Head and neck squamous cell carcinoma (HNSC) is the sixth most common form of malignant tumor occurring in the epithelial tissues of the head and neck regions [1–3]. Despite the advancement in treatment approaches, such as surgical resection and multidisciplinary treatments involving radiotherapy and chemotherapy, the overall 5-year survival rate remains below 70%, particularly for patients in advanced stages of the disease [2–4]. Hence, comprehending the tumor microenvironment of HNSC, identifying therapeutic targets, and formulating novel treatment approaches are imperative. Myeloid-derived suppressor cells (MDSCs) are a heterogeneous population of pathologically activated neutrophils and monocytes that accumulate in the tumor microenvironment and play a critical role in promoting tumor growth and immune evasion [5]. MDSCs have been identified in various types of cancer, including lung cancer, breast cancer, melanoma, and others [6–8]. MDSCs promote tumor growth by suppressing the immune response, particularly T cell activation, and function, which are critical for controlling cancer growth [9, 10]. Immunotherapies that target MDSCs include monoclonal antibodies and small molecule inhibitors. These therapies work by blocking the recruitment of MDSCs to the tumor microenvironment, inhibiting their function, and promoting T cell activation [11–13]. A recent clinical study revealed that the administration of tadalafil, a phosphodiesterase-5 (PDE5) inhibitor, in HNSC patients resulted in a decrease in circulating MDSCs [14, 15]. Moreover, the treated patients exhibited reduced expression of iNOS and arginase in these cells, along with an increased presence of spontaneously generated tumorspecific T cells [14, 15]. Notably, in a co-culture system, HNSC cells had the potential to induce MDSCs differentiation from peripheral blood mononuclear cells and upregulate the expression of iNOS and arginase [16], which further indicated that immunotherapy strategies targeting MDSCs hold great promise. In this study, we utilized a public single-cell RNA sequencing (scRNA-seq) dataset to investigate the heterogeneity of myeloid cells in HNSC. Our analysis revealed that SPP1+ tumorassociated macrophages (TAMs) and MDSCs were the most abundant myeloid cells in the tumor microenvironment. We also discovered that highly infiltrated MDSCs were associated with poorer overall survival rates in HNSC patients. To further explore the potential clinical application of MDSCs as a prognostic marker, we identified a set of six MDSC-related genes that could be used as a prognostic signature. Using this signature, we were able to predict patients’ prognosis outcomes with promising accuracy in both training and test cohorts. By examining intercellular communications, we found that MDSCs were able to suppress the activity of cytotoxic CD8+ T cells and recruit SPP1+ TAMs to shape an immunosuppressive microenvironment that promoted tumor angiogenesis. Overall, our findings suggest that targeting MDSCs may provide a promising therapeutic strategy for the immunotherapy of HNSC.