Distinct pathophysiological cytokine profiles for discrimination between autoimmune pancreatitis, chronic pancreatitis, and pancreatic ductal adenocarcinoma

Banked serum (drawn upon admission or prior to surgery) and pancreatic tissue samples (obtained during surgery) were collected between 2001 and 2009, in the Biobank of the European Pancreas Center, Department of Surgery, Heidelberg University Hospital, Heidelberg. All sera were stored at −80 °C. Age, pathologic diagnosis, and serological analyses at the time of sample acquisition were obtained for all groups. Prior to analysis, all pancreatic tissues were reviewed by a pathologist to confirm disease diagnosis (AIP-1, AIP-2, CP, and PDAC).

A multiplex assay for simultaneous quantitative determination of proteins in diverse and complex biofluids was applied to assess concentrations of the selected cytokines in serum and tissue samples. The analysis was performed using the Bio-Plex Pro Human Cytokine 17-Plex panel (Bio-Rad Laboratories GmbH, Munich, Germany), a kit composed of 17 determinants. The following cytokines were simultaneously detected and standard curve range and limit of detection (LOD) all expressed in pg/ml are shown in parenthesis: IL-1β (0.34–4.991; LOD 0.06), IL-2 (2.27–9.531; LOD 0.62), IL-4 (0.22–3.207; LOD 0.15), IL-5 (3.25–13.069; LOD 2.3), IL-6 (1.41–23.178; LOD 0.61), IL-7 (0.68–11.063; LOD 0.66), IL-8 (1.38–2.183; 0.95), IL-10 (3.39–5.521; LOD 0.22), IL-12 (2.35–38.575; LOD 0.29), IL-13 (0.36–5.908; LOD 0.19), IL-17 (2.00–32.716; LOD 0.68), G-CSF (1.73–28.267; LOD 1.47), GM-CSF (2.45–10.848; LOD 1.47), IFN-γ (2.57–14.845; LOD 0.96), MCP-1 (3.72–55.124; LOD 2.65), MIP-1β (1.134–4.543; LOD 1.06), and TNF-α (3.18–52.115; LOD 1.88). A series of calibrators was analyzed with the patient samples to convert the fluorescence ratio to international units per milliliter. Concentration of each analyte was obtained by interpolating fluorescence intensity to eight-point dilution standard curve supplied by the kit. Values that were outside the standard curve range were calculated by the Brendan Scientific weighted five parameter fit algorithm as best estimation of the analyte concentrations within the samples and were obtained by extrapolation beyond the dynamic range of the standard curve and calculated by the Bio-Plex™ software (Bio-Plex Manager version 6.1 BioRad). All serum samples were measured in triplicate.”

Multiplexing was performed according to the manufacturer’s instruction manual. Briefly, for cytokine determination in serum, magnetic beads coated with antibodies against the examined antigens were mixed with 200 μl of each 1 : 4 v/v diluted patient sample (50 µl serum and 150 µl dilution buffer, Bio-Rad) and then incubated for 30 min. After a wash cycle followed by the addition of a detection antibody, another 30 min incubation, and a second wash cycle, streptavidin-PE was added to the beads. With the third wash cycle the excess conjugate was removed, and the bead mixture was analyzed on a Bio-Rad Bio-Plex 200 system.

For cytokine determination in pancreatic tissue, 60–80 mg of frozen tissue samples was cut with a cryomicrotome into 9 μm-thick slices (Leica CM 3050 S Leica Biosystems, Nussloch, Germany), collected in 15 ml polypropylene tubes, and mixed with 500 µl of lysis buffer (Bio-Rad). Each tube was vigorously vortexed then shock-frozen in liquid nitrogen and stored overnight at −80 °C. The day after, using an ultrasonic homogenizer (SonoPuls mini20 Bandelin^®, Berlin, Germany), the suspensions were subjected to a 30 s sonication step on ice (ampl. 80%, 0.99 kJ) and subsequently centrifuged at 16,000×g for 10 min. Supernatants were collected and divided into aliquots, and the total protein concentration was determined using a Pierce BCA assay (Thermo, Rockford, IL, USA). For multiplexing, the supernatants were adjusted with a dilution buffer (Bio-Rad) to a total protein concentration of 600 μg/ml and analyzed using the Bio-Plex Pro Human Cytokine 17-Plex panel kit (Bio-Rad) according to the manufacturer’s protocol as described above.

Statistical analyses were performed using GraphPad Prism software (version 5; La Jolla, CA, USA) and IBM SPSS Statistics version 22 (IBM Corp, Armonk, NY, USA). As variables were not normally distributed, the nonparametric Mann–Whitney test was used to assess the significant differences in multiplexing assays of serum and tissue extracts. The quantitative variables are graphically presented as box-and-whisker plots. Values of p ≤ 0.05 were considered to be significant. All tests were used two-sided.

Receiver operating characteristic (ROC) curves were calculated to analyze the test performance of serum cytokines G-CSF and IL-7 for predicting AIP using SAS software (release 9.4, SAS Institute, Inc., Cary, NC, USA). Logistic regression analysis was performed to generate sensitivity, specificity, and area under curve (AUC) values. Youden’s J statistic was used to select the optimal predicted probability cut-off.

All the comparisons on different cytokines were executed twice. At first, evaluations were performed between all AIP vs PDAC and CP patients and the reference category was always AIP. The comparison of serum cytokine expression levels in the three different groups is summarized in Table 2. The statistical analysis of serum comparing AIP to PDAC revealed a significantly higher concentration in AIP patients for IL-1β (p = 0.0221), for IL-7 (p = 0.0003), for IL-13 (p = 0.0337), and for G-CSF (p = 0.0105). The G-CSF levels in AIP (median 14.23 pg/ml) were also significantly discriminatory between AIP and CP (1.47 pg/ml) (p = 0.0006).

Subsequently, comparisons were performed among the AIP-1, AIP-2, PDAC, and CP groups using AIP-1 and AIP-2 separately as reference categories. When AIP-1, AIP-2, CP, and PDAC were compared and AIP-1 was used as a reference category, significantly higher concentrations of IL-7 (p = 0.0012), IL-13 (p = 0.0162), and G-CSF (p = 0.0425) were found in AIP-1 compared to PDAC. Also, a significantly higher level of G-CSF (p = 0.0039) in AIP-1 compared to CP was noticed. When we compared AIP-2 with PDAC, we found significantly higher levels of IL-1β (p = 0.0217), IL-7 (p = 0.005), and G-CSF (p = 0.032) in AIP-2. Comparing AIP-2 with CP, significantly higher levels in AIP-2 were found for IL-6 (p = 0.0361), for IL-17 (p = 0.0377), and G-CSF (p = 0.0034). The range of cytokine distribution in the analyzed groups is presented in Fig. 1a–i and the median concentrations with the IQR of analyzed samples are summarized in Additional file 1: Table S1.

In order to evaluate the diagnostic utility of cytokines IL-7 and G-CSF for predicting AIP, a ROC curve analysis was carried out using the AIP, CP, and PDAC cytokine serum data. The obtained results are summarized in Table 3. IL-7 discriminated better than G-CSF AIP from PDAC (AUC of 0.780 vs 0.686). Combination of both IL-7 and G-CSF for differentiation only marginally improved the diagnostic value of the two markers (AUC = 0.782) as presented in Fig. 2; addition of IL-1β showed no further improvement (results not shown). Of note, the comparison between patients with CP and those with AIP, G-CSF alone performed better (AUC = 0.804) than the combination of both G-CSF and IL-7 (AUC = 0.787). Using G-CSF as reference curve the differences in the AUC reached significance in the AIP vs. CP comparisons (p = 0.0052) but not in the AIP vs PDAC using G-CSF as reference curve (p = 0.395).

Similarly to the serum cytokine profiling, a tissue homogenate analysis was performed. The levels of cytokines in CP, PDAC, and both AIP forms were compared. Here, significantly higher concentrations (9.56 pg/mg) of G-CSF (p = 0.0138) were measured in AIP vs PDAC (3.44 pg/mg) and significantly lower levels were measured for IL-8 and TNF-α in the AIP group.

On the other hand, comparing the AIP and CP groups, AIP revealed significantly higher concentrations for IL-17 (p = 0.0120) and MIP-1β (p = 0.0326) and lower levels of TNF-α (p = 0.0038). The median concentration values and IQR are presented in Table 4.

Subsequently, all groups were separately compared either with subgroup AIP-1 or with AIP-2. Cytokines with interesting differences between the analyzed groups are presented in Fig. 3a–k and Additional file 2: Table S2. Of note, both AIP subtypes were distinguished from another by two cytokines IL-8 (p = 0.0152) and IFN-γ (p = 0.0376) showing higher levels in the AIP-2 subtype. IL-10 also showed slightly higher levels in AIP-2 (p = 0.0542).