From 2005 to 2011, 28 patients with newly diagnosed PDA were enrolled in the study approved. Inclusion criteria consisted of an ECOG performance status 0–2, signing of a written informed consent form and no previous CT or radiotherapy. Exclusion criteria consisted of a lack of ability to perform follow-up (1 year) and any previous malignant tumors, with the following exception: adequately treated basal/spino-cell carcinomas, cervical carcinoma in situ, and patients affected by other malignancies, but disease-free for at least 5 years at the date of enrollment.

In this cohort, 14 patients had been subjected to total surgical resection, 4 patients had received palliative resection only, and 10 patients had not been resected. All patients were treated with CT, 26 with GEM-based CT (GEM or GEM plus oxaliplatin) and 2 with capecitabine and bevacizumab (table 1).

Peripheral blood mononuclear cells (PBMC) and sera samples were isolated from venous blood before CT, and after each round of CT, and stored until use.

PDA cell lines CFPAC-1, CAPAN-2, DT6606, and K8484 were cultured at 37°C in a humidified 5% CO₂ atmosphere in DMEM (Invitrogen, San Giuliano Milanese, Italy) supplemented with 20 mM glutamine (Invitrogen), 10% certified fetal bovine serum (FBS) (Invitrogen), and 40 µg/mL gentamycin (Schering-Plow, Milan, Italy).

For proteomics studies, CFPAC-1 cells were untreated or treated with 0.25 µg/µL GEM for 24 hours. Procedure details are described in the online supplemental file. For serological proteome analysis (SERPA), CFPAC-1 cells were solubilized and proteins were quantified to perform two-dimensional electrophoresis (DE) and Coomassie staining, as previously described.18 19 2-DE gels were transferred to nitrocellulose membranes and incubated with sera from patients before and after CT. Detailed procedures are described in the online supplemental file.

CFPAC-1 or CAPAN-2 cells were seeded in 96-well plates (5×10⁴ cells/well). Cells were washed with warm PBS, incubated with sera diluted in PBS (1:1000) for 1 hour at 4°C, then washed again, followed by incubation with fresh reconstituted human complement (Cedarlane, Euroclone, Milan, Italy) diluted at 1:10 in PBS for 1 hour at 37°C. The CytoTox 96 non-radioactive cytotoxicity kit (Promega, Milan, Italy) is based on the ELISA, and detects the release of lactate dehydrogenase (LDH) after 4 hours of incubation from cells that have lost membrane integrity (lysed cells). A LDH positive control was added to empty wells of each plate, and all tests were performed in triplicate.

The quantity of total IgG in 12 patients was measured using an ELISA kit before and after CT (Bethyl Laboratories, Montgomery, Texas, USA), following the manufacturer’s instructions. TAA (ENO1, FUBP1, K2C8, and G3P)-specific IgG responses of sera from 28 healthy subjects and 28 patients with PDA, before and after each round of CT, were analyzed by ELISA (see online supplemental file). Anti-ENO1 and anti-G3P IgG in mouse sera, at different weeks of age, were measured by ELISA using ENO1 or G3P, as previously described.

After thawing, PBMC from 13 patients, before and after each round of CT, were cultured at a density of 5×10⁴/well in RPMI (Invitrogen) plus 5% of FBS in 96-well microplates with 5 µg/mL of ENO1, FUBP1, K2C8, or G3P. At day 3, supernatants were collected and IFN-γ and IL-10 levels were measured by ELISA (BioLegend, Campoverde, Milan, Italy), following the manufacturer’s instructions. Specific production of IFN-γ and IL-10 in the supernatants was calculated by subtracting the concentration of each cytokine in the medium only from that in PBMC stimulated with TAA. These specific concentration values were pseudocounted (+1) and then expressed as an IFN-γ/IL-10 ratio. At day 5, proliferation was evaluated using a tritiated-thymidine (3H-TdR) incorporation assay (see online supplemental file). Stimulation index (SI) T cell proliferation was calculated by dividing the cpm from PBMC stimulated with each antigen by the cpm of PBMC cultured in medium only. PBMC from five patients were cultured 5×10⁵ in 48-well plates in the presence of each TAA, as described above, and at day 3, cells were collected and stained with anti CD4-PercP, CD8-APC, and FoxP3-FITC to perform flow cytometer analysis (see online supplemental file).

Pancreatic cancer-prone KC mice were generated by crossing single-mutated KrasG12D with C57BL/6 mice expressing Cre recombinase (see online supplemental file). KC mice were treated twice at week 8 by i.p. injection with GEM (on Monday and Friday) using 1 mg/injection/mouse. The ENO1 vaccine was administered at 9 weeks of age, and every 2 weeks for a total of three rounds of vaccination (see online supplemental file). Mice were sacrificed at 24 weeks of age, and pancreases were fixed in formalin and embedded in paraffin for analyzing tumor lesions and immunohistochemistry (see online supplemental file). Different experimental groups of mice were depleted of CD4, CD8 or CD20 (Bioxcell, West Lebanon, New Hampshire, USA) every 3 days (200 µg/mouse/injected i.p.) until the week after the final vaccination. At 14 weeks, depleted mice were sacrificed, and pancreases were fixed in formalin and embedded in paraffin for evaluating tumor lesions and anti-ENO1 antibody titers. To test the effectiveness of GEM+ENO1 vaccination on established tumors, C57BL/6 mice were injected subcutaneously with pancreatic tumor cells in the right flank (K8484, 1×10⁵ cells/mouse) and treated with GEM+ENO1 vaccination when tumors reached 0.2 cm in diameter, as described above.

Splenocytes from treated or untreated mice were stimulated in vitro for 4 days in the presence of 10 µg/mL of selected TAA proteins. Recovered T cells were seeded at a density of 1×10⁵/well in triplicate and stimulated with DT6606 cells (with a stimulator/effector ratio of 1:100) or 5 µg/mL TAA for 48 hours to evaluate specific IFN-γ production by ELISpot assay (see online supplemental file).

Data from SERPA experiments, protein differential expression, overall survival, immunohistochemical and FACS staining, ELISA, and ELISpot were analyzed, as described in the online supplementary file.

Circulating IgG from the sera of 28 patients with PDA before and after one or two rounds of CT were profiled by SERPA. In CT-treated patients with PDA, a clear enhancement of the IgG-specific response in both the number and intensity of recognized TAA was observed (figure 1A, B). This increase was not due to a global enhancement of total IgG after two rounds of CT as no difference in sera IgG levels was detected by ELISA (online supplemental figure 1A).

The CDC of sera from patients with PDA before and after CT was evaluated in CFPAC-1 cells (complement sensitive), which do not express complement activity inhibitor CD55, and in CAPAN-2 cells (complement resistant), which do express CD55 (online supplemental figure 1B). A significant difference between sera-induced LDH release in CFPAC-1 and CAPAN-2 cells after the first and second rounds, but not before CT, was observed. A significant increase of sera-induced LDH release against CFPAC-1, but not in CAPAN-2 cells, was found after CT rounds (figure 1C).

Mass spectrometry revealed 91 proteins from 160 isoforms, highlighted by SERPA as distinct spots (online supplemental table 1A and online supplemental figure 1C). Following CT, a widespread and progressive increase of the autoantibody response to TAA was observed (figure 2). There was an increase in 32 TAA recognized in at least three patients after CT (online supplemental figure 1D), among which ENO1 was the most recognized TAA (online supplemental table 1B), in accordance with our previous studies.20

Proteomic analysis of GEM-treated CFPAC-1 cells revealed that CT modified the expression of 34 proteins, of which 12 corresponded to TAA whose autoantibody recognition was increased in patients with PDA following CT (online supplemental figure 2A,B and supplemental table 2).

The 32 protein targets of autoantibody responses following CT (online supplemental figure 1D) were coded by 23 genes transcriptionally overexpressed in The Cancer Genome Atlas (TCGA) patients with PDA (online supplemental figure 3) and gene-set enrichment analysis (online supplemental table 3).

There was considerable TAA recognition by autoantibodies of patients, and 11 TAA were recognized by at least 7 patients before and/or after CT (online supplemental table 1B). Four TAA, namely ENO1, FUBP1, K2C8 and G3P, were selected to represent metabolic-related, cytoskeletal-related and transcription-related molecules. A statistically significant correlation was found between patient survival and the increased autoantibody recognition of G3P and K2C8 after the first round of CT (figure 3A,B). Data from ELISA showed that a group of patients with PDA displayed a significantly higher level of autoantibodies to all four TAA compared with healthy subjects (figure 3C). A significant increase of autoantibodies to ENO1, K2C8 and G3P was measured in patients with PDA after CT (figure 3C), confirming the SERPA results.

Peripheral blood T lymphocytes from patients with PDA were stimulated with each of the four selected TAA, and the proliferation, regulatory (IFN-γ/IL-10 ratio <1), effector (IFN-γ/IL-10 ratio >1), or null cytokine responses (no production of either IFN-γ or IL-10) was monitored (table 2).

Overall, the number of patient T cell proliferative responses (evaluated as SI≥2) to selected TAA were significantly increased after CT rounds (online supplemental figure 3A).

The relationship between proliferative and regulatory/effector responses to each TAA were also analyzed. Before CT, only one patient (pt. 24) displayed an effector response to ENO1. By contrast, seven patients (pt. 3, 9, 10, 24, 30, 32, 87) showed effector responses after CT, with just two of them (pt. 32, 87) being associated with a proliferative response (table 2). Before CT, four patients (pt. 3, 19, 41, 87) displayed effector responses to FUBP1, but only two showed a concomitant proliferative response (pt. 41, 87). After CT, the number of patients that displayed an effector response increased to 11 (pt. 3, 5, 9, 10, 19, 24, 25, 27, 30, 41, 84), with 8 of them (pt. 5, 9, 10, 25, 27, 30, 41, 84) being associated with a proliferative response (table 2). Before CT, six patients (pt. 3, 19, 25, 41, 84, 87) displayed an effector response to K2C8, with four of them (pt. 25, 41, 84, 87) showing concomitant proliferative responses. After CT, there was a regulatory-to-effector shift in three patients (pt. 9, 10, 24); an increase in effector responses in three patients (pt. 3, 19, 87); and a decrease in regulatory responses in one patient (pt. 30). Among these seven patients, five (pt. 9, 10, 24, 30, 87) were associated with proliferation (table 2). Before CT, three patients (pt. 3, 24, 87) displayed a G3P-induced effector response, with only one (pt. 87) showing a concomitant proliferative response. After CT, there was a regulatory-to-effector shift in two patients (pt. 9, 19); an increase in effector responses in four patients (pt. 3, 10, 24, 87); and a reduction in regulatory responses in two patients (pt. 30, 41). Among these eight patients, seven (pt. 3, 9, 10, 24, 30, 41, 87) showed T cell proliferation.

A significant increase of the ratio between the percentage of CD8 and Treg after one and two CT rounds was also observed in TAA-stimulated T lymphocytes from patients with PDA (online supplemental figure 3B and supplemental table 4).

The effect of the combined treatment of CT and DNA vaccination to ENO1, a TAA expressed in the early phases of PDA carcinogenesis (online supplemental figure 5A) was evaluated in KC mice.

KC mice were treated with a suboptimal dose of GEM and then vaccinated with ENO1 (figure 4A). The size of all neoplastic lesions (from PanINs to invasive PDA) was evaluated by H&E staining. Both ENO1 and GEM+ENO1 treatment induced a significant reduction of PDA lesions compared with untreated mice and GEM+ENO1 treatment further reduced tumor lesions compared with the ENO1 vaccine alone (figure 4B). The evaluation of both the number and size of invasive PDA and PanIN lesions showed that ENO1 vaccination and combined treatment reduced the number of both invasive and PanIN lesions, whereas only the size of PanIN lesions were reduced by ENO1 vaccination alone and to a larger extent by combined treatment (online supplemental figure 6).

Smaller tumor lesions were accompanied with an increase of anti-ENO1 IgG antibodies (figure 4C). ELISpot revealed that splenocytes from ENO1-vaccinated mice displayed a significantly higher number of IFN-γ-secreting cells, which was increased in the GEM+ENO1 group (figure 4E). These results were confirmed when syngeneic PDA murine DT6606 cells were used as a stimulus (not shown). GEM+ENO1 treatment caused an increase of both IgG production and IFN-γ secretion specific to G3P, enhancing the epitope spreading effect of ENO1 (figure 4D and F). No increase in anti-K2C8 or anti-FUBP1 IgG and IFN-γ was observed in GEM+ENO1-treated mice (data not shown).

Immunohistochemistry demonstrated that only GEM+ENO1 caused a significant increase of tumor-infiltrating CD4 and CD8 T cells (figure 4G–H and online supplemental figure 5B,C). Compared with untreated mice, an increase of macrophage infiltration was observed in ENO1-vaccinated mice, but not in GEM or GEM+ENO1 treated mice (figure 4I and online supplemental figure 5D). No differences in PD-L1 and FoxP3 expression were observed in tumor lesions among the different groups (data not shown). At 24 weeks, GEM+ENO1 mice showed a significant reduction of circulating CD4 and CD8 T cells, which supported the increased infiltration of these cells in tumor lesions (online supplemental figure 5E,F). In the GEM+ENO1 group, a significant correlation between CD4 and CD8 infiltration was observed (online supplemental figure 5G). Depletion of CD4, but not of CD8 or B cells, completely reverted the antitumor effect of the combined treatment (figure 4L). Anti-ENO1 IgG titers decreased significantly in all subset-depleted mice (online supplemental figure 5H). Finally, a significant delay of the growth of established 0.2 cm tumors was observed in GEM+ENO1 treated mice (figure 4M).