Alterations in single-cell transcriptomic levels and cellular localisation of hnRNP proteins in Frontotemporal Lobar Degeneration

Frontotemporal dementia (FTD) is a neurodegenerative disorder with a strong heritable component. Frontotemporal lobar degeneration (FTLD) refers to the pathological changes seen in FTD, characterised by atrophy of the frontal and temporal lobes and the presence of abnormal protein inclusions. In the case of FTLD with hyperphosphorylated TDP-43 positive inclusions (FTLD-TDP), five pathological subtypes (A, B, C, D, and E) are observed based on the types and distribution of inclusions found in the brain. In all subtypes, there tends to be a large variability in the amount of pathological inclusions observed between cases, with limited correlation to clinical manifestations. TDP-43 is an RNA binding protein belonging to the heterogeneous nuclear ribonucleoprotein (hnRNP) family which along with other hnRNPs modulates multiple aspects of RNA processing. HnRNPs other than TDP-43 have been implicated in several neurological diseases, including ALS, FTLD-TDP, FTLD-FUS and Alzheimer's disease. Multiple hnRNPs have been found in pathological inclusions in specific subtypes of FTLD-TDP, suggesting potential roles in the disease process. The role of the hnRNP network in FTLD disease pathogenesis has not yet been investigated. This study aimed to comprehensively evaluate the presence and expression of hnRNP proteins in two pathological subtypes of sporadic FTLD-TDP (A and C) as well as the genetic form FTLD-TDP A C9orf72 using immunohistochemistry and gene expression analysis by single cell RNA-sequencing. We sought to determine whether hnRNPs exhibited aberrant expression at the protein and transcript level in FTLD cases with TDP-43 pathology. Overall design: Nuclei were isolated from frozen frontal cortex grey matter from 3 FTLD-TDPA, 3 TDPA-C9, and 5 TDPC, and 5 healthy controls. Tissue was transferred directly onto ice-cold sucrose buffer and homogenised by hand. The homogenate was layered on top of a sucrose cushion and nuclei were recovered after centrifugation. Nuclei were resuspended in PBS containing BSA (to prevent clumping) and RNase inhibitors, counted and diluted to 1,000 nuclei/µl. Sequencing libraries were prepared using the “single cell 3' library kit v3” with the Chromium instrument (10X Genomics). Aiming to obtain libraries for ~5,000 independent nuclei per sample. Libraries were sequenced on an Illumina NovaSeq6000 instrument (UCL Genomics), to an average of at least 100,000 reads/nucleus. Raw sequencing data were processed through the software CellRanger (10X Genomics) to assign each transcript read to its respective nucleus, we performed alignment to the transcriptome, counting reads, and processing of the unique molecular identifier tag to reduce PCR bias. The resulting gene count table was analysed using the Seurat package (version 4.4). Quality control was performed to remove low-quality cells based on the number of detected genes and unique molecular identifiers (UMIs). Cells with an unusually high number of UMIs, suggesting potential doublet events, were also excluded using scDblFinder. Using SCTransform, gene expression data were normalised to adjust for differences in sequencing depth between cells and scaled to correct for cell-to-cell variation in total UMI count. Data were also normalised to remove confounding sources of variation such as mitochondrial mapping percentage, post-mortem delay, age of death, and batch number.