A Vision Transformer-Based Framework for Knowledge Transfer From Multi-Modal to Mono-Modal Lymphoma Subtyping Models.

Determining lymphoma subtypes is a crucial step for better patient treatment targeting to potentially increase their survival chances. In this context, the existing gold standard diagnosis method, which relies on gene expression technology, is highly expensive and time-consuming, making it less accessibility. Although alternative diagnosis methods based on IHC (immunohistochemistry) technologies exist (recommended by the WHO), they still suffer from similar limitations and are less accurate. Whole Slide Image (WSI) analysis using deep learning models has shown promising potential for cancer diagnosis, that could offer cost-effective and faster alternatives to existing methods. In this work, we propose a vision transformer-based framework for distinguishing DLBCL (Diffuse Large B-Cell Lymphoma) cancer subtypes from high-resolution WSIs. To this end, we introduce a multi-modal architecture to train a classifier model from various WSI modalities. We then leverage this model through a knowledge distillation process to efficiently guide the learning of a mono-modal classifier. Our experimental study conducted on a lymphoma dataset of 157 patients shows the promising performance of our mono-modal classification model, outperforming six recent state-of-the-art methods. In addition, the power-law curve, estimated on our experimental data, suggests that with more training data from a reasonable number of additional patients, our model could achieve competitive diagnosis accuracy with IHC technologies. Furthermore, the efficiency of our framework is confirmed through an additional experimental study on an external breast cancer dataset (BCI dataset).

Performance of non-invasive tests for liver fibrosis resolution after bariatric surgery.

BACKGROUND & AIMS: The value of non-invasive tests for monitoring the resolution of significant liver fibrosis after treatment is poorly investigated. We compared the performances of six non-invasive tests to predict the resolution of significant fibrosis after bariatric surgery. METHODS: Participants were individuals with obesity submitted to needle liver biopsy at the time of bariatric surgery, and 12 and/or 60 months after surgery. We calculated the fibrosis-4 index (FIB-4), NAFLD fibrosis score (NFS), AST to platelet ratio index (APRI), Hepatic fibrosis score (HFS), Fibrotic NASH index (FNI), and Liver risk score (LRS) at each time point, and compared their performances for predicting significant fibrosis (F ≥ 2) and its resolution following surgery. RESULTS: At baseline, 2436 patients had liver biopsy, including 261 (10.7 %) with significant fibrosis. Overall, 672 patients had pre- and post-operative biopsies (564 at M12 and 328 at M60). The fibrosis stage decreased at M12 and M60 (p < 0.001 vs M0). Resolution of significant fibrosis occurred in 58/121 (47.9 %) at M12 and 32/50 (64 %) at M60. The mean value of all tests decreased after surgery, except for FIB-4. Performances for predicting fibrosis resolution was higher at M60 than at M12 for all tests, and maximal at M60 for FNI and LRS: area under the curve 0.843 (95%CI 0.71-0.95) and 0.92 (95%CI 0.84-1.00); positive likelihood ratio 3.75 (95 % CI 1.33-10.59) and 4.58 (95 % CI 1.65-12.70), respectively. CONCLUSIONS: Results showed the value and limits of non-invasive tests for monitoring the evolution of liver fibrosis after an intervention. Following bariatric surgery, the best performances to predict the resolution of significant fibrosis were observed at M60 with tests combining liver and metabolic traits, namely FNI and LRS.

Time-of-day-dependent variation of the human liver transcriptome and metabolome is disrupted in MASLD.

Background & Aims: Liver homeostasis is ensured in part by time-of-day-dependent processes, many of them being paced by the molecular circadian clock. Liver functions are compromised in metabolic dysfunction-associated steatotic liver disease (MASLD) and metabolic dysfunction-associated steatohepatitis (MASH), and clock disruption increases susceptibility to MASLD progression in rodent models. We therefore investigated whether the time-of-day-dependent transcriptome and metabolome are significantly altered in human steatotic and MASH livers. Methods: Liver biopsies, collected within an 8 h-window from a carefully phenotyped cohort of 290 patients and histologically diagnosed to be either normal, steatotic or MASH hepatic tissues, were analyzed by RNA sequencing and unbiased metabolomic approaches. Time-of-day-dependent gene expression patterns and metabolomes were identified and compared between histologically normal, steatotic and MASH livers. Results: Herein, we provide a first-of-its-kind report of a daytime-resolved human liver transcriptome-metabolome and associated alterations in MASLD. Transcriptomic analysis showed a robustness of core molecular clock components in steatotic and MASH livers. It also revealed stage-specific, time-of-day-dependent alterations of hundreds of transcripts involved in cell-to-cell communication, intracellular signaling and metabolism. Similarly, rhythmic amino acid and lipid metabolomes were affected in pathological livers. Both TNFα and PPARγ signaling were predicted as important contributors to altered rhythmicity. Conclusion: MASLD progression to MASH perturbs time-of-day-dependent processes in human livers, while the differential expression of core molecular clock components is maintained. Impact and implications: This work characterizes the rhythmic patterns of the transcriptome and metabolome in the human liver. Using a cohort of well-phenotyped patients (n = 290) for whom the time-of-day at biopsy collection was known, we show that time-of-day variations observed in histologically normal livers are gradually perturbed in liver steatosis and metabolic dysfunction-associated steatohepatitis. Importantly, these observations, albeit obtained across a restricted time window, provide further support for preclinical studies demonstrating alterations of rhythmic patterns in diseased livers. On a practical note, this study indicates the importance of considering time-of-day as a critical biological variable which may significantly affect data interpretation in animal and human studies of liver diseases.

A novel anti-galectin-9 immunotherapy limits the early progression of pancreatic neoplastic lesions in transgenic mice.

Background: Pancreatic adenocarcinoma (PDAC) is a devastating disease with an urgent need for therapeutic innovation. Immune checkpoint inhibition has shown promise in a variety of solid tumors, but most clinical trials have failed to demonstrate clinical efficacy in PDAC. This low efficacy is partly explained by a highly immunosuppressive microenvironment, which dampens anti-tumor immunity through the recruitment or induction of immunosuppressive cells, particularly regulatory T cells (Tregs). In this context, our laboratory has developed a novel immunotherapeutic strategy aimed at inhibiting the suppressive activity of Tregs, based on a patented (EP3152234B1) monoclonal antibody (mAb) targeting galectin-9 (LGALS9). Materials and methods: CD4+ conventional T cells (TCD4 or Tconv), Treg ratio, and LGALS9 expression were analyzed by immunohistochemistry (IHC) and cytometry in blood and pancreas of K-rasLSL.G12D/+;Pdx-1-Cre (KC) and K-rasWildType (WT);Pdx1-Cre (WT) mice aged 4-13 months. Pancreatic intraepithelial neoplasm (PanIN) progression and grade were quantified using FIJI software and validated by pathologists. The anti-galectin-9 mAb was validated for its use in mice on isolated murine C57BL/6 Treg by immunofluorescence staining and cytometry. Its specificity and functionality were validated in proliferation assays on rLGALS9-immunosuppressed murine Tconv and in suppression assays between murine Treg and Tconv. Finally, 2-month-old KC mice were treated with anti-LGALS9 and compared to WT mice for peripheral and infiltrating TCD4, Treg, and PanIN progression. Results: IHC and cytometry revealed a significant increase in LGALS9 expression and Treg levels in the blood and pancreas of KC mice proportional to the stages of precancerous lesions. Although present in WT mice, LGALS9 is expressed at a basal level with low and restricted expression that increases slightly over time, while Treg cells are few in number in their circulation and even absent from the pancreas over time. Using our anti-LGALS9 mAb in mice, it is shown that (i) murine Treg express LGALS9, (ii) the mAb could target and inhibit recombinant murine LGALS9, and (iii) neutralize murine Treg suppressive activity. Finally, the anti-LGALS9 mAb in KC mice reduced (i) LGALS9 expression in pancreatic cancer cells, (ii) the Treg ratio, and (iii) the total surface area and grade of PanIN. Conclusion: We demonstrate for the first time that an anti-LGALS9 antibody, by specifically targeting endogenous LGALS9 tumor and exogenous LGALS9 produced by Treg, was able to limit the progression of pancreatic neoplastic lesions in mice, opening up new prospects for its use as an immunotherapeutic tool in PDAC.