A Pilot Study of the Predictive Potential of Chemosensitivity and Gene Expression Assays Using Circulating Tumour Cells from Patients with Recurrent Ovarian Cancer

Circulating tumour cells (CTCs) from liquid biopsies are under current investigation in several cancers, including epithelial ovarian cancer (EOC) but face significant drawbacks in terms of non-standardised methodology, low viable cell numbers and accuracy of CTC identification. In this pilot study, we report that chemosensitivity assays using liquid biopsy-derived metastatic EOC CTCs, from 10 patients, nine with stage IIIC and one with stage IV disease, in progression after systemic chemotherapy, submitted for hypoxic isolated abdominal perfusion (HAP), are both feasible and useful in predicting response to therapy. Viable metastatic EOC CTCs (>5 cells/mL for all 10 blood samples), enriched by transient culture and identified by reverse transcription polymerase chain reaction (RT-PCR) and indirect immunofluorescence (IF), were subjected to flow cytometry-based Annexin V-PE assays for chemosensitivity to several chemotherapeutic agents and by RT-PCR for tumour gene expression profiling. Using a cut-off value of >80% cell death, CTC chemosensitivity tests were predictive of patient RECIST 1.1 responses to HAP therapy associated with 100% sensitivity, 50% specificity, 33% positive predictive, 100% negative predictive and 60% accuracy values. We propose that the methodology employed in this study is feasible and has the potential to predict response to therapy, setting the stage for a larger study.

IV EOC, followed by immediate chemo-filtration, has been reported to result in median survival times of 10 and 12 months, respectively, and an improved QoL [14].
In addition, precision oncotherapy based upon tumour chemosensitivity assays has been under evaluation as an alternative therapeutic approach for the treatment of platinum-resistant EOC. Empiric therapies are chosen from the current literature based upon outcomes achieved for a particular tumour-type with single and combinations of chemotherapeutic agents, whereas drug-selection based upon chemosensitivity assays takes cues from the sensitivity of tumour tissues, tumour cell cultures [15][16][17][18][19][20][21][22][23][24][25][26][27] or purified circulating tumour cells (CTCs) to a panel of chemotherapeutic agents in in vitro cytotoxicity assays. In addition to these approaches, important information predicting a potential drug-response can also be gleaned from standard immunohistochemical, gene expression and transcription profiling of non-viable tumour tissues.
With respect to CTCs, flow cytometry studies have detected CTCs with a high degree of sensitivity and specificity in blood samples from melanoma, breast, prostate, pancreatic and colon carcinoma patients [28]. As CTCs represent potential metastatic precursors, CTC purification provides a unique opportunity to gain a more accurate assessment of the genomic, transcriptomic and chemosensitivity characteristics of individual tumours and has prompted the development of several purification techniques based on tumour cell density, size, deformability and biological properties that facilitate positive or negative label-dependent immunoaffinity purification, which now include an antibody-based herringbone-chip CTC purification technique [29,30], with EOC CTCs identified by immunocytochemistry (ICC), reverse transcription polymerase chain reaction (RT-PCR) and fluorescent in situ hybridisation (FISH) for novel gene fusions [31,32]. In spite of no standardised methodology for CTC purification, characterisation or enrichment, CTC quantification and gene/protein expression profiling are already used to predict disease progression and select treatment strategies [33]. However, PCR analysis precludes additional analysis of living CTC and antibodies for CTC purification against epithelial cell-specific proteins such as epithelial cell adhesion molecule (EpCAM) would not purify metastatic CTC populations that have undergone epithelial to mesenchymal transition, which no longer express typical epithelial cell markers [34].
Considering that not all treatments are available in every institution and no single treatment strategy fits all, for reasons of lesion size and number, anatomical location, regional lymph node involvement, distant metastases, biomolecular aspects, concomitant disease and previous therapy, it is our considered opinion that treatment strategies for advanced EOC should be multidisciplinary and could benefit greatly from detailed biomolecular characterisation and chemosensitivity assessment of liquid biopsy derived CTCs from individual patients, already USA Food and Drug Administration (FDA)-approved for prognostics [37]. Furthermore, despite the current lack of methodological consensus, CTC-based analytical methodologies are already under investigation for selecting therapeutic strategies in different cancers, including EOC [30,35].
Recently, we reported a method for liquid biopsy predictive oncotherapy based upon purified metastatic CTCs, transiently cultured in vitro, permitting both gene expression profile and chemosensitivity analysis [38][39][40]. Here, we report a pilot study of metastatic CTCs purified from a homogeneous group of stage IIIC and IV EOC patients, presenting with disease relapse in progression after surgery and two lines of systemic chemotherapy, submitted for locoregional HAP chemotherapy, with the aim of confirming the feasibility and utility of assessing chemosensitivity and biological characterisation of liquid biopsies-derived purified CTCs, as a predictive test for selecting appropriate drugs for HAP and for further therapeutic strategies. Table 1 reports biological and clinical characteristics, RECIST 1.1 tumour responses, and survival of 10 advanced stage EOC patients submitted for multidisciplinary treatments.

Chemosensitivity and Tumour Gene Expression Assays Using EOC CTCs
Viable metastatic EOC CTCs (>5 cells/mL; median 9.4 cells/mL and interquartile range 8.2-9.6) ( Figure 1), were isolated from liquid biopsies from 10 patients with recurrent EOC and CTC chemosensitivities presented in Table 2. With respect to the two drugs used for HAP, a chemosensitivity cell-death cut-off value of >80% was chosen. This value was achieved in EOC CTCs from six patients who subsequently received both chemotherapeutic agents. Three patients with CTC chemosensitivity values of >80% for one but not the other drug, received the agent achieving cut-off and received a second agent that did not achieve the cut-off, based on a multidisciplinary decision. The remaining patient, for whom chemosensitivity assay cut-off values failed to reach >80%, received the drug pair inducing the highest levels of cell-death. Based upon responses to previous systemic platinum-based chemotherapy, one patient was platinum-resistant, six were partially platinum-sensitive and three were fully platinum-sensitive. Chemosensitivity assays confirmed ≥80% cell death of CTCs from six patients for at least one of three platinum-compounds (carboplatin, cisplatin and oxaliplatin), whereas drug-induced CTC cytotoxicity in the remaining four patients failed to reach the 80% cut-off chemosensitivity value.

CTC Chemosensitivity Test Accuracy in Relation to "two-drug" HAP
Positive (complete or partial) and negative (stable disease or progression) RECIST 1.1 responses, following "two-drug" HAP, associated with CTC chemosensitivity of >80% or < 80% drug-induced cell death, are displayed in Table 4. A 100% sensitivity value was observed for complete or partial RECIST 1.1 responses, following 2 drug HAP selected by > 80% CTC chemosensitivity, and a specificity value of 50% was observed for RECIST 1.1 disease responses characterised as stable or in progression, following 2 drug HAP selected by <80% CTC chemosensitivity. A 33% positive predictive value (PPV) for a positive RECIST 1.1 response was associated with >80% CTC chemosensitivity to both drugs and a 100% negative predictive value (NPV) for a negative RECIST 1.1 response was associated with CTC chemosensitivities of <80%. The overall of value of CTC chemosensitivity tests in predicting the patient response to 2 drug HAP (accuracy value) was 60% and was calculated from the ratio of positively and negatively-corrected classified patients, using the chemosensitivity cut-off value of 80%, RECIST 1.1 criteria and the number of patients treated with 2 drug HAP. Table 4. Positive (complete or partial) and negative (stable disease or progression) RECIST 1.1 responses to 2-drug HAP, selected by liquid biopsy CTC chemosensitivity assay and associated with either > 80% (positive) or ≤ 80% (negative) CTC death, for both drugs.

Treatments Following HAP, and Patient Follow-up
Additional multidisciplinary treatments were also based upon the results from CTC chemosensitivity assays, CTC gene expression profiles and BRCA mutational status. Based on gene expression assays, two patients received bevacizumab targeted-therapy (Patients 3 and 10). Patient 3 exhibited a CR of 48-month duration, subsequently received rucaparib for 36 months, in accordance with a mutated BRCA status. This patient is still alive and continues to exhibit a CR. In contrast, patient 10 exhibited a PR of 15-month duration and unfortunately died at 24 months. Three patients with locoregional and distant relapsed disease received systemic therapy, one patient received both locoregional and systemic therapy and four patients received best supportive care, only ( Table 1). The three patients with locoregional relapsed disease received additional locoregional treatments based on CTC chemosensitivity assays. Of the 10 patients enrolled in this cohort study, nine died as the result of disease progression, one remains alive today with no evidence of disease and the median OS for the cohort was 15 months (interquartile ranges 12-24 months).

Discussion
In this study, we report that the methods employed for blood sampling, storage, transport and subsequent CTC purification and enrichment are feasible and reproducible, and that the subsequent CTC chemosensitivity and gene expression assays employed to select chemotherapeutic strategies are predictive of therapeutic response. Our results confirm that the numbers of EOC CTCs purified from liquid biopsies, expanded by temporary culture, are sufficient for flow cytometry-based Annexin V-PE chemosensitivity assays, which predict RECIST 1.1 responses to "two-drugs" locoregional HAP with 60% accuracy.
The interval between blood sampling and quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis did not exceed 80 h, previously reported to minimise alterations in gene and protein expression [41] and metastatic EOC CTC numbers in 10 liquid biopsies were greater than the 5 CTCs/mL cut-off value required for chemosensitivity and tumour gene expression analysis [41]. In our study, CTC detection rate was 100%, which is significantly higher than previous reports of CTCs purified from liquid biopsies obtained from patients with primary tumour [30] and can potentially be explained by the increased numbers of CTC that would be expected to associate with the advanced stage IIIc and IV EOC metastatic tumour burden.
Considering the relatively low numbers of CTCs isolated from individual patients, isolates were expanded in appropriate media, at 37 • C and 5% CO 2 , to densities sufficient for chemosensitivity assays. This raises the possibility for phenotypic alteration and enrichment of subpopulations less representative of the bulk tumour. Since, less-differentiated cancer stem cell-like subpopulations, present in the vast majority of CTC isolates [42], expand more rapidly than more differentiated CTC counterparts [43], expanded CTC cultures are more likely to represent more aggressive components of the bulk tumour and, therefore, a more relevant cell population for predictive oncotherapy. Expanded cultures, however, did not exhibit phenotypic or genotypic alterations, assessed using short tandem repeats (STRs) or by analysing biomarker mRNA and protein expression prior to and post expansion.
Flow cytometry Annexin V-PE CTC chemosensitivity assays provided important information on several chemotherapeutic agents but in contrast to tissue-validated chemosensitivity assays, do not preserve cell-to-cell or cell-to-matrix interactions [3]. For this reason, a cut-off value of >80% cell death was chosen for drug selection based upon CTC chemosensitivity, which is far higher than the > 30% cell death cut-off normally used in predictive chemosensitivity assays employing tumour tissues that better resemble in vivo tumour architecture [3].
According to the confusion matrix displayed in Table 4, the potential of in vitro CTC chemosensitivity assays with an 80% cut-off value to predict therapeutic efficacy, evaluated by RECIST response to 2 drug HAP, resulted in values of 100% for sensitivity, 50% for specificity, 33% for PPV, 100% for PNV and a more moderate value of 60% for accuracy, all of which represent better overall values than those reported for in vitro tissue-validated chemosensitivity assays, employing a cut-off value of 30% to predict response to therapy in ovarian cancers, reported as 85.7% for sensitivity, 18.2% for specificity, 40% for PPV, 66.7% for PNV and 44.44% for accuracy [3].
The scope of this EOC small cohort pilot study was to provide a feasibility evaluation of our predictive oncotherapy CTC-based approach and was not designed to evaluate treatment safety, efficacy or effectiveness [48]. This not only justifies the small sample size but also the absence of inferential statistical analysis. Clinical covariates were not considered in the analyses. Furthermore, the stringent recruitment parameters employed, including sample homogeneity, also mitigate the small sample size and are a pre-requisite for feasibility [48]. However, we caution that the percentage values for sensitivity, specificity, PPV, PNV and accuracy are indicative, at best, and should not be extrapolated to inclusion and exclusion criteria not used in this study. The efficacy data presented here are also uncontrolled and, therefore, observational and future studies, with adequate sample sizes and controlled, should aim to validate a standardised methodology for CTCs isolation, purification, enrichment, characterisation for use in chemosensitivity and tumour gene expression assays in order to elaborate a multidisciplinary treatment strategy.

Patients
This project was performed in accordance with the Declaration of Helsinki and approved by the ethics committee of ASL n.1, Abruzzo, Italy (10/CE/2018, 19 July, n.1419). Ten patients were prospectively enrolled from 2011 to 2016 and written informed consent was obtained from each patient. Clinical characteristics of the 10 patients with advanced stage EOC are presented in Table 1. Median age was 59 years, with an interquartile range of 56-68. All 10 patients presented with histologically diagnosed high-grade serous carcinoma and one patient exhibited a BRCA1 nonsense mutation. One patient was classified as platinum-resistant after 1st line systemic chemotherapy and was in progression after the 2nd line chemotherapy. Six patients were classified as partially platinum-sensitive and all were in progression after at least two lines of systemic chemotherapy. Three patients were classified as fully platinum-sensitive, two of whom were in progression after several lines and one after one line of systemic chemotherapy.

HAP
HAP procedures were performed under general anaesthesia, as previously described [14,52,53]. Briefly, after systemic heparinisation (150 IU heparin/kg), the common femoral artery and saphenous vein (or femoral vein, or external iliac artery and vein, when necessary) were isolated and cross-clamped. Two, previously heparinised, triple-lumen 12-F gauge balloon catheters (PfM, Cologne, Germany and Dispomedica, Hamburg, Germany), were X-ray guided into both the artery and vein and aortic and inferior cava balloons positioned, by fluoroscopy, just above the celiac trunk, above the confluence of the right hepatic vein and below the right atrium, respectively. After blocking and checking their position with contrast medium, balloons were unblocked and thighs initially blocked with pneumatic cuffs. Under temporary hyperoxygenation, chemotherapeutic agents not influenced by hypoxia and pH < 7.1 [54], such as cisplatin, carboplatin, vinorelbine, etoposide, paclitaxel and docetaxel, were introduced through the aortal catheter, as a bolus, and balloon catheters immediately blocked again. It is mandatory to inject the drugs under prior hyperoxygenation directly before balloon-blocking in order to avoid ischaemic bowel complications and for drug action [53,54]. After balloon catheter inflation, isolated perfusion of the abdominal cavity was performed by way of extracorporeal peristaltic pumps for approximately 15-20 min, under hypoxic conditions to enhance doxorubicin (and other drugs such as mitomycin) cytotoxicity [54], and at a pH of < 7.1 to enhance topotecan cytotoxicity [55]. The extracorporeal circuit (Figure 2) also contained a heating element and a hemofiltration module. Temperature loss was approximated to be 1 • C per meter of tubing and the length of the tubing was 5 m. Therefore, to ensure normothermia the element outlet port was pre-heated to a temperature of approximately 42 • C, ensuring both patient well-being and drug action. At termination of perfusion, the venous balloon was deflated prior to the arterial balloon and lower extremity tourniquets sequentially released upon attainment of a stable hemodynamic profile and the extracorporeal circuit was then subjected to chemo-filtration to reduce systemic drug access in order to reduce side effects. After chemo-filtration, catheters were withdrawn and vessels repaired by 5/0 Prolene suture. Protamine (200 IU/kg) was then injected to reverse the effects of heparin. the confluence of the right hepatic vein and below the right atrium, respectively. After blocking and checking their position with contrast medium, balloons were unblocked and thighs initially blocked with pneumatic cuffs. Under temporary hyperoxygenation, chemotherapeutic agents not influenced by hypoxia and pH < 7.1 [54], such as cisplatin, carboplatin, vinorelbine, etoposide, paclitaxel and docetaxel, were introduced through the aortal catheter, as a bolus, and balloon catheters immediately blocked again. It is mandatory to inject the drugs under prior hyperoxygenation directly before balloon-blocking in order to avoid ischaemic bowel complications and for drug action [53,54]. After balloon catheter inflation, isolated perfusion of the abdominal cavity was performed by way of extracorporeal peristaltic pumps for approximately 15-20 min, under hypoxic conditions to enhance doxorubicin (and other drugs such as mitomycin) cytotoxicity [54], and at a pH of < 7.1 to enhance topotecan cytotoxicity [55]. The extracorporeal circuit (Figure 2) also contained a heating element and a hemofiltration module. Temperature loss was approximated to be 1°C per meter of tubing and the length of the tubing was 5 m. Therefore, to ensure normothermia the element outlet port was preheated to a temperature of approximately 42 °C, ensuring both patient well-being and drug action. At termination of perfusion, the venous balloon was deflated prior to the arterial balloon and lower extremity tourniquets sequentially released upon attainment of a stable hemodynamic profile and the extracorporeal circuit was then subjected to chemo-filtration to reduce systemic drug access in order to reduce side effects. After chemo-filtration, catheters were withdrawn and vessels repaired by 5/0 Prolene suture. Protamine (200 IU/kg) was then injected to reverse the effects of heparin.

Statistical Analysis
According to Connelly [57], a pilot study should be 10% of the sample projected for the larger parent study. Therefore, considering the small sample size in this pilot study, statistical analysis is descriptive, with purified CTC numbers, progression-free survival (PFS) and overall survival (OS) times presented as medians with interquartile ranges. The relationship between CTC chemosensitivity and response to HAP are presented without confidence intervals, as sensitivity,

Statistical Analysis
According to Connelly [57], a pilot study should be 10% of the sample projected for the larger parent study. Therefore, considering the small sample size in this pilot study, statistical analysis is descriptive, with purified CTC numbers, progression-free survival (PFS) and overall survival (OS) times presented as medians with interquartile ranges. The relationship between CTC chemosensitivity and response to HAP are presented without confidence intervals, as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy percentages, and all computations were performed using STATA statistical software.

Conclusions
We conclude that the methodology for liquid biopsy, samples storage and transport, CTC purification and transient in vitro culture, and subsequent chemosensitivity and gene expression assays are both feasible and reproducible. However, issues that remain to be addressed, include: (i) the use of miniaturised single cell technology for low CTC numbers [58], (ii) testing with increased numbers of drugs, as on average only 10 drugs have been tested in existing reports; (iii) multiple testing prior to and following selected therapies, to take neoplastic evolution into account; (iv) analysis of primary and metastatic tumour CTCs compared to the same patient's untreated cancer and non-cancer cells, as more appropriate controls; (v) use of more relevant 3D cell cultures; (vi) better standardisation of methodology and assays with a greater number of clinical studies to fully confirm predictive potential and importance in the selection of cancer therapy.

Conflicts of Interest:
The authors declare no conflict of interest.