Fig 2: Manual versus machine gating of MDSC populations. (A) Representative manual gating strategy used to identify MDSC. M-MDSCs are CD11b+, HLA-DR-, CD15-, CD14+; PMN-MDSCs are CD15+, CD14-, CD11b+, SSChi; e-MDSCs are CD14-, CD15-, CD3-, CD19-, CD20-, CD56-, HLA-DR-, CD33+, CD11b+. (B) Stepwise integration of supervised, machine learning techniques facilitates analysis of high parameter flow cytometric data. Rudimentary gating isolates viable singlets, and subsequent manual gating for the ‘lowest common denominator’ (CD11b+ surface marker, in this case) enriches the population of interest. Downsizing to a minimal, common cell count eliminates sampling bias. (C) cyCombine corrects for technical artifacts arising from different experiments. (D) User optimizes number of FlowSOM clusters to appropriately fit the model to the data set. FlowSOM, generates a minimal spanning tree and illustrates marker positivity of each identified cluster. (E) Dimensionality reduction allows two-dimensional data representation while preserving the global data structure. Proximal clusters on the 2-dimensional plot have closely related surface marker profiles. (F) FlowSOM and UMAP data are integrated in Cluster Explorer, allowing in-depth data exploration and cluster interrogation. PMN-MDSC, M-MDSC and e-MDSC were determined based on the definitions established in A, and by using a conservative estimate of >10% of maximum positive/negative intensity to denote marker positive/negative status respectively. Unannotated cell populations do not meet criteria for M-, PMN- or e-MDSC. (G) Hyperfinder optimizes polygon gating of populations of interest. (H) The Hyperfinder algorithm is then applied to flow data from step (B) Displayed is an example of Hyperfinder gating for e-MDSC cluster.

Fig 3: MPA is associated with coronary pathology and serves as a biomarker for coronary artery aneurysm. A) Flowchart depicting the overall experimental design of this study. B) Gating strategy for MPA (CD14+CD41+) and box plots showing the cell proportion of MPA (CD14+CD41+) in PBMC from participants. Lineage‐APC indicates anti‐human Lineage Cocktail (CD3, CD19, CD20, CD56). Kruskal–Wallis test and Dunn's multiple comparisons test. C) Gating strategy for CD14+CD16+ monocytes and box plots showing the cell proportion of CD14+CD16+ monocytes in PBMC from participants. Lineage‐APC indicates anti‐human Lineage Cocktail (CD3, CD19, CD20, CD56). One‐way ANOVA and Tukey's multiple comparisons test. D,E) Correlation of MPA ratio to (D) IL‐1β (n = 18), (E) TNF‐ɑ (n = 18) in individuals with Acute KD. F) The proportion (%) of MPA in HS (n = 60), Acute KD (n = 32), and Recovered KD (n = 60) in the diagnostic cohort was measured relative to Lin− (CD3, CD19, CD20, CD56) CD11b+ cells. Kruskal–Wallis test and Dunn's multiple comparisons test. G) The proportion (%) of MPA in patients with NCAA (n = 16) and CAA (n = 16) during Acute KD was measured relative to Lin− (CD3, CD19, CD20, CD56) CD11b+ cells. Unpaired t test. H) A receiver operating characteristic (ROC) analysis was performed to evaluate the ability of the MPA ratio during the acute phase to distinguish KD patients with NCAA from CAA. HS, healthy subject; KD, Kawasaki disease; MPA, platelet–monocyte aggregate; NCAA, noncoronary artery aneurysm; CAA, coronary artery aneurysm; ANOVA, analysis of variance.

Fig 4: Flow cytometric identification and phenotypic characterization of human circulating MDSC subsets in prostate cancer. (A) Sequential gating workflow applied to PBMC samples. Singlets were first isolated by FSC-A x SSC-A discrimination, followed by selection of the PBMC compartment. Within PBMCs, lineage-negative (Lin−; CD3− CD19− CD20− CD56−) myeloid cells were separated into CD15+/CD14− (P1) and CD15−/CD14+/dim (not P1) fractions. From the CD15+/HLA-DR−/CD33+ compartment (P2), PMN-MDSCs were identified. The CD15− fraction was further resolved into HLA-DR−/CD33+ M-MDSCs (P3) and the CD14−/Lin−/HLA-DR−/CD11b+ population corresponding to e-MDSCs (P4) [87]. (B) Expression of immunoregulatory checkpoints in e-MDSCs. Representative dot plots show surface staining for CD47, SIRPα, Galectin-9, CD73, and PD-L1, demonstrating differential expression patterns across the early-stage suppressor subset. Gates reflect the threshold of positive staining defined by matched FMO controls. Collectively, these data illustrate a harmonized, reproducible gating strategy that enables high-fidelity discrimination of PMN-MDSC, M-MDSC, and e-MDSC subsets and their associated inhibitory receptor profiles.

Fig 5: UMAP analysis of non-lymphoid, HLA-DR− PBMCs identifies prostate cancer-associated myeloid clusters. UMAP dimensionality reduction was performed on peripheral blood mononuclear cells (PBMCs) after exclusion of lymphoid populations (CD3+, CD19+, CD20+, and CD56+) and antigen-presenting cells (HLA-DR+), thereby enriching for the mononuclear myeloid compartment. The resulting UMAP projection reveals discrete clusters differentially enriched in healthy controls (HC, n = 15) or prostate cancer patients (PC, n = 12). Two PC-enriched clusters (black and red) and one HC-enriched cluster (green) were annotated based on marker-expression profiles. Histogram overlays show increased expression of CD15, CD11b, CD47, PD-L1, CD73, and Galectin-9 within PC-enriched clusters, a phenotype consistent with immunosuppressive, MDSC-like myeloid states and supporting systemic myeloid remodeling in prostate cancer.