Transcriptional Heterogeneity in ALS: Patients with Retrovirus-mediated Disease and Potential Cell-type Specific Therapeutic Approaches

Amyotrophic lateral sclerosis (ALS) is a universally fatal neurodegenerative disease affecting both upper and lower motor neurons. There is currently no cure for ALS and most clinical trials treat all ALS patients as a single group. The hypothesis addressed in this thesis is that ALS is a transcriptionally heterogeneous disease with distinct biological pathways converging on a similar phenotype. To address this hypothesis, a bulk RNA sequencing (RNA-seq) dataset from almost 2,000 ALS and unaffected control samples was leveraged to uncover transcriptionally distinct patient subpopulations. Using non-negative matrix factorization (NMF), an unsupervised clustering algorithm, four distinct clusters of ALS patients and controls were identified in the cortex (CTX) and spinal cord (SC). Differential expression analysis (DEA) revealed that certain loci associated with human endogenous retrovirus K (HERV-K), a type of transposable element (TE), as well as other non-TE features were dysregulated in patient samples compared to controls. In both the CTX and SC, around 20% of ALS patients had higher overall HERV-K expression than other patients and controls (high HERV-K ALS). Additionally, in both the CTX and SC, there was an NMF cluster that was enriched for high HERV-K ALS patients relative to other patients. Moreover, the high HERV-K ALS patients had unique transcriptomic signatures at the individual feature and biological pathway levels relative to other ALS patients and controls. There were four robustly upregulated HERV-K loci (1q21.3, 5q15, 8q24.3, and 11q13.4) that can encode for truncated HERV-K envelope protein (env) in high HERV-K ALS patients. These truncated peptides generally corresponded to the transmembrane domain of the full-length env, implying it is neuropathological in ALS. Moreover, the DEA results were validated using a regularised logistic regression, a type of supervised machine learning, classifier for ALS patient or control samples. Finally, the cell type specificity of these results was determined using a publicly available single nucleus RNA sequencing (snRNA-seq) dataset consisting of 23 ALS patients and 17 age- and biological sex- matched controls. Carbonic anhydrase 1, CA1, and the sulfate transporter, SLC13A4, are the most promising targets as they were the most robustly dysregulated protein coding genes across analyses and are also specifically upregulated in upper motor neurons in ALS patients. In summary, these results demonstrate that ALS is a transcriptionally heterogeneous disease, and patient subpopulation- and cell-type-specific approaches will be needed to effectively treat ALS neuropathology.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2023

Author dc:creator

• Pasternack, Nicholas

Advisors dc:contributor.advisor

• Paulsen, Ole
• Nath, Avindra

• Amyotrophic lateral sclerosis
• Endogenous retroviruses
• Machine learning
• Neuroscience
• RNA sequencing

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/9fb93a33-049b-48c7-bc64-ff3b1849a085/download
• https://www.rioxx.net/licenses/all-rights-reserved/

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.102260

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/358838