Background
Myxoid periductal stroma occurs in a subset of ductal carcinoma in situ (DCIS) of the breast. Previous studies identified myxoid stroma as a risk factor for earlier ipsilateral locoregional recurrence after surgery for DCIS. However, its pathogenesis is poorly understood.
Purpose
To explore the underlying mechanisms of myxoid stroma development in breast biopsies containing pure DCIS.
Methods
The tissue samples of the ASNoPro study cohort, comprising 262 biopsy-diagnosed DCIS, were subjected to multiplex immunofluorescence (IF), aiming to visualise twelve different proteins in the neoplastic cells, the extracellular matrix (ECM) and the tumour-infiltrating lymphocytes (TILs). The outcome was the presence or absence of myxoid stroma. Digital image analysis with QPath was performed to objectify IF protein expression.
Results
Eighty-eight DCIS (34%) presented with myxoid stroma, whereas 174 DCIS (66%) had a predominantly sclerotic stroma. Myxoid stroma was significantly associated with high-grade nuclear atypia (p<0.001), lack of calcifications (p=0.016), ER negativity (p<0.001), PR negativity (p<0.001), HER2 positivity (p<0.001), post-operative larger DCIS size (p<0.001), high TILs density (p<0.001), and p53 mutant-type staining (p<0.001). Myxoid stroma was significantly associated with increased stromal expression of versican (p<0.001), and decreased stromal expression of biglycan (p<0.001), decorin (p=0.034) and lumican (p<0.001). Myxoid stroma correlated with increased B-lymphocytes (p=0.028). CD8 and FoxP3 expression were not associated with myxoid stroma, nor were cyclin-D1, synaptophysin and EGFR.
Conclusions
DCIS with myxoid stroma present more often with ‘high-grade’ histopathological features. Modified ECM protein expression strongly correlates with the presence of myxoid stroma in DCIS. We hypothesize that the periductal stroma may indirectly reflect the aggressiveness of the neoplastic cells: some DCIS might be capable of preparing their microenvironment for invasion. Future in vitro studies should explore the precise underlying mechanisms to increase our understanding of the pathogenesis of myxoid stroma, and the transition of DCIS to invasive cancer.

Background:
TDP-43 proteinopathy is a histopathological hallmark of a class of neurodegenerative disorders termed 'TDP-43 proteinopathies', including motor neuron disease (MND) and frontotemporal lobar degeneration (FTLD). Within disease-affected CNS regions, nearly all MND patients and half of all FTLD patients feature TDP-43 proteinopathy, namely including nuclear loss, cytoplasmic relocation, aggregation, and post-translational modification of the TDP-43 protein.

Purpose:
Our study aimed to investigate the relationship between TDP-43 proteinopathy and epigenetic methylation processes, with a focus on DNA methylation in MND. Additionally, our study sought to explore the metabolic pathways underlying methylation (one-carbon metabolism: composed of the methionine cycle, folate cycle, and transsulfuration pathway), as possible modulators of TDP-43 proteinopathy.

Methods:
We used serial immunohistochemistry to visualise TDP-43 and epigenetic methylation markers (5-methylcytosine and H3K27me3) in formalin-fixed paraffin-embedded MND spinal cord, MND motor cortex, and FTLD hippocampus tissue. Furthermore, we quantified one-carbon metabolites in frozen MND brain tissue using a liquid-chromatography mass spectrometry approach. Finally, we utilised an inducible NSC-34 cell model of TDP-43 proteinopathy to study the influence of one-carbon metabolite supplementation on TDP-43.

Results:
Our study found a significant association between TDP-43 proteinopathy and both 5-methylcytosine and H3K27me3 in all disease-affected CNS regions studied. Further to this, we found evidence for deficiency of key one-carbon metabolites in MND brain, and that supplementation of an NSC-34 cell model of TDP-43 proteinopathy with these deficient metabolites prevented formation of TDP-43 proteinopathy in vitro.

Conclusions:
To conclude, we present evidence of a relationship between TDP-43 proteinopathy and aberrant methylation of DNA and histones, and evidence for one-carbon metabolism dysregulation that may contribute to this. We also highlight the possible therapeutic avenues of manipulating one-carbon metabolism and epigenetic methylation to target TDP-43 proteinopathy.

Background
There is a global shortage of pathologists, driving interest in machine learning (ML) approaches to assist diagnostic workload. Approximately 75% of duodenal biopsies are histologically normal, representing a large volume of low-complexity work. An ML system capable of screening out normal biopsies could significantly reduce the number of cases requiring manual review.

Purpose
We set out to develop and validate a generalisable AI triage system for automated sign-out of normal duodenal biopsies.

Methods
We developed AutoDuo, a multi-stage classification pipeline trained on haematoxylin & eosin-stained duodenal biopsies. The system employs classifiers that iteratively identify: (1) neoplastic tissue, (2) gastric metaplasia or heterotopia, (3) coeliac disease, and (4) inflammatory changes. The training and validation dataset comprised 8730 whole-slide images (WSIs) from 7,769 patients (75% normal), acquired from five hospitals using five different scanners. Generalisability was assessed using leave-one-centre-out cross-validation. The system was further evaluated on an independent test set of 1,328 consecutive biopsies (2,396 WSIs, 74% normal), from a separate hospital using a different scanner manufacturer.

Results
On leave-one-centre-out cross-validation, AutoDuo achieved 94% sensitivity detecting abnormal biopsies and 96% negative predictive value (NPV) whilst signing out half of normal biopsies. Fewer than 1% of coeliac disease and ulceration cases were incorrectly signed out as normal, and no neoplasms were missed. On the independent test set, AutoDuo achieved 94% sensitivity and 97% NPV, signing out 68.4% of normal biopsies. All cases of neoplasia, coeliac disease, and ulceration were correctly retained for pathologist review.

Conclusions
We developed and validated AutoDuo, a multi-centre AI system for triage of duodenal biopsies. AutoDuo safely and efficiently signs out normal biopsies, substantially reducing routine workload while maintaining high sensitivity for clinically significant disease.

Background
Membranous nephropathy (MN) is a common cause of nephrotic syndrome. The phospholipase A2 receptor (PLA2R) is the target antigen in 55-60% of cases, and anti-PLA2R testing is now integral to clinical practice in the United Kingdom (UK). However, many patients, including up to 83% with secondary MN, are PLA2R-negative.

Recently identified glomerular antigens, including thrombospondin type-1 domain-containing 7A 47 (THSD7A), neural epidermal growth factor-like 1 (NELL1), and exostosin 1/2 (EXT1/2) have been implicated in PLA2R-negative MN, with links to malignancy (NELL1) and lupus nephritis (EXT1/2). There are currently no protocols for their detection in UK practice.

We aimed to optimise immunohistochemistry (IHC) for THSD7A, NELL1 and EXT2 in native renal biopsies, and apply these to a UK retrospective cohort.

Methods
IHC protocols were optimised and applied to 86 retrospective renal biopsies collected between 2018 and 2025 at a UK tertiary centre. These comprised 41 cases of non-lupus PLA2R-negative MN and 45 cases of Class V lupus nephritis. Sections underwent antigen retrieval and staining with commercially available monoclonal antibodies. Positivity was defined as granular capillary wall staining. Samples without glomeruli were excluded.

Results
In PLA2R-negative MN, THSD7A was positive in 2/33 samples (6.1%), NELL1 in 10/30 (33%), and EXT2 in 1/32 (3.1%). In the lupus nephritis cohort, EXT2 was positive in 10/40 cases (25%), NELL1 in 2/36 (5.6%), and THSD7A in none.

Conclusion
This is among the first UK cohorts to systematically apply IHC for THSD7A, NELL1, and EXT2 in PLA2R-negative MN. Preliminary positivity rates align with international data. We are now analysing further retrospective cases and patient data to confirm clinical associations; investigating a prospective cohort to assess impact on clinical decision-making; and developing serological assays to support integration into UK diagnostic pathways.