Expression of CAMK1 and its association with clinicopathologic characteristics in pancreatic cancer

Calcium/calmodulin‐dependent protein kinase (CAMKs) can control a wide range of cancer‐related functions in multiple tumour types. Herein, we explore the expressions and clinical significances of calcium/calmodulin‐dependent protein kinase 1 (CAMK1) in pancreatic cancer (PC). The expression of CAMK1 in PC was analysed by Gene Expression Profiling Interactive Analysis 2 (GEPIA 2) database and the Oncomine database. For further validation, the protein level of CAMK1 in PC tissues was also detected in the Human Protein Atlas (HPA) database and the tissue microarray (TMA)‐based immunohistochemistry (IHC). GEPIA 2 and Kaplan‐Meier Plotter (KM Plotter) databases were used to explore the prognostic significances of CAMK1 in overall survival (OS) and disease‐free survival (DFS) of PC at mRNA level. The relationship between CAMK1 expression and the clinicopathological characteristics of PC was further explored. Additionally, the Search Tool for the Retrieval of Interacting Genes (STRING) database was used to analyse protein‐protein interactions (PPI). We found CAMK1 was highly expressed in PC both in bioinformatics analyses and TMA‐IHC results. The prognostic analyses from the public databases also showed consistent results with follow‐up data. The PPI network suggested that CALM1, CALM3, CREB1, CALM2, SYN1, NOS3, ATF1, GAPDH, PPM1F and FBXL12 were important significant genes associated with CAMK1. Our finding revealed CAMK1 has prognostic value in PC patients, suggesting that CAMK1 may has a distinct role in PC patients and can be used as a candidate marker for investigating clinical prognosis of PC.

only chance for cure, less than 20% of patients are even surgical candidates. 5 In addition, despite the completion of surgical resection and adjuvant chemotherapy, nearly 60% of patients relapse within 2 years after surgery. 6 CAMKs are serine/threonine kinases that are activated by increased intracellular calcium concentration and can mediate subsequent cell activity. Ca2 + binding greatly changes the conformation of CaM and increases its affinity for some CaMKs including CaMKK, CaMKI, CaMKII and CaMKIV. These CaM kinases are widely expressed and can participate in a variety of cancer-related functions. 7 Their potential in anti-cancer treatment interventions has gradually begun to receive attention. It was reported that targeting Ca2 + signalling may provide therapeutically useful options, such as inducing epigenetic reactivation of tumour suppressor genes in cancer patients. 8 CaMKI family consists of 4 members including CaMKIα, CaMKIβ/Pnck, CaMKIγ/CLICK3 or CaMKIδ/CKLiK, which are coded for by CAMK1, PNCK, CAMK1G and CAMK1D, respectively. 7 CAMK1 is known to play important roles in Ca2 + signalling pathways and it is also involved in multiple cell functions, including ATP binding, signal transduction, cell differentiation, et al 9 Despite the importance of CAMK1 in cell functions, it is also faced with some intriguing questions and challenges in tumour field. In this present study, we plan to illustrate the presence and importance of CAMK1 in PC through bioinformatic mining analysis and the samples presented in TMAs.

| Bioinformatics mining methods
The GEPIA 2 database (http://gepia2.cance r-pku.cn) could analyse the gene expression profiles from the Cancer Genome Atlas (TCGA) dataset and the Genotype-Tissue Expression (GTEx) projects. The expression level of one gene in different types of cancer could be achieved by Boxplot. 10 We identified the expression levels of CAMK1 in PC based on TCGA normal and GTEx data.
The cut-off value of log2FC was set as 1, and P value was set to 0.01. Next, Oncomine (www.oncom ine.org), a cancer microarray database and integrated data-mining platform, 11,12 was used to compare CAMK1 expression in PC tissues with that in normal tissues. In this study, we chose mRNA levels of cancer vs. normal patient datasets, 1.5-fold change and P value = 0.01 as threshold.
We also retrieved the data from the HPA database (http://www. prote inatl as.org). The HPA database was made available freely to provide the expression profiles at protein levels, as well as IHC images for a wide variety of cancer tissues. In the HPA database, genome-wide transcriptomics data and clinical metadata of almost 8000 patients were used in order to analyse the proteome of 17 major cancer types. The IHC analysis in the HPA database is also presented for many protein-coding genes in respective cancer patients, the antibody information used for each IHC analysis can also be obtained in the HPA database. The IHC score is mainly classified into strong, moderate, weak and negative based on the staining intensity and fraction of stained cells. 13,14 Furthermore, we also used KM Plotter database (http://kmplot.com/analysis), an online database is capable to assess the effect of any gene on survival in cancer patients 15 and GEPIA 2 database to evaluate the OS and DFS of PC patients. In order to assess the prognostic values of CAMK1, the patient samples were divided into two cohorts based on the median expression (high expression and low expression) of CAMK1. CAMK1 was uploaded respectively to obtain the survival plots, in which Logrank P value and hazard ratio (HR) with 95% confidence intervals(CI) were calculated and showed on the webpage.

| Tissue microarray construction
TMA is a high throughput tool that allows hundreds of tissue samples to be analysed quickly, and conveniently, this method allows all tissue samples in an experiment to be analysed under standardized conditions. In our study, each TMA was constructed in the way described many times before. 16 The sections were placed on slides coated with 3-aminopropyltriethoxysilane. The non-cancer tissue samples were taken at a distance of > 3 cm from the tumour margin. All these human tissue samples were obtained with appropriate bioethics approvals and informed consents. Diagnoses of PC were confirmed on the basis of pathological evidence. These PC patients had not received any preoperative anti-cancer therapy before surgery.
All clinicopathological features of these PC patients were provided (Table 2), and tumour differentiation grades and clinical stages were classified based on the 7th American Joint Committee on Cancer (AJCC) TNM classification. A pathologist participated in reviewing the process.

| Immunohistochemistry
Immunohistochemistry technology can detect antigens in tissue sections through immunological and chemical reactions, and this technique has high sensitivity and specificity and can detect a variety of antigens in tissue. 17 We placed the para°ffin-coated microarray sections on a 60°C heating block for 30 min and continuously washed with xylene. The slides were rehydrated in different concentrations of alcohols and boiled in a pressure cooker containing 6.5 mm sodium citrate buffer to restore the antigen. 18 Then, we used 3% hydrogen peroxide to block the endogenous peroxidase activity for about 30 min at room temperature. Pre-incubate the slides with bovine serum albumin (BSA) in 0.1-mM Tris-buffered saline (TBS) for 2 hour to reduce non-specific background. Then we used rabbit monoclonal CAMKI antibody (ab68234, abcam) diluted 1:1000 in BSA to incubate slides at 4°C overnight. After incubation with antibodies and BSA, we rinsed the slides with 0.05% Tween-20 three times, 5 min each time and secondary antibody was used to incubate with the slides for 2 h at room temperature. The slides were developed in diaminobenzidine solution and stained with haematoxylin. 3 representative fields of each case were collected by Leica Aperio Image Scope software to ensure homogeneity and representativeness. The immunoreactivity score (IRS) assessments of CAMKI were performed by two independent pathologists without knowing the clinical pathological data. The immunohistochemical staining results were considered both the intensity of staining and the score for positive area. The scoring criteria for staining intensity were as follows: 0(negative), 1(weak), 2 (moderate) and 3 (strong). The criteria for the score for positive area were 0 (<10%), 1 (11-25%), 2 (26-50%), 3 (51-75%) and 4 (76-100%). Then the final expression score was calculated as the staining intensity score × positive area score, ranging from 0 to 12. A total score of 6 or higher were grouped as high expression group, and less than 6 was grouped as low expression group. The above criteria for the score were performed according to a previously described published literature. 19

| PPI network construction and KEGG pathway analysis
STRING database (https://strin g-db.org/cgi/input.pl) can collect and integrate known and predicted protein-protein association data. The associations in STRING database include direct (physical) interactions and indirect (functional) interactions, as long as both are specific and biologically meaningful. 20 The identification and characterization of protein-protein interactions will be necessary to better understand the functions and efficacy of CAMK1.
In this study, we used STRING database to construct PPI network of CAMK1 with minimum required interaction score 0.7 and the interaction predictions were mainly derived from textmining, experiments, databases, co-expression and co-occurrence, et al The KEGG pathway analysis was also constructed by STRING database.

| Statistical analysis
The CAMK1 expression levels between PC tissue and normal tissue were evaluated by the GEPIA 2 and the Oncomine database.

| CAMK1 was highly expressed in pancreatic cancer in bioinformatics database
The GEPIA 2 database was used to determine CAMK1 expression in PC and normal tissues. This results showed that CAMK1 expression was higher in PC tissue (red box) than normal tissue (grey box )(*P < .01, Figure 1). Next, the expression of CAMK1 was further validated  Figure   S1C). We also yielded and analysed IHC data for PC tissues from the HPA database, and the CAMK1 staining showed moderate to strong cytoplasmic immunoreactivity in most PC tissues (Antibody HPA051409) ( Table 1).

| Predicting the prognostic values of CAMK1 in pancreatic cancer based on GEPIA 2 and KM Plotter database
To

| Independent validation of prognostic value of CAMKI by TMA-based IHC
Considering the results of prognostic value of CAMK1 in database, we further validated the prognostic value of CAMK1 expression by using TMA-based IHC in 90 paired PC tissues and P=.002, respectively), but not CAMK1 ( Figure 4C). Furthermore, the association between clinicopathological variables and CAMK1 immunostaining was also analysed using Pearson's chi-square test (

| PPI network and KEGG pathway analysis
The PPI information about CAMK1 can be evaluated by STRING da-  Table 1). The candidate genes in these pathways included CALM1, CREB1, ATF1 and NOS3, and they were all up-regulated in PC (P<.05, Figure S2).

| D ISCUSS I ON
Calcium is a widespread second messenger which controls vari-  In conclusion, it seemed that CAMK1 might be a promising biomarker for a better prognosis in PC patients, although the potential effect of CAMK1 expression on the biological function of PC and the reason for better prognosis remains to be further investigated.
The PPI data only provided potential probabilities for interactions between genes based on different sources of information, and the underlying molecular mechanisms of CAMK1 in PC would be further explored.

This work was supported by the Ring Chang Special Fund of Shanghai
Charity Foundation (Q2015-024).

CO N FLI C T O F I NTE R E S T
The authors declare that they have no known competing financial interests or personal relationships that would influence the work reported in this paper. Writing-review & editing (equal).

DATA AVA I L A B I L I T Y S TAT E M E N T
Some publicly available datasets were analysed in this study. The authors confirm that these data can be found here: http://gepia2.