American Association for Cancer Research® (AACR) Annual Meeting 2024

Neoadjuvant and perioperative immunotherapy with checkpoint inhibitors (ICIs), combined with chemotherapy (CT), have improved outcomes for patients with resectable non-small cell lung cancer (NSCLC). Pathologic tumor response has been utilized in clinical trials as a surrogate of clinical efficacy. Yet, over half of treated patients experience non-response, highlighting the need for novel markers to identify those who are most likely to derive therapeutic benefit.
To dissect the spatial complexity of the tumor-associated immune landscape following neoadjuvant chemoimmunotherapy (CT+ICI), we selected patients (n =3) with a range of residual viable tumor (%RVT) – pathological complete response (pCR; 0% RVT), partial response (20% RVT), and no-response (100% RVT). We performed multimodal spatial transcriptomics (ST) and proteomics (using a 35-plex protein panel) on the same tissue sections using the Visium CytAssist from 10X Genomics. We also examined spatial patterns of 20-plex panel of immune and stromal markers at single-cell resolution in the same samples using the COMET Lunaphore platform. In the pCR patient, we studied two tumor areas to assess the impact of tumor response heterogeneity on the immune landscape.
Our analysis identified distinct immune profiles across three patients with varying %RVT, reflecting the heterogeneity of the immune response. We found robust gene-protein correlation, such as MS4A1 gene expression and CD20 protein levels, in our multimodal spatial-omics analysis. Intriguingly, the arrangement of B cells and their interaction with surrounding plasma cells (PCs) varied notably across patients. Unlike in the patients with partial or no response, pronounced tertiary lymphoid structure (TLS) formation and signatures were observed in the pCR patient, with the CXCL13-CXCR5 axis overlapping with the TLS profile, suggesting a functional interplay that may be crucial in mediating response to therapy. Intriguingly, in the pCR patient, PCs uniquely arrayed around dense B cell aggregates, marked by MS4A1, CR2, FCER1, and CXCL13. PCs were not only less abundant in non-PCR patients but they also did not array around lymphoid aggregates.
Our high-resolution and multimodal analysis identified spatially-resolved expression and patterns for B lineage cells that may be associated, and thus, underlie response of resectable NSCLC to neoadjuvant CT+ICI. Ongoing efforts on this expanding cohort will shed further light on transcriptional states and immunogenomic roles of B lineage cells in neoadjuvant CT+ICI-treated NSCLC.

Z. Rahal¹, T. Zhou¹, S. Yang¹, A. Serrano¹, J. Feng¹, A. Sinjab¹, J. T. Le¹, X. Sun¹, M. Wang¹, W. Hu¹, J. Zang¹, T. Bruno², H. T. Tran¹, S. G. Swisher¹, C. H. Leung¹, H. Y. Lin¹, J. J. Lee¹, J. Wang¹, J. V. Heymach¹, D. L. Gibbons¹, I. I. Wistuba¹, A. Weissferdt¹, J. K. Burks¹, L. M. Solis Soto¹, H. Kadara¹, T. Cascone¹;
¹The University of Texas MD Anderson Cancer Center, Houston, TX, ²University of Pittsburgh, Pittsburgh, PA

Spatial biology techniques have revolutionized our approach towards the study of the tumor microenvironment (TME) and its complex cellular interplay. On the one hand, multiplex immunofluorescence (mIF) methods have enabled a precise profiling of immune cells and other key cellular players of the TME while uncovering their spatial distribution and interactions. On the other hand, in situ hybridization (ISH) technologies have shown to provide complementary information to protein profiling such as the mapping of cytokine- and chemokine-expressing cells, essential for comprehending signaling networks and immune activation statuses.
Here, we made use of a novel multiomics approach that combines these two biological inputs by integrating RNAscope™ and sequential immunofluorescence (seqIF™) protocols to achieve same-slide co-detection of RNA and protein targets. The combined workflow is automated on COMET™, an advanced tissue staining and imaging platform. Through precise control over temperature and reagent distribution, the instrument ensures maximum efficiency and reproducibility of the assays. The integrated multiomics protocol allows for up to three RNAscope™ detection cycles combined with twelve seqIF™ cycles, for a final 12-plex RNA + 24-plex protein panel.
We first showed the capacity of COMET™ platform to fully automate the RNAscope™ protocol and demonstrated its sensitivity and specificity with the analysis of positive and negative control genes on HeLa cell pellets, highlighting the flexibility of sample selection. To illustrate the potential multiomics in unravelling the complexity of the TME, we designed a panel of 12-RNA-targeting probes for key biomarkers of tumor-infiltrating lymphocytes and their activation status, including multiple secreted molecules (e.g., cytokines and proteases). This was combined with a 24-antibody panel for the detection of protein biomarkers, selected to enable the single-cell profiling of different players within the TME. We applied this panel to characterize different human FFPE tumor tissues and showed that the co-detection of RNA and protein biomarkers on the same slide allows a better characterization of key cellular components involved in tumor progression and immune response.
Our results highlight the potential of spatial multiomics technologies in enhancing the efficiency of investigations of immune cell behavior and generally the understanding of cellular interplay in the TME. Their full automation on a platform like COMET™ will accelerate analysis and increase robustness by minimizing user intervention, and ultimately helping in the development of prognostic and predictive biomarkers, in the refinement of cancer diagnoses, and in the selection of new personalized therapies.

P. Juricic¹, A. Manoukian¹, A. Comberlato¹, P. Bordignon¹, A. Failletaz¹, A. Dikshit², E. Doolittle², R. Delvillar², S. Zhou², C.-w. Chang², L.-c. Wang², S. Brajkovic¹, M. Srinivasan², D. Dupouy¹;
¹Lunaphore, Lausanne, Switzerland, ²Advanced Cell Diagnostics, Newark, NJ

Understanding the earliest changes during lung adenocarcinoma (LUAD) development can set the stage for discovery of fertile targets for disease interception, thereby mitigating the dire public health burden of LUAD. We previously identified bulk-level molecular and immunological changes that are enriched in normal-appearing tissue (NAT) in the local niche of human LUAD, as well as those that commence in adenomatous premalignant lesions (aPMLs, LUAD precursors) and that are further enriched in LUADs. Yet, in-depth understanding of the identities, states, and properties of specific cell subsets in NAT and precancer that trigger LUAD remain largely elusive due to inherent roadblocks to sampling and characterizing aPMLs that are at the center of this trajectory. Here, we aimed to map molecular profiles, states, and interactions of cell subsets that underlie initiation in lesion-adjacent NAT, and to determine how features of these cells evolve along the aPML-LUAD spectrum. We analyzed an expanding cohort of archived matching NATs, aPMLs and LUADs, including challenging small samples, from up to 17 patients. Samples were analyzed using combined high-resolution, multi-modal Visium spatial transcriptomics (ST, n=17) and proteomics (n=12), as well as single-cell RNA-sequencing of fixed cells (scFFPE-seq) from consecutive sections of the same tissues from 15 patients. We also studied a subset of samples by high-plex spatial proteomics (COMET) or subcellular spatial gene expression analysis (Xenium). In total, we sequenced more than half a million single cells comprising diverse epithelial, immune, and stromal subsets, which were concordant with lineage clusters identified by ST analysis of the same samples. Loss of alveolar differentiation was a hallmark of tumor cell-enriched areas in aPMLs and LUADs, with a more pronounced effect in the LUADs. Alveolar intermediate cells (AICs), which we have previously shown to be involved in LUAD initiation, were found in lesion-adjacent NAT, and their signature was enriched in matching aPMLs and further in LUADs. In line with these findings, trajectory analysis showed that AICs are derived from normal alveolar type 2 cells and likely progress towards tumor cells. Spatial neighborhood analysis pointed towards strong crosstalk between AICs and macrophages. We also noted heterogeneity in immune infiltration including patterns indicative of progressive features along the NAT-aPML-LUAD continuum. For example, B cell lineages were not only increased in abundance in LUADs relative to aPMLs, but they also mobilized into relatively more mature tertiary lymphoid aggregates/structures. While T regulatory cells aggregated close to LUADs, they were rather scattered in matching aPMLs. Our multimodal and spatial atlas of NAT, nearby aPML and LUAD elucidates the earliest cellular events underlying transition of NAT to LUAD and that could inform of targets for interception of this trajectory.

A. Sinjab¹, F. Peng¹, Y. Liu¹, S. Yang¹, T. Zhou¹, A. G. Serrano¹, J. Feng¹, L. Gomez Bolanos¹, G. Han¹, D. G. Rosen¹, S. G. Swisher¹, A. Spira², S. M. D. Dubinett³, L. M. Solis Soto¹, M. Li⁴, J. Fujimoto⁵, J. Burks¹, I. I. Wistuba¹, L. Wang¹, H. Kadara¹;
¹UT MD Anderson Cancer Ctr., Houston, TX, ²Boston University, Boston, MA, ³The University of California Los Angeles, Los Angeles, CA, ⁴University of Pennsylvania, Philadelphia, PA, ⁵Hiroshima University Hospital, Hiroshima, Japan

Desmoplastic small round cell tumor (DSRCT) is a rare and usually incurable aggressive sarcoma subtype. All malignant cells harbor a pathognomonic EWSR1::WT1 fusion protein (FP). However, FP-targeted agents are nonexistent; less than 20% of patients survive beyond five years. The spatial organization of DSRCT, consisting of tumor nests surrounded by desmoplastic stroma, is a hallmark of this disease. Given this context, we seek to understand the spatial heterogeneity within DSRCT and the tumor-stroma interactions to reveal potential therapeutic vulnerabilities. We generated a 20-panel marker developed to interrogate the cellular constituents of the tumor microenvironment (TME) and to characterize the multilineage expression DSRCT, including androgen receptor (AR) and neuroendocrine signatures (NE). The following markers were used: DSRCT (ST6GALNAC, pan-Cytokeratin), Fibroblasts (Collagen, α-smooth muscle actin), Endothelial Cells (CD31, α-smooth muscle actin), T-cells (CD45, CD4, CD8), Macrophages (CD45, CD68, CD163, CD11c), AR/NE markers (AR, neural-specific enolase, synaptophysin). We explored the expression of these markers in a nine-patient cohort from 9 x 9 mm² tissue sections imaged by the Lunaphore COMET. We used a deep learning model in Visiopharm to extract high-quality areas and segment cells based on their nucleus. The spatial image data was converted into matrix data, which could be further analyzed in R. We confirmed that the tumor nests are surrounded by dense desmoplasia by collagen expression. We identified tumor cells, fibroblasts, immune cells, and endothelial cells by thresholding for classical cell markers. ST6GALNAC, a sialyltransferase that was found to be highly expressed in DSRCT based on RNA-seq data, marked the tumor nests. Previous data from single-nucleus RNA-sequencing (snRNA-seq) showed that DSRCT exhibited three possible subtypes: AR+/NE-, AR-/NE+, and Hybrid AR+/NE+. We found concordance between the prior transcriptomic signatures and the proteomic markers. In tumor nests that were positive for AR, we found AR localized to the nucleus, suggesting downstream activation of this pathway. High AR expression was also correlated with positive expression of pan-cytokeratin in the tumor nests. Future work will focus on quantifying spatial relationships and patterns. This includes characterizing the spatial distribution of all cell types found in DSRCT, including fibroblasts, endothelial cells, and immune cells. We will also address whether tumor nest characteristics are associated with cell phenotype or DSRCT subtype.

D. Truong, S. Krishnan, C. Agyemang, D. Ingram, R. Lascano, A. Basi, J. Gomez, J. Burks, A. Lazar, J. Ludwig;
UT MD Anderson Cancer Center, Houston, TX

Solid tumor complexity extends beyond the genetic and functional landscapes of heterogenous cancer cells, encompassing the tumor microenvironment (TME). Elucidating the TME’s complexity requires a comprehensive assessment of its cellular composition, functional states, and spatial distributions. We developed the Harmonic Output of Stromal Traits (HOST) to identify TME cells, and the HOST-Factor to quantify their functional states. The HOST-Factor is a numerical value that reflects the relative contribution of cancer-associated fibroblasts (CAFs) to tumor-suppressive or tumor-promoting functions.
Our workflow combines automated cycling highplex immunofluorescent microscopy with artificial intelligence (AI)-guided image analysis. This generates HOST-Factor values for each identified TME cell within selected regions of interest, providing spatial distribution data. The TME signature encompasses 15 immune cells and 14 CAF antibody-detected biomarkers.
We applied our workflow to ten human pancreatic cancer specimens, generating OME-TIFF output images. This cancer model was used due to its significant TME makeup. The 29 highplex AI-based digital image analysis was conducted using the Phenoplex™ workflow from Visiopharm. The workflow included deep-learning-based tissue morphologic and cellular segmentations, cellular phenotyping, and integration of spatial/location data. Biomarker subsets were visualized, and a user-trained algorithm was used for tissue segmentation. Nuclear segmentation was done using a pre-trained algorithm on a DAPI-labeled DNA channel. Cellular phenotyping was performed using thresholds based on visual assessment and confirmation of positivity. Spatial neighborhood plots and metrics, heatmaps and partitioned t-SNE plots were generated for the dataset for downstream analysis. Importantly, the workflow’s visualization templates, pre-trained nuclear/cytoplasm segmentation tools, and neighborhood plots and metrics, are reusable and fully customizable for new datasets.
Using HOST-Factor values, we successfully classified cancer and TME cells, along with their functional states, and spatial distributions.
This AI-based computational approach and user-friendly workflow provides a simple and effective way to obtain single-cell information from multiplex immunofluorescence images, making spatial phenotyping of several cell populations in tissues more accessible to researchers, providing a fully amendable means for future clinical translation.

J. Franco-Barraza¹, F. Schneider², R. Norré-Sørensen², R. Ahrenkiel-Lyngby², C. Brown¹, D. Winkowski², J. R. Mansfield², E. Cukierman¹;
¹Fox Chase Cancer Center/Lewis Katz School of Medicine, Philadelphia, PA, ²Visiopharm, Horsholm, Denmark