Myofibroblast to T-Zone Reticular Cell Ratio Identifies Unicentric Castleman Disease Lymph Nodes

Castleman Disease (CD) is a polyclonal lymphoproliferative disorder presenting with enlarged lymph nodes (LNs), constitutional symptoms, and inflammatory laboratory abnormalities. Idiopathic multicentric Castleman Disease (iMCD) can be acutely life-threatening, but patients often face diagnostic delays due to conflicting pathologic interpretations and the need for second opinions from expert pathologists. Fibroblasts and endothelial cells – non-hematopoietic lymph node stromal cells (LNSCs) – are prominent and contribute to the unusual histologic features of CD, with hyalinized vessels and prominent follicular dendritic cells comprising two of its pathognomonic microscopic features. Recent genetic evidence of non-hematopoietic clonality has emerged in the unicentric form of CD (UCD) where a single LN station is enlarged, including the discovery of a somatic kinase-domain PDGFRB mutation, suggesting that LNSCs may play an active role in disease pathogenesis. We hypothesized that a better understanding of the changes in morphology and distribution of LNSCs could improve the diagnosis of UCD, and eventually iMCD. We speculated that the use of advanced transcriptomic tools could sensitively capture LNSC changes for the development of immunohistochemical stains that can be more widely employed to improve the CD diagnosis.

We established a patient-to-data pipeline to sort-purify LNSCs (CD45-negative fraction) for single-cell RNA sequencing (scRNAseq) from fresh patient LNs. LNSCs, adherent to extracellular matrix, are lost by traditional mechanical LN digestion and require specialized enzymatic digestion for extraction. We processed 21 LN samples: five confirmed to have histological features of hyaline vascular CD and five with a reactive histology. The remaining 11 LNs were interpreted as “indeterminate” with features that included some CD characteristics, but could not be classified as clearly CD or reactive, reflecting a common real-life clinical scenario.

We used a Seurat-based pipeline to compare UCD LNs with cells from LNs with a reactive histology. Using expression of canonical markers, fibroblastic reticular cells, blood endothelial cells, and lymphatic endothelial cells could be resolved in addition to fibroblast subsets, including vascular smooth muscle cells/myofibroblasts, ACTA2⁺ perivascular reticular cells (ACTA2⁺ PRCs), T-zone reticular cells (TRCs), B-zone reticular cells (BRCs), and PI16⁺ reticular cells (PI16⁺ RCs).

Proportionally, there were significantly more myofibroblasts in the UCD LNs compared to reactive histology LN controls. We also observed a decrease in TRCs in UCD LNs. Myofibroblasts, which express contractile genes (ACTA2, MYH11), are PDGFRB-expressing fibroblasts that encircle blood vessels and are known to play inflammatory roles in other contexts. The increase in myofibroblasts with concomitant decrease in TRCs was used as a ratio to distinguish LNs diagnosed as UCD from controls (p=0.0011 on student’s t-test after one-way ANOVA, n=5/group). This ratio identified a subset of 5 of the 11 “Other” LNs with intermediate histology (neither clearly CD or not CD) that had a myofibroblast:TRC ratio similar to UCD. This analysis reveals potential UCD cases that might be missed on simple histopathologic evaluation.

We observed increased TNF and angiogenic gene expression within differentially abundant myofibroblasts in UCD compared to reactive LN controls. By in silico interaction analysis, we identified an increase in predicted stromal-stromal interactions in UCD compared to reactive controls, particularly via the VEGFA-FLT1, COL18A-integrin-α11β1 complex, and JAG1-NOTCH3 signaling axes.

We used hybrid-capture genomic sequencing to identify the presence of the kinase-domain PDGFRB mutation in two of five UCD LNs at low allele frequencies consistent with the proportion of LNSCs within the LN. Moving forward, we will employ spatial transcriptomics (10X Genomics Xenium) to localize the expanded myofibroblasts within the broader LN microanatomy and include a probe to identify the cells harboring the PDGFRB mutation.

Altogether, by defining LNSC changes with advanced transcriptomic approaches in UCD, we hope to identify immunohistochemical stains and other tools that can be more widely employed to facilitate UCD or iMCD diagnosis and prevent the diagnostic delays that currently plague patients and clinicians.