Increased frequency of CD117+ G-MDSCs in the peripheral blood of patients with newly diagnosed non-metastatic BC
Based on previous observations made in metastatic BC patients (34), we hypothesized that an increased frequency of circulating CD11b+ cells expressing CD117 and/or displaying a M2 activation phenotype, may also occur in patients with early BC. To test this hypothesis, we monitored the frequency of MDSC cells in the blood of non-metastatic BC patients (cT1-4, N0-1, M0) at time of diagnosis using Flow Cytometry. Aged-matched women without BC served as control population (healthy donors – HDs). Monocytic-MDSC (Mo-MDSC) were defined as CD11b+CD33+CD14highCD15- and granulocytic-MDSC (G-MDSC) as CD11b+CD33+CD14lowCD15+ cells. In both cell populations we monitored the expression of CD117, the receptor for Kit-ligand/stem cell factor widely present in hematopoietic progenitors cells (35,36), and CD163, a M2 polarization marker in monocytes (37). In order to avoid investigator-associated biases and variability in the results inherent to supervised manual analysis of flow cytometry data, we develop a minimally supervised, standardized analytical workflow based on the FlowSOM algorithm (Supplementary Fig. 1), in complement to conventional manual gating and supervised analysis.
Cells clusters revealed by standardized analytical workflow, were considered of interest when their frequency was more than 10% different between HDs and cancer patients. Some of the 14 analysed clusters corresponded to non-standard populations, for example those negative for all tested markers or expressing unanticipated marker combinations (Fig. 1A-C). These populations would have been missed by conventional supervised gating and analysis driven by the marker combination of interest. Interestingly, we observed a significant increase in the frequency of CD117+ cells among the circulating G-MDSC population in cancer patients relative to HDs (Fig. 1D). We observed a similar (but non-significant) trend in the frequency of Mo-MDSC cells, albeit at lower frequency. In addition, the frequencies of some non-classical cell populations, such as those expressing none of the markers of interest (Cluster 13 and 3) or a CD11b+CD15low cell population (Cluster 22 + 9), were significantly different between BC patients and HDs (Fig. 1E and F). No significant changes were observed for CD163+ cells in both G-MDSC and Mo-MDSC populations (Supplementary Figure S2).
Non-standard CD3 expressing cells are present with increased frequency in the peripheral blood of newly diagnosed BC patients
In parallel we monitored the presence of selected lymphocyte populations in both groups. By conventional supervised analysis we observed no differences in classical CD3+CD4+ T cells, CD3+CD8+ T cells and CD3+CD4+CD25+CD127- regulatory T cells (Tregs). Likewise, we observed no changes in the frequency of memory (CD45RO+CD45RA-) or naïve (CD45RA+CD45RO-) T cells within the same lymphocyte populations (Supplementary Figure S3).
In contrast, FlowSOM analysis performed on lymphocytes revealed 21 populations that were more then 10% differentially represented between HDs and cancer patients (Fig. 2A-C). A population of cells of the size of lymphocytes, but negative for CD3, CD4 or CD8 expression (Cluster 24 + 29) was significantly less represented in cancer patients relative to HDs. CD3+CD8+ T cells expressing CD127 and CD45RO markers (Cluster 3 + 7) are present at significantly higher frequency in cancer patients. A cell cluster expressing CD3, but not CD4 or CD8 (Cluster 20) was fond more represented in cancer patients relative to HDs (Fig. 2D-F).
Taken together these results reveal an increased frequency of peripheral blood CD117+ G-MDSC in non-metastatic BC patients at time of diagnosis, as well as significant changes in the frequency of myeloid and lymphocytic cell populations expressing unconventional marker combinations. They also demonstrate that FlowSOM-based analysis can identify cell populations that would have been likely missed by supervised analysis.
No detectable changes in the expression level of transcripts for M2 polarization markers
We previously reported that transcripts of M2-associated genes were expressed at higher levels circulating CD11b+ cells in metastatic BC patients compared to HDs (34). We therefore analysed expression of mRNA for CD117 and the M2 markers IL-10, fibronectin-1 (FN1) and arginase 1 (ARG1) in total leukocytes from cancer patient and HDs. No differences in expression levels were observed (Supplementary Figure S4).
A proof-of-concept study to monitor the effects of surgical tumor removal and adjuvant radiotherapy on circulating immune cells in BC patients
The differences observed in myelomonocytic and lymphocytic populations in BC patients at time of diagnosis relative to HDs, raised the question whether tumor removal and/or adjuvant therapy may reverse these changes, or induce additional ones. To address this question, we performed a proof-of-concept study, by taking advantage of the fact that the investigated patients were scheduled for breast conservative tumor removal and adjuvant radiotherapy as part of their standard treatment. Adjuvant radiotherapy was selected as therapy of choice as systemic effects on the immune system have been reported (38–40), while on the other side chemotherapy was excluded in order to avoid that myelosuppressive effects induced by chemotherapy could non-specifically impact the results (41). To search for potential changes in cell populations in response to surgery and radiotherapy we analysed G-MDSC and Mo-MDSCs as well as lymphocytes at three time points : after surgery/before radiotherapy start (1_PostOP), at the end of radiotherapy (6 weeks; 2_Post_RTX_6w) and at 6–8 weeks after the end of radiotherapy (12–14 weeks; 3_Post_RTX_12w). Results were compared to values obtained at time of diagnosis (0_PreOP) (Fig. 3).
Tumor removal increased the frequency Mo-MDSCs and G-MDSCs but decreased CD117+ G-MDSCs and radiotherapy induced changes in non-standard myeloid cell populations
Using FlowSOM workflow of analysis we observed distinct expression profiles at the four time-points globally visualized by tSNE. 17 cell clusters were found highly differentially represented in one group compared to the other groups. Surprisingly, when looking at the expression profiles of each cluster of interest, the majority of them was lacking CD33 expression, suggesting that this marker may not be suitable to analyse the monocytes fraction (Fig. 4A-B).
After tumor removal and at the end of radiotherapy the frequency of both Mo-MDSCs and G-MDSCs was significantly increased relative to values at time of diagnosis, and returned to pre-therapy levels 6–8 weeks after the end of radiotherapy (Fig. 4C-D). Strikingly, within the G-MDSCs population, the fractions of CD117 expressing cells significantly decreased after tumor removal and this decrease persisted after the end of radiotherapy (Fig. 4E). Surgery had no impact on the fraction of CD163+ G-MDSCs population (Fig. 4F). Radiotherapy itself had an impact on G-MDSC expressing CD163, but not CD117 (Fig. 4E-F).
The presence of one particular cell cluster expressing only CD11b and CD15 (Pop 1, 6 and 7) clearly decreased after radiotherapy. Another population expressing CD11b but lacking expression of all tested markers significantly increased during and after radiotherapy (Pop 16, 18, 12, 15 and 26) (Fig. 4G-H). The latter observation suggest that some populations defined by non-standard marker combinations may be potentially interesting candidates to investigate further with an extended panel or markers.
Analysis of total blood leukocytes for CD117, IL-10, FN1, and ARG1 mRNA expression by RT-qPCR revealed no observable differences in their expression levels (Supplementary Figure S5).
Adjuvant radiotherapy increases the frequency of CD4+ memory and regulatory T cells, and induces changes in non-standard lymphocytic populations
Likewise, we performed unsupervised analysis of the lymphocyte populations at the three time-points after surgery and radiotherapy. Visualization by tSNE revealed distinctive changes in marker expression profiles. Eleven cells clusters were found highly differentially represented in one group compared to the other ones (Fig. 5A-B). After tumor removal we observed highly variable effects on the frequency of T lymphocyte subpopulations, most of which were inconsistent and statistically non-significant.
Among the stably differentially represented clusters at the various time-points, a CD3+ cell population positive for CD4 and CD8 (Cluster 25), and a CD3+CD4+CD127+CD45RO+ population (Cluster 41) appeared at higher frequency after treatment (Fig. 5C-D). Strikingly, the frequency of this CD45RO+RA- memory subset within the CD3+CD4+ lymphocyte population was significantly and consistently increased at the end of radiotherapy and this increase was still evident 6–8 weeks later. A similar increase was also present among CD4+ regulatory T cells, corresponding to cluster 23 and 30, which also persisted after the end of radiotherapy (Fig. 5E-G).