Protein S100A8/A9: A Potential New Biomarker for Pancreatic Diseases

Objectives: S100A8/A9 expression has been linked to carcinogenesis and infl ammation. We hypothesized that S100A8/A9 protein serum levels are a useful stratifi cation marker for patients with pancreatic ductal adenocarcinoma (PDAC), intraductal papillary mucinous neoplasm (IPMN) or chronic pancreatitis (CP). Methods: S100A8/A9 serum levels were analysed in PDAC, CP and IPMN patients and compared to S100A8/A9 healthy donor controls (HD) using ELISA. S100A8/A9 levels and clinical data were statistically analysed. Results: Out of 134 patients included, 84 were diagnosed with PDAC (46%), 30 patients with IPMN (16%) and 20 patients with CP (11%), 50 patients were HD (27%). When compared to HD (343.3 ng/ml), S100A8/A9 serum concentration was elevated in PDAC (402.0 ng/ml, p = 0.001) and CP patients (426.51 ng/ml p < 0.001). Also, S100A8/A9 levels were elevated in PDAC compared to IPMN group (369.0 ng/ml p = 0.026) and in CP compared to IPMN group (p = 0.001). A multivariate model including age, gender, leukocyte levels, C-reactive protein (CrP), S100A8/A9 and CA 19-9 concentrations reported a diagnostic sensitivity of 74.8%. Conclusion: S100A8/A9 serum levels are increased in patients with PDAC, CP, and IPMN and might be useful to distinguish malignant and infl ammatory diseases from normal and non-malignant pathological conditions. Research Article Protein S100A8/A9: A Potential New Biomarker for Pancreatic Diseases Alexander T El Gammal1*, Jörn H Sturm1*, Hans O Pinnschmidt2, Bianca T Hofmann1, Eugen Bellon1, Tarik Ghadban1, Nathaniel Melling1, Kai Bachmann1, Jakob R Izbicki1, Maximilian Bockhorn1, Cenap Güngör1# and Daniel R Perez1# 1Department of General, Visceral, and Thoracic Surgery, University Medical Center HamburgEppendorf, Hamburg, Germany 2Department of Medical Biometry and Epidemiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany *Shared fi rst authorship #Shared senior authorship Dates: Received: 29 July, 2017; Accepted: 19 September, 2017; Published: 19 September, 2017 *Corresponding author: Alexander T El Gammal, Department of General, Visceral, and Thoracic Surgery, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany, Tel: +49 152 22817908; Fax: +49 40 7410 44995, E-mail: https://www.peertechz.com


Introduction
Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal cancers worldwide and is still substantiated by insuffi cient diagnostic tools and therapeutic options. The overall 5-year survival rate among PDAC patients is less than 5%, which is partially due to an almost symptomless progression. Today, PDAC is ranked as the 4 th leading cause of cancer related death worldwide.
The S100 protein family is the largest subgroup of the Ca 2+binding EF-hand (helix E-loop-helix F) proteins [1]. First being identifi ed by Moore in 1965 [2], S100 proteins are expressed in a wide range of tissues featuring multiple cellular functions such as contraction, motility, cell growth, differentiation, and secretion [1,3]. As a member of the S100 family, S100A8/A9 is composed of two subunits, S100A8 and S100A9, both forming heterogenous multimers. The S100A8/A9 protein expression is primarily found in cells of the innate immune response [4][5][6], as well as in infl ammatory endothelial and epithelial cells [7,8].

Statistical analysis
Metric variables were tested for normality via Kolgomorov-Smirnov and Shapiro-Wilk tests. Their distributions were also assessed via histograms and boxplots. Means, medians and 1 st and 3 rd quartiles (interquartile range; IQR) of metric variables were reported. The variables CrP and CA 19-9, which displayed heavily right-skewed distributions, were ln-transformed (ln[x+1]) prior to further analyses. The means of variable S100A8/A9 were compared across diagnostic groups using student's t-test. Variables CA 19-9, CrP and leukocytes had about 27%, 8% and 6% missing values, respectively. Multiple imputations were therefore performed using the monotone method for data having a monotone pattern of missing and the Markov chain Monte Carlo method for data with nonmonotone pattern of missing, thus yielding 10 imputed data sets. Diagnosis, as a dependent nominally scaled variable consisting of 3 categories (0= PDAC 1=IPMN and 2=CP), was then subjected to multinomial logistic regression modelling, using the variables ln(CA 19-9+1), S100A8/A9, ln(CrP+1), leukocytes, age and gender of the 10 imputed data sets as independent variables (covariates). All independent variables were tested in univariate models as well as in multivariate models. The multivariate approach started with an initial model containing all main effect terms of the independent variables plus all their two-way interaction terms. Non-signifi cant terms were then removed, based on the p-values of likelihood-ratio tests ( = 0.05), following a stepwise hierarchical backward elimination procedure sensu Kleinbaum & Klein [40]. However, the main effect terms of all independent variables were kept in the fi nal multivariate model, even if they were not signifi cant. Modelpredicted diagnosis was cross-tabulated versus observed diagnosis for comparison of models and Nagelkerke's pseudo R 2 was computed to indicate strength of association and model fi t. All statistical analyses were done using SPSS version 22. In 73% (61/84) of the investigated PDAC patients, the pT No difference was detected in both PDAC and CP patient serum levels. Also, there was no signifi cant difference between HD and IPMN S100A8/A9 serum levels ( Table 1,3).

Results
Additionally, we investigated a possible relation between serum S100A8/A9 protein levels and histological grading, TNM-staging, survival time, serum CA 19-9 levels, leukocytes and CrP as well as patients' age and gender.
We found that leukocytes and CrP serum levels correlate with S100A8/A9 protein serum levels on a statistically signifi cant level (p < 0.01, data not shown).
However, no statistically signifi cant difference was found between well (G1-G2) and poorly (G3) differentiated cancers as well as between T3 and T4 staging and lymph node status.
Statistical analyses and correlation of S100A8/A9 protein levels with survival time, CA 19-9, patients' age or gender revealed no correlation at all.
To determine whether S100A8/A9 protein levels might be suitable as a possible stratifi cation marker for pancreatic cancer, we chose a panel of four serum markers (Table 4).
The fi nal multivariate model after removal of redundant  variables yielded 74.8% correctly classifi ed diagnoses. In comparison, a univariate model based on ln(CA19-9 + 1) alone yielded 61.3% correctly classifi ed cases (Table 4). Nagelkerke's pseudo-R 2 indicated that only the multivariate model fi ts the data relatively well.

Discussion
The current study reports for the fi rst time serum protein levels of S100A8/A9 in PDAC and IPMN patients. Members of the S100 protein family are frequently up regulated in various types of autoimmune diseases such as rheumatoid arthritis, psoriasis or infl ammatory bowel disease, but also in numerous cancer types [5,9,10,33].
In order to evaluate the diagnostic value of S100A8/A9 in pancreatic tumors and as an easy to obtain resource in clinical practice, patient serum samples were analysed. Elevated protein levels in PDAC sera as well as in sera obtained from patients diagnosed solely with CP were detected. While there was no signifi cant difference in S100A8/A9 between PDAC and CP, S100A8/A9 levels were elevated in PDAC, compared to IPMN patients and HD.
Chen et al. previously found an overexpression of S100A8/ A9 protein levels in pancreatic main duct fl uid and reported that its concentration correlates with median survival [34]. In the present study S100A8/A9 serum level was not associated with outcome. Statistically signifi cant correlations existed between S100A8/A9 protein levels and CrP as well as leukocyte levels, however, this is not surprising since S100A8/A9 protein is known to be an infl ammatory-linked protein.
In colorectal cancer, Kim et al. identifi ed S100A8/A9 protein plasma levels to be more specifi c and sensitive compared to the established tumor marker carcinoembryonic antigen CEA [26]. Also, in prostate cancer, Hermani et al. reported serum levels of S100A9 to be more sensitive than PSA when discriminating between prostate cancer and benign prostate hyperplasia [31]. Likewise, S100A8 and S100A9 serum levels were recently identifi ed as potential biomarkers for renal cell cancer earlydetection [48].
Our multivariate multinomial logistic regression model (containing the biomarkers leukocytes, CrP, CA 19-9 and S100A8/A9 protein levels, gender and age), performed better in estimating different histologic subtypes of pancreatic lesions than any univariate model, yielding about 75% correctly estimated diagnoses. In contrast, the CA 19-9-based model correctly classifi ed only about 61% of the diagnoses.
The S100A8/A9 protein has been identifi ed to be a potent chemoattractant for myeloid-derived suppressor cells (MDSC) [49][50][51][52][53]. While MDSC seem to be the main source of S100A8/A9 protein in cancer patients, it remains unclear to which degree malignant cells express the S100A8/A9 protein.
Immunohistochemical studies of pancreatic cancer tissue identifi ed the S100A8/A9 protein as exclusively expressed in myeloid cells infi ltrating the tumor-stroma [36], although several pancreatic cancer cell lines are already known to express S100A8/A9 [47,54,55].
In conclusion, our study identifi ed increased S100A8/A9 expression levels in patients suffering from PDAC, CP and IPMN. Also, the correlation of S100A8/A9 protein serum levels with commonly available clinical data revealed that it could help stratifi cation to distinguish patients with malignant and/  or infl ammatory disease from normal and non-malignant pathological conditions. The combined quantifi cation of S100A8/A9 and CA-19-9 serum levels in patients points to a higher sensitivity for diagnosis. The latter, regarding the crux of late pancreatic cancer diagnosis, might be another important and necessary step towards early cancer detection.