Transcriptional profiling of peripheral blood mononuclear cells identifies inflammatory phenotypes in Ataxia Telangiectasia

Abstract

Introduction: Ataxia telangiectasia (A-T) is an autosomal recessive neurodegenerative disease with widespread systemic manifestations and marked variability in clinical phenotypes. In this study, we sought to determine whether transcriptomic profiling of peripheral blood mononuclear cells (PBMCs) defines subsets of individuals with A-T beyond mild and classic phenotypes, enabling identification of novel features for disease classification and treatment response to therapy.

Methods: Participants with classic A-T (n = 77), mild A-T (n = 13), and unaffected controls (n = 15) were recruited from two outpatient clinics. PBMCs were isolated and bulk RNAseq was performed. Plasma was also isolated in a subset of individuals. Affected individuals were designated mild or classic based on ATM mutations and clinical and laboratory features.

Results: People with classic A-T were more likely to be younger and IgA deficient and to have higher alpha-fetoprotein levels and lower % forced vital capacity compared to individuals with mild A-T. In classic A-T, the expression of genes required for V(D)J recombination was lower, and the expression of genes required for inflammatory activity was higher. We assigned inflammatory scores to study participants and found that inflammatory scores were highly variable among people with classic A-T and that higher scores were associated with lower ATM mRNA levels. Using a cell type deconvolution approach, we inferred that CD4 + T cells and CD8 + T cells were lower in number in people with classic A-T. Finally, we showed that individuals with classic A-T exhibit higher SERPINE1 (PAI-1) mRNA and plasma protein levels, irrespective of age, and higher FLT4 (VEGFR3) and IL6ST (GP130) plasma protein levels compared with mild A-T and controls.

Conclusion: Using a transcriptomic approach, we identified novel features and developed an inflammatory score to identify subsets of individuals with different inflammatory phenotypes in A-T. Findings from this study could be used to help direct treatment and to track treatment response to therapy.

Keywords: Bioinformatics; Genetic diseases; T cells.

Fig. 1

Study design and cohort phenotype/genotype characteristics. (A) Diagram outlining study design. (B) Boxplot showing rlog-normalized ATM expression in PBMCs stratified by A-T classification, showing decreasing ATM mRNA detection with increasing disease severity. C,E) Pie charts showing number of mutations with a high impact consequence detected in a given study participant for classic A-T (C) and mild A-T (E) affected individuals. D,F) Barplot showing most frequent mutation consequences for mutations detected in classic A-T (D) and mild A-T (F) affected individuals

Fig. 2

Bulk RNAseq of PBMCs in classic A-T suggests loss of T-cell V(D)J recombination activity and enrichment in monocyte-associated inflammatory pathways. A) Volcano plot showing differences in gene expression in PBMCs from classic A-T vs. unaffected control individuals. B,C) Over representation analysis (ORA) of GO Biological Processes (GO:BP) for genes increased (B) and decreased (C) in expression in classic A-T vs. unaffected control individuals. D) Matrixplot of selected MSigDB hallmark geneset scores. E,F,G) Matrix plots showing mean Z-scored expression of the differentially expressed genes involved in the GO:BPs DNA repair (E), neurogenesis (F), or vascular development (G) (FDR < 0.05)

Fig. 3

A-T classic affected individuals show heterogeneous inflammatory pathway activities. (A) Kernel density estimate plot showing distribution of hallmark inflammatory response scores stratified by condition. (B) Stacked barplot showing proportion of samples with low [0,0.5), moderate [0.5, 0.8), or high [0.8, 1] inflammatory response levels, based on scores in (A). (C) Matrixplot showing expression of genes differentially expressed across inflammatory response levels in classic A-T affected individuals only (Wilcoxon rank-sum, FDR < 0.05). D,E,F) Boxplots of measured alpha-fetoprotein (D), % forced vital capacity (E), and rlog-normalized ATM mRNA expression (F) in mild A-T and classic A-T affected individuals, stratified by inflammatory response level

Fig. 4

Inference of cell type proportions from bulk RNAseq indicates loss of CD4+/CD8 + T-cell populations in A-T affected individuals. (A) Uniform Manifold Approximation and Projection (UMAP) plot of subset of data from (Hagemann-Jensen et al., 2022) Smart-seq3xpress atlas of human PBMCs, showing parent cell type labels mapped here for deconvolution. (B) Dotplot showing expression of a subset of identified marker genes across parent cell type groups. (C) Dotplot showing expression of A-T related genes of interest in this study across parent cell type groups. (D) Stacked barplot of CellAnneal-inferred cell type proportions in bulk RNAseq samples stratified by condition. Single asterisk indicates p < 0.05 between unaffected at classic A-T, double asterisk additionally indicates p < 0.05 between unaffected and mild A-T. (E) As in (D) but for classic A-T affected individuals only, stratified by inflammatory response level. Single asterisk indicates p < 0.05 between low and high inflammatory response levels, double asterisk additionally indicates p < 0.05 between low and moderate inflammatory response levels

Fig. 5

Identification and characterization of protein biomarkers of A-T disease severity. A) Matrixplot of expression of differentially expressed genes across classic A-T and unaffected individuals, restricted to those genes whose protein products are detectable by blood plasma immunoassay or mass-spectrometry according to the Human Protein Atlas. B,D,E) Boxplots of protein concentrations in blood plasma by ELISA for PAI-1 (SERPINE1) (B), VEGFR3 (FLT4) (D), and GP130 (IL6ST) (E), stratified by condition. C) Linear regression of PAI-1 plasma concentration versus age in individuals with classic A-T