Vimentin and Ki67 expression in circulating tumour cells derived from castrate-resistant prostate cancer

Circulating tumor cells (CTCs), captured as a ‘liquid biopsy’ from blood for enumeration and biological characterization of cancers, have the potential to replace biopsy and provide important clinical information on prognosis, therapeutic choice, and drug resistance, while also being of interest for drug development and biomarker discovery. They may represent an alternative source of tumor tissue which is easily accessible using a simple blood test, allowing longitudinal monitoring of tumor aggression and biology at different timepoints to guide therapeutic decisions in a patient’s treatment course [1‐3]. Prostate cancer was one of the first malignancies where the prognostic value of monitoring CTC numbers was demonstrated in patients with advanced disease, both before and during treatment for castrate-resistant prostate cancer (CRPC) using the FDA-approved CellSearch technology [4, 5]. Moreover, a recent prospective trial demonstrated that CTC count and LDH value could be a surrogate of overall survival in a population of mCRPC patients treated with abiraterone in the COU-301 trial [6]. The potential of CTC for molecular characterization has been demonstrated on a number of occasions, most recently in CRPC with the demonstration of their variable androgen receptor (AR) expression [7‐12].

Ki-67 is a nuclear protein that is associated with ribosomal RNA synthesis and may be necessary for cell cycle proliferation. Its tissue staining has consistently demonstrated prognostic value in prostate cancer, and has been tested in men managed with radiation and surgery, as well as in those conservatively managed without definitive therapy [13‐16]. One group of studies used pretreatment biopsies of patients undergoing radiation and androgen deprivation as part of the Radiation Therapy Oncology Group (RTOG) 92–02 trial, defining a cut-off of 11.3 % high Ki-67 staining as independently correlated with an increase risk in distant metastasis, cancer-specific death, and overall death [13, 14]. A Mayo clinic study described similar outcomes from patient tissue taken before or during definitive prostatectomy, using a staining threshold of 6 % high Ki67 (present in 11 % of patients) which correlated with increased risks of cancer progression and cancer-specific mortality [16]. The phase III GETUG 12 study, which assessed androgen-deprivation therapy in high risk localized prostate cancer, showed that high Ki67 predicted an unfavourable PFS using a median Ki67 cutoff of 1 % [17]. Finally, 243 conservatively-treated patients from the Transatlantic Prostate Group were shown to have significantly increased risk of cancer-specific mortality in the 5 % of patients who had >10 % Ki67 staining [15]. The detection of Ki67 in CTCs has been previously reported using a microfluidic CTC detection method, offering the opportunity to assess if longitudinal quantification of proliferative and non-proliferative subpopulations may be more clinically informative than examination of total CTC count and/or biopsy alone [18].

Vimentin is a filamental protein expressed in mesenchymal cells, often recorded as a marker of tumour cell invasion via its expression during activation of the epithelial-mesenchymal transition (EMT) [19, 20]. Compared to Ki67, its use in biological interrogation of CTC is more commonplace. Interest in EMT/vimentin in the CTC field continues due to the continuing biological challenge it represents: CTC undergo EMT in order to enter the bloodstream where they are captured, yet many current enrichment methods and definitions of CTC do not incorporate mesenchymal markers such as vimentin. It is hypothesized that, in cancers such as non-small cell lung cancer which have a low yield of CTC via EPCAM (epithelial)-antibody enrichment in CellSearch, many ‘EMT’ and mesenchymal CTC are missed as a consequence of sub-optimal enrichment methods [21‐24]. Despite this potential problem, the validity of EMT analysis using the CellSearch platform has previously been confirmed by a report documenting the co-expression of vimentin with cytokeratin in samples from 10/10 CRPC patients, with 108/126 CTCs in total co-expressing these markers [25].

Given the association of Ki67 and vimentin with poor prognosis in metastatic CRPC biopsy, we hypothesized that vimentin or Ki67 expression in CTC would offer a poorer prognostic picture compared to that observed in patients with no vimentin- or Ki67-positive CTCs. We analysed Ki67 and vimentin expression in CTC derived from a total of 144 advanced prostate cancer patients using CellSearch technology, optimized and validated to assay Ki67 and vimentin in separate patient cohorts. Our results corroborate those seen in tumour biopsies, with a diminished OS observed in patients with vimentin-expressing CTCs and those with Ki67-expressing CTCs.

Here we report CTC expression of vimentin and Ki67 in advanced/metastatic prostate cancer, offering a clinical insight into their mesenchymal and proliferative properties. We have established a method for testing CTC expression of Ki67 and vimentin in the normal pathway of prostate cancer care. In line with previous studies, we confirmed the prognostic value of detection of ≥5 standard CTCs at baseline through analysis of PSA and OS. No significant difference was evident when baseline PSA was split into vimentin-positive and Ki67-positive CTC cohorts, but there was a significantly reduced OS in the vimentin-positive and Ki67-positive CTC cohorts compared to those patients with no vimentin/Ki67-positive CTCs, respectively. In both Ki67 and vimentin cohorts, no additional prognostic significance of Ki67- or vimentin-positive CTCs was evident when cohorts were subdivided into the unfavourable prognostic category of ≥5 standard CTCs overall, or when patients with high proportions of Ki67/vimentin-positive CTCs were compared to patients with high proportions of Ki67/vimentin-negative CTCs. This suggests that the presence of vimentin-positive or Ki67-positive CTC is more clinically important than their relative numbers or proportions, and that their presence in the early stages of advanced disease could be of greater prognostic significance than in later stages. Finally, there was no change in patient percentage or CTC proportion with Ki67- or vimentin-expressing CTCs before and after treatment, suggesting the potential of these techniques to offer predictive benefits of treatment remains unestablished.

It is important to reflect on the deficiencies of this dataset and its analysis. The results here are hypothesis-generating and should be interpreted within the context of a retrospective observational design. Survival data reported was not adjusted for patient treatment, and mCRPC patient numbers tested were relatively small, particularly for the Ki67 cohort. The variability of planned treatments was the reason progression-free survival was not evaluated in our study: with small patient numbers, an imbalance of treatment line and therapy was inevitable between our patient cohorts and would have confounded such an analysis. Finally, an initial biological selection during CellSearch should also be noted: it will only test for Ki67 and vimentin after patient CTC have been initially isolated using EPCAM-coated ferromagnetic beads, an epithelial-antibody based selection. This process is consistent across all patient cohorts analysed, but a fully representative picture of mesenchymal and proliferative CTCs may be impaired as a consequence. As an alternative to antibody-based CTC selection via CellSearch, a number of other CTC detection techniques are in different stages of development, some with the aim of isolating CTC by their large size rather than their biological phenotype: a similar analysis of CTC vimentin expression in mCRPC using these techniques would be of future interest, and could offer a more accurate picture of its clinical relevance.

We noted a significant reduction of OS in the patient cohort harbouring vimentin-expressing CTC: HR was 1.8, with a median of 305 days OS in the vimentin-positive CTC population compared to 453 days for those patients with no vimentin-positive CTCs. Given some of the study limitations noted in the previous paragraph (ie the possibility that a high percentage of mesenchymal CTC are missed by the CellSearch technology), this finding merits further examination and has future potential to offer important insight regarding mesenchymal CTC and their treatment. It is possible that this difference could have been more profound using CTC enrichment platforms that are not antibody-based. More opportunity to perform detailed study of vimentin co-localisation with other CTC markers should hopefully arise soon, given the recent identification and clinical characterisation of a non-commercial cell surface vimentin antibody that detects CTCs from epithelial cancers with high sensitivity and specificity [28].

Mesenchymal or EMT CTC, as represented by vimentin expression in this study, have previously been associated with tumour invasiveness, chemoresistance, CTC clustering, stem cell characteristics, and poorer clinical outcomes [29, 30]. Although not validated by our study, the results generated here are generally consistent with this proposed biological and clinical phenotype. In parallel with the poorer prognosis seen in this study for those patients with vimentin-positive CTCs at baseline, the lack of survival difference observed in those patients whose samples contained the highest proportion of vimentin-positive CTCs is consistent with (but does not confirm) the stem cell theory that just one vimentin-positive cell will suffice to contribute to poorer outcomes. Armstrong and colleagues previously used the CellSearch platform to offer a detailed biological examination of vimentin and other markers of the EMT program in samples from patients with breast and prostate cancer, revealing the existence and high frequency of CTCs that can co-express epithelial and mesenchymal proteins [25]. Using a cutoff of 1 vimentin-positive CTC, 30/93 CRPC patients from our study had baseline samples with CTCs that co-expressed cytokeratin with vimentin, compared to 10/10 CRPC patients documented previously. This disparity could be accounted for by differences between patient characteristics and staining methods employed in the two studies, but here we have also been able to build on the clinical aspect of vimentin analysis by offering the larger clinical cohort with longer-term follow-up that was recommended in the previous study.

A significant reduction of OS was also observed in the patient cohort harbouring Ki67-expressing CTC: HR was 2.37, with a median of 512 days OS in the Ki67-positive CTC population compared to 751 days for those patients with no Ki67-positive CTCs. One previous study has explored the biological character of Ki67-positive CTC using an alternative CTC isolation platform, also offering some preliminary insight from CTC enumeration of 10 patient samples: 12.3 % of CTCs were Ki67-positive in our study, compared to 7–73 % from 4 CRPC patients reported previously [18]. We have thus been able to consolidate the results of this initial study using larger patient numbers and the CellSearch platform, suggesting again that there is potential prognostic clinical insight to be gained by examining proliferative and non-proliferative CTC subpopulations. Future studies can now focus prospectively on prognostic validation of Ki67-positive CTCs and their response to particular treatment types.

A number of other prostate cancer CTC studies have recently elaborated on the original work which showed the prognostic value of CTC in mCRPC [4, 5]. In terms of CTC enumeration, a putative predictive role for ‘categorical’ CTC enumeration (ie whether a patient is CTC positive or not) has been suggested after 1 cycle of chemotherapy, an evaluation that can be performed weeks ahead of standard response approaches such as PSA and CT [31]. A recent large prospective study suggested that CTC count and LDH value could be a surrogate of overall survival in mCRPC treated with abiraterone [6]. On the other end of the research spectrum, the biological heterogeneity of prostate cancer with treatment has also recently been re-illustrated despite its seemingly predictable clinical pattern of metastatic spread compared to many other cancers: CTC measurement and next generation single cell sequencing of metastatic prostate cancer patient revealed a vast complexity of clonal evolution, a finding consistent with the metastasis-to-metastasis spread of multiple tumour clones recently revealed by prostate tumour biopsies [32‐34]. Many of these studies have concentrated on androgen receptor expression and its predictive capacity [11, 12, 35]. Overall, the vast range of CTC enumeration, genomic and biological progress achieved in these studies offer an encouraging window of insight towards the future for circulating biomarkers in prostate cancer.

Here we show that routine clinical testing of vimentin and Ki67 expression in CTC derived from mCRPC patients is feasible. A significant reduction in OS for patients with baseline CTC vimentin and Ki67 expression suggests that largescale prospective analyses would be important to further explore these potential links. Future studies will also require further detailed correlation of vimentin/Ki67 CTC expression and response to certain treatment types: for example, the predictive capacity of lost or increased Ki67 CTC-expression in a large cohort of mCRPC patients treated exclusively with anti-proliferative chemotherapy agents such as docetaxel or cabazitaxel. Finally, a number of alternative CTC-isolation techniques have been developed and merit further analysis for such clinical correlation studies: size-based filtration techniques may be particularly helpful for vimentin-expressing CTC which may not be efficiently captured by techniques based on epithelial-antibody expression.