Introduction

SOR102 is a novel, orally delivered, bispecific antibody construct that neutralizes TNFα and IL-23 activity within GI tissue, targeting the site of inflammation and potentially limiting systemic exposure and immunosuppression¹. Systemic TNFα/IL-23 inhibition recently demonstrated encouraging clinical efficacy in patients with ulcerative colitis² suggesting additive benefit of dual TNFα/IL-23 inhibition, observed as early as 4 weeks of dosing³, although safety and efficacy of long-term systemic combination therapy is unknown. SOR102 contains two humanized single domain antibodies (SDA) targeting TNFα and IL-23p19 connected by a trypsin-labile linker, enabling monomer separation within the small intestine. SOR102 inhibits TNFα and IL-23 activity pre- and post-trypsin cleavage. SOR102 SDAs were engineered for high stability among intestinal and inflammatory proteases to enable oral dosing. This Phase 1, first-in-human study (SOR102-101; NCT06080048) consists of 3 parts. Parts 1 and 2 enrolled healthy subjects (HS) and evaluated single ascending (SAD) and multiple doses (MD) of SOR102, respectively.

Aims & Methods

SOR102-101, Parts 1 and 2, was a randomized, double-blind, placebo-controlled, SAD and MD study in HS aged 18 to 55 years that evaluated the safety and tolerability of SAD (Part 1) and MD (Part 2) of SOR102, the concentration of intact SOR102 and its monomers (SOR102-TNFα and SOR102-IL-23) in serum and faeces (Parts 1 and 2) and urine (Part 2), and the incidence of anti-drug antibodies (ADA) (Part 2). In Part 1, HS were randomized 6:2 to active and placebo treatment into one of 4 ascending dose cohorts. HS were administered a single oral (PO) SOR102 dose ranging approximately between 100mg and 4000mg, or placebo. In Part 2, HS were randomized 8:2 to active (approximately 1200mg PO BID) or placebo (PO BID) administered for 7 days. Safety and tolerability, faecal, serum and urine concentrations of SOR102 and its monomers, and ADA levels were assessed at predetermined times.

Results

All 42 enrolled participants completed the study. Seven adverse events (AE) were observed, all occurring in Part 1. AE severity was mild or moderate. There was one treatment-related AE (loose stools) of mild intensity and 17 hours duration occurring in the lowest dose cohort.
SOR102 and its monomers were not detected in serum. In Part 1, faecal concentrations of active (via antigen-binding ELISA) monomers were detected at all dose levels, peaking on Day 2 and 3 post-dose, with generally higher concentrations observed in higher-dose cohorts. The highest faecal concentrations were observed in the third dose cohort on Day 2 post-dose (SOR102-TNFα ≈ 260 mcg/mL; SOR102-IL-23 ≈ 130 mcg/mL). In Part 2, active monomer faecal concentrations were observed during all dosing days; the highest faecal concentrations observed were SOR102-TNFα ≈ 1,700 mcg/mL and SOR102-IL-23 ≈ 5,400 mcg/mL. In Parts 1 and 2, intact SOR102 faecal concentrations were on average ≥ 800-fold lower than monomer concentrations.

Conclusion

SOR102 was well tolerated at all tested doses. Systemic exposure of SOR102 and its monomers was not observed, supporting its gut selectivity. Low levels of intact SOR102 and high levels of SOR102 monomers in faeces confirm that SOR102 is efficiently cleaved after oral dosing, as designed. Monomer faecal concentrations were observed at all doses in Part 1 and generally increased with each ascending dose, peaking on Day 2 and 3 after administration. In Part 2, monomer faecal concentrations were observed across all dosing days, peaking at >1700 mcg/mL.

References

1. Roberts KJ et al. Preclinical development of a bispecific TNFα/IL-23 neutralising domain antibody as a novel oral treatment for inflammatory bowel disease. Sci Rep 2021;11:19422.
2. Feagan BG et al. Guselkumab plus golimumab combination therapy versus guselkumab or golimumab monotherapy in patients with ulcerative colitis (VEGA): a randomised, double-blind, controlled, phase 2, proof-of-concept trial. Lancet Gastroenterol Hepatol 2023;8:307-320.
3. Richards D et al. Guselkumab and golimumab combination induction therapy in ulcerative colitis results in early local tissue healing that is sustained through guselkumab maintenance therapy. J Crohns Colitis 2024;18(Supplement_1):i171–i172.

Disclosure

VJ has received has received consulting/advisory board fees from AbbVie, Alimentiv, Arena pharmaceuticals, Asahi Kasei Pharma, Asieris, Astra Zeneca, Avoro Capital, Bristol Myers Squibb, Celltrion, Eli Lilly, Endpoint Health, Enthera, Ferring, Flagship Pioneering, Fresenius Kabi, Galapagos, Gilde Healthcare, GlaxoSmithKline, Genentech, Gilead, Innomar, JAMP, Janssen, Merck, Metacrine, Mylan, MRM Health, Pandion, Pendopharm, Pfizer, Protagonist, Prometheus Biosciences, Reistone Biopharma, Roche, Roivant, Sandoz, Second Genome, Sorriso, Synedgen, Takeda, TD Securities, Teva, Topivert, Ventyx, Vividion; speaker’s fees from, Abbvie, Ferring, Bristol Myers Squibb, Galapagos, Janssen Pfizer Shire, Takeda, Fresenius Kabi
CS and PW are employees of Sorriso Pharmaceuticals
EW and CG are employees of MAC Clinical Research

Introduction

Deep learning systems in gastrointestinal endoscopy are traditionally pretrained on large, diverse datasets comprising generic, non-medical images before being trained on scarce endoscopic data. Foundation models are models pretrained based on domain-specific datasets. For endoscopy, foundation models require large datasets of general endoscopic images. Yet datasets for developing such models remain limited. In this study, we present GastroNet-5M, a dataset comprising 4,820,653 endoscopic images of approximately 500,000 procedures.

Aims & Methods

GastroNet-5M is composed of anonymized endoscopic images collected from eight Dutch hospitals between 2012 and 2020. Using a self-supervised learning approach, this dataset enabled the development of a foundation model for various endoscopic AI applications. In this study, we compared GastroNet-5M-pretrained models against current state-of-the-art models with traditional pretraining approaches and all publicly available endoscopic foundation models.
We evaluated diagnostic performance on 9 classification and 9 segmentation tasks, including Barrett’s neoplasia detection, colorectal polyp characterization, and gastric cancer invasion depth prediction. All models were trained from scratch for each task. Data efficiency was assessed by repeating all classification tasks with progressively smaller training sets (i.e. 100%, 50%, 25%, 10% and 5%). Robustness was evaluated across 12 datasets reflecting data heterogeneity such as variation in endoscope manufacturer, imaging modality (e.g., virtual chromoendoscopy), and image quality.

Results

GastroNet-5M-pretrained models outperformed all benchmarks in 8 of 9 classification tasks, with an average AUC improvement of 6.8% (Table 1). For segmentation, they achieved higher Dice scores compared to all other models across all 9 tasks, with an average gain of 15.4%. In data efficiency experiments, GastroNet-5M models achieved the highest AUC in 5 of 9 tasks when training data was limited, while no other model led in more than one. In robustness testing, they outperformed other models in 8 of 12 tasks evaluating heterogeneous data conditions.

Table 1. Area-Under-the-Curve scores of the GastroNet-5M-pretrained model and benchmark models for all classification tasks.

Conclusion

GastroNet-5M, a multicenter dataset of over 5 million unlabeled endoscopic images, offers a valuable resource for pretraining deep learning models in endoscopy. The use of GastroNet-5M enhances diagnostic accuracy, reduces the need for scarce application-specific endoscopic imagery, and improves robustness of endoscopic AI systems to the inevitable data heterogeneity in clinical practice. GastroNet-5M is publicly accessible for further research and development.

References

1. Sanderson E, Matuszewski BJ. arXiv preprint arXiv:2401.06278. 2024.
2. van der Putten J, de Groof J, Struyvenberg M, et al. Artif Intell Med. 2020;107.
3. Wang Z, Liu C, Zhang S, et al. arXiv preprint arXiv:2306.16741. 2023.
4. Liu Z, Mao H, Wu C-Y, et al. arXiv preprint arXiv:2201.03545. 2022.
5. Dosovitskiy A, Beyer L, Kolesnikov A, et al. arXiv preprint arXiv:2010.11929. 2020.
6. Oquab M, Darcet T, Moutakanni T, et al. arXiv preprint arXiv:2304.07193. 2023.
7. Jocher G, Qiu J. Ultralytics YOLOv11, v11.0.0. 2024.
8. de Groof AJ, Struyvenberg MR, van der Putten J, et al. Gastroenterology. 2020;158(4):915–29.e4.
9. Fockens KN, Jukema JB, Boers T, et al. United European Gastroenterol J. 2023;11(4):324–36.
10. Jukema J, Kusters C, Jong M, et al. Endoscopy (ESGE Days). 2023.
11. GIANA Challenge. Available from: https://giana.grand-challenge.org
12. Borgli H, Thambawita V, Smedsrud PH, et al. Sci Data. 2020;7(1):283.
13. Polat G, Kani HT, Ergenc I, et al. Inflamm Bowel Dis. 2023;29(9):1431–9.
14. van der Zander QEW, Schreuder RM, Fonollà R, et al. Endoscopy. 2021.
15. Tang D, Zhou J, Wang L, et al. Front Oncol. 2021;11:654093.
16. Houwen BB, Hazewinkel Y, Giotis I, et al. Endoscopy. 2022;54(Suppl 1):eP133.

Introduction

Microscopic colitis (MC) refers to two pathologies – Collagenous colitis (CC) and Lymphocytic colitis (LC) that cause chronic watery diarrhoea, urgency, incontinence leading to significantly decreased quality of life¹. It is thought that MC develops due to a pathological immune response to intestinal luminal antigens in genetically predisposed individuals, with microbiota playing an important role^{2, 3}. However, the pathogenesis of MC is still unclear.

Aims & Methods

The aim of this study was to identify and characterize changes in bacterial composition of MC patients in terminal ileum and different colonic segments and determine its relation to the clinical course of disease. Biopsies from terminal ileum and six segments of the colon were collected during diagnostic colonoscopies. DNA extraction was performed using AllPrep DNA/RNA Micro Kit (Qiagen) following the manufacturer’s instructions. To determine bacterial composition, sequencing of the 16S rRNA variable region V3-V4 was conducted using the 27F and 338R polymerase chain reaction primers and sequenced on a MiSeq (2×250 bp; Illumina, Hayward, CA). Bioinformatic analysis was performed using DADA2, and taxonomic classification was carried out with the RDP classifier.

Results

Biopsy samples from 20 active CC patients, 20 active LC patients, and 50 controls demonstrated distinct microbial profiles across different intestinal segments, as revealed by beta-diversity and differential compositional analyses. Compared with controls, CC patients had consistent alterations in three bacterial families throughout the entire colon, with Erysipelatoclostridiaceae and Coriobacteriaceae decreased and Pasteurellaceae increased. Additionally, 35 families had segment-specific changes, and 25 families displayed altered abundance in the terminal ileum.
LC patients exhibited a consistent increase in Erisypelotrichaceae across the entire colon, with 38 other families showing segmental changes, and 12 families displaying altered abundance in the terminal ileum. Comparing LC and CC revealed a significantly different bacterial profile. LC showed lower abundance across many families, particularly in the right colon and terminal ileum. After sample collection, 15 CC patients and 12 LC patients received budesonide, with 3 CC and 2 LC patients having incomplete response. Beta diversity analysis revealed significant differences in the ascending colon and left side of the colon of CC patients based on response to budesonide. Patients with incomplete response to budesonide had an increased abundance of the genus Paucibacter. Similar tendencies were not observed in LC.
1 CC patient exhibited a quiescent disease course, 6 reached a sustained remission after treatment, 7 had a relapsing and 6 had a chronic active disease. Significant differences in beta diversity were observed in transverse, descending colon and rectum. Beta diversity of the descending colon had a medium correlation with disease behaviour. In the LC group 6 patients had a quiescent disease, 8 had a sustained remission after treatment, 3 exhibited a relapsing and 4 had a chronic active disease. No significant correlation between microbiota and disease behaviour was observed in LC.

Conclusion

This study showed significant differences in bacterial composition between CC, LC and controls in the terminal ileum and colon. In both CC and LC, bacterial alterations were not consistent throughout the entire colon but varied between different colonic segments. Microbiome differences based on response to budesonide and disease behaviour were observed in CC, but not LC.

References

1. Verhaegh BPM, Münch A, Guagnozzi D, et al. Course of Disease in Patients with Microscopic Colitis: A European Prospective Incident Cohort Study. J Crohns Colitis. 2021;15(7):1174-1183.
2. Miehlke S, Guagnozzi D, Zabana Y, et al. European guidelines on microscopic colitis: United European Gastroenterology (UEG) and European Microscopic Colitis Group (EMCG) statements and recommendations. United European Gastroenterol J. Published online 2020.
3. Burke KE, D’Amato M, Ng SC, Pardi DS, Ludvigsson JF, Khalili H. Microscopic colitis. Nat Rev Dis Primers. 2021;7(1).

Introduction

The role of the gut microbiota in inflammatory bowel disease (IBD) pathogenesis is widely acknowledged, with numerous bacteria identified as contributing factors. Little however is known about the part viruses play – a likely major component in the structuring and functioning of the gut microbiota. Understanding how viruses interact with the gut microbiota and host cells in the context of IBD could facilitate innovative treatment approaches and personalized interventions.

Aims & Methods

We explore the colonic and upper intestinal virome in patients with IBD compared with general population (GP) controls. Faecal (colonic) and effluent (upper intestinal) derived shotgun metaviromes of 100 IBD individuals (UC = 50, CD = 50) and 50 IBD individuals with an ileostomy, respectively were compared to faecal derived metaviromes of 50 GP individuals. Metaviromes were generated using a combination of shotgun methods and viral particle enrichment, allowing for identification of both eukaryotic and prokaryotic, dsDNA, ssDNA and RNA viruses. Viral diversity, composition and prevalence were compared by calculating Shannon diversity, Bray-Curtis distances and performing logistic regression analyses. P-values were adjusted using false discovery rate (FDR) to correct for multiple testing, where relevant; FDR<0.05 was considered significant.

Results

Similar to bacterial diversity in IBD versus GP samples, we see a significantly lower viral diversity in CD and UC colonic samples compared with the GP, and an even lower diversity when comparing upper intestinal samples to colonic samples (FDR<0.05). Comparing viral populations between groups, we observe that, whilst there is evident interindividual variability within groups, distinct clustering between colonic and upper intestinal samples can be seen, with colonic IBD samples clustering slightly more towards upper intestinal samples. When comparing viral group prevalences between GP and IBD colonic samples, 83 out of 260 viral groups present in IBD-UC samples are absent in the GP samples and 125 out of 302 viral groups present in the GP were absent in IBD-UC. For IBD-CD samples these numbers are 79 out of 248 and 133 out of 302, respectively. The top 3 viral groups that positively associated with IBD-UC samples are: one unknown phage group (predicted bacterial host: Parasutterella spp.), a group containing Lactobacillus phages (bacterial host: Lactobacillus spp.) and a group containing Lactococcus phages (bacterial host: primarily Lactococcus spp). The top 3 viral groups that positively associated with IBD-CD samples are: the same unknown phage group as for IBD-UC, and two other unknown phage groups (predicted bacterial hosts: primarily Enterocloster & Bacteroides spp, respectively).

Conclusion

This first comprehensive description of the gut virome across multiple gastrointestinal sites shows specific alterations in IBD. Many viral groups differentially present in the IBD intestine infect bacteria often associated with gut health and significantly reduced in IBD patients. Dysregulated bile acid signaling has also been implicated in the pathogenesis of IBD, with bacterial dysbiosis being linked to its cause. In IBD populations, we observe phages with the host E.bolteae – a bacteria increased in dysbiosis due to their potential to metabolise bile acids. We speculate that alterations in the gut virome composition are key events that disrupt the equilibrium between gut microbiota and host, promoting a pro-inflammatory state and driving IBD pathogenesis. The next steps will be to investigate the mechanisms underlying these complex interactions.

Disclosure

RKW unrestricted grants: Takeda, Johnson and Johnson, Tramedico, Ferring. Consultant: Takeda