Related Experiment Videos
Integration of bulk and single-cell transcriptomics with spatial analysis identifies candidate hub genes and inferred
Yuqiao Chen1,2,3,4, Yulong Pan1,2,3, Jiajie Chen1,2,3
1Department of Dermatology, The First Affiliated Hospital of Anhui Medical University, Hefei, Anhui, China.
Background:
Atopic dermatitis (AD) is a common chronic inflammatory disorder of the skin. Despite the availability of targeted biologics like dupilumab, heterogeneous treatment responses persist, highlighting the need for reliable biomarkers and a deeper understanding of cell-type-associated molecular programs within the tissue microenvironment.
Methods:
This exploratory study integrated public microarray, single-cell RNA sequencing, and spatial transcriptomic datasets. GSE130588, comprising 124 lesional AD skin samples, was used for weighted gene co-expression network analysis (WGCNA). GSE59294 comprised 40 skin biopsies from 18 patients, including 16 pre-treatment lesional, 12 post-treatment lesional, 7 pre-treatment non-lesional, and 5 post-treatment non-lesional samples, and was used for differential expression analysis. The single-cell query cohort comprised 66,846 cells from 25 samples, with downstream analyses restricted to 19 blister-derived samples. Regulon activity was computationally inferred. GSE197023 comprised 20 spatial transcriptomic samples, including 7 lesional AD, 7 non-lesional AD, and 6 healthy-control samples, and was analyzed using cell-population deconvolution. CCR7 immunoreactivity was assessed by immunohistochemistry in independent healthy-control and lesional AD skin samples (n = 5 per group) analyze.
Results:
WGCNA identified disease-severity-associated gene modules enriched in immune and cell-cycle-related processes. Nominal-P-value-based integration with differential expression results identified seven exploratory candidate hub genes (GZMB, CCR7, GPR183, MMP12, IL7R, RGS1, and KLHDC7B) associated with immune- cell abundance patterns. FDR-adjusted sensitivity analysis supported MMP12 and RGS1 in the lesional-versus-non-lesional comparison, whereas none of the seven genes reached nominal or FDR-adjusted significance in the paired treatment analysis. Single-cell reference mapping showed cell-type-associated expression of GZMB in cytotoxic T-cell populations and CCR7 and MMP12 in dendritic-cell subsets. Computational regulon analysis identified candidate cell-type-associated regulatory signals and differences in inferred regulon activity across treatment-state groups. Spatial deconvolution indicated predominantly dermal distributions of inferred immune-cell populations. Immunohistochemical assessment showed higher relative CCR7 immunoreactivity in lesional AD skin than in healthy-control skin.
Conclusion:
These cross-modal analyses linked candidate molecular signals todendritic-cell- and T- cell-associated patterns in AD. These findings are exploratory and hypothesis-generating and do not establish validated biomarkers, causal regulatory mechanisms, or predictors of treatment response. Independent cohort replication and functional and cell-type-specific validation are required.