsies performed in the early post-transplant period (<90 days) CD3⁺ T cells were increased in patients with T cell mediated rejection compared to those without rejection (Figure 3A), with a tendency towards an increase also in patients with indeterminate for rejection. Within this period T cells (AUC 0.80 [95% CI 0.61–1.0]; p = 0.012), both CD4⁺ (AUC 0.79 [95% CI 0.59–0.99]; p = 0.014) and CD8⁺ (AUC 0.80 [95% CI 0.62–0.99]; p = 0.012), and B cells (AUC 0.86 [95% CI 0.69–1.0]; p = 0.003) outperformed lipase for the diagnosis of acute rejection (AUC 0.62 [95% CI 0.38–0.86]; p = 0.12) (Figure 3B).

Since patients included in the analysis were recipients of simultaneous kidney-pancreas transplantation, we assessed whether there were concomitant kidney graft rejections at the time points evaluated, either concordant with pancreas graft rejection in BPAR group, or discordant in the no rejection group. Since per hospital policy simultaneous biopsies to both organs are not performed, we observed that in only one case there was a concomitant (on consecutive days) kidney graft biopsy, which was concordant with severe T cell mediated rejection - grade IIA in the kidney and grade 3 in the pancreas graft. Though serum creatinine is not a sensitive marker for kidney graft rejection, in particular subclinical rejection, we explored whether there were unperceived differences between groups. Both groups presented similar serum creatinine (BPAR 1.14 ± 0.71 vs. no AR 1.40 ± 0.37 mg/dl) at time of pancreas graft biopsy (Table 1). Moreover, no differences between groups were observed when patients were stratified by indication for and histological classification of pancreas graft biopsy (for cause vs. surveillance; BPAR vs. no rejection; p > 0.05).

Having identified that the discriminative ability of T cells in peripheral blood was particularly pronounced in the early post-transplant period, we explored whether these could be associated with an increased relative number prior to transplantation, or a resistance to T-cell depleting therapy. To do so we performed the immune profile from 15 patients who received a first simultaneous pancreas-kidney transplantation (SPK) and had a biopsy performed during the first 3 months after transplantation. Five had BPAR (3M-AR), whereas the remaining 10 had normal graft biopsies (No-AR).

At the time of transplantation, 3M-AR patients showed a significant increase of T cell populations (CD3⁺, CD8⁺ and CD4⁺; p < 0.005) (Figure 4A). However, the percentage of B and NK cells were comparable in both groups (Figure 4A). More detailed analyzes on T and B cell subsets using surface markers revealed that 3M-AR patients had higher levels of TCRαβ⁺ naive T cells, either CD4⁺ and CD8⁺ cells, and memory CD4⁺ T cells compared to patients without rejection (Figure 5).

To evaluate the response to T-cell depleting therapy, thymoglobulin, we firstly analyzed the total lymphocyte depletion in peripheral blood during the first days after transplantation. The reduction and subsequent lymphocyte recovery was similar between both groups (Figure 4B) during the first 14 days after transplant. Nonetheless, at time of first biopsy (BPAR median 0.7 months, no rejection median 1.2 months; p = 0.12), the reduction in T cells, particularly in CD8⁺ T cells, was higher in those with BPAR compared to no rejection group (Figures 4C,D).

We then explored whether immune profiling at pancreas transplantation (D0) could correlate with the risk of acute rejection early after transplantation. Having identified that in peripheral blood both CD8⁺ and CD4⁺ T cells, as well as B cells, were increased during early acute rejection episodes, we explored their ability to classify those at risk for early acute rejection prior to transplantation. In a receiver operating curve analysis, both CD4⁺ (AUC 0.94 [95% CI 0.82–1.0]; p = 0.007) and CD8⁺ (AUC 0.88 [95% CI 0.70–1.0]; p = 0.020) presented good discriminative capacity, whereas B cells failed (AUC 0.70 [95% CI 0.42–0.98]; p = 0.221) (Figure 6A). At cut-off of 6%, CD3⁺ T cells presented a sensitivity of 100% and a specificity of 70% for the diagnosis of acute rejection. We then stratified patients according to CD3⁺ T cells at time of transplantation. Those with CD3⁺ T cells >6% presented an inferior rejection-free graft survival (at 3 months 43% vs. 100% in those with CD3⁺ =<6%; Log-rank p = 0.037), and a 15 times superior risk for an acute rejection during the first year (HR 14.9 [95% CI 2.4–92.4]; p = 0.04) (Figure 6B).