Carboplatin with Decitabine Therapy, in Recurrent Platinum Resistant Ovarian Cancer, Alters Circulating miRNAs Concentrations: A Pilot Study

Objective Plasma miRNAs represent potential minimally invasive biomarkers to monitor and predict outcomes from chemotherapy. The primary goal of the current study—consisting of patients with recurrent, platinum-resistant ovarian cancer—was to identify the changes in circulating miRNA concentrations associated with decitabine followed by carboplatin chemotherapy treatment. A secondary goal was to associate clinical response with changes in circulating miRNA concentration. Methods We measured miRNA concentrations in plasma samples from 14 patients with platinum-resistant, recurrent ovarian cancer enrolled in a phase II clinical trial that were treated with a low dose of the hypomethylating agent (HMA) decitabine for 5 days followed by carboplatin on day 8. The primary endpoint was to determine chemotherapy-associated changes in plasma miRNA concentrations. The secondary endpoint was to correlate miRNA changes with clinical response as measured by progression free survival (PFS). Results Seventy-eight miRNA plasma concentrations were measured at baseline (before treatment) and at the end of the first cycle of treatment (day 29). Of these, 10 miRNAs (miR-193a-5p, miR-375, miR-339-3p, miR-340-5p, miR-532-3p, miR-133a-3p, miR-25-3p, miR-10a-5p, miR-616-5p, and miR-148b-5p) displayed fold changes in concentration ranging from -2.9 to 4 (p<0.05), in recurrent platinum resistant ovarian cancer patients, that were associated with response to decitabine followed by carboplatin chemotherapy. Furthermore, lower concentrations of miR-148b-5p after this chemotherapy regimen were associated (P<0.05) with the PFS. Conclusions This is the first report demonstrating altered circulating miRNA concentrations following a combination platinum plus HMA chemotherapy regiment. In addition, circulating miR-148b-5p concentrations were associated with PFS and may represent a novel biomarker of therapeutic response, with this chemotherapy regimen, in women with recurrent, drug-resistant ovarian cancer.


Introduction
Each year, over 200,000 women are diagnosed with ovarian cancer worldwide; more than 21,000 of those are American. Approximately 15,000 American women die from epithelial ovarian cancer (EOC) every year; of the various types of ovarian cancer, high grade serous ovarian cancer is the most common (HGSOC) [1,2]. For HGSOC identification, a wellestablished serum biomarker is the cancer antigen 125 (CA-125). Although CA-125 identifies most late stage primary and relapsed disease and largely correlates with disease burden, CA-125, in terms of sensitivity and specificity, is not predictive of the risk of relapse or chemotherapeutic response [2,3]. In general, EOC biomarkers are needed, and circulating miRNAs, which have associated with disease outcome [4], may represent a novel source of clinically informative ovarian cancer biomarkers.
MicroRNAs (miRNAs) are short RNAs differentially expressed in many disease states, including cancer. They alter the target protein synthesis by regulating post-transcriptional gene expression [5][6][7]. miRNAs are released into the plasma by tissues and tumors [6,8]. miRNAs are relatively heat stable in both whole blood and plasma [9,10]. Circulating miRNAs have been associated with early detection and prognosis of EOC [11,12] and may predict therapeutic response.
First-line EOC treatments utilize platinum based chemotherapies. These chemotherapies have been shown to change, within tissues, miRNA expression [13][14][15][16]. In this study, plasma miRNA concentrations were measured from a previously completed phase II clinical trial [17]. In this trial, 17 subjects with platinum resistant recurrent ovarian cancer were tested for efficacy and tolerability of a combination regimen of low dose decitabine (5-aza-2'-deoxycytidine; a DNA methyltransferase inhibitor) and carboplatin. The main objective of the clinical trial was to test the hypothesis that treatment with decitabine would resensitize ovarian tumors to carboplatin. The trial was based on previous preclinical studies indicating that reversal of DNA hypermethylation induces re-expression of tumor suppressor and differentiation associated genes that would restore sensitivity to platinum [18]. The results of the trial demonstrated prolonged progression free survival (PFS of~10 months) and a promising rate of objective response rates of 35% [17]. Taking advantage of the clinically annotated plasma samples collected for this trial, we determined in response to decitabine followed by carboplatin chemotherapy the changes in circulating miRNAs concentrations. Secondarily, miRNA changes were associated with clinical response.

Patient population
The open label phase II clinical trial and response rates were previously described by Matei et al [17]. Briefly, all patients had platinum resistant EOC and all but one patient had measurable disease as defined by Response Evaluation Criteria in Solid Tumor (RECIST). The patient with non-measureable disease had a CA-125 greater than twice the upper limit of normal and ascites. All patients in this study were older than 18 years, had greater than 3 months life expectancy, and had an Eastern Cooperative Group performance status of 0 or 1. Patients could have received unlimited number of prior cytotoxic therapies; however, they were excluded if they had a prior history of brain metastases, an allergy to carboplatin, uncontrolled medical issues, or grade 2 or higher neuropathy. All patients signed informed consent, and the protocol was approved by the Indiana University Institutional Review Board.

Treatment plan
Each chemotherapy cycle lasted 4 weeks and consisted of decitabine (Eisai, Tokyo, Japan) given intravenously at 10 mg/m 2 daily for 5 days; followed by, carboplatin (Bristol Meyers Squibb, New York, New York) given intravenously at an AUC5 on day 8 [17]. Additionally, peg-filgastrim (Amgen, Thousand Oaks, CA) was given on day 9 to prevent prolonged myelosuppression. Treatment was continued until disease progression or intolerable toxicity.
RNA extraction from plasma miRNAs were extracted from the plasma following a previously published protocol [9]. Briefly, plasma samples were thawed on ice to~4°C, and centrifuged at 1500 RCF, 4°C for 10 minutes. Fifty μL of plasma was extracted following the Qiagen miRNeasy plasma protocol (Qiagen, Valencia, CA). Following the Qiazol lysis step, C. elegans miR-39 miRNA (cel-miR-39) was added (5 μL of 5 nM stock) as an internal control for extraction efficiency and normalization. Additionally, 0.3 μL of a stock (0.8 μg/μL) Bacteriophage M2 RNA was added per 200 μL Qiazol buffer to increase miRNA yield. miRNAs were eluted from the purification column twice in 30 μL RNase/DNase free water (total 60 μL) and stored at -80°C.

Statistical analysis
In this pilot study, the primary outcome was the analysis of the difference in plasma miRNA concentrations between the before and after chemotherapy treatment. All values were normalized to the cel-miR-39 spiked in control. Statistical significance (p-value) in the comparisons of groups was determined by paired T-test. In Table 1, the baseline comparison was two tailed Ttest. The heatmaps were completed in Partek software using the delta delta C T (ddC T ) values. The following parameters were used for the heatmap: no scale; the distance vector = Spearman rank correlation; and the clustering = average. Of note, 1 is no fold-change, while 2 is a doubling of fold change.

Kaplan Meier curve
Progression free survival (PFS) analysis was conducted with "survival" package in R. Samples were divided into 2 groups by the median value of the "-delta delta CT" values. The log-rank test was performed in R to assess the association between miRNA concentration (high vs. low) with the time to progression, and p-value less than 0.05 was considered significant. Kaplan-Meier curves were used to visually examine the cumulative probability of remaining progression free over time for miRNA sub-groups.

Plasma sample selection
Briefly, whole blood was collected in EDTA containing tubes, and then centrifuged to plasma. Samples used for microRNA analysis were collected at two time points: at baseline (prior to the start of the first cycle of treatment); and on day 29 after treatment initiation (post the first cycle of chemotherapy; Fig 1). Samples from 14 of the 17 patients were available for this miRNA analysis. The remaining 3 samples were not analyzed because of insufficient plasma. Samples were analyzed after the first cycle to allow for early miRNA analysis, which may provide evidence toward their future use in early clinical decision-making.

miRNA selection and concentration
Ninety-three miRNAs were selected for measurement in the 28 samples (14 baseline and 14 after treatment). To select the miRNAs, real-time qPCR miRNA OpenArrays (Life Technologies, Carlsbad, CA), which measure 756 miRNAs, were used to analyze a pooled baseline sample (14 samples) versus a pooled day 29 sample (14 after treatment samples). RNA was extracted from each individual sample and pooled. The miRNAs that were easily quantifiable and with the largest difference between the baseline and day 29 were selected for further analysis of individual samples. The 93 miRNAs selected were analyzed by a custom real-time PCR Taqman miRNA low density Subgroup analysis of non-responsive subjects (progressed before 7 cycles of therapy) and responsive subjects (did not progress before 7 cycles of therapy), respectively. 4 Plasma miRNA concentrations are fold differences between responders and non-responders using the after chemotherapy treatment miRNA concentrations. Non-responsive subjects are those that had < 6 months of treatment prior to progression of ovarian cancer; Responsive subjects are those that had > 6 months prior to progression of ovarian cancer based on RECIST. *p<0.05

Patient characteristics
Fourteen of the original 17 patients in the parent trial had plasma samples available for miRNA analysis. Disease status was assessed using RECIST and CA-125 criteria every 2 cycles of chemotherapy. Chemotherapy was discontinued with recording of an intolerable toxicity or disease progression. Disease progression in patients prior to 6 cycles (n = 8) of chemotherapy were considered non-responders; and disease progression after 6 cycles (n = 6) of chemotherapy were considered responders. Of the responders, one patient remained without progressive disease for~17 months. Each patient completed the number of cycles of chemotherapy shown in Fig 2. The responders and non-responders are separated by the red line.

Circulating miRNAs change in response to chemotherapy
The primary endpoint of our study was to determine changes in circulating miRNAs following decitabine/carboplatin chemotherapy. Prior to chemotherapy (baseline) and prior to the start of cycle 2 (day 29), plasma concentration were compared for 78 miRNAs. All 78 miRNA were visualized using a Spearman ranked heatmap (Fig 3; same patient identification numbers as in Fig 2). In the heatmap, patients tended to discriminate, after chemotherapy, into groups. Three distinct patient groups were visualized: group1 (patients #1, #3, #4, #7, #10, and #11); group 2 (patients #8, #13, and #14); and group 3 (patients #2, #5, #6, and #12). Groupings were based on the comparison of within patient similarities of the relative changes in miRNA concentrations. For instance, group 3, patient #12 had increased, with treatment, total miRNA concentrations compared to other patients within the group. Patient #12's miRNAs concentrations changed in the same relative proportions as those recorded for patient #6. Thus, patient #12 and #6 were similarly grouped. As these visualized groupings suggest, this chemotherapy regimen may induce distinct changes in circulating miRNA concentrations in subpopulations. Next, the changes in the concentrations of individual circulating miRNAs were evaluated after chemotherapy treatment. Following decitabine/carboplatin therapy as shown in Table 1 (individual patient's miRNA changes shown in S1 Fig), two microRNAs decreased in concentration. miR-193a-5p and miR-375 had respective fold-changes of -1.7 and -2.0 (p<0.05). Additionally, the fold change increased 1.9 (P<0.05) for miR-339-3p.
To determine if individual miRNAs were associated with treatment response, we compared in responders with non-responders the post-treatment concentrations of circulating miRNAs (Fig 2). Three miRNAs, miR-616, and miR-532-3p, miR-148b-5p, were increased in responders (Table 1; S4 Fig). Of these miRNA, miR-148b-5p increase (3.2 fold) in responders was consistent with the subgroup analysis, noted above, revealing its decrease in concentration (2.86) in non-responders.
Lower miR-148b-5p concentrations predicted worse progression free survival Next, a Kaplan Meier PFS analysis was used to evaluate if any of the changes in miRNA concentrations predicted treatment response. To make more applicable to the clinical setting, patients who progressed during the 1st cycle (4 patients) were removed from the analysis. For those patients, progression would be observed using standard diagnostics, and a predictive biomarker would be unnecessary. As shown in Fig 4 of the Kaplan Meier analysis, lower (p = 0.015) miR-148b-5p concentration on day 29 associated with disease progression. Baseline miR-148b-5p concentration was low (Fig 4, inset); and in non-responders, miR-148-5p decreased in concentration largely after 1 month of treatment.

Discussion
Our study has several important findings. First, decitabine/carboplatin chemotherapy induced changes in circulating miRNA concentrations in patients with platinum-resistant recurrent ovarian cancer. Second, subgroup analysis revealed miRNAs changed distinctly in nonresponders vs. responders. Third, miR-148-5p predicts possible clinical outcome.  Consistently in all patient samples, this chemotherapy regimen induced changes in three circulating miRNAs miR-375, miR-339, and miR-193a. Of these miRNA, miR-193a has been studied previously in human EOC. miR-193a is a tumor suppressor, and overexpression of miR-193a in human EOC cells leads to increased apoptosis [19]. The circulating concentration of miR-193a, in our study, decreased (-1.7-fold) after chemotherapy treatment. The relationship between circulating and tumor miR-193a may warrant further study. Changes in circulating miRNAs common to all patients may be a marker of chemotherapy compliance and dose response.
The observed changes in circulating miRNA concentrations in patients with recurrent ovarian cancer enrolled in this study are similar to previously reported chemotherapy-induced changes in circulating miRNA in patients with melanoma. In that setting, circulating miRNAs predicted drug dose response to dacarbazine and interferon-alfa-2b [20]. miRNA fold changes compared favorably to our study's -2 to 4 fold changes. The fold changes observed in our study appear reasonable and consistent with those recorded by others.
Subgroup analysis revealed distinctions in miRNA concentrations between non-responders and responders. These distinctions in miRNAs may be a marker of specificity for a subgroup.
In the subgroup analysis of non-responders, chemotherapy decreased two additional miRNAs, miR25-3p, and miR148b-5p. Suggesting that these may be specific for a non-responder.
miR-148b-5p predicts possible clinical outcomes. The circulating miR-148b-5p concentration was 3.2 fold higher in our study after decitabine/carboplatin chemotherapy in patients with longer PFS: miR-148b-5p decreased in patients with shorter PFS. Circulating miR-148b-5p has not been reported previously; though, intracellular miR-148-5p has been studied. miR-148b-5p is overexpressed in ovarian cancer tissue; but its overexpression is not associated with histological subtypes [21]. Other studies have shown that changes in intracellular miR-148 (also called miR-148a, and miR-148a-5p) expression, located on chromosome 7, are associated with ovarian cancer outcomes [22][23][24][25][26][27]. To our knowledge, this is the first study to show that miR-148b, which is located on chromosome 12, is associated with clinical outcomes after treatment with any chemotherapy.
Our study has a couple of strengths. The first strength was a well-described group of patients. This study focused on decitabine/carboplatin chemotherapy induced changes in circulating miRNA focused in recurrent ovarian cancer. Many other studies have been confounded, for example, by either heterogeneous cancer patient population or additional treatments in conjunction with the chemotherapy. For instance, measuring circulating miR-NAs as biomarkers for neoadjuvant chemotherapy in breast cancer, one study included surgery and a very heterogeneous population, and the other also included surgery [28,29]. The results of the studies were not consistent with each other. In the study with chemotherapy followed by surgery, the fold changes in circulating miRNA ranged from 1.5 to 4. In the other study that included patients with the many different stages of breast cancer, the subjects had miRNA fold changes from 1 to 15,000 [28]. Additionally from these studies, circulating miRNA changes induced by chemotherapy were difficult to differentiate from changes caused by surgery. A second strength of our study was the timing of sample collection. Our study samples were collected early in the treatment regimen compared to the studies listed. Our study evaluated changes in circulating miRNA after 1 cycle of chemotherapy; these other studies analyzed changes in circulating miRNAs after 4 cycles of treatment. We propose that evaluating circulating miRNAs after only 1 cycle allows for early clinical decision-making. An important early decision is whether to continue the current chemotherapy regimen or change to another regimen. This decision is especially important for patients with recurrent ovarian cancer, as the progression of disease may be rapid.
A secondary issue is the appropriate utilization of the heatmap. Our heatmap visually identified groups of patients by differences in miRNA concentrations. Though descriptive, the heatmap highlights circulating miRNAs are numerous and potentially altered by chemotherapy. Circulating miRNAs in combination may be a biomarker of chemotherapy response worth exploring in future studies.
We recognized the limitations of our study, including the small sample size. However, despite the small number of patients analyzed, we were still able to detect differences in chemotherapy-induced changes in circulating miRNAs. We acknowledge that these specific miRNA changes may be limited to this regimen, which is not a standard therapy for ovarian cancer patients, and that the patients had previously received a variety of different treatment regimens before starting this trial. Finally, this study included a single time point, a common limitation of these types of studies, and the results suggests that a time course analysis may be warranted in future studies of chemotherapy induced changes in circulating miRNAs.
In conclusion, our study indicates that a carboplatin and decitabine chemotherapy regimen induces changes in circulating miRNAs concentrations that may be predictors of clinical response to this regimen. Our results further indicate that circulating miRNAs may not directly reflect the tumor miRNA expression. These studies indicate that future studies focused on identifying circulating miRNAs predictive of chemotherapy response are warranted.