Abstract Immune checkpoint inhibition (ICI) has produced a paradigm shift in treatment across a range of cancer subtypes. However, our understanding of why only certain patients respond, as well as our ability to predict this response, is still quite limited. This is exacerbated by the complexity of the tumor microenvironment (TME), which contains cancer cells, immune cells, and stromal cells. These differing cell types can have context-dependent roles in either promoting or inhibiting anti-tumor immune responses, making it challenging to identify features which reliably correlate with patient outcome. In addition, the roles of these diverse cell types are often influenced by their location within the TME, meaning that assays without spatial information cannot fully resolve this complexity. To address these gaps, we mapped the spatial distribution and phenotype of 21 cell populations across 117 patients with metastatic triple negative breast cancer that received nivolumab (anti-PD1) through the TONIC clinical trial (NCT02499367). We collected metastatic samples prior to and during treatment to understand how anti-PD1 reshaped the TME. We also collected archival material from the original primary tumor for each patient. We generated multiplexed imaging data from these samples, identified the location and phenotype of each cell, and then quantified the spatial distribution, diversity, and functional marker status of all cell populations in the TME. We detected numerous features indicative of a productive immune response that correlated positively with clinical benefit, whereas features indicative of a suppressive microenvironment correlated negatively with clinical benefit. We observed significant temporal effects, where many pre-treatment predictive features were not predictive on-treatment; similarly, many on-treatment predictive features had no predictive power prior to treatment. Intriguingly, we found features in the original primary tumors which predicted subsequent (metastatic) response to immunotherapy. We observed substantial differences in predictive accuracy of multivariate models based on timepoint, with on-treatment samples exhibiting the best performance and primary tumor samples exhibiting the worst. Of note, we observed substantially worse predictive performance when using bulk DNA sequencing data to predict outcome from the same patients, highlighting the importance of spatial information. Taken together, we show that the features associated with a productive anti-tumor immune response are temporally structured, and differ based on what phase of response is profiled. Furthermore, using the primary tumor to assess whether metastatic TNBC patients are good candidates for ICI may miss important predictive signals. These findings shed new light on the determinants of ICI outcome, and may shape the design of subsequent trials to better understand the temporal aspects of ICI. Citation Format: Noah F. Greenwald, Iris Nederlof, Hugo Horlings, Marleen Kok, Christina Curtis, Michael Angelo. The temporal influence of the tumor microenvironment in response to checkpoint blockade [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3894.