BMP/TGF-β signaling as a modulator of neurodegeneration in ALS

Abstract

This commentary focuses on the emerging intersection between BMP/TGF-β signaling roles in nervous system function and the amyotrophic lateral sclerosis (ALS) disease state. Future research is critical to elucidate the molecular underpinnings of this intersection of the cellular processes disrupted in ALS and those influenced by BMP/TGF-β signaling, including synapse structure, neurotransmission, plasticity, and neuroinflammation. Such knowledge promises to inform us of ideal entry points for the targeted modulation of dysfunctional cellular processes in an effort to abrogate ALS pathologies. It is likely that different interventions are required, either at discrete points in disease progression, or across multiple dysfunctional processes which together lead to motor neuron degeneration and death. We discuss the challenging, but intriguing idea that modulation of the pleiotropic nature of BMP/TGF-β signaling could be advantageous, as a way to simultaneously treat defects in more than one cell process across different forms of ALS.

Bone morphogenetic protein (BMP) and transforming growth factor-β (TGF-β)/Activin signaling pathways are known to impact synaptogenesis, axonal and dendritic growth, synaptic transmission, and neuronal survival. Given their involvement in nervous system development and function, there is considerable potential that disruptions in signaling activity and efficacy result in neurological disease. Moreover, it is conceivable that targeted manipulations of these pathways may be effective in abrogating cellular dysfunctions typical of nervous system disorders. The BMP and TGF-β/Activin intercellular signaling pathways represent two evolutionarily related branches, composed of a family of ligands, transmembrane receptors, and signal transducers, that display significant sequence and functional conservation, all together often referred to as TGF-β superfamily, or BMP/TGF-β, signaling.^1–4 BMP/TGF-β signaling plays critical roles in a multitude of developmental and homeostatic processes in invertebrates and vertebrates alike.^1,5 BMP ligands were first identified in demineralized bone with remarkable capabilities as inducers of bone deposition,⁶ and TGF-β as a growth factor that could stimulate anchorage-independent growth of fibroblasts in soft agar (reviewed in [Moses, 2016 #2016]). In addition to their role in the development and homeostasis of bone,⁷ BMPs are well known to provide instructional cues in many aspects of development, as well as in tissue homeostasis later in life. Similarly, TGF-β signaling regulates cell proliferation, differentiation, and apoptosis during development and also acts as a critical regulator of immune function.^8–11 While these two related branches share some pathway elements, they often antagonize one another, inducing quite different outcomes in the same cellular or tissue environment.^12,13

The roles of BMP/TGF-β signaling in nervous system function represent a substantial gap in our knowledge of how these remarkably pleiotropic pathways tightly regulate cell communication between both neuronal and non-neuronal components of dynamic neural circuits. Early roles for both pathway branches in neural patterning of both invertebrates and vertebrate embryos are well known,^14–18 but an appreciation of the nuances of BMP/TGF-β signaling in later events, such as axon guidance, synaptic growth, neural plasticity, and axonal regeneration has now gained traction with more research in these areas of study.^19–21 Likewise, more reports are appearing in the literature that document disruptions in BMP/TGF-β signaling in a variety of neurological diseases, implicating both branches in, Alzheimer’s disease, Parkinson’s disease, Angelman Syndrome, multiple sclerosis (MS), and amyotrophic lateral sclerosis (ALS).^22–25 Here, we will focus on our current understanding of the intersection of BMP/TGF-β signaling with ALS, a lethal neurodegenerative disease, for which diagnosis is challenging for the clinician, and devastating for the patient, as no effective treatments or therapeutics are yet available. We will not attempt a comprehensive review of ALS or BMP/TGF-β signaling, but will instead summarize what is currently known about how BMP/TGF-β signaling is altered in ALS, and how it may be possible for modulation of these pathways to alter the outcome of ALS. We raise questions that remain to be addressed, in hopes that they will help drive future research towards a more thorough understanding of the underlying mechanisms of the progressive degeneration associated with ALS. Importantly, we suggest avenues of research that may aid in the identification of both biomarkers and therapeutics, so desperately in need for overcoming the loss of neuron function, motor activity and ultimately life, all hallmarks of ALS.^26,27

1 |. THE ALS PHENOTYPE

ALS is characterized by a progressive loss of motor control accompanied by the death of motor neurons whose cell bodies reside in the brain (upper motor neurons) and/or the spinal cord (lower motor neurons).^28,29 Neuronal death leads to muscle wasting and atrophy, manifesting as weight loss, muscle weakness, fasciculations, and changes in gait or speech. As the disease progresses, everyday motor tasks become more difficult, and the degeneration of diaphragm innervation ultimately leads to respiratory failure and death.^30–32 In addition to motor loss, approximately 15% of ALS patients experience frontotemporal dementia (FTD), implicating ALS as a multisystem neurodegenerative disease, known as ALS/FTD.³³ (Here, we refer to the ALS/FTD spectrum, as simply ALS.) The onset of ALS typically occurs in mid- to late in life, at approximately 50–70 years of age, at an incidence of ~2 cases per 100 000 individuals. Survival time post diagnosis typically ranges from 2 to 3 years.^30,34 Currently, there are two Federal Drug Administration (FDA) approved treatments for ALS: Edaravone and Riluzole. These treatments slow the progression of the disease by no more than 6 months.^35,36 With no known cure, or therapeutic that provides a considerable extension in lifespan, the diagnosis of ALS is devastating to the patient, their family members and caregivers, as a rapid irreversible decline is inevitable. There is a pressing need to identify treatments that reduce, stop degeneration, or even restore motor function.

ALS is classified as familial (fALS) or sporadic (sALS), depending on whether a patient does, or does not, have a family history of the disease, respectively. Clinically, both types of ALS patients present with nearly identical pathologies, and mutations in fALS genes are often identified in sALS patients.^26,27,37 Familial ALS (10% of total ALS cases) has been associated with primarily autosomal dominant mutations in at least 24 different genes which encode a variety of proteins whose functions range from roles in RNA biology to fundamental cellular metabolism, some of which have previously been associated with FTD, reinforcing an ALS-FTD spectrum.^31,38–40 The remaining 90% of ALS cases, classified as sporadic, may arise from de novo mutations, previously unknown genetic mutations, and/or, non-genetic associations. Environmental factors that are disproportionately correlated with sALS cases include, but are not limited to, military service, low body fat, and a variety of external and chemical insults.^27,41 ALS diagnosis is often reached when no other definitive explanation for the progressive loss of motor control is found.³² As previously noted, motor neuron death is a common feature in all ALS cases. Thus, a question plaguing the field for some time is: how do varied insults, either environmental or genetic, result in the same phenotype: the demise of motor neurons and loss of motor function?

Prior to motor neuron death, neuromuscular junction (NMJ) structure and function is compromised, axonal projections retract and degenerate, inclusions appear in the cytoplasm of aggregated proteins and RNAs, and an increased number of neuroinflammatory cells infiltrate the brain and spinal cord.^42–46 Both patients, and animal models of ALS, exhibit NMJ abnormalities at the morphological and functional level. ALS patients exhibit a dying back of motor axons in the spinal cord coupled with NMJ degeneration.⁴⁷ A mouse model of ALS has shown impaired neurotransmission accompanied by disrupted end plates and innervation at the NMJ prior to evidence of motor deficits or motor neuron loss.⁴⁸ From an electrophysiological perspective, a zebrafish model provided evidence of decreased quantal neurotransmission.⁴⁹ Further, a Drosophila knock-in model of ALS showed a reduction in organismal locomotion accompanied by electrophysiological changes in the sensory feedback component of the motor circuit prior to significant NMJ abnormalities.⁵⁰

2 |. ALS MODELS

Directed studies of the fALS genes most affected worldwide have enabled the research community to gain considerable insight into the molecular/cellular basis of cellular pathologies associated with ALS. These studies have made use of patient tissues, induced pluripotent stem cells (iPSCs) derived from patients, as well as animal models of ALS.^51–54 Genetic models in organisms, such as rodents (mice, Mus musculus and rats, Rattus norvegicus), nematode (Caenorhabditis elegans), fruit fly (Drosophila melanogaster), and zebrafish (Danio rerio), have been able to reproduce ALS neurodegenerative phenotypes, and in some cases, impaired locomotor function.^52,55–60 The majority of models to date induce ALS-like phenotypes by overexpressing a mutant form, or even wild type form, of a human, or an orthologous fALS gene, in select cells, that is, motor neurons or astrocytes.^61,62 More recently, as technological advances have been made in genome editing, ALS mutations can be “knocked-in” to endogenous orthologous loci, enabling an animal model to replicate gene dosage and proper spatial and temporal gene expression patterns, while also recapitulating progressive motor deficits and the neurodegenerative pathogenesis of ALS.^63–66 Advances, that eliminate the use of heterologous systems, have allowed patient derived iPSCs to be differentiated into a cell type of choice, producing a “purified” culture of human cells that harbor the relevant genetic background, that is, an “ALS genome”.^67–69 The primary disadvantage, of course, is that the uniform population of differentiated iPSCs have lost their in vivo context. To rectify this shortcoming, 3D co-cultures and microfluidic systems aim to replicate the in vivo environment, addressing the importance of spatial context and cell-cell communication in ALS.^70–73

An important element in the function of neural circuits is the synapse which constitutes the juncture between the presynaptic neuron and its postsynaptic target. The structure of the synapse facilitates neurotransmission at the NMJ, for example, to enable the efficient release of neurotransmitter from vesicle fusions at the presynaptic membrane and their reception by neurotransmitter receptors localized on the opposing muscle cell membrane, leading to subsequent activation of contractile machinery. Motor neuron degeneration in ALS is accompanied by a disruption in NMJ structure and function. While differences exist in the organization and specific neurotransmitters used at NMJs of different species, that is, Drosophila vs mice vs human,^74,75 many molecular details are similar, with findings of the Drosophila glutamatergic NMJ advancing our knowledge of glutamatergic synapse in the vertebrate CNS.⁷⁶ Nevertheless, when considering the behavioral consequences of ALS-related disruptions at a mammalian NMJ vs that at the Drosophila NMJ, we need to keep in mind that more than one motor neuron can innervate a single muscle in Drosophila, while vertebrates exhibit a one neuron to one muscle fiber relationship. Further, different types of synaptic boutons are found in Drosophila, with no evidence of bouton diversity in vertebrates.^77,78 Despite these differences Drosophila ALS models have successfully identified cellular pathways shown to be compromised in ALS patients or mammalian models, as well as revealing genetic suppressors that have translated to human models.^55,79–81

Regardless of classification (fALS or sALS) or heterogeneity in disease presentation, research using genetic animal models and patient derived cells has advanced the field in its conceptual understanding of the molecular underpinnings of ALS-associated degeneration. The extreme view that each mutated gene or environmental factor uniquely triggers a different process resulting in motor neuron death is unlikely. In reality, it is more likely that individual triggers (gene mutations and environmental insults) effect one of only a handful of biological processes, such as protein homeostasis and quality control, RNA biology, or cytoskeletal dynamics,⁸² that initially function somewhat independently of one another (Figure 1). As each potentially minor defect triggers their own disruption in a molecular pathway, multiple failings begin to impinge upon one another, together causing a more global imbalance and dysfunction of cellular physiology, resulting in the “perfect storm” of cellular pathologies characteristic of ALS (Figure 1). A better understanding of the molecular relationships between affected cellular pathways will reveal if common targets exist and whether a single therapeutic for all ALS patients is possible. If such interventions are not effective or cannot be made early enough, a combination of treatments or therapeutics, tailored to each of the initial dysfunctional processes will be necessary, ideally prior to significant decline, requiring a “personalized” approach to treatment.

FIGURE 1.

Conceptual depiction of cellular pathways impacted in ALS leading to motor neuron degeneration and death. (Top) Healthy motor neurons (green) in an environment of microglia (yellow) and astrocytes (blue). Gene mutations associated with ALS and/or environmental insults usually lead to disruptions in three general categories of cellular processes, that is, protein homeostasis, RNA biology, and cytoskeletal dynamics. The paths initiating cellular decline appear to trigger additional shared dysfunctions (mitochondrial or cellular metabolic dysfunction, impaired neurotransmission and the formation of protein and/or RNA aggregates and inclusions) that converge and accumulate, eventually leading to degeneration of the neuron, neuroinflammation (indicated by swollen microglia, yellow), astrocytosis (purple astrocytes), and finally, cell death

3 |. DO COMMON CELLULAR PATHWAYS UNDERLIE ALL FORMS OF ALS?

Four genes with the highest worldwide prevalence for fALS mutations are C9orf72, superoxide dismutase 1 (SOD1), Transactive Response DNA-binding protein 43 (TARDBP encoding TDP-43), and fused in sarcoma (FUS).³⁹ Expansions of a GGGGCC (G₄C₂) hexanucleotide repeat sequence, which normally exist as ~10 copies within the intron of the chromosome 9 open reading frame 72 (C9orf72) gene, expand to 400 to 2000 repeats in most (39.6%) fALS cases, with concomitant haploinsufficiency of C9orf72.^83,84 Of the people who carry the G₄C₂ hexanucleotide repeat expansion, 25.1% also exhibit cognitive decline and memory loss typical of FTD.⁸⁵ Toxicity associated with the expansion of G₄C₂ repeats has also been attributed to multiple aspects of the hexanucleotide expansion. The increase in G₄C₂ repeats in RNA transcribed from the C9orf72 locus is thought to result in increased sequestration of protein binding partners from their normal cellular functions. Translation of the expansion of repeats via RAN (repeat associated non-AUG) translation results in long stretches of dipeptide repeats which exhibit a high propensity to for accumulations and aggregations within the cell.^86,87 The second most common gene associated with ALS is superoxide dismutase 1 (SOD1). Mutations in SOD1 not only affect the protein’s ability to prevent the accumulation of superoxide radicals that wreak havoc on a number of cellular processes (DNA and RNA metabolism, mitochondrial function and protein folding),^88,89 but also result in the aggregation of mutant SOD1 protein with wild type SOD1 protein, further diminishing the enzyme’s role in reactive oxygen species (ROS) abatement.⁹⁰ SOD1 inclusions have been observed in both neurons and glia,⁹¹ with a more rapid disease progression associated with misfolded SOD1 in microglia.⁸⁹ Mutations in TDP-43 and in FUS account for approximately 4.5% and 5% of all fALS cases, respectively. Both TDP-43 and FUS are RNA binding proteins and mutant forms of these proteins sequester transcriptional regulators, disrupting multiple aspects of RNA biology, including transcription, splicing, and RNA trafficking, all thought to contribute to ALS pathologies.³⁷ Like SOD1, cytoplasmic inclusions are often associated with the co-localization of mutant FUS and TDP43 with stress granules.^92–96 A prominent disease signature also shared with other neurodegenerative diseases is the increase in stress granule formation and protein inclusions, indicative of general disruptions in both protein and RNA folding and turn-over.³⁷ In roughly 97% of all ALS patients, TDP43 inclusions are found in cells of the spinal cord and motor cortex, irrespective of a fALS or sALS diagnosis. In general, TDP43 aggregates are thought to be toxic to neurons, but their potential protective role is important to consider.⁹⁷ Whether cytoplasmic inclusions initiate decline, or they arise as a secondary consequence of upstream pathway dysfunctions, the question remains, does preventing their formation have a significant impact on disease progression and patient outcome? Relevant to this area of research is the implication that such molecular instabilities and aggregations could be responsible for the prion-like propagation of ALS pathologies, a molecular mechanism that could account for efficient spread from one cell to another and the rapid rate of disease progression.^98–100

Efforts to uncover common pathways at the root of ALS decline are starting to emerge from comparative studies. In three different iPSC-derived motor neuron models harboring fALS mutations in SOD1, FUS, and C9ORF72, an increase in hyperexcitability suggests that toxicity associated with superfluous glutamatergic activity could be a common mechanism leading to abnormal increases in calcium that ultimately contribute to neuron death.⁵¹ In a similar vein, attempts to identify genetic variants that suppress neurodegenerative phenotypes across multiple ALS models is a powerful approach for revealing shared pathways that underlie the demise of neural function.^101–103 There are also efforts being made to identify the molecular signatures associated with dysfunctions in ALS using RNA sequencing, proteomics, and metabolomic approaches.^104–106 The field would benefit from comparative omic analyses across different mutations.¹⁰⁷ One important goal will be to compare presymptomatic and post-symptomatic samples to determine how changes in gene expression, protein modifications, and metabolites correlate with ALS pathologies, such as, hyperexcitability, synaptic and axonal instability, or cytoplasmic inclusions. Ideally, such studies will examine molecular changes in specific cell types in an in vivo context through time. Such studies of the shifts in molecular signatures early in the manifestation of ALS hold greater promise for (1) identifying potential biomarkers as predictors of disease risk, (2) determining whether commonalities exist early in disease onset, and (3) uncovering targets for the discovery of therapeutics that can prevent disease progression.

It is likely that ALS treatment and/or prevention will require targeting multiple molecular pathways, either simultaneously or in succession. However, it is conceivable that targeting a single pathway could be effective, if that pathway is pleiotropic, affecting multiple cell types and cellular processes, in which case modifying its activity or efficacy could attenuate the multi-pronged reach of ALS. BMP/TGF-β signaling is one such pathway, notorious for its involvement in a dizzying array of cellular functions.^10,21,108 If BMP/TGF-β signaling is affected in multiple ALS pathologies, it may be possible to take advantage of its known regulatory mechanisms, to dial up, or dial back, signaling activity, modifying multiple dysfunctional processes in order to prevent neurodegeneration and restore motor function.

4 |. BMP/TGF-B SIGNALING OVERVIEW

TGF-β superfamily ligands are expressed broadly in the brain and spinal cord of the adult central nervous system, as well as in the key musculoskeletal components of the motor circuit.^25,109 Extracellular ligands that comprise the large family of dimeric BMP/TGF-β signaling molecules, bind the extracellular domain of a heterotetrameric receptor complex, composed of type I and type II transmembrane serine/threonine kinases, to transduce a signal (Figure 2). Constitutively active type II receptor kinases phosphorylate and activate the type I receptor kinase, enabling its phosphorylation of cytoplasmic proteins that act to alter the receiving cell’s physiology. The pathway is highly regulated at all levels, from processing, maturation and availability of homo- or heterodimeric ligands,^110–115 to different constellations of receptors in the signaling complex,¹ to a variety of intracellular effectors¹¹⁶ (Figure 2). In general, the BMP and TGF-β pathways are based on the evolutionary relatedness of ligands, type I and type II serine/threonine receptors, receptor mediated cytoplasmic Smads (R-Smad) proteins, in association with the co-Smad (Smad4) to regulate transcription in nuclei of responding cells. Phosphorylated R-Smads, the effector in Smad-dependent signaling, are generally distinct between the two branches with pSmad1/5/8 mediating signaling induced by BMP and growth differentiation factor (GDF) ligands, and pSmad2/3 mediating signaling initiated by TGF-β, Activin, and Nodal ligands,¹ with inhibitory I-Smads able to negatively regulate Smad-dependent signaling.^117,118 In some contexts, pSmads act independently of co-Smad4, such as in their role in chromatin remodeling and microRNA processing.^119–121 R-Smads are not the only effectors of BMP/TGF-β signaling. In Smad-independent signaling, a number of different kinases including p38 MAPK, PI3K, JNK, and LIMK, are activated in response to the ligand-bound receptor complex, extending the range of targets and cellular functions effected by BMP/TGF-β ligands.^2,122

FIGURE 2.

Bmp/TGF-β signaling consists of two evolutionarily related branches with sequences and structural similarities between ligands, type I and type II receptors, and Smad effectors. The pathways are regulated at multiple levels including ligand maturation and availability, receptor complex composition and effector activity. Signal transduction depends on both Smad-dependent and Smad-independent (non-Smad) mediation. p38 and LIMK are two examples of non-Smad signaling effectors. The response of specific signaling components in ALS patients and/or models are shown as, elevated (up arrow) or reduced (down arrow) levels. The consequence of a direct experimental manipulation of BMP or TGF-β signaling in the context of an ALS model is highlighted in chartreuse (see text). Various properties of signaling that are altered or may be affected in ALS are indicated, that is, the different response of signaling (pMad) in different cellular compartments of the motor neuron, differential inclusion of pSmad2 vs pSmad3 dependent on importin-β, and the potential for BMP-mediated LIMK regulation of Actin dynamics in ALS

5 |. ALS-ASSOCIATED DISRUPTIONS IN BMP/TGF-B SIGNALING

Despite the identified roles for BMP/TGF-β signaling in nervous system function and the suggestions that dysregulated signaling may underlie neurodegenerative disorders, relatively few studies have directly addressed BMP/TGF-β signaling in ALS.^{1,22,23,25,123} A common pattern across the pathway has yet to emerge with regard to whether signaling is generally upregulated, or down-regulated, in specific tissues or cells. A preferential involvement of Smad-dependent vs Smad-independent signaling in ALS is also not yet clear. However, to date, the findings that report a change in the TGF-β branch in the ALS disease state, all report an upregulation. A preferential elevation in TGF-β1 ligand level, compared to those of TGF-β2 and 3, is seen in the spinal cord, muscles, and astrocytes of individuals carrying a SOD1-ALS mutation, as well as, in sALS patients.^{109,124–126} An increase in pSmad2/3 has been detected in both neurons and glial cells of sALS and fALS patients, thus the TGF-β/Activin pathway is indeed activated in both of these cell types.¹²⁵ Furthermore, pSmad2/3 was found to localize to TDP-43 aggregates in lower motor neurons of sALS patients.¹²⁷ Significant increases in Smad2 RNA in muscle tissue of a transgenic SOD1^G93A ALS mouse model¹²⁶ and the orthologous dSmad2/Smox RNA in the central nervous system of a TDP-43 Drosophila knock-out model, while alone is not indicative of an increase in signaling per se, is suggestive of altered TGF-β/Activin signaling.¹²⁸ Consistent with the idea that an upregulation of TGF-β/Activin signaling could underlie ALS pathologies, administration of a TGF-β inhibitor (SB-431542) was found to extend the survival of SOD1^G93A ALS mice.¹²⁴

With regard to BMP signaling, a simple up- or down-regulation, does not consistently correlate with ALS. In a rat SOD1^H46R ALS model, an elevated level of BMP4 was detected in reactive astrocytes in the lumbar ventral spinal horns of late symptomatic animals,¹²⁹ consistent with reports that particular BMP ligands, including BMP4, increase following neuronal injury, in animal models of spinal cord injury (SCI), traumatic brain injury (TBI), or demyelinating disorders, accompanied by a neuroinflammatory response (reviewed in [Hart, 2020 #1712]). In addition to an activation of Smad-dependent signaling (ie, increased pSmad1/5), astrocytes in the rat SOD1^H46R ALS model showed an increase in p38MAPK, indicating an accompanying upregulation of Smad-independent signaling.¹²⁹ As a means to block BMP signaling, intrathecal injection of Noggin, an extracellular antagonist of the BMP ligands, was shown to reduce markers of neuroinflammation in the rat SOD1^H46R ALS model, and to help maintain myelin integrity in independent neuronal injury models.²⁵

Proper synapse growth, neurotransmission, and synaptic plasticity in the Drosophila larval nervous system requires Smad-dependent BMP signaling, mediated by the BMP5/6/7 ortholog, gbb, the type II BMP receptor 2 (BMPR2), wit, and the Smad1/5/8, ortholog, Mad.^123,130,131 The loss of gbb results in small NMJs, and overexpression of gbb and hyperactivation of BMP signaling produces larger NMJs, with an increase in synaptic transmission.^123,132,133 Reduced locomotor activity and severe NMJ degeneration exhibited by a dSod1^G85R knock-in ALS model, could both be abrogated by BMP signaling activation in response to the overexpression of gbb, or SaxA, a constitutively active form of the BMP type I receptor (ACVR1).⁵⁰ In this case, ALS-associated defects were attenuated by activation of BMP signaling, in contrast to the results from rodent models where down-regulation of BMP signaling was able to reduce neuroinflammation, and an inhibition of TGF-β signaling extended survival in a SOD1 model.^124,129 In both cases the Drosophila and rodent models are SOD1-ALS mutants, however the nature of the genetic manipulation of SOD1 differed. Does the consequence of BMP/TGF-β signaling manipulation depend on whether the ALS model is a knock-in vs overexpression model? Or does the ability of signaling to suppress various pathologies depend on the cell or tissue in which signaling is activated?

To address the latter question, cell-type specific and cell-autonomous activation of BMP signaling was tested for rescue of multiple phenotypes in the dSod1^G85R Drosophila knock-in model.⁵⁰ Specific activation of signaling in motor neurons, sensory neurons, or a subpopulation of sensory neurons, the proprioceptors, showed significant rescue, but activation in muscle or glial cells did not, indicating that BMP signaling rescue of motor dysfunction is cell-type specific. While underappreciated, the involvement of sensory neurons in ALS is known from the clinical standpoint, as sensory neuropathies are evident in patients.^134–136 Similarly, sensory neurons are found to be stress-sensitive in ALS animal models.¹³⁷ The failure of muscle or glial-specific signaling to suppress degeneration could not be simply explained by detrimental effects of high signaling in those cells, as none were detected. The potential for ectopic activation of signaling, in and of itself, to produce negative consequences is an important consideration, especially in the experimental manipulations of potent pathways such as BMP and TGF-β/Activin signaling. For example, overexpression of BMP ligands in the Drosophila eye, a popular tissue to assay for modification of human disease genes, produces disorganized and roughened eyes, a phenotype similar to that seen when a variety of human disease genes are expressed in this tissue, precluding any assessment of modification.^138,139 It is important to keep in mind the limitations of various assays and models, while also recognizing that in some cases, while one may not be able to directly assess molecular functions specific to a disease, the epistatic relationships between genes and/or pathways can be explored, and then tested further in other disease models.

Another important consideration with regard to BMP/TGF-β signaling is their potential for crosstalk in part due to the evolutionary relatedness of the signaling components comprising the two BMP and TGF-β/Activin branches. The ability of receptors and Smad effectors from one pathway to form non-functional complexes with those typically promoting signaling could underlie competition or antagonism between the two pathways, observed in a variety of developmental and disease contexts (reviewed in (Hudnall, 2016 #2009), and most recently in MS and neuroinflammation.¹³ Certainly, a single response, such as, up- or downregulation of signaling is not associated with either the ALS disease state, or the ability to ameliorate ALS-associated degeneration. More studies elucidating the action of each BMP and TGF-β/Activin signaling branch in ALS, and comparisons in the same cellular context, are necessary for future research.

6 |. IMPORTANCE OF SPATIAL AND COMPONENT-SPECIFIC ACTIVATION OF BMP SIGNALING

An interesting and potentially important aspect of BMP signaling that has emerged from detailed studies using the Drosophila larval motor circuit, is the presence of what appears to be two distinct pools of the phosphorylated Smad1/5/8 transducer, pMad, an indicator of active signaling. In filleted preparations of the intact wild type motor circuit, pMad is evident within motor neuron nuclei, in the ventral nerve cord (VNC), the spinal cord equivalent, as expected given its function as transcriptional regulator. pMad also accumulates in the presynaptic space of NMJs.^140,141 The functional difference between “nuclear pMad” and “synaptic pMad” has not been fully elucidated; however, several studies suggest that roles for BMP signaling in synaptic growth and synaptic function are separable.^{131,133,141,142} Interestingly, in the dSod1^G85R knock-in model, “synaptic pMad” is elevated indicating that BMP signaling is upregulated at the NMJ. However, no change in pMad levels is apparent in motor neuron nuclei.⁵⁰ Yet in the context of dSod1^G85R locomotor rescue by gbb (BMP5/6/7) expression, a significant increase in pMad within the VNC is observed but no change to the already elevated levels of “synaptic pMad”. A differential response of pMad in these two subcellular compartments is also apparent in two other Drosophila ALS models (TDP43 and VAPB), where “nuclear pMad” is also unaffected and “synaptic pMad” is changed, but in contrast to dSod1^G85R, in these models synaptic pMad is reduced.^143,144 In the TDP43 model, a degenerative phenotype elicited by the overexpression of wild type hTDP43 is associated with a defect in endosomal trafficking of the BMP type I Tkv receptor at the synapse thought to be responsible for the observed reduction in “synaptic pMad”.¹⁴⁴ Currently, few details about differential regulation of BMP signal transduction in distinct subcellular domains are known, especially in a cell as asymmetric as a neuron. The precise function of ‘synaptic pMad’ is not known albeit its presence is associated with postsynaptic clustering of GluR receptors.¹⁴¹ Are there distinct subcellular pools of pSmad1/5/8 or pSmad2/3 in vertebrate motor neurons? A major gap in our knowledge is how BMP/TGF-β signaling is differentially regulated in discrete subcellular domains and how those differences manifest themselves with regard to neuronal function.

Two other aspects of BMP/TGF-β pathway regulation in the nervous system that require further investigation are the mechanisms underlying and consequences of ligand-specific signaling, and transduction through Smad-independent pathways. In the lumbar spinal cords of SOD1^H46R rats, researchers have found that BMP4 levels are preferentially elevated, but levels of BMP2, 6, and 7 were unchanged.¹²⁹ In the Drosophila nervous system, pMad-dependent BMP signaling is activated in multiple cell types: muscles, motor neurons, and glia. The BMP5/6/7 ortholog, gbb, and the BMP2/4 ortholog, dpp, are each expressed by distinct sets of cells, and contribute differentially to the activation of signaling in target cells.^145,146 This kind of spatial difference in ligand production will be first regulated at the level of gene transcription, then by dimer formation, proprotein processing, post-translational modifications during trafficking, and secretion.^110,147 As such, in addition to BMP ligand maturation, their structural properties, and availability for receptor binding are also known to alter signaling output,^148,149 as demonstrated by the requirement for a specific Gbb proprotein proconvertase processing event that impacts CCAP (crustacean cardioactive peptide) neuropeptide production in CNS neurons.¹⁵⁰ Differential reception of each ligand type is further impacted by the presence/absence of different receptor types and their contribution to receptor complexes expressed at the cell. In addition to these forms of regulation, crosstalk between pathways is evident, as in the case of glial-expressed Activin, encoded by maverick (mav), where its contribution to the complexity of BMP signaling regulation and its influence on NMJ growth appears to occur through its modulation of gbb gene expression in muscles.¹⁵¹ Thus, the constellation of ligands, their bioavailability, and composition of receptor complexes, contribute to the tight regulation of signaling in neuronal function, providing a mechanism for a fine level of control.

7 |. INVOLVEMENT OF SMAD-INDEPENDENT SIGNALING

In addition to our growing knowledge of varied roles for different ligand forms, evidence for their ability to initiate signaling via Smad-independent pathways is also growing. While our current understanding is limited, links between BMP/TGF-β signaling and p38/MAPK in the nervous system are emerging. BMPs are important in inducing neuronal differentiation of PC12 cells, a process also shown to require p38 activation.¹⁵² Reduced neurodegeneration associated with an upregulation in BMP7 has been linked to an activation of pro survival p38/MAPK signaling in neurons.¹⁵³ Hyper-phosphorylation of neurofilaments is a hallmark of ALS and the finding that p38 phosphorylates neurofilaments in culture is intriguing.¹⁵⁴ Furthermore, p38 activation is observed in both patients and SOD1-ALS mutant mouse models,^155,156 and inhibitors of p38 have been shown to correct a defect in retrograde axonal transport present in SOD1^G93A mice.¹⁵⁷ To date, no direct connections have been made between upstream BMP/TGF-β signaling components, such as ligands and receptors, with p38 activity in ALS. p38 is a known regulator of Rab5, a key factor in early endosome biology,^158,159 and given the disrupted endocytic trafficking of the BMP Tkv receptor proposed in a Drosophila TDP43 model,¹⁴⁴ it would be interesting to measure p38 levels in that context.

Smad-independent BMP signaling has also been implicated in the regulation of actin dynamics at the tips of dendrites and axons in part through the direct association of LIMK with the cytoplasmic domain of the BMPRII receptor.¹⁶⁰ Stimulation of LIMK occurs in response to BMP ligand-receptor interactions in the absence of Smad involvement. As discussed below, activation of LIMK in response to BMP signaling has the potential to alter actin polymerization dynamics in both axons and dendrites.

8 |. CELLULAR PROCESSES AFFECTED BY BOTH ALS AND BMP/TGF-B SIGNALING

Overall, a better understanding of how components and mediators of BMP/TGF-β signaling are affected across a number of ALS mutations and models will be important for constructing the array of molecular consequences that link the two. When considering the possibility of amelioration of ALS-associated neurodegeneration by BMP/TGF-β pathway activation, it is also important to consider the overlap of cellular processes in which TGF-β/BMP signaling is known to act and those that are dysfunctional at early stages of ALS. In order to address the effectiveness of BMP/TGF-β signaling as a modulator of ALS, we must consider two important questions: Does amelioration of a single cellular process by BMP/TGF-β signaling modulation, provide phenotypic rescue across a number of different ALS mutations and models? Can the functions of multiple defective processes, that together contribute to neuronal degeneration in ALS, be restored by increasing or decreasing BMP/TGF-β signaling?

There are several areas for intersection of BMP/TGF-β signaling in processes commonly affected in ALS that are certainly worthy of further investigation. The role for BMP signaling in regulating actin dynamics through interactions between LIMK with the cytoplasmic domain of the type II receptor BMPRII receptor¹⁶⁰ is of interest. BMP7 stimulates LIMK activity and dendrite outgrowth, through a Smad-independent pathway that is activated within a specific subcellular domain. Phosphorylation of cofilin by LIMK inactivates its ability to sever and depolymerize F-actin to enrich the G-actin pool. The yang to cofilin’s yin, is profilin, which facilitates the polymerization of actin filaments by acting on the G-actin pool, balancing the actions of cofilin to control actin filament dynamics. Mutations in profilin have been identified in ALS patients¹⁶¹ implicating a disruption in actin polymerization as a mechanistic failure that could result in defective axon and dendrite growth and transport.^162,163 Consistent with such a model, phosphorylation of cofilin is enhanced in both ALS patients and C9orf72-ALS animal models. Further, C9orf72 gene function can modulate LIMK activity.¹⁶⁴ The conservation of process is evident in Drosophila where the interaction between the orthologous BMPRII receptor, Wit, and LIMK, is critical for NMJ stabilization,¹⁶⁵ and LIMK-dependent BMP signaling is required for axonal growth in the mushroom body.¹⁶⁶

Another cellular process impacted by both BMP/TGF-β signaling and ALS that requires more investigation is nucleocytoplasmic transport. Disruptions in nucleocytoplasmic transport in ALS are multifold and involve a failure in the proper nuclear localization of RNA-binding proteins, resulting in cytoplasmic aggregation of factors such as, TDP43, and peptides composed of dipeptide repeats (DPRs) produced by RAN translation of the G4C2 expansions in C9orf72-ALS. Abnormal nuclear membrane structure was first observed in spinal motor neurons of fALS and sALS patients with SOD1 mutations.¹⁶⁷ An association of the expanded G₄C₂ hexanucleotide repeat RNA with RanGAP, a mediator of nuclear pore dynamics, has been documented in a Drosophila ALS model, iPSC-derived neurons, as well as in C9orf72 ALS patient tissue and shown to rescue ALS-associated defects when over expressed.^55,168,169 Importin-β is a nuclear import receptor that regulates transport through RanGAP-regulated nuclear pores from the cytoplasm to the nucleus of a number of RNA-binding proteins, such as TDP-43 and FUS which exhibit aggregation-prone properties in ALS.^170,171 Mutations in importin-β result in defective NMJ structure and synaptic transmission in Drosophila.¹⁴⁰ Synaptic pMad is reduced in importin-β mutant animals and a genetic interaction identified between importin-β mutations and mutations in BMP signaling components, highlight the connection between importin-β function and BMP signaling in the motor neuron.¹⁴⁰ How importin-β, as a regulator of nucleocytoplasmic transport, specifically interferes with synaptic pMad abundance and not nuclear pMad levels in not clear. One key towards understanding the impact of BMP signaling on its effectors in different cellular domains may be an elucidation of how the cytoplasmic and nuclear transport of different pMad pools is regulated, coupled with what controls its localization at the NMJ, in wild type and diseased motor neurons. An intriguing parallel is the observation in both sALS and SOD1 fALS that the differential localization of pSmad2 vs pSmad3 to motor neuron inclusions is dependent on importin-β function.¹²⁵

9 |. EXCITING NEW RESEARCH TO BE DONE

In conclusion, the intersection between BMP-TGF-β signaling roles in nervous system function and the cellular disturbances arising from ALS is becoming more robust. Research directed at understanding the molecular underpinnings responsible for fine tuning signaling output will reveal how disruptions or breaks in this system arise in a diseased state. Importantly, this knowledge will identify entry points for signaling manipulation through targeted modulation, as well as markers of disease risk or onset that have the potential to aid in diagnosis. Important questions to be addressed include: Can modulation of BMP/TGF-β signaling alleviate dysfunction of multiple cellular processes affected by ALS? And, is modulation of the pleiotropic nature of BMP/TGF-β signaling effective in its suppression of neurodegeneration across all forms of ALS? We identify several particularly intriguing connections between processes regulated by BMP/TGF-β signaling, such as the dynamics of actin polymerization, and nucleocytoplasmic transport, with their known dysfunction in ALS. Research elucidating the mechanistic basis of these intersections of ALS compromised cellular processes which roles for BMP/TGF-β signaling, is a top priority.