Background
Breast cancer (BC) is the most frequent cancer and the main cause of cancer-related mortality for women in industrialized countries [
1]. Three clinically relevant biological BC subtypes (i.e., estrogens/progesterone receptor positive (ER+/PR
+), human epidermal growth factor receptor 2 (HER2) amplified, and triple negative) and multiple molecular subtypes (e.g., Luminal A/B, HER2, basal like, normal like) with distinct features and clinical outcomes have been defined and characterized [
2‐
5].
Early detection and surgery in combination with adjuvant treatments tailored on biological and molecular subtypes have improved patients’ survival by about 30% in the past three decades [
6]. The goal of adjuvant therapy, including radiotherapy, is the eradication of tumor cells that are disseminated before diagnosis and surgery. Some of these disseminated tumor cells (DTC), however, will escape therapy and later progress to form metastases, which in most patients represent the main cause of cancer-related death. After breast-conserving surgery, radiotherapy reduces the risk of BC recurrence and death. Among women with operable BC, randomized trials have demonstrated equivalent disease-free and overall survival between mastectomy and breast-conserving surgery followed by radiotherapy alone and/or hormonal, anti-HER2, or chemotherapy [
7‐
16].
Mammography is the standard approach for the detection of asymptomatic BC [
17]. In spite of its benefits in reducing BC specific mortality, mammography has some important limitations [
18]: low specificity and sensitivity; risk of over-diagnosis; risk of inducing BC due to X-ray exposure, particularly in patients with defective DNA repair genes [
19]; and not recommended before the age of 50 in spite of the fact that 20–25% of all BCs appear before this age. There is therefore an unmet need for complementary or alternative methods for the detection of asymptomatic, early BC [
20‐
22]. Circulating tumor cells (CTC), cell-free tumor-derived DNA, mRNA and miRNA, proteins, autoantibodies, and metabolites are being explored as candidate blood-based biomarkers for BC detection, diagnosis, or monitoring, but so far none entered routine clinical practice [
23‐
27]. Similarly, there are no effective blood-based biomarkers to actively assess patients’ response to treatment and monitoring disease state after therapy. Also, the most used blood biomarker in clinical practice, CA 15-3, is not specific and sensitive in early breast cancer diagnosis [
28].
Tumors, including BC, mobilize and recruit immuno-inflammatory cells to their microenvironment [
29‐
31]. Monocytic and granulocytic cells, mostly immature forms, as well as lymphocytes, contribute to cancer progression by promoting immunosuppression, angiogenesis, cancer cell survival, growth, invasion, and metastasis [
32,
33]. We have previously shown that metastatic BC patients have elevated frequencies of TIE2
+CD11b
+ and CD117
+CD11b
+ leukocytes circulating in the blood, and that circulating CD11b
+ cells express higher mRNA levels of the M2 polarization markers CD163, ARG1, and IL-10 [
34]. Treatment with paclitaxel in combination with bevacizumab decreased the frequency CD117
+CD11b
+ leukocytes, IL-10 mRNA levels in CD11b
+ cells, and IL-10 protein in plasma. We therefore considered that blood circulating leukocytes, or sub-population, thereof, may reflect cancer-relevant immuno-inflammatory events that may be further explored as BC-associated biomarkers.
Here, we analyzed the phenotype of blood leukocytes of patients with early BC at time of diagnosis, after surgery, and after adjuvant radiotherapy (RTX), relative to healthy donors (HD), using flow cytometry, and a minimally supervised analytical approach based on FlowSOM algorithm and manual validation. We identified with both approaches cell populations associated with the presence of a primary BC, tumor removal, and adjuvant radiotherapy. These results indicate that phenotypical analysis of peripheral blood leukocytes, with a minimally supervised analytical approach, may be a clinically-relevant strategy for the identification of cellular biomarkers for BC detection and therapy monitoring.
Materials and methods
Patients and clinical study
The study was approved by the Cantonal ethic commission for human research on Humans of Canton Ticino (CE 2967) and extended to Vaud-Fribourg-Neuchâtel, Switzerland. The study includes 13 female patients (Table
1) who were diagnosed with primary, non-metastatic BC (stage T1–4, N0–N1, M0,). All patients underwent conservative surgery and received standard fractionated adjuvant radiotherapy (2 Gy per session, total dose : 50 + 10 Gy). For the analysis at time of primary detection, only 11 patients were included for comparison with 11 age-matched healthy donors (HDs). Blood samples were collected at the following time-points (Fig.
3): after diagnosis was confirmed histologically but before surgery (Sample 0); after surgery the day of radiotherapy start (immediately before first irradiation, Sample 1); at the last day of radiotherapy (6 weeks after starting radiotherapy, Sample 2); and 6–8 weeks after the end of the radiotherapy (for the majority to the patients this was 12 weeks after starting radiotherapy, Sample 3). All Patients and HDs gave written informed consent before study entry. Patients were recruited before surgery at Clinica Luganese Moncucco, Lugano, and at Hôpital Neuchâtelois, La Chaux-de-Fonds, once diagnosis was histologically confirmed. Mean age for cancer patients was 60.6 years (all patients were between 43 and 73 years old). HDs were recruited along the study, based on the following criteria: age-matched relative to BC patients, no regular medications in the last 6 months, no previous cancer diagnosis, no chronic diseases, and normal blood analyses at time of recruitment.
Table 1
Clinical-pathological data of breast cancer patients included in the study
1 | 62 | 95 | 70 | − | 5 | 1 | pT1b | pN1a | − |
2 | 58 | 95 | 95 | − | 10 | 2 | T1b | N0 | − |
3 | 50 | 100 | 100 | + | 5 | 1 | pT1 | pN0 | Tamoxifen |
4 | 73 | 95 | 2 | − | 25 | 2 | pT2 | pN1a | − |
5 | 49 | 90 | 80 | − | 10 | 2 | T1b | pN0 | Tamoxifen |
6 | 69 | 100 | 100 | − | 15 | 2 | pT1a | pN0 | Tamoxifen |
7 | 53 | 95 | 60 | − | 10 | 2 | pT1c | pN0 | Letrozole |
8 | 73 | 100 | 0 | − | 20 | na | pT1c | pN0 | Tamoxifen |
9 | 67 | 90 | 80 | − | 5 | 1 | pT1b | pN0 | Tamoxifen |
10 | 66 | 100 | 100 | − | 5 | 2 | pT1c | pN0 | Letrozole |
11 | 64 | 95 | 80 | − | 10 | 1 | pT1b | pN0 | Letrozole |
12 | 61 | 80 | 100 | − | 10 | 2 | pT1b | pN0 | Anastrozole |
13 | 43 | 95 | 95 | − | 10 | 2 | pT1c | pN0 | Tamoxifen |
Blood processing
Twenty milliliters of peripheral venous blood was collected using BD Vacutainer® Blood Collection EDTA Tubes (Becton Dickinson, Franklin Lakes, NJ, USA) following the manufacturer’s instructions and immediately shipped by courier at room temperature to the laboratory. All analyses were performed within 24 h after blood collection. Antibody staining was performed in whole blood. Plasma and total leukocytes were isolated from the remaining blood using BD Vacutainer® CPT™ Cell Preparation Tube (Becton Dickinson) with sodium heparin following the manufacturer’s instructions. Plasma fraction was frozen at – 80 °C and isolated leukocytes were lysed in RA1 lysis buffer (Macherey-Nagel, Düren, Deutschland) and stored at – 80 °C.
Flow cytometry
Whole blood stainings were performed within 24 h after blood collection. Leukocytes were counted using Cell-Dyn Sapphire Hematology System (Abbott Diagnostics, Chicago, IL, USA). For staining, 1 million cells per tube were used based on direct blood cell count. Directly labeled antibodies were added to whole blood and incubated for 20 min at 4 °C, followed by 10 min red-blood-cells lysis (Bühlmann Laboratories, Schönenbuch, Switzerland) and subsequently washed using cold PBS. All anti-human antibodies were used at the concentrations recommended by the manufacturer: anti-CD15-PeCy7 (clone HI98), anti-CD14-Pe (clone MφP9), anti-CD163-FITC (clone GHI/61), anti-CD11b-BV510 (clone ICRF44), anti-CD33-V450 (clone WM53), anti-CD64-APCH7 (clone 10.1), anti-CD117-APC (clone YB5.B8), anti-CD45RA-PeCy7 (clone HI100), anti-CD25-Pe (clone M-A251), anti-CD4-FITC (clone RPA-T4), anti-CD8-V500 (clone SK1), anti-CD45RO-BV421 (clone UCHL1), anti-CD3-APCH7 (clone SK7), and CD127-Alexa Fluor 647 (clone HIL-7R-M21) and 7AAD (all from Becton Dickinson). BD FACSCanto II (Becton Dickinson) instrument was used to analyze samples and FlowJo 10.6.2 (Treestar Inc., Ashland, OR, USA) software and several software plugins (FlowCLEAN, downsample_V3, FlowSOM, tSNE) were used to analyze all data.
Reverse transcription real-time PCR (RT-qPCR)
Total mRNA from total white blood cells was extracted using the NucleoSpin RNA kit from Macherey-Nagel following the manufacturer’s instructions (Düren, Germany). The purity and quantity of all RNA samples were examined by NanoDrop (Witec AG, Luzern, Switzerland). Total RNA was retro-transcribed using M-MLV reverse transcriptase kit following the manufacturer’s instructions (ThermoFisher Scientific, Waltham, Massachusetts, USA) using 500 ng of total RNA. cDNA was subjected to amplification by real-time qPCR with the StepOne SYBR System (Life Technologies) using the following primer pairs (Eurofins Genomics, Huntsville, AL, USA) at the indicated hybridization temperatures: GAPDH 58 °C (Fw-TCTTCTTTTGCGTCGCCAGC, Rev-GATTTTGGAGGGATCTCGCTCCT), ARG1 58 °C (Fw-GGAGTCATCTGGGTGGATGC, Rev-CTGGCACATCGGGAATCTTTC), IL-10 58 °C (Fw-CGAGATGCCTTCAGCAGAGT, Rev-AATCGATGACAGCGCCGTAG), CD117 57 °C (Fw-GATTATCCCAAGTCTGAGAATGAA, Rev-CGTCAGAATTGGACACTAGGA), FN1 52 °C (Fw- ACTTCGACAGGACCACTTGA, Rev-TCAAATTGGAGATTCATGGGA). Real-time PCR data were then analyzed using the comparative Ct method [
35].
Statistical analysis
Acquired data were analyzed and graphics were generated using Prism Software (GraphPad, La Jolla, CA, USA). Samples with incomplete staining due to technical problems during the antibody staining process or during acquisition were excluded from the statistical analysis. Statistical comparisons between cancer patients and healthy donors were performed by T test assuming non-homogenous variance. Normality distribution of the samples was checked in case of significance and, if non-Gaussian, a Mann-Whitney replaced the T test results. Statistical comparisons of all time-points to observe the effect of radiotherapy were performed by one-way ANOVA assuming non-homogenous variance using Tukey correction. Normality distribution of the samples was checked in case of significance and, if non-Gaussian, a Kruskal-Wallis assay replaced the ANOVA results. Results were considered to be significant from p < 0.05. In the figures, the various p values thresholds are presented as follows: ≤ 0.05 = *, ≤ 0.01 = **, ≤ 0.001 = ***, ≤ 0.0001 = ****.
Discussion
Mammography-based screening significantly reduces BC-related mortality, but intrinsic and practical limitations call for novel screening approaches [
17–20]. Blood-based biomarkers exploiting CTC, cancer-derived DNA or RNA, are being explored, but so far none reached clinical routine practice [
22,
43]. Similarly, there are no validated biomarkers for monitoring patients’ response to treatment or detecting relapses before they become symptomatic.
In this study, we pursued the use of flow cytometry to analyze the phenotype and frequency of blood leukocytes in patients with non-metastatic BC at the time of diagnosis, after surgical tumor removal, and after adjuvant radiotherapy. Using a combination of minimally supervised (FlowSOM algorithm), and supervised (manual) analytical approaches, we report that (i) at diagnosis, BC patients have an increased frequency of circulating CD117+ granulocytes relative to age-matched healthy donors; (ii) surgical tumor removal causes a transient increase of monocytes and granulocytes, and a long-lasting decrease of CD117+ granulocytes; (iii) radiotherapy significantly increases CD45R0+ memory T cells and CD4+ Treg cells; and (iv) with the FlowSOM algorithm, we identified additional unanticipated, non-classical cell populations differentially represented between HD and BC patients and in BC patients in response to therapy.
Traditionally, flow cytometry results are analyzed manually, which has the potential to introduce investigator-specific biases. To tested for this, we asked a technician expert with flow cytometry analysis but who was not involved in this project, to reanalyze the raw data and to find four specific populations of interest. Reassuringly this analysis confirmed the original results (not shown). A full reanalysis, however, is time-consuming and hard to implements in a routine (clinical) setting. As an alternative to avoid potential human biases and variability of the result, we choose to perform unsupervised, algorithm-based analyses. Also, algorithms are more effective in finding potentially interesting marker combinations. Populations of interest were then investigated more in detail manually. Here too, we performed several times the same unsupervised analysis and we could reproducibly find the same populations (not shown). This confirms the value of the approach and of the populations identified, further supporting the use of an unsupervised analysis approach to prospectively identify robust changes among complex populations.
One question that emerged during the unsupervised analysis is how to set the cutoff between samples to discriminate robust marker combinations (clusters of interest) form from unstable ones. As no standard or optimal method is defined for this kind of analyses, we tested cut-off values of 5%, 10%, and 20%. The 5% cutoff was not stringent enough, because many identified populations could not be confirmed. With a 10% cutoff, we observed a good balance between identify robust, reproducible populations, and non-interesting or unstable ones. With a 20% cutoff, we missed small but robust differences seen with 5 and 10% cutoff. We therefore decided to use the 10% cutoff.
CD117, the receptor for Kit-ligand/stem cell factor, is widely expressed in hematopoietic progenitor cells in the bone marrow, while CD117
+ leukocytes are rarely detectable in the circulation under homeostatic conditions [
36]. We have previously reported a role of CD117
+ leukocytes in metastasis in the murine 4T1 metastatic BC model [
44] and the presence of CD117
+CD11b
+ cells in the blood of mBC patients [
34]. Here, we observed an increased frequency of a CD117
+ population among total granulocytes in the peripheral blood of non-metastatic BC patients at the time of diagnosis, compared to HD. Interestingly, the frequency of CD117
+ granulocytes significantly dropped upon tumor removal and remained below pre-treatment levels after radiotherapy. Thus, the increased frequency of CD117
+ granulocytes may reflect the presence of the primary tumor. No changes were observed in CD117 mRNA expression in total leukocytes. This could be due to the fact that CD117
+ cells are lost during leukocyte isolation for RNA extraction (flow cytometry was performed in non-separated total whole blood), or that CD117 mRNA expression has ceased upon cell mobilization (while CD117 protein persisted at the cell surface). The latter possibility is consistent with our previous observation that mobilized CD117
+ cells adoptively transferred to a recipient mouse, rapidly became CD117 negative [
44]. In contrast, the frequency of CD163
+ granulocytes remained rather constant, with only a transient decrease during radiotherapy. The implication of this decrease is unclear as CD163 expression did not significantly differ between HD and BC patients at the time of diagnosis. After surgery and radiotherapy also no change in the mRNA expression of CD117 and ARG1, FN1, IL-10 (i.e., M2 polarization markers) was observed, owing probably to the lack of enrichment of CD11b
+ cells for PCR analysis.
In cancer patients at the time of diagnosis, we observed a higher frequency of atypical T lymphocytes (CD3+CD4−CD8−) and of a population of the size of lymphocytes lacking expression of all the tested markers. These observations suggest that some potentially interesting changes may occur in atypical T cells or in non-T cell populations such as B cells or NK cells. Strikingly, after radiotherapy, we observed a steady and significant increase of the fraction of CD45RO+ memory T cells within total CD4+ T cells and within CD4+ Treg. We also observed the increased presence of a T cell population expressing both CD4 and CD8 markers. This suggests that radiotherapy may cause T cell activation leading to the subsequent generation of memory T cell subsets. All patients included in this study had ER+ cancer, and the majority (10/13) received concomitant anti-hormonal treatment. We were thoughtful to the potential effect of anti-hormonal therapy and we analyzed our results by stratifying patients based on anti-hormonal therapy. No significative changes in the cell populations were observed (data not shown). However, as there were only three patients without anti-hormonal treatment, the significance of these results has to be considered with care.
There is increasing evidence that radiotherapy exerts its therapeutic effects, not only in the local treatment field, but also at distant sites (i.e., the so-called abscopal effect), at least in part, by eliciting a T cell immune response [
39]. The recent observation that the combination of radiotherapy with immune checkpoint inhibitors in experimental models and cancer patients results in potent synergistic therapeutic effects further supports the involvement of T cell-dependent events and the therapeutic effects of radiotherapy [
45–48]. Through experimental work and mathematical modeling, it has been proposed that anti-tumor T cells may be mobilized by radiotherapy toward peripheral tissues to eliminate DTC [
49–51]. However, to date, there is paucity of human data demonstrating specific changes in circulating T lymphocytes to support such a model. Radiotherapy was reported to cause a global reduction in circulating lymphocyte subsets in patients treated for stage I–II prostate cancer [
52] or to induce an increase in CD4
+ Treg in the peripheral blood of patients with diverse solid cancers [
53]. Low-dose radon therapy for chronic inflammatory diseases was shown to induce a long-lasting increase in circulating T cells paralleled with a reduced expression of activation markers [
54]. Thus, the observed effect of adjuvant radiotherapy on memory CD4
+ T cells is novel and should be further explored in conjunction with patients’ outcome, as possible biomarkers of therapy response or efficacy.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.