Drosophila Vap-33 is required for axonal localization of Dscam isoforms

Gene targeting in amyotrophic lateral sclerosis using causality-based feature selection and machine learning

BACKGROUND

Amyotrophic lateral sclerosis (ALS) is a rare progressive neurodegenerative disease that affects upper and lower motor neurons. As the molecular basis of the disease is still elusive, the development of high-throughput sequencing technologies, combined with data mining techniques and machine learning methods, could provide remarkable results in identifying pathogenetic mechanisms. High dimensionality is a major problem when applying machine learning techniques in biomedical data analysis, since a huge number of features is available for a limited number of samples. The aim of this study was to develop a methodology for training interpretable machine learning models in the classification of ALS and ALS-subtypes samples, using gene expression datasets.

METHODS

We performed dimensionality reduction in gene expression data using a semi-automated preprocessing systematic gene selection procedure using Statistically Equivalent Signature (SES), a causality-based feature selection algorithm, followed by Boosted Regression Trees (XGBoost) and Random Forest to train the machine learning classifiers. The SHapley Additive exPlanations (SHAP values) were used for interpretation of the machine learning classifiers. The methodology was developed and tested using two distinct publicly available ALS RNA-seq datasets. We evaluated the performance of SES as a dimensionality reduction method against: (a) Least Absolute Shrinkage and Selection Operator (LASSO), and (b) Local Outlier Factor (LOF).

RESULTS

The proposed methodology achieved 85.18% accuracy for the classification of cerebellum or frontal cortex samples as C9orf72-related familial ALS, sporadic ALS or healthy samples. Importantly, the genes identified as the most determinative have also been reported as disease-associated in ALS literature. When tested in the evaluation dataset, the methodology achieved 88.89% accuracy for the classification of sporadic ALS motor neuron samples. When LASSO was used as feature selection method instead of SES, the accuracy of the machine learning classifiers ranged from 74.07 to 96.30%, depending on tissue assessed, while LOF underperformed significantly (77.78% accuracy for the classification of pooled cerebellum and frontal cortex samples).

CONCLUSIONS

Using SES, we addressed the challenge of high dimensionality in gene expression data analysis, and we trained accurate machine learning ALS classifiers, specific for the gene expression patterns of different disease subtypes and tissue samples, while identifying disease-associated genes.

Trans-splicing facilitated by RNA pairing greatly expands sDscam isoform diversity but not homophilic binding specificity

The Down syndrome cell adhesion molecule 1 (Dscam1) gene can generate tens of thousands of isoforms via alternative splicing, which is essential for nervous and immune functions. Chelicerates generate approximately 50 to 100 shortened Dscam (sDscam) isoforms by alternative promoters, similar to mammalian protocadherins. Here, we reveal that trans-splicing markedly increases the repository of sDscamβ isoforms in Tetranychus urticae. Unexpectedly, every variable exon cassette engages in trans-splicing with constant exons from another cluster. Moreover, we provide evidence that competing RNA pairing not only governs alternative cis-splicing but also facilitates trans-splicing. Trans-spliced sDscam isoforms mediate cell adhesion ability but exhibit the same homophilic binding specificity as their cis-spliced counterparts. Thus, we reveal a single sDscam locus that generates diverse adhesion molecules through cis- and trans-splicing coupled with alternative promoters. These findings expand understanding of the mechanism underlying molecular diversity and have implications for the molecular control of neuronal and/or immune specificity.

May the Odds Be Ever in Your Favor: Non-deterministic Mechanisms Diversifying Cell Surface Molecule Expression

How does the information in the genome program the functions of the wide variety of cells in the body? While the development of biological organisms appears to follow an explicit set of genomic instructions to generate the same outcome each time, many biological mechanisms harness molecular noise to produce variable outcomes. Non-deterministic variation is frequently observed in the diversification of cell surface molecules that give cells their functional properties, and is observed across eukaryotic clades, from single-celled protozoans to mammals. This is particularly evident in immune systems, where random recombination produces millions of antibodies from only a few genes; in nervous systems, where stochastic mechanisms vary the sensory receptors and synaptic matching molecules produced by different neurons; and in microbial antigenic variation. These systems employ overlapping molecular strategies including allelic exclusion, gene silencing by constitutive heterochromatin, targeted double-strand breaks, and competition for limiting enhancers. Here, we describe and compare five stochastic molecular mechanisms that produce variety in pathogen coat proteins and in the cell surface receptors of animal immune and neuronal cells, with an emphasis on the utility of non-deterministic variation.

Axonal Endoplasmic Reticulum Dynamics and Its Roles in Neurodegeneration

The physical continuity of axons over long cellular distances poses challenges for their maintenance. One organelle that faces this challenge is endoplasmic reticulum (ER); unlike other intracellular organelles, this forms a physically continuous network throughout the cell, with a single membrane and a single lumen. In axons, ER is mainly smooth, forming a tubular network with occasional sheets or cisternae and low amounts of rough ER. It has many potential roles: lipid biosynthesis, glucose homeostasis, a Ca2+ store, protein export, and contacting and regulating other organelles. This tubular network structure is determined by ER-shaping proteins, mutations in some of which are causative for neurodegenerative disorders such as hereditary spastic paraplegia (HSP). While axonal ER shares many features with the tubular ER network in other contexts, these features must be adapted to the long and narrow dimensions of axons. ER appears to be physically continuous throughout axons, over distances that are enormous on a subcellular scale. It is therefore a potential channel for long-distance or regional communication within neurons, independent of action potentials or physical transport of cargos, but involving its physiological roles such as Ca2+ or organelle homeostasis. Despite its apparent stability, axonal ER is highly dynamic, showing features like anterograde and retrograde transport, potentially reflecting continuous fusion and breakage of the network. Here we discuss the transport processes that must contribute to this dynamic behavior of ER. We also discuss the model that these processes underpin a homeostatic process that ensures both enough ER to maintain continuity of the network and repair breaks in it, but not too much ER that might disrupt local cellular physiology. Finally, we discuss how failure of ER organization in axons could lead to axon degenerative diseases, and how a requirement for ER continuity could make distal axons most susceptible to degeneration in conditions that disrupt ER continuity.

Sphingolipid-dependent Dscam sorting regulates axon segregation

Neurons are highly polarized cells with distinct protein compositions in axonal and dendritic compartments. Cellular mechanisms controlling polarized protein sorting have been described for mature nervous system but little is known about the segregation in newly differentiated neurons. In a forward genetic screen for regulators of Drosophila brain circuit development, we identified mutations in SPT, an evolutionary conserved enzyme in sphingolipid biosynthesis. Here we show that reduced levels of sphingolipids in SPT mutants cause axonal morphology defects similar to loss of cell recognition molecule Dscam. Loss- and gain-of-function studies show that neuronal sphingolipids are critical to prevent aggregation of axonal and dendritic Dscam isoforms, thereby ensuring precise Dscam localization to support axon branch segregation. Furthermore, SPT mutations causing neurodegenerative HSAN-I disorder in humans also result in formation of stable Dscam aggregates and axonal branch phenotypes in Drosophila neurons, indicating a causal link between developmental protein sorting defects and neuronal dysfunction.

Little is known about the initial segregation of axonal and dendritic proteins during the differentiation of newly generated neurons. Here authors use a forward genetic screen to identify the role of sphingolipids in regulating the sub-cellular distribution of Dscam for neuronal patterning in Drosophila Mushroom Bodies

Amyotrophic Lateral Sclerosis Model

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease that affects upper and lower motor neurons in the brain and the spinal cord. Due to the progressive neurodegeneration, ALS leads to paralysis and death caused by respiratory failure 2-5 years after the onset of symptoms. There is no effective cure available. Most ALS cases are sporadic, without family history, whereas 10% of the cases are familial. Identification of variants in more than 30 different loci has provided insight into the pathogenic molecular mechanisms mediating disease pathogenesis. Studies of a Drosophila melanogaster model for each of the ALS genes can contribute to uncovering pathophysiological mechanism of ALS and finding targets of the disease-modifying therapy. In this review, we focus on three ALS-causing genes: TAR DNA-binding protein (TDP-43), fused in sarcoma/translocated in liposarcoma (FUS/TLS), and chromosome 9 open reading frame 72 (C9orf72).

Insights into ALS pathomechanisms: from flies to humans

Amyotrophic Lateral Sclerosis (ALS) is a devastating neurodegenerative disease causing the death of motor neurons with consequent muscle atrophy and paralysis. Several neurodegenerative diseases have been modeled in Drosophila and genetic studies on this model organism led to the elucidation of crucial aspects of disease mechanisms. ALS, however, has lagged somewhat behind possibly because of the lack of a suitable genetic model. We were the first to develop a fly model for ALS and over the last few years, we have implemented and used this model for a large scale, unbiased modifier screen. We also report an extensive bioinformatic analysis of the genetic modifiers and we show that most of them are associated in a network of interacting genes controlling known as well as novel cellular processes involved in ALS pathogenesis. A similar analysis for the human homologues of the Drosophila modifiers and the validation of a subset of them in human tissues confirm and expand the significance of the data for the human disease. Finally, we analyze a possible application of the model in the process of therapeutic discovery in ALS and we discuss the importance of novel “non-obvious” models for the disease.

Autophagy and Neurodegeneration: Insights from a Cultured Cell Model of ALS

Autophagy plays a major role in the elimination of cellular waste components, the renewal of intracellular proteins and the prevention of the build-up of redundant or defective material. It is fundamental for the maintenance of homeostasis and especially important in post-mitotic neuronal cells, which, without competent autophagy, accumulate protein aggregates and degenerate. Many neurodegenerative diseases are associated with defective autophagy; however, whether altered protein turnover or accumulation of misfolded, aggregate-prone proteins is the primary insult in neurodegeneration has long been a matter of debate. Amyotrophic lateral sclerosis (ALS) is a fatal disease characterized by selective degeneration of motor neurons. Most of the ALS cases occur in sporadic forms (SALS), while 10%–15% of the cases have a positive familial history (FALS). The accumulation in the cell of misfolded/abnormal proteins is a hallmark of both SALS and FALS, and altered protein degradation due to autophagy dysregulation has been proposed to contribute to ALS pathogenesis. In this review, we focus on the main molecular features of autophagy to provide a framework for discussion of our recent findings about the role in disease pathogenesis of the ALS-linked form of the VAPB gene product, a mutant protein that drives the generation of unusual cytoplasmic inclusions.

Amyotrophic lateral sclerosis-linked mutant VAPB inclusions do not interfere with protein degradation pathways or intracellular transport in a cultured cell model

VAPB is a ubiquitously expressed, ER-resident adaptor protein involved in interorganellar lipid exchange, membrane contact site formation, and membrane trafficking. Its mutant form, P56S-VAPB, which has been linked to a dominantly inherited form of Amyotrophic Lateral Sclerosis (ALS8), generates intracellular inclusions consisting in restructured ER domains whose role in ALS pathogenesis has not been elucidated. P56S-VAPB is less stable than the wild-type protein and, at variance with most pathological aggregates, its inclusions are cleared by the proteasome. Based on studies with cultured cells overexpressing the mutant protein, it has been suggested that VAPB inclusions may exert a pathogenic effect either by sequestering the wild-type protein and other interactors (loss-of-function by a dominant negative effect) or by a more general proteotoxic action (gain-of-function). To investigate P56S-VAPB degradation and the effect of the inclusions on proteostasis and on ER-to-plasma membrane protein transport in a more physiological setting, we used stable HeLa and NSC34 Tet-Off cell lines inducibly expressing moderate levels of P56S-VAPB. Under basal conditions, P56S-VAPB degradation was mediated exclusively by the proteasome in both cell lines, however, it could be targeted also by starvation-stimulated autophagy. To assess possible proteasome impairment, the HeLa cell line was transiently transfected with the ERAD (ER Associated Degradation) substrate CD3δ, while autophagic flow was investigated in cells either starved or treated with an autophagy-stimulating drug. Secretory pathway functionality was evaluated by analyzing the transport of transfected Vesicular Stomatitis Virus Glycoprotein (VSVG). P56S-VAPB expression had no effect either on the degradation of CD3δ or on the levels of autophagic markers, or on the rate of transport of VSVG to the cell surface. We conclude that P56S-VAPB inclusions expressed at moderate levels do not interfere with protein degradation pathways or protein transport, suggesting that the dominant inheritance of the mutant gene may be due mainly to haploinsufficiency.

A genetic screen identifies Tor as an interactor of VAPB in a Drosophila model of amyotrophic lateral sclerosis

Amyotrophic Lateral Sclerosis (ALS) is a progressive neurodegenerative disorder characterized by selective death of motor neurons. In 5–10% of the familial cases, the disease is inherited because of mutations. One such mutation, P56S, was identified in human VAPB that behaves in a dominant negative manner, sequestering wild type protein into cytoplasmic inclusions.

We have conducted a reverse genetic screen to identify interactors of Drosophila VAPB. We screened 2635 genes and identified 103 interactors, of which 45 were enhancers and 58 were suppressors of VAPB function. Interestingly, the screen identified known ALS loci – TBPH, alsin2 and SOD1. Also identified were genes involved in cellular energetics and homeostasis which were used to build a gene regulatory network of VAPB modifiers. One key modifier identified was Tor, whose knockdown reversed the large bouton phenotype associated with VAP(P58S) expression in neurons. A similar reversal was seen by over-expressing Tuberous Sclerosis Complex (Tsc1,2) that negatively regulates TOR signaling as also by reduction of S6K activity. In comparison, the small bouton phenotype associated with VAP(wt) expression was reversed with Tsc1 knock down as well as S6K-CA expression. Tor therefore interacts with both VAP(wt) and VAP(P58S), but in a contrasting manner. Reversal of VAP(P58S) bouton phenotypes in larvae fed with the TOR inhibitor Rapamycin suggests upregulation of TOR signaling in response to VAP(P58S) expression.

The VAPB network and further mechanistic understanding of interactions with key pathways, such as the TOR cassette, will pave the way for a better understanding of the mechanisms of onset and progression of motor neuron disease.

A fruitful endeavor: modeling ALS in the fruit fly

For over a century Drosophila melanogaster, commonly known as the fruit fly, has been instrumental in genetics research and disease modeling. In more recent years, it has been a powerful tool for modeling and studying neurodegenerative diseases, including the devastating and fatal amyotrophic lateral sclerosis (ALS). The success of this model organism in ALS research comes from the availability of tools to manipulate gene/protein expression in a number of desired cell-types, and the subsequent recapitulation of cellular and molecular phenotypic features of the disease. Several Drosophila models have now been developed for studying the roles of ALS-associated genes in disease pathogenesis that allowed us to understand the molecular pathways that lead to motor neuron degeneration in ALS patients. Our primary goal in this review is to highlight the lessons we have learned using Drosophila models pertaining to ALS research. This article is part of a Special Issue entitled ALS complex pathogenesis.

Gain-of-function mutations in the ALS8 causative gene VAPB have detrimental effects on neurons and muscles

Amyotrophic Lateral Sclerosis (ALS) is a motor neuron degenerative disease characterized by a progressive, and ultimately fatal, muscle paralysis. The human VAMP-Associated Protein B (hVAPB) is the causative gene of ALS type 8. Previous studies have shown that a loss-of-function mechanism is responsible for VAPB-induced ALS. Recently, a novel mutation in hVAPB (V234I) has been identified but its pathogenic potential has not been assessed. We found that neuronal expression of the V234I mutant allele in Drosophila (DVAP-V260I) induces defects in synaptic structure and microtubule architecture that are opposite to those associated with DVAP mutants and transgenic expression of other ALS-linked alleles. Expression of DVAP-V260I also induces aggregate formation, reduced viability, wing postural defects, abnormal locomotion behavior, nuclear abnormalities, neurodegeneration and upregulation of the heat-shock-mediated stress response. Similar, albeit milder, phenotypes are associated with the overexpression of the wild-type protein. These data show that overexpressing the wild-type DVAP is sufficient to induce the disease and that DVAP-V260I is a pathogenic allele with increased wild-type activity. We propose that a combination of gain- and loss-of-function mechanisms is responsible for VAPB-induced ALS.

Regulation of Dscam exon 17 alternative splicing by steric hindrance in combination with RNA secondary structures

The gene Down syndrome cell adhesion molecule (Dscam) potentially encodes 38 016 distinct isoforms in Drosophila melanogaster via mutually exclusive splicing. Here we reveal a combinatorial mechanism of regulation of Dscam exon 17 mutually exclusive splicing through steric hindrance in combination with RNA secondary structure. This mutually exclusive behavior is enforced by steric hindrance, due to the close proximity of the exon 17.2 branch point to exon 17.1 in Diptera, and the interval size constraint in non-Dipteran species. Moreover, intron-exon RNA structures are evolutionarily conserved in 36 non-Drosophila species of six distantly related orders (Diptera, Lepidoptera, Coleoptera, Hymenoptera, Hemiptera, and Phthiraptera), which regulates the selection of exon 17 variants via masking the splice site. By contrast, a previously uncharacterized RNA structure specifically activated exon 17.1 by bringing splice sites closer together in Drosophila, while the other moderately suppressed exon 17.1 selection by hindering the accessibility of polypyrimidine sequences. Taken together, these data suggest a phylogeny of increased complexity in regulating alternative splicing of Dscam exon 17 spanning more than 300 million years of insect evolution. These results also provide models of the regulation of alternative splicing through steric hindrance in combination with dynamic structural codes.

The ALS8 protein VAPB interacts with the ER-Golgi recycling protein YIF1A and regulates membrane delivery into dendrites

The vesicle-associated membrane protein (VAMP) associated protein B (VAPB) is an integral membrane protein localized to the endoplasmic reticulum (ER). The P56S mutation in VAPB has been linked to motor neuron degeneration in amyotrophic lateral sclerosis type 8 (ALS8) and forms ER-like inclusions in various model systems. However, the role of wild-type and mutant VAPB in neurons is poorly understood. Here, we identified Yip1-interacting factor homologue A (YIF1A) as a new VAPB binding partner and important component in the early secretory pathway. YIF1A interacts with VAPB via its transmembrane regions, recycles between the ER and Golgi and is mainly localized to the ER-Golgi intermediate compartments (ERGICs) in rat hippocampal neurons. VAPB strongly affects the distribution of YIF1A and is required for intracellular membrane trafficking into dendrites and normal dendritic morphology. When VAPB-P56S is present, YIF1A is recruited to the VAPB-P56S clusters and loses its ERGIC localization. These data suggest that both VAPB and YIF1A are important for ER-to-Golgi transport and that missorting of YIF1A may contribute to VAPB-associated motor neuron disease.