Measurement of circulating tumor cells determines treatment response and exitus in patients with hepatocellular carcinoma submitted to transarterial chemoembolization

Circulating tumor cells (CTCs) enumeration is a promising technique to predict cancer prognosis and treatment response. CTCs were evaluated in healthy subjects, cirrhotic controls and hepatocarcinoma (HCC) patients. CTCs were isolated using microfluidic system based on the expression of EpCAM, EGFR and three epithelial to mesenchymal transition (EMT) markers. Patients were stratified according to disease progression and exitus. Although counts of individual CTCs, clustered CTCs and α-fetoprotein (AFP) at basal level in patients with HCC were significantly increased compared with the values obtained in cirrhotic patients and control subjects, only individual CTCs (p=0.027), but not clustered CTCs (p=0.063) and AFP (p=0.072), were independent predictors of HCC development. The univariate regression model showed that basal levels of CTCs46 were related to high risk of HCC (Odds Ratio 3.467, p=0.011). The stratification of our cohort according to disease progression and death showed that basal individual CTCs 76 (Hazard Ratio 5.131, p=0.004) were related to disease progression, as well as the difference of clustered CTCs between 1-month and baseline levels 1.5 were related to death (Hazard Ratio 10.204, p=0.036). In conclusion, the preoperative and 1-month measurements of CTCs in blood constitute useful markers to predict the outcome of patients under TACE treatment.

for therapeutic recommendations according to the criteria of the Barcelona Clinic Liver Cancer (BCLC) [1]. The curative treatments (radiofrequency, liver resection, and orthotopic liver transplantation or OLT) are indicated during the early stages of the disease (BCLC 0-A), which show a high 5-years survival (50-80%), representing an average survival of more than 60 months [3].
However, nowadays, two third of patients diagnosed with HCC are in more advanced stages of the disease (BCLC B or C) [4]. Transarterial chemoembolization (TACE), which consists tumor arteria embolization with drug-eluting beads loaded with doxorubicin, is the recommended treatment for patients at the intermediate stages (BCLC B) [3]. The average survival of these patients is 26 months [3]. The intermediate stage appears to be a critical phase with high risk of evolving to more advanced stages of the disease (BCLC C) with poor prognostic factors such as vascular invasion, extrahepatic metastases and/or impaired hepatic function. Sorafenib is the standard of care for advanced HCC stage, as demonstrated in two large-scale trials such as the SHARP [5] and the Asia-Pacific [6] clinical trials. Nowadays, the average survival of patients in advanced stages treated with first and second line therapies ranges from 7 to 26 months [7].
Early screening of patients with HCC has been reported to improve the impact of therapeutic strategy and was an independent predictor of survival [8]. Therefore, to provide the best outcome of treatments, precise and effective biomarkers are urgently needed. Circulating tumor cells (CTCs) refer to cancer cells that are disengaged from the primary lesion, delivered to blood circulation, bone marrow, or lymphatic vessels and represent tumor progression and metastases [9]. The results of broad meta-analysis showed that the presence of CTC in blood was associated with poor recurrence free survival (RFS) and overall survival (OS), significantly increased risk of disease recurrence and death in patients with HCC [10]. Another metanalysis showed that CTCs positivity was significantly associated with relapse-free survival, overall survival and some clinical characteristics in patients with HCC [11].
The aim of the study was to correlate the levels of individual CTCs, clusters and clustered CTCs before and 1-month after TACE treatment in peripheral blood with clinical progression or effectiveness of treatment in patients with HCC at intermediate stage (BCLC B). Our data showed that the number of CTCs, clusters and clustered CTCs in blood at preoperative and 1-month posttreatment were predictive markers of disease progression and exitus in patients with HCC. Table 1 shows demographic and epidemiological basal data of our cohort (n=82) from Hospital University "Virgen del Rocío" (Seville, Spain). The percentage of exitus was higher in HCC patients (35.3%, p<0.001). Liver function of patients with HCC measured by Child-Pugh staging system was better than that observed in cirrhotic patients (p<0.001). The basal values of glucose, α-fetoprotein (AFP) and individual CTCs were significantly higher in patients with HCC than cirrhosis (p=0.006, p<0.001 and p=0.018, respectively). (Table 1). An additional analysis was carried out to identify the increased power of CTCs compared with AFP measured at baseline in predicting the development of HCC. Univariate regression models showed that individual CTCs but not AFP were associated with the presence of HCC (p=0.027 and p=0.072, respectively) ( Table 2).  (Figure 1e). Next, analysis using the cut-off value of 46 individual CTCs further confirmed that this parameter was able to increase the probability of HCC (Odds Ratio=3.467, p=0.011) ( Table 3).

High baseline levels of individual CTCs were predictors of tumor progression after TACE
Next, we studied the relationship between baseline levels of CTCs and tumor progression during follow-up in patients with intermediate HCC (n=51). Therefore, we divided the overall cohort into progressors (n=21) and non-progressors (n=30) patients (Table 4). We found no statistical differences in several demographic or clinical parameters (sex, etiology, Child-Pugh cirrhosis stage, number, size or location of tumors, glucose, alanine transaminase or AFP) between both groups ( Univariate and multivariate models were constructed based on CTCs quantitative data ( Table   6). Cox univariate regression models considering the time to progression (TTP) showed that baseline individual CTCs, clusters and clustered CTCs increased the probability of progression after TACE (p=0.008, p=0.015 and p=0.016, respectively) ( Table 6). Binary regression models showed that baseline individual CTCs and clusters were related to disease progression (p=0.028 and p=0.022, respectively) (

3.The increase of CTCs at 1-month after TACE versus baseline constituted a risk factor for patient death
We analysed the changes of CTCs levels 1 month after TACE versus baseline in a subpopulation of HCC patients (n=28). Table 8 shows demographic and clinical parameters of the cohort of patients who were died (n=6) and alive (n=22) at the time of evaluation. We found no statistical differences in several demographic or clinical parameters (sex, etiology, Child-Pugh, number, size or location of tumors, development of metastases, glucose, ALT and AFP) between both groups ( Table 8). The changes in CTCs 1 month after TACE versus baseline were strongly associated with death (  (Table 9).     were important predictors of death at any time point in our cohort (Table 11). Multivariate analysis was not carried out due to the high dependence of variables.

Discussion
HCC represents 80% of the primary hepatic neoplasms that appears mainly in the context of chronic liver cirrhosis (1). TACE has shown a beneficial impact on patients with unresectable HCC by increasing long-term survival and TTP (12,13). Although it remains advantageous, there is still a low to moderate degree of evidence of its effectiveness and many combinations with other therapies have recently emerged (14). Therefore, biomarkers of treatment response are needed. Several studies have identified radiological patterns that correlate with response after treatment. Preoperative magnetic resonance image (MRI) features based on irregular tumor margins and abnormal AFP levels in tumors >5cm could predict non-complete response after TACE coupled to high intensity focused ultrasound in an Asiatic cohort (15). Similarly, another study set up on MRI, has indicated that large tumors and intense arterial enhancement increase probabilities of incomplete response (16).
Nevertheless, these studies have not identified predictors of complete disease remission. In addition, they are observative approaches that do not consider molecular characteristics of tumors.
EMT has been established as one of the main mechanisms in HCC progression. During these biological process, epithelial cells, which are usually attached to a basement membrane, lose their adhesion molecules and acquire mesenchymal characteristics that allow them to become migratory (17). Expression of cytoskeletal markers such as vimentin, β-catenin, α-SMA, extracellular matrix proteins like fibronectin or transcription factor Twist are typical hallmarks of EMT (18). Cell transdifferentiation is thought to play an important role in metastasis and tumor progression, contributing to generation of CTCs. In this multistep process, epithelial cancer cells detach from primary tumors and get access to general circulation, acquire anoikis resistance, and extravasate to generate secondary tumors at distance (19). Thus, CTCs with EMT markers might be considered relevant markers in liquid biopsies, which allow a minimally invasive access and that might be useful for monitoring the evolution of malignancies. According to their molecular phenotype, CTCs can be divided into epithelial, epithelial to mesenchymal or mesenchymal cells, having the second classification a stronger association with tumor dissemination (20).
Liquid biopsy has also been applied to study the prognosis and priority of patients with HCC in the waiting list for OLT. The procedure used in the present study, IsoFlux, has resulted more sensitive in detecting CTCs than CellSearch in patients in the waiting list for OLT (21). In addition, Ramírez et al. showed that CTCs measured by IsoFlux procedure were detected in 21 out of 24 patients, and ranged from 2 to 1768 counts per 10 mL of blood, and being associated with the time in the waiting list (22). A postoperative CTC count ≥3 appears to be a surrogate marker for the prediction of postoperative extrahepatic metastases after curative surgical resection of HCC (23). The CanPatrol CTC enrichment system based on cellular filtration using membranes with 8 µm diameter pores and tricolor RNA hybridization in situ (RNA-ISH) was not able to find any correlation between CTCs counts and tumor recurrence after OLT (24). In this line, other research using a CanPatrol-based approach has found no correlation between CTCs and recurrence in HCC (25). CTCs enriched by membrane filtration detected counts ranging between 0 and 31 that appeared to be predictors of clinical outcome in patients submitted to surgical intervention for HBV-related HCC (26). All these studies suggest that a great variability exists on data mainly due to technical issues and markers used to discriminate CTCs. We have demonstrated herein that our system has a great sensitivity in terms of detection, as CTCs were present in clinically relevant numbers in all blood samples. Regarding CTCs counts, IsoFlux provided higher CTCs counts than CellSearch, with established cutoffs of no more than 5 CTCs for bad prognosis (27)(28)(29). In particular, the use of EMT markers for the  (Table 7). To our knowledge, there are few publications proving that CTCs shedding correlates and predicts TACE response. Fang et al. (33) have shown that CTCs isolated from peripheral or central blood by magnetic bead-labeled anti-human EpCAM monoclonal antibody using a magnetic cell separator were not associate to TTP after TACE. In contrast, serum levels of dickkopf-1, a negative regulator of the Wnt-beta signaling pathway, and the presence of CTCs detected by immunomagnetic beads combined with fluorescence in situ hybridization (FISH), have been connected to efficacy (34).
These results support our findings although they are not fully comparable due to differences in CTCs enrichment procedures.
The effectiveness of treatment was addressed comparing changes in CTCs at 1-month after  (Table 11).
Therefore, our results showed that increasing baseline levels of individual CTCs were able to predict disease progression. Furthermore, changes in clusters and clustered CTCs 1-month after TACE were foretold patient death. Our study suggests that those patients who fulfilled criteria of baseline CTCs>76 and clustered>14.5, and 1-month changes in clusters>0.5 or clustered CTCs>1.5 might be redirected to systemic therapy due to their intrinsic probabilities of bad prognosis. Thus, ideally, our variables aspire to be decisive checkpoints in algorithms for helping in clinical decision making and treatment allocation. scans by qualified radiologists following mRECIST criteria. Complete response was defined when target lesions did not show contrast enhancement nor washout, being tissue totally necrotized. Partial response was considered when there was still contrast enhancement but a reduction in size of the viable target lesions were evident, being tissue partially necrotized.  Patients were distributed according to the treatment response or disease progression, and exitus.

Enrichment of Circulating
The univariate and multivariate binary logistic regression analysis identified different variables that were significantly associated with the progression of the disease or exitus in patients with intermediate HCC. We identified the cut-off values of the selected variables that give the best sensibility and specificity in the ROC curve for the prediction of disease progression or exitus. The presence of basal clusters >4.5 and basal CTCs in cluster>14.5 were modelled for the significant independent prediction of disease progression in the multivariate logistic regression analysis. The presence of 1-month after treatment clusters >0.5 and 1-month after treatment CTCs in cluster>1.5 were modelled for the significant independent prediction of exitus in the multivariate logistic regression analysis. A significance level of 5% was used throughout the variables studied. Statistical analyses were performed by SPSS statistical software v. 11.0 (SPSS Inc., Chicago, Illinois, USA).

Conclusions
The study correlated the baseline levels of CTCs with the presence of HCC and the disease progression after TACE. In addition, comparison of changes in the number of clusters and clustered CTCs at 1-month after TACE versus baseline identified patients with increased risk of death after TACE. The information might be relevant for clinical decision making in order to select patients that might not benefit from TACE, and consequently receiving systemic therapy at first might be alternative recommended.