Mid-Gestation lethality of Atxn2l-Ablated Mice

Depletion of yeast/fly Ataxin-2 rescues TDP-43 overexpression toxicity. In mouse models of Amyotrophic Lateral Sclerosis via TDP-43 overexpression, depletion of its ortholog ATXN2 mitigated motor neuron degeneration and extended lifespan from 25 days to >300 days. There is another ortholog in mammals, named ATXN2L (Ataxin-2-like), which is almost uncharacterized but also functions in RNA surveillance at stress granules. We generated mice with Crispr/Cas9-mediated deletion of Atxn2l exons 5-8, studying homozygotes prenatally and heterozygotes during aging. Our novel findings indicate that ATXN2L absence triggers mid-gestational embryonic lethality, affecting female animals more strongly. Weight and development stages of homozygous mutants were reduced. Placenta phenotypes were not apparent, but brain histology showed lamination defects and apoptosis. Aged heterozygotes showed no locomotor deficits or weight loss over 12 months. Null mutants in vivo displayed compensatory efforts to maximize Atxn2l expression, which were prevented upon nutrient abundance in vitro. Mouse embryonal fibroblast cells revealed more multinucleated giant cells upon ATXN2L deficiency. In addition, in human neural cells, transcript levels of ATXN2L were induced upon starvation and glucose and amino acids exposure, but this induction was partially prevented by serum or low cholesterol administration. Neither ATXN2L depletion triggered dysregulation of ATXN2, nor a converse effect was observed. Overall, this essential role of ATXN2L for embryogenesis raises questions about its role in neurodegenerative diseases and neuroprotective therapies.


Introduction
In all eukaryotic organisms, at least one copy of the Ataxin-2 gene (gene symbol ATXN2 in humans) is phylogenetically conserved and serves roles during nutrient stress for RNA surveillance [1]. A conserved Lsm and LsmAD motif enables direct interaction with RNAs, and a PAM2 motif mediates association with the poly(A)-binding protein PABPC1 [2,3]. Thus, most Ataxin-2 protein localizes with mRNAs at the rough endoplasmic reticulum with marker Ribosomal Protein S6 (RPS6 aka S6R) during cell growth periods [4], where its absence leads to expression adaptations of the associated ribosomal translation machinery [5] and modulates the phosphorylation control of translation [6]. During cell stress, e.g., from nutrient deprivation, Ataxin-2 is transcriptionally induced [6] and relocates with the small ribosomal subunit and PABPC1 to stress granules (SG) [7]. The RNA surveillance function of Ataxin-2 seems to be relevant to protect against the translation of viral RNAs, given that poliovirus is optimized to cleave Ataxin-2 [8]. In mammals, all these protein structure domains are also found in its paralog Ataxin-2-like (gene symbol ATXN2L in humans). Both ATXN2 and ATXN2L mRNAs also conserved an alternatively spliced exon, which encodes a proline-rich domain (PRD) that mediates its direct association with SH3 motifs in components of the growth factor receptor endocytosis apparatus [9][10][11]. Abnormal ATXN2 splicing and alternative polyadenylation were documented in diseases with RNA toxicity, such as amyotrophic lateral sclerosis (ALS) [12]. The common ancestor of both proteins in yeast and Caenorhabditis elegans was observed to suppress growth signaling via mTORC1, modulating cell size, and lipid stores [13][14][15]. This reprogramming of nutrient metabolism is accompanied by an important influence on the mitochondrial breakdown of fatty acids and amino acids, as well as glucose utilization [16][17][18][19], probably mediated by the direct protein interaction of ATXN2 with the cytosolic enzyme BCAT1 [20] as the rate-limiting factor in the breakdown of leucine, isoleucine, and valine.
The role of ATXN2 in neurodegenerative diseases has triggered intense research over the past 25 years. Exclusively in human ATXN2, an N-terminal domain with 22 consecutive glutamines (polyQ) exists, which can undergo expansion mutations across generations. Large expansions beyond the size of 32Q trigger the multi-system nervous tissue atrophy Spinocerebellar ataxia type 2 (SCA2), while intermediate expansions of sizes 27Q-32Q increase the risk to be affected by motor neuron diseases such as ALS, fronto-temporal lobar dementia (FTLD) [21][22][23][24][25] or by other tauopathies and Parkinson's disease variants like progressive supranuclear palsy (PSP) [26,27]. Conversely, the depletion of ATXN2 by knock-out or by injection of antisense-oligonucleotides has a massive neuroprotective effect in yeast/fly/mouse models of ALS and FTLD, as well as in SCA2 and SCA1 fly models [24,[27][28][29]. In addition, in yeast, depletion of the ATXN2/ATXN2L ortholog PBP1 rescues the lethal effect of poly(A)-binding protein deletions [30]. The constitutive knock-out of Atxn2 in mice leads to progressive weight gain with excessive storage of lipid droplets and glycogen in the liver, elevated cholesterol and other lipids in the blood, beta-cell hyperplasia in the pancreas with hyperinsulinemia and insulin resistance, increased ganglioside and sulfatide lipids in the brain myelin, locomotor hyperactivity, and mild infertility with gender-dependent impairment of embryogenesis [31,32].
In view of the importance of ATXN2 orthologs for stress response, redundancy occurred in land plants and in vertebrates (except birds) by the co-existence of two homologous genes, named CID3-CID4 in Arabidopsis thaliana weed and ATXN2-ATXN2L in humans and rodents [33]. ATXN2L protein dimerizes with ATXN2 in yeast-two-hybrid tests, and is also a regulator of SGs and mRNA processing during starvation periods, but shows more co-localization with the nuclear splice apparatus than ATXN2 due to an arginine-dimethylation [34,35]. Similar to ATXN2, ATXN2L associates with plasma membrane receptors in dependence on their phosphorylation status, is involved in epidermal-growth-factor (EGF)-receptor signaling, and exists in several isoforms [36][37][38]. Little more is known at present about ATXN2L. Database mining at the STRING web platform for Protein-Protein Interaction Networks and Functional Enrichment Analysis, available online at: https://string-db.org/ [39] confirms that human and mouse ATXN2L show direct protein-protein-interaction with the poly(A)-binding-proteins PABPC1/2, PABPC1L/PABPC3/PABPC4/PABPC6, with further stress-granule-components such as G3BP2, DDX6, and LSM12 [40,41], as well as nuclear RNA-binding protein NUFIP2 and nuclear transcription modulator NFATC2IP. The RNA helicase activity of DDX6 enhances the expression of growth factor receptors such as HER2 and FGFR2 [42]. Possibly by this mechanism, ATXN2 was found to modulate surface receptors [43]. A recent global protein interactome study via FLAG-tag affinity purification and mass spectrometry in HEK293 cells also showed ATXN2L to associate with DDX6 and the proline-rich-domain-containing RNA-binding Ras-GTPase G3BP2, but in addition, demonstrated ATXN2L binding to KLHL20 [20], which is responsible for the degradation of the autophagy/growth modulators DAPK1 and ULK1, as well as of the Rho Guanine nucleotide exchange factor and glutamate-transport-/neurite-growth-modulator ARHGEF11. STRING also lists a potential interaction (based on association in lower species) with SLC9A3R2/NHERF2 that is known to act as PSD-95 scaffold to control albumin endocytosis, the glutamate transporter GLAST, and the metabotropic glutamate receptor mGluR5, in addition to activating Src phosphorylation and modulating the high density lipoprotein receptor SR-B1 that is crucial for cholesterol metabolism [44][45][46][47][48].
Further data mining at the Allen brain atlas, available online at: https://mouse.brain-map.org/ [49] retrieves evidence that Atxn2l mRNA has a particularly strong expression in cerebellar Purkinje neurons. Protein data in different cell types from the Human Protein Atlas, available online at: https://www.proteinatlas.org/ [50] demonstrate its prominent abundance in testis. Its moderate abundance in postmitotic neurons from the cortex, cerebellum, hippocampus, and caudate nucleus, as well as glia and muscle cells were detected by the Atlas antibody HPA041506 targeting ATXN2L at aa 456-547 (in the Uniprot Q8WWM7-1 sequence), which exhibits a single immunoblot band of predicted size and has siRNA-controlled specificity. However, it was not detected by Atlas antibody HPA043391, which targets human ATXN2L aa 572-644, but does not exhibit immunoblot signals of correct size and was not tested by siRNA, as can be seen within the Human Protein Atlas online at: https://www.proteinatlas.org/ENSG00000168488-ATXN2L/tissue/primary+data [50]. Subcellular analyses with both antibodies showed a diffuse cytosolic signal. Among cell lines, a particularly strong expression is detected under lymphoid differentiation, in good agreement with the importance of RNA surveillance for the innate and adaptive immune system. The upregulation of ATXN2L in cancer tissue is an unfavorable marker, in agreement with its published induction by EGF growth signaling [38]. A co-expression survey at the Coexpedia database, online available at: www.coexpedia.org [51] shows that mouse Atxn2l mRNA is regulated together with pathways of thin myocardium, open neural tube, improved glucose tolerance, immune defense, and cell cycle. The RNA binding spliceosome component Srrm2 and the proline-rich inflammatory splicing factor Prrc2a show the highest co-expression score.
In a recently generated mouse with the aggregation of ATXN2 due to a polyglutamine (Q100) expansion knock-in [52], we observed a significant accumulation of ATXN2L peptides in cerebellar global proteome profiles from old animals. It was difficult to know if the impact of this anomaly has pathogenic or compensatory roles, and it was impossible to study it further in view of the unavailability of antibodies shown to be specific, of tagged recombinant constructs and mutant cells. We decided to start an exploratory project into the physiological functions of ATXN2L. Working at organism level, we generated novel Atxn2l-ablated mice to assess their phenotypes in growth, histology, and behavior, in comparison to our previous study of Atxn2-ablated mice [31]. Working at cell culture level, we examined the cellular phenotype of Atxn2l −/− mouse embryonal fibroblasts (MEF), and whether the regulation of human ATXN2L and ATXN2 mRNA levels is similar in response to stressors for neural cells. Overall, the data indicate that ATXN2L is much more important than ATXN2 for growth at early embryonal age, particularly in females. Atxn2l exon 5-8 deletion triggered a growth deficit phenotype rather than the weight excess observed in mice with Atxn2 exon 1 deletion. Apparently, ATXN2L has a similar function as ATXN2, which is not only transcriptionally induced upon deprivation from nutrients and especially lipids, but also upon exposure to excess glucose or amino acids. Hence, we believe that elucidating the native role of ATXN2L throughout development and in response to stress provides important insights into the complementary spatio-temporal roles of ATXN2-ATXN2L and will further our understanding of neurodegenerative pathomechanisms.

Results
We obtained a null allele of murine Atxn2l via Crispr/Cas9-mediated deletion between two sgRNA sites flanking the exons 5-8 of the gene (outsourced work by TIGM, College Station, TX, USA) (see Figure 1). The absence of ATXN2L protein epitopes from residue 150 until residue 1050 was confirmed in Atxn2l −/− mouse embryonal fibroblast (MEF) (Figure 2A). ATXN2L abundance was higher in cerebral cortex than in cerebellum. While MEFs showed two specific ATXN2L bands in immunoblots, in brain tissue the larger band dominated ( Figure 2B). The separate quantification of several Atxn2l mRNA exon junctions, namely 7-8 to represent the deleted region, 10-11 further downstream, and 1-2 upstream versus the alternatively spliced 1c-2 suggested the Atxn2l −/− limb cells to sense the loss-offunction and either trigger the Atxn2l promoter or minimize the Atxn2l mRNA decay, thus maximizing the levels of the main Atxn2l mRNA variant. This feedback did not occur in MEF lines that are maintained in nutrient abundance ( Figure 2C), suggesting that deprivation of overall energy or specific nutrients modulates Atxn2l transcription. The absence of ATXN2L protein epitopes from residue 150 until residue 1050 was confirmed in Atxn2l −/− mouse embryonal fibroblast (MEF) (Figure 2A). ATXN2L abundance was higher in cerebral cortex than in cerebellum. While MEFs showed two specific ATXN2L bands in immunoblots, in brain tissue the larger band dominated ( Figure 2B). The separate quantification of several Atxn2l mRNA exon junctions, namely 7-8 to represent the deleted region, 10-11 further downstream, and 1-2 upstream versus the alternatively spliced 1c-2 suggested the Atxn2l −/− limb cells to sense the loss-of-function and either trigger the Atxn2l promoter or minimize the Atxn2l mRNA decay, thus maximizing the levels of the main Atxn2l mRNA variant. This feedback did not occur in MEF lines that are maintained in nutrient abundance ( Figure 2C), suggesting that deprivation of overall energy or specific nutrients modulates Atxn2l transcription. The protein was completely absent in three assessed epitopes from residue 150 until the most C-terminal region (aa: amino acid). (B) Protein abundance of ATXN2L in heterozygous Atxn2l +/− cerebellum (cb) compared to wildtype littermates, or in heterozygous Atxn2l +/− cortex (ctx), in comparison to one wildtype Atxn2l +/+ MEF line. MEFs exhibit two specific bands of ATXN2L, in contrast to brain tissue where the larger band dominates. The overall ATXN2L levels were decreased to approx. 50% in heterozygous tissues (n = 2). Detection was done for the epitope aa 712-1050. (C) In Atxn2l −/− MEF (upper row, n = 4vs4), RT-qPCR showed that mRNA expression of Atxn2l was absent at the boundary between exons 7-8 that was genetically deleted, and significantly reduced at the boundaries of exons 1-2 and 10-11, possibly due to nonsense-mediated RNA decay. In tissue from fore-/hind-limbs [lower row, 3 wildtype (Atxn2l +/+ , white bars), 4 heterozygous (Atxn2l +/− , grey bars), 2 Atxn2l-null embryos (Atxn2 −/− , black bars)], the Atxn2l mRNA expression measured at the boundary of exons 7-8 was again absent in Atxn2l −/− and reduced by 50% in Atxn2l −/+ mice, whereas 5′-upstream and 3′-downstream from the deletion the increased Atxn2l expression level of null mutants suggested transcript upregulation efforts either via the Atxn2l promoter or via altered mRNA stabilization/degradation to compensate the loss of protein function. The exon 1c-2 assay detects an alternatively spliced isoform. Asterisks represent significance (p < 0.01 **, p < 0.0001 ****).
While the heterozygous mice bred normally, a lethality at the embryonic/prenatal stage was observed for the homozygous mutants. There was also reduced viability of heterozygous mice, particularly for females at the prenatal stage (see Tables 1,2). The protein was completely absent in three assessed epitopes from residue 150 until the most C-terminal region (aa: amino acid). (B) Protein abundance of ATXN2L in heterozygous Atxn2l +/− cerebellum (cb) compared to wildtype littermates, or in heterozygous Atxn2l +/− cortex (ctx), in comparison to one wildtype Atxn2l +/+ MEF line. MEFs exhibit two specific bands of ATXN2L, in contrast to brain tissue where the larger band dominates. The overall ATXN2L levels were decreased to approx. 50% in heterozygous tissues (n = 2). Detection was done for the epitope aa 712-1050. (C) In Atxn2l −/− MEF (upper row, n = 4vs4), RT-qPCR showed that mRNA expression of Atxn2l was absent at the boundary between exons 7-8 that was genetically deleted, and significantly reduced at the boundaries of exons 1-2 and 10-11, possibly due to nonsense-mediated RNA decay. In tissue from fore-/hind-limbs [lower row, 3 wildtype (Atxn2l +/+ , white bars), 4 heterozygous (Atxn2l +/− , grey bars), 2 Atxn2l-null embryos (Atxn2 −/− , black bars)], the Atxn2l mRNA expression measured at the boundary of exons 7-8 was again absent in Atxn2l −/− and reduced by 50% in Atxn2l −/+ mice, whereas 5 -upstream and 3 -downstream from the deletion the increased Atxn2l expression level of null mutants suggested transcript upregulation efforts either via the Atxn2l promoter or via altered mRNA stabilization/degradation to compensate the loss of protein function. The exon 1c-2 assay detects an alternatively spliced isoform. Asterisks represent significance (p < 0.01 **, p < 0.0001 ****).
While the heterozygous mice bred normally, a lethality at the embryonic/prenatal stage was observed for the homozygous mutants. There was also reduced viability of heterozygous mice, particularly for females at the prenatal stage (see Tables 1 and 2). The dissection at different stages of pregnancy showed the death of homozygous mutants to have mostly occurred by embryonic day E14-16. Female Atxn2l −/− embryos were more severely affected, most of them did not reach E11 ( Table 2). These observations were in good agreement with reports in C. elegans worms that embryonal patterning is arrested, stem cell proliferation is reduced, and germline abnormally masculinized upon depletion of the ortholog ATX-2 [53,54]. Interestingly, at the 2-4 cell stage of bovine embryos, a 2.4-fold expression difference of Atxn2l was recently observed to correlate with fertility and that the knock-down of Atxn2l enhanced the development of blastocysts [55]. Table 2. Embryonic day E11-20 genotype distribution for offspring from Atxn2l +/− intercrosses. Most lethality occurred before E11, with females dying before males. In the male homozygotes surviving beyond E11, there was a weight loss of 25% on average (Table 3). Table 3. Weight of embryos in mg directly after dissection at three different embryonal ages. n: Only one dead male Atxn2l −/− embryo was identified at E20 ( Figure 3A, right side), exhibiting disintegrating tissue surrounded by dense liquid as expected during resorption processes [56]. An Atxn2l −/− embryo at E16 was observed to display retarded growth ( Figure 3B, right side). The Atxn2l −/− embryos were not pale, had livers of usual size, and showed normal blood-filling of liver tissue ( Figure 3C, right side), so there was no evidence for the cardio-vascular or blood circulation deficits that are frequent causes of early or mid-gestational lethality [57]. Strong developmental differences between Atxn2l −/− versus WT littermates ( Figure 3D) sometimes amounted to several Theiler stages [58][59][60].

Weight of Embryos [mg]
No selective organ affection in gross morphology was noticed during embryogenesis ( Figure 3A-D), so an evaluation by high-resolution episcopic microscopy [61] was deemed unnecessary. In general, many fetal sacs in utero were found to contain only placenta rests and a fluid-filled cavity. Histological analyses at different embryonic stages from approximately E13 via E14 to E20 ( Figure 4) confirmed that the growth deficit is generalized throughout the embryos. Histological analyses at different embryonic stages from approximately E13 via E14 to E20 ( Figure 4) confirmed that the growth deficit is generalized throughout the embryos.
The most frequent causes of early/mid-gestation embryonic lethality before E14 include deficient placentation, followed by blood vessel/heart, skeleton/joint and nervous system malformations [62]. In Atxn2l −/− embryos, no obvious phenotypes apart from developmental delay were detected upon hematoxylin/eosin (H&E) staining of placenta (Supplementary Materials Figure S1), heart, and bones. Interestingly, mice with genetic ablation of the EGF receptor, which was observed in cell lines to influence ATXN2L expression, also exhibit a mid-gestational lethality. However, they display reduced placenta size [63,64] in contrast to the normal Atxn2l −/− placentas. Mid-gestation embryonic lethality was recently also documented for factors of endocytic uptake, such as DENND1A and VPS54 [65,66].
In Atxn2l −/− embryos, preferentially the nervous tissue showed frequent apoptosis, together with reduced cortical layer formation ( Figure 5). Similar mid-gestation lethality with abnormal development of the brain cortex was observed, e.g., upon deficient methylation of eukaryotic mRNAs [67]. Neurons more than other cells have a need to modify the 3 -untranslated regions of mRNAs via alternative polyadenylation and microRNA repression [68,69] in a pathway that involves the ATXN2/ATXN2L indirect interactor TDP-43 and stress granules. The most frequent causes of early/mid-gestation embryonic lethality before E14 include deficient placentation, followed by blood vessel/heart, skeleton/joint and nervous system malformations [62]. In Atxn2l −/− embryos, no obvious phenotypes apart from developmental delay were detected upon hematoxylin/eosin (H&E) staining of placenta (Supplementary Materials Figure S1), heart, and bones. Interestingly, mice with genetic ablation of the EGF receptor, which was observed in cell lines to influence ATXN2L expression, also exhibit a mid-gestational lethality. However, they display reduced placenta size [63,64] in contrast to the normal Atxn2l −/− placentas. Mid-gestation embryonic lethality was recently also documented for factors of endocytic uptake, such as DENND1A and VPS54 [65,66].
In Atxn2l −/− embryos, preferentially the nervous tissue showed frequent apoptosis, together with reduced cortical layer formation ( Figure 5). Similar mid-gestation lethality with abnormal development of the brain cortex was observed, e.g., upon deficient methylation of eukaryotic mRNAs [67]. Neurons more than other cells have a need to modify the 3′-untranslated regions of mRNAs via alternative polyadenylation and microRNA repression [68,69] in a pathway that involves the ATXN2/ATXN2L indirect interactor TDP-43 and stress granules. . H&E histology at E14 showed good separation of neuron layers in the cortex of Atxn2l +/+ brain (left); in littermate Atxn2l −/− mice (right), less lamina definition and many neurons were observed with nuclear condensation, which that is indicative of apoptosis; the diameter of brain cortex was also thinner.
In view of the postnatal survival of Atxn2l −/+ mice, a cohort of 12 WT animals and 12 heterozygous littermates was aged until 12 months and assessed every 3 months regarding survival, growth, and locomotor behavior. This preliminary survey found no increased mortality or weight loss, and detected no prominent deficits in spontaneous movements in the open-field paradigm or in muscle coordination under stress on an accelerating rotarod (Supplementary Materials Figure S2). Analyzing cell growth in four ATXN2L-null MEF lines versus four sex-matched WT littermates that were derived from dissected embryos, we noticed the presence many multinucleated giant cells. This Figure 5. H&E histology at E14 showed good separation of neuron layers in the cortex of Atxn2l +/+ brain (left); in littermate Atxn2l −/− mice (right), less lamina definition and many neurons were observed with nuclear condensation, which that is indicative of apoptosis; the diameter of brain cortex was also thinner.
In view of the postnatal survival of Atxn2l −/+ mice, a cohort of 12 WT animals and 12 heterozygous littermates was aged until 12 months and assessed every 3 months regarding survival, growth, and locomotor behavior. This preliminary survey found no increased mortality or weight loss, and detected no prominent deficits in spontaneous movements in the open-field paradigm or in muscle coordination under stress on an accelerating rotarod (Supplementary Materials Figure S2).
Analyzing cell growth in four ATXN2L-null MEF lines versus four sex-matched WT littermates that were derived from dissected embryos, we noticed the presence many multinucleated giant cells. This cell type occurs in primary cell cultures when nuclear division is not followed by cytokinesis [70]. In ATXN2L-null MEF lines, their frequency was consistently increased, with a >2-fold effect size across all lines ( Figure 6A,B). Cholesterol is necessary for cytokinesis, its deficiency induces polyploid cell formation, so it is noteworthy that cholesterol anomalies are well documented in patients and mice with ATXN2 mutations [31,[71][72][73]. In addition, the furrow progression during cytokinesis depends on receptor tyrosine kinase activation of SRC and on vesicle endocytosis, where ATXN2 is known to play a modulating role [9,10,[74][75][76]. , while the diagram on the right side represents the overall increase across all lines. Statistical testing was done by two-way ANOVA and t-test with Welch's correction, respectively, significance levels were illustrated by ** for p < 0.01, by T for 0.05 < p <0.1.
Then, we assessed whether the deprivation of nutrients, such as glucose and lipids, indeed alter the expression of Atxn2l, as was previously observed for its homolog Atxn2 [31]. The human neuronal cell line SH-SY5Y, first cultured in high-glucose DMEM medium with fetal calf serum (FCS) supplement, was switched to the starvation medium HBSS without FCS for a time course of 48 h. The sudden exposure to conditions of low glucose, no amino acids, no FCS (containing growth factors/apolipoproteins with lipids/transferrin with iron, etcetera) triggered a phasic induction that peaked at 8-24 h with a 3-fold Atxn2l mRNA increase, similar to the induction of Atxn2 (Figure 7A,B). Preliminary testing in MEF cells suggest that RNA levels of Atxn2 levels decreased slightly upon to stress by a toxic RNA-analogon, Poly(I:C), while Atxn2l failed to respond (Supplementary Materials  1-4, while the diagram on the right side represents the overall increase across all lines. Statistical testing was done by two-way ANOVA and t-test with Welch's correction, respectively, significance levels were illustrated by ** for p < 0.01, by T for 0.05 < p <0.1. Then, we assessed whether the deprivation of nutrients, such as glucose and lipids, indeed alter the expression of Atxn2l, as was previously observed for its homolog Atxn2 [31]. The human neuronal cell line SH-SY5Y, first cultured in high-glucose DMEM medium with fetal calf serum (FCS) supplement, was switched to the starvation medium HBSS without FCS for a time course of 48 h. The sudden exposure to conditions of low glucose, no amino acids, no FCS (containing growth factors/apolipoproteins with lipids/transferrin with iron, etcetera) triggered a phasic induction that peaked at 8-24 h with a 3-fold Atxn2l mRNA increase, similar to the induction of Atxn2 ( Figure 7A,B). Preliminary testing in MEF cells suggest that RNA levels of Atxn2 levels decreased slightly upon to stress by a toxic RNA-analogon, Poly(I:C), while Atxn2l failed to respond (Supplementary Materials Figure S3E-H). These MEF analyses provided no evidence that the depletion of ATXN2 triggers compensatory dysregulation of its homolog Ataxin-2-like at mRNA or protein level, or that conversely the depletion of ATXN2L triggers expression or abundance changes of Ataxin-2 (Supplementary Materials Figure S3E-J).
Further analysis of individual nutrients demonstrated that the transcriptional induction of ATXN2L in HBSS starvation medium could be diminished by the supplementation with FCS (which contains components such as insulin, transferrin, and apolipoproteins), and by low level cholesterol administration, two effects that were similarly observed for ATXN2. This was in contrast to supplementation with excess glucose, amino acids, or excess cholesterol that were unable to downregulate ATXN2L or further enhanced its levels ( Figure 7C,E). The increased transcription of ATXN2 upon glucose and amino acid addition was significant, as well as the induction by cholesterol excess (Figure 7D,F).
Further analyses about components of FCS that could be responsible for the rescue effect in the induction of ATXN2L are shown in Supplementary Materials Figure S3. Different cell culture media were used as basal media (DMEM having the highest abundance of nutrients, MEM containing fewer nutrients such as vitamins or amino acids, and HBSS containing no nutrients apart from low glucose amounts). The supplements included SPITE (a mixture of sodium selenite, sodium pyruvate, bovine insulin, human transferrin, and ethanolamine), ITS (a mixture of bovine insulin, human transferrin, and sodium selenite), and SITE+3 (containing analogous components as SPITE, but 5 mg/mL BSA, linoleic acid and oleic acid as a lipid energy source instead of pyruvate). However, none of these supplements was sufficient for a significant rescue of the starvation-triggered ATXN2L induction, to a degree that approached FCS. The comparison of SPITE versus SITE+3 supplementation suggested that fatty acid administration is more helpful than pyruvate for the reduction of ATXN2L and ATXN2 mRNA levels (Supplementary Materials Figure S3C,D). Overall, these in vitro observations confirm the notion derived from limb tissues in vivo and from MEF cells, that either the promoter activity of the Ataxin-2-like gene is upregulated or the Atxn2l/Atxn2 mRNA decay is downregulated during periods of growth deficits and then this regulation can be reversed upon availability of specific nutrients. In this context, it is noteworthy that the RegRNA2.0 webtool predicts both transcripts to contain AU-rich elements in their 3 -untranslated region, which are known to modulate the stability of selected mRNA according to growth needs. It is also relevant to mention that the promoters/enhancers of human Ataxin-2/Ataxin-2-like share 476 transcription factor binding sites, according to predictions found at the GeneCards website, including stress-response regulator families that are strongly implicated in neurodegeneration, such as ATF1/2/3/4/7, E2F1/5/6/8/ELF1/3/4/ETS1/ETV1/4/5/6, HNRNP-H1/K/L/LL/UL1, or IRF1/2/3/4.
A dysregulation of ATXN2L levels upon depletion of its ortholog ATXN2 was not observed in MEF cells, nor a converse dysregulation of ATXN2 upon depletion of ATXN2L, neither at protein nor at mRNA level (Supplementary Materials Figure S3E-J). As mentioned in the Introduction however, unpublished global proteome mass spectrometry data in our lab documented a significant increase of ATXN2L in the brain of Atxn2-CAG100-knockin mice. This finding may therefore be specific to the CAG-repeat expansion mutation and be caused by the polyQ aggregation process in stress granules, where ATXN2 and ATXN2L have common protein interactors such as DDX6 and may co-precipitate.

Discussion
ATXN2L was described as direct RNA-binding protein with sequence homology to nuclear spliceosome factors [37,77]. Its alternatively spliced isoforms may interact with tyrosine kinase receptor signaling at the plasma membrane [36]. The constitutive isoform of ATXN2L was mainly characterized for its role as surveillance factor during transient periods of cell damage, which promotes the formation of stress granules in the cytosol [34]. Now, our ATXN2L-deficient mouse data demonstrate that homozygous loss of its RNA-interaction domains is incompatible with embryonic development from early stages. Their absence delays the cytokinesis stage of cell division and impairs the layer formation and survival of brain neurons.
Our studies of the expression regulation of Atxn2l and Atxn2 by diverse nutrients indicate a strong similarity in their role, with a transcriptional induction being triggered by nutrient deprivation as well as by an excess of cholesterol and glucose (both of which have a role for plasma membrane flexibility/stickiness), while the transcription was reduced upon the supplementation of FCS. Given that the depletion of the ATXN2L ortholog in flies is neuroprotective, there may be therapeutic value in the knowledge that its expression can be downregulated in mammalian cells by the abundance of specific nutrients. The observations here may be relevant for the treatment of ALS, where hypercaloric feeding was shown to be beneficial, and evaluations on the relative contribution from carbohydrates versus lipids are ongoing [93,94]. An experimental artifact was observed upon nutrient deprivation and upon Poly(I:C) stress in vitro: Although the induction strength of Atxn2 and Atxn2l showed limited variance during repeated experiments by a student with a SH-SY5Y aliquot and a specific FCS batch, it could however show quite different fold-changes months later in different hands using a separate SY-SY5Y stock with another FCS batch. It remained unclear to what degree passage number, clonal mutation drift or handling variance played a role here (compare Figure 7C vs. E, and Figure S2 F vs. H).
Publicly available expression data on Atxn2l and Atxn2 by RNA sequencing and further processing at the Broad Institute within the Genotype-Tissue-Expression project database, online available at: https://gtexportal.org/home/ [95] with normalization to transcripts per million reads (TPM) confirm that Atxn2 and Atxn2l are both preferentially expressed in cerebellar hemispheres and total cerebellum (median TPM value = 30/33 vs. 184/229, respectively), in comparison with other regions of the nervous system. Interestingly, our observations on sex differences in the phenotype of null mutants are also reflected by these public expression data: they document that Atxn2 expression is stronger in most female organs (36/31/30/29/28/24 TPM in ovary/uterus/endocervix/FallopianTube/ectocervix/vagina, respectively) than in cerebellum. In contrast, Atxn2l shows the highest expression in testis tissue (TPM 296), as opposed to the average expression in female organs (e.g., ovary TPM 124). Overall, Atxn2l in cerebellum shows a 6-fold higher expression than Atxn2, and the preferential expression of Atxn2l in male testis contrasts with the pronounced expression of Atxn2 in several female sexual organs. Given that the biosynthesis of all sex hormones starts from cholesterol, the effects of ATXN2L and ATXN2 on the cholesterol homeostasis may account for the sex difference of Atxn2l-null phenotypes, the increased formation of giant multinucleated MEF cells, and the expression regulation by cholesterol.
It is important to note that Atxn2-mutations were already observed to affect cholesterol homeostasis and trigger gender-dependent or hormonal effects [31,73,96,97]. Giant multinucleated cells can be due to increased mTOR activity [98][99][100]. Given that ATXN2 was reported to inhibit mTOR in yeast, nematodes and mice; also the absence of ATXN2L might lead to enhanced mTOR signaling and, in this way, increase the appearance of multinucleated giant cells in MEF cultures. These phenotypes may be downstream effects of the reported protein interaction between ATXN2L and NHERF2 as a known modulator of cholesterol metabolism and Src-dependent phosphorylation signals for growth. While a gene duplication event in mammals led to the coexistence of two homologous genes, in plants a coexistence of four homologous genes from the ataxin-2 family raises the possibility that mTOR-dependent growth can be regulated by each of them in response to different stressors or to different nutrients, via an RNA-binding mechanism.
Clearly, the regulation of Atxn2l and Atxn2 mRNA levels by nutrients is almost parallel in our cell culture analyses, so it seems paradoxical that our null mutant of ATXN2L has a growth deficit phenotype, while the null mutant of ATXN2 was found by two independent teams to show a converse nutrient excess phenotype [31,32]. An important difference in the recombinant design of these mutants should be noted here. The KO of Atxn2 was targeting exon 1 and removed all of the protein domains, while the ablation in Atxn2l targeted exons 5-8, with a resulting shift of reading frame downstream. In the mutant ATXN2L, the N-terminal Pro-rich domain across residues 4-61 and the MPL interaction domain across residues 96-119 would still be synthesized as a fragment, the available commercial antibodies cannot detect this N-terminal region. In view of the increase of Atxn2l non-targeted exons in the limb tissue of homozygous mutants ( Figure 2B), this hypothetical fragment could exist in overdose. The excess amounts of such an N-terminal fragment would interact with plasma membrane receptors and with the endocytosis machinery. In analogy to the published observations for ATXN2 overexpression, it is expected to impair trophic uptake mechanisms [9]. At the same time, the absent middle and C-terminal part of ATXN2L would fail to interact with RNAs in the appropriate manner. Overall, our mouse mutant might represent not only a loss of the C-terminal functions of ATXN2L; it is conceivable that it also involves a toxic gain of N-terminal functions of ATXN2L. Perhaps this speculation provides a clue why our mouse mutant has this unexpected phenotype, which is converse to Atxn2-KO mice.
In the future, antibodies against the N-terminal portion of ATXN2L should be generated to obtain evidence regarding this doubt, and a new ATXN2L mutant that targets exons 1-4 would also help to clarify this issue. It will be crucial to generate conditional mutants where the postnatal role of ATXN2L for tissues such as the aging brain can be studied selectively. Moreover, it is interesting to note that commercial antibodies such as HPA041506/PA5-59601 in U2OS cells showed only the band of predicted size, and our immunoblots in brain tissue also detected this putative full-length band, while we observed two ATXN2L bands with three different antibodies in MEF cells, and also the A301-369A and 24822-1-AP antibodies detected a smaller band in HeLa/HEK293T and Jurkat cells, according to manufacturer datasheets. Thus, further studies will be necessary to understand the usage of the two alternative start codons and of the alternative splice variants in various tissues/cell populations and after different stressors. Clearly, our data indicated that the depletion of ATXN2 or ATXN2L in MEF cells fail to trigger a compensatory upregulation of the other homolog, so the unpublished accumulation of ATXN2L in Atxn2-CAG100-KIN brain appears to be an expansion-driven effect and ATXN2L immunoreactivity should be assessed within the Q100-ATXN2 aggregates. Since TDP-43 is essential for embryonic development in mouse and its aggregation drives the decisive toxicity for motor neuron degeneration, it is conceivable that ATXN2L-a factor that is also essential for embryonic development-could co-aggregate with ATXN2 and drive toxicity in cerebellar Purkinje neurons where it has selective abundance.

Genotyping
DNA was extracted from adult ear punches or embryo tail biopsies for genotyping by incubating at 95 • C in 100 µL of 25 mM NaOH/0.2 mM EDTA for 30 min before neutralizing with 100 µL 40 mM Tris-HCl, pH 5.0. For the subsequent genotyping reactions, 1 µL of lysed tissue sample was used per reaction. The Atxn2l alleles were amplified by PCR with the primers Atxn2l-F (GAGAGTGTGTGT TGGGTGAC), Atxn2l-F2 (GAAGCACAGTGCTGTTCAGC) and Atxn2l-R (GCTCTACTGTTGCT GCAAATTAC), the PCR products 217 bp and 309 bp corresponding to Atxn2l wt and Atxn2l mutant respectively were separated in 1.5% agarose gel.
For the determination of the sex, primers Rmb31F (CACCTTAAGAAGCCAATACA) and Rbm31R (GGCTTGTCCTGAAAACATTTGG) were used to yield a 269 bp product from X chromosome and a 353 bp product from the Y chromosome [102].

Behavioral Tests
Behavioral tests were done as described in [103]. Twelve Atxn2l +/+ and twelve Atxn2l +/− male mice were aged and assessed at ages between 3 months and 12 months.

RNA Extraction and Expression Analysis
RNA extraction from limbs and MEF was performed with TRIzol Reagent (Sigma Aldrich) according to the user manual. RNA from cell culture experiments with SH-SY5Y cells was performed with Extractme Total RNA Kit (7Bioscience EM09.1-250). Synthesis of cDNA from 1 µg of total RNA template was performed with the SuperScript IV VILO kit (ThermoFisher) according to the manufacturer's instructions. cDNA from 25 ng total RNA was used for each RT-PCR reaction with 0.5 µL TaqMan ® Assay, 5 µL FastStart Universal Probe Master 2x (Rox) Mix and ddH 2 O up to 10 µL of total volume. The mouse-specific TaqMan ® Assays utilized for this study: Atxn2l (Mm00805548_m1, Mm00805539_m1, Mm01165459_m1, Mm01276208_m1), Atxn2 (Mm01199894_m1), and Tbp (Mm00446973_m1).

Sections and Staining
All tissue specimens were fixed in 4% paraformaldehyde (PFA) for at least 48 h. All specimens were embedded in paraffin and aligned as displayed on the respective images (see Figures 4 and 5 and Figure S1). The 4-µm thick sections were made using a microtome. Slides were stained with hematoxylin and eosin according to standard protocols. Images were taken using an Olympus BX50 microscope and an Olympus DP72 camera.

Microscopy and Cell Counting of MEF
175.000 MEF cells were seeded in 6-well plates and grown overnight (n = 4 Atxn2l +/+ vs. 4 Atxn2l −/− lines in 3 technical replicates each). For each well, 5 random areas were microscopically visualized (Leica) and all giant cells were counted. Means for each cell line was calculated and statistically evaluated. Images were taken with a Nikon microscope after changing the cell culture medium to PBS.

Cell Culture Experiments
SH-SY5Y cells were purchased from ECACC, and cultured as described [6] Figure S3, cells were incubated with the indicated basal media (DMEM, MEF, or HBSS), supplemented with SPITE medium supplement 1x (Sigma S5666), ITS liquid media supplement 1x (Sigma I3146), or SITE+3 liquid media supplement 1x (Sigma S5295). After 24 h, cells were collected and RNA was extracted with EXTRACTME (Blirt, Gdańsk, Poland). Synthesis of cDNA and RT-qPCR was performed as previously described with human-specific TaqMan ® Assay ATXN2L (Hs00944485_g1), ATXN2 (Hs00268077_m1) and HPRT1 (Hs99999909_m1). The numbers of cell lines used is stated in respective figure legends.

Statistical Analysis
All statistical tests were performed as unpaired Student's t-test with Welch's correction or one-way or two-way ANOVA using GraphPad Prism software version 7 after establishing that each population was normally distributed (one-sided Kolmogorov-Smirnov test). Figures display mean values and standard error of the mean (s.e.m.). Values of p < 0.05 were considered significant and marked with asterisks p < 0.05 *, p < 0.01 **, p < 0.001 ***, p < 0.0001 ****.

Conclusions
Overall, this first analysis of ATXN2L-depletion in mouse demonstrates its essential role for embryonic development, due to growth delays with prominent vulnerability of brain lamination, especially in females. A preferential affection of female embryos and an increased occurrence of multinucleated giant cells in MEF culture were observed, which may reflect a role of ATXN2L in cholesterol and hormone homeostasis. This mouse mutant will provide an important tool to dissect the role of ATXN2L domains, such as the impact of LSM and PAM2 motifs on RNA surveillance, which were ablated here, versus the effects of PRD and MPL interaction motifs on trophic endocytosis, during stress periods and aging. Since the data show ATXN2L to be more important than ATXN2 for cell growth, it will be highly interesting to study its role in neurodegenerative disorders due to RNA surveillance problems, such as ALS, where the depletion of mammalian ATXN2 and of the yeast/fly ortholog also of ATXN2L was already shown to be neuroprotective.

Acknowledgments:
We are grateful to Gabriele Köpf, Aleksandar Arsovic, and Tatjana Starzetz for technical assistance, and to the staff at the animal facility ZFE at Frankfurt university school of medicine. The assistance of Dr. Kader Thiam and his team at Genoway (Lyon, France) in the design and evaluation of targeted mutations within Atxn2l (Figure 1) is highly appreciated.

Conflicts of Interest:
The authors declare no conflict of interest. G.A. advises RochePharma and TakedaPharma regarding ATXN2 research, receiving honoraria from them. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.