Serum calretinin as an independent predictor for platinum resistance and prognosis in ovarian cancer

Calretinin (CRT) is a calcium‐binding protein that controls intracellular calcium signaling. Besides its prominent expression in neurons, serum CRT (sCRT) has recently been suggested as blood‐based biomarker for prediagnostic mesothelioma detection. CRT is expressed in ovarian cancer tissues in up to 40% of cases; however, its clinical relevance as blood‐based biomarker for ovarian cancer is unknown. sCRT was determined by calretinin enzyme‐linked immunoabsorbent assay (Calretinin‐ELISA, DLD Diagnostika GmbH, Hamburg, Germany) in a total of 515 serum samples from 116 healthy controls and 134 ovarian cancer patients (thereof 86% with Fédération Internationale de Gynécologie et d'Obstétrique [FIGO] III/IV), including samples at primary diagnosis and at four longitudinal follow‐up time points in the course of treatment and at recurrence. sCRT level was significantly increased in ovarian cancer patients compared to healthy controls (estimated difference = 0.3 ng/ml, p < 0.001), was mostly independent from CA125 (r ≤ 0.388) and enabled accurate discrimination between cases and controls (area under the curve = 0.85). Higher sCRT level at primary diagnosis predicted suboptimal debulking (p < 0.001) and was associated with advanced FIGO‐stage (p < 0.001) and increased amount of ascites (p < 0.001). sCRT levels at primary diagnosis and its dynamics in the course of chemotherapy were independent predictors for poor progression‐free survival (hazard ratio [HR] = 1.99, confidence interval [CI] = [1.13–3.52], p = 0.0181) and overall survival (HR = 15.4, CI = [1.92–124], p = 0.0099). Furthermore, sCRT at primary diagnosis or a relative sCRT increase in the time interval between surgery and the onset of chemotherapy were both independent predictors of platinum resistance (OR = 4.99, CI = [3.50–16,001], p = 0.0016; OR = 2.41, CI = [1.37–6,026], p = 0.0271, respectively). This is the first study that suggests sCRT as liquid biopsy marker for independent prediction of platinum resistance and prognosis.


Introduction
Epithelial ovarian cancer is the leading cause of death among patients with gynecologic malignancies. At primary diagnosis, more than 70% of ovarian cancer patients present with advanced disease. 1 Standard treatment of advanced ovarian cancer consists of primary radical debulking surgery aiming at macroscopic complete tumor resection followed by platinum-and paclitaxel-based chemotherapy, which extends progression-free survival (PFS) and overall survival (OS). 2,3 Postoperative residual tumor burden is one of the most important prognostic factors in advanced ovarian cancer. 4 Despite improved primary radical surgery and the implementation of innovative targeted therapies into standard treatment, such as anti-angiogenetic therapy with Bevacizumab or PARP-inhibitors, ovarian cancer patients still have a poor overall prognosis. 5,6 Patients initially respond well to platinum-based chemotherapy with response rates up to 80%; however, virtually all patients develop platinum resistance over time, that is, the probability of acquiring resistance continuously increases with every event of clinical relapse. Moreover, one of the most relevant clinical challenge is that about 15-20% of patients initially do not respond to platinum drugs and are therefore considered to have primary platinum resistant tumors. [7][8][9] Considering this clinical challenge, the identification of innovative blood-based biomarkers predicting prognosis and response to platinum-based chemotherapy is highly desirable.
Calretinin (CRT) was originally discovered in 1987 and was named based on its structural similarity with calbindin-1 and its initial site of characterization, the chicken retina. 10 CRT belongs, together with calbindin and secretagogin, to the "EFhand family" of calcium binding proteins, which contain a helix-loop-helix motif that can act as calcium-binding site. 11 Human CRT has a molecular weight of approximately 31 kDa and is predominantly located in the cytosol. It contains six helix-loop-helix domains (also referred to as EF-hand motifs), five of which are able to bind Ca 2+ ions. 12 Cellular functions of CRT are closely related to calcium signaling. CRT can act as a calcium buffer, a calcium sensor or a modulator of neuronal excitability. 13,14 It has a broad tissue distribution and is expressed in a subset of neurons of the central and peripheral nervous system. 12,15 However, physiological CRT expression has been observed in nonneural cells, particularly mesothelial cells, but also in adipocytes or ovarian stroma and during embryonic development. [15][16][17] CRT is expressed in several epithelial and nonepithelial tumor entities, especially in mesotheliomas, in which it is used as an immunohistochemical marker to differentiate mesotheliomas from adenocarcinomas. 18 The diagnostic impact of CRT for mesothelioma has recently been transferred to a liquid biopsy approach, in which plasma CRT was suggested as blood-based marker for mesothelioma detection, particularly in a prediagnostic setting. 19,20 Among malignancies of the female genital tract, CRT expression has commonly been reported in sex cord stromal tumors with an expression rate ranging from 40% up to 100%. 15 There is evidence that CRT is variably expressed in ovarian cancer, with highest prevalence among serous histotypes (0-40%) followed by endometrioid subtypes (0-33%) and clear cell carcinomas (0-14%). 15 In contrast to mesothelioma, clinical relevance of CRT as potential blood-based biomarker for ovarian cancer is completely unknown. Therefore, we interrogated (i) whether we can detect serum CRT (sCRT) in ovarian cancer patients at a level that significantly differs from healthy controls and (ii) whether sCRT level at primary diagnosis and in the course of surgery and adjuvant chemotherapy could potentially serve as a predictive (and/or prognostic) liquid biopsy marker for ovarian cancer.

Patient characteristics
The present study was conducted at the Department of Gynecology and Obstetrics at the Carl Gustav Carus University of Dresden, Technische Universität Dresden, Dresden, Germany. In total, 134 patients with histologically confirmed primary epithelial ovarian cancer and adjuvant treatment were included. Informed written consent was obtained from all patients, and the study was approved by the Local Research Ethics Committee in Dresden (EK74032013). The patients' clinical data are reported in Supplementary Table S1. Tumors were classified in line with the WHO-classification of tumors derived from female genital tract. Grading was conducted using the grading system established by Silverberg 21 and tumor staging was classified according to the Fédération Internationale de Gynécologie et d'Obstétrique (FIGO), 22 which was revised in 2014. 23 The revised version of the FIGO classification was used for all patients, who underwent primary surgery in 2014 or later. The whole study population received primary radical surgery aiming at macroscopic complete tumor resection followed by platinum-and paclitaxel-based adjuvant chemotherapy. Bevacizumab, which is approved for patients with a tumor stage of at least FIGO IIIb, was additionally administered to 91/134 patients (68%).

Healthy donors
In total, 116 female healthy individuals without any history of benign or malignant disease were recruited for obtaining control serum samples. Informed written consent was obtained from all donors, and the study was approved by the Local What's new? Calretinin is a calcium-binding protein with diagnostic implications in cancer, particularly in mesothelioma. This study shows that serum calretinin (sCRT) also has strong potential as a liquid biopsy marker in ovarian cancer. In ovarian cancer patients, sCRT level was found to be elevated at primary diagnosis. While sCRT levels declined following surgery and the initiation of platinum-based chemotherapy, baseline levels were reestablished upon disease recurrence. Moreover, sCRT level at primary diagnosis predicted both poor survival and platinum resistance. The findings support the incorporation of sCRT as an auxiliary marker in diagnostic and prognostic assessments of ovarian cancer.
Research Ethics Committee in Dresden (EK74032013). There was no correlation between sCRT and age as confirmed by our own data set (p = 0.15) and by previous studies. 20 Nevertheless, individuals for blood donation were selected in an age range matching the included ovarian cancer patients, with the majority of donors ranging between 50 and 75 years. Furthermore, serum preparation of control samples was performed with exactly the same protocol as for the patient samples to ensure comparability.

Serum preparation
After blood withdrawal with a 7.5 ml S-Monovette ® (Sarstedt AG & Co., Nuembrecht, Germany), blood samples were incubated at room temperature for at least 30 min to allow complete blood coagulation. Within 1 hr after blood drawing, serum was prepared by centrifugation for 8 min at 1,800g at room temperature. The obtained cell free serum fraction was immediately frozen at −80 C until further processing. Unnecessary freeze-thaw cycles were avoided. Samples were blinded so that neither time of blood drawing nor any other information was disclosed during the investigation. Samples were thawed on ice and were immediately processed after complete thawing.

Detection of sCRT
Concentrations of sCRT were determined as described by Johnen et al., using the commercially available Calretinin enzyme-linked immunoabsorbent assay kit by DLD Diagnostika GmbH. 19 The assay is based on polyclonal antibodies developed by Raiko et al. 20,24 All reagents and samples were equilibrated to room temperature (20-25 C) and the incubations were performed at room temperature (20-25 C) using a plate shaker. Serum samples (2 × 15 μl) were diluted 1:5 in the provided dilution buffer. The diluted samples were determined in duplicate. Details of the procedure are outlined in the manufacturer's instructions. Optical densities at 450 nm were measured by a photometer (Molecular Devices, Sunnyvale, CA). A standard curve was obtained by four-parameter curve fitting using SoftMax Pro 4.0 from Molecular Devices.

Statistical analysis
The statistical analysis was conducted with R, Version 3.5.1 and GraphPad Prism version 8.0.2 (GraphPad Software, La Jolla, CA). p-Values <0.05 were considered statistically significant. All confidence intervals (CIs) were specified as 95% CI. Nonparametric two-sided Mann-Whitney U-test was used to compare sCRT levels of ovarian cancer patients at primary diagnosis with those of healthy controls. Using receiver operating characteristic (ROC) curve analysis, we assessed the capacity of sCRT concentrations to discriminate between ovarian cancer patients and healthy controls. By Spearman's rank-order test, the correlation between sCRT and CA125 was evaluated. For analyzing sCRT level change in the course of treatment, two-sided Wilcoxon rank-sum tests for paired data were applied. Nonparametric two-sided Mann-Whitney U-tests were used to analyze the associations of sCRT and clinicopathological parameters. Uni-and multivariable Cox regression analyses were performed to study the prognostic relevance of sCRT in ovarian cancer. For all Cox regression analyses, respective hazard ratios (HRs) are indicated. In order to assess prognostic characteristics of sCRT for early recurrences, multivariable logistic regression analysis was applied and odds ratios (ORs) are provided. Firth's biasreduced logistic regression analysis was used to determine whether sCRT can predict platinum resistance. Additionally, Kaplan-Meier analyses and log-rank tests were performed. All models were tested for OS and PFS as separate outcome variables. The correlation between sCRT levels at primary diagnosis and at recurrence was assessed by Spearman's rank-order test. For all regression and Kaplan-Meier analyses, the cutoff determination for stratifying the patients either into a sCRT high or a sCRT low group (or sCRT-area under the curve [AUC]-high vs. sCRT-AUC-low) was conducted by Maximally Selected Rank Statistics. Cutoff selection was supported by a test for independence between the outcome variables OS/PFS and sCRT/sCRT-AUC.

Data availability
The article contains all relevant data, including the patient's clinicopathological parameters. The original set of raw data will be made available upon reasonable request.

sCRT level is elevated in ovarian cancer and declines in the course of treatment
We analyzed levels of sCRT in a comprehensive set of clinically documented ovarian cancer patients at primary diagnosis (n = 133) and compared it to healthy controls (n = 116). Moreover, we tracked sCRT in the course of primary surgery and adjuvant chemotherapy, represented by four additional longitudinal follow-up samples, obtained (i) one week after primary surgery (n = 68), (ii) before the onset of platinum-based chemotherapy (n = 76), (iii) after the third cycle of chemotherapy (n = 55) and (iv) after the completion of chemotherapy (n = 48).
sCRT was highly upregulated at primary diagnosis of ovarian cancer (estimated difference [ED] = 0.3 ng/ml, CI = [0.243-0.359], p < 0.0001 compared to healthy controls (Fig. 1). In patients with exclusively low-stage disease (FIGO I or II; n = 19), ROC curve analysis showed that sCRT at primary diagnosis enabled discrimination between ovarian cancer patients and healthy controls with an AUC of 0.67 (CI = [0.524-0.823]; Supplementary Fig. S1). In the total patient cohort, irrespective of FIGO stage, discrimination was possible with an AUC of 0.85 (CI = [0.803-0.900]; Fig. 2). Choosing 0.35 ng/ml sCRT as a diagnostic cut-off from ROC analysis, ovarian cancer was detected by sCRT assessment in the total cohort with a sensitivity of 74% and a specificity of 90% (Supplementary Table S2).
After surgery, there was a strong and highly significant decline of sCRT (ED = 0.22 ng/ml, CI = [0.124-0.315], p < 0.0001) to a level, which remained grossly stable in the course of platinum-based chemotherapy, however, which remained significantly higher than the sCRT levels of healthy controls. After the completion of chemotherapy, sCRT was still slightly increased compared to healthy controls (ED = 0.098, CI = [0.054-0.142], p < 0.0001; Fig. 1).
Conclusively, sCRT level is significantly elevated in serum of ovarian cancer patients at primary diagnosis, sharply discriminates ovarian cancer from controls and strongly declines in the course of surgery and adjuvant chemotherapy.
Higher sCRT level parallels advanced ovarian cancer and predicts suboptimal primary debulking To sum up, higher sCRT levels at primary diagnosis of ovarian cancer correlate with advanced disease and predict suboptimal primary debulking surgery without achieving macroscopically complete tumor resection.
High sCRT level at baseline and in the course of treatment is an independent predictor of poor survival in ovarian cancer In order to study the prognostic relevance of sCRT in ovarian cancer, we performed survival analysis at primary diagnosis and at all four follow-up time points in the course of surgery   Tables S3-S8). Prognostic relevance of sCRT, concluded from the analysis above, was additionally demonstrated by Kaplan-Meier analysis and the log-rank test (Fig. 4).
We subsequently performed multivariable Cox regression analysis with PFS or OS as separate outcome variables, including sCRT levels and established risk factors of ovarian cancer, that is, age at primary diagnosis, amount of malignant ascites, FIGO stage, tumor grading and residual tumor burden left after primary debulking. High sCRT levels at primary diagnosis were an independent predictor for a poor PFS (HR = 1.99, CI = [1.13-3.52], p = 0.0181). In addition, multivariable logistic regression analysis revealed that sCRT levels at primary diagnosis and before the onset of chemotherapy were independent predictors for increased risk of recurrence  Tables S3-S8).
In conclusion, high sCRT levels in the course of surgery and adjuvant platinum-based chemotherapy is an accurate and independent blood-based predictor for poor survival in ovarian cancer. Its strongest prognostic significance was observed immediately before the onset of chemotherapy.

Individual sCRT dynamic in the course of treatment is an independent predictor for poor survival
From 48/134 patients, a complete set of longitudinal serum samples was available. In this context, we interrogated, whether the individual dynamics of sCRT across these samples is of prognostic relevance. Therefore, assuming a linear and continuous change of sCRT values between the investigated time points, we plotted a dynamic curve for each patient by interconnecting sCRT values from all longitudinal blood samples. Considering the different time points of blood   Tables S10 and S11). Moreover, prognostic relevance of sCRT-AUC, as reported above, was additionally demonstrated by Kaplan-Meier analysis and the log-rank test (Fig. 5). According to multivariable Cox regression, the sCRT-AUC-high condition was an independent predictor for poor PFS (HR = 2.90, CI = [1.08-7.82], p = 0.035) and OS (HR = 7.38, CI = [1.17-46.6], p = 0.0336; Supplementary Tables S10 and S11).
Diagnostic capacity of the patient's individual sCRT dynamics, as reported above, could similarly be reproduced when only the first three time points of serum collection (primary diagnosis, one week after primary surgery and before the onset of platinum-based chemotherapy; available from 66/134 patients) were considered for sCRT-AUC calculation ( Supplementary Fig. S3, Tables S12 and S13).
We conclude that longitudinal sCRT values, considered separately at the given time points or in terms of individual sCRT dynamics, have a strong and independent prognostic relevance in ovarian cancer.
sCRT level in recurrent ovarian cancer From 16 patients, corresponding serum samples at primary diagnosis and at first recurrence of ovarian cancer were available. Interestingly, patients with a relatively higher baseline sCRT level experienced sCRT decline due to surgery and adjuvant chemotherapy and later recovered similarly higher sCRT levels in the presence of a recurrent tumor (Figures 6a and  6b). In 3/16 patients, a further sample at second recurrence was available. In two of those patients, sCRT at second recurrence again matched the respective baseline level at primary diagnosis (Fig. 6a). There was as strong correlation between sCRT level at primary diagnosis and at first recurrence (r = 0.834, CI = [0.567-0.943], p < 0.0001; Fig. 6c).
To sum up, we demonstrate that sCRT levels significantly drop in the course of primary surgery and platinum-based chemotherapy but individual baseline levels reestablish at recurrence.
sCRT level is an independent predictor for primary platinum resistance We interrogated, whether sCRT level (i) at separate individual follow-up time points or (ii) its dynamics in the course of surgery and chemotherapy may predict response to platinumbased chemotherapy (platinum resistance defined as PFS < 6 months). According to Firth's bias-reduced logistic regression analysis, we revealed that sCRT level at primary diagnosis was highly predictive for platinum resistance (univariate: OR = 32.  Table S14). Additionally, a relative sCRT rise  For comparison, there was no statistically significant association between CA125 values at primary diagnosis and platinum resistance.
Conclusively, we show that sCRT level at primary diagnosis and a postsurgery sCRT rise independently predict primary platinum resistance in ovarian cancer.

Comparison between sCRT and CA125
There was a weak correlation between sCRT and CA125 at primary diagnosis ( . Therefore, assuming that both markers provide independent and possibly complementary diagnostic information, we were particularly interested in sCRT level in patients with low CA125. In total, matched sCRT and CA125 values at primary diagnosis were available from 128/133 patients ( Supplementary Fig. S5). Of those, 20 patients had a relatively low CA125 value (<100 U/ml, 15th percentile of the CA125 range) and were selected for comparison. Interestingly, 50% of these patients (10/20) had a higher sCRT value than the majority of healthy controls (75%) and tended to have a higher FIGO stage.
Taken together, we report a low correlation between sCRT and CA125 at primary diagnosis and in the course of chemotherapy in ovarian cancer and identify a subgroup of patients with low CA125 but conspicuous sCRT level.

Discussion
In the present study, we systematically analyzed clinical relevance of sCRT in ovarian cancer and report as our key finding that sCRT level at primary diagnosis as well as its dynamics in the course of surgery and platinum-based chemotherapy are independent predictors for poor prognosis and platinum resistance.
sCRT was highly elevated at primary diagnosis of ovarian cancer and accurately discriminated cases from healthy controls (AUC = 0.85), which is comparable to a study on German and Australian mesothelioma patients, reporting on sCRT elevation in patient blood and a discrimination of cases and controls with an AUC of 0.83 and 0.91, respectively. 20 We additionally reported that sCRT allows, albeit with less discriminative power, detection of particularly early disease (FIGO I-II; AUC = 0.67). However, this analysis was based on only 19 patients and does not allow a statistically substantiated conclusion. Considering sCRT as potential screening marker for ovarian cancer, our results encourage future investigation with independent patient cohorts enriched by FIGO I-II ovarian cancer patients (in the present study 86% had FIGO III or IV). Such an early detection approach has already been published for mesothelioma, reporting that sCRT (in combination with serum mesothelin) enabled prediagnostic detection of mesothelioma with an AUC of 0.74. 19 The biology and origin of CRT release into the bloodstream remains an open question. Since immunohistochemical studies on serous ovarian cancer reported variable CRT positivity in 0-40%, 15 it is possible that CRT promotes growth and survival of ovarian cancer cells, as it has been shown in vitro for mesothelioma. 25 In this scenario, CRT could possibly be released by tumor cells themselves, either by active secretion or by passive release upon tumor cell apoptosis, which is consistent with the generally high apoptotic index of malignant tumors. 26 However, this is not necessarily the most likely way of interpretation, since CRT expression in epithelial ovarian cancer is by far lower than in sex cord stromal tumors (0-40% vs. up to 100%) and staining has often been described as rather focal than homogeneous. 15 Moreover, global gene expression analysis in ovarian/peritoneal serous carcinoma (O/P-SC) versus diffuse peritoneal malignant mesothelioma (DMPM) revealed that CRT was overexpressed in DMPM but not in O/P-SC, suggesting that CRT could be (along a potential gene-expression signature) a marker for differentiating DMPM from O/P-SC. 27 In breast cancer, for comparison, strong CRT expression was predominantly observed in high-grade tumors with basal-like phenotypes and paralleled poor prognosis. 28,29 Moreover, an alternatively spliced form of CRT (CRT-22k) was detected in serum of breast cancer patients. 30 Considering that CRT is also highly expressed in mesothelial cells of the peritoneum, 15,16 it is alternatively possible that elevated sCRT in advanced ovarian cancer patients is related to the tumor microenvironment within peritoneal metastases. In this context, increased CRT-release from mesothelial peritoneal cells could be due to either an "educational niche response," in which ovarian cancer cells reprogram mesothelial cells toward an active CRT secretion or it could be due to an unspecific release of CRT from apoptotic/necrotic mesothelial cells, which are superseded by ovarian cancer cells.
High sCRT level at primary diagnosis predicted advanced (platinum resistant) disease and poor overall prognosis. After primary debulking, we observed a strong sCRT decline, approaching the level of controls. We therefore conclude that sCRT could be a sensitive blood-based biomarker for assessing individual tumor load and the patient's initial risk profile at primary diagnosis, including platinum sensitivity status. Considering that higher sCRT level was predictive for inefficient primary debulking, the detection of this marker may additionally help to identify tumors with aggressive and complex growth patterns, which are more difficult to debulk. sCRT levels at primary diagnosis were mostly independent from CA125 and were elevated in a subset of patients with low CA125 and high FIGO stage, suggesting that sCRT is partly complementary to CA125 and could be useful as auxiliary tumor marker in ovarian cancer, particularly for those patients with CA125 failure.
The strength of our study is that we analyzed sCRT not only at primary diagnosis but also in a comprehensive serum set of four longitudinal follow-up samples per patient in the course of surgery, chemotherapy and at recurrence. We report that sCRT level provides independent prognostic information not only at primary diagnosis but also among longitudinal sampling in the course of surgery and chemotherapy. Therefore, sCRT could additionally be useful as blood-based biomarker for therapy monitoring, which identifies aggressive disease with high risk of recurrence and poor prognosis (particularly in patients with macroscopically complete resection). This hypothesis is further supported by the fact that sCRT predicted platinum resistance not only at primary diagnosis but also in patients, who showed a relative sCRT rise between surgery and the onset of chemotherapy. Therefore, postsurgery sCRT may indicate active disease, possibly associated with a platinum resistant phenotype, which is subsequently triggered by the directional selection pressure of chemotherapy, leading to the outgrowth of possibly resistant tumor cells. 31 However, all cutoff values for survival analysis, reported herein, are exploratory and need to be validated in independent patient cohorts. Furthermore, we reported, that individual baselines of sCRT at primary diagnosis dropped after surgery and chemotherapy. In most cases, sCRT levels recovered to this individual baseline at recurrence. This points to unknown and patient intrinsic characteristics of CRT release, which recovers after chemotherapy and obviously remains stable across tumor evolution. Overall, our finding suggests that sCRT could be useful as blood-based biomarker for monitoring (early) recurrence.

Conclusion
This is the first study reporting that serum CRT (sCRT) has strong potential as liquid biopsy marker for ovarian cancer with a wide field of potential clinical application, particularly independent prediction of primary platinum resistance and prognosis. Along our previous liquid biopsy approaches in ovarian cancer, suggesting innovative biomarkers concepts on the level of circulating tumor cells or microvesicle-associated microRNAs, 32,33 our results show clinical relevance of CRT as a single and easily detectable serum parameter, which could straightforwardly be implemented into routine diagnostics as a CA125 auxiliary tumor marker for improving personalized medicine in ovarian cancer. High-risk patients, as identified by sCRT, could potentially benefit from an intensified therapy regime, including PARP inhibitors or immunotherapy.