MAPT haplotype-associated transcriptomic changes in progressive supranuclear palsy

Abstract

Progressive supranuclear palsy (PSP) is a neurodegenerative movement and cognitive disorder characterized by abnormal accumulation of the microtubule-associated protein tau in the brain. Biochemically, inclusions in PSP are enriched for tau proteoforms with four microtubule-binding domain repeats (4R), an isoform that arises from alternative tau pre-mRNA splicing. While preferential aggregation and reduced degradation of 4R tau protein is thought to play a role in inclusion formation and toxicity, an alternative hypothesis is that altered expression of tau mRNA isoforms plays a causal role. This stems from the observation that PSP is associated with common variation in the tau gene (MAPT) at the 17q21.31 locus which contains low copy number repeats flanking a large recurrent genomic inversion. The complex genomic structural changes at the locus give rise to two dominant haplotypes, termed H1 and H2, that have the potential to markedly influence gene expression. Here, we explored haplotype-dependent differences in gene expression using a bulk RNA-seq dataset derived from human post-mortem brain tissue from PSP (n = 84) and controls (n = 77) using a rigorous computational pipeline, including alternative pre-mRNA splicing. We found 3579 differentially expressed genes in the temporal cortex and 10,011 in the cerebellum. We also found 7214 differential splicing events in the temporal cortex and 18,802 in the cerebellum. In the cerebellum, total tau mRNA levels and the proportion of transcripts encoding 4R tau were significantly increased in PSP compared to controls. In the temporal cortex, the proportion of reads that expressed 4R tau was increased in cases compared to controls. 4R tau mRNA levels were significantly associated with the H1 haplotype in the temporal cortex. Further, we observed a marked haplotype-dependent difference in KANSL1 expression that was strongly associated with H1 in both brain regions. These findings support the hypothesis that sporadic PSP is associated with haplotype-dependent increases in 4R tau mRNA that might play a causal role in this disorder.

Keywords: KANSL1; MAPT haplotype; 17q21.31; Progressive supranuclear palsy; RNA-seq; Tauopathy.

Fig. 1

RNA sequencing pipeline using RAPiD-nf aligns and quality controls 161 samples for differential expression and splicing analysis. a RNA-seq samples were obtained from Synapse.org [50]. The cohort included 161 PSP cases and controls, with samples taken from the cerebellum and temporal cortex. Samples were processed using RAPiD-nf, a processing pipeline in the NextFlow framework. RAPiD-nf uses Trimmomatic, STAR, FASTQC, featureCounts, Pathogen, and Picard for pre-processing and quality control. RSEM and Limma were used for differential gene expression, and Regtools and LeafCutter were used for differential intron excision. b, c Principal component analysis of the RNA-seq expression matrix after covariate adjustment in cerebellum and temporal cortex shows no remaining outliers. 161 samples were included in downstream analysis

Fig. 2

Genes within the 17q21.31 locus are differentially expressed between progressive supranuclear palsy cases and controls. 10,011 genes were differentially expressed in the cerebellum (a) and 3579 genes were differentially expressed in the temporal cortex (b; p < 0.05, FDR adjusted). (c, d) MAPT, KANSL1, NSF, PLEKHM1, LRRC37A, and ARHGAP27 were differentially expressed in the cerebellum. KANSL1, PLEKHM1, LRRC37A, and ARHGAP27 were differentially expressed in the temporal cortex. ARL17A and CRHR1 were not differentially expressed in either brain region. e Map of the MAPT locus with location of genes examined. f IPA pathway analysis of top 5 canonical pathways from both the cerebellum and temporal cortex in order of p value, with z-score displayed

Fig. 3

Differential alternative pre-mRNA splicing of MAPT and KANSL1 in PSP cases compared to controls. a Full MAPT gene with differential inclusion of exon 10 in both brain regions for PSP cases compared to controls. (FDR adj. p = 0.0007 in cerebellum, FDR adj. p = 0.0003 in temporal cortex). Red lines represent introns annoted in GENCODE, pink lines are novel introns. Each number corresponding to a line represents the relative junction usage within the cluster. b Full KANSL1 gene with differential inclusion of an intron–exon junction in both brain regions (FDR adj. p = 3.0 × 10⁻¹³ in cerebellum, FDR adj. p = 0.000001 in temporal cortex). Hg38 Chr17: 46,094,701–46,170,855. c Schematic of H1/H2 haplotype breakpoints occurring within the KANSL1 gene, with potential KANSL1 pseudogene overlaid

Fig. 4

17q21.31 H1 haplotype and progressive supranuclear palsy are both associated with increased 4R tau mRNA and KANSL1 mRNA isoform expression. Patient haplotype information was determined using the SNP tag at rs1800547. H1 homozygous individuals are shown in orange (n = 77 PSP, n = 44 controls), H1/H2 heterozygous in green (n = 7 PSP, n = 22 controls), and H2 homozygous in blue (n = 0 PSP, n = 4 controls). q values are FDR-adjusted p values and were used to determine significance between cases and controls for each isoform. For haplotype-associated isoform and pseudogene expression analysis, a one-way ANOVA was performed