Autoantibodies to Ezrin are an early sign of pancreatic cancer in humans and in genetically engineered mouse models

Abstract

Background: Pancreatic Ductal Adenocarcinoma (PDAC) is a highly aggressive malignancy with only a 5% 5-year survival rate. Reliable biomarkers for early detection are still lacking. The goals of this study were (a) to identify early humoral responses in genetically engineered mice (GEM) spontaneously developing PDAC; and (b) to test their diagnostic/predictive value in newly diagnosed PDAC patients and in prediagnostic sera.

Methods and results: The serum reactivity of GEM from inception to invasive cancer, and in resectable or advanced human PDAC was tested by two-dimensional electrophoresis Western blot against proteins from murine and human PDAC cell lines, respectively. A common mouse-to-human autoantibody signature, directed against six antigens identified by MALDI-TOF mass spectrometry, was determined. Of the six antigens, Ezrin displayed the highest frequency of autoantibodies in GEM with early disease and in PDAC patients with resectable disease. The diagnostic value of Ezrin-autoantibodies to discriminate PDAC from controls was further shown by ELISA and ROC analyses (P < 0.0001). This observation was confirmed in prediagnostic sera from the EPIC prospective study in patients who eventually developed PDAC (with a mean time lag of 61.2 months between blood drawing and PDAC diagnosis). A combination of Ezrin-autoantibodies with CA19.9 serum levels and phosphorylated α-Enolase autoantibodies showed an overall diagnostic accuracy of 0.96 ± 0.02.

Conclusions: Autoantibodies against Ezrin are induced early in PDAC and their combination with other serological markers may provide a predictive and diagnostic signature.

Figure 1

SERPA analysis of KC and KPC serum reactivity against K8484 cell line 2DE map. Total lysates from the K8484 cell line were separated by 2DE as described in the Methods section. Samples were focused in the first dimension using a gradient spanning the indicated pH range, separated in the second dimension in 4-12% acrylamide gels and subsequently Blue Coomassie stained (A) or transferred to a nitrocellulose membrane and probed with mouse sera. Three representative Western blot images show the immunoreactivity of control (B), KC (C) or KPC (D) serum. Immunoreactive protein spots were determined for each serum by superimposition of immunoblot signal pattern with the spot pattern of the corresponding Blue Coomassie stained gel using the “ProFinder 2D” software. Numbered circles indicate immunoreactive proteins specifically recognized by KC and KPC sera and identified by MALDI-TOF MS. Immunoreactive protein names are listed in Additional file 1: Table S1.

Figure 2

Individual KC and KPC serum reactivity against the identified antigens. The intensity of reactivity of each control Cre, KC (A) and KPC (B) serum against each MALDI-TOF MS identified protein is represented as a gray gradient scale of color as described in the legend. The volume (Vol) of each immunoreactive spot was calculated after background subtraction with the image analysis software “ProFinder 2D” and reported as arbitrary units (AU). For proteins represented from more than one spot the volume was expressed as mean AU value.

Figure 3

Antigen validation in resectable PDAC patients. (A) The graph shows the frequency of autoantibodies against mouse and human common immunoreactive antigens in the group of resectable patients who underwent surgery with curative intent (n = 38), analyzed by SERPA against CF-PAC-1 cell line 2DE map. P-values were calculated vs. control frequencies listed in Table 3 by Fisher's exact test (** P < 0.005). (B) Proteins were extracted from eight frozen PDAC tissues from surgically-treated patients (stage IIA and IIB), separated by 2DE, transferred to a nitrocellulose membrane and probed with the autologous serum. A representative Western blot is shown; circles indicate the presence of autoantibodies against the mouse and human common immunoreactive antigens.

Figure 4

Diagnostic performance of EZR-autoantibodies captured by ELISA. (A) Scatter plots show the reactivity of PDAC (n = 69), healthy subject (HS, n = 45), non-PDAC (n = 28), autoimmune disease (AD, n = 12) and chronic pancreatitis (CP, n = 37) patient sera to EZR recombinant protein as assessed by ELISA: PDAC vs. HS, non-PDAC and CP P < 0.0001; PDAC vs. AD P = 0.0006. (B) Scatter plots show the reactivity of prediagnostic PDAC patient (n = 16) and matched control (n = 32) sera from the EPIC cohort to EZR recombinant protein as assessed by ELISA: PDAC vs. controls P = 0.0002. Reactivity is expressed as optical density (O.D.) read at 450 nm, P-values were calculated by Student's t-test. (C) ROC analysis of EZR-autoantibody sensitivity and specificity using O.D. obtained in ELISA as a continuous variable (cut-off value: O.D. = 0.1183). (D) Classification and regression tree (CART) analysis of CA19.9 serum levels (≥ 37 IU/ml), EZR-autoantibody reactivity (O.D. ≥ 0.1183) and ENOA1,2-autoantibody reactivity (expressed as 2DE WB positivity) with 93 PDAC patients and controls where all parameters were available. The number and percentage of PDAC patients and controls are shown for each node. (E) ROC analysis of sensitivity and specificity of EZR-autoantibody detection in combination with CA19.9 and ENOA1,2-autoantibodies in the cohort of samples where all three parameters were available (PDAC patients: n = 45; benign controls: HS, AD, CP, n = 48). The applied diagnostic algorithm assigns patients to the PDAC group when both EZR-autoantibodies and CA19.9 are positive, and separates discordant cases into PDAC or controls based on the presence or absence of ENOA1,2-autoantibodies.