Spatial multiplex analysis of lung cancer reveals that regulatory T cells attenuate KRAS-G12C inhibitor–induced immune responses

Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 1 of 19 IMMUNOLOGY Spatial multiplex analysis of lung cancer reveals that regulatory T cells attenuate KRASG12C inhibitor–induced immune responses Megan Cole1, Panayiotis Anastasiou1, Claudia Lee2, Xiaofei Yu3,4,5, Andrea de Castro1, Jannes Roelink3, Chris Moore1, Edurne Mugarza1, Martin Jones6, Karishma Valand1, Sareena Rana1, Emma Colliver2, Mihaela Angelova2, Katey S. S. Enfield2, Alastair Magness2, Asher Mullokandov1, Gavin Kelly7, Tanja D. de Gruijl4,5,8, Miriam MolinaArcas1, Charles Swanton2,9, Julian Downward1*†, Febe van Maldegem1,3,4,5*† Kirsten rat sarcoma virus (KRAS)–G12C inhibition causes remodeling of the lung tumor immune microenvironment and synergistic responses to anti–PD1 treatment, but only in T cell infiltrated tumors. To investigate mechanisms that restrain combination immunotherapy sensitivity in immuneexcluded tumors, we used imaging mass cytometry to explore cellular distribution in an immuneevasive KRAS mutant lung cancer model. Cellular spatial pattern characterization revealed a community where CD4+ and CD8+ T cells and dendritic cells were gathered, suggesting localized T cell activation. KRASG12C inhibition led to increased PD1 expression, proliferation, and cytotoxicity of CD8+ T cells, and CXCL9 expression by dendritic cells, indicating an effector response. However, suppressive regulatory T cells (Tregs) were also found in frequent contact with effector T cells within this community. Lung adenocarcinoma clinical samples showed similar communities. Depleting Tregs led to enhanced tumor control in combination with anti–PD1 and KRASG12C inhibitor. Combining Treg depletion with KRAS inhibition shows therapeutic potential for increasing antitumoral immune responses. INTRODUCTION Recent years have seen a transformation in the treatment of non– small cell lung cancer (NSCLC), with the introduction of immune checkpoint blockade, which has increased survival rates of patients with a previously poor prognosis. Despite this, only a subset of patients responds, and many responders acquire resistance over time (1,2). In 2021, a further breakthrough occurred when the Kirsten rat sarcoma virus (KRAS) inhibitor sotorasib was approved for the treatment of locally advanced or metastatic KRASG12C mutant NSCLC. This followed successful clinical trials where 80% of patients achieved temporary disease control following sotorasib treatment. However, despite a modest improvement in progressionfree survival, sotorasib gave no improvement in overall survival compared to docetaxel (3), demonstrating its limitations for use as a monotherapy. Therefore, strategies for combination with other therapies are being urgently sought (4,5). The importance of the immune system in the response to KRASG12C inhibition was revealed when Canon etal. (6) showed that T cell presence was essential for durable responses in subcutaneous tumors of the colon cancer model CT26 treated with sotorasib. In addition, Briere et al. (7) using the same model demonstrated a switch in the tumor microenvironment (TME) from immunosuppressive in the vehicle setting, with high presence of M2 macrophages and myeloidderived suppressor cells, to favoring antitumoral immune response following KRASG12C inhibition with adagrasib, another clinically approved KRAStargeted drug. CT26 tumors show an immune hot TME (i.e., high T cell infiltration rates) and are responsive to singleagent immune checkpoint inhibition (ICI). Similar to the proinflammatory responses observed in this colon cancer model (6,7), our previous work also revealed remodeling of the TME following KRASG12C inhibition in multiple lung tumor models, including the immune cold orthotopic Lewis lung murine NSCLC tumor model that was genetically engineered to disrupt the NRAS gene, avoiding redundancy in RAS signaling (3LL ΔNRAS, here referred to as 3LL in short) (8,9). We found that oncogenic KRAS suppresses interferon signaling within the tumor cells via MYC, leading to a proinflammatory cascade upon KRAS inhibition (9). KRASG12C inhibitors were also shown to synergize well with ICI in lung cancer models, but only in immune hot TME settings (8,10). These findings highlight that while KRASG12C inhibitors specifically target tumor cells, this results in profound secondary effects on the TME and T cells are crucial for durable responses. Our previous analysis established that following seven consecutive days of treatment with the KRASG12C inhibitor MRTX1257, although 3LL tumor growth was inhibited, tumors did not regress, indicating that KRASG12C inhibitors alone are not sufficient to cause tumor regression in this model (9). Combination of this KRASG12C inhibitor with anti–programmed cell death protein 1 (PD1) therapy did not increase responses in this model (8). This is reflective of the failure to see a beneficial effect on response of combined KRAS inhibition and PD1 blockade in the clinical setting, either due to combination toxicities or lack of efficacy (11). The unmet need 1Oncogene Biology laboratory, Francis Crick institute, london, UK. 2Cancer evolution and Genome instability laboratory, Francis Crick institute, london, UK. 3department of Molecular Cell Biology and immunology, Amsterdam UMC, vrije Universiteit Amsterdam, Amsterdam, Netherlands. 4Cancer Center Amsterdam, Cancer Biology and immunology, Amsterdam, Netherlands. 5Amsterdam institute for immunology and infectious diseases, Amsterdam, Netherlands. 6electron Microscopy, Francis Crick institute, london, UK. 7Bioinformatics and Biostatistics, Francis Crick institute, london, UK. 8department of Medical Oncology, Amsterdam UMC, vrije Universiteit Amsterdam, Amsterdam, Netherlands. 9Cancer Research UK lung Cancer Centre of excellence, UCl Cancer institute, london, UK. *Corresponding author. email: julian. downward@ crick. ac. uk (J.d.); f. vanmaldegem@ amsterdamumc. nl (F.v.M.) †these authors contributed equally to this work. Copyright © 2024 the Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. distributed under a Creative Commons Attribution license 4.0 (CC BY). Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 2 of 19 for new combinatorial treatment options that improve T cell–mediated antitumoral immune responses in immune cold tumor models has hence become increasingly clear. We therefore used the 3LL immuneevasive orthotopic lung tumor model, in which effector immune cells are excluded from the tumor, to seek more effective therapeutic combinations with KRASG12C inhibition. We used imaging mass cytometry (IMC) to analyze the makeup of these tumors in situ. This method is particularly useful for study of the TME due to its ability to capture up to 40 markers simultaneously. Spatial information remains intact, meaning that cell phenotypes can be analyzed in the context of their spatial neighbors (12). Obtaining spatial information is essential when studying the TME, as it provides insight into the cellular interactions dictating local immune activation or suppression, as mechanisms of immune response or resistance to treatment, and has been shown to improve prediction of clinical outcomes compared to cell frequency alone in patients with NSCLC (13). Here, we present data on the identification of neighborhood communities through singlecell spatial analysis to investigate which cellular interaction patterns may restrain antitumoral immune responses in the 3LL tumor model. A community resembling a T cell activation hub was identified, where regulatory T cell (Treg) interactions may play a key role in dampening antitumoral immune responses following KRASG12C inhibition. In parallel, cellular community analysis of treatmentnaïve human lung adenocarcinoma (LUAD) patient samples from the TRAcking Cancer Evolution through therapy (Rx) (TRACERx) IMC cohort (14) suggested that similar local Treg control may be restraining immune responses in a subset of patients. Together, this led us to explore the effect of combining KRASG12C inhibitor with a Treg depleting anti–cytotoxic T lymphocyte– associated protein 4 (CTLA4) antibody in the in vivo orthotopic setting, where markedly improved responses were noted. This opens up the perspective of combining KRASG12C inhibitors with Treg targeting to improve durable response rates. RESULTS Introduction to spatial communities and validation Our previous IMC analyses on 3LL lung tumors have shown that there are clear patterns in the arrangements of cells in the lung cancer tissues and changes to those patterns occur in response to KRASG12C inhibition (9). For example, we saw one subset of macrophages lining the tumornormal interface, while another type of macrophages was intermixed with the tumor cells. Effector cells such as T cells and B cells were excluded from the tumor domain, while treatment with KRASG12C inhibitor MRTX1257 induced movement of T cells and antigenpresenting cells into the tumor domain. We also noted that the phenotypes of cells differed depending on their location in the tissue and hypothesized that the local neighborhood is likely to strongly influence the activation or inhibition of immune cells. Therefore, we adopted a cellular community analysis to cluster cells based on the composition of their local neighborhood [adapted with modifications from (15)]. Analysis focused on two previously published datasets (8,9), both generated from an in vivo experiment in which the Lewis lung (3LL) carcinoma model was treated with the KRASG12C inhibitor MRTX1257 or vehicle for 7 days before harvesting of the lungs. The tumors were stained with two partially overlapping antibody panels to give two datasets, with the dataset 2 panel being more T cell oriented (fig. S1, A and B). The cell typing was derived from our previous analyses and based on lineage markers only but independent of maturation or activation markers. Notably, the subsequent cellular communities were therefore blind to cell phenotypes. We identified neighbors by a 15pixel expansion of the cell boundary through segmentation in CellProfiler. A 15pixel radius was chosen as it depicts the average size of a cell, and, therefore, cells identified as neighbors would be those “up to one cell away.” As a result, neighborhoods were not equal in cell number but, instead, reflected the local density surrounding each cell. Louvain clustering using Rphenograph was then run on the neighbor proportion information per cell to identify recurring spatial patterns in the tissue, labeled “spatial communities” or just “communities” in short (Fig. 1A). Graph building with a knearest neighbor input value of 250 yielded 62 communities for dataset 1; these were agglomerated to 30 and subsequently 18 communities. Agglomeration merged the communities with the lowest cell diversity, representing mainly small variations in the tumor cell neighbors (e.g., agglomerated community 3), while the highly diverse communities, such as those with a high proportion of immune cells, remained stable (Fig. 1B and fig. S1C). We were most interested in the immunedominated communities for this analysis and therefore decided that 18 communities were optimal to carry forward for our investigation into antitumoral immune response. To determine whether our method to identify spatial communities was robust to altered clustering input conditions, we also ran Rphenograph clustering on the neighborhood information for all cells in dataset 1 with a kinput value of 350. Dimensionality reduction using tdistributed stochastic neighbor embedding (tSNE) of the 62 communities identified from kinput value of 250 and the 47 communities identified from a kinput value of 350 revealed very similar patterns of community phenotypes (Fig. 1C). In addition, there were multiple overlaps between communities identified using the different k values for clustering, suggesting matching phenotypes. The 254 communities identified from neighbor clustering of dataset 2 using a kinput of 250 were also agglomerated to 18 communities to enable parallel analysis with the communities identified in dataset 1. The communities in both datasets varied largely in size, with some containing fewer than 2000 cells and others comprising over 50,000 cells (Fig. 1, D and E). There was also differing heterogeneity of these spatial groups, with some being dominated by a single cell type, such as tumor cells in community 3 from dataset 1 and community 9 from dataset 2, while others comprised a mixture of various cell types in more balanced proportions (fig. S1, D and E). Datasets 1 and 2 were based on different antibody panels and therefore not identifying all the same cell types. For example, lack of markers epithelial cellular adhesion molecule and platelet endothelial cell adhesion molecule1 for dataset 2 meant that endothelial and epithelial cells could not be identified and, thus, a large number of cells were labeled as “unclassified.” Nevertheless, a separate community clustering analysis on both datasets based on shared cell types only demonstrated that the method to generate communities was stable to altered input data (fig. S1F). Tissue architecture reflected in spatial communities As the spatial communities are based on local neighborhoods but these local neighborhoods are likely to be different within different regions of the tissue, we further explored the link with tissue Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 3 of 19 A D B 17 18 E C B IMC Tissue slice Neighbor clustering Neighborhood identification Neighborhood Community Spatial communities 1 2 3 5 6 7 8 9 10 11 12 13 14 15 16 Size 10,000 20,000 30,000 40,000 50,000 Communities Segmentation and cell typing Fig. 1. Clustering of cells based on their neighbors yield spatial communities. (A) Workflow of generating spatial communities using the data generated from lung tissue slices. (B) Cluster tree of 62 communities generated using Rphenograph with a kinput value of 250, agglomerated to 30 communities and subsequently agglomerated again to 18 communities, where each circle represents a community and lines indicate communities that were merged during agglomeration. (C) tSNe plot of 62 communities generated with a kinput value of 250 and 47 communities generated with a kinput value of 350 into Rphenograph using dataset 1, where tSNe analysis was run on the basis of the proportion of each cell type contributing to each community. (D and E) eighteen spatial communities generated from clustering based on neighbor proportions of each cell type for (d) dataset 1 and (e) dataset 2, with the size of each bar representing cell count of that community and colors indicating the contribution of each cell type. Bars are ordered by decreasing tumor cell count. NK, natural killer; dN, double negative. Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 4 of 19 architecture. For dataset 1, we also had information about the three tissue domains that each cell had been assigned to during image segmentation, i.e., tumor, normal, and interface (9). As the communities had a nonrandom distribution across the domains, we visualized the distribution of each community relative to the cross section through the tissue to further expand on this spatial organization of the communities (Fig. 2A). For example, in the vehicle setting, community 18, with high endothelium and B cell portion, was restricted to the normal nontumor region as its cell count diminished going into the tumor bulk (fig. S2A). In addition, community 10, with high type 1 macrophage contribution, peaked in cell count at the tumor boundary, demonstrating a clear interface region between normal tissue and the tumor bulk (fig. S2B). There were also further communities, numbers 2 and 3, that were found only in the tumor region, albeit at very different frequencies between vehicle and MRTX1257 conditions (fig. S2, C and D). We previously observed that following treatment with MRTX1257, the immuneexcluded phenotype of this tumor model was remodeled into a more inflammatory immune–infiltrated TME (9). Here, we could see that this conversion was also reflected in spatial distribution of the communities, suggesting that tissue domain definition was lost following KRASG12C inhibition. Not only did the concentration of communities such as 10 and 18 to the interface and normal regions become less pronounced, but also spatial patterns within the tumor bulk changed following KRASG12C inhibition, as the frequency of many communities, such as 2 (type 2 macrophage dominant), 3 (tumor dominant), and 16 (mixed phenotype with high type 2 macrophage portion) was altered, suggesting a transition to a new organization of the TME (Fig. 2B and fig. S2E). Comparing the relative contribution of each community between the vehicle and MRTX1257 treatment groups revealed that prevalence of many spatial patterns was altered following KRASG12C inhibition. Some of these shifts in communities captured changes that we previously described at the singlecell level (9). For example, communities 5, 13, 14, and 15 from dataset 1 were found solely in tumors treated with vehicle. These communities represent abundant interactions between tumor cells and neutrophils, which were lost or dispersed following treatment with MRTX1257 (Fig. 2C and fig. S1D). This is in line with a previous observation that the number of neutrophils within the tumor domain was substantially reduced following treatment (9). Alternatively, communities 2, 4, and 16 from dataset 1 and communities 2, 6, and 8 from dataset 2 were almost exclusively found in tumors following treatment with MRTX1257. A shared feature for these communities was a neighborhood involving high numbers of type 2 macrophages (F4/80+ CD206+), which we previously showed to become greatly increased in number upon AC Cross section through tissue Vehicle MRTX1257 Percentage distribution Dataset 1 Dataset 2 Cell count B Cell count Cross section through tissue DE Community BRAC4002.3c_ROI2_t2 BRAC3495.3f_ROI1_t1 BRAC3326.4e_ROI1 BRAC3438.6f_ROI1 BRAC3438.6f_ROI3 BRAC3495.3f_ROI1_t2 BRAC3529.2d_ROI1 BRAC3438.6f_ROI2 BRAC3529.2b_ROI1 BRAC4002.3c_ROI1 BRAC4002.3c_ROI2_t1 BRAC4002.3c_ROI3 3 18 2 10 16 11 17 8 9 1 12 6 7 5 13 14 15 4 T reatment Heatmap −2 0 2 4 Treatment MRTX Vehicle Fig. 2. Spatial communities reflect refined tissue architecture and have functional relevance. (A and B) Cell count per community relative to cross section through the tissue, where 0 represents the center point of the tumor in (A) vehicle and (B) MRtX1257 treatment settings. (C) Percentage distribution of each community across vehicle (left) and MRtX1257 (right) treatment groups. Bars are ordered by increasing percentage distribution in vehicle setting. (D) hierarchical clustering of community proportion per ROi for dataset 1, with the use of dendrograms to show relationships between similar ROis, similar communities, and community distribution across the treatment groups. (E) Pearson correlation calculation on the proportion of each cell type pair within each community. *P<0.05, **P<0.01, and ***P<0.001. Cell types were clustered on the basis of correlation value. MRtX, MRtX1257. Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 5 of 19 KRASG12C inhibition (9). This agreement with previous observations suggests that the communities are reflecting relevant biological processes. Further supporting this notion was the ability to largely separate tissues into the two treatment groups when clustering the frequency of each spatial community per region of interest (ROI) (Fig. 2D and fig. S2F). One way to infer potential cellular relationships is by correlating cell frequencies, measuring cooccurrences. Therefore, we calculated the correlation of each cell type pair within communities. Strong positive correlations were seen between certain cell types, such as T cells among themselves, T cells with dendritic cells (DCs), type 2 macrophages with fibroblasts, and CD8+ T cells with B cells when calculated per community (Fig. 2E and fig. S2G). By contrast, a number of these, such as Tregs with CD4+and CD8+ T cells, and most T cell–DC relationships showed lack of significance when quantified in the ROIs (fig. S2, H and I). This demonstrates the benefit of studying cellular relationships through the identification of localized spatial patterns, as they provide increased statistical power about the interactions occurring in the TME in comparison to measuring interaction of cells across a whole tissue. Spatial communities that are abundant in CD8+ T cells Because the abundance of CD8+ T cells is associated with positive outcomes in relation to antitumoral immune response, we decided to investigate the T cell–rich communities from this tumor model. Previous analysis revealed increased numbers of CD8+ T cells inside the tumor domain following KRASG12C inhibition in this model (9). The top five communities with the highest CD8+ T cell count were identified from dataset 1 and dataset 2 (fig. S3, A and B). These five CD8+ T cell–rich communities from both datasets paired up phenotypically, particularly when only shared cell types between datasets were considered (fig. S3C). Therefore, unique colors and the names T cell/normal adjacent community (T/NA), T cell/type 1 macrophage community (T/M1), T cell/DC community (T/DC), T cell/ type 2 macrophage community 1 (T/M2_1), and T cell/type 2 macrophage community 2 (T/M2_2) were assigned to each pair for parallel analysis going forward (Fig. 3A). These five communities contained over 75% of the total CD8+ T cell population from both datasets, suggesting a representative population of the overall cohort (fig. S3D). These communities had widely different compositions, placing the T cells in very diverging spatial contexts (Fig. 3A). The T/NA Vehicle C AB Vehicle MRTX1257 Dataset 1 Dataset 2 Dataset 1 Dataset 2 T/NA T/M1 T/DC T/M2_1 T/M2_2 Cross section through tissue Cell count Dataset 1 Dataset 2 MRTX Percentage distribution T/NA T/M1 Dataset 1 Name Color 18 Community Dataset 2 11 10 10 11 14 16 1 22 T/M1 T/DC T/M2_1 T/M2_2 Fig. 3. Spatial communities that are abundant in CD8+ T cells are diverse in composition and spatial distribution. (A) Percentage distribution of cell types contributing to the top five communities with the highest Cd8+ t cell count from dataset 1 (left) and dataset 2 (right), ordered in pairs based on proportions of shared cell types and labeled with the names t/NA, t/M1, t/dC, t/M2_1, and t/M2_2. (B) Cell count of each of the top five communities relative to the cross section through the tissue, where 0 represents the center point of the tumor. (C) visualization of cell outlines from cells assigned to t/NA, t/M1, t/dC, t/M2_1, and t/M2_2 communities, with each outline filled with a color, associating it to one of the five communities, in vehicleand MRtX1257treated tumors for datasets 1 and 2. Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 6 of 19 community was characterized by high endothelial cell content; in the T/M1 community, the type 1 macrophages (CD11c+ CD68+) (9) were the most abundant cell type; the T/DC community was strongly enriched in various T cell subsets and DCs; and T/M2_1 and T/M2_2 communities were dominated by tumor cells and type 2 macrophages (F4/80+ CD206+) (9) in different ratios. These CD8+ T cell–rich communities also differed in presence between treatment groups. The T/NA community was more frequently found following treatment with vehicle, while the T/M2_2 community was almost exclusively detected in MRTX1257treated tissues (fig. S3E). Following this, most of CD8+ T cells in the vehicletreated tumors were found within T/NA and T/M1 communities, while CD8+ T cells in the MRTX1257 treatment group were more likely to reside within T/DC, T/M2_1, and T/M2_2 communities (fig. S3F). Furthermore, the spatial location of the top five communities also varied largely in relation to three assigned tissue domains: normal, interface, and tumor (Fig. 3, B and C). In particular, in the vehicle setting, the T/NA community was situated predominantly in the “normal” nontumor region, and the T/M1 community was restricted to the “interface,” situated as a clear ring around the tumor bulk (fig. S3G). The T/DC community was located just inside the tumor bulk in the vehicle setting but increased in size, and its position moved toward the tumor core following treatment with MRTX1257. The T/M2_1 and T/M2_2 communities were concentrated within the tumor domain and predominantly found within the MRTX1257treated tumors. Evidently, treatment with MRTX1257 led to a shift in spatial distribution and neighborhood environment of the CD8+ T cells. Responses of T cell–rich communities to KRASG12C inhibition Communities were defined on the basis of cell types using lineage markers independently of maturation or activation markers. We next sought to explore how cells within the communities responded to KRASG12C inhibition based on the maturation and activation markers included in both datasets. Previously, we described how the two subsets of macrophages in this tumor model differed in spatial location and response to treatment with MRTX1257 (9). The most notable observation was that the type 2 macrophages increased in cell number and upregulated activation markers such as programmed cell death ligand 1 (PDL1) and major histocompatibility complex II. Zooming in on the communities here revealed that the upregulated expression could be largely attributed to the type 2 macrophages in the T/DC community (fig. S4, A and B). Similarly, we looked at expression of PDL1 and costimulatory receptor CD86 on DCs. Differences were more noticeable across the communities than between treatment groups, suggesting that DC phenotypes were more influenced by surrounding neighbors than treatment with MRTX1257 (Fig. 4, A and B, and fig. S4C). The behavior of expression varied slightly across these markers, with T/M1 community harboring high PDL1 but low CD86 expression in the vehicle setting but less so under inhibitortreated conditions. This high PDL1 and low CD86 expression is thought to be a characteristic of regulatory or migratory tolerogenic DCs, with a potential to inhibit immune responses (16). Alternatively, in T/DC, T/M2_1, and T/M2_2 communities, increased PDL1 expression on DCs in the MRTX1257 treatment group was accompanied by high CD86 expression, indicative of a more activated DC phenotype, making these cells better equipped to facilitate T cell activation. This differential in activation status was further supported by the expression levels of major histocompatibility complex II, being the highest in the T/DC community (fig. S4D). We previously reported increased expression of T cell chemoattractants chemokine (CXC motif) ligand 9 (CXCL9) and CXCL10 in KRASG12C inhibitor–treated tumors (8). MRTX1257 treatment increased the expression of CXCL9, with the highest overall expression in the T/DC community, suggesting that most of the T cell attraction was occurring within this environment (Fig. 4C). This agrees with our finding that the T/DC community contains the largest T cell density. We also compared proliferative marker Ki67 and apoptotic marker cleaved caspase3 (ccasp3) expression on tumor cells in the tumorassociated communities (T/DC, T/M2_1, and T/M2_2) and found the highest expression of both markers in the T/DC community (Fig. 4, D and E, and fig. S4, E and F). This demonstrates the high level of activity occurring within this community, where some tumor cells were thriving, while others were dying. We expected that these different neighborhoods would also have an impact on the phenotype of the T cells, so we investigated their PD1 expression to understand how T cell activation state changed following KRASG12C inhibition, per community. In the T/NA and T/M1 communities, the PD1 expression was negligible in both treatment groups, compared to the T/DC, T/M2_1, and T/M2_2 communities, where an increase occurred following treatment with MRTX1257, most pronounced in the T/DC community, indicating a switch in cell state from naïve to activated in these communities following KRASG12C inhibition (Fig. 4F and fig. S4G). A similar pattern could be seen for CD4+ T cells, whereas the Tregs had higher PD1 expression mainly in the T/DC community following MRTX1257 treatment (fig. S4, H and I). However, the expression of lymphocyteactivation gene 3 (LAG3) protein was also distinctly higher in CD8+ T cells, specifically in the tumorassociated communities, in which ~30% of PD1+ T cells were also LAG3+ following treatment with MRTX1257. This suggests that activation of the T cells was, in part, accompanied with induction of T cell exhaustion following KRASG12C inhibition (fig. S4, J and K). Positive and negative regulations of antitumoral immune responses While the T/DC, T/M2_1, and T/M2_2 communities contained most of the CD8+ T cells in the tumor tissue, these cells expressed significant levels of potential exhaustion markers such as PD1 and LAG3. Colocalization of PDL1–expressing macrophages and PD1+ CD8+ T cells has recently been highlighted as associated with good response to ICI (17,18). However, previous attempts to reinvigorate these T cells using MRTX1257 in combination with ICIs anti–PD1 or anti– PDL1 and antiLAG3 had failed to achieve any improved tumor control in our 3LL model (8). We therefore wondered whether we could gain any insight into the signals provided to the T cells before moving into the core of the tumor and becoming incapacitated by exhaustion. The high proportion of activated DCs and increased expression of markers associated with T cell attraction and activation following KRASG12C inhibition, as well as evidence for local tumor cell death, pointed toward the T/DC community as a potential cytotoxic T cell activation hub. As noted, CXCL9 expression was the highest within antigenpresenting cells in the T/DC community (Fig. 4C). Separating DCs based on their CXCL9 expression into “CXCL9low” and “CXCL9high” Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 7 of 19 groups showed that CXCL9high DCs had a significantly shorter distance to their nearest PD1+ CD8+ T cell, PD1+ CD4+ T cells, and PD1+ Tregs (Fig. 5A and fig. S5, A and B). Visual inspection indeed confirmed that CXCL9high DCs were frequently found in proximity to and interacting with PD1+ CD8+ T cells (Fig. 5B). CXCL9 expression could therefore be one of the key mediators driving the aggregation of the T cells and DCs within the T/DC community, as part of a mechanism to draw in the activated T cells to launch an antitumor immune response. This would be in agreement with our previous work showing that CXCL9 expression was one of the strongest predictors of ICI (19). We sought to determine further indications of an active antitumoral immune response within this community. A significant increase in Ki67 expression was identified for CD4+ and CD8+ T cells on MRTX1257 treatment, as well as a similar trend for Tregs, suggesting that the T cells had increased proliferation following KRASG12C inhibition (Fig. 5C). Frequency of casp3+ tumor cells was the highest in the T/DC community, and the occurrences in which a ccasp3+ tumor cell was found in the 15pixel neighborhood of a CD8+ T cell increased following treatment with MRTX1257 (Fig. 5D). These interactions were primarily found within the T/DC community (fig. S5C). There was no such increase in spatial interactions identified for CD4+ or Tregs with ccasp3+ tumor cells (fig. S5D). These increased interactions point to increased cytotoxicity of the T cells against the tumor cells following KRASG12C inhibition, suggesting that the neighborhood of the T/DC community was likely to be able to support the effector function of the CD8+ T cells. We then wondered why such a cytotoxic response was not effective enough to mediate clinical benefit and what could be driving the induction of T cell exhaustion or dysfunction. To identify potential negative regulatory influences of the immune response, we used another way of interrogating spatial relationship by calculating enrichment scores for cells in the near neighborhood, compared to randomized data (9,20). Following MRTX1257 treatment, it was determined that CD4+ T cells and DCs were significantly enriched in the neighborhood of a CD8+ T cell within the T/DC community in at least five of the six images, compared to random permutation (Fig. 5E). This is supportive of a microenvironment promoting AB DEF Dataset 1 Dataset 1 Dataset 1 Dataset 1 Dataset 1 Dataset 2 C Fig. 4. T cell–rich communities respond differently to KRASG12C inhibition. (A and B) Mean expression of (A) Pdl1 and (B) Cd86 on dCs and Cd103+ dCs in t/NA, t/M1, t/dC, t/M2_1, and t/M2_2 communities following vehicle and MRtX1257 treatments for dataset 1 only. values were log2 scaled. (C) Mean expression of CXCl9 on dCs, Cd103+ dCs, macrophages type 1, and macrophages type 2 combined for t/NA, t/M1, t/dC, t/M2_1, and t/M2_2 communities in vehicleand MRtX1257treated groups in dataset 2. values were log2 scaled. Center line shows median expression for each treatment group. (D and E) Mean expression of (d) Ki67 and (e) ccasp3 on tumor cells in t/dC, t/M2_1, and t/M2_2 communities following treatment with MRtX1257 for dataset 1. values were log2 scaled. (F) Mean expression of Pd1 on Cd8+ t cells in communities t/NA, t/M1, t/dC, t/M2_1, and t/M2_2 in vehicleand MRtX1257treated groups for dataset 1. values were log2 scaled. Center line shows median expression for each treatment group. MRtX, MRtX1257. Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 8 of 19 B D G 12345678910 11 12 13 14 15 16 1817 F Dendritic cells cDC1 T cells CD4 T cells CD8 T regs Other P value ≤0.01 >0.01 Relative count of c-casp3 + tumor cell and CD8 + T cell interaction Proportion of regulatory T cells per community (T/M2_2) (T/M2_2) (T/M1) (T/M1) (T/DC) (T/DC) (T/M2_1) (T/M2_1) (T/NA) (T/NA) A E cDC1 CXCL9 high Dendritic cell CXCL9 high T cell CD8 PD-1 high Other C Fig. 5. Positive and negative regulations of antitumoral immune responses come together in the T/DC community. (A) Minimum distance of dCs and Cd103+ dCs that have “low” or “high” CXCl9 expression (threshold=0.5) to Pd1+ Cd8+ t cells within 800 pixels in t/dC community from dataset 2. distance values were log2 scaled. ***P < 0.001. (B) visualization of cell outlines for cells assigned to t/dC community in MRtX1257treated tissues from dataset 2, with CXCl9high dCs, Cd103+ dCs, and Pd1+ Cd8+ t cells colored in to show spatial proximity of these cell phenotypes. Some regions were expanded for easier visualization. (C) Mean expression of Ki67 on Cd4+ and Cd8+ t cells and tregs within the t/dC community for vehicle and MRtX1257 treatment groups from dataset 2. values were log2 scaled. (D) the number of times a ccasp3+ tumor cell is found in the 15pixel neighborhood of a Cd8+ t cell within the t/dC community, compared across vehicle and MRtX1257 treatment groups for dataset 2, averaged per ROi. Count is relative to the proportion of tumor cells that were ccasp3+ in vehicle versus MRtX1257 treatment groups. each dot represents the value of one ROi. (E) log2 fold changes (log2FC) in enrichment from neighbouRhood analysis for Cd8+ t cells in the t/dC community following treatment with MRtX1257. Filled circles represent images from which enrichment was statistically significant compared to randomized spatial arrangements following treatment with MRtX1257 for dataset 2. (F) Scaled proportion of tregs contributing to each of the 18 original communities for dataset 2. (G) visualization of cell outlines for cells assigned to the t/dC community in MRtX1257treated tissues for dataset 2. dCs, Cd103+ dCs, Cd4+ and Cd8+ t cells, and tregs were filled in to illustrate spatial proximity of these cell types. Some regions were expanded for easier visualization. MRtX, MRtX1257. Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 9 of 19 antitumoral immune response, which was not apparent from T/NA, T/M1, T/M2_1, and T/M2_2 communities (fig. S5, E to H). However, Tregs were also significantly enriched in the neighborhood of CD8+ T cells in three of the six images, indicating the presence of immune suppressor cells in the vicinity where T cell activation and effector functions may be taking place (Fig. 5E). Treg frequencies were the highest within the T/DC and T/M2_2 communities (Fig. 5F and fig. S5I). This T cell subset also increased in size, most notably in the T/DC and T/M2_2 communities, and, as we saw previously, was showing evidence of increased activation after MRTX1257 treatment (figs. S4I and S5J). Tregs were seen intermixed with effector T cells and DCs, suggesting a potential role in locally dampening antitumoral immune responses (Fig. 5G). Role for Tregs in dampening antitumoral immune responses Upon revealing the high presence of Tregs within the T/DC community and showing that they are enriched within the neighborhood of CD8+ T cells and neighboring DCs and CD4+ T cells following treatment with MRTX1257, we decided to explore their role within this community in relation to antitumoral immune response. We therefore split up the T/DC community into two neighborhoods: those with presence of Tregs or those with absence of Tregs. The neighborhoods differed slightly in their composition of cell types, with the Treg neighborhood comprising a higher proportion of DCs and CD4+ T cells, whereas the no Treg neighborhood contained a higher tumor and type 2 macrophage portion (Fig. 6A). While the frequency of CD8+ T cells within both the Treg and no Treg neighborhoods were similar (fig. S6A), the cellular interactions for CD8+ T cells differed substantially between these environments. Despite the slightly higher DC frequency in the Treg neighborhoods, we saw a lack of spatial enrichment between CD8+ T cells and both DC subsets when Tregs were present, in contrast to a positive spatial enrichment of DCs in the CD8+ T cell neighborhood when Tregs were absent (Fig. 6B and fig. S6B). In addition, when Tregs were not present in the CD8+ T cell neighborhood, a strong enrichment of tumor cells was identified following MRTX1257 treatment, compared to slight depletion when the Tregs were nearby. Furthermore, the frequency of ccasp3+ tumor cells in the CD8+ T cell close neighborhood increased on MRTX1257 treatment when Tregs were absent (Fig. 6C). We therefore only saw interactions with CD8+ T cells indicative of an active antitumoral immune response when the Tregs were absent, suggesting an inhibitory role for the Tregs. These changes in neighborhood enrichment were not observed when comparing CD4+ T cells with or without Tregs in their neighborhood, indicating that Treg presence may affect CD8+, but not CD4+, T cell relationships (fig. S6, C and D). However, there were several other changes to the cellular interactions when subsetting the T/DC community based on presence or absence of Tregs, suggesting that Tregs may be affecting the local milieu of this community in many ways, further pointing toward their potential negative influence on antitumoral immune response (fig. S6B). Overall, these analyses indicated that the T/DC community potentially could provide an activating environment for P value ≤0.01 >0.01 A - with Tregs - without Tregs C P value ≤0.01 >0.01 B No. of interactions between c-casp3 + tumor cells and CD8 + T cells Fig. 6. Tregs dampen local antitumoral immune responses. (A) Percentage distribution of cell types contributing to neighborhoods with tregs (“tregs”) and neighborhoods without tregs (“No tregs”) within the t/dC community following treatment with MRtX1257. (B) log2 fold changes in enrichment from neighbouRhood analysis for Cd8+ t cells in tregs (top) and no tregs (bottom) neighborhoods within the t/dC community following treatment with MRtX1257. Filled circles represent images from which enrichment value was statistically significant compared to randomization of the spatial arrangements within the t/dC community following treatment with MRtX1257 for dataset 2. (C) Number of times a ccasp3+ tumor cell is found in the 15pixel neighborhood of a Cd8+ t cell within the t/dC community, compared across treg and no treg neighborhoods in dataset 2, averaged per ROi. Count is relative to the proportion of tumor cells that were ccasp3+ in treg versus no treg groups. Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 16 of 19 Briefly, the method is as follows: A window was defined around every cell in an image and its 10 nearest neighboring cells including the center cell. These windows were clustered by their composition with respect to the 18 cell types in the panimmune panel and the 20 cell types in the T cells and stroma panel (with at least 10 cells on average per image) using MiniBatchKMeans. We optimized the parameters of the method by Schurch etal. (15) and identified 10 spatial cellular communities from the panimmune panel and 30 spatial cellular communities using the T cells and stroma panel. Communities were then assigned representative names based on the enrichment of cell densities within them. Spatial community identities were mapped onto segmented cells and visualized using Cytomapper (71), which were then validated by a pathologist’s assessment of serial hematoxylin and eosin–stained tissue sections. The cell density of spatial cellular communities was calculated by taking the number of cells assigned to a spatial cellular community divided by the total tissue area (in cells per square millimeter). Association between cell densities of cell subtypes (T cells and stroma panel) and spatial cellular communities (panimmune panel)—TRACERx data We correlated the cell density of stromalocalized T cell subtypes detected using the T cells and stroma antibody panel and the cell density of 10 spatial communities detected using the panimmune antibody panel in cores from the same tumor region. For comparisons with multiple tumor cores/regions per tumor, we used linear mixedeffects model analysis to incorporate patient ID as a random effect. We report the T score and P value of the model. Cooccurrence of spatial cellular communities—TRACERx data The lowest quartile of the cell densities of Treg communities was approximately 25 cells/mm2. Therefore, we used 25 cells/mm2 as the threshold to determine whether a spatial community was present or absent in a tumor core. We reported the proportion of paired tumor cores (T cells and stroma panel and panimmune panel) with at least 25 cells/mm2 in p2_C1: tumor border communities and in p1 Treg communities. In vivo survival experiment 3LLΔNRAS (37) was cultured in RPMI 1640 supplemented with 10% fetal bovine serum, 4 mMlglutamine (SigmaAldrich), penicillin (100 U/ml), and streptomycin (100 mg/ml; SigmaAldrich). Cell lines were tested for mycoplasma and authenticated by shorttandem repeat DNA profiling by the Francis Crick Institute Cell Services Facility. Cells were allowed to grow for not more than 20 subculture passages. Intravenous tail vein injections of 106 3LLΔNRAS cells were carried out for orthotopic studies using 8to 12weekold male C57BL/6J mice. Mice were euthanized, with an overdose of pentobarbitone, when a humane end point of 15% weight loss was reached or any sign of distress was observed (i.e., hunched, piloerection, and difficulty of breathing). In addition, if a mouse was observed to have a tumor burden in excess of 70% of lung volume when assessed by μCT scanning, they were deemed at risk of rapid deterioration in health and euthanized immediately. Mice were anesthetized by isoflurane inhalation and scanned using the Quantum GX2 μCT imaging system (PerkinElmer) at a 50μm isotropic pixel size. Serial lung images were reconstructed and analyzed using Analyze12 (AnalyzeDirect) as previously described in Zaw etal. (72). Tumor volume changes between time points were calculated as follows: (tumor volume time point 2 − tumor volume time point 1) / tumor volume time point 1 × 100%. Bristol Myers Squibb antibodies anti–PD1 (clone 4H2, g1D265A) and anti–CTLA4 (clone 9D9, mlgG2a), with mlgG1D265A and mlgG2a isotype controls, were given twice weekly at 200 μg per dose by intraperitoneal injection. MRTX849 (adagrasib) was given by oral gavage daily at 100 mg/kg for a total of 2 weeks. The mouse work was carried out with approval of the Francis Crick Institute Animal Welfare and Ethical Review Body under UK Home Office Project License P19FC0E42. Flow cytometry Flow cytometry was performed as previously (8) using the antibody mixes listed in table S1. Details of staining protocol, data acquisition, and analysis can be found in the Supplementary Materials. Supplementary Materials This PDF file includes: Supplementary Materials and Methods Figs. S1 to S8 table S1 References REFERENCES AND NOTES 1. h. Borghaei, l. PazAres, l. horn, d. R. Spigel, M. Steins, N. e. Ready, l. Q. Chow, e. e. vokes, e. Felip, e. holgado, F. Barlesi, M. Kohlhäufl, O. Arrieta, M. A. Burgio, J. Fayette, h. lena, e. Poddubskaya, d. e. Gerber, S. N. Gettinger, C. M. Rudin, N. Rizvi, l. Crinò, G. R. Blumenschein Jr., S. J. Antonia, C. dorange, C. t. harbison, F. Graf Finckenstein, J. R. Brahmer, Nivolumab versus docetaxel in advanced nonsquamous nonsmallcell lung cancer. N. Engl. J. Med. 373, 1627–1639 (2015). 2. S. N. Gettinger, A. Wurtz, S. B. Goldberg, d. Rimm, K. Schalper, S. Kaech, P. Kavathas, A. Chiang, R. lilenbaum, d. Zelterman, K. Politi, R. S. herbst, Clinical features and management of acquired resistance to Pd1 axis inhibitors in 26 patients with advanced nonsmall cell lung cancer. J. Thorac. Oncol. 13, 831–839 (2018). 3. A. J. de langen, M. l. Johnson, J. Mazieres, A. C. dingemans, G. Mountzios, M. Pless, J. Wolf, M. Schuler, h. lena, F. Skoulidis, Y. Yoneshima, S. W. Kim, h. linardou, S. Novello, A. J. van der Wekken, Y. Chen, S. Peters, e. Felip, B. J. Solomon, S. S. Ramalingam, C. dooms, C. R. lindsay, C. G. Ferreira, N. Blais, C. C. Obiozor, Y. Wang, B. Mehta, t. varrieur, G. Ngarmchamnanrith, B. Stollenwerk, d. Waterhouse, l. PazAres, K. i. CodeBrea, Sotorasib versus docetaxel for previously treated nonsmallcell lung cancer with KRAS(G12C) mutation: A randomised, openlabel, phase 3 trial. Lancet 401, 733–746 (2023). 4. F. Skoulidis, B. t. li, G. K. dy, t. J. Price, G. S. Falchook, J. Wolf, A. italiano, M. Schuler, h. Borghaei, F. Barlesi, t. Kato, A. CurioniFontecedro, A. Sacher, A. Spira, S. S. Ramalingam, t. takahashi, B. Besse, A. Anderson, A. Ang, Q. tran, O. Mather, h. henary, G. Ngarmchamnanrith, G. Friberg, v. velcheti, R. Govindan, Sotorasib for lung cancers with KRAS p.G12C mutation. N. Engl. J. Med. 384, 2371–2381 (2021). 5. M. MolinaArcas, A. Samani, J. downward, drugging the undruggable: Advances on RAS targeting in cancer. Genes 12, 899 (2021). 6. J. Canon, K. Rex, A. Y. Saiki, C. Mohr, K. Cooke, d. Bagal, K. Gaida, t. holt, C. G. Knutson, N. Koppada, B. A. lanman, J. Werner, A. S. Rapaport, t. S. Miguel, R. Ortiz, t. Osgood, J. R. Sun, X. Zhu, J. d. McCarter, l. P. volak, B. e. houk, M. G. Fakih, B. h. O’Neil, t. J. Price, G. S. Falchook, J. desai, J. Kuo, R. Govindan, d. S. hong, W. Ouyang, h. henary, t. Arvedson, v. J. Cee, J. R. lipford, the clinical KRAS(G12C) inhibitor AMG 510 drives antitumour immunity. Nature 575, 217–223 (2019). 7. d. M. Briere, S. li, A. Calinisan, N. Sudhakar, R. Aranda, l. hargis, d. h. Peng, J. deng, l. d. engstrom, J. hallin, S. Gatto, J. FernandezBanet, A. Pavlicek, K. K. Wong, J. G. Christensen, P. Olson, the KRASG12C inhibitor MRtX849 reconditions the tumor immune microenvironment and sensitizes tumors to checkpoint inhibitor therapy. Mol. Cancer Ther. 20, 975–985 (2021). 8. e. Mugarza, F. van Maldegem, J. Boumelha, C. Moore, S. Rana, M. llorian Sopena, P. east, R. Ambler, P. Anastasiou, P. RomeroClavijo, K. valand, M. Cole, M. MolinaArcas, J. downward, therapeutic KRASG12C inhibition drives effective interferonmediated antitumor immunity in immunogenic lung cancers. Sci. Adv. 8, eabm8780 (2022). 9. F. van Maldegem, K. valand, M. Cole, h. Patel, M. Angelova, S. Rana, e. Colliver, K. enfield, N. Bah, G. Kelly, v. S. K. tsang, e. Mugarza, C. Moore, P. hobson, d. levi, M. MolinaArcas, Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 17 of 19 C. Swanton, J. downward, Characterisation of tumour microenvironment remodelling following oncogene inhibition in preclinical studies with imaging mass cytometry. Nat. Commun. 12, 5906 (2021). 10. J. Boumelha, S. de Carne trecesson, e. K. law, P. RomeroClavijo, M. A. Coelho, K. W. Ng, e. Mugarza, C. Moore, S. Rana, d. R. Caswell, M. Murillo, d. C. hancock, P. P. Argyris, W. l. Brown, C. durfee, l. K. larson, R. i. vogel, A. SuarezBonnet, S. l. Priestnall, P. east, S. J. Ross, G. Kassiotis, M. MolinaArcas, C. Swanton, R. harris, J. downward, An immunogenic model of KRASmutant lung cancer enables evaluation of targeted therapy and immunotherapy combinations. Cancer Res. 82, 3435–3448 (2022). 11. A. Chour, J. denis, C. Mascaux, M. Zysman, l. BigayGame, A. Swalduz, v. Gounant, A. Cortot, M. darrason, v. Fallet, e. Auclin, C. Basse, C. tissot, C. decroisette, P. Bombaron, e. Girouxleprieur, l. Odier, S. Brosseau, Q. Creusot, M. Gueçamburu, C. Meersseman, A. Rochand, A. Costantini, C. M. Gaillard, e. Wasielewski, N. Girard, J. Cadranel, C. lafitte, F. lebossé, M. duruisseaux, Brief report: Severe sotorasibrelated hepatotoxicity and nonliver adverse events associated with sequential antiprogrammed cell death (ligand) 1 and sotorasib therapy in KRASG12Cmutant lung cancer. J. Thorac. Oncol. 18, 1408–1415 (2023). 12. C. Giesen, h. A. Wang, d. Schapiro, N. Zivanovic, A. Jacobs, B. hattendorf, P. J. Schuffler, d. Grolimund, J. M. Buhmann, S. Brandt, Z. varga, P. J. Wild, d. Gunther, B. Bodenmiller, highly multiplexed imaging of tumor tissues with subcellular resolution by mass cytometry. Nat. Methods 11, 417–422 (2014). 13. M. Sorin, M. Rezanejad, e. Karimi, B. Fiset, l. desharnais, l. J. M. Perus, S. Milette, M. W. Yu, S. M. Maritan, S. dore, e. Pichette, W. enlow, A. Gagne, Y. Wei, M. Orain, v. S. K. Manem, R. Rayes, P. M. Siegel, S. CamilleriBroet, P. O. Fiset, P. desmeules, J. d. Spicer, d. F. Quail, P. Joubert, l. A. Walsh, Singlecell spatial landscapes of the lung tumour immune microenvironment. Nature 614, 548–554 (2023). 14. K. S. S. enfield, e. Colliver, C. lee, A. Magness, d. A. Moore, M. Sivakumar, K. Grigoriadis, O. Pich, t. Karasaki, P. S. hobson, d. levi, S. veeriah, C. Puttick, e. l. Nye, M. Green, K. K. dijkstra, M. Shimato, A. U. Akarca, t. Marafioti, R. Salgado, A. hackshaw; tRACeRx consortium, M. Jamalhanjani, F. van Maldegem, N. McGranahan, B. Glass, h. Pulaski, e. Walk, J. l. Reading, S. A. Quezada, C. t. hiley, J. downward, e. Sahai, C. Swanton, M. Angelova, Spatial architecture of myeloid and t cells orchestrates immune evasion and clinical outcome in lung cancer. Cancer Discov. 14, 1018–1047 (2024). 15. C. M. Schurch, S. S. Bhate, G. l. Barlow, d. J. Phillips, l. Noti, i. Zlobec, P. Chu, S. Black, J. demeter, d. R. Mcilwain, S. Kinoshita, N. Samusik, Y. Goltsev, G. P. Nolan, Coordinated cellular neighborhoods orchestrate antitumoral immunity at the colorectal cancer invasive front. Cell 182, 1341–1359.e19 (2020). 16. B. Maier, A. M. leader, S. t. Chen, N. tung, C. Chang, J. leBerichel, A. Chudnovskiy, S. Maskey, l. Walker, J. P. Finnigan, M. e. Kirkling, B. Reizis, S. Ghosh, N. R. d’Amore, N. Bhardwaj, C. v. Rothlin, A. Wolf, R. Flores, t. Marron, A. h. Rahman, e. Kenigsberg, B. d. Brown, M. Merad, A conserved dendriticcell regulatory program limits antitumour immunity. Nature 580, 257–262 (2020). 17. A. Färkkilä, d. C. Gulhan, J. Casado, C. A. Jacobson, h. Nguyen, B. Kochupurakkal, Z. Maliga, C. Yapp, Y. A. Chen, d. Schapiro, Y. Zhou, J. R. Graham, B. J. dezube, P. Munster, S. Santagata, e. Garcia, S. Rodig, A. lako, d. Chowdhury, G. i. Shapiro, U. A. Matulonis, P. J. Park, S. hautaniemi, P. K. Sorger, e. M. Swisher, A. d. d’Andrea, P. A. Konstantinopoulos, immunogenomic profiling determines responses to combined PARP and Pd1 inhibition in ovarian cancer. Nat. Commun. 11, 1459 (2020). 18. M. Bortolomeazzi, M. R. Keddar, l. Montorsi, A. AchaSagredo, l. Benedetti, d. temelkovski, S. Choi, N. Petrov, K. todd, P. Wai, J. Kohl, t. denner, e. Nye, R. Goldstone, S. Ward, G. A. Wilson, M. Al Bakir, C. Swanton, S. John, J. Miles, B. larijani, v. Kunene, e. Fontana, h. t. Arkenau, P. J. Parker, M. RodriguezJusto, K. K. Shiu, J. Spencer, F. d. Ciccarelli, immunogenomics of colorectal cancer response to checkpoint blockade: Analysis of the KeYNOte 177 trial and validation cohorts. Gastroenterology 161, 1179–1193 (2021). 19. K. litchfield, J. l. Reading, C. Puttick, K. thakkar, C. Abbosh, R. Bentham, t. B. K. Watkins, R. Rosenthal, d. Biswas, A. Rowan, e. lim, M. Al Bakir, v. turati, J. A. GuerraAssuncao, l. Conde, A. J. S. Furness, S. K. Saini, S. R. hadrup, J. herrero, S. h. lee, P. van loo, t. enver, J. larkin, M. d. hellmann, S. turajlic, S. A. Quezada, N. McGranahan, C. Swanton, Metaanalysis of tumorand t cellintrinsic mechanisms of sensitization to checkpoint inhibition. Cell 184, 596–614.e14 (2021). 20. d. Schapiro, h. W. Jackson, S. Raghuraman, J. R. Fischer, v. R. t. Zanotelli, d. Schulz, C. Giesen, R. Catena, Z. varga, B. Bodenmiller, histoCAt: Analysis of cell phenotypes and interactions in multiplex image cytometry data. Nat. Methods 14, 873–876 (2017). 21. A. G. dykema, J. Zhang, l. S. Cheung, S. Connor, B. Zhang, Z. Zeng, C. M. Cherry, t. li, J. X. Caushi, M. Nishimoto, A. J. Munoz, Z. Ji, W. hou, W. Zhan, d. Singh, t. Zhang, R. Rashid, M. MitchellFlack, S. Bom, A. tam, N. ionta, t. h. K. Aye, Y. Wang, C. A. Sawosik, l. e. tirado, l. M. tomasovic, d. vandyke, J. B. Spangler, v. Anagnostou, S. Yang, J. Spicer, R. Rayes, J. taube, J. R. Brahmer, P. M. Forde, S. Yegnasubramanian, h. Ji, d. M. Pardoll, K. N. Smith, lung tumorinfiltrating treg have divergent transcriptional profiles and function linked to checkpoint blockade response. Sci. Immunol. 8, eadg1487 (2023). 22. N. A. Rizvi, B. C. Cho, N. Reinmuth, K. h. lee, A. luft, M. J. Ahn, M. M. van den heuvel, M. Cobo, d. vicente, A. Smolin, v. Moiseyenko, S. J. Antonia, S. le Moulec, G. Robinet, R. Natale, J. Schneider, F. A. Shepherd, S. l. Geater, e. B. Garon, e. S. Kim, S. B. Goldberg, K. Nakagawa, R. Raja, B. W. higgs, A. M. Boothman, l. Zhao, U. Scheuring, P. K. Stockman, v. K. Chand, S. Peters; MYStiC investigators, durvalumab with or without tremelimumab vs standard chemotherapy in firstline treatment of metastatic nonsmall cell lung cancer: the MYStiC phase 3 randomized clinical trial. JAMA Oncol. 6, 661–674 (2020). 23. M. Jamalhanjani, G. A. Wilson, N. McGranahan, N. J. Birkbak, t. B. K. Watkins, S. veeriah, S. Shafi, d. h. Johnson, R. Mitter, R. Rosenthal, M. Salm, S. horswell, M. escudero, N. Matthews, A. Rowan, t. Chambers, d. A. Moore, S. turajlic, h. Xu, S. M. lee, M. d. Forster, t. Ahmad, C. t. hiley, C. Abbosh, M. Falzon, e. Borg, t. Marafioti, d. lawrence, M. hayward, S. Kolvekar, N. Panagiotopoulos, S. M. Janes, R. thakrar, A. Ahmed, F. Blackhall, Y. Summers, R. Shah, l. Joseph, A. M. Quinn, P. A. Crosbie, B. Naidu, G. Middleton, G. langman, S. trotter, M. Nicolson, h. Remmen, K. Kerr, M. Chetty, l. Gomersall, d. A. Fennell, A. Nakas, S. Rathinam, G. Anand, S. Khan, P. Russell, v. ezhil, B. ismail, M. irvinSellers, v. Prakash, J. F. lester, M. Kornaszewska, R. Attanoos, h. Adams, h. davies, S. dentro, P. taniere, B. O’Sullivan, h. l. lowe, J. A. hartley, N. iles, h. Bell, Y. Ngai, J. A. Shaw, J. herrero, Z. Szallasi, R. F. Schwarz, A. Stewart, S. A. Quezada, J. le Quesne, P. van loo, C. dive, A. hackshaw, C. Swanton; tRACeRx Consortium, tracking the evolution of nonsmallcell lung cancer. N. Engl. J. Med. 376, 2109–2121 (2017). 24. M. J. Selby, J. J. engelhardt, M. Quigley, K. A. henning, t. Chen, M. Srinivasan, A. J. Korman, AntiCtlA4 antibodies of igG2a isotype enhance antitumor activity through reduction of intratumoral regulatory t cells. Cancer Immunol. Res. 1, 32–42 (2013). 25. F. A. vargas, A. J. S. Furness, K. litchfield, K. Joshi, R. Rosenthal, e. Ghorani, i. Solomon, M. h. lesko, N. Ruef, C. Roddie, J. Y. henry, l. Spain, A. B. Aissa, A. Georgiou, Y. N. S. Wong, M. Smith, d. Strauss, A. hayes, d. Nicol, t. O’Brien, l. Martensson, A. ljungars, i. teige, B. Frendeus; tRACeRx Melanoma; tRACeRx Renal; tRACeRx lung consortia, M. Pule, t. Marafioti, M. Gore, J. larkin, S. turajlic, C. Swanton, K. S. Peggs, S. A. Quezada, Fc effector function contributes to the activity of human antiCtlA4 antibodies. Cancer Cell 33, 649–663.e4 (2018). 26. A. Marabelle, h. Kohrt, i. SagivBarfi, B. Ajami, R. C. Axtell, G. Zhou, R. Rajapaksa, M. R. Green, J. torchia, J. Brody, R. luong, M. d. Rosenblum, l. Steinman, h. i. levitsky, v. tse, R. levy, depleting tumorspecific tregs at a single site eradicates disseminated tumors. J. Clin. Invest. 123, 2447–2463 (2013). 27. d. S. hong, M. G. Fakih, J. h. Strickler, J. desai, G. A. durm, G. i. Shapiro, G. S. Falchook, t. J. Price, A. Sacher, C. S. denlinger, Y. J. Bang, G. K. dy, J. C. Krauss, Y. Kuboki, J. C. Kuo, A. l. Coveler, K. Park, t. W. Kim, F. Barlesi, P. N. Munster, S. S. Ramalingam, t. F. Burns, F. MericBernstam, h. henary, J. Ngang, G. Ngarmchamnanrith, J. Kim, B. e. houk, J. Canon, J. R. lipford, G. Friberg, P. lito, R. Govindan, B. t. li, KRASG12C inhibition with sotorasib in advanced solid tumors. N. Engl. J. Med. 383, 1207–1217 (2020). 28. P. A. Janne, G. J. Riely, S. M. Gadgeel, R. S. heist, S. i. Ou, J. M. Pacheco, M. l. Johnson, J. K. Sabari, K. leventakos, e. Yau, l. Bazhenova, M. v. Negrao, N. A. Pennell, J. Zhang, K. Anderes, h. dertorossian, t. Kheoh, K. velastegui, X. Yan, J. G. Christensen, R. C. Chao, A. i. Spira, Adagrasib in nonsmallcell lung cancer harboring a KRASG12C mutation. N. Engl. J. Med. 387, 120–131 (2022). 29. J. K. Sabari, h. Park, A. W. tolcher, S. h. i. Ou, e. B. Garon, B. George, P. A. Janne, S. e. Moody, e. Y. tan, S. K. Sen, d. Peters, X. h. Yan, J. G. Christensen, A. S. Chi, R. S. heist, KRYStAl2: A phase i/ii trial of adagrasib (MRtX849) in combination with tNO155 in patients with advanced solid tumors with KRAS G12C mutation. J. Clin. Oncol. 39, tPS146 (2021). 30. S. Suzuki, K. Yonesaka, t. teramura, t. takehara, R. Kato, h. Sakai, K. haratani, J. tanizaki, h. Kawakami, h. hayashi, K. Sakai, K. Nishio, K. Nakagawa, KRAS inhibitor resistance in METAmplified KRASG12C nonsmall cell lung cancer induced by RASand nonRASmediated cell signaling mechanisms. Clin. Cancer Res. 27, 5697–5707 (2021). 31. t. vaclova, A. Chakraborty, J. Sherwood, S. Ross, d. Carroll, J. C. Barrett, J. downward, e. C. de Bruin, Concomitant KRAS mutations attenuate sensitivity of nonsmall cell lung cancer cells to KRAS G12C inhibition. Sci. Rep. 12, 2699 (2022). 32. N. tanaka, J. J. lin, C. li, M. B. Ryan, J. Zhang, l. A. Kiedrowski, A. G. Michel, M. U. Syed, K. A. Fella, M. Sakhi, i. Baiev, d. Juric, J. F. Gainor, S. J. Klempner, J. K. lennerz, G. Siravegna, l. BarPeled, A. N. hata, R. S. heist, R. B. Corcoran, Clinical acquired resistance to KRASG12C inhibition through a novel KRAS switchii pocket mutation and polyclonal alterations converging on RASMAPK reactivation. Cancer Discov. 11, 1913–1922 (2021). 33. M. M. Awad, S. liu, i. i. Rybkin, K. C. Arbour, J. dilly, v. W. Zhu, M. l. Johnson, R. S. heist, t. Patil, G. J. Riely, J. O. Jacobson, X. Yang, N. S. Persky, d. e. Root, K. e. lowder, h. Feng, S. S. Zhang, K. M. haigis, Y. P. hung, l. M. Sholl, B. M. Wolpin, J. Wiese, J. Christiansen, J. lee, A. B. Schrock, l. P. lim, K. Garg, M. li, l. d. engstrom, l. Waters, J. d. lawson, P. Olson, P. lito, S. i. Ou, J. G. Christensen, P. A. Janne, A. J. Aguirre, Acquired resistance to KRASG12C inhibition in cancer. N. Engl. J. Med. 384, 2382–2393 (2021). 34. h. Y. Khan, M. Nagasaka, Y. li, A. Aboukameel, M. h. Uddin, R. Sexton, S. Bannoura, Y. Mzannar, M. N. Alhallak, S. Kim, R. Beydoun, Y. landesman, h. Mamdani, d. Uprety, P. A. Philip, R. M. Mohammad, A. F. Shields, A. S. Azmi, inhibitor of the nuclear transport protein XPO1 enhances the anticancer efficacy of KRAS G12C inhibitors in preclinical models of KRAS G12Cmutant cancers. Cancer Res. Commun. 2, 342–352 (2022). Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 18 of 19 35. S. Khan, J. Wiegand, P. Zhang, W. hu, d. thummuri, v. Budamagunta, N. hua, l. Jin, C. J. Allegra, S. e. Kopetz, M. ZajacKaye, F. J. Kaye, G. Zheng, d. Zhou, BClXl PROtAC degrader dt2216 synergizes with sotorasib in preclinical models of KRASG12Cmutated cancers. J. Hematol. Oncol. 15, 23 (2022). 36. S. l. tammaccaro, P. Prigent, J. C. le Bail, O. dosSantos, l. dassencourt, M. eskandar, A. Buzy, O. venier, J. C. Guillemot, Y. veeranagouda, M. didier, e. Spanakis, t. Kanno, M. Cesaroni, S. Mathieu, l. Canard, A. Casse, F. Windenberger, l. Calvet, l. Noblet, S. Sidhu, l. debussche, J. Moll, i. valtingojer, teAd inhibitors sensitize KRASG12C inhibitors via dual cell cycle arrest in KRASG12Cmutant NSClC. Pharmaceuticals 16, 553 (2023). 37. M. MolinaArcas, C. Moore, S. Rana, F. van Maldegem, e. Mugarza, P. RomeroClavijo, e. herbert, S. horswell, l. S. li, M. R. Janes, d. C. hancock, J. downward, development of combination therapies to maximize the impact of KRASG12C inhibitors in lung cancer. Sci. Transl. Med. 11, eaaw7999 (2019). 38. P. Anastasiou, C. Moore, S. Rana, A. de Castro, M. tomaschko, J. Boumelha, e. Mugarza, C. Blaj, S. de Carné trécesson, R. Goldstone, J. Smith, e. Quintana, M. MolinaArcas, J., downward, Combining RASG12C (ON) inhibitor with ShP2 inhibition sensitises immune excluded lung tumours to immune checkpoint blockade: A strategy for turning cold tumours hot. bioRxiv 2024.01.15.575765 [Preprint] (2024). https://doi.org/10.1101/2024.01.15.575765. 39. S. B. Kemp, N. Cheng, N. Markosyan, R. Sor, i. K. Kim, J. hallin, J. Shoush, l. Quinones, N. v. Brown, J. B. Bassett, N. Joshi, S. Yuan, M. Smith, W. P. vostrejs, K. Z. Perezvale, B. Kahn, F. Mo, t. R. donahue, C. G. Radu, C. Clendenin, J. G. Christensen, R. h. vonderheide, B. Z. Stanger, efficacy of a smallmolecule inhibitor of KrasG12d in immunocompetent models of pancreatic cancer. Cancer Discov. 13, 298–311 (2023). 40. v. Kumarasamy, J. Wang, C. Frangou, Y. Wan, A. dynka, h. Rosenheck, P. dey, e. v. Abel, e. S. Knudsen, A. K. Witkiewicz, the extracellular niche and tumor microenvironment enhance KRAS inhibitor efficacy in pancreatic cancer. Cancer Res. 84, 1115–1132 (2024). 41. K. K. Mahadevan, K. M. McAndrews, v. S. leBleu, S. Yang, h. lyu, B. li, A. M. Sockwell, M. l. Kirtley, S. J. Morse, B. A. Moreno diaz, M. P. Kim, N. Feng, A. M. lopez, P. A. Guerrero, F. Paradiso, h. Sugimoto, K. A. Arian, h. Ying, Y. Barekatain, l. K. Sthanam, P. J. Kelly, A. Maitra, t. P. heffernan, R. Kalluri, KRASG12d inhibition reprograms the microenvironment of early and advanced pancreatic cancer to promote FASmediated killing by Cd8+ t cells. Cancer Cell 41, 1606–1620.e8 (2023). 42. M. MolinaArcas, J. downward, exploiting the therapeutic implications of KRAS inhibition on tumor immunity. Cancer Cell 42, 338–357 (2024). 43. J. G. Mayo, Biologic characterization of the subcutaneously implanted lewis lung tumor. Cancer Chemother. Rep. 3, 325–330 (1972). 44. e. danenberg, h. Bardwell, v. R. t. Zanotelli, e. Provenzano, S. F. Chin, O. M. Rueda, A. Green, e. Rakha, S. Aparicio, i. O. ellis, B. Bodenmiller, C. Caldas, h. R. Ali, Breast tumor microenvironment structures are associated with genomic features and clinical outcome. Nat. Genet. 54, 660–669 (2022). 45. e. Karimi, M. W. Yu, S. M. Maritan, l. J. M. Perus, M. Rezanejad, M. Sorin, M. dankner, P. Fallah, S. dore, d. Zuo, B. Fiset, d. J. Kloosterman, l. Ramsay, Y. Wei, S. lam, R. Alsajjan, i. R. Watson, G. Roldan Urgoiti, M. Park, d. Brandsma, d. l. Senger, J. A. Chan, l. Akkari, K. Petrecca, M. C. Guiot, P. M. Siegel, d. F. Quail, l. A. Walsh, Singlecell spatial immune landscapes of primary and metastatic brain tumours. Nature 614, 555–563 (2023). 46. d. tavernari, e. Battistello, e. dheilly, A. S. Petruzzella, M. Mina, J. Sordetdessimoz, S. Peters, t. Krueger, d. Gfeller, N. Riggi, e. Oricchio, i. letovanec, G. Ciriello, Nongenetic evolution drives lung adenocarcinoma spatial heterogeneity and progression. Cancer Discov. 11, 1490–1507 (2021). 47. K. M. van Pul, M. F. Fransen, R. van de ven, t. d. de Gruijl, immunotherapy goes local: the central role of lymph nodes in driving tumor infiltration and efficacy. Front. Immunol. 12, 643291 (2021). 48. A. Magen, P. hamon, N. Fiaschi, B. Y. Soong, M. d. Park, R. Mattiuz, e. humblin, l. troncoso, d. d’Souza, t. dawson, J. Kim, S. hamel, M. Buckup, C. Chang, A. tabachnikova, h. Schwartz, N. Malissen, Y. lavin, A. SoaresSchanoski, B. Giotti, S. hegde, G. ioannou, e. GonzalezKozlova, C. hennequin, J. le Berichel, Z. Zhao, S. C. Ward, i. Fiel, B. Kou, M. dobosz, l. li, C. Adler, M. Ni, Y. Wei, W. Wang, G. S. Atwal, K. Kundu, K. J. Cygan, A. M. tsankov, A. Rahman, C. Price, N. Fernandez, J. he, N. t. Gupta, S. KimSchulze, S. Gnjatic, e. Kenigsberg, R. P. deering, M. Schwartz, t. U. Marron, G. thurston, A. O. Kamphorst, M. Merad, intratumoral dendritic cellCd4+ t helper cell niches enable Cd8+ t cell differentiation following Pd1 blockade in hepatocellular carcinoma. Nat. Med. 29, 1389–1399 (2023). 49. A. Schmidt, N. Oberle, P. h. Krammer, Molecular mechanisms of tregmediated t cell suppression. Front. Immunol. 3, 51 (2012). 50. M. tekguc, J. B. Wing, M. Osaki, J. long, S. Sakaguchi, tregexpressed CtlA4 depletes Cd80/Cd86 by trogocytosis, releasing free Pdl1 on antigenpresenting cells. Proc. Natl. Acad. Sci. U.S.A. 118, e2023739118 (2021). 51. A. Kennedy, e. Waters, B. Rowshanravan, C. hinze, C. Williams, d. Janman, t. A. Fox, C. Booth, A. M. Pesenacker, N. halliday, B. Soskic, S. Kaur, O. S. Qureshi, e. C. Morris, S. ikemizu, C. Paluch, J. huo, S. J. davis, e. Boucrot, l. S. K. Walker, d. M. Sansom, differences in Cd80 and Cd86 transendocytosis reveal Cd86 as a key target for CtlA4 immune regulation. Nat. Immunol. 23, 1365–1378 (2022). 52. i. Yofe, t. landsberger, A. Yalin, i. Solomon, C. Costoya, d. F. demane, M. Shah, e. david, C. Borenstein, O. Barboy, i. Matos, K. S. Peggs, S. A. Quezada, i. Amit, AntiCtlA4 antibodies drive myeloid activation and reprogram the tumor microenvironment through FcγR engagement and type i interferon signaling. Nat. Cancer 3, 1336–1350 (2022). 53. B. t. li, G. S. Falchook, G. A. durm, t. F. Burns, F. Skoulidis, S. S. Ramalingam, A. Spira, C. M. Bestvina, S. B. Goldberg, R. veluswamy, W. t. iams, A. A. Chiappori, C. R. lemech, A. R. Meloni, v. ebiana, t. dai, d. M. Gauto, t. l. varrieur, W. J. Snyder, R. Govindan, OA03.06 CodeBreaK 100/101: First report of safety/efficacy of sotorasib in combination with pembrolizumab or atezolizumab in advanced KRAS p.G12C NSClC. J. Thorac. Oncol. 17, S10–S11 (2022). 54. t. M. Pentimalli, S. Schallenberg, d. leónPeriñán, i. legnini, i. theurillat, G. thomas, A. Boltengagen, S. Fritzsche, J. Nimo, l. Ruff, G. dernbach, P. Jurmeister, S. Murphy, M. t. Gregory, Y. liang, M. Cordenonsi, S. Piccolo, F. Coscia, A. Woehler, N. Karaiskos, F. Klauschen, N. Rajewsky, highresolution molecular atlas of a lung tumor in 3d. bioRxiv 2023.05.10.539644 [Preprint] (2024). https://doi.org/10.1101/2023.05.10.539644. 55. K. Shimizu, M. Nakata, Y. hirami, t. Yukawa, A. Maeda, K. tanemoto, tumorinfiltrating Foxp3+ regulatory t cells are correlated with cyclooxygenase2 expression and are associated with recurrence in resected nonsmall cell lung cancer. J. Thorac. Oncol. 5, 585–590 (2010). 56. d. h. Kang, C. Chung, P. Sun, d. h. lee, S. i. lee, d. Park, J. S. Koh, Y. Kim, h. S. Yi, J. e. lee, Circulating regulatory t cells predict efficacy and atypical responses in lung cancer patients treated with Pd1/Pdl1 inhibitors. Cancer Immunol. Immunother. 71, 579–588 (2022). 57. R. Saleh, e. elkord, FoxP3+ t regulatory cells in cancer: Prognostic biomarkers and therapeutic targets. Cancer Lett. 490, 174–185 (2020). 58. M. van Gulijk, A. van Krimpen, S. Schetters, M. eterman, M. van elsas, J. Mankor, l. Klaase, M. de Bruijn, M. van Nimwegen, t. van tienhoven, W. van ijcken, l. Boon, J. van der Schoot, M. verdoes, F. Scheeren, S. h. van der Burg, B. N. lambrecht, R. Stadhouders, F. dammeijer, J. Aerts, t. van hall, Pdl1 checkpoint blockade promotes regulatory t cell activity that underlies therapy resistance. Sci. Immunol. 8, eabn6173 (2023). 59. l. van hooren, S. M. handgraaf, d. J. Kloosterman, e. Karimi, l. van Mil, A. A. Gassama, B. G. Solsona, M. h. P. de Groot, d. Brandsma, d. F. Quail, l. A. Walsh, G. R. Borst, l. Akkari, Cd103+ regulatory t cells underlie resistance to radioimmunotherapy and impair Cd8+ t cell activation in glioblastoma. Nat. Cancer 4, 665–681 (2023). 60. M. Sorin, e. Karimi, M. Rezanejad, M. W. Yu, l. desharnais, S. A. C. Mcdowell, S. dore, A. Arabzadeh, v. Breton, B. Fiset, Y. Wei, R. Rayes, M. Orain, F. Coulombe, v. S. K. Manem, A. Gagne, d. F. Quail, P. Joubert, J. d. Spicer, l. A. Walsh, Singlecell spatial landscape of immunotherapy response reveals mechanisms of CXCl13 enhanced antitumor immunity. J. Immunother. Cancer 11, e005545 (2023). 61. J. h. Chen, l. t. Nieman, M. Spurrell, v. Jorgji, l. elmelech, P. Richieri, K. h. Xu, R. Madhu, M. Parikh, i. Zamora, A. Mehta, C. S. Nabel, S. S. Freeman, J. d. Pirl, C. lu, C. B. Meador, J. l. Barth, M. Sakhi, A. l. tang, S. Sarkizova, C. Price, N. F. Fernandez, G. emanuel, J. he, K. van Raay, J. W. Reeves, K. Yizhak, M. hofree, A. Shih, M. SadeFeldman, G. M. Boland, K. Pelka, M. J. Aryee, M. MinoKenudson, J. F. Gainor, i. Korsunsky, N. hacohen, human lung cancer harbors spatially organized stemimmunity hubs associated with response to immunotherapy. Nat. Immunol. 25, 644–658 (2024). 62. d. Planchard, N. Reinmuth, S. Orlov, J. R. Fischer, S. Sugawara, S. Mandziuk, d. MarquezMedina, S. Novello, Y. takeda, R. Soo, K. Park, M. McCleod, S. l. Geater, M. Powell, R. May, U. Scheuring, P. Stockman, d. Kowalski, ARCtiC: durvalumab with or without tremelimumab as thirdline or later treatment of metastatic nonsmallcell lung cancer. Ann. Oncol. 31, 609–618 (2020). 63. e. B. Garon, B. C. Cho, N. Reinmuth, K. h. lee, A. luft, M. J. Ahn, G. Robinet, S. le Moulec, R. Natale, J. Schneider, F. A. Shepherd, M. C. Garassino, S. l. Geater, Z. P. Szekely, t. van Ngoc, F. liu, U. Scheuring, N. Patel, S. Peters, N. A. Rizvi, Patientreported outcomes with durvalumab with or without tremelimumab versus standard chemotherapy as firstline treatment of metastatic nonsmallcell lung cancer (MYStiC). Clin. Lung Cancer 22, 301–312.e8 (2021). 64. M. d. hellmann, l. PazAres, R. Bernabe Caro, B. Zurawski, S. W. Kim, e. Carcereny Costa, K. Park, A. Alexandru, l. lupinacci, e. de la Mora Jimenez, h. Sakai, i. Albert, A. vergnenegre, S. Peters, K. Syrigos, F. Barlesi, M. Reck, h. Borghaei, J. R. Brahmer, K. J. O’Byrne, W. J. Geese, P. Bhagavatheeswaran, S. K. Rabindran, R. S. Kasinathan, F. e. Nathan, S. S. Ramalingam, Nivolumab plus ipilimumab in advanced nonsmallcell lung cancer. N. Engl. J. Med. 381, 2020–2031 (2019). 65. A. v. Serritella, N. K. Shenoy, Nivolumab plus ipilimumab vs nivolumab alone in advanced cancers other than melanoma: A metaanalysis. JAMA Oncol. 9, 1441–1446 (2023). 66. A. elKhoueiry, A. Bullock, A. tsimberidou, d. Mahadevan, B. Wilky, P. twardowski, B. Bockorny, J. Moser, W. O. Feliu, J. Grossman, K. Rosenthal, S. O’day, M. Gordon, 479 AGeN1181, an Fcenhanced antiCtlA4 antibody, alone and in combination with balstilimab (antiPd1) in patients with advanced solid tumors: initial phase i results. J. Immunother. Cancer 9, A509–A509 (2021). Downloaded from https://www.science.org on November 15, 2024 Cole et al., Sci. Adv. 10, eadl6464 (2024) 1 November 2024 SCieNCe AdvANCeS | ReSeARCh ARtiCle 19 of 19 67. t. Zhang, J. Sun, J. li, Y. Zhao, t. Zhang, R. Yang, X. Ma, Safety and efficacy profile of mogamulizumab (Poteligeo) in the treatment of cancers: An update evidence from 14 studies. BMC Cancer 21, 618 (2021). 68. F. A. vargas, A. J. S. Furness, i. Solomon, K. Joshi, l. Mekkaoui, M. h. lesko, e. M. Rota, R. dahan, A. Georgiou, A. Sledzinska, A. B. Aissa, d. Franz, M. W. Sunderland, Y. N. S. Wong, J. Y. henry, t. O’Brien, d. Nicol, B. Challacombe, S. A. Beers, t. C. Melanoma, t. C. Renal, t. C. lung, S. turajlic, M. Gore, J. larkin, C. Swanton, K. A. Chester, M. Pule, J. v. Ravetch, t. Marafioti, K. S. Peggs, S. A. Quezada, Fcoptimized antiCd25 depletes tumorinfiltrating regulatory t cells and synergizes with Pd1 blockade to eradicate established tumors. Immunity 46, 577–586 (2017). 69. K. M. van Pul, J. C. l. Notohardjo, M. F. Fransen, B. d. Koster, A. G. M. Stam, d. Chondronasiou, S. M. lougheed, J. Bakker, v. Kandiah, M. P. van den tol, K. Jooss, R. vuylsteke, A. J. M. van den eertwegh, t. d. de Gruijl, local delivery of lowdose antiCtlA4 to the melanoma lymphatic basin leads to systemic treg reduction and effector t cell activation. Sci. Immunol. 7, eabn8097 (2022). 70. l. J. P. van der Maaten, G. e. hinton, visualizing highdimensional data using tSNe. J. Mach. Learn. Res. 9, 2579–2605 (2008). 71. N. eling, N. damond, t. hoch, B. Bodenmiller, cytomapper: An R/Bioconductor package for visualization of highly multiplexed imaging data. Bioinformatics 36, 5706–5708 (2020). 72. M. Zaw thin, C. Moore, t. Snoeks, t. Kalber, J. downward, A. Behrens, MicroCt acquisition and image processing to track and characterize pulmonary nodules in mice. Nat. Protoc. 18, 990–1015 (2022); https://doi.org/10.1038/s41596022007695. 73. J. h. levine, e. F. Simonds, S. C. Bendall, K. l. davis, e.- A. d. Amir, M. d. tadmor, O. litvin, h. G. Fienberg, A. Jager, e. R. Zunder, R. Finck, A. l. Gedman, i. Radtke, J. R. downing, d. Pe’er, G. P. Nolan, datadriven phenotypic dissection of AMl reveals progenitorlike cells that correlate with prognosis. Cell 162, 184–197 (2015). Acknowledgments: We thank the science technology platforms at the Francis Crick institute including Biological Resources, Scientific Computing, Bioinformatics and Biostatistics, and Flow Cytometry and the Microscopy and Cytometry Core Facility at the Amsterdam UMC, location vUmc. Funding: this work was supported by the Francis Crick institute that receives its core funding from Cancer Research UK (FC001070), the UK Medical Research Council (FC001070), and the Wellcome trust (FC001070). this work also received funding from the european Research Council Advanced Grant RASimmune and from a collaborative research agreement with Bristol Myers Squibb. F.v.M. is the recipient of the Amsterdam UMC fellowship and receives funding from european Union (grant agreements 101079113 and 101136552). M.C. is funded by a Phd studentship from AstraZeneca. K.S.S.e. reports support from european Union’s horizon 2020 research and innovation program under the Marie SkłodowskaCurie grant agreement no. 838540 and grants from the Royal Society (RF\eRe\210216). Author contributions: Writing—original draft: M.C., P.A., J.d., and F.v.M. Conceptualization: M.C., C.l., e.M., M.M.- A., t.d.d.G., C.S., J.d., and F.v.M. investigation: P.A., e.M., A.d.C., C.M., K.v., S.R., M.A., M.M.- A., C.S., and F.v.M. Writing—review and editing: M.C., K.S.S.e., e.C., M.M.- A., t.d.d.G., J.d., and F.v.M. Methodology: M.C., P.A., C.M., K.v., e.M., A.Ma., M.J., G.K., J.d., and F.v.M. Resources: C.S., J.d., and F.v.M. data curation: X.Y., J.R., K.S.S.e., M.A., and F.v.M. validation: M.C., C.M., and F.v.M. Supervision: M.J., J.d., and F.v.M. Formal analysis: M.C., P.A., C.l., X.Y., J.R., C.M., A.Mu., M.J., G.K., and F.v.M. Software: M.C., e.C., K.v., A.Mu., G.K., M.J., and F.v.M. Project administration: J.d. and F.v.M. visualization: M.C., P.A., C.l., X.Y., J.R., G.K., J.d., and F.v.M. Funding acquisition: J.d. Competing interests: J.d. has acted as a consultant for AstraZeneca, Jubilant, theras, Roche, and vividion and has funded research agreements with Bristol Myers Squibb, Revolution Medicines, and AstraZeneca. C.S. acknowledges grant support from Pfizer, AstraZeneca, Bristol Myers Squibb, Rocheventana, Boehringeringelheim, Archer dx inc. (collaboration in minimal residual disease sequencing technologies), and Ono Pharmaceutical; is an AstraZeneca advisory board member and chief investigator for the MeRmaid1 clinical trial; has consulted for Pfizer, Novartis, GlaxoSmithKline, MSd, Bristol Myers Squibb, Celgene, AstraZeneca, illumina, Genentech, Rocheventana, GRAil, Medicxi, Bicycle therapeutics, and the Sarah Cannon Research institute; has stock options in Apogen Biotechnologies, epic Bioscience, and GRAil; has stock options; and is cofounder of Achilles therapeutics. t.d.d.G. is advisor to lAvA therapeutics, Ge healthCare, and Mendus; received research funding from idera Pharmaceuticals; and holds stocks in lAvA therapeutics. M.C., P.A., J.d., and F.v.M. are the authors on a patent application related to this work filed by Bristol Myers Squibb (no. 63/616,357, filed 29 december 2023). the authors declare that they have no other competing interests. Data and materials availability: dataset 1 (https://hdl.handle.net/10779/crick.c.5270621.v2), dataset 2 (https://doi.org/10.25418/crick.19590259), and dataset 3 (https://doi.org/10.5281/ zenodo.12566209) tRACeRx iMC data used or analyzed during this study are available through the CRUK and UCl Cancer trials Centre (ctc. tracerx@ ucl. ac. uk) for academic noncommercial research purposes. Access will be granted upon review of a project proposal, which will be evaluated by a tRACeRx data access committee, and entering into an appropriate data access agreement, subject to any applicable ethical approvals. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Source code is available from https://github.com/FrancisCrickinstitute/Cole_2024 and https://doi. org/10.5281/zenodo.12566630. Submitted 30 October 2023 Accepted 27 September 2024 Published 1 November 2024 10.1126/sciadv.adl6464 Downloaded from https://www.science.org on November 15, 2024