Chronic inflammatory demyelinating polyneuropathy (CIDP) is an immune-mediated peripheral neuropathy with heterogeneous clinical responses to standard immunotherapies. Although CIDP and multiple sclerosis (MS) both involve immune-mediated demyelination, they affect different anatomical compartments and show divergent therapeutic responses, implying distinct underlying immune programs. This study sought to define the peripheral immune architecture of CIDP using an integrated, multi-modal approach and to contextualize CIDP immune states against MS as a cross-disease reference.
Peripheral blood was obtained from 20 patients with CIDP and 20 age- and sex-matched healthy controls (HCs). Participants with conditions affecting peripheral nerve function or systemic immune status were excluded. The study protocol was approved by the Institutional Review Board of Asan Medical Center and written informed consent was obtained from all participants. A focused scRNA-seq discovery subset comprised cryopreserved PBMCs from two CIDP patients and two matched HCs and was integrated with publicly available MS and HC scRNA-seq datasets for cross-disease comparison.
Cryopreserved PBMCs were thawed and analyzed by flow cytometric immunophenotyping using commercial DuraClone reagent panels. Regulatory B cells were defined as CD19+ CD24hi CD38hi and regulatory T cells as CD3+ CD4+ CD25+ FoxP3+ using manufacturer panels without additional antibodies. Data acquisition used a 10-color Navios flow cytometer and Kaluza software. Serum cytokines (including GM-CSF, Granzyme B, IFN-γ, IL-1β, IL-6, IL-10, TNF-α, TGF-β1 and others) were measured externally using Luminex-based assays.
Up to 10,000 PBMCs per sample were processed with the Chromium Single Cell 3′ platform (10X Genomics) and sequenced on an Illumina NextSeq 500. Raw data were processed with CellRanger aligned to GRCh38 to generate gene-cell UMI count matrices. In-house scRNA-seq from two CIDP and two HC samples were integrated with three relapsing–remitting MS samples and three additional HCs from public repositories, resulting in a total of 10 datasets for integration.
Quality filters removed cells with <500 or >2,500 detected genes, UMIs outside 1,000–15,000, or >15% mitochondrial transcripts. Ambient RNA contamination and predicted doublets were removed, leaving 23,717 high-quality cells and 11,455 genes. Data were log-normalized; the top 2,000 variable genes were selected. Batch correction used FastMNN; dimensionality reduction used PCA and UMAP. Initial cell identities were assigned via Azimuth CITE-seq reference label transfer, followed by high-resolution clustering and majority-vote reannotation. Canonical marker expression and pseudobulk correlations validated annotations.
Module scores were calculated for curated immune-related gene signatures including type I/II interferon responses, inflammasome activation, and cytotoxicity. Statistical comparisons across CIDP, MS, and HC used Kruskal–Wallis tests with post-hoc Dunn’s tests and effect-size thresholds to ensure biologic relevance. Cell-cycle S phase scoring was performed using a standard gene set. Pathway activity at single-cell resolution used GSVA with MSigDB KEGG MEDICUS gene sets; Wilcoxon rank-sum tests and FDR correction were applied. Pathways were ranked by aggregated effect sizes across cell types to identify those most differentially active in CIDP relative to HC.
CellPhoneDB was used per donor with 1,000 permutations to infer significant ligand–receptor interactions (P < 0.05). For each sender–receiver pair, counts of significant versus tested interactions were compared across disease groups using Fisher’s exact tests and FDR adjustment. Aggregate mean interaction scores were compared to identify interactions exclusive to or enriched in CIDP versus HC or MS. Results were ranked by magnitude of difference to highlight the most dysregulated intercellular communications.
All B cells were isolated and reclustered into B Naïve and B Memory subsets with resolution 0.5 to define subclusters. Several B Naïve subclusters showed disease-biased distributions (CIDP-enriched, MS-enriched, HC-enriched). Differentially expressed genes for CIDP-enriched clusters were identified using Seurat’s FindAllMarkers and interpreted via over-representation analysis to define molecular programs associated with CIDP-biased B-cell states.
Across modalities, CIDP showed broad inflammatory activation with preferential enrichment of type I interferon and inflammasome-related programs compared with MS. Although total B-cell frequencies were reduced in CIDP, B cells exhibited transcriptional enrichment of germinal center–associated activation programs, indicating a dissociation between cell number and activation state. In parallel, CD8 effector T cells in CIDP displayed enhanced cytotoxicity and cytoskeletal remodeling programs, with core actin-regulatory genes (for example, ACTR2, ACTR3, ARPC1B) implicated in CD8 effector clusters. Cell–cell interaction analyses supported strengthened intercellular signaling involving cytotoxic CD8 T cells in CIDP. By contrast, MS demonstrated greater enrichment of integrin–talin–vinculin signaling pathways in B cells and CD4 T-cell subsets, consistent with trafficking-related immune mechanisms described in central nervous system disease.
This integrated peripheral immune profiling identified candidate CIDP-associated immune signatures: dysregulated B-cell activation despite numerical reduction and a prominent cytotoxic CD8 T-cell program within an inflammatory milieu skewed toward type I interferon and inflammasome pathways. The single-cell component was performed as an exploratory, hypothesis-generating step. The authors report these signatures as candidates that warrant further translational validation in larger, treatment-stratified cohorts to inform biomarker development and targeted therapeutic strategies for CIDP.