Significant outcomes
-
1. We present the first transcriptome-wide, transcript-level, differential expression analysis of post-mortem Lewy body dementia (LBD) brains, and we identified 169 differentially expressed transcripts (DET) and 228 alternatively spliced genes after multiple testing corrections in LBD brains.
-
2. Identified DET may contribute to LBD pathology by impacting DNA repair, apoptosis, protein phosphorylation, and transcription regulation.
-
3. Therapeutic and biomarker potential of identified DET, especially those from TMEM18, MICB, MPO and GABRB3, warrant further investigation.
Limitations
-
1. Transcriptome-wide transcript-level differential expression data analysis algorithms are still evolving.
-
2. Functional annotations of individual RNA transcripts are limited. Our functional enrichment analysis was based on translated proteins, and it did not include the effects of identified non-coding DET.
-
3. Small sample size.
Introduction
Dementia is the seventh leading cause of global mortality (Patterson, Reference Patterson2018). Lewy body dementias (LBD) are the second most common type of neurodegenerative dementias (Kane et al., Reference Kane, Surendranathan, Bentley, Barker, Taylor, Thomas, Allan, McNally, James, McKeith, Burn and O’Brien2018). They include two overlapping dementias: dementia with Lewy bodies (DLB) and Parkinson’s disease (PD) dementia (PDD) (Kane et al., Reference Kane, Surendranathan, Bentley, Barker, Taylor, Thomas, Allan, McNally, James, McKeith, Burn and O’Brien2018). When compared to Alzheimer’s disease (AD), people with LBD experience more rapid cognitive decline, shorter life expectancy, greater care costs, and more frequent and more severe neuropsychiatric symptoms (Svendsboe et al., Reference Svendsboe, Terum, Testad, Aarsland, Ulstein, Corbett and Rongve2016). However, LBD remain relatively under-researched. The molecular mechanisms underlying neurodegeneration in LBD are uncertain, and we do not have any disease-modifying treatment for LBD (Watts et al., Reference Watts, Storr, Barr and Rajkumar2023). There is no reliable biological fluid-based biomarker for LBD, and nearly 50% of people with LBD in the UK may remain misdiagnosed as other dementias (Freer, Reference Freer2017). There is an urgent need for further research that advances our understanding of LBD molecular pathology. Such research is prerequisite for the discovery of novel therapeutic targets and diagnostic biomarkers that can improve clinical diagnosis and management of LBD.
We have systematically reviewed prior genetic (DNA) (Sanghvi et al., Reference Sanghvi, Singh, Morrin and Rajkumar2020) and gene expression (RNA) (Chowdhury and Rajkumar, Reference Chowdhury and Rajkumar2020) studies that investigated people with LBD. Genetic studies have consistently replicated the associations between LBD and variants in APOE, SNCA, and GBA, and have reported genetic associations with variants in several genes including BCL7C, CNTN1, GABRB3, and MAPT (Sanghvi et al., Reference Sanghvi, Singh, Morrin and Rajkumar2020). DNA exert their functions by RNA transcription. Prior RNA expression and transcriptomic studies have identified many differentially expressed genes (DEG) that may contribute to LBD pathology by impacting mitochondrial dysfunction, immunosenescence, the ubiquitin-proteasome system (UPS), the autophagy lysosomal pathway (ALP), RNA-mediated gene silencing, oxidative stress and signal transduction (Chowdhury and Rajkumar, Reference Chowdhury and Rajkumar2020, Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020). However, each gene may transcribe multiple RNA transcripts with unique functions. Alternative splicing can change the proportions of expressed transcripts of a gene and consequent functions without changing the overall expression level of that gene (Wang et al., Reference Wang, Kumar, Olson and Ware2019). Gene-level DEG analyses investigate only aggregated expression levels of all transcripts from individual genes. They cannot provide information regarding differentially expressed transcripts (DET), and the extent of alternative splicing in LBD (Wang et al., Reference Wang, Kumar, Olson and Ware2019). Identifying DET and alternatively spliced isoforms has provided novel insights into the molecular pathology of AD (Raj et al., Reference Raj, Li, Wong, Humphrey, Wang, Ramdhani, Wang, Ng, Gupta, Haroutunian, Schadt, Young-Pearse, Mostafavi, Zhang, Sklar, Bennett and De Jager2018), but there has not been any transcriptome-wide study investigating transcript-level differential expression analysis in LBD.
The only available transcriptome-wide alternative splicing investigation of LBD brains (n = 14) has highlighted widespread dysregulation of alternative splicing (Feleke et al., Reference Feleke, Reynolds, Smith, Tilley, Taliun, Hardy, Matthews, Gentleman, Owen, Johnson, Srivastava and Ryten2021). That study investigated only one brain region (Anterior Cingulate Cortex; ACC), did not report the most significant alternatively spliced genes, and did not conduct DET analyses. There have been candidate gene expression studies investigating alternatively spliced isoforms of SNCA, SNCB, PRKN, FYN, APP, RELA, and ATXN2 in LBD (Beyer et al., Reference Beyer, Lao, Carrato, Mate, Lopez, Ferrer and Ariza2004; Beyer et al., Reference Beyer, Domingo-Sabat, Santos, Tolosa, Ferrer and Ariza2010; Funahashi et al., Reference Funahashi, Yoshino, Yamazaki, Mori, Mori, Ozaki, Sao, Ochi, Iga and Ueno2017; Low et al., Reference Low, Lee, Lim, Lee, Ballard, Francis, Lai and Tan2021). Their findings support the importance of alternative splicing and transcript-level alterations in the molecular pathology of LBD (Beyer et al., Reference Beyer, Lao, Carrato, Mate, Lopez, Ferrer and Ariza2004; Beyer et al., Reference Beyer, Domingo-Sabat, Santos, Tolosa, Ferrer and Ariza2010; Funahashi et al., Reference Funahashi, Yoshino, Yamazaki, Mori, Mori, Ozaki, Sao, Ochi, Iga and Ueno2017; Low et al., Reference Low, Lee, Lim, Lee, Ballard, Francis, Lai and Tan2021). There is a clear need to analyse transcriptome-wide data at a transcript-level resolution to provide further insight into the molecular pathology of LBD. We have previously reported transcriptome-wide DEG and consequent metabolic reprogramming in post-mortem ACC and dorsolateral prefrontal cortices (DLPFC) of people with pathology-verified LBD (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020). There is a clear requirement to analyse this next-generation RNA-Sequencing (RNA-Seq) data (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020) at transcript-level resolution. We aimed to perform the first transcriptome-wide transcript-level differential expression analysis of post-mortem LBD brains for identifying DET and alternatively spliced genes that may facilitate discovery of novel therapeutic targets and biomarkers for LBD.
Materials and methods
Post-mortem brain tissue
We analysed data from our prior RNA-Seq study (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020). We provide a summary of research methods here. Further details of our methodology have been published elsewhere (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020). The Brains for Dementia Research (BDR) network of brain banks (https://bdr.alzheimersresearchuk.org/researchers/), UK, provided post-mortem brain tissue, and generic ethical approval for this study (Approval 13/SC/0516 by the Oxford C Committee of the National Research Ethics Service). Post-mortem ACC (Brodmann area 24) (Pietrzak et al., Reference Pietrzak, Papp, Curtis, Handelman, Kataki, Scharre, Rempala and Sadee2016) and DLPFC (Brodmann area 9) (Bronnick et al., Reference Bronnick, Breitve, Rongve and Aarsland2016) tissue samples from people with pathologically-verified LBD (n = 14), and from people without PD or dementia (NDC; n = 7) were included in this study. Supplemental Digital Content (SDC-1) presents the sample characteristics.
RNA extraction
50 mg of tissue per brain region was obtained from each sample. They were homogenised with the T10-basic ultra-turrax and S10D-7G-KS-110 dispersing tool (Ika, USA). We extracted total RNA using the RNeasy Plus Universal Mini Kit (Qiagen, Germany).
RNA-seq
cDNA libraries were prepared from RNA samples using TruSeq RNA sample preparation kit (Illumina, San Diego, USA). The cDNA libraries were sequenced (paired-end; 75 base pairs/read; minimum 30 million clean reads per sample) using the Illumina HiSeq-4000 (Illumina, San Diego, USA) in the Wellcome Centre for Human Genetics, Oxford, UK. Figure 1. presents an overview of study methods.
Quantification of RNA transcripts
Figure 2. presents an overview of our data analysis pipeline. We excluded RNA-seq reads that had ambiguous bases or had more than 1% sequencing error in more than 10% bases. We quantified transcript abundance by Salmon (Patro et al., Reference Patro, Duggal, Love, Irizarry and Kingsford2017), a quasi-alignment quantification tool capable of transcriptome-wide bias calculation. Salmon compares RNA-seq reads to a transcriptome index and performs equivalence class calculations for estimating abundance of each transcript in an RNA sample. We used the Gencode.v38 transcript fasta file (Frankish et al., Reference Frankish, Carbonell-Sala, Diekhans, Jungreis, Loveland, Mudge, Sisu, Wright, Arnan, Barnes, Banerjee, Bennett, Berry, Bignell, Boix, Calvet, Cerdan-Velez, Cunningham, Davidson, Donaldson, Dursun, Fatima, Giorgetti, Giron, Gonzalez, Hardy, Harrison, Hourlier, Hollis, Hunt, James, Jiang, Johnson, Kay, Lagarde, Martin, Gomez, Nair, Ni, Pozo, Ramalingam, Ruffier, Schmitt, Schreiber, Steed, Suner, Sumathipala, Sycheva, Uszczynska-Ratajczak, Wass, Yang, Yates, Zafrulla, Choudhary, Gerstein, Guigo, Hubbard, Kellis, Kundaje, Paten, Tress and Flicek2023) as the index transcriptome, performed G-C content bias correction, and calculated transcript abundance. We annotated estimated transcript counts with the GRCh38.p13 genome reference.
Transcriptome-wide DET analysis
We identified DET in LBD brains using edgeR (Robinson et al., Reference Robinson, McCarthy and Smyth2010), and Benjamini-Hochberg false discovery rate (FDR) correction at 5%. edgeR employs negative binomial distribution for calculating differential expression from count data without degrees of freedom (df) (Robinson et al., Reference Robinson, McCarthy and Smyth2010). The transcript count matrix was analysed for differential expression, while using a transcript length matrix as an offset within the analysis. SDC-2 presents our command scripts for supporting reproducibility.
Alternative splicing analysis
We identified alternatively spliced genes using the transcript count matrix and DRIMseq (Nowicka and Robinson, Reference Nowicka and Robinson2016). DRIMseq employs a Dirichlet-multinomial model for comparing the relative ratio of expressed isoforms between conditions, while accounting for differential gene expression. DRIMseq analyses have n-1 df, while comparing two conditions. Post-hoc filtering of transcript-level tests was applied using StageR (Van den Berge et al., Reference Van den Berge, Soneson, Robinson and Clement2017) and an alpha of 5% for correcting the overall false discovery rate (OFDR) of DRIMseq. StageR was preferred for this correction as DRIMseq may exceed its FDR bounds, and application of post-hoc filtering can improve accuracy significantly (Nowicka and Robinson, Reference Nowicka and Robinson2016).
Functional enrichment analysis
We investigated functional implications of the identified DET using the Database for Annotation, Visualization and Integrated Discovery (DAVID) (da Huang et al., Reference da Huang, Sherman and Lempicki2009, Sherman et al., Reference Sherman, Hao, Qiu, Jiao, Baseler, Lane, Imamichi and Chang2022). DAVID groups input terms into biological modules, and identifies enriched biological processes, molecular functions and Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathways. We combined all nominally significant (edgeR p < 0.05) DET from both brain regions into a single list. We converted their Ensembl transcript IDs to UniProt Accession numbers using the biological database network (bioDBnet; https://biodbnet-abcc.ncifcrf.gov) and analysed the list of UniProt Accession numbers using DAVID. Such systematic functional analysis of alternatively spliced genes is not possible because of the incomplete functional annotation of the effects of alternative splicing within individual genes and the lack of a comprehensive functional annotation database for alternative splicing.
Secondary analyses
The LBD group including both DLB and PDD samples was compared with the NDC group for identifying DET and alternatively spliced genes in LBD brains. Later, we conducted pairwise subgroup analyses comparing DLB, PDD and NDC groups using similar edgeR and DRIMseq algorithms with appropriate FDR corrections. We conducted these analyses for ACC and DLPFC separately.
Results
196,916 and 196,360 unique RNA transcripts were expressed in ACC and DLPFC of the study samples, respectively.
DET in LBD brains
We identified 74 FDR-corrected DET in the ACC of LBD brains. Of these, 30 were upregulated and the remaining 44 were downregulated. There were 96 FDR-corrected DET, including 31 upregulated and 65 downregulated DET, in the DLPFC of LBD brains. Table 1 presents the top 10 upregulated and downregulated DET, ranked by their FDR-corrected p-values, in ACC. Table 2 shows the top 10 upregulated and downregulated DET in DLPFC of LBD brains. Moreover, SDC-3 presents the details of all nominally significant (edgeR p-values < 0.05) DET in both brain regions. The ENST00000432667.5 transcript, transcribed by TMEM18, was significantly upregulated and expressed in all LBD DLPFC samples, but not in any NDC DLPFC sample. Furthermore, ENST00000225275.4, transcribed by the AD-associated proinflammatory MPO, was significantly downregulated in both regions of LBD brains after Benjamini-Hochberg FDR correction (5%).
aTop differentially expressed transcripts as ranked by their adjusted p-values (Benjamini-Hochberg false discovery rate correction at 5%); bBase of 2; Analysis was completed using edgeR. edgeR estimates dispersion with a negative binomial distribution and calculates differential expression from count data. Transcript length was used as an offset and this analysis was performed without df.
aTop differentially expressed transcripts as ranked by their adjusted p-values (Benjamini-Hochberg false discovery rate correction at 5%); bBase of 2; Analysis was completed using edgeR. edgeR estimates dispersion with a negative binomial distribution and calculates differential expression from count data. Transcript length was used as an offset and this analysis was performed without df.
DET in DLB and PDD brains
SDC-3 presents the details of all nominally significant DET that were identified by DET subgroup analyses, comparing the transcriptomes of both brain regions of the DLB samples with corresponding NDC and PDD samples. When compared to the NDC samples, there were 129 and 121 FDR-corrected DET in the ACC and DLPFC of DLB samples, respectively. While comparing the PDD samples, we identified 143 and 114 FDR-corrected DET in the ACC and DLPFC of DLB samples, respectively. ENST00000262327.9, transcribed by DNA ligase LIG3, was the top DET with the least FDR-corrected p-value (8.30*10-23) in the ACC of DLB brains. ENST00000682046.1, transcribed by DNA-binding THAP12, was the top DET (FDR-corrected p-value = 3.32*10-10) in the DLPFC of DLB brains, when compared to PDD brains. Moreover, when compared to the NDC samples, we identified 130 and 156 FDR-corrected DET in the ACC and DLPFC of PDD samples, respectively.
Alternative splicing in LBD brains
We detected 135 significantly alternatively spliced genes in the ACC of LBD brains after OFDR correction (5%). There were 98 significantly alternatively spliced genes including TMEM18, GOLGA2, CTTN, ARHGEF4 and SHC2 in the DLPFC of LBD brains after similar OFDR correction. Table 3 presents the top 20 alternatively spliced genes, ranked by their OFDR-corrected p-values, in both brain regions. SGSH, NAV2, ZC3H7A, FAM76A and SURF1 were the top five alternatively spliced genes in ACC of LBD brains. SDC-4 presents further results from the DRIMseq transcriptome-wide alternative splicing analysis of both LBD brain regions. Four genes, MYL6, CTTN, ING3, and LOC105374338, were significantly alternatively spliced after OFDR correction in both LBD brain regions.
aTop 20 alternatively spliced genes as ranked by their adjusted p-values; bAnterior cingulate cortex; cDorsolateral prefrontal cortex; dP-values adjusted with StageR, alpha = 5%. Alternative splicing analysis was performed with DRIMSeq. DRIMSeq utilises a Dirichlet-multinomial model to compare the relative ratio of expressed isoforms between conditions. Degrees of freedom was set as q-1.
Alternative splicing of TMEM18 and MICB
Figure 3. presents an overview of possible mechanisms by which alternative splicing of TMEM18 and of MICB may contribute to LBD pathology. TMEM18 is the top alternatively spliced gene in the DLPFC of LBD brains (Table-3; SDC-4). It also transcribes the top upregulated DET and the top downregulated DET in the same region (Table-2; SDC-3). The top downregulated transcript, TMEM18-202, translates into functional Transmembrane Protein 18 (UniProt:Q96B42; TMM18), and the top upregulated transcript, TMEM18-205, leads to a nonsense mediated decay product (UniProt:F8WBA6). TMEM18-202 represented 18.64% of all TMEM18 transcripts within the NDC brains, and it was not detected in any LBD brains. TMEM18-205 represented 22.08% of all TMEM18 transcripts within the LBD brains, and it was not detected in any NDC brain. Similarly, MICB is an OFDR-corrected alternatively spliced gene (OFDR-corrected p-value = 1.85*10-3) in the DLPFC of LBD brains. MICB transcribes MICB-202 and MICB-204 that translate into an MHC class 1 polypeptide-related sequence with (UniProt:Q29980) or without (UniProt:F5H7Q8) a signal peptide, respectively. In LBD brains, 26.96% of all MICB transcripts were MICB-204, whilst none of the NDC brains expressed MICB-204. Conversely, MICB-202 represented 47.70% of all MICB transcripts within the NDC group, and it only represented 4.47% of transcripts within the LBD group.
Alternative splicing in DLB and PDD brains:
SDC-4 presents DRIMseq transcriptome-wide alternative splicing subgroup analyses, comparing the transcriptomes of both DLB brain regions with corresponding NDC and PDD samples. When compared to the NDC group, there were 49 and 84 OFDR-corrected significantly alternatively spliced genes in the ACC and DLPFC of DLB brains, respectively. When compared to the PDD group, we found 96 and 100 OFDR-corrected alternatively spliced genes in the ACC and DLPFC of DLB brains, respectively. SGSH, associated with a lysosomal storage disease, was the top alternatively spliced gene with the lowest OFDR-corrected p-value (6.06*10-7) in the ACC of DLB brains. TMEM18 was one of the top five alternatively spliced genes (OFDR-corrected p-value = 8.88*10-6) in the DLPFC of DLB brains, when compared to NDC brains. Furthermore, when compared to the NDC group, there were 133 and 110 OFDR-corrected alternatively spliced genes in the ACC and DLPFC of PDD brains, respectively.
Functional enrichment analysis of identified DET
We analysed the functional implications of all unique proteins that are translated by the identified nominally significant (edgeR p < 0.05) DET from both brain regions of LBD brains. Table-4 presents the top 10 enriched biological processes and molecular functions, ranked by their FDR-corrected p-values, in LBD brains. SDC-5 presents all FDR-corrected significantly enriched biological processes, molecular functions and KEGG pathways in LBD, DLB and PDD brains, when compared to NDC brains. It also presents the FDR-corrected enriched biological processes, molecular functions and KEGG pathways in DLB brains, when compared to PDD brains. Proteins, translated by the identified LBD brain DET, were significantly enriched for 70 biological processes, 46 molecular functions and 63 KEGG pathways after FDR correction. They may contribute to LBD pathology by impacting DNA repair, signal transduction, apoptosis, protein phosphorylation, regulation of RNA transcription, vesicle-mediated transport, regulation of cell cycle, histone deacetylation, UPS, ALP, and the Wnt signalling pathway. Similarly, proteins, translated by the identified DLB brain DET, are likely to contribute to DLB pathology by affecting cell cycle, apoptosis, protein phosphorylation, proteolysis, regulation of RNA transcription, vesicle-mediated transport, ALP, and signal transduction.
aTop ten biological process and molecular function terms as ranked by their adjusted p-values; bNominally significant differentially expressed transcripts in either anterior cingulate or dorsolateral prefrontal cortices; cp-values adjusted by Benjamini-Hochberg false discovery rate correction at 5%; Analysis was completed using DAVID: https://david.ncifcrf.gov/. DAIVD functional analysis utilises a modified Fisher’s exact test without df to create an EASE score for gene enrichment analysis.
Discussion
This is the first transcriptome-wide study that analysed transcript-level differential expression and investigated alternatively spliced genes at transcript-level resolution in post-mortem LBD brains. It has identified 169 novel DET and 228 alternatively spliced genes in post-mortem LBD brains after appropriate FDR corrections. It has found specific transcripts, TMEM-205 and MICB-204, that were expressed exclusively in LBD samples. It confirms widespread alternative splicing in ACC of LBD brains (Feleke et al., Reference Feleke, Reynolds, Smith, Tilley, Taliun, Hardy, Matthews, Gentleman, Owen, Johnson, Srivastava and Ryten2021), and presents the first transcriptome-wide transcript-level differential expression analysis of LBD prefrontal cortices. However, the limitations of this study include its small sample size, lack of replication experiments such as quantitative real-time PCR, and the lack of single cell RNA-seq data. Moreover, functional enrichment analysis was based on translated proteins and that did not include the effects of identified non-coding DET. The current knowledge on the functions of individual non-coding RNA transcripts is very limited. When there are better methods for predicting the targets of identified non-coding DET, and a database for transcript-level functional knowledge of non-coding RNA, further functional enrichment analysis can be performed by including the identified non-coding DET.
Prior gene-level analysis of these RNA-seq data has identified 12 FDR-corrected DEG (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020). This transcriptome-wide DET analysis demonstrates the limitations of gene-level DEG analysis and the extent of novel insights that can be gained by transcript-level analyses. MPO was the top DEG, ranked by FDR-corrected p-values, in the ACC and it was one of the top three FDR-corrected DEG in the DLPFC of these LBD brains (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020). An MPO transcript, ENST00000225275.4 (MPO-201; Uniprot:P05164) was the only FDR-corrected DET that was significantly differentially expressed in both LBD brain regions. Downregulation of MPO-201 is likely to cause reduced translation of Myeloperoxidase protein. Myeloperoxidase plays an important role in oxidative stress and neuroinflammation, and it mediates proteolysis. MPO variants have been associated with AD (Reynolds et al., Reference Reynolds, Rhees, Maciejewski, Paladino, Sieburg, Maki and Masliah1999). Myeloperoxidase co-localizes with amyloid plaques in AD brains, and its plasma levels were reportedly higher in people with AD (Tzikas et al., Reference Tzikas, Schlak, Sopova, Gatsiou, Stakos, Stamatelopoulos, Stellos and Laske2014). Chronic neuroinflammation is well established in AD brains but several lines of evidence indicate the absence of chronic neuroinflammation in LBD, especially in DLB (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020; Amin et al., Reference Amin, Holmes, Dorey, Tommasino, Casal, Williams, Dupuy, Nicoll and Boche2020; Rajkumar et al., Reference Rajkumar, Hye, Lange, Manesh, Ballard, Fladby and Aarsland2021). The downregulation of MPO-201 may contribute to the differences in neuroinflammation in LBD brains, and functional consequences of this downregulated DET warrant further research. Moreover, GABRB3 encodes a subunit of gamma-aminobutyric acid (GABA) type-A receptor. A GABRB3 transcript, ENST00000638099.1 (GABRB3-223; Uniprot:A0A1B0GVW3), was significantly downregulated in the ACC of LBD brains (SDC-3). This finding, and previously reported genome-wide significant association of a GABRB3 variant (rs1426210) with LBD (Sanghvi et al., Reference Sanghvi, Singh, Morrin and Rajkumar2020), support the importance of GABAergic dysfunction in LBD pathology.
Designing oligonucleotide probes for measuring specific RNA transcripts is relatively easier than measuring overall expression levels of genes with multiple transcripts through targeted gene expression assays. Post-mortem DLB brain DEG have been found significantly enriched among the DEG identified in serum small extracellular vesicles (SEV) of people living with DLB (Rajkumar et al., Reference Rajkumar, Hye, Lange, Manesh, Ballard, Fladby and Aarsland2021). Statistically significant differential expression of many DEG in DLB brains could be measured in the blood-based SEV of people living with DLB (Rajkumar et al., Reference Rajkumar, Hye, Lange, Manesh, Ballard, Fladby and Aarsland2021). Hence, the identified FDR-corrected DET, especially those expressed exclusively in LBD brains, not only advance our understanding of LBD molecular pathology, but also warrant further investigation of their biomarker potential in people living with LBD.
TMEM18 was the top OFDR-corrected alternatively spliced gene in the DLPFC of LBD brains. TMEM18 encodes TMM18 protein. It regulates adipogenesis, gene silencing, and neuronal migration, as well as promoting neuroplasticity (Jurvansuu and Goldman, Reference Jurvansuu and Goldman2011, Jurvansuu et al., Reference Jurvansuu, Zhao, Leung, Boulaire, Yu, Ahmed and Wang2008; Luck et al., Reference Luck, Kim, Lambourne, Spirohn, Begg, Bian, Brignall, Cafarelli, Campos-Laborie, Charloteaux, Choi, Coté, Daley, Deimling, Desbuleux, Dricot, Gebbia, Hardy, Kishore, Knapp, Kovács, Lemmens, Mee, Mellor, Pollis, Pons, Richardson, Schlabach, Teeking, Yadav, Babor, Balcha, Basha, Bowman-Colin, Chin, Choi, Colabella, Coppin, D’Amata, De Ridder, De Rouck, Duran-Frigola, Ennajdaoui, Goebels, Goehring, Gopal, Haddad, Hatchi, Helmy, Jacob, Kassa, Landini, Li, van Lieshout, MacWilliams, Markey, Paulson, Rangarajan, Rasla, Rayhan, Rolland, San-Miguel, Shen, Sheykhkarimli, Sheynkman, Simonovsky, Taşan, Tejeda, Tropepe, Twizere, Wang, Weatheritt, Weile, Xia, Yang, Yeger-Lotem, Zhong, Aloy, Bader, De Las Rivas, Gaudet, Hao, Rak, Tavernier, Hill, Vidal, Roth and Calderwood2020). The TMEM18-205 transcript was found only in LBD samples. Alternative splicing of TMEM18 and significant upregulation of TMEM18-205 are likely to reduce TMM18 expression by increasing nonsense mediated decay in LBD brains. Impaired transcription repression due to reduced TMM18 levels may lead to upregulation of other pathogenic transcripts. Such gene regulatory changes, and the consequent dysfunctional molecular networks in LBD brains, should be investigated further. Moreover, TMM18 interacts with two proteins, RETR3 (Uniprot:Q86VR2) and REEP4 (Uniprot:Q9H6H4) (Luck et al., Reference Luck, Kim, Lambourne, Spirohn, Begg, Bian, Brignall, Cafarelli, Campos-Laborie, Charloteaux, Choi, Coté, Daley, Deimling, Desbuleux, Dricot, Gebbia, Hardy, Kishore, Knapp, Kovács, Lemmens, Mee, Mellor, Pollis, Pons, Richardson, Schlabach, Teeking, Yadav, Babor, Balcha, Basha, Bowman-Colin, Chin, Choi, Colabella, Coppin, D’Amata, De Ridder, De Rouck, Duran-Frigola, Ennajdaoui, Goebels, Goehring, Gopal, Haddad, Hatchi, Helmy, Jacob, Kassa, Landini, Li, van Lieshout, MacWilliams, Markey, Paulson, Rangarajan, Rasla, Rayhan, Rolland, San-Miguel, Shen, Sheykhkarimli, Sheynkman, Simonovsky, Taşan, Tejeda, Tropepe, Twizere, Wang, Weatheritt, Weile, Xia, Yang, Yeger-Lotem, Zhong, Aloy, Bader, De Las Rivas, Gaudet, Hao, Rak, Tavernier, Hill, Vidal, Roth and Calderwood2020), both of which are associated with Endoplasmic Reticulum (ER) morphology (Kumar et al., Reference Kumar, Golchoubian, Belevich, Jokitalo and Schlaitz2019). ER dysfunction and consequent increased activation of the unfolded protein response may contribute to LBD pathology (Baek et al., Reference Baek, Whitfield, Howlett, Francis, Bereczki, Ballard, Hortobagyi, Attems and Aarsland2016). Furthermore, reduced expression of TMM18 is likely to impair migration of neural stem cells and neuroplasticity (Jurvansuu et al., Reference Jurvansuu, Zhao, Leung, Boulaire, Yu, Ahmed and Wang2008). Such impaired neuroplasticity may lead to neurodegeneration and cognitive decline in LBD (Toricelli et al., Reference Toricelli, Pereira, Souza Abrao, Malerba, Maia, Buck and Viel2021).
Identified alternative splicing of MICB may lead to a greater proportion of MHC proteins without a signal peptide. This may lead to signal peptide-dependent inhibition of protein translocation, and reduced expression of the functional MHC class-I protein (Powers and Fruh, Reference Powers and Fruh2008) in the DLPFC of LBD brains. MICB was significantly alternatively spliced only in DLB brains and not in PDD brains. MICB shares location (6p21.33) and function with a previously reported DEG, HLA-B, in LBD brains (Feleke et al., Reference Feleke, Reynolds, Smith, Tilley, Taliun, Hardy, Matthews, Gentleman, Owen, Johnson, Srivastava and Ryten2021; Cunningham et al., Reference Cunningham, Allen, Allen, Alvarez-Jarreta, Amode, Armean, Austine-Orimoloye, Azov, Barnes, Bennett, Berry, Bhai, Bignell, Billis, Boddu, Brooks, Charkhchi, Cummins, Da Rin Fioretto, Davidson, Dodiya, Donaldson, El Houdaigui, El Naboulsi, Fatima, Giron, Genez, Martinez, Guijarro-Clarke, Gymer, Hardy, Hollis, Hourlier, Hunt, Juettemann, Kaikala, Kay, Lavidas, Le, Lemos, Marugán, Mohanan, Mushtaq, Naven, Ogeh, Parker, Parton, Perry, Piližota, Prosovetskaia, Sakthivel, Salam, Schmitt, Schuilenburg, Sheppard, Pérez-Silva, Stark, Steed, Sutinen, Sukumaran, Sumathipala, Suner, Szpak, Thormann, Tricomi, Urbina-Gómez, Veidenberg, Walsh, Walts, Willhoft, Winterbottom, Wass, Chakiachvili, Flint, Frankish, Giorgetti, Haggerty, Hunt, IIsley, Loveland, Martin, Moore, Mudge, Muffato, Perry, Ruffier, Tate, Thybert, Trevanion, Dyer, Harrison, Howe, Yates, Zerbino and Flicek2022). The MICB protein (Uniprot:Q29980) is stress induced and it activates the cytolytic response of natural killer cells (Baranwal and Mehra, Reference Baranwal and Mehra2017). Upregulation of MICB and consequent neuroinflammation have been reported in AD brains (Garranzo-Asensio et al., Reference Garranzo-Asensio, San Segundo-Acosta, Martinez-Useros, Montero-Calle, Fernandez-Acenero, Haggmark-Manberg, Pelaez-Garcia, Villalba, Rabano, Nilsson and Barderas2018). Alternative splicing of MICB may lead to reduced cytolytic response of natural killer cells in the DLPFC of DLB brains, and this may contribute to the absence of chronic neuroinflammation in DLB brains (Rajkumar et al., Reference Rajkumar, Bidkhori, Shoaie, Clarke, Morrin, Hye, Williams, Ballard, Francis and Aarsland2020; Amin et al., Reference Amin, Holmes, Dorey, Tommasino, Casal, Williams, Dupuy, Nicoll and Boche2020).
Prior candidate gene expression studies that investigated the alternatively spliced isoforms of SNCA, SNCB, and APP have reported significant associations with LBD (Beyer et al., Reference Beyer, Lao, Carrato, Mate, Lopez, Ferrer and Ariza2004; Beyer et al., Reference Beyer, Domingo-Sabat, Santos, Tolosa, Ferrer and Ariza2010; Funahashi et al., Reference Funahashi, Yoshino, Yamazaki, Mori, Mori, Ozaki, Sao, Ochi, Iga and Ueno2017; Low et al., Reference Low, Lee, Lim, Lee, Ballard, Francis, Lai and Tan2021), and those alternative splicing findings have not been replicated by our transcriptome-wide analysis. However, it replicated significant alternative splicing of another previously reported alternatively spliced gene, FYN, in LBD brains (Beyer et al., Reference Beyer, Lao, Carrato, Mate, Lopez, Ferrer and Ariza2004). FYN is associated with cell survival and immune response (Low et al., Reference Low, Lee, Lim, Lee, Ballard, Francis, Lai and Tan2021). Prior studies have suggested that FYN alternative splicing is driven by increased expression of isoforms that are primarily expressed in T-cells (FynT) (Low et al., Reference Low, Lee, Lim, Lee, Ballard, Francis, Lai and Tan2021). We did not find statistically significant differential expression of FynT transcripts, and alternative splicing of FYN was likely due to upregulation of FYN-218 (ENST00000517419.5). This transcript was an FDR-corrected DET in the DLPFC of DLB brains, and little is known about its function. Moreover, identified alternatively spliced genes include GOLGA2, ABCB9 and RHBDD1, which may contribute to protein aggregation in LBD brains (Baek et al., Reference Baek, Whitfield, Howlett, Francis, Bereczki, Ballard, Hortobagyi, Attems and Aarsland2016).
Identified protein coding DET in LBD brains were significantly enriched for several biological processes and molecular pathways that are relevant to α-synuclein aggregation, Lewy body formation, and neurodegeneration. α-synuclein modulates DNA repair, and DNA repair deficits contribute to neurodegeneration in LBD (Schaser et al., Reference Schaser, Osterberg, Dent, Stackhouse, Wakeham, Boutros, Weston, Owen, Weissman, Luna, Raber, Luk, McCullough, Woltjer and Unni2019). Identified DET are also likely to impact Ubiquitin protein ligase binding, which is essential for the post-translational modification of proteins and the removal of misfolded protein aggregates (Zhang et al., Reference Zhang, Li and Li2019). Changes in Ubiquitin ligase binding and UPS dysfunction may impact the post-translational modification of α-synuclein, and could promote Lewy body formation (Zhang et al., Reference Zhang, Li and Li2019). Further research investigating the contributions of the identified DET towards dysfunctional apoptosis, protein phosphorylation, RNA transcription, vesicle-mediated transport, the UPS, the ALP, and Wnt signalling in LBD may facilitate discovery of novel therapeutic targets. When compared to the functional understanding of genes and proteins, the current understanding of functions of individual RNA transcripts is limited. Fortunately, transcript-level functional knowledge is quickly expanding, and this will help functional interpretation of DET analysis in the future. The DET identified within this study should be verified with independent biological replicates and be the subject of future investigations focussing on their dysfunctional molecular networks and functional consequences. Further functional insight can be gathered by multi-omic analysis that combines transcript-level differential expression analysis with genomic, proteomic, and/or epigenetic data. Moreover, investigating the potential of the identified DET in mitigating α-synuclein aggregation in appropriate in vitro and in vivo models will enhance their clinical translation into potential novel therapeutic targets for LBD and other synucleinopathies.
We cannot overemphasise the need for adequately powered transcriptome-wide transcript-level studies investigating LBD brains and blood-based SEV of people living with LBD. If such studies include comparisons with AD or other dementia, they can facilitate discovery of diagnostic biomarkers for LBD. Plasma SEV RNA can be novel blood-based diagnostic biomarkers for DLB (Rajkumar et al., Reference Rajkumar, Hye, Lange, Manesh, Ballard, Fladby and Aarsland2021). A prior study has demonstrated statistically significant overlap between post-mortem DLB brain DEG and the DEG, identified in blood-based SEV from people living with DLB (Rajkumar et al., Reference Rajkumar, Hye, Lange, Manesh, Ballard, Fladby and Aarsland2021). Parallelly measuring multiple RNA biomarkers together may improve their diagnostic accuracy, and the identified DET in LBD set the stage for developing potential multiplex RNA diagnostic biomarker assays for LBD. Moreover, unlike DNA polymorphisms, RNA expression levels remain dynamic during disease progression. Hence, investigating blood-based expression levels of the identified DET at various clinical stages of LBD may aid early diagnosis of LBD and may lead to the discovery of novel prognostic biomarkers for LBD.
Acknowledgements
We thank the Brains for Dementia Research (BDR) network of brain banks for providing the necessary post-mortem brain tissues. The Southwest dementia brain bank, Bristol, UK, is a part of the BDR programme, jointly funded by the Alzheimer’s Research UK and the Alzheimer’s Society, and is supported by the Bristol Research into Alzheimer’s and Care of the Elderly, and the Medical Research Council (MRC), UK. We thank the donors whose donation of brain tissue to the London Neurodegenerative Diseases Brain Bank allowed this work to take place. The London brain bank is supported by the MRC and the BDR. The Manchester brain bank, a part of the BDR, receives service support costs from the MRC. We acknowledge the Oxford brain bank, supported by the MRC, the NIHR Oxford Biomedical Research Centre, and the BDR programme for providing post-mortem specimens.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/neu.2024.65.
Data availability
Additional data including raw RNA-seq data files, transcript count matrix files and transcript level alternative splicing proportions are available upon reasonable request to the corresponding author.
Authors’ contributions
TG, KB, and APR were involved in the conception and design of this research. PF and APR obtained the post-mortem brain tissues and extracted RNA samples. TG and APR analysed the RNA-sequencing data and completed subsequent functional analysis. TG, KB, and APR drafted the initial manuscript. All authors were involved in the critical revisions and final approval of the manuscript.
Funding statement
RNA-seq of the post-mortem brains was funded by the Biomedical Research Unit for Dementia (BRU-D) and the Maudsley Biomedical Research Centre – dementia theme at the King’s College London, London, UK. Transcriptome-wide alternative splicing and transcript level differential expression analysis of the RNA-seq data was supported by a Biotechnology and Biological Sciences Research Council Doctoral Training Programme (BBSRC-DTP) Ph.D. fellowship in the University of Nottingham, Nottingham, UK. The Newcastle brain tissue resource, another part of the BDR, is funded in part by a grant from the MRC (G0400074).
Competing interests
Prof. Dag Aarsland has received research support and/or honoraria from Astra-Zeneca, H. Lundbeck, Novartis Pharmaceuticals, and GE Health, and serves as paid consultant for H. Lundbeck, Eisai, and Axovant. Other authors do not have any competing interests to declare. The funding bodies did not play any role in the design, collection, analysis, and interpretation of data, and in writing of the manuscript.