Circulating tumour cell enumeration does not correlate with Miller–Payne grade in a cohort of breast cancer patients undergoing neoadjuvant chemotherapy

Breast cancer is a major public health issue globally, representing the most common cancer in women and one in ten of all newly diagnosed cancers. It is also the main cause of female cancer death globally, with 2,088,849 cases diagnosed in 2018 and 626,679 deaths [1]. Neoadjuvant chemotherapy (NAC) is now a standard treatment for breast cancer, often shrinking the tumour and allowing a less aggressive surgical approach for the patient [2]. 7–27% of new breast cancers are treated with NAC [3]. Between 10 and 40% of patients, depending on tumour subtype, receiving NAC can have a pathologic complete response (pCR) to chemotherapy [4] as determined by the Miller–Payne grading system [5].

While mortality has decreased dramatically due to earlier diagnosis and advances in treatment, metastatic disease represents the main cause of breast cancer-related morbidity and death [6], and approximately 20% of breast cancers will experience metastatic relapse. Disease spread via circulating tumour cells in the bloodstream may explain how metastasis occurs [7, 8]. At present, enumeration of CTCs is limited in the clinical setting to predicting clinical outcome [9, 10]. However, future potential applications include their use for both determining and monitoring efficacy of personalised treatment, and predicting and detecting metastases [11, 12]. In order that the full prognostic and predictive power of CTCs is realised, there are a number of issues to examine in terms of limitations in how CTCs are defined, detected and isolated [9], as EpCAM-based detection excludes what is widely considered the most clinically relevant subsets of CTCs. The gold standard CellSearch™ for CTC identification approved by U.S. Food and Drug Administration (FDA) uses fluorescently labelled cytokeratin monoclonal antibodies, the nuclear stain DAPI and the absence of staining by CD45 (pan-leukocyte stain), resulting in the overall selection of EpCAM⁺, CK8⁺, CK18⁺, CK19⁺ and CD45⁻ cells [9]. A major disadvantage of a number of CTC detection techniques is that they are dependent on capture based on epithelial marker expression, i.e., EpCAM and cytokeratins [13]. EpCAM is not a universal CTC biomarker [14], and therefore, detection is limited if expression has been downregulated due to epithelial mesenchymal transition (EMT) [15], while circulating tumour cells of mesenchymal origin are also not captured [16]. It is well recognised that CTCs are heterogeneous, with certain subgroups of CTCs harbouring higher metastatic potential. CTC clusters/microemboli have been associated with a worse clinical outcome in breast and lung cancer [17‐19]. Current technologies underestimate CTC clusters because few specialised devices exist for the detection of CTC clusters and microemboli [20, 21].

The relationship between the presence of CTCs in the circulation and the response to NAC is currently an area of interest. If CTCs had the ability to predict those patients that have a complete pathological response, this could have a major impact on breast cancer treatment. A recent meta-analysis looking at the utility of CTCs assessed using the CellSearch™ system in non-metastatic breast cancer patients receiving neoadjuvant chemotherapy did not find any correlation with CTC count and response to chemotherapy [22]. A second meta-analysis concluded that CTC count as measured by multiple devices had utility in predicting therapy response in breast cancer [23]. The isolation technologies and characterisation of CTCs are clearly recognised as a limitation in these studies.

This current study focussed on evaluation of all physical forms of CTCs, using a non-marker based approach, ScreenCell (Paris, France), pre- and post-neoadjuvant chemotherapy for breast cancer in order to ascertain the utility of CTCs to predict response to neoadjuvant chemotherapy. The ScreenCell size-selective method takes advantage of the larger size of CTCs compared with nucleated blood cells for isolation, with its circular pores of 7.5 ± 0.36 μm randomly distributed throughout the filter with a pore density of 1 × 10⁵ pores/cm² [24]. It avoids the bias introduced by antibodies, and false negatives/positives associated with these methods. Response to chemotherapy was assessed using the Miller–Payne grading system [5] and radiological assessment.

This study did not show any benefit for CTC counts prior to treatment or prior to surgery in assessing pathological response to neoadjuvant chemotherapy in breast cancer patients. It is likely that more in depth analysis of CTCs and larger studies will unlock their true potential in the clinic but currently many limitations exist in terms of their isolation and characterisation.

CTCs were detected in 85% of patients in this study prior to treatment, which is higher than some of the quoted studies in the literature which vary from 31 to 61% [22, 23]. However, many of the published studies used the CellSearch™ system, which we know underestimates CTC numbers due to its reliance on the presence of EpCAM. In addition, some meta-analyses published [22, 23] have included early breast cancer patients whereas in this study we focussed on those with locally advanced disease (including some triple positive/negative with N0 disease n = 6) who were undergoing NAC followed by surgery, a cohort we would expect to be higher CTC traffickers. The GeparQuattro trial did focus on a neoadjuvant cohort using the CellSearch™ system and found prognostic ability in pre-treatment CTC counts for HER2 positive and triple negative patients using a cut-off of 2 CTCs but similar to our findings this prognostic ability was independent of the primary tumour response [25]. The patient population selected and the ScreenCell isolation device used may explain the higher percentage of patients having CTCs at the outset.

Significant heterogeneity was observed in the CTC phenotypes isolated, single cells, doublets and clusters. At baseline, clusters were isolated in 50% of our cohort. This is slightly higher than recently reported by Vetter et al., who found 35% of their cohort had clusters present prior to treatment [26], using the Parsortix microfluidic device. Studies using the CellSearch™ system report cluster rates around 20% [27]. Current technologies significantly under-call the number of clusters because few specialised devices exist for their detection. The definition of clusters also varies in the literature from > 1 CTC to > 2 or 3 cells [19]. In this study, we define clusters as ≥ 3 CTCs but we have also conducted analyses using the other cut-offs of > 1/2 cells but no significance was seen. Clusters are more aggressive than single CTCs [18, 19] and a lot of effort is now focussed on their isolation and characterisation [20, 21]. We have previously reported the importance of platelets in cancer metastasis, which are a major component of these clusters and microemboli [28] and may explain the increased aggressiveness observed. In addition, we have shown that the platelet cloak can inhibit immune surveillance by NK cells enabling the CTCs to establish metastasis [29].

No association was seen between CTC counts and clinicopathological details apart from BMI and pre-treatment CTC counts. BMI has been shown to mediate the prognostic significance of CTCs in inflammatory breast cancer [30]. In a recent publication, patient-derived xenograft (PDX) model, tumours grown in the presence of obesity-altered adipose stem cells in SCID/beige mice had increased circulating HLA1⁺ human cells as well as increased numbers of CD44⁺CD24⁻ cancer stem cells in the peripheral blood, the authors concluded leptin produced by obesity-altered adipose stem cells promotes metastasis. Others have found a negative association with BMI and CTC counts using the CellSearch device [31, 32]. Further work is needed to assess this correlation between CTCs and BMI. Most studies including two meta-analyses [22, 23], did not observe any correlation with clinicopathological details but they did find CTC counts to have prognostic ability. The prognostic potential of CTC counts in our study has not yet been statistically assessed due to the low number of patients, presenting with recurrences but it is our intention to carry out a 5-year follow-up on these patients. It does, however, seem from the observational data that the presence of CTCs post-chemotherapy may have some prognostic potential.

Some studies have assessed correlation of CTC counts with complete pathological response. The method of measuring the pCR is not detailed in all studies. In this study, the CTC count before and after chemotherapy was correlated with radiological findings and also to the pathological score of Miller–Payne grading but no correlation was seen, 19% of the patients in our cohort had a pCR and all still had CTCs following neoadjuvant chemotherapy. Other studies report a decline in CTC counts post-neoadjuvant treatment [33] while some report an increase [34]. While CTCs were present post-treatment, we do not know the metastatic potential of these CTCs and this is needed to unlock the true potential of CTCs in the clinic. In our study, we observed an increase of 58% in CTCs and 42% in CTC clusters in the patients post-chemotherapy which is similar to the observations in a prostate cancer study where 42% of patients had an increase in CTCs post-treatment [35]. There may be many reasons for the increase in CTC counts/clusters observed post-treatment. This may be explained by the rate at which the various tumour subtypes shed tumour cells into the circulation. The chemotherapy may have caused the vessels to become leaky and shed more cells into the vasculature. It may be that the chemotherapy has selected for a clonal population of stem like cells that have survived the chemotherapy. Other mechanisms that may explain this increased number include autophagy, aniokis, expansion of pro-metastatic variants or a potential drive towards dormancy, which would be of interest for future studies. Clearly, further work is needed on assessing the viability and molecular characterisation of the CTCs that remain post-treatment. We hypothesise that CTC clusters are more likely to be viable than single CTCs. While these reasons may explain the findings in the cohort of patients who still have residual disease, it does not explain the situation for those who have had a complete pathological response. It suggests a differential response between the primary tumour and the CTCs and is similar to the findings in the Geparquattro trial [18]. Follow-up of these patients is needed to determine the long-term significance of these CTCs. One study has reported the detection of CTCs 8–22 years out from treatment, despite no clinical evidence of disease [36]. It is not known if this represents tumour dormancy and persistence of disseminated disease in the bone marrow as suggested by some investigators [37, 38] or whether a proportion of these patients will go on to develop metastatic disease. This stresses the need for molecular characterisation and long-term follow-up of patients.

Many groups are now focussing on characterising and dissecting the metastatic potential of CTCs. Our group has established a CTC-5 program which allows us to merge a morphology image of the isolated CTCs with an immunofluorescent profile for the same cells which will give us a better insight into their biology (manuscript in prep). A focus on single cell genomics and cluster dissection may reveal mechanisms of how the cells in a cluster co-operate and metastasise, enabling CTCs to have a more proactive role in the clinic. Unlocking the full potential of the liquid biopsy to monitor treatment response in breast cancer will allow us ultimately to deliver a more personalised medicine approach for cancer patients, improving therapeutic outcomes.

The finding that CTCs still exist post-neoadjuvant chemotherapy in patients who have a complete pathological response is a significant finding in the setting of what this study was trying to achieve, demonstrating that CTC enumeration may not be suitable to determine if patients can avoid surgery. Larger studies will be needed to further evaluate this finding. Additional characterisation of CTCs and CTC clusters is needed to assess the true potential of CTCs in this cohort of patients. Long-term follow-up of patients is needed to assess the significance of CTC counts.