(P 061) UNRAVELLING THE HETEROGENEITY OF THE IMMUNE MICROENVIRONMENT IN STS USING QUANTITATIVE MRI

Objective: Quantitative imaging can interrogate the physiology of tumours both spatially and temporally with the potential to provide an assessment of the underlying biology via surrogate imaging biomarkers. Soft tissue sarcomas (STS) are notoriously heterogeneous and carry a dismal prognosis. STS located within the retroperitoneum can present late and carry a high propensity for local relapse. Combining a greater biological understanding of tumours, including their immune microenvironment, with powerful imaging tools may lead to the development of clinically applicable biomarkers to support risk stratification of patients and to allow improved individualised management. In a cohort of patients with retroperitoneal sarcoma, we aimed to characterise the immune microenvironment of tissue samples matched to regions of interest (ROIs) on quantitative MRI images and assess for correlation of immune cell infiltrate with quantitative MRI and histopathological parameters.

Methods: 22 patients with retroperitoneal sarcoma were included in this retrospective study, recruited as part of a previous prospective imaging study, NCT01902667. Subtypes were well-differentiated liposarcoma (WDLPS, n=7), dedifferentiated liposarcoma (DDLPS, n=11), leiomyosarcoma (LMS, n=3) and spindle cell sarcoma (SCS, n=1). Patients underwent quantitative MRI, 12 were treatment naïve and 10 had received radiotherapy. ROIs were selected in morphologically distinct areas on MRI prior to surgery and their quantitative MRI parameters calculated. These were enhancing fraction (EF, defined as percentage of enhancing voxels within a given ROI), fat fraction (FF, ratio of signal from fat image over signal from sum of fat and water images) and apparent diffusion coefficient (ADC). Tissue samples matched to the ROIs from the surgical excision sample were available as fixed formalin paraffin embedded (FFPE), with a total of 40 blocks from the 22 patients. Histopathological parameters (nuclear to stromal ratio, Ki67, fat content, vessel density) were scored at the time of the original study by a specialist sarcoma pathologist.

Single stain immunohistochemistry for CD3+, CD4+, CD8+, CD20+ and CD68+ was undertaken, followed by manual and automated counting of immune cell infiltrate. 10 corresponding high powered fields (HPF) were selected on each immune cell slide, counting undertaken and the average of the 10 HPFs calculated to generate an immune cell infiltrate per HPF for the entire tissue block. Correlations between average immune cell infiltrate per HPF and the quantitative MRI and histopathological parameters were assessed using Spearman’s rank correlation coefficient, and Spearman rank co-correlation matrix was constructed.

Results: Immune cell density exhibited variability on an inter- and intra-patient level, with differences in the number of immune cells per HPF between blocks. Within a tumour, the immune cell density between blocks ranged from 1 cell per HPF to 504 cells per HPF. The largest differences were observed for T cells (CD3+, CD4+ and CD8+). Of the 4 different subtypes, LMS had the highest median immune cell infiltrate for T cells with 260 CD3+ cells per HPF (range 15-364), 77 CD4+ cells per HPF (range 5-150) and 117 CD8+ cells per HPF (range 10-289). WDLPS had the highest quantity of CD20+ cells with an median of 4 cells per HPF (range 0-157) When CD20+ cells were present in high density, these consistently appeared as part of lymphoid aggregate structures and these were found in 9 of the 40 total blocks. Finally SCS had the highest quantity of CD68+ cells with a median of 270 cells per HPF (range 214-327), however, of note, there was only 1 case of SCS. MRI EF showed a significant positive correlation with T cells: CD3+ with ρ= 0.55 (p=0.0003), CD4+ with ρ=0.49 (p=0.0017) and CD8+ with ρ= 0.56 (p=0.0004). FF showed a significant inverse correlation with CD68+ cells with ρ= -0.50 (p=0.0180), this was not found for FF with any other immune cell type. Ki67 and nuclear to stromal ratio also demonstrated correlations with CD3+, CD4+, CD8+ and CD68+ cells. Spearman rho values for Ki67 and T cells were as follows: CD3+ ρ=0.50 (p- 0.0014), CD4+ ρ=0.54 (0.0005), CD8+ ρ=0.57 (p= 0.0002). For nuclear to stromal ratio and T cells, spearman rho values were as follows: CD3+ ρ=0.41 (p=0.009), CD4+ ρ=0.40 (0.0130), CD8+ ρ=0.47 (p=0.0033). For CD68+ cells, spearman rho values were 0.52 and 0.32 (p-values 0.0007 and 0.0500) respectively for Ki67 and nuclear to stromal ratio.

Conclusion: We have characterised the immune microenvironment in a cohort of patients with retroperitoneal sarcoma and have demonstrated a variability in immune cell infiltrate between tissue blocks for 4 subtypes. We hypothesise that the correlation of MRI EF with T cells may reflect areas of higher vascular perfusion with higher T cell infiltrate. This is a limited and retrospective cohort however results encourage further exploration in future prospective studies which are planned.

Figure 1.pdf