Intron retention serves as a blueprint for RNA binding protein mislocalisation in amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis (ALS) is incurable and fatal. Familial (10%) and sporadic (90%) ALS are largely clinically indistinguishable, and are characterised by premature degeneration of upper and lower motor neurons. Average survival is between three to five years and patients become increasingly paralysed, losing the ability to speak, eat and breathe. The devastating nature of ALS means there is an urgent need to understand disease mechanisms in order to develop effective therapies.

Development and homeostasis of neurons relies on the precise implementation of cell-type and stage-specific alternative RNA processing, coordinated by the action of trans-acting RNA binding proteins (RBPs), that bind specific sites within precursor RNAs and mRNA itself to form ribonucleoprotein complexes (RNPs). RNPs mediate splicing, nuclear export and localisation of mRNAs; accumulating evidence implicates RNP formation as a key regulator of both neurodevelopment and specific forms of neurodegeneration, including ALS.

More than 20 mutations in genes of varying function have been identified in familial ALS, but the common downstream molecular denominators in both familial and sporadic ALS are defective RNPs.

There is strong genetic evidence implicating aberrant RBPs in ALS, including disease-causing mutations in the RBPs Fused in Sarcoma (FUS) and Transactive response DNA Binding protein 43 (TDP-43). Further, the pathological hallmark in 97% of all ALS cases, irrespective of their original mutation, is mislocalisation of TDP-43 from the nucleus to the cytoplasm, where it becomes abnormally phosphorylated and cleaved, and forms insoluble protein inclusions.

Studies of ALS using animal models and human or animal cells in tissue culture rely on overexpression or knockdown of known mutated genes. While useful, these surrogates cannot precisely recapitulate the clinical pathophysiological state of ALS patients. In 2007, a new approach became possible with the advent of human induced pluripotent stem cells (iPSCs). It was now feasible to make patient-specific iPSCs from donated fibroblasts, and use directed differentiation to make spinal cord motor neurons in vitro. Such cells are the closest possible model for an individual ALS patient’s disease, and provide a platform to conduct primary discovery research in a patient model and then do secondary validation in mouse or tissue culture, rather than the opposite, as had been the case previously.

Generation of motor neurons from iPSCs proved to be beset with difficulties. There was low enrichment of desired cell types — 20-50% — and it typically took between 70 and 100 days to differentiate fibroblasts into motor neurons. However, a paper in 2017 from Rickie Patani and his Crick colleagues, including fellow neurologist Sonia Gandhi and RNA specialist Jernej Ule, solved the problem (Hall et al, 2017). By forcing differentiation by innovative use of small molecule inhibitors of the activin/nodal, BMP4 and GSK3β pathways, followed by patterning with retinoic acid and a sonic hedgehog antagonist, Patani’s method took only three weeks, and produced more than 85% pure motor neurons, which were functionally equivalent to motor neurons in situ, as assessed by stringent analysis of molecular markers, electrical activity, and formation of neuromuscular junctions when co-cultured with skeletal muscle.

With a protocol now in place, tests of patient-derived iPSCs from familial ALS subtypes showed that those from patients with mutations in Valosin-Containing Protein (VCP) gave the most robust and severe phenotype in the differentiated motor neurons. VCP-mutant MNs displayed increased cell death, disrupted synapse formation with concomitant transcriptional perturbation of genes encoding ion channels and synapse structure or assembly, and electrophysiologically, an overall decrease in activity and burst firing.

The developmental nature of the model meant that in addition to documenting these processes, it was also possible to pinpoint the earliest events in MN degeneration, and then track them over time. Although no defects were observed during initial differentiation, within three days of terminal VCP-mutant MN differentiation, there was an increase in aberrant cytoplasmic TDP-43, the hallmark of ALS, together with an increase in ER stress, later followed by mitochondrial dysfunction, increased oxidative stress, reduced synaptic density and cell death. Interestingly, non-cell-autonomous effects were also detected as VCP-mutant astrocytes also degenerated, and their ability to promote survival of MNs was compromised.

This landmark study demonstrated the potential of using directed differentiation of iPSCs to confidently identify primary molecular pathogenic events in ALS patient-derived MNs. Building on this work, Patani, again with Ule’s support, was now in a position to examine the RNA processing events thought to be disturbed in ALS. Accordingly, his group began a detailed analysis of post-transcriptional processing of mRNA over the timecourse of differentiation of VCP-mutant iPSCs to motor neurons, in collaboration with Crick bioinformatics and computational biology expert Nick Luscombe, and Luscombe’s postdoc Raphaelle Luisier (Luisier et al, 2018).

Using RNA-seq to examine alternative RNA processing events, Patani and co-workers found that differential use of exons and intron skipping events increased during progressive lineage restriction to MNs, but there were no significant differences between normal and VCP-mutant MNs. However, intron retention (IR), which accounted for more than 65% of alternative splicing events as cells differentiated from neuronal precursors to MNs, was strikingly increased in VCP mutant samples. While 60% of IR events matched those in normal cultures in both timing and identity, the remainder were either unique to VCP mutant cultures, or occurred at an earlier timepoint.

Rigorous manual curation of all the IR events generated a shortlist of 167 introns, all of which were more highly retained in VCP-mutant neuronal precursors than in control samples. All were longer and more conserved than non-retained introns from the same gene set.

To determine whether aberrant IR was found not just in VCP-mutant cells but in other genetic forms of ALS, Patani and colleagues looked at datasets from mutant cells carrying the SOD1 and FUS ALS mutations. Confirming that the IR phenotype was generalisable, they found aberrantly increased IR in both subtypes, and were further able to show that there was significant overlap in the genes involved, which encoded proteins predominantly involved in RNA splicing and translation. These data were of particular significance as the canonical ALS hallmark of TDP-43 mislocalisation is absent in the SOD1 and FUS forms of the disease, implying that aberrant IR might be a more universal marker.

Despite these striking post-transcriptional defects, the overall gene and protein expression programme during motor neurogenesis was not markedly affected by the VCP mutation. Therefore, the team turned to analysis of the most highly upregulated IR event shared between all the ALS subtypes examined: the intron 9/9-retaining form of the mRNA for the SFPQ gene, which encodes a protein with key roles in transcription, splicing, 3’ end processing and axon viability. Using 176 eCLIP datasets (which identify RBPs bound to RNA by crosslinking and immunoprecipitation analysis) generated from human cell lines, they showed that the SFPQ protein itself bound directly to its own 9/9 intron, and also to another aberrantly retained intron in the FUS gene.

There was little change in SFPQ protein expression in VCP mutant cultures relative to controls, but together with its intron-retaining mRNA isoform, SFPQ’s distribution had changed: there was aberrant accumulation of both the protein and mRNA members of the complex in the cytoplasm, accompanied by nuclear loss, and this was generalisable across all in vivo mouse transgenic models tested, and in human post-mortem tissue from sporadic ALS cases.

Nuclear-to-cytoplasmic mislocalisation of SFPQ appears to be a unifying hallmark of ALS, across diverse in vitro and in vivo models representing familial and sporadic forms of ALS, as well as in patient post-mortem tissue. Further, the export of the 9/9 intron isoform of SPQR to the cytoplasm is not an isolated event: recent work from Patani’s and Luscombe’s groups (Tyzack et al, 2021) has shown that there are over 100 cytoplasmic IR transcripts with increased abundance in ALS samples. Remarkably, TDP-43, SFPQ and FUS avidly and specifically bind to this aberrant cytoplasmic pool, suggesting an intriguing and hitherto unrecognised regulatory role for cytoplasmic IR transcripts in both neurological health and disease. The team has also demonstrated that FUS mislocalisation is widespread in ALS, something that had previously been missed, as FUS does not form TDP-43-like cytoplasmic aggregates (Tyzack et al, 2019).

These findings are clearly of great importance from both neurobiological and clinical standpoints, and Patani’s and Luscombe’s groups are now focussed on resolving the many mechanistic questions that have been raised by their discoveries.