An integrated single-cell transcriptomic pipeline identifies ZNF740/BRD3 and Cathepsin S as novel therapeutic targets in chronic active rim smoldering multiple sclerosis

Background. Smoldering multiple sclerosis (MS), characterised by slowly expanding chronic active rim (CA-RIM) lesions, is responsible for a substantial proportion of irreversible disability yet has no approved disease-modifying therapy. The molecular drivers of the CA-RIM lesion microenvironment are poorly characterised. Methods. We developed a four-phase computational pipeline integrating bulk RNA-seq differential expression (Phase 1), single-cell variational autoencoder (scVI) atlas construction and cross-modal integration (Phase 2), multi-database target validation (Phase 3), and STRING-DB network medicine proximity scoring (Phase 4). Bulk DEGs from CA-RIM vs. control white matter (GSE108000) were mapped onto a 32,239-cell scVI atlas combining CELLxGENE Census MS data with single-cell CD8⁺ T-cell profiles from the 6 MS patients and 4 healthy controls in GSE193770 (an earlier version of this atlas pooled in 25 unrelated COVID-19 patients from the same GEO submission's merged file; see §2.1 and Data/Code Availability). Results. Phase 1 identified 394 upregulated candidates (log₂FC ≥ 0.5, FDR < 0.1) from 1,065 total CA-RIM DEGs. Phase 2 constructed a 32,239-cell, 11,811-gene atlas with 19 Leiden clusters; ZNF740 was enriched in the CD8⁺ T-cell cluster, while Cathepsin S (CTSS) was most highly expressed in VIP GABAergic cortical interneurons (see Discussion §4.2 for why the CTSS cell-type attribution changed from an earlier analysis). Phase 2 priority scoring ranked DNMT1 first overall (score 0.715, log₂FC = +1.59), ZNF740 second (0.548, STRING partner BRD3 pairwise score 0.960), and CTSS 15th (0.379, ChEMBL pChEMBL 10.0); 20/25 candidates were blood-accessible by GTEx criteria. Phase 4 network proximity scoring identified CTSS as the strongest MS network medicine candidate among the three focal targets (connects to CSF1R and PTPRC seed genes; combined repurposing score 0.040, second overall behind LINGO1), ZNF740 as directly connected to IFNG (STRING score 0.447) and the BRD2/BRD3/BRD4 bromodomain complex, and DNMT1 as connected to STAT3. Conclusions. This pipeline identifies three previously unreported therapeutic axes in smoldering MS: BET bromodomain inhibition via the ZNF740/BRD3 transcriptional complex; CTSS inhibition combined with a blood-accessible CTSS liquid biopsy (cell-type locus of action not yet resolved — see Discussion); and DNMT1-targeted epigenetic reprogramming of CA-RIM lesion CD8⁺ T cells. All three targets are novel, blood-accessible, and supported by orthogonal computational evidence.