Integrated molecular landscape of amyotrophic lateral sclerosis provides insights into disease etiology

Abstract

Amyotrophic lateral sclerosis (ALS) is a severe, progressive and ultimately fatal motor neuron disease caused by a combination of genetic and environmental factors, but its underlying mechanisms are largely unknown. To gain insight into the etiology of ALS, we here conducted genetic network and literature analyses of the top‐ranked findings from six genome‐wide association studies of sporadic ALS (involving 3589 cases and 8577 controls) as well as genes implicated in ALS etiology through other evidence, including familial ALS candidate gene association studies. We integrated these findings into a molecular landscape of ALS that allowed the identification of three main processes that interact with each other and are crucial to maintain axonal functionality, especially of the long axons of motor neurons, i.e. (1) Rho‐GTPase signaling; (2) signaling involving the three regulatory molecules estradiol, folate, and methionine; and (3) ribonucleoprotein granule functioning and axonal transport. Interestingly, estradiol signaling is functionally involved in all three cascades and as such an important mediator of the molecular ALS landscape. Furthermore, epidemiological findings together with an analysis of possible gender effects in our own cohort of sporadic ALS patients indicated that estradiol may be a protective factor, especially for bulbar‐onset ALS. Taken together, our molecular landscape of ALS suggests that abnormalities within three interconnected molecular processes involved in the functioning and maintenance of motor neuron axons are important in the etiology of ALS. Moreover, estradiol appears to be an important modulator of the ALS landscape, providing important clues for the development of novel disease‐modifying treatments.

Introduction

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by the gradual degeneration of upper motor neurons in the cerebral motor cortex, and lower motor neurons in the brainstem and spinal cord. The subsequent muscle weakness and atrophy leads to death from respiratory failure, usually within two to five years after the first symptoms arise 17, 30, 62, 89. ALS has a worldwide incidence of 1‐3 per 100,000 person‐years, a mean age of onset between 50 and 65 years 30, 48, 58, 59, and a male to female ratio of approximately 2:1 36, 97. Based on the neuroanatomical substrate from which the initial symptoms originate, ALS can be categorized as of either bulbar‐ or spinal‐onset, which are characterized by initial speech/swallowing problems or limb‐related symptoms, respectively 30.

Previous ALS research has mainly focused on glutamate toxicity, or deficits in protein degradation, oxidative stress, mitochondrial function, axonal transport of organelles and RNA processing 104, 110 as underlying disease‐causing mechanisms. The familial ALS genes that cause the disorder when mutated often served as a starting point for these studies. Thus far, at least twelve familial genes causing ALS have been unequivocally identified (C9ORF72, CCNF, CHCHD10, FUS, OPTN, PFN1, SOD1, SQSTM1, TARDBP, TBK1, UBQLN2, VCP), whereas mutations in a number of other genes (e.g., ALS2, ANG, ATXN2, CHMP2B, HNRNPA1, HNRNPA2B1, NEFH, VAPB) have also been associated with ALS 77, 101. Approximately 10% of the ALS cases are classified as “familial,” i.e. following a Mendelian inheritance pattern. However, this classification is no longer so clear because mutations in “familial” genes also explain up to 11% of the sporadic, non‐inherited cases of ALS 77, 110. Nevertheless, ALS is still considered to be mainly a sporadic disease that, together with environmental and lifestyle risk factors 37, 86, is associated with a large number of common genetic variants (typically single nucleotide polymorphisms or SNPs), each with a slightly increased disease risk 8, 64. In recent years, genome‐wide association studies (GWASs) of ALS have identified many of these SNPs for sporadic ALS 9, 12, 19, 55, 56, 83, 84, 93, 94, 95.

In this study, we have integrated the most significant findings from six published GWASs of sporadic ALS, through genetic network and elaborate literature analyses, into a molecular landscape that also includes proteins encoded by familial genes and therefore covers both familial and sporadic ALS‐linked signaling cascades. The constructed landscape reveals the involvement of deficits in the functioning and maintenance of motor neuron axons as well as estradiol signaling in ALS etiology, and provides important clues for new ALS treatments.

Materials and Methods

ALS GWAS gene selection

ALS candidate genes were selected based on GWAS SNPs and their corresponding P‐values. All GWASs of sporadic ALS published to date were considered. Criteria for study inclusion were a publicly available independent GWAS discovery sample, with (at least) all SNPs associated at P < 0.0001. From the GWASs for which these data were available, SNPs were selected that were associated with ALS at P < 0.0001 to compile a list of associated genes. The selected genes either contained a SNP that was located within an exonic, intronic or untranslated region of the gene, or were found within 100 kilobases (kb) downstream or upstream of the SNP. The latter was based on the fact that the vast majority of expression quantitative trait loci (eQTL) for a given gene are located within 100 kb downstream and/or upstream of a gene 26, 74, 99 and because trait‐associated SNPs are more likely to be eQTL 72. The chosen statistical cut‐off for association (P < 0.0001) has been employed to designate “suggestive” evidence of association before 57, 61, 103. Subsequently, the literature was searched for additional (genetic) evidence linking the proteins encoded by the selected GWAS candidate genes to ALS.

Genetic network enrichment analysis

To identify enriched protein networks in the ALS GWAS candidate genes, a network analysis using the Ingenuity Pathway Analysis (IPA) software package (http://www.ingenuity.com) was performed, using default parameters. For each network, the Ingenuity software generates an enrichment score, i.e. the negative logarithm of the right‐tailed Fisher's exact test result.

Molecular landscape building

Guided by the results of the network enrichment analysis, the literature was extensively searched for the (putative) functions of all proteins encoded by the ALS GWAS candidate genes using the Uniprot Protein Knowledgebase (UniProtKB) (http://www.uniprot.org/uniprot) 91 and PubMed (http://www.ncbi.nlm.nih.gov/sites/entrez). Furthermore, the literature was searched for interactions between the ALS GWAS candidate gene‐encoded proteins and additional ALS candidate genes implicated in the disease through other (genetic) evidence, as well as genes/proteins and metabolites that have no known link with ALS, but have extensive functional connections with other proteins in the landscape.

Analysis of data from a cohort of sporadic ALS patients

We used epidemiological data from a cohort of sporadic ALS patients that were diagnosed and followed at the Department of Neurology, University Hospital Gasthuisberg (KU Leuven, Belgium) until death to analyze possible gender effects, linked to the relative abundance of estradiol in pre‐menopausal women. In this respect, the male:female ratios before and after the start of the menopause (corresponding to 51 years of age on average in Western European women 16, 18, 65, 73) were compared for the disease‐related variables “age at first symptoms” and “age at death” for the whole cohort of ALS patients and separately for the cases with spinal and bulbar onset. For these comparisons, a χ² test was used and P‐values <0.05 were considered statistically significant. Furthermore, possible gender effects on the mean disease duration – the mean time in years from age at first symptoms to age at death – before and after the menopause was assessed for all ALS patients and separately for the cases with spinal and bulbar onset. A Student's t‐test was used and P‐values <0.05 were considered statistically significant.

Results

Selected ALS GWAS genes and genetic network enrichment analysis

Six of the eleven published ALS GWASs met our inclusion criteria (Supporting Information Table 1) and were used to compile a list of 197 unique ALS candidate genes (Supporting Information Table 2). The most significantly enriched genetic network (P < 1.00E‐43; Supporting Information Figure 1) served as a starting point for building the molecular landscape.

The molecular landscape of ALS

Guided by the most significantly enriched genetic network and extensive literature searches, we built a molecular landscape that contains interacting proteins encoded by 121 of the 197 GWAS genes (61%; Supporting Information Table 2), 92 proteins (and protein complexes) implicated in ALS etiology through (familial) candidate gene, mRNA/protein expression and/or functional studies (Supporting Information Table 3), and 12 proteins that have not been directly linked to ALS (yet) but have extensive functional interactions within the landscape (Supporting Information Table 3).

Supporting Information Figures 2 and 3 show all relevant protein interactions that constitute the landscape. In the Supporting Information Data, a detailed description of the evidence linking all the proteins in the ALS landscape is provided, together with the respective references. The description here will be restricted to the overview of the ALS landscape as shown in Figure 1, depicting the main biological processes and signaling cascades in the landscape. Three interconnected main signaling cascades are present in the molecular ALS landscape: 1 regulation of Rho‐GTPase signaling, 2 signaling involving three regulatory molecules (estradiol, folate, and methionine) and their metabolites, and 3 ribonucleoprotein (RNP) granule functioning and axonal transport. These signaling cascades will be briefly discussed below.

Figure 1.

Overview of the molecular landscape of ALS. The different pathways and signaling cascades of the molecular landscape are shown in a healthy motor neuron (left) and a defective motor neuron from an ALS patient (right). Red crosses indicate the dysregulation of pathways and cascades that may result in motor neuron destabilization and death. See text for details.

The first cascade involves signaling through the Rho‐GTPases (CDC42, RAC1, and RHOA), and controls cytoskeletal dynamics and neurite outgrowth, and is regulated by growth factors (e.g., EGF, NGF, and VEGF), axonal guidance factors (e.g., CXCL12, netrin) and familial ALS proteins (e.g., SOD1 and TARDBP). In addition, the Rho‐GTPase RAC1 is part of the NADPH oxidase complex that produces reactive oxygen species (ROS) and regulates neurite outgrowth. Therefore, dysregulated or deficient Rho‐GTPase signaling affects cytoskeletal dynamics and neurite outgrowth of motor neurons and may increase ROS‐mediated oxidative stress. Of note, two very recent genetic studies have identified novel ALS genes, C21ORF2 and NEK1 47, 96. The proteins encoded by these genes both fit within our ALS landscape, i.e. C21orf2 is a direct functional interactor of the NEK1 kinase and a regulator of the cytoskeleton. Moreover, NEK1 is also involved in regulating cytoskeletal dynamics, binds the kinesin‐II motor complex and is a downstream target of the estradiol receptor ESR1.

Second, the folate and methionine cycles, linked with each other through vitamin B12, are implicated in ALS through multiple metabolites (e.g., carnitine, homocysteine, and S‐adenosylmethionine) and affect estradiol metabolite levels and estradiol‐mediated transcription (e.g., through transcriptional regulation by the methyltransferase MLL). The levels of folate and estradiol metabolites are regulated by the multidrug resistance transporters ABCB1 and ABCG2. Tetrahydrofolate (THF), the active form of folate, upregulates ABCG2 expression in the cell membrane and downregulates this transporter in intracellular organelles, which makes its localized expression dependent on THF availability. Furthermore, folate metabolites are involved in the synthesis of co‐activators of the NMDA glutamate receptor, whereas polyglutamation – the binding of multiple glutamate groups – of folate metabolites affects their kinetics in the cell. Riluzole, the only FDA‐approved drug to treat ALS, is an antiglutamatergic compound that reduces glutamate‐induced excitotoxicity by inhibiting the NDMA receptor and increasing the expression of astrocytic glutamate transporters. However, riluzole itself is also transported by ABCB1 and ABCG2, and it increases the expression of ABCG2. Hence, riluzole not only affects and regulates glutamatergic signaling, but also the intracellular levels of estradiol and folate (metabolites). Estradiol metabolites have different affinities for ESR1 and their regulation (e.g., through efflux and conversion) affects ESR1‐dependent transcription and activation. Taken together, a complex interaction exists between folate, methionine, and estradiol metabolites, associated with glutamate‐induced excitotoxicity and effects of riluzole.

Third, most of the proteins encoded by genetically linked ALS genes appear to be involved in RNA processing and transport. These proteins are located in RNPs, which are complexes of mRNAs and RNA‐binding proteins. RNPs regulate the processing, transport and immediate local translation of their constituent mRNAs, enabling the neuron to quickly react to environmental cues and/or damage in the axon and/or distant synapse. Motor neurons have exceptionally long axons and therefore rely heavily on the RNP system to locally regulate protein expression far away from the cell body. Cellular stress causes RNPs to stop the translation of certain mRNAs and keep them dormant until future demand. The formation and aggregation of these stress RNPs, or stress granules, are increased in ALS motor neurons. Moreover, mutations in e.g. the familial ALS genes FUS and TARDBP result in increased stress granule formation and affect (local) mRNA translation. The RNP system does also rely on the above‐mentioned Rho‐GTPase signaling, which is required for cytoskeletal maintenance and neuronal (out)growth. RNPs are transported via this “cytoskeletal framework” to their site of action in/along the axon. Therefore, deficient Rho‐GTPase signaling also negatively affects the axonal transport and function of RNPs.

Of note, estradiol and ESR1 control axonal (out)growth by regulating Rho‐GTPases and their expression, physically interact with RNPs and familial ALS proteins, and affect the regulation of glutamate‐induced excitotoxicity and the NADPH oxidase complex. Moreover, estradiol and ESR1 interact with and/or regulate (the expression of) multiple other proteins in the landscape. Therefore, estradiol‐related signaling appears to be an important modulator within the ALS landscape, through regulating axonal function and maintenance and hence motor neuron function and survival.

Analysis of data from a cohort of sporadic ALS patients

Our molecular landscape of ALS pointed towards an important regulatory role of estradiol in the etiology of the disease. Therefore, we used epidemiological data of a cohort of sporadic ALS patients to analyze possible gender effects – linked to the relative abundance of estradiol in pre‐menopausal women – on the age at onset, disease duration and age at death of bulbar‐ and spinal‐onset ALS patients. An overview of the cohort of sporadic ALS patients is shown in Supporting Information Table 4. The male:female ratio of all ALS patients is significantly lower after the start of the menopause – indicating a relatively increased number of affected females – for the disease‐related variable “age at first symptoms” (P = 0.016; Figure 2). Furthermore, the mean duration of disease is longer in younger ALS patients (all <51 vs. all ≥51 years old; P = 5.3E‐06), and in spinal‐onset vs. bulbar‐onset ALS patients (P = 6.5E‐11), but is independent of gender for both. However, when analyzing bulbar‐ and spinal‐onset separately, postmenopausal women with bulbar‐onset ALS have a longer mean duration of disease compared with male bulbar‐onset ALS patients (P = 0.029). Furthermore, the parameter “age at death” showed that the male:female ratio of all patients and especially that of bulbar‐onset ALS patients is reduced (from a ratio of 3:1 to 1:1) after the start of the menopause (P = 0.011 and P = 0.022, respectively).

Figure 2.

Analysis of data from a cohort of sporadic ALS patients for possible gender effects on the prevalence, onset and disease duration of (bulbar‐ and spinal‐onset) ALS. The male:female ratio for the disease‐related variable “age at first symptoms” (left), the “mean duration of disease” (middle), and “age at death” (right) was compared for all ALS cases, and separately for the cases with spinal or bulbar onset, before and after the start of the menopause (corresponding to 51 years of age in Western European women). A χ² test was used for the analyses of “age at first symptoms” and “age at death,” and the Student's t‐test was used for the analysis of “mean duration of disease.” P‐values are indicated above the bars and considered significant when P < 0.05.

Discussion

In this study, we integrated available ALS data into a molecular landscape that reveals the main biological processes that are affected in ALS, i.e. Rho‐GTPase signaling, signaling involving estradiol, folate, and methionine, and RNP granule functioning and axonal transport, that may contribute to motor neuron dysfunction and, ultimately, death. The molecular ALS landscape represents processes and cascades that may be affected in both the monogenic, familial and the more prevalent polygenic, sporadic forms of ALS. In this respect, the landscape includes processes and signaling cascades reported to be involved in familial ALS such as oxidative stress and RNA processing, as well as processes of “classical” ALS theories such as glutamate toxicity. However, the landscape also comprises processes that have been less well studied before – e.g. growth‐ and guidance factor signaling and cytoskeletal dynamics – and also sheds further light on the functional relationships between the various ALS‐linked processes. Interestingly, estradiol signaling is functionally involved in all main processes and as such an important modulator of the ALS landscape.

It is important to note that the molecular landscape is not intended to imply a fixed “sequence of events” that eventually leads to motor neuron death in all patients, i.e. a number of biological processes that occur in a temporally and/or spatially distinct order. Instead, we propose that deficits in any of the main landscape processes, either by themselves or in combination with others, can cause motor neurons to die. Of note, the familial proteins are often involved in more than one of these main processes and/or of crucial importance to these processes. Consequently, mutations in the familial ALS genes have a functionally “high impact” in the landscape and directly lead to ALS, whereas in sporadic ALS patients multiple functionally “lower‐impact” genetic variations are required to develop the disease. For example, mutations in the familial gene SOD1 result in oxidative stress 2, affect Rho‐GTPase signaling 31, and impair axonal transport of the enzyme choline acetyltransferase 90 that synthesizes acetylcholine, the main and essential neurotransmitter at the neuromuscular synapse 79. Thus, both familial mutations and sporadic variations may result in disorganization of the cytoskeleton, defects in axon maintenance and motor neuron death.

Within our ALS landscape, Rho‐GTPases are important regulators of neuronal development and survival as well as cytoskeleton dynamics 85, and as such they are crucial for axonal maintenance, regeneration, and transport. This view is supported by reported defects in the regulation of motor neuron axonal regeneration of motor neurons in ALS patients 5, 50 and ALS mouse models 29, 88. In addition, regulation of neurofilaments – the major building blocks of the cytoskeleton and crucial for axonal regeneration 100 – is affected in ALS patients 63 and in ALS mouse models 81, 106, and multiple studies have shown defects in axonal transport 1, 6, 7, 11 and dysregulation of motor proteins 39, 40, 54 in motor neurons of ALS patients.

Of note, axonal regeneration is dependent on DNA methylation, which is regulated by the landscape's methionine and folate cycles and enhanced by folic acid (folate) supplementation 42, 43, 52. Because the DNA methylation cascade regulates – through MLL activation – the transcriptional activity of the estradiol receptor ESR1 102, these findings together imply that methionine, folate, and estradiol are important modulators of ALS pathogenesis.

The vitamin B12‐linked folate and methionine cycles are also implicated in ALS through multiple previous studies. For example, homocysteine is increased in the plasma and CSF of ALS patients 92, 107, 109, whereas carnitine 80 and methionine 41 are decreased in the plasma of ALS patients. Furthermore, folic acid supplementation is neuroprotective 105, S‐adenosylmethionine delays disease onset 87 and carnitine decreases disease progression and increases survival 49 in a mouse model for ALS. Carnitine supplementation may also be beneficial in ALS patients 4 and methyl‐vitamin B12 may delay the motor symptoms of ALS 38, 45, 107. Nevertheless, the benefits of supplementing these metabolites may not be universal, as ALS patients may have different nutritional deficiencies. For example, hyperhomocysteinemia has been noted in rats on a low methionine diet 21, as well as in human vegetarians and vegans along with decreased vitamin B12 and increased folate levels compared with controls on an omnivorous diet 51. Another environmental factor that affects these pathways is endurance exercise, associated with increased homocysteine plasma levels 34, 76 together with low vitamin B12 and folate levels 23, 34, 35. Vigorous physical activity may therefore – at least partially – affect pathways that are part of ALS etiology, and it is tempting to speculate that this may be the link between a highly active lifestyle and/or high level of physical fitness (e.g., athletes, blue‐collar workers) and increased ALS incidence 3, 20, 24, 32, 37, 66. Therefore, it is worthwhile to further investigate whether individuals on a specific diet, with or without an active lifestyle, have an increased risk to develop ALS due to dysregulation of their folate and methionine cycles.

Another important factor in the ALS landscape is estradiol signaling that functionally integrates, regulates and is regulated by key landscape processes. Estradiol mediates cytoskeleton dynamics through interacting with Rho‐GTPases and the NADPH oxidase complex, and as such may protect motor neurons against degeneration. This apparent critical role of estradiol and related signaling in ALS etiology is corroborated by several lines of evidence. First, epidemiological studies have demonstrated that men have an approximately three times higher risk to develop ALS before the age of 50 years than women 33, a gender difference that gradually decreases with increasing age 65. Our epidemiological analysis of a large cohort of sporadic ALS patients confirms these findings and suggests that premenopausal women are – to a certain extent – protected against ALS. Furthermore, for bulbar‐onset – but not spinal‐onset – ALS, the mean duration of disease is longer in postmenopausal women than in men, and also the male:female ratio for the age at death is drastically decreased after the menopause. This indicates that estradiol may exert a protective effect especially for bulbar‐onset ALS and that the effect on the disease duration may be due to residual estradiol slowing disease progression. To our knowledge, this is the first study that shows an association between estradiol and the primary neuroanatomical substrate (spinal vs. bulbar) from which the initial ALS symptoms originate.

The involvement of estradiol signaling in ALS pathogenesis is also in line with functional studies in ALS animal models showing that estradiol delays disease onset and progression, and increases survival 10, 28, 98. Moreover, estradiol protects cultured spinal motor neurons against excitotoxicity and rescues these neurons from degenerating and dying 14, 44, 71, 75. Taken together, these findings imply that estradiol and estradiol‐related signaling have an important modulatory role in motor neuron function and survival.

In this respect, it is of note that gender differences can also be observed at the genetic level, i.e. SNPs in the gene encoding the ESR1 co‐activator PPARGC1A are associated with age of ALS onset and survival in males specifically 22. Moreover, genetic variations in MTHFR 53, 82 – which encodes an important enzyme in the folate cycle – and in the promoter region of the growth factor VEGFA 25 have been associated with ALS in women. These notions imply that the presumed hormonal neuroprotective advantage of women could be counteracted by genetic variations in, for example, the folate cycle and/or VEGF signaling.

Currently, the antiglutamatergic drug riluzole is the only FDA‐approved drug to treat ALS and extends the life expectancy of ALS patients by approximately 2‐3 months 69. Multiple other antiglutamatergic compounds have been tested in clinical trials – either by themselves or in combination with riluzole – but were unsuccessful 13, 15, 27, 60, 70, 78, 108. These findings imply that the beneficial function of riluzole may not be limited to regulating glutamate toxicity only, as explained by our molecular landscape riluzole regulates and is regulated by transporters that also control folate and estradiol metabolite levels in the cell 46, 67, 68, which may therefore be an additional mechanism through which riluzole conveys neuroprotection in ALS. It is tempting to speculate that such a multi‐modal effect of riluzole explains its effectiveness in ALS, while other antiglutamatergic compounds fail.

In conclusion, our integrated molecular landscape of ALS highlights the involvement of processes that lead to deficient axonal functioning of motor neurons (i.e., axonal transport, local translation, regeneration, and outgrowth) and points toward estradiol‐related signaling as an important mediator of ALS pathological mechanisms. Consequently, the landscape not only yields in‐depth insights into the etiology of ALS but also provides new clues for the development of disease‐modifying ALS treatments.

Supporting information

Additional Supporting Information may be found in the online version of this article at the publisher's web‐site:

Data file S1:

List of abbreviations

Figure S1. Top enriched Ingenuity genetic network.

Figure S2. ALS landscape (1/2).

Figure S3. ALS landscape (2/2).

Table S1. Published ALS GWASs.

Table S2. List of ALS GWAS candidate genes with corroborating evidence.

Table S3. Other genes implicated in ALS with corroborating evidence.

Table S4. Overview of ALS patient cohort.

Detailed description of the ALS landscape