An immunometabolic companion biomarker to enhance FDG-PET interpretation and guide frontline therapy in follicular lymphoma

Abstract

Abstract Introduction: Bendamustine-based immunochemotherapy (ICT) is a standard frontline regimen for follicular lymphoma (FL) but has limited efficacy in high-grade lymphomas characterised by rapid B cell proliferation. In clinical practice, Standardized Uptake Value (SUVmax) on pre-treatment PET is often interpreted as a surrogate for proliferative activity to guide the use of bendamustine, although evidence for this practice is conflicting (Mir et al, Blood, 2020). The inconsistent prognostic impact of pre-treatment SUVmax in FL may be explained by ex vivo studies demonstrating that the SUV reflects glucose uptake, not only of malignant B cells, but also by intratumoral T cells (Nath et al, Blood Adv, 2021). Thus, there is a need for an ‘immunometabolic’ companion biomarker to refine PET interpretation and identify FL subsets that may benefit from alternative ICT to bendamustine. Methods: A retrospective, real-world cohort of treatment-naive FL patients was assembled from 13 academic institutions across Australia. Baseline PET indices (Standardized Uptake Value [SUVmax], Total Metabolic Tumour Volume [TMTV]) were reported centrally and clinical and laboratory characteristics collected. Targeted next-generation sequencing (NGS) and bulk RNA sequencing (RNAseq) were performed on diagnostic nodal specimens. Immune cell proportions within the lymphoma microenvironment (LME) were quantified using RNASeq cell deconvolution and LME subtyping performed. B cell functional gene expression signatures (fGES) were curated using non-overlapping gene signatures expressed in purified B cell populations (centroblast “CB”, centrocyte “CC”, memory B cell “MBC”). Log-rank test was used to assess prognostic significance of variables. Spatial proteomic analysis used the Akoya PhenoCycler-Fusion platform, applying a custom 55-protein immunometabolic panel including markers of glucose uptake (GLUT1). Cellular phenotypes were determined using fluorescent marker profiles. Community analysis was performed using a graph neural network, integrating cell types with spatial architecture. Results: 129 bendamustine ICT treated patients were included between 2003 and 2021. Median follow-up was 46.4 months. Median age was 61 years and 62 (48%) had high-risk FLIPI. The median PFS was 136.8 months and 14 (11%) experienced POD24 events. Progression free survival (PFS) did not differ based on clinical PET indices, SUVmax >12 (43%, p=0.59) or TMTV >510cm3 (27%, p=0.069). A high centroblast (CBhigh) score (59.4%), was a strong predictor of inferior PFS (HR 2.51 95% CI 1.29–4.87; p <0.01) and POD24/histological transformation (HT) events (OR 4.03, p=0.03). The sensitivity and specificity for HT following bendamustine ICT was 88.9% and 63.9% respectively. CBhighwas associated with higher Ki67 (30% vs 15%, p<0.001) and grade 3a FL (p=0.014) on histopathology. The CBhigh subset was associated with a depleted LME (p<0.001) including reduction in CD4 helper (p<0.001), follicular helper (p<0.001) and CD8 T cells (p=0.03). There was no association with individual mutations or translocations by NGS. LME subtyping and CC and MBC fGES were not significantly associated with PFS. On multivariate analysis adjusting for grade, Ki67, and SUVmax, CBhigh remained significant for PFS (HR 3.31, p=0.012). A combined model of CBhigh and SUVmax further improved prognostic accuracy (p=0.005). Specifically in patients with SUVmax>12, CBhigh was associated with significantly shorter PFS than the CBlow group, in whom no early treatment failure events occurred(4yr PFS 48% vs 100%, p<0.001). Spatial analysis identified 20 unique cellular communities. Concordant with the RNASeq findings, communities enriched in FL B cells within intrafollicular (p<0.05) and marginal zone (p<0.01) regions were more frequent in high SUVmax cases. In these communities, FL B cells demonstrated increased GLUT1 expression (p<0.001) and Ki67 (p=0.014) compared with low SUVmax cases, consistent with high glucose utilisation and proliferation. Conclusion: While often applied in clinical practice, pre-treatment SUVmax alone is an unreliable predictor of response to bendamustine ICT. We demonstrate that a CBhigh FL subset, marked by proliferative malignant B cells and a depleted LME, is highly prognostic in bendamustine ICT treated FL patients. This tissue-based biomarker is independent of and additive to SUVmax, and should be tested prospectively for its ability to inform personalised frontline treatment decisions in FL.