Immune Checkpoint and EMT-Related Molecules in Circulating Tumor Cells (CTCs) from Triple Negative Breast Cancer Patients and Their Clinical Impact

Simple Summary A better understanding of the molecular mechanisms that govern metastasis and the identification of early therapeutic approaches to prevent the dissemination of tumor cells in triple negative breast cancer (TNBC) patients is highly important. The present study focuses on investigating the expression of immune checkpoint molecules (PD-L1, CTLA-4) and epithelial to mesenchymal transition (EMT)-related proteins (detyrosinated α-tubulin (GLU) and vimentin (VIM)) in TNBC patients’ CTCs and assess their relations to disease severity and clinical outcome. All the examined biomarkers were found to be expressed in CTCs, whereas PD-L1, GLU, and VIM were related to worse overall survival (OS) in TNBC patients. Our data demonstrate the importance of these four biomarkers for TNBC patients and provide an interesting tool for stratifying patients that could benefit from a potential combination of novel therapies. Abstract Triple negative breast cancer (TNBC) is the most aggressive breast cancer subtype. There are few targeted therapies for these patients, leading to an unmet need for new biomarkers. The present study aimed to investigate the expression of PD-L1, CTLA-4, GLU, and VIM in CTCs of TNBC patients. Ninety-five patients were enrolled in this study: sixty-four TNBC and thirty-one luminal. Of these patients, 60 were in the early stage, while 35 had metastatic disease. Protein expression was identified by immunofluorescence staining experiments and VyCAP analysis. All the examined proteins were upregulated in TNBC patients. The expression of the GLU+VIM+CK+ phenotype was higher (50%) in metastatic TNBC compared to early TNBC patients (17%) (p = 0.005). Among all the BC patients, a significant correlation was found between PD-L1+CD45−CK+ and CTLA-4+CD45−CK+ phenotypes (Spearman test, p = 0.024), implying an important role of dual inhibition in BC. Finally, the phenotypes GLU+VIM+CK+ and PD-L1+CD45−CK+ were associated with shorter OS in TNBC patients (OS: log-rank p = 0.048, HR = 2.9, OS: log-rank p < 0.001, HR = 8.7, respectively). Thus, PD-L1, CTLA-4, GLU, and VIM constitute significant biomarkers in TNBC associated with patients’ outcome, providing new therapeutic targets for this difficult breast cancer subtype.


Introduction
Breast cancer (BC) is a multifactorial disease and accounts for 30% of cancers in women [1]. Despite the latest progress in diagnostic and therapeutic approaches, breast Immune checkpoint molecules work as protective factors for the body's immune system, and their role is to regulate the immune system to avoid autoimmune responses caused by excessively activated immune cells [21]. In case of overexpression or overactivation of these molecules, immune function is inhibited. Tumor cells can exploit this event by excessively activating immune checkpoint molecules to prevent local immune cells from escaping surveillance and clearance, thus accelerating tumor growth [21]. Programmed death ligand 1 (PD-L1), which is an immune checkpoint molecule, is expressed in tumor cells and interacts with programmed cell death protein 1 (PD-1) on the surface of immune cells negatively regulating the immune system [22]. An association of poor prognosis in patients with advanced non-small cell lung cancer (NSCLC) has been indicated [23]. In general, elevated levels of PD-L1 in tumor cells have been linked with poor prognosis in different tumor types, such as breast cancer, head/neck cancer, gastric cancer, and prostate cancer [24][25][26][27][28][29]. Atezolizumab, avelumab, durvalumab, and cemiplimab constitute four monoclonal anti-PD-L1 antibodies that are approved by the FDA for the treatment of different types of cancer [30]. In addition, there are FDA-approved anti-PD-1 antibodies such as nivolumab, pembrolizumab, and dostarlimab for the treatment of different types of cancer, including melanoma, renal cell carcinoma, squamous lung cancer, and metastatic NSCLC [31][32][33]. Interestingly, recent studies have demonstrated that dual inhibitors, combining anti-PD1/PD-L1 antibodies with cytotoxic agents such as tubulin inhibitors, accomplish synergistic effects and show better anti-tumor efficacy than chemotherapy alone [34].
Cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) is considered the "leader" of the immune checkpoint inhibitors, as it stops potentially autoreactive T-cells at the initial stage of naive T-cell activation, typically in lymph nodes [35]. CTLA-4 is poorly understood in the BC context, and the clinical impact of CTLA-4 expression on BC treatment is still not clear [35]. There have been some studies reporting an association of its increased levels with advanced disease clinical stage, emphasizing CTLA-4's importance in the development and progression of breast cancer [35]. Thus, CTLA-4 expression in BC can be considered a potential prognostic biomarker as well as a possible therapeutic target in the emerging field of BC immunotherapy [35]. Immunotherapies targeting PD-1/ PD-L1 and CTLA-4/B7 pathways have been shown to be clinically efficient against various cancer types [30]. An extensive summary of PD-L1 and CTLA-4 molecules and their common regulatory mechanisms could be significant for the identification of patients with favorable responses to anti-PD-L1 and anti-CTLA-4 treatments or even possible targeting of these two with dual inhibitors. Dual inhibitors represent a novel and promising approach that can show synergistic and efficient effects compared to chemotherapy [34].
The expression of all these molecules has not been elucidated to date in CTCs from TNBC. Therefore, the aim of the present study was to investigate PD-L1, CTLA-4, GLU, and VIM in CTCs of TNBC patients and to assess their relationship with the severity of disease and clinical outcome.

Patients' Samples and Cytospins' Preparation
Ninety-five blood samples were obtained from BC patients (sixty-four with TNBC and thirty-one with luminal A or B). Patients were excluded from the study if they met any of the following criteria: a. Age < 18 years, b. Patients without histologically and cytologically confirmed primary breast cancer, c. Patients who had received at least one cycle of therapy, and d. Patients without a signed informed consent form. All patients gave their written informed consent, and the study was approved by the Ethics and Scientific Committees of our institution (15/12/21-6734). Patients' characteristics are presented in Table 1. The primary endpoint of the study was the identification of CTCs belonging to the examined phenotypes. The secondary endpoint was the investigation of the potential clinical relevance of the observed biomarkers for TNBC patients' outcome. Peripheral blood (10 mL in EDTA) was obtained by venipuncture after discarding the first 5 mL to avoid contamination from skin epithelial cells during sample collection. Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Hypaque density, after centrifugation at 1800 rpm for 30 min at 4 • C. PBMCs were washed twice with PBS and centrifuged at 1500 rpm for 10 min. Aliquots of 500,000 cells were centrifuged at 2000 rpm for 2 min on glass slides [17,36]. Cytospins were dried up and stored at −80 • C. In this study, we did not use any magnetic isolation with EpCAM beads because the use of two different epithelial markers (EpCAM and CK) would decrease the recovery rate of CTCs. In addition, it has been shown that EpCAM is downregulated in many CTCs [37]. Instead, we followed the published methodology used in the past in many publications of our team [17,[38][39][40].

Statistical Analysis
Spearman analysis, a Mann-Whitney test, and χ 2 tests were used to compare the two subtypes of cancer and the status of disease with the expression of PD-L1, CTLA-4, GLU, and VIM, as well as available clinical data. Kaplan-Meier survival tests for OS and PFS were performed. All analyses were performed on IBM SPSS statistics version 27 software (IBM, Armonk, NY, USA). A value of p ≤ 0.05 was used to identify significant results.

Expression of Immune Checkpoints PD-L1 and CTLA-4 in the CTCs of Luminal BC Patients Compared to TNBC Patients
To explore potential differences in the expression pattern of the examined immune checkpoint molecules between luminal and TNBC patients, thirty-one patients with luminal subtypes were also examined for the expression of immune checkpoint molecules. Among BC subtypes, the phenotype of PD-L1 + CD45 − CK + was present in 41% (26 out 64) of TNBC vs. 29% (9 out of 31) of luminal patients, while the phenotype of CTLA-4 + CD45 − CK + was present in 36% (23 out of 64) of TNBC and 23% (7 out of 31) of luminal patients ( Figure 4A and Table S1). Furthermore, the frequency of the PD-L1 + CD45 − CK + phenotype was 78% vs. 65% among TNBC and luminal patients' CTCs, respectively, while the frequency of the CTLA-4 + CD45 − CK + phenotype was 69% vs. 59% among TNBC and luminal patients' CTCs, respectively ( Figure 4C). Consequently, expression of PD-L1 and CTLA-4 was higher in TNBC compared to luminal patients, although the differences do not reach statistical significance (Table S1).
metastatic TNBC compared to early TNBC patients' CTCs, whereas the phenotype PD-L1 − CD45 − CK + was higher in early TNBC patients' CTCs. Representative images of all the different phenotypes of CTCs are illustrated ( Figure 3, Table S2).

Expression of Immune Checkpoints PD-L1 and CTLA-4 in the CTCs of Luminal BC Patients Compared to TNBC Patients
To explore potential differences in the expression pattern of the examined immune checkpoint molecules between luminal and TNBC patients, thirty-one patients with luminal subtypes were also examined for the expression of immune checkpoint molecules. Among BC subtypes, the phenotype of PD-L1 + CD45 − CK + was present in 41% (26 out 64) of TNBC vs. 29% (9 out of 31) of luminal patients, while the phenotype of CTLA-4 + CD45 − CK + was present in 36% (23 out of 64) of TNBC and 23% (7 out of 31) of luminal patients ( Figure 4A and Table S1). Furthermore, the frequency of the PD-L1 + CD45 − CK + phenotype was 78% vs. 65% among TNBC and luminal patients' CTCs, respectively, while the frequency of the CTLA-4 + CD45 − CK + phenotype was 69% vs. 59% among TNBC and luminal patients' CTCs, respectively ( Figure 4C). Consequently, expression of PD-L1 and CTLA-4 was higher in TNBC compared to luminal patients, although the differences do not reach statistical significance (Table S1).  Concentrating on the status of the disease, the PD-L1 + CD45 − CK + phenotype was present in 38% (16 out of 42) of early and 45% (10 out of 22) of metastatic TNBC patients, while it was present in 17% (3 out of 18) of early and 46% (6 out of 13) of metastatic luminal patients ( Figure 4B and Table S1).
On the other hand, the CTLA-4 + CD45 − CK + phenotype was present in 26% (11 out of 42) of early and 50% (11 out of 22) of metastatic TNBC patients (p = 0.025), while it was present in 17% (3 out of 18) of early and 31% (4 out of 13) of metastatic luminal patients ( Figure 4B and Table S1). Therefore, expression of CTLA-4 was higher in metastatic com- Concentrating on the status of the disease, the PD-L1 + CD45 − CK + phenotype was present in 38% (16 out of 42) of early and 45% (10 out of 22) of metastatic TNBC patients, while it was present in 17% (3 out of 18) of early and 46% (6 out of 13) of metastatic luminal patients ( Figure 4B and Table S1).
On the other hand, the CTLA-4 + CD45 − CK + phenotype was present in 26% (11 out of 42) of early and 50% (11 out of 22) of metastatic TNBC patients (p = 0.025), while it was present in 17% (3 out of 18) of early and 31% (4 out of 13) of metastatic luminal patients ( Figure 4B and Table S1). Therefore, expression of CTLA-4 was higher in metastatic compared to early patients with both subtypes.

PD-L1, CTLA-4, GLU, and VIM in the CTCs of BC Patients and Clinical Outcome
Clinical data regarding follow-up were available for 60 patients (48 out of 64 TNBC and 12 out of 31 luminal). After a median follow-up period of 7 months for TNBC and 53 months for luminal patients, 13 deaths and 1 death, respectively, were registered due to disease progression.
Among all BC patients, as expected, the metastatic ones were associated with a shorter OS (log-rank p < 0.001, HR = 9.7).

Discussion
In the present study, we characterized the CTCs from TNBC patients regarding the immune checkpoint antigens PD-L1 and CTLA-4 plus the expression of the EMT-related molecules GLU and VIM in CTCs of TNBC patients. In addition, we compared the expression of PD-L1 and CTLA-4 in CTCs of TNBC vs. luminal patients. We also examined the correlation of all these molecules with the severity of the disease and clinical outcome.
Based on the disease status, 17% vs. 50% of patients with early and metastatic TNBC, respectively, expressed the phenotype GLU + VIM + CK + (p = 0.005). Therefore, the expression of this phenotype was higher in the metastatic setting. This is also supported by another study from our group including mainly luminal patients, where expression levels of GLU and VIM were higher in metastatic BC patients and this expression was related to patients' outcome [8]. Tubulin detyrosination, which results from the suppression of tubulin tyrosine ligase and the resulting unbalanced activity of tubulin-carboxypeptidase, apparently represents a strong selective advantage for cancer cells and has been linked to poor prognosis in BC [42]. Interestingly, our findings indicated that the phenotype GLU + VIM + CK + was associated with shorter OS of TNBC patients (log-rank p = 0.048). This finding is also similar to the results observed in NSCLC patients, where the OS has been shown to significantly decrease in patients with high GLU (3.8 vs. 7.9 months; p = 0.018) and/or high VIM (3.2 vs. 7.1 months; p = 0.029) expression in their CTCs [43]. Furthermore, the detection of GLU + VIM + CK + tumor cells in peripheral blood has been correlated to shorter progression free survival (PFS) of BC patients [8]. Overexpression of GLU and VIM has been shown in cytoskeletal structures called microtentacles, which represent an important mechanism for metastatic dissemination and are associated with EMT pathways [8]. The findings of that study showed that CTCs could interact with one another through filamentous protrusions that were supported by TUB, VIM, and GLU [8]. Previous research has revealed that invasive breast carcinomas expressing vimentin have a higher rate of tubulin microtentacles after detachment than non-invasive cell lines that do not express vimentin [44]. Additionally, it has been demonstrated that GLU levels significantly rise after epithelial cell detachment, and the microtentacles that are formed due to detachment are enriched in GLU [20]. The preservation of microtentacles after detachment suggests that vimentin is aligned with GLU in microtentacles, while cytokeratin is not [44]. Based on the above evidence, GLU and VIM are associated with disease progression and constitute important biomarkers in TN breast cancer.
One of the main objectives of this study was also to evaluate the expression of immune checkpoint molecules such as PD-L1 and CTLA-4 in TNBC and luminal patients and associate these results with their clinical outcome. Among BC subtypes, the phenotype of PD-L1 + CD45 − CK + was present in 41% of TNBC and 29% of luminal patients, while the phenotype of CTLA-4 + CD45 − CK + was present in 36% of TNBC and 23% of luminal patients, implying that immunosuppressive phenotypes potentially predominate in TN breast cancer. In addition, the percentage of the PD-L1 + CD45 − CK + phenotype was 78% vs. 65% among TNBC and luminal patients' CTCs, respectively, while the percentage of the CTLA-4 + CD45 − CK + phenotype was 69% vs. 59% among TNBC and luminal patients' CTCs, respectively. Hence, the expression of PD-L1 and CTLA-4 seemed to be higher in TNBC compared to luminal patients, possibly linked to the higher aggressiveness of the disease, although this finding did not reach statistical significance. In agreement with these findings, it has been shown that CTLA-4 level was significantly higher in TNBC than in the luminal subtype and HER2+ subtype (p = 0.019 and p < 0.001, separately), and was significantly higher in ER-and PR-negative samples than in ER-and PR-positive samples (p < 0.001) [45]. Similarly, it has been demonstrated that PD-L1 expression levels are higher in TNBC than any other breast cancer subtype [46].
Based on the status of the disease, the percentage of the PD-L1 + CD45 − CK + phenotype was 70% vs. 91% among early and metastatic TNBC patients' CTCs, respectively, while in luminal patients the same percentages were 60% vs. 70%, respectively ( Figure 4D). The CTLA-4 + CD45 − CK + phenotype was present in 26% of early vs. 50% of metastatic TNBC patients (p = 0.025), while it was present in 17% of early vs. 31% of metastatic luminal patients ( Figure 4B). These findings indicate a slightly higher prevalence of both immune checkpoint phenotypes (PD-L1 + CD45 − CK + and CTLA-4 + CD45 − CK + ) in metastatic settings of both TNBC and luminal patients, implying that these CTCs during disease evolution possess an increased capacity to suppress the immune system and therefore represent potential suitable targets for anti-PDL1 and anti-CTLA-4 therapies. Similar findings demonstrated significant evidence that PD-L1 was frequently expressed on metastatic cells circulating in the blood of HR (+) and HER2 (−) breast cancer patients, whereas other studies have suggested that higher levels of CTLA-4 may be linked to more advanced stages of breast cancer, highlighting the significance of CTLA-4 in the growth and progress of the disease [24,35]. Interestingly, the phenotype PD-L1 + CD45 − CK + was associated with shorter OS of TNBC patients. Specifically, all TNBC patients, including early TNBC patients with two or more PD-L1+ CTCs, had shorter OS (log-rank p = 0.048 and p < 0.001 log-rank respectively) than those with one or no PD-L1+ CTCs. This finding is consistent with other studies which have indicated that high levels of PD-L1 expression are linked to a significantly worse overall survival rate in breast cancer and NSCLC patients [43,[47][48][49]. Additionally, it has been demonstrated that patients with metastatic breast cancer (MBC) harboring PD-L1 + CTCs had a shorter PFS, indicating that PD-L1 expression in MBC acts as a negative prognostic biomarker and highlighting the importance of this as a biomarker [50]. It is also noteworthy that the poorer OS (p < 0.001) was also confirmed in early TNBC patients, implying that PD-L1 could be a useful biomarker in the early stages of the disease for the TNBC subtype.
Among TNBC patients, significant correlation was found between PD-L1 + CD45 − CK + and GLU + VIM − CK + phenotypes (Spearman test, p = 0.007). This possibly explains the efficiency of the new dual inhibitors targeting both tubulin and the PD-1/PD-L1 axis. More specifically, a series of novel CA-4 analogs as dual inhibitors of tubulin polymerization and PD-1/PD-L1 were designed, synthesized, and bioevaluated. Among these CA-4 analogs, TP5 was the most effective as it exhibited moderate anti-PD-1/PD-L1 activity, while it could also effectively inhibit tubulin polymerization, suppressing HepG2 cell migration and colony formation, causing cell arrest at G2/M phase and inducing apoptosis [34]. Dual inhibitors or multitargeting drugs possess the prospect to complement or even replace chemotherapy or therapeutic regimens based on drug combinations, as they can show either additive or synergistic effects and guarantee the simultaneous presence of the molecule at the sites of action as well as interaction with its multiple targets [51].
Notably, a significant correlation was also found between PD-L1 + CD45 − CK + and CTLA-4 + CD45 − CK + phenotypes among all the BC patients (Spearman test, p = 0.024). This finding demonstrates a connection between CTLA-4/B7 and PD1/PD-L1 pathways, the two most representative immune checkpoint molecules, which negatively regulate T-cell immune function during different phases of T-cell activation [52]. Targeting checkpoints of immune cell activation has been demonstrated to be the most effective approach for the activation of anti-tumor immune responses [30]. Inhibitors targeting these pathways have revolutionized immunotherapies for several cancer types. Ipilimumab, an FDA-approved antibody targeting CTLA-4, seems to restore tumor immunity at the priming phase, whereas anti-PD-1/PD-L1 antibodies, such as pembrolizumab and nivolumab, restore immune function in the tumor microenvironment [53]. A novel and really promising approach is the use of dual inhibitors which can show synergistic and efficient effects compared to chemotherapy. An example is the development of indole alkaloid-type dual immune checkpoint inhibitors against CTLA-4 and PD-L1 based on diversity-enhanced extracts which were indicated to suppress not only CTLA-4 and PD-L1 gene expression but also protein expression on the cell surface [54].
Our study highlighted the importance of PD-L1, CTLA-4, GLU, and VIM as biomarkers in TNBC, the correlation among them, and their potential clinical significance. Furthermore, our analysis provided evidence that supports the necessity of novel and prospective dual inhibitors targeting both tubulin and the PD-1/PD-L1 axis as well as dual inhibitors against CTLA-4 and PD-L1. These findings could be of clinical importance in preventing the dissemination of cancer cells in the bloodstream.

Conclusions
TNBC is the most aggressive subtype and there is a limitation in the targeted therapeutic approaches for these patients. The field of liquid biopsy can be very beneficial in terms of stratifying patients in real time and identifying subpopulations that can benefit from specific treatments. To the best of our knowledge, this is the first study focusing simultaneously on PD-L1, CTLA-4, GLU, and VIM expression in CTCs of TNBC patients. The findings of this study demonstrate the importance of these four biomarkers which are shown to be relevant to TNBC patients' outcomes and provide an interesting tool for stratifying patients that could benefit from the novel and prospective dual inhibitors for therapy.