Cell Profiling Coupled with TCR Clonotype Characterization: Discovery of Predictive Biomarkers of Response to T Cell

Summary

The discovery of predictive biomarkers of drug response is critical for forecasting patient benefit from novel immune-modulatory therapeutics. However, such discovery is challenging due to the heterogeneous nature of the tumor microenvironment (TME) as well as the lack of an approach to analyze drug-responsive immune cells that impact tumor progression.

We developed a novel Drug Intelligent Science (DIS™) approach to define biomarkers of response by subjecting dissociated human tumor tissues to drug treatments, followed by single-cell transcriptomic profiling and TCR clonotype characterization. Responding T cells are identified as those showing shifts in gene expression consistent with known T cell activation signatures. TCR clonotypes can be used to match T cells in treatment conditions to their sister clones in the baseline state. Comparing the baseline gene expression profiles between T cell clonotypes that responded to the treatment and those that did not makes possible the discovery of gene expression signatures that predict response (Lee et al.).

We processed 17 tumor samples obtained from cancer patients and treated them ex vivo with two novel biologics currently under clinical development - HFB301001, a potentially best-in-class $2^{th}$ generation X40 agonist, and HFB200301, a first-in-class TNFR2 agonist. We applied our biomarker discovery strategy to the integrated scRNA-seq and scTCR-seq data from these samples to define predictive signatures of response to these drugs. To further validate this strategy, we also generated single-cell data with our ex vivo culture system to characterize response to anti-PD-1 treatment. In the anti-PD-1 predictive response signature, we identified genes involved in inflammatory response and genes in the pathway of other co-inhibitory checkpoints. Application of the anti-PD-1 predictive response signature to bulk transcriptomic data from clinical studies with checkpoint inhibitors successfully stratified patients into two groups with significantly different risks of progression. The predictive response signatures for the two novel agonistic antibodies shared genes involved in inflammatory pathways with the anti-PD-1 signature, but also contained other distinct gene sets.

Ex vivo single-cell profiling coupled with TCR clonotype characterization enables the discovery of predictive response signatures that inform patient selection strategies for the early clinical development of novel therapeutics. These results suggest the potential for additional patient populations that may respond to these treatments.

Experimental Setup

Tumor Samples from Surgical Resections

• Primary tumor samples obtained from Discovery Life Sciences and the Cooperative Human Tissue Network.
• A total of 17 tumor samples (8 RCC, 7 NSCLC, 1 MEL, and 1 UC) were selected to represent key indications for HFB200301 and HFB301001.

scRNA-seq, scTCR-seq

• Single-cell suspensions from dissociated tumor samples.
• Baseline flow cytometry characterization of immune cell composition and target expression.
• Ex vivo culture of all cells, including tumor cells, without exogenous TCR cross-linking.
• 24h ex vivo treatment with controls (isotype, $$CD3, $$PD1) and drug candidates (HFB200301 and HFB301001).
• HiFiBiO DIST *scRNA-seq platform: 156,522 single cells.

Computational Analysis Workflow

• Pipeline for read mapping and clonotype calling.
• Data integration and batch correction using Variational Autoencoder (Lopez et al.).
• Automatic cell type classification based on transfer learning (Lotfollahi et al.) from internal reference atlas.

Results

Single-cell Profiling of T Cells from Ex Vivo Culture

Response of Single T Cells to Treatment

Responding cells are CD8+ T cells characterized by an activation signature and coincide with large TCR clonotypes.

Figure 2. Characterization of T cells that responded to treatment ex vivo. a, UMAP embedding of scRNA-seq transcriptomes from T cells as seen in Figure 1b. Leiden clusters identified. Clusters 3, 4, 10, 14, 16 and 18 correspond to T cells and are circled in red. b, Box plot showing the expression of an activation gene signature, computed as an activation score for each Leiden cluster. The activation score is a composite of published signatures that define primary human T cells cultured in culture by αCD3 and αCD28 (Sabat et al.). Clusters with median activation score above that of primary human T cells are colored red. c, Box plots showing the relative distribution of various sizes are colored red to show their relative distribution. Activated the UMAP embedding. Clonotype size is represented as a percentage of the total number of T cells with TCR read from the originating sample. Activated the T cell clusters are circled in red and coincide with large T clonotypes.

TCR Clonotype Barcode Analysis Defines αPD1, HFB200301, and HFB301001 Biomarkers

Figure 3. Predictive biomarkers of response to αPD1, HFB200301, and HFB301001. a, UMAP embedding of T cells as seen in Figure 1b. Responsive (R) and non-responsive (NR) T cells at baseline for αPD1, HFB200301, and HFB301001 are colored red to show their relative distribution on the UMAP embedding. Heatmaps showing the log-normalized and z-score expression of genes (columns) from the αPD1, HFB200301, and HFB301001 biomarkers from differential gene expression analysis of R versus NR cells (rows). c, Venn diagram showing the number of genes shared between the three biomarkers.

Validation of Predictive Biomarker of Response to αPD1

αPD1 Biomarker Derived from Ex Vivo Culture Predicts ICB-specific Patient Survival in Clinical Trials