Idiopathic multicentric Castleman disease (iMCD) is a heterogeneous cytokine-mediated lymphoproliferative disorder with distinct clinical subtypes (iMCD-TAFRO, iMCD-IPL, iMCD-NOS). The authors performed bulk RNA sequencing and targeted gene expression profiling on lymph node (LN) tissue from multiple iMCD clinical subtypes, sentinel controls, and selected disease comparators (SLE and DLBCL). They identified hundreds of differentially expressed genes in iMCD-TAFRO versus sentinel controls and pathway enrichment in angiogenesis, cell proliferation, and immune responses. Clusterin (CLU) emerged as a gene upregulated across iMCD subtypes and, by immunohistochemistry, was elevated in germinal center and mantle zone regions. A composite model incorporating Clusterin expression and germinal center morphometry discriminated iMCD from selected lymphadenopathies, suggesting Clusterin plus GC features as a candidate biomarker.
iMCD is a rare polyclonal lymphoproliferative disorder characterized by systemic inflammation and multicentric lymphadenopathy. Clinical phenotypes include iMCD-TAFRO (thrombocytopenia, anasarca, fever/CRP elevation, reticulin myelofibrosis/renal dysfunction, organomegaly), iMCD-IPL (plasmacytic lymphadenopathy with thrombocytosis and hypergammaglobulinemia), and iMCD-NOS. Histopathological subtypes (hyaline vascular/hypervascular, plasmacytic, mixed) are recognized but their clinical implications remain unclear. Approximately half of iMCD patients respond to anti–IL-6 therapy, highlighting the need for additional diagnostic markers and therapeutic targets.
The discovery cohort included sentinel control lymph nodes (resected, metastasis-negative breast cancer sentinel nodes; n=7) and iMCD-TAFRO samples (n=7). Additional LN samples from DLBCL (n=5) were included in discovery analyses. The targeted NanoString cohort comprised iMCD-TAFRO (n=12), iMCD-IPL (n=3), iMCD-NOS (n=4), SLE (n=4), and DLBCL (n=5), with overlap between some discovery and targeted samples noted. Validation cohorts and supplementary published samples were incorporated for expanded analyses; detailed composition is provided in the article’s supplementary material.
RNA was extracted from FFPE tissue sections using a commercial RNeasy DSP FFPE kit with proteinase K digestion and DNase I treatment. RNA quality was assessed using TapeStation and concentration by Qubit; DV200 values were used to quantify the proportion of fragments >200 bp. Libraries for whole transcriptome sequencing were prepared with a Stranded Total RNAseq with Ribo-zero Plus kit, using 100 ng input and 15 PCR cycles. Sequencing was performed as paired-end 2x100 bp on an Illumina NextSeq 2000 with a minimum of 30 million reads per sample.
For the NanoString nCounter assays, 100 ng of RNA fragments ≥200 bp were hybridized to the Immunology panel (v2), measuring 594 immune-related genes plus 15 internal reference genes. Raw counts were analyzed using NanostringQCPro and normalized with RUVSeq, leveraging the 15 housekeeping genes for normalization prior to differential expression testing with DESeq2.
RNA-seq reads were pseudo-aligned to GRCh38 with Kallisto, and gene counts were quantified in R. DESeq2 was used for quality control and differential expression analysis. Genes were filtered to include those with read count ≥20 in >33% of samples. Differentially expressed genes (DEGs) in the discovery cohort were defined by |log2 fold change| >1 and Benjamini-Hochberg–adjusted P < 0.05. GSEA used Hallmark gene sets with FDR q-value < 0.25 considered significant. The validation cohort underwent batch correction using limma and independent DEG analysis following the same criteria.
GSEA of the iMCD-TAFRO versus sentinel control comparison identified enrichment in angiogenesis, cell proliferation, and humoral and innate immune response pathways. To estimate cell-type composition from bulk RNAseq, publicly available single-cell LN atlases were used as reference; Bisque deconvolution produced inferred proportions for 13 annotated cell types (including B cells, T cells, plasma cells, dendritic cells, macrophages, etc.). Comparative statistics used Wilcoxon rank-sum tests with Bonferroni correction for multiple comparisons.
Top DEGs from the iMCD-TAFRO versus sentinel control signature (± log2 FC >1.5, adjusted P <0.05) were entered into L2S2 querying of the LINCS1000 database to predict drug perturbations that might reverse the signature. Predicted reversing drugs were grouped by mechanism of action (MOA) and ranked by aggregated significance across cell lines; outputs were reported in contexts including all lines and immune-restricted lines. The procedural details for score calculation and ranking are described in the article.
Clusterin protein localization was assessed by immunohistochemistry (DAB chromogen with hematoxylin counterstain). A Python-based algorithm separated DAB and hematoxylin channels (rgb2hed from scikit-image) and quantified DAB optical density. Lymph node area was segmented using an Otsu threshold on the hematoxylin channel. Germinal centers (GCs), mantle zones (MZs), and interfollicular regions were manually segmented blinded to Clusterin staining; DAB thresholds were used to define positive versus negative Clusterin regions.
Manual segmentation and digital quantification produced morphometric measures of GCs and Clusterin localization within GC and MZ compartments. The authors report that a composite model combining Clusterin expression in these compartments with germinal center features showed strong discriminatory performance for distinguishing iMCD from selected non-CD lymphadenopathies. Specific performance metrics for the composite model (AUC, sensitivity, specificity) were not provided in the text excerpt.
In the discovery iMCD-TAFRO versus sentinel control comparison, 249 genes were upregulated and 42 downregulated. Shared DEGs across all three iMCD clinical subtypes included SPP1, XBP1, PRDM1, and CLU. Subtype-unique DEGs included PLA2G2A, MSR1, and IL8 in iMCD-TAFRO; IL9, KIR3DL1, and IL18RAP in iMCD-IPL; and IL7, TLR8, and CX3CR1 in iMCD-NOS. Prior studies identified increased XBP1, CXCL13, CLU, angiogenesis genes, mTORC1 pathway genes, and complement cascade signatures in MCD LN tissue; the present study extends these observations with subtype-specific and comparative analyses.
The work implicates Clusterin as a potential lymph node biomarker for iMCD when evaluated alongside germinal center morphometry. Clusterin was not elevated in the related comparators included in the targeted cohort, and its localization to GC and MZ compartments in iMCD supports a tissue-level signal consistent with follicular or stromal compartment involvement. The identification of shared and unique DEGs across clinical subtypes suggests both common and subtype-specific biology, with pathway enrichment consistent with angiogenesis, proliferation, and humoral/innate immune responses. The LINCS1000 query nominates candidate perturbations to reverse the iMCD-TAFRO signature, although further validation is required.
The provided text reports cohort sizes and analytic methods but does not include specific numerical classifier performance statistics for the composite model, nor the list of top LINCS-predicted compounds in the excerpt. Details on statistical effect sizes for all DEGs in the full validation cohort are referenced in supplementary tables but not reproduced here. The sample sizes for some subgroups (e.g., iMCD-IPL) are small and were combined in validation analyses where noted.
This study identifies transcriptomic signatures in iMCD lymph nodes with upregulated Clusterin across iMCD subtypes and demonstrates that Clusterin localization to germinal centers and mantle zones, combined with GC morphometry, can aid in distinguishing iMCD from select lymphadenopathies. These findings nominate Clusterin plus histomorphometric features as candidate diagnostic markers and motivate further validation and exploration of therapeutic perturbations suggested by gene-signature reversal analyses.