The Prognostic Roles of Pretreatment Circulating Tumor Cells, Circulating Cancer Stem-Like Cells, and Programmed Cell Death-1 Expression on Peripheral Lymphocytes in Patients with Initially Unresectable, Recurrent or Metastatic Head and Neck Cancer: An Exploratory Study of Three Biomarkers in One-ti

Circulating tumor cells (CTCs) and immune status are strongly related to cancer prognosis, although few studies have examined both factors. This prospective observational study (ClinicalTrials.gov: NCT02420600) evaluated whether CTCs, circulating cancer stem-like cells (cCSCs), and peripheral lymphocytes with/without Programmed cell death protein 1 (PD-1) expression were associated with prognosis among patients receiving palliative chemotherapy for initially unresectable, recurrent/metastatic head and neck squamous cell carcinoma (rmHNSCC). Thirty-four patients were enrolled between January 2015 and June 2016. Overall survival (OS) was associated with a higher CTC number (hazard ratio [HR]: 1.01, p = 0.0004) and cCSC ratio (HR: 29.903, p < 0.0001). Progression-free survival (PFS) was also associated with CTC number (HR: 1.013, p = 0.002) and cCSC ratio (HR: 10.92, p = 0.003). A CD8+ proportion of ≥ 17% was associated with improved OS (HR: 0.242, p = 0.004). A CD4: CD8 ratio of >1.2 was associated with poorer trend of PFS (HR: 2.12, p = 0.064). PD-1 expression was not associated with survival outcomes. Baseline CTCs, cCSC ratio, and CD8+ ratio may predict prognosis in rmHNSCC.

However, no studies have examined the associations of survival with PD-1 expression on peripheral lymphocytes (i.e., host immune status) plus the patient's CTC and cCSC status (i.e., markers for cancer aggressiveness). Therefore, this study aimed to evaluate whether CTCs, cCSCs and peripheral lymphocytes with/without PD-1 expression were associated with prognosis and therapy response among patients who were receiving palliative chemotherapy for initially unresectable, recurrent/metastatic HNSCC.

Patient Enrollment
Between January 2015 and June 2016, 34 patients were enrolled after receiving a detailed explanation regarding the study's design, methods, and potential risks ( Figure 1). The patients' basic characteristics are shown in Table 1, and all patients were followed until September 2017. The median follow-up time was 5.6 months ranging from 1.0-29.0 months. The median patient age was 50 years (range: 37-73 years), and most patients were male (85.3%). The most common primary sites were the oral cavity (55.9%) and oropharynx (23.5%). Most patients had an Eastern Cooperative Oncology Group performance status of 0-1 (61.4%), and a smaller subgroup had a performance status of ≥2 (38.6%). The cases were classified as initially unresectable or recurrent (n = 4) or metastatic (n = 30), and no radiation was delivered during this study, although all patients received ≥2 full cycles of palliative chemotherapy. The most common sites of distant metastasis were the lungs (53.3%) and distant lymph nodes or soft tissue metastasis (36.7%). The chemotherapy regimens mainly involved cisplatin-based regimens with or without cetuximab (82.4%). Not all patients were evaluated for p16 expression, which was only detected in 8.8% (3/34) of the total population and 27.3% (3/11) in evaluated cases. However, no studies have examined the associations of survival with PD-1 expression on peripheral lymphocytes (i.e., host immune status) plus the patient's CTC and cCSC status (i.e., markers for cancer aggressiveness). Therefore, this study aimed to evaluate whether CTCs, cCSCs and peripheral lymphocytes with/without PD-1 expression were associated with prognosis and therapy response among patients who were receiving palliative chemotherapy for initially unresectable, recurrent/metastatic HNSCC.

Patient Enrollment
Between January 2015 and June 2016, 34 patients were enrolled after receiving a detailed explanation regarding the study's design, methods, and potential risks ( Figure 1). The patients' basic characteristics are shown in Table 1, and all patients were followed until September 2017. The median follow-up time was 5.6 months ranging from 1.0-29.0 months. The median patient age was 50 years (range: 37-73 years), and most patients were male (85.3%). The most common primary sites were the oral cavity (55.9%) and oropharynx (23.5%). Most patients had an Eastern Cooperative Oncology Group performance status of 0-1 (61.4%), and a smaller subgroup had a performance status of ≥2 (38.6%). The cases were classified as initially unresectable or recurrent (n = 4) or metastatic (n = 30), and no radiation was delivered during this study, although all patients received ≥2 full cycles of palliative chemotherapy. The most common sites of distant metastasis were the lungs (53.3%) and distant lymph nodes or soft tissue metastasis (36.7%). The chemotherapy regimens mainly involved cisplatin-based regimens with or without cetuximab (82.4%). Not all patients were evaluated for p16 expression, which was only detected in 8.8% (3/34) of the total population and 27.3% (3/11) in evaluated cases.

The cCSC Ratio May Contribute to Chemoresistance
We had hypothesized that a high cCSC ratio might predict a poor response to chemotherapy, based on the chemoresistance of cCSCs. Figure 5 and Table 2 show that a higher baseline cCSC ratio predicted disease progression within the first three months of chemotherapy (p = 0.003 based on the Mann-Whitney U test).

The cCSC Ratio May Contribute to Chemoresistance
We had hypothesized that a high cCSC ratio might predict a poor response to chemotherapy, based on the chemoresistance of cCSCs. Figure 5 and Table 2 show that a higher baseline cCSC ratio predicted disease progression within the first three months of chemotherapy (p = 0.003 based on the Mann-Whitney U test).

The cCSC Ratio May Contribute to Chemoresistance
We had hypothesized that a high cCSC ratio might predict a poor response to chemotherapy, based on the chemoresistance of cCSCs. Figure 5 and Table 2 show that a higher baseline cCSC ratio predicted disease progression within the first three months of chemotherapy (p = 0.003 based on the Mann-Whitney U test).

Discussion
This prospective study evaluated 34 patients with initially unresectable, recurrent/metastatic HNSCC to examine whether chemotherapy response could be predicted using CTC, cCSC, and peripheral lymphocyte expression of PD-1. The results indicate that CTCs and cCSCs were independent predictors of cancer progression and OS and that patients with a high baseline CD8 + proportion had longer OS, while patients with a lower CD4: CD8 ratio had longer PFS. However, PD-1 expression on peripheral lymphocytes was not a significant predictor of the survival outcomes. To the best of our knowledge, this is the first study to examine CTC/cCSC and immune status using a single blood sample from patients with HNSCC.
Many studies have addressed the role of CTCs in HNSCC [25][26][27][28][29][30], including our previous study [8]. The results of the present study support the findings of the previous studies and indicate that CTC independently predicts prognosis in cases of HNSCC. In a previous stud [8], we cannot identify the prognostic role from CTC alone. One of the plausible reasons might be the improved stability and internal controls for the platform, which makes the enumerated CTC number more clinically and biologically significance by exclusion false positive signal. In addition, several studies have examined the role of cCSCs in cancer cases, which were examined using RT-PCR [31,32], counting with a gradient protocol [17,33], or flow cytometry [34,35]. Those results indicate that cCSCs are linked to a poor prognosis in cases of colorectal cancer [31,36], gastric cancer [17], prostate cancer [37], lung cancer [34,38], and hepatocellular carcinoma [32]. However, to the best of our knowledge, this is the first report to indicate that cCSCs reflect a positive prognosis in cases of HNSCC. In contrast, Park et al. [39] used a similar methodology and reported that the presence of CD133 + EpCAM + cells was associated with a poor prognosis in 10 cases of colorectal cancer. It is possible that the small sample sizes and different cancer types may explain the discrepancy between our findings and those of Park et al. Furthermore, studies have used different definitions of cCSCs, which we defined as expressing both CD133 and EpCAM after CD45 depletion. Thus, the different definitions might also explain the conflicting findings regarding the roles of cCSCs. For example, Toyoshima et al. defined cCSCs as CD44 + cells and reported that they had tumorigenicity [17], while Rentala et al. [40] defined cCSCs as cells that were CD133 + CD44 + . The most commonly used marker for identifying cCSCs is likely CD133 [31], and we included this marker to improve the ability to compare and validate our findings.
The present study revealed that high baseline peripheral CD8 + cells and low CD4: CD8 ratio were associated with a good prognosis. However, there is debate regarding the usefulness of peripheral lymphocytes versus tumor-infiltrating lymphocytes (TILs), as some investigators believe that TILs reflect the true tumor microenvironment and are correlated with prognosis and treatment response, despite the difficulty of repeated biopsy [41,42]. However, other reports have indicated that CD8 + TILs have no prognostic value in cases of HNSCC [39]. Therefore, we selected peripheral lymphocytes rather than TILs for the present study, as they are easier to monitor and may be useful as a follow-up marker (we hope to publish follow-up data in a later report).
Our results support the positive effects of higher baseline peripheral CD8 + T-cell proportions [40,43] and lower CD4:8 ratios [44,45] in cancer cases, especially among patients with HNSCC [46]. Nevertheless, our results are different from those reported by Dewyer et al., who found that pre-induction chemotherapy CD4 + T-cells but not CD8 + T-cells were correlated with a good prognosis and treatment response [47]. These differences may be related to the fact that we mainly enrolled patients with oral cavity cancer, while Dewyer et al. enrolled patients with laryngeal cancer, which suggests that the primary site may influence the prognostic markers. In addition, we examined the relative proportion of CD8 + cells, while Dewyer et al. performed absolute counting of lymphocytes. Thus, the different measurement units may explain the different findings.
The expression of PD-1 on lymphocytes has been widely discussed during recent years, and Waki et al. have reported that PD-1 expression on CD4 + and CD8 + T lymphocytes was correlated with OS among patients with lung cancer who received personalized peptide vaccination [22]. However, our study failed to detect significant relationships between the survival outcomes and the expression of PD-1 on CD4 + , CD8 + , or CD56 + T lymphocytes. It is possible that our negative findings are related to the small sample size and/or an insufficient cut-off for PD-1 positivity. For example, it may be more appropriate to use the difference between PD-1 staining and polyhydroxyalkanoate induction of control cells (Figure S1), and further adjustment of PD-1 staining or methodology standardization should be considered in future studies. However, Lyford-Pike et al. found that PD-1 expression on CD4 + T-cells and CD8 + T-cells was higher in tonsil tissue, compared to the peripheral blood, of patients with HNSCC [48], which indicates that PD-1 expression on peripheral blood lymphocytes may not be clinically significant, and our results support that possibility. Thus, PD-1 expression may only have prognostic value when it is observed on TILs.
The present study has several limitations. First, the sample size was small and a larger prospective study is needed to validate our findings. Second, despite the benefits of dynamic follow-up using peripheral lymphocytes, there is debate regarding whether they reflect the whole body's immune status, as their properties are easily disturbed. Third, although we used a clear and strict definition for cCSCs, we did not isolate these cells and confirm that they possessed in-vivo tumorigenicity, invasion, or self-renewal, which is why we referred to them as "stem-like" cells.

Study Design
This prospective observational study examined baseline CTCs, cCSCs, and PD-1 expression on peripheral lymphocytes, as well as progression-free survival (PFS) and overall survival (OS) among patients with initially unresectable, recurrent/metastatic HNSCC. The analyses were performed when more than half of the events had occurred, and the ratio of cCSC to CTC was selected to evaluate its ability to predict cancer progression and cancer-related death. Two CTC samplings were performed, with the first being performed at baseline (seven days before the first dose of chemotherapy) and the second being performed at 14-28 days after the first dose of chemotherapy. Study results were reported according to the Reporting Recommendations for Tumour Marker Prognostic Studies (REMARK) guidelines.

Patient Enrollment
Patients were enrolled at the Chang Gung Memorial Hospital (Linkou, Taiwan), and the observational protocol was approved by the institutional review board of Chang Gung Memorial Hospital (103-5322B). All patients provided written informed consent before being enrolled. Patients were considered eligible if they had: (1) histologically or cytopathologically confirmed stage IVb HNSCC (initially unresectable or recurrent) or stage IVc HNSCC (metastatic), based on the 7th edition of the American Joint Committee on Cancer criteria, (2) were ≥20 years old, (3) provided informed consent, and (4) had adequate liver and renal functions with a sufficient white blood cell count (WBC) to tolerate the anticancer therapies (especially chemotherapy). Patients were excluded if they had synchronous cancer or previous cancers within the last five years.
Blood samples were drawn within seven days before the first dose of chemotherapy. Disease staging, disease management, and treatment response evaluations followed the standard treatment protocols and institutional guidelines and were based on the findings from computed tomography, pan-endoscopy, magnetic resonance imaging, and positron emission tomography. Chemotherapy was scheduled and delivered by medical oncologists who were part of the head and neck tumor board at Chang Gung Memorial Hospital. Based on the Response Evaluation Criteria in Solid Tumors guidelines (version 1.0), treatment response was classified as complete remission, partial response, stable disease, or progressive disease. Disease-specific PFS was calculated from the date of the first CTC sampling to the first instance of cancer-specific disease progression or death, and OS was calculated from the date of the first CTC sampling to the date of death from any cause.

Identifying CTCs and cCSCs
The CTCs were identified using negative selection and positive detection strategies, which were validated in our previous studies [49]. The first step involves a negative selection protocol that depletes red blood cells using lysis and leukocytes using a CD45 depletion kit (STEMCELL Technologies Inc., Vancouver, BC, Canada). The second step involves using flow cytometry to quantitatively identify CTCs (EpCAM + Hochest + CD45 − ) and cCSCs (CD133 + EpCAM + Hochest + CD45 − ). The ratio of cCSCs to CTCs was calculated and reported as a percentage.
The CTC tests were performed using 4 mL of peripheral blood after discarding the first 4 mL of blood to avoid epithelial contamination. Red blood cell lysis was performed within 72 h, and further negative selection was performed using the EasySep CD45 Depletion Cocktail (25 µL/mL; STEMCELL Technologies Inc., Vancouver, BC, Canada) and EasySep Magnetic Nanoparticles (50 µL/mL; STEMCELL). The immunomagnetically enriched samples were subsequently spiked with OECM1 cells, labelled using an Alexa Fluor ® 488-conjugated monoclonal antibody to EpCAM (1:400; Cell Signaling Technology Inc., Danvers, MA, USA), and stained with Hoechst 33342 (1:500 in washing solution; Thermo Scientific, Waltham, MA, USA) for the nuclear staining. An isotype control antibody was used as an internal control, as well as peripheral blood samples from healthy individuals (4 mL) that were and were not spiked with 1000 OECM1 cells, which had been purchased from Taiwan's Food Industry Research and Development Institute (Hsinchu, Taiwan). Performance recovery was defined as the proportion of OECM1 cells detected using flow cytometry (BD FACSCalibur; BD Biosciences, San Jose, CA, USA) to the number of spiked OECM1 cells, and a stable coefficient of variation (CV) value has been calculated in a previous report [49,50]. To be brief, the platform can have a recovery rate of 44.6 ± 9.1% and a % coefficient of variation (CV) of 20.4%. The previous platform reported in 2015 would detect 13.1 ± 0.9 cells/mL in healthy individuals (n = 20) [49], which was confusing and might be background signal (false positive). In this revised platform, isotype control was used for each sample, which resulted in a range of 0.0-3.0 cells/mL in healthy individuals in this study cohort (n = 20). CTCs were defined as cells that were positive for both EpCAM and Draq5. Immunohistochemistry was performed to detect p16 using a prediluted mouse monoclonal antibody (CINtec p16INK4a [clone E6H4]; Roche Laboratories Inc., Westborough, MA, USA) [51] and a Dako Autostainer (Dako North America Inc., Carpinteria, CA, USA), according to the manufacturer's protocol.

PD-1 Staining and Controls
Peripheral blood mononuclear cells were isolated from 4 mL of healthy volunteers' blood using a Ficoll-Hypaque density gradient. The cells were then cultured in RPMI medium containing 10% fetal bovine serum (FBS) for 72 h with or without stimulation using 2% phytohemagglutinin (M form; Thermo Fisher GIBCO, Waltham, MA, USA.) ( Figure S1). The harvested cells were subsequently used in the immunoassays as controls that were positive and negative for PD-1 expression.

Immune Activity Assay
After red blood cell lysis, WBCs were collected from 2 mL of the patients' blood and subjected to flow cytometry after surface staining using a CD4-FITC antibody (eBiosciences; San Diego, CA, USA), a CD8a-PE-Cy7 antibody (eBiosciences), and a PD-1-PE antibody (BD Pharmingen, San Jose, CA, USA). A total of 2 × 10 5 WBCs were resuspended in 100 µL of RPMI medium with 10% FBS, 0.25 µg of the CD4-FITC antibody, 0.06 µg of the CD8a-PE-Cy7 antibody, and 20 µL of the PD-1-PE antibody. The corresponding isotype control antibodies were used under the same conditions. After staining for 1 h, the cells were washed using 2 mL of phosphate-buffered saline solution containing 10% FBS and 1 mM EDTA. The fluorescence measurements were performed using a CytoFLEX Flow cytometer (Beckman Coulter, Brea, CA, USA). The gating strategies were (1) 2 × 10 4 events for CD4 + /CD8 + -gated cells, (2) positive and negative signals for PD-1 expression were based on the CD4 + or CD8 + cursors to yield positive signals after subtracting the signal from the PE-conjugated isotype control, and (3) the percentages of CD4 + and CD8 + cells were calculated, as well as the proportions of CD4 + and CD8 + cells that were positive for PD-1.

Statistical Analysis
The patients' demographic data were reported as number (percentage) for categorical variables, and median (range) for continuous variables. Univariate and multivariate analysis were used, all factors used in the univariate analysis were examined in multivariate analysis but only those factors with statistical significance were displayed. A multivariate Cox's proportional hazard model using the forward stepwise approach was used to examine the variables' associations with OS and PFS. A risk model was also developed from the multivariate analysis. Kaplan-Meier curves and the log-rank test were used to examine any differences in the OS and PFS outcomes and determine the cutoffs, including the ratio of CD8 + and CD4: CD8. All analyses were performed using SPSS software (version 18; SPSS Inc., Chicago, IL, USA), and p-values of <0.05 were considered statistically significant.

Conclusions
This prospective study revealed that baseline CTCs and the cCSC ratio (i.e., cancer status), as well as the CD8 + proportion and the CD4: CD8 ratio (i.e., host immune status), were significantly associated with the prognosis of HNSCC. Furthermore, the baseline cCSC ratio could predict chemotherapy response during the first three months, which suggests that these cells are involved in chemoresistance.

Acknowledgments:
The authors thank all members of the Cancer Centre, Chang Gung Memorial Hospital, for their invaluable help. We also gratefully acknowledged the anonymous reviewer's invaluable advice.

Conflicts of Interest:
The authors have no conflict of interest and no competing financial interests to declare.