9P - Identification of a μCT-based radiomic signature of CD8+ tumour infiltrating lymphocytes in an orthotopic murine model

Background

In the era of immunotherapy (IT), radiomics emerged as a non-invasive tool to decode tumor immune microenvironment (TIME) and predict IT response. Intrinsic patient and tumor variability challenging the explainability of radiomic readouts might be overcome in preclinical settings. Thus, we aimed to develop a μCT radiomic platform in a CD8 T cell functionally manipulated orthotopic murine model, closely mimicking human head and neck cancer, to validate the biological reality of radiomics in assessing tumor infiltrating lymphocytes (TILs).

Methods

Set-up, training and validation experiments (N Tot = 50 C57BL/6 mice) were conducted as follows: submucosal injection of 0.5 x 10⁶ TC1-luc cells in the right inner lip and generation of an orthotopic model of head and neck cancer (day [D] -7); tailored treatment consisting of irradiation [D0] ± anti-CD8 antibody [D3] to selectively enrich or deplete CD8+ TILs; in vivo preclinical imaging through Quantum FX μCT technology [D3, D4]; mice euthanasia and tumor sampling [D4], followed by immunohistochemical staining for CD8; μCT image pre-processing (voxel resampling, image discretization and delineation of the volume of interest) and radiomic features (RFs) extraction (pyradiomics).

Results

We developed and optimized an efficient quantitative μCT imaging approach. Overall, 106 μCT RFs (shape, first- and second- order) were extracted and correlated with distinct TIME. Following redundant feature elimination (Spearman correlation, cut-off = 0.99) and Z-score standardization, we identified 12 μCT-RFs differentially regulated in T CD8 enriched vs depleted TIME (P < 0.05, Mann Whitney). The statistical performance of our CD8 signature, documented in both training and validation sets, was further implemented when pooled data were analysed (P < 0.01). Finally, by applying a reconstruction algorithm, we obtained a 3D map of CD8+ TILs, which enabled us to correlate the spatial distribution of immune cells with μCT textural parameters.

Conclusions

Our results document the feasibility and accuracy of radiomics to detect dramatic changes in T cells within the TIME, thus providing effective radio-immune signatures potentially translatable into clinical practice.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Institut Gustave Roussy.

Funding

ESMO (Translational Research Fellowship; Recipient: Dr. Giulia Mazzaschi).

Disclosure

All authors have declared no conflicts of interest.