Background: Immune evasion is one of the mechanisms by which cancer cells gain the ability to metastasize from the primary tumor to distant sites. In triple-negative breast cancer (TNBC), metastatic tumors are shown to be more immunologically silent than primary tumors. As a result, there are varying degrees of responses to immunotherapy between early-stage and metastatic tumors. Positive clinical responses to immune checkpoint inhibitors (ICIs) are seen in patients with early-stage TNBC regardless of PD-L1 expression. In this study, we investigated the differences in the immune signatures of primary and metastatic breast cancer in a real-world patient population. Methods: We analyzed retrospective comprehensive genomic and immune profiling (CGIP) results from 529 breast tumors tested in the real-world clinical setting. Tumor specimens were classified as primary breast (PB), any lymph nodes (LN; regional and non-regional) or metastatic visceral (MV) sites. LN samples were chosen as positive controls due to expected elevated inflammatory signaling. PD-L1 (CPS positive ≥10) and TMB (high ≥10 Mut/Mb) were determined using IHC and DNA sequencing, respectively. mRNA expression signatures of tumor inflammation (TIGS, strong/moderate/weak) and expression of LAG3, TIGIT and TIM3 were determined by RNA-sequencing from a 395-gene panel. LAG3, TIGIT and TIM3 were selected due to emerging clinical trials in solid tumors. We used over-representation and proportion analysis using chi-squared test to determine the association of specimen sites to various genomic and immune correlates. Results: Among 529 cases, the median patient age was 63.2 years (25.5-93.5). The majority of patients were female (519, 98%), and the most common tumor histology was invasive ductal carcinoma (287, 54%). A total of 224 (42%), 72 (14%), and 232 (44%) of patients had specimens from PB, LN, MV, respectively, while 1 (0.19%) patient had an unknown source. Samples of PB harbored greater degree of immune infiltration, demonstrating higher TIGS score than MV (p=0.014). Primary lesions also demonstrated a greater proportion of PD-L1 positive than metastatic lesions (44% vs 21%, p < 0.001) and higher expressions of immune checkpoints such as TIGIT (p<0.001), LAG3 (p=0.037) and TIM3 (p<0.001). Conclusions: We found that non-lymph node breast cancer metastases harbor a less active immune response than primary breast lesions, showing a lower degree of immune cell infiltration and decreased expression of immune checkpoint markers. These findings support the notion that the immune microenvironment of breast cancer metastases is immunosuppressive and may exhibit a tempered response to ICIs. Therefore, combination treatments of ICIs with chemotherapy, targeted therapies or cancer vaccines may be promising therapeutic approaches to enhance the immune responses and potentially overcome resistance to ICIs in metastatic breast cancer.