Background
Pancreatic ductal adenocarcinoma (PDAC) is the fourth leading cause of cancer death in Western countries. Upon diagnosis, less than 20% of patients present localized, potentially curable tumors. The overall 5-year survival rate is <5% [
1,
2]. This poor prognosis has been attributed to failure in early disease diagnosis, when the tumor may still be resectable, along with its propensity to disseminate and its resistance to systemic treatment [
3]. CA19.9 is the only biomarker that has demonstrated clinical value for therapeutic monitoring and early detection of recurrent disease after treatment in patients with known pancreatic cancer. However, its use as a screening tool has proved unsuccessful, thus other biomarkers alone or in combination with it are required for early diagnosis of PDAC [
1].
Autoantibody levels can function as diagnostic and prognostic markers [
4,
5]. By SERological Proteome Analysis (SERPA) we have previously identified a number of PDAC-associated antigens that are specifically recognized by circulating autoantibodies present in the serum of PDAC patients [
6‐
9]. However as these autoantibodies were discovered in sera from patients at advanced stages of PDAC, earlier diagnostic markers would not have been identified.
Genetically engineered mice (GEM) that spontaneously develop PDAC may be used to facilitate the development of novel tests for the early detection and treatment of PDAC [
10].
LSL-Kras
G12D/+
;
Pdx-1-Cre mice (KC) develop the entire histologic compendium of pancreatic intraepithelial neoplasia (PanIN) lesions observed in the human disease, and a subset of mice also develop invasive pancreatic carcinomas.
LSL-Kras
G12D/+
;
LSL-Trp53
R172H/+
; Pdx-1-Cre double mutant mice (KPC), develop a more aggressive invasive and metastatic PDAC with an earlier time of onset, and display a reduced survival rate compared to KC mice [
11,
12].
In the present study, we used SERPA to identify TAAs eliciting an early humoral response in KC and KPC. Results from two-dimensional electrophoresis (2DE), Western blotting (WB) and mass spectrometry (MS) were combined to compare the reactivity of KC and KPC sera to that of corresponding matched controls. Antigens recognized by autoantibodies in KC and KPC at PanIN stages were identified and validated in a set of resectable and advanced PDAC patients. Ezrin (EZR), the protein with the highest frequency of autoantibodies in both early stage GEM and resectable PDAC patients, was validated by ELISA test using PDAC sera either collected at the time of diagnosis or several months before cancer onset (prediagnostic PDAC). The sensitivity and specificity of EZR-autoantibodies for discriminating PDAC was evaluated together with other serological markers.
Methods
Murine study
All animals were treated in accordance with European and institutional guidelines (Legislative Order No. 116/92). 129SvJae/B6 H-2D
b mice carrying mutated Kras
G12D and Trp53
R172H under the endogenous promoter, and flanked by Lox-STOP-Lox cassettes (
LSL-Kras
G12D/+
and
LSL-Trp53
R172H/+
) were kindly provided from Dr. D.A. Tuveson (Cancer Research UK, Cambridge Research Institute, Cambridge, UK). C57BL/6 mice expressing Cre recombinase under a specific pancreatic transcriptional factor Pdx-1 (pancreatic duodenum homeobox 1) promoter
(Pdx-1-Cre) were obtained from Dr. A.M. Lowy (University of California, San Diego, CA). Conditional
LSL-Kras
G12D/+
,
LSL-Trp53
R172H/+
and
Pdx-1-Cre strains were bred to obtain
LSL-Kras
G12D/+
;
Pdx-1-Cre single mutant (KC) and
LSL-Kras
G12D/+
;
LSL-Trp53
R172H/+
; Pdx-1-Cre double mutant (KPC) mice [
11,
12]. To collect serum, mice were euthanized and blood was collected by cardiac puncture using a 22-gauge needle and 1 ml syringe. Mice were surgically and pathologically examined to confirm the presence of pancreatic tumors and metastases.
Human studies
Cross-sectional clinical study
The study was approved by the Ethical Committees of: Azienda Ospedaliera Città della Salute e della Scienza di Torino, Turin; Policlinico G.B. Rossi, Verona; Regina Elena National Cancer Institute, Rome and Ordine Mauriziano Hospital, Turin. Serum samples were isolated from venous blood at time of diagnosis with the informed consent of patients and control subjects and stored at -80°C until use. De-identified numeric specimen codes were used to protect the identity of the individuals. Diagnosis of PDAC or any other cancer was consistently confirmed by histological or cytological analysis. Sera from 120 PDAC patients (M/F: 67/53; median age, 67 y; range, 32–86 y) with clinical features previously described [
9] were analyzed by SERPA, and sera from 69 PDAC patients with clinical features described in Table
1 were tested by ELISA. Reactivity of these sera was compared, in both SERPA and ELISA studies, with that of control sera from the following sources: 60 healthy subjects (HS, M/F: 25/35; median age, 70 y; range, 49 - 90 y) with no prior history of cancer or autoimmune disease; 50 non-PDAC cancer patients (9 liver, 12 breast, 9 colon, 19 lung and 1 ovarian; M/F: 24/26; median age, 69 y; range, 44 - 86 y); 46 chronic pancreatitis patients (CP, M/F: 26/20; median age, 58 y; range, 22 - 74 y); 12 autoimmune diseases patients (AD, 3 Mixed Cryoglobulinemia, 2 Meniere's Syndrome, 4 Rheumatoid Arthritis, 2 Systemic Lupus Erythematosus, and 1 Autoimmune Pancreatitis; M/F: 3/9; median age, 49 y; range, 38 - 79 y).
Table 1
Clinical features of PDAC patients analyzed by ELISA
Gender
| | | | |
| Male | | 39 | 57 |
| Female | | 30 | 43 |
Age (y)
| | | | |
| Mean | 63 | - | - |
| Range | 42-84 | - | - |
Stage
b
| | | | |
| IB | | 1 | 2 |
| IIA | | 7 | 10 |
| IIB | | 29 | 42 |
| III | | 10 | 14 |
| IV | | 22 | 32 |
Grading
| | | | |
| Not reported | | 32 | 46 |
| 1 | | 4 | 6 |
| 2 | | 16 | 23 |
| 3 | | 17 | 25 |
Primary site
| | | | |
| Head | | 49 | 71 |
| Body | | 6 | 9 |
| Tail | | 5 | 7 |
| Body-Tail | | 9 | 13 |
ECOG PS
| | | | |
| Not reported | | 10 | 14 |
| 0 | | 32 | 47 |
| 1 | | 25 | 36 |
| ≥2 | | 2 | 3 |
Surgery with radical intent
| | | | |
| Yes | | 39 | 57 |
| No | | 30 | 43 |
Baseline CA19.9 (IU/ml)
| | | | |
| Evaluable | | 63 | 91 |
| Mean | 3052 | - | - |
| Median | 500 | - | - |
| Range | 2- > 12000 | - | - |
First-line chemotherapy
c
| | | | |
| Evaluable | | 59 | 86 |
| Gem | | 43 | 73 |
| Gem/Oxal | | 10 | 17 |
| Gem/5-FU | | 3 | 5 |
| Non-Gem | | 1 | 2 |
| No CT | | 2 | 3 |
ENOA1,2 Reactivity
| | | | |
| Evaluable | | 50 | 73 |
| Positive | | 34 | 68 |
| Negative | | 16 | 32 |
Prospective pre-clinical study
Prediagnostic serum samples of PDAC patients and matched controls were obtained from the Turin European Prospective Investigation into Cancer and Nutrition (EPIC) cohort that includes samples from 10 604 healthy subjects at the moment of enrolment (6 047 males and 4 557 females, aged 35–65 y) recruited in the city of Turin. Recruitment took place between 1993–1998 and involved blood donors and other healthy volunteers. After blood donation, samples were stored at 5–10°C, protected from light, and transported to local laboratories for processing and dividing into aliquots. Blood was separated into 0.5-ml fractions (serum, plasma, red cells, and buffy coat for DNA extraction) and stored in heat-sealed straws in liquid nitrogen (-196°C). Subjects were monitored longitudinally for cancer or other disease development. Co-operation with the local cancer registry and the local health authority enabled access to hospital discharge information and all newly diagnosed cancer cases. Study design, population and baseline data collection have previously been described in detail [
13,
14]. Sixteen PDAC patients identified from the Turin EPIC cohort are included in the present study. Controls were matched by age, sex, and date at entry in the cohort, and did not develop any cancer or autoimmune disease. Characteristics of subjects are summarized in Table
2. Each participant provided informed consent, and the local Ethics Review Committees approved this study.
Table 2
Characteristics of the EPIC cohort subjects
Total
| 16 | 100 | 32 | 100 |
Age (y)
| | | | |
Mean | 54.9 | | 55.1 | |
SD | 7.3 | | 7.5 | |
Sex
| | | | |
Female | 7 | 44 | 14 | 44 |
Male | 9 | 56 | 18 | 56 |
Time span to diagnosis (mo)
| | | | |
Mean | 61.2 | | | |
Range | 5–117.1 | | | |
Two-dimensional electrophoresis and western blot analysis
Cells (10
7) from the CF-PAC-1 (ECACC ref. No. 91112501) and K8484 isolated from a tumor arising in KPC mice, kindly provided by Dr. K.P. Olive (Columbia University, New York, NY), were solubilized, subjected to 2DE and electro-transferred onto a nitrocellulose membrane (GE Healthcare Bio-Sciences, Uppsala, Sweden) as previously described [
6]. Frozen PDAC tissues from eight surgically-treated patients (stage IIA and IIB of PDAC) were homogenized in 2DE lysis buffer, subjected to 2DE and electro-blotted onto a nitrocellulose membrane (GE Healthcare) as previously described [
9]. Sera from KC, KPC, PDAC patients and controls were tested to determine mouse and human IgG concentrations using commercial kits (IgG ELISA Quantitation Set from Bethyl Laboratories - Montgomery, TX, USA). Sera were individually tested on 2DE maps at a working dilution of 0.1 mg/ml IgG for 4 h, followed by incubation with horseradish peroxidase (HRP)-conjugated rabbit anti-human IgG (90 minutes, 1:1000; Santa Cruz Biotechnology, Santa Cruz, CA, USA) or sheep anti-mouse Ig (90 minutes, 1:5000; GE Healthcare) as a secondary antibody. Ezrin spots were revealed with anti-Ezrin antibody (1 hour incubation, 1:5000; Abcam, Cambridge, MA, USA) and HRP-conjugated donkey anti-rabbit IgG (1 hour incubation, 1:2000; GE Healthcare) as a secondary antibody. Immunodetection was accomplished by ECL PLUS (Enhanced Chemiluminescence, GE Healthcare). The chemifluorescent signals were scanned with “ProXPRESS 2D” (PerkinElmer, Waltham, MA, USA) with an excitation/emission filter setting of 460/80 and 530/30, respectively, for an exposure time of 12 s. Images were recorded in TIFF format. The volume of each spot recognized by autoantibodies was calculated after background subtraction using “ProFinder 2D” (PerkinElmer) software and reported as arbitrary units (AU). For proteins represented from more than one spot the volume was expressed as a mean value.
Protein identification by mass spectrometry
Coomassie G-stained spots were excised from 2DE preparative gels; destaining and in-gel enzymatic digestion were performed as previously described [
15]. Briefly each spot was destained with 100 μl of 50% vol/vol acetonitrile in 5 mmol/l ammonium bicarbonate and dried with 100 μl of acetonitrile. Each dried gel piece was rehydrated for 40 minutes at 4°C in 10 μl of a digestion buffer containing 5 mmol/l ammonium bicarbonate, and 10 ng/μl of trypsin. Digestion was allowed to proceed overnight at 37°C and peptide mixtures were stored at 4°C until assayed. All digests were analyzed by a MALDI micro MX - TOF Mass Spectrometer (Waters, MA, USA) equipped with a delayed extraction unit. Peptide solution was prepared with equal volumes of saturated α-cyano-4-hydroxycinnamic acid solution in 40% vol/vol acetonitrile-0.1% vol/vol trifluoroacetic acid. The MALDI-TOF was calibrated with a mix of PEG (PEG 1000, 2000 and 3000 with the ratio 1:1:2) and mass spectra were acquired in the positive-ion mode. Peak lists were generated with ProteinLynx Data Preparation (ProteinLynx Global Server 2.2.5) using the following parameters: external calibration with lock mass using mass 2465.1989 Da of ACTH, background subtract type adaptive combining all scans, performing deisotoping with a threshold of 1%. The 25 most intense masses were used for database searches against the SWISSPROT database (Release 2011_12 of 14-Dec-11) using the free search program MASCOT 2.3.02 (
http://www.matrixscience.com/cgi/search_form.pl?FORMVER=2&SEARCH=PMF). The following parameters were used in the searches: taxa
Homo sapiens or
Mus musculus, trypsin digest, one missed cleavage by trypsin, carbamidomethylation of cysteine as fixed modification, methionine oxidation as variable modifications and maximum error allowed 100 ppm. Only proteins with a Mascot score >55 were taken into consideration.
Anti-Ezrin autoantibody capture by enzyme-linked immunosorbent assay
Purified recombinant protein of Homo sapiens Ezrin, transcript variant 1 (OriGene, Rockville, MD, USA) was used to capture autoantibodies to Ezrin. Briefly, the protein was coated (0.5 μg/ml in PBS) on 96-well micro-plates overnight at room temperature, followed by blocking with PBS containing 4% bovine serum albumin for 2 hours at room temperature. Sera (working dilution 0.01 mg/ml) were then added to the coated wells for 2 hours at room temperature. After washing with PBS-Tween-20, microplates were incubated with HRP-conjugated rabbit anti-human IgG (dilution 1:1000; Santa Cruz Biotechnology) for 1 hour at room temperature and TMB One Solution (Promega, Madison, WI, USA) was added to each well. The reaction was stopped by 2N HCl and the optical density (OD) value was measured at 450 nm. The corresponding background values of the sera on uncoated wells were subtracted. All samples were assayed in triplicate and the results represent mean values.
Statistical analysis
Statistical analysis was performed using GraphPad (Version 4, San Diego, CA), MedCalc (Version 11.4.2.0, Mariakerke, Belgium) and SPSS (Version 18.0, Chicago, IL, USA) software packages. Mouse survival was estimated by Kaplan-Meier analysis and compared with Log-rank tests. Receiver operating characteristic (ROC) curve analysis was performed in order to find the optimal cut-off levels capable of splitting patients into groups with different outcome probabilities. Specificity, sensitivity and area under curve (AUC) were estimated considering histology results as the gold standard. The classification and regression tree (CART) analysis, a type of decision tree methodology, is a nonparametric statistical procedure that identifies mutually exclusive and exhaustive subgroups of a population whose members share common characteristics that influence the dependent variable of interest. CART uses a binary recursive partitioning method that produces a decision tree that identifies subgroups of patients with a higher likelihood of being found positive in a test for a disease state. The exhaustive CHAID method was used for CART analysis. Correlations and associations between variables were tested by Pearson’s test, Student's t-test, χ2 test or Fisher’s exact test, as appropriate. For all tests, 2-sided P < 0.05 (*), P < 0.005 (**) and P < 0.0005 (***) values were considered significant.
Discussion
This study identifies autoantibodies to EZR as early markers in mouse and human PDAC. Of clinical relevance, we also show that EZR-autoantibodies efficiently complement the diagnostic performance of CA19.9.
To identify early immune response markers we applied SERological Proteome Analysis (SERPA) in KC and KPC mice spontaneously developing PDAC. As GEM can be sampled at defined stages of tumor development and under controlled breeding conditions, greater standardization is possible when using mouse models as opposed to human studies. GEM allowed us to identify EZR-autoantibodies as early biomarkers in PDAC, since precociously detected in their serum when the disease stage was limited to PanIN. Through this approach, we also identified additional antigens (VCL, PDC6I, FUBP2, hnRNPL, VIM, K2C8, ANXA1 e ANXA2) recognized at high frequencies by both KC and KPC sera.
Reactivity against some of these antigens was present in control mice, but the intensity of WB recognition was much greater in GEM. A clear example is represented by EZR, faintly recognized by a number of control mice but strongly evident in all KPC. Despite the fact that auto-reactive lymphocytes should have been removed from the repertoire before maturation into naïve B cells, a large number of circulating IgG
+ memory B cells produce low affinity antibodies to self-antigens [
17,
18]. The humoral response against these self-antigens is strongly increased in tumor conditions, as demonstrated in this work both in humans and in mice (Figure
1, Table
3 and Figure
4). Some differences in the pattern of recognition were present between mice of the same age, probably due to the molecular heterogeneity of tumor progression in this model that fully recapitulates the genetic and molecular features of human PDAC [
12]. Importantly, all the identified antigens, except for FUBP2 and ANXA1, induced a powerful humoral response not only in KPC but also in KC bearing PanIN lesions, indicating that the antibody response to these TAAs is already occurring when
Kras is the only genetic alteration in the tumor, independently of
p53 mutation, which is a later event in PDAC development. This reflects previous studies reporting how the immune response to TAAs in humans occurs at an early stage during tumorigenesis, as illustrated by the detection of high titers of autoantibodies, as early as 5 years before disease onset [
19,
20].
By comparing the 2DE WB reactivity of GEM with that of a large cohort of PDAC patients and controls, six proteins, namely: EZR, ANXA2, VCL, hnRNPL, ANXA1 and PDC6I were common to both human and mouse signatures. EZR and ANXA2 were recognized by most PDAC patients who underwent surgery with curative intent. ELISA confirmed the diagnostic value of anti-EZR but not anti-ANXA2 autoantibodies, which were also present in control groups. Other studies have indeed reported the presence of autoantibodies against ANXA2 in systemic autoimmune diseases and lung cancer, as well as pancreatic cancer [
21‐
23], suggesting that the humoral response to ANXA2 is not specific for PDAC transformation.
EZR is a member of the ezrin-radixin-moesin (ERM) family and a link between a number of growth factor receptors/adhesion molecules and the actin cytoskeleton. It is localized to the cytoplasm as an inactive form. Upon threonine and tyrosine phosphorylation, EZR is transported to the cell membrane whereupon it tethers F-actin [
24]. It works downstream of cell-surface receptors through the activation of Rho and PI3K/Akt signaling pathways [
25,
26], and in physiological conditions, EZR is required for macropinocytosis, cell adhesion, and membrane ruffling in epithelial cells, whereas in tumor cells it is an important metastatic regulator [
27]. EZR is overexpressed in many cancers, including PDAC, even in PanIN lesions [
28‐
30], and it interacts with cortactin to form podosomal rosettes in PDAC cells, which may play an important role in tumor invasion [
31]. These observations support the immunogenicity of EZR that we observed in the present study, even if it is not clear how TAAs overcome self-tolerance and thus become autoantibody targets in cancer patients, as many of those discovered so far are intracellular proteins [
4,
32,
33]. Interestingly, EZR has been identified both in exosomes secreted by mesothelioma cells [
34] and as a substrate of matrix metalloproteinases able to generate neo-epitopes from self-antigens [
35].
The most important observation of our study is that autoantibodies against EZR were present also in prediagnostic PDAC samples from the prospective EPIC cohort that were collected several months or years before PDAC diagnosis. The EPIC study recruited over half a million healthy volunteers in ten European countries, including Italy, monitored longitudinally for cancer or other disease development [
36]. Since it has been estimated that the elapsed time between PDAC initiation to metastatic spread is at least 10 years [
37], our results strongly support the hypothesis that EZR-autoantibody development is an early event in PDAC. Notably, prediagnostic patients with the highest levels of EZR-autoantibody in the ELISA test were the ones with an intermediate time lag to diagnosis (69.3 and 56.9 mo, Additional file
1: Table S3). This observation supports the hypothesis that autoantibody levels decrease closer to diagnosis due to immune complex formation [
20].
Although EZR-autoantibody testing has displayed a high diagnostic performance, especially in resectable PDAC patients (Additional file
1: Figure S4), a single TAA may lack adequate sensitivity and specificity, and the combination of a panel of autoantibodies and serological markers can improve the overall accuracy of a diagnostic assay for cancer detection. We therefore assessed the diagnostic performance of combined dichotomized EZR-autoantibody levels, CA19.9, the only PDAC marker currently in clinical use, and ENOA1,2-autoantibodies. We have previously demonstrated that autoantibodies against Ser-419-phosphorylated ENOA isoforms (ENOA1,2) complement the performance of CA.19.9 [
9]. Interestingly, a diagnostic algorithm separating CA19.9 and EZR-autoantibodies discordant cases into PDAC or controls based on the presence or absence of ENOA1,2-autoantibodies respectively, resulted in an overall diagnostic accuracy of 0.96. Notably, the algorithm here applied is more stringent than the one previously described by our group [
9], where tested cases were assigned to the PDAC group when either ENOA1,2-autoantibodies or CA19.9 were positive, as their values were inversely correlated. This finding is of real translational relevance, since CA19.9 is the only biomarker with demonstrated clinical value for therapeutic monitoring and detection of recurrent PDAC, but its use as a screening tool has proved unsuccessful until now [
3].
Further validation studies, performed in a large and independent patient cohort, are warranted to establish the diagnostic performance of this multiplexed analysis and of the identified TAA panel tested alone or in combination.
Competing interests
FN, MC, and PC are inventors of an Italian patent application No: TO2012A000523 entitled “Kit for in vitro diagnosis and predisposition assessment of pancreatic ductal adenocarcinoma”. Potential investigator conflict of interest has been disclosed to study participants.
Authors’ contributions
MC designed the study, performed human SERPA and ELISA experiments, analyzed the data and wrote the manuscript; PC designed the study, coordinated and performed GEM breeding, murine sample collection and analyzed data; FCL performed murine SERPA studies and GEM serum collection; MG contributed to GEM breeding and analyzed data; RC performed GEM histological and immunohistochemical analysis; IS performed statistical analysis; GM performed mass spectrometry analysis; SB and SB performed human histological and immunohistochemical analysis; SB performed microarray analysis; AN, PN, PS, AS, CB and MM recruited patients and contributed to experimental design and analysis of data; AN, CS and PV provided samples from the Turin EPIC cohort and analyzed data; FN supervised the project and wrote the manuscript. All authors read and approved the final manuscript.