Parallel fluctuation of circulating PMN-MDSCs and M-MDSCs in patients with mHSPC in response to ADT
We investigated three MDSC subsets, e-MDSCs, PMN-MDSCs, and M-MDSCs from PBMCs in patients with newly diagnosed mHSPC. In a total of 74 metastatic PC patients, 43 patients were excluded for the following reasons: they were CRPC patients, had no prostatic biopsy or did not consent to this study. A total of 31 patients with untreated metastatic PC patients were included in this study (Fig. 1). We first evaluated the therapeutic effect of ADT in patients with untreated metastatic PC and confirmed that all patients were sensitive to hormonal therapy. As shown in Suppl. Fig. 1, the PSA levels in all 31 patients declined from pre-ADT to 3 months post-treatment, indicating that all subjects were mHSPC. Fig. 2-A shows the representative dot plots for one of the patients in the study to illustrate the gating strategy. Since an expansion of MDSCs has been reported in patients with diabetes and also in smokers [18, 19], this was considered in the population of MDSCs in this study. No effect on MDSC frequencies from either diabetes or smoking was found in the subjects in our study (Suppl. Fig. 2). We next investigated the cell fractions from hormone-sensitive patients before and 3 months after treatment. As shown in Fig. 2-B, while there was no significant difference in the e-MDSC subset between mHSPC patients (median value :0.76%) and the healthy control (0.67%), the PMN/M-MDSC fraction was significantly elevated in patients relative to the controls before the treatment (PMN-MDSCs: 0.28% vs 0.11%, M-MDSCs: 1.46% vs 0.58%) (Fig. 2-B). The percentage of PMN/M-MDSCs in patients declined to almost the same level as in the healthy subjects after treatment, suggesting that these MDSC subsets fluctuated in conjunction with the therapeutic effects of the GnRH antagonist.
Relevance of MDSC fractions with serological and hematological examination
While it has been reported that PSA levels are associated with the frequencies of M-MDSCs [14, 16], their correlation with e-MDSCs and/or PMN-MDSCs has not been investigated in PC patients. Furthermore, there are no reports showing an association between each MDSC subset and other serological/hematological tests in PC patients. Therefore, we investigated the correlation between each of the MDSC subsets and the results of serological and hematological examinations (PSA, ALP, LDH, peripheral blood factions, biopsy Gleason score) in patients with mHSPC at the start of the treatment (Fig 3, and Suppl. Fig. 2, 3). The frequency of e-MDSCs did not correlate with serum PSA and ALP levels, but not with PSA levels. M-MDSC fractions correlated with PSA levels as previously reported [14, 16], in addition to ALP levels. MDSC have been reported to differentiate into osteoclast-like cells and cause bone destruction in mice with bone metastatic carcinoma [20], which may be reflected in the correlation with increased ALP levels. On the other hand, no correlation was found between the percentages of neutrophils, monocytes, and lymphocytes and the ratios of each of the MDSC subsets. According to these results, there is a clear correlation between PMN-MDSC and M-MDSC frequencies and the PSA and/or ALP levels in patients with newly diagnosed mHSPC.
Prognostic value of MDSCs for patients with mHSPC
The high frequency of MDSCs was shown to be associated with a poor prognosis with various malignant neoplasms via the immunosuppressive effect in the tumor-associated immune system [11, 12]. While the association between the total MDSC subsets and OS in PC patients has been reported in an earlier study [15], no studies have examined the impact of each of the MDSC subsets on the prognosis of patients with mHSPC. Therefore, we assessed the prognostic value of pretreatment MDSCs in mHSPC patients (The mean follow-up time: 567 days). During the follow-up period, 19 patients with mHSPC had PSA progression, and 3 patients expired due to tumor aggravation. In order to evaluate the prognostic value of MDSCs, we compared mHSPC patients with high and low percentages of each of the MDSC subsets. While cut-off values from retrospective studies have been used in previous studies [21, 22], it is necessary to set a more sensitive cut-off value for a prospective therapeutic intervention study in untreated subjects. We determined the cut-off value for the high percentage of PMN-MDSCs and M-MDSCs, except for e-MDSCs, at three times the median value of normal subjects, and a value well above the standard deviation (SD) of normal subjects. The cut-off value for the high percentage of e-MDSCs was set to be approximately above median values of the patient’s e-MDSCs because there were no significant differences between patients and healthy subjects. Therefore, we specifically defined the following values as cut off values for each of the MDSC subsets: e-MDSCs 0.75%, PMN-MDSCs 0.30%, M-MDSCs 1.70%. As shown in Fig. 4-A, patients with a high ratio of PMN-MDSCs displayed a lower PSA–PFS than those with a low ratio. In addition, a high frequency of PMN-MDSCs was associated with a lower OS in subjects (Fig. 4-B). However, high percentages of e-MDSCs and/or M-MDSCs did not relate with PSA–PFS and OS (Fig. 4). These results suggested that PMN-MDSCs subset gain a prognostic value for metastatic PC patients receiving first-line treatment.
Different cell fractions and gene expressions in mHSPC revealed by scRNA-seq analysis
Since the fraction of MDSCs was significantly increased in mHSPC patients, we investigated the immune cell dynamics and gene expressions of MDSCs using an scRNA-seq analysis of PBMCs from a patient with an average percentage increase in the fraction of PMN-MDSCs (Neutrophil; 64.4%, Lymphocyte; 26.3%, Monocyte; 8.4%, e-MDSCs; 1.10%, PMN-MDSCs; 0.79%, M-MDSCs 3.24%) and a healthy volunteer (Neutrophil; 51.6%, Lymphocyte; 43.3%, Monocyte; 3.4%, e-MDSCs; 0.74%, PMN-MDSCs; 0.04%, M-MDSCs 2.76%). The transcripts of cells were obtained by the 10X Genomics platform, and we acquired 21,280 and 21,697 genes of single cells from the subject and control, respectively. Fig.5-A showed the t-SNE plots displaying clusters of all immune cells of the mHSPC patient (3885 cells) and the healthy control (4639 cells). We obtained 14 clusters by unbiased graph-based clustering, and identified cell groups consisting of T cells, B cells, innate immune cells (γδT cells, NK cells, and NKT cells), and myeloid cells. A greater increase in the frequencies of innate immune cells and myeloid cells was shown in the patient than in the control (Fig. 5-A), and the gene expressions of these immune cells differed among subjects (Suppl. Table 2). For example, the expression of S100 proteins, which relates to the development and progression of various cancers, was increased in the innate immune cells and myeloid cells of the PC patient. We also confirm that the gene expressions were different even in the same cell subset by the reclustering of T cells, innate lymphoid cells and myeloid cells (Supple. Fig. 4). While the population of effecter memory CD4/8 T cells was almost the same between PC patient and control, the ratio of central memory CD4/8 T cells was decreased remarkably in the patient, inferring that the differentiation of T cells into effector cells was enhanced in the tumor-bearing state. Immune checkpoint molecules were not detected as differentially expressed genes in T cell subsets (Suppl. Table 2), whereas the S100 gene-expressing γδT cells and monocytes were increased in the mHSPC sample (Suppl. Fig. 4). In addition, we were able to detect small cell groups of PMN-MDSCs and found the enhanced expression of genes involved in tumor progression (CXCL8, IL-1b, proteoglycan versican (VCAN), and Transforming growth factor b induced (TGFBI)) (Fig. 5-B). While we acknowledge that further data accumulation is needed, our present RNA-seq data were largely consistent with previous reports on other cancers [23, 24], suggesting that these genes could also be potential therapeutic targets in mHSPC.