Are ATXN2 variants modifying our understanding about neural pathogenesis, phenotypes, and diagnostic?

ATXN2 gene encodes a cytosolic protein (ataxin-2) with pleiotropic functions (see below). This gene contains a number of exonic Cytosine-Adenine-Guanine (CAG)-repeats which encodes a polyglutamine tract (polyQ) in the N-terminal intrinsically disordered region (IDR) of the protein. ATXN2 CAG repeats are interrupted by CAA codons which is relevant only for DNA and RNA but not for protein since CAA also encodes glutamine (Q).

The ataxin-2 polyQ lengths vary in the healthy population, with 22Q and 23Qs as the most frequents variants. ATXN2 polyQ length variation is associated with smaller volume of some brain structures (putamen, thalamus, amygdala and globus pallidus) potentially affecting cognition in old age. Genome Wide Association Studies have shown that genetic variants in ATXN2 and its surrogate ATXN2L are associated with cognition and educational attainment (reviewed in Laffita-Mesa et al., 2021a).

When polyQ length surpasses certain thresholds, it becomes toxic to specific neuronal populations (main target Purkinje cells) reverberating in multiple neurological diseases. For instance, the presence of at least one allele of > 22polyQ lowers the Age at Onset (AO) in patients with transthyretin familial amyloid polyneuropathy with the Val30Met variant (Santos et al., 2019).

Intermediate-length ATXN2 polyQ (29-to-31Q) modulate AO in SCA3/MJD and is a proven disease modifier for C9ORF72-ALS (van Blitterswijk et al., 2014; de Mattos et al., 2019). Hence, intermediate-length ATXN2 polyQ confers an increased risk for a broad number of neurodegenerative diseases including amyotrophic lateral sclerosis (ALS), Parkinson’s disease, progressive supranuclear palsy, Alzheimer’s disease, frontotemporal dementia and multiple system atrophy (Figure and ) (Lee et al., 2011; Ross et al., 2011).

ATXN2, genotype-phenotype relationship and molecular genetics.

(A) Neurological diseases and variable phenotypes associated with CAG expansions in ATXN2. PubMed search (1992–2021) was made included the terms: ATXN2, SCA2 and Ataxia (ATX), Amyotrophic Lateral Sclerosis, (ALS), Parkinson’s disease and Parkinsonism (PD), Frontotemporal Dementia (FTD), Alzheimer’s Disease (AD), Multisystem atrophy (MSA), Open Angle Glaucoma (OAG), Epilepsy (EP), Schizophrenia (SCZ), Obesity (OB), Diabetes (DB). For plotting purposes, the searched data was divided in two periods 1992–2006 and 2007–2021. (B) Genotype phenotype relationship in healthy population and ATXN2 related diseases. CAG repeats range 29–34 is found for ALS, FTD, Cortico Basal Syndrome (CBD), MSA and familial amyloid polyneuropathy (FAP). (C) Genetic anticipation in SCA2 and intergenerational instability. (D) ATXN2-S/AS locus with diverse genetic variants located in the close vicinity of the CAG repeats (star colored in orange gradient). Genetic alterations fuchsia colored is in the same locus 12–111599394. Variants colored in green are included or overlapped with the UH13/UH10 primer binding site. At the protein level all these variants have consequences in the ataxin-2 protein, potentially modifying the N-terminal intrinsically disorder region. (E) Map for the ataxin-2 functional domains, disordered regions, interactors molecules and physiological functions. (F) ATXN2 splice variants with the highest expression in the brain tissue and spinal cord of healthy donors (Gtex). Only the longest ATXN2 splice variants were included in the scheme.

PolyQ length ≥ 32Qs is the monogenic cause for spinocerebellar ataxia Type 2 (SCA2) (Imbert et al., 1996; Pulst et al., 1996). Furthermore, the polyQ repeat ATXN2 variation explains ~40–60% of Age of disease Onset (AO) and others SCA2 phenotype markers (). In SCA2 CAG-repeats are unstable in both mitotic and germline tissues showing a marked bias toward larger repeats expansions in paternal transmissions which underlies the genetic anticipation phenomenon ().

Therefore, the disruption of ataxin-2 functions caused by polyQ expansion, may affect multiple brain structure/circuits involved in motor, emotions, cognitive functions and may also affect metabolic pathways.

Is the ataxin-2 polyQ tract the only quilt for different neural phenotypes? PolyQ expansions are present in the coding regions of other eight genes causing other neurodegenerative diseases such as Ataxin-1/SCA1, Ataxin-3/SCA3, CACNA1A/SCA6, Ataxin-7/SCA7, TATA-binding protein/SCA17, atrophin-1/dentatorubral-pallidoluysian atrophy, Huntingtin/Huntington disease and androgen receptor/spinobulbar muscular atrophy. However, none of those genes does have the pleiotropism and association with multiple neurological diseases compared to ATXN2 polyQ tracts. Lee et al. (2011) hypothesized that this effect is related to the biology of ataxin-2 or ataxin-2/specific interactions, rather than the presence of an expanded polyQ repeat per se. Although their hypothesis is restricted to explain the associated increased risk for ALS conferred by intermediate-lengths, it could be generalized to other diseases associated with variants in ATXN2. Another aspect to consider is the presence of different ATXN2 allelic variants, in cis or trans to the polyQ, accounting to phenotype variability and disease severity. This perspective uses our (Laffita-Mesa et al., 2021b) and other (Inada et al., 2021) recent findings on novel ATXN2 variants, outside of the polyQ domain, but located in the N-terminal IDR, aiming at understanding its contribution to phenotypes variability, modifying effect, ataxin-2 function and diseases management.

Ataxin-2 is a cytosolic protein with pleiotropic functions: Human ataxin-2 is a large 140 kD cytosolic protein with ubiquitous expression, high expression in neural tissues, and the cerebellum having one of the highest levels (https://gtexportal.org/home/gene/ATXN2). This protein is mainly localized in the trans-Golgi network (main functional domains are shown in Figure and ). It has been difficult to elucidate its 3D structure and physiological functions as it is not possible to obtain crystal structures. The current heterologous expression systems are not capable to express this large protein. In silico 3D predictions are consistent with an intrinsically disordered protein (). There are multiple mouse models that, together with fly and yeast, demonstrate that ataxin-2 is a cytosolic protein crucial for RNA metabolism. Ataxin-2 binds to polysomes while activating and stabilizing targets transcriptomes by polyadenylation. Ataxin-2 is also constitutive factor for RNA and stress granules. RNA granules are very dynamic higher order structures regulating the timing, trafficking, and transcriptomic expression. The main interacting proteins for these functions are polyadenylate-binding protein 1, TAR DNA-binding protein 43 (TDP-43), tia1 cytotoxic granule-associated rna-binding protein-like 1, Eukaryotic initiation factors, Staufen, fragile X mental retardation 1, Ataxin-2-binding protein 1 (A2BP1 also known as RBFOX1), RNA-binding protein FUS/TLS (Fused in Sarcoma/Translocated in Liposarcoma) and Ataxin-2-like protein. Furthermore, ataxin-2 functions as activator of mRNA that controls circadian rhythm, but also repress CamKII translation influencing synaptic plasticity and memory (Lim and Allada, 2013). Yeast ataxin-2 proteins were observed as cloud-like structures surrounding the mitochondria functioning as redox sensor, coupling autophagy to the metabolic state of mitochondria (Lin et al., 2020). Moreover, ataxin-2 is involved in endocytic processes (CIN85 and endophilins-1/2 as interactors) and lipid metabolism () (reviewed in Scoles and Pulst, 2018).

Funetional consequences and genetic mechanisms of ataxin-2 variants in the W-terminal IDR.

(A) 3D Alpha Fold prediction for the human ataxin-2. The polyQ (Q166-Q188) is indicated as an alfa helix in light blue. The LSM motifs harbor residues from Q254-S345, the LSM-AD F353-Q475. (B) 3D structure prediction for the region R37-S38-G39 where 9bp duplication is located. (C) Prediction for the region P145-A151 with 21bp duplication. The model confidence for regions dark blue colored is very high (predicted Local Distance Difference Test, pLDDT > 90), light blue Confident (90 > pLDDT > 70), yellow Low (70 > pLDDT > 50) and very low (pLDDT < 50) for orange colored. The regions below 50 pLDDT may be unstructured in isolation. (D) ATXN2 genetic variations and allelic interaction, polyQ and hybrids repeats. (E) Phenotype modifier effect of the 9bp dup, R37-S38-G39 -R40-S41-G42-R. The mutation is in the IDR located in the N-terminus and it was found to decrease the age at onset for SCA3 and C9ORF72 ALS. (F) For the ATXN2-AS it was found aberrant splicing in carriers of the 9bp duplication. (G) Several nucleotide changes are in the primer binding site for the most common used primer set limiting the accuracy in the genetic testing services. (H) Liquid-Liquid transition for hydrogel formation aided by IDR in ataxin-2. (I) Observed effect of the 21-bp dup in cell viability. Inada et al., 2021 found an ataxic case with normal ataxin-2 polyQ expansions. The genotype was 22/22CAG, but one allele had a 21-bp duplication doubling the P145-A151 tract close to the polyQ tail in ataxin-2. This makes one allele with 43 repeats units of hybrid stretch, polyQ and a double P145-A151 tract and the other allele with 22 units only. In a set of functional experiments, they transfected cells separately with wild type and the 21-bp dup followed by treatment with increasing amounts of H2O2. Cells with the duplication showed decreased survival, TDP43 mislocalization and increased susceptibility to cellular stress. (J) Other studies shown that the oxidation of ataxin-2 IDR may disrupt the ataxin-2 function in coupling mitochondrial redox state and autophagy (Lin et al., 2020).

Emerging Current pathophysiological mechanisms are built on models with either hyperexpanded polyQ repeats (gain-of-function) or ablation of the whole gene (loss-of-function). Other studies have focused in ataxin-2’s Like-Sm and Like-Sm Associated domains. The SCA2 field has been CAG/polyQ centric and apart the interspersed CAA interruptions within the CAG repeats and the two single nucleotide polymorphisms (SNPs) rs695871 and rs695872 no other variants were shown as relevant contributors to neural phenotypes.

A recent work examined ATXN2 variants as modulators of phenotype variability in different neurodegenerative disease cohorts (SCAs, C9ORF72-ALS, SCA3, and Parkinson’s Disease) (Laffita-Mesa et al., 2021b). That work identified a 9bp-duplication (c.109_117dupCGGAGCGGG) in the ATXN2-S/ATXN2-AS region as potential modulator for C9ORF72-ALS and SCA3 which are otherwise modified by intermediate ATXN2-CAG repeats. The structural change encodes for the R37-S38-G39 resulting in R- R37-S38-G39 -R40-S41-G42-R, located in the ataxin-2 N-terminus (Figures , and ). The variant was in trans with intermediate alleles, suggestive for pathological allelic interaction (lower section in ).

In C9ORF72-ALS is co-expressed with CAG/CAA sequence mosaic consisting of 8CAG–1CAA–4CAG–1CAA–8CAG and 7CAG–2CAA–4CAG–1CAA–8CAG. This may have potential toxic consequences at DNA/RNA levels. McGurk et al. (2021) showed that mechanisms associated with the purity of the sequence of the ataxin-2 CAG repeats tract contribute to disease (McGurk et al., 2021).

Droplet digital PCR in retrotranscribed mRNA supported 9bp-dup expression consistent with use of the ATG at 495bp upstream the CAG repeat (Transcript IDs in Gtex: ENST00000608853.5, ENST00000483311.5, ENST00000616825.4, ENST00000550104.5, ENST00000377617.7). However, there is another ATG 15 bp upstream to the CAG repetitions encoding for (ENST00000535949.5, ENST00000389153.8, ENST00000549455.1, ENST00000542287.6) which may not include the variant and produce a smaller ataxin-2. As shown in , the longest splice variants ENST00000608853.5 and ENST00000535949.5 are widely expressed in brain tissues and spinal cord in healthy donors (https://gtexportal.org/home/gene/ATXN2).

The locus for the 9-bp duplication (12-111599394, GRCh38) is variable in different European populations at GnomAD and different protein alterations are predicted. For instance, the variant rs776634254, 12-111599058-C-A GRCh38 may introduce stop codons (p. Glu153Ter). Hence, the different variants may modify the expressivity of other neurological diseases. Our ongoing studies are aimed for the above and other variants in the close vicinity rs776634254, rs764588879, rs552050500 and rs769578492 interfering with the primer binding sites (UH13/UH10) ().

We have provided evidence that the small 9-bp duplication could be a phenotype modifier in neural diseases (SCA3 and C9ORF72-ALS) (). For this we ruled out other modifiers and found that cases with the duplication develop SCA3 and ALS earlier than others with the same mutation (Laffita-Mesa et al., 2021b). Nonetheless, larger cohort studies are necessary to better control for effects of the environment and to define the contribution to the overall disease variance in SCA2, SCA3 or other diseases connected with ATXN2. This limitation may be sorted out through international collaborations as SCAs, and genetically determined ALS are rare diseases.

The transcriptomic profile for the Natural Antisense Transcript ATXN2-AS in carriers shows differences compared to controls (Laffita-Mesa et al., 2021a) (). Ongoing studies are aimed to better define sequence differences for these ATXN2-AS splice variants throughout brain tissues targeted in SCA2 and ALS.

All these genetic alterations are in the intrinsically disordered domain (Figures and ). These regions are crucial for liquid-liquid phase transitions, for hydrogel formation and RNA granules formation. Liquid-liquid phase transitions are determined by weak interactions among intrinsically disordered regions of proteins with mRNAs resulting in condensed proteinaceous membrane-less organelles containing halted transcriptomes bound to ribosomal units, translation factors and RNA binding proteins (RBP). The vast majority are cytosolic transcriptomes which are condensed with RNA binding proteins and are transported bound to motor proteins along axons to their destination at the synaptic terminals and dendrites for further processing, catalysis and/or trafficking. IDR are also prone to assemble into amyloid like fibers and is increased when RBP are mutated ().

Genetic alterations in the ataxin-2 N-terminal IDR could be in cis or trans with the polyQ repetitions with different interacting allelic effects (). At the protein level this is translated into hybrids repetitive stretches with and/or without non-expanded polyQ. These polypeptides may aggregate, disturbing TDP43 localization, and harmfully sensitizing cells at the same extent of full CAG expansions. For instance, the combination of the 21bp duplication c.432_452dup, Pro145-Ala151 in cis with 22CAG repeats in ATXN2 was presented in an ataxic individual with no other causative mutation (Inada et al., 2021) (Figure and ).

Finally, the oxidation of ataxin-2 C-terminal IDR may also disrupts the ataxin-2 function in coupling mitochondrial redox state and autophagy (Lin et al., 2020) ().

Future perspectives Abnormal expansions of CAG, CUG and polyQ tracts located in the ATXN2 remain as the main causative and modifying factor for several diseases. Notwithstanding the above, other genetic changes in the same region deserve attention as our knowledge about ATXN2 pathogenesis and pleiotropism is still limited. Structural genetic changes may contribute to disease presentation when accompanying either intermediate or full expanded ATXN2 allele. This is a general genetic mechanism that may also involve protective genetic variants. The latter is still an unexplored area in the field. The use of CRISPR Cas-9 libraries combining the different allelic versions, SNPs, CAG/CAA mosaics, and the hitherto discussed variants in Induced Neuronal Stem Cells will aid to understand their role in ATXN2 related diseases and traits. Genetic testing will be also benefited as some alterations may interfere with primer binding sites ().

We apologize to authors whose work were either not referenced in this perspective or cited as part of review articles due to space limitations.

Parkinsonfonden and ALF support from Region Stockholm. PS was a Wallenberg Clinical Scholar. MP’s research was supported by Region Stockholm.