CAPRIN1P512L causes aberrant protein aggregation and associates with early-onset ataxia

CAPRIN1 is a ubiquitously expressed protein, abundant in the brain, where it regulates the transport and translation of mRNAs of genes involved in synaptic plasticity. Here we describe two unrelated children, who developed early-onset ataxia, dysarthria, cognitive decline and muscle weakness. Trio exome sequencing unraveled the identical de novo c.1535C > T (p.Pro512Leu) missense variant in CAPRIN1, affecting a highly conserved residue. In silico analyses predict an increased aggregation propensity of the mutated protein. Indeed, overexpressed CAPRIN1P512L forms insoluble ubiquitinated aggregates, sequestrating proteins associated with neurodegenerative disorders (ATXN2, GEMIN5, SNRNP200 and SNCA). Moreover, the CAPRIN1P512L mutation in isogenic iPSC-derived cortical neurons causes reduced neuronal activity and altered stress granule dynamics. Furthermore, nano-differential scanning fluorimetry reveals that CAPRIN1P512L aggregation is strongly enhanced by RNA in vitro. These findings associate the gain-of-function Pro512Leu mutation to early-onset ataxia and neurodegeneration, unveiling a critical residue of CAPRIN1 and a key role of RNA–protein interactions. Supplementary Information The online version contains supplementary material available at 10.1007/s00018-022-04544-3.


Introduction
Cell Cycle-Associated Protein 1 (CAPRIN1 [MIM: 601178]) is a ubiquitously expressed protein, whose levels are high in tissues characterized by an elevated cell turnover, but is also abundant in the brain [1][2][3][4]. There, it plays a crucial role as an RNA-binding protein (RBP), which regulates the transport and translation of mRNAs of synaptic proteins [2,5]. CAPRIN1 binds to mRNAs via C-terminal RGG motifs and contains a prion-like domain (PrLD) [6,7]. PrLDs are characterized by the lack of a defined three-dimensional structure and a low complexity sequence composition [8,9]. They interact dynamically with other proteins and RNAs, and these interactions can trigger phase transitions as well as protein aggregation [10]. Indeed, CAPRIN1 is a component of stress granules (SGs), cytoplasmic assemblies of RBPs and stalled mRNAs that form under stress conditions [6,11]. The N-terminal part of CAPRIN1 harbors a dimerization domain and the binding sites for the RBPs G3BP1 and FMR1 [6,12,13].
Here we describe a novel early-onset ataxia and neurodegenerative disorder caused by a single recurrent missense variant in CAPRIN1. We propose a gain-of-function mechanism mediated by increased protein aggregation propensity and highlight the crucial role of RNA-protein interactions in its pathophysiology.

Genetic studies
Genetic studies were performed as part of clinical and/or research investigations dependent on clinical presentation and family history. DNA was extracted from blood. DNA samples from A-I.1, A-I.2, A-II.3, B-I.1, B-I.2 and B-II.2 were prepared for whole exome sequencing (WES) and analyzed as described in detail in Note S1. CAPRIN1 exon 14 genomic region was amplified by PCR following the manufacturer's protocol (Multiplex PCR Kit-QIAGEN) with specific primers (CAPRIN1-E14, Table S1).

In silico CAPRIN1 P512L modeling
Human CAPRIN1 (Q14444-1) and CAPRIN1 P512L FASTA protein sequences were pasted in PLAAC [18]. The Relative weighting of background probabilities (α) was set to 0 and Homo Sapiens was selected as organism background frequency.
The CamSol intrinsic solubility profile was obtained from the CamSol web server using the same FASTA sequences [19].

Solubility analysis
Sequential extraction of proteins from the different soluble fractions was performed following a published protocol [21]. Briefly, 24 h post transfection, cells were washed twice with PBS, lysed in cold RIPA buffer (SIGMA), and sonicated. Cell lysates were centrifuged at 100,000 g for 30 min at 4 °C to generate RIPA-soluble samples. Pellets were washed, sonicated and centrifuged twice with PBS. RIPA-insoluble pellets were then extracted with urea buffer (8 M urea, 4% CHAPS, 30 mM Tris, pH 8.5), sonicated, and centrifuged at 100,000 g for 30 min at 22 °C. Protease inhibitors were added to all buffers before use. Protein concentration was determined with the Bradford method.

Immunoblotting
Protein lysates (10 μg) in Laemmli Buffer were heated at 95 °C for 5 min, then separated on 12% polyacrylamide gels and transferred onto nitrocellulose membranes (Merck Millipore). Membranes were blocked for 1 h in 5% BSA in TBS-T, incubated overnight with primary antibodies (Table S2) in 2.5% BSA in TBS-T at 4 °C, washed three times in TBS-T and incubated with HRP-conjugated secondary antibodies (Table S2) for 1 h at room temperature. Proteins were visualized using the Immobilon Western chemiluminescent HRP substrate (Merck Millipore). Quantification was performed with ImageLab (Bio-Rad).

Immunofluorescence
Coverslips were washed in PBS and fixed in 4% PFA for 10 min. They were then washed three times in PBS, permeabilized in PBS-T for 10 min and blocked for 1 h at room temperature in 5% BSA in PBS-T. Coverslips were then incubated overnight with primary antibodies (Table S2) in 5% BSA in PBS-T at 4 °C, washed three times in PBS-T, incubated with conjugated secondary antibodies (Table S2) for 1 h at room temperature, washed three times in PBS-T, rinsed in water and mounted with ProLong Gold antifade reagent with DAPI (Invitrogen) on Polylysine slides (Ther-moFisher Scientific). Images were acquired with a Zeiss AxioImager M2 microscope equipped with ApoTome2 system and processed using ZEN software (Zeiss). Pearson's colocalization coefficient was obtained using the Coloc 2 tool in ImageJ. Stress granules were quantified with a selfcompiled macro in ImageJ, which uses the Weka Trainable Segmentation plugin on the G3BP1 signal. A cell was considered positive if ≥ 3 signals were detected.
Off-t argets for bot h CAPRIN1 W T/P512L and CAPRIN1 P512L/P512L lines were excluded by Sanger sequencing in the first five gRNA off-targets sites predicted using CRISPOR. Expression of pluripotency markers was confirmed by immunofluorescence.

Microelectrode array recordings
On D26, 7.5 × 10 4 cells / well were seeded in a 24-well epoxy plate with a microelectrode array (MEA) (Multi Channel Systems). Cells were recorded with a Multiwell-MEA-System (Multi Channel Systems) for 3 min at 37 °C between D29 and D57 using the Multiwell-Screen software (Multi Channel Systems) and analyzed with the Multiwell-Analyser software (Multi Channel Systems). When no activity was recorded, the parameter was set to 0. The following parameters were used: 2nd order low-pass filter frequency: 3500 Hz; 2nd order high-pass filter frequency: 100 Hz; rising/falling edge of the automatic threshold estimation: 5.5/− 5.5 SD; minimum spike count in burst: 3; minimum channels participating in a network burst: 5; minimum simultaneous channels for a network burst; minimum spikes per minute: 5; minimum amplitude: 10 µV.

SG dynamics study
On D36, iPSC-derived neurons plated onto coverslips were treated for 1 h with 0.5 mM Sodium Arsenite (SA, NaAsO 2 , Sigma-Aldrich), then washed in PBS and incubated in N2/ B27 medium for 0-240 min to study SG resolution. Coverslips were then processed for Immunofluorescence.

Recombinant protein expression and purification
mGFP-CAPRIN1 WT and P512L, as well as CAPRIN1 WT were purified from Sf9 insect cells using a baculovirus expression system [26,27]. Cells expressing recombinant TwinstrepII-MBP-mGFP-CAPRIN1-6xHis were lysed in 50 mM Tris-HCl pH 7.5, 300 mM KCl, 150 mM Arginine-HCl, 1 mM DTT and 1 × EDTA-free protease inhibitor cocktail (Roche Applied Sciences) using a LM10 Microfluidizer (Microfluidics) at 5000 psi. The lysate was cleared by centrifugation at a maximum speed for 1 h at 4 °C. The supernatant was applied to a 5 ml Strep-Tactin®XT 4Flow® column (IBA Lifesciences GmbH) using an ÄKTA pure 25 (GE Healthcare). The column was washed with 10 column volumes (CV) of 50 mM Tris-HCl pH 7.5, 300 mM KCl, 150 mM Arginine-HCl and 1 mM DTT and the protein was eluted with 3 CV of 50 mM Tris-HCl pH 7.5, 300 mM KCl, 1 mM DTT and 50 mM biotin. The eluted protein was applied to a 5 mL HiTrap Q HP column (GE Healthcare). The column was washed with 20 CV of 50 mM Tris-HCl pH 7.5, 50 mM KCl and 1 mM DTT. Elution was achieved with a linear gradient of 20 CV of 50 mM Tris-HCl pH 7.5, 1000 mM KCl and 1 mM DTT. The elution fractions containing MBP-mGFP-CAPRIN1-6xHis were pooled and incubated 6 h at RT with 6xHis-Prescission protease (1:300 w/w) to cleave off the MBP and 6xHis tags. The sample was then applied to a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) equilibrated in 50 mM Tris-HCl pH 7.5, 300 mM KCl and 1 mM DTT. mGFP-CAPRIN1 fractions were pooled, concentrated to ~ 100 µM with Amicon Ultra centrifugal 30 KDa MWCO filters (Merck Millipore), flashfrozen with liquid nitrogen and stored at − 80 °C. Purified proteins were quality controlled by SDS-PAGE and Coomassie staining. The homogeneity of GFP-tagged proteins was determined by imaging the gels using the Amersham typhoon scanner.

Nano-differential scanning fluorimetry (nanoDSF)
mGFP-CAPRIN1 (WT or P512L) was diluted to a final protein concentration of 5 µM in 50 mM Tris/KOH pH 7.5 and 75 mM KCl. Unfolding transitions were recorded with a Prometheus Panta (Nanotemper) in high sensitivity capillaries (Nanotemper) at 0.3 °C min −1 . Data analysis and plotting were with the R/RStudio software package.

Fluorescence anisotropy measurements
100 nM of ATTO590-labelled RNA were mixed with increasing concentrations of the mGFP-CAPRIN1 or mGFP-CAPRIN1 P512L , in 25 mM HEPES-KOH, pH 7.5, 75 mM KCl and incubated for 10 min at 25 °C. Fluorescence anisotropy was measured with a Tecan Spark plate reader in 384-well plates (Greiner bio-one). Fluorescent excitation was at 598 nm/20 slit width and emission was recorded at 664 nm/30 slit width. Fluorescence anisotropy was calculated using the manufacturer's software. For RNA competition assay, x100 nM of ATTO590-labelled RNA and 3 μM of mGFP-CAPRIN1 was preincubated for 10 min at 25 °C and then competed for with an increasing concentration of unlabeled long homopolymeric polyA RNA. The complex then competed with an increasing concentration of unlabeled long homopolymeric polyA RNA. Fluorescence anisotropy data were fitted to Eq. (1), where r is the determined fluorescence anisotropy at a given protein concentration (c), L t is the ligand concentration, K D is the apparent dissociation constant and r free and r bound the fluorescence anisotropy of the free and protein-complexed ATTO590-labelled RNA, respectively. Data analysis and plotting were carried out with R/RStudio software package.

FCS measurements
FCS measurements were carried out using a LSM780 (Zeiss) confocal microscope. The system and measurements were calibrated using ATTO488 (D = 4.0 ± 0.1 × 10 -6 cm 2 s −1 at 25 °C) [29]. Diffusion times were recorded using a 488 nm argon laser for excitation and a 495-555 nm bandpass filter for emission, at 3.5 μW laser power, AOTF dampening factor of 10% with 15 rep for 10 s. The diffusion coefficients were calculated using Eq. (2). (1) The radius of the proteins was then calculated using the Stokes-Einstein Eq. (3).
All FCS measurements were carried out with 50 nM of mGFP-CAPRIN1 WT or P512L or mGFP in 25 mM Tris-KOH, pH 7.5, 75 mM KCl and 1 mM DTT at 22 °C. For oligomerization experiments 50 nM mGFP-CAPRIN1 WT were mixed with either 100 or 2000 nM unlabelled CAPRIN1. ATTO488 dye and mGFP were used as standards to calculate the diffusion volume and the number of GFP molecules in the diffusion volume respectively. Data analysis and plotting were carried out with R/RStudio software package.

Statistical analysis
All statistical analyses were performed using the software Prism 9 (GraphPad). Unpaired t test were used for the comparison of two groups (solubility analysis, aggregates number/size, circularity, colocalization, biochemical properties), while one-way ANOVA was used for the comparison of three groups. Chi-square test was applied to compare frequencies of aggregates in transfected SH-SY5Y cells.

Two independent ataxic individuals carry the same de novo Pro512Leu mutation in CAPRIN1
Affected individual II.3 of family A was referred for genetic counseling at the age of 10 years because of the development of gait abnormalities and predominantly proximal muscle weakness (positive Gower's sign). She was born to non-consanguineous parents of Turkish descent and had two older healthy siblings (Fig. 1a). Her symptoms worsened over the following years, with increased muscle weakness and the development of ataxia with light tremor and dysdiadochokinesis. The progressive trunk instability and scoliosis lead her to be first confined to a wheelchair and later to a bed. The motor deficits were accompanied by bulbar symptoms (dysphagia and dysarthria), deficits in sustained attention, social withdrawal and cognitive decline. She attended a mainstream school but later changed to a special needs school. Formal testing revealed an intelligence quotient of 64. Standard laboratory and metabolic tests were negative. Electromyography and nerve conduction studies identified a sensorimotor axonal neuropathy. Muscle biopsy revealed neurogenic fiber atrophy and suralis nerve biopsy uncovered a chronic axonal neuropathy with loss of small-and bigcaliber nerve fibers. Magnetic resonance imaging (MRI) at 16 years of age displayed cerebral and cerebellar atrophy (Fig. 1b). Deletions of SMN1 were excluded and no causative variant was found using an in-house NGS gene panel covering 62 genes associated with lower motor neuron disorders [30].
Affected individual II.2 of family B was born to nonconsanguineous parents of Italian descent and had a healthy sister (Fig. 1a). He articulated his first words with slight phonetic problems and presented with dysarthria at the age of 4 years. At the age of 7 years, he developed slowly progressive ataxia and learning difficulties (IQ: 77). By the age of 11 years, his trunk stability worsened and standing up became more difficult. At the age of 12 years, MRI showed global cerebellar atrophy (Fig. 1b). At the age of 13 years, he showed increased muscle fatigue and muscle hypotrophy, with absent deep tendon reflexes in all four limbs. He also became increasingly anxious but improved with psychotherapy. Standard laboratory and metabolic tests were negative. Somatosensory evoked potentials (SSEPs) were reduced in the lower limbs.
Note added in proof: Just after the acceptance of our manuscript, we were notified by GeneMatcher of a patient identified at NINDS, NIH (female, 14 yrs old) with exactly the same de-novo variant and an identical phenotype (cerebellar Pherograms confirm the heterozygous de novo c.1535C > T variant in CAPRIN1 in the affected individuals and its absence in the unaffected parents. d CAPRIN1 Pro512 residue is highly conserved. Protein sequence alignment of CAPRIN1 orthologues displays high conservation in the region of the P512L mutation (red). e HMM logo of CAPRIN1 protein sequence alignment confirms the high conservation of the Pro512 residue. f Schematic representation of CAPRIN1. Highlighted are: homology region 1 and 2 (HR-1, residues 56-248; HR-2, residues 352-685) with CAPRIN2; CAPRIN1 dimerization region (residues 132-251, //) [12]; FMR1 binding region (residues 231-245, blue) [13]; G3BP1 binding region (residues 352-380, green) [6]; RGG motifs (RGG) [1]. g CAPRIN1 C-terminal region is a PrLD. PLAAC application predicts a PrLD between residues 537 -709. The position of the CAPRIN1 P512L mutation is highlighted in red. h PLAAC in silico modeling of the P512L mutation. The CAPRIN1 P512L mutation lowers the -4*PAPA score (green solid line), crossing the cutoff (green dashed line) and indicating an increased aggregation propensity. i CamSol in silico modeling of the CAPRIN1 P512L mutation. The CAPRIN1 P512L mutation lowers CAPRIN1 solubility

CAPRIN1 P512L forms insoluble aggregates
To investigate the potentially increased aggregation propensity of CAPRIN1 P512L , we overexpressed V5-tagged CAPRIN1 and CAPRIN1 P512L in HEK293T cells and sequentially extracted proteins from a more soluble (RIPA) and less soluble (urea) fraction. CAPRIN1-V5 was mainly eluted in the RIPA-soluble fraction, while CAPRIN1 P512L -V5 exhibited reduced solubility and was recovered in the ureasoluble fraction (Fig. 2a), a behavior found in other mutant PrLD-containing proteins related with degenerative disorders [21].

CAPRIN1 P512L aggregates are positive for typical NDs markers
Since protein misfolding and impairment of the protein quality control (PQC) are widely recognized pathomechanism of NDs [45], we investigated protein homeostasis markers, such as ubiquitin and p62. Under physiological conditions, the formation of aggregates is prevented by the activity of molecular chaperons of the PQC, which are also able to unfold misfolded proteins. When folding is not possible, misfolded proteins are ubiquitinated by E3 ubiquitin ligases and directed to proteasomal degradation via the ubiquitin-proteasome system (UPS) [46,47]. Indeed, bulky CAPRIN1 P512L aggregates were positive for ubiquitin (Fig. 2c). Moreover, since insoluble aggregates can inhibit the 26S proteasome and be targeted for lysosomal degradation by macroautophagy [48,49], we stained CAPRIN1 P512L aggregates for p62/SQSTM1 positivity and we could indeed detect a strong signal, as reported for other NDs (Fig. 2d) [50]. Taken together, these results suggest that CAPRIN1 P512L misfolds and becomes targeted for degradation.
Due to the progressive muscle atrophy of the affected individuals, we additionally investigated if the aggregates contained SNRNP200, another CAPRIN1 interacting partner that has been reported in the cortical and spinal motor neurons of ALS cases and indeed we could detect it ( Fig. 3d;  Table 1) [36]. Taken together, these results indicate that CAPRIN1 P512L inclusions are able to sequester multiple ND and ataxia-related proteins.

CAPRIN1 P512L iPSC-derived neurons show reduced neuronal activity
To study the effects of the P512L substitution in a human neuronal cell model, we generated the heterozygous CAPRIN1 WT/P512L and the homozygous CAPRIN1 P512L/P512L isogenic cell lines from the CAPRIN1 WT/WT HUVEC iPSC line using CRISPR/Cas9 genome editing (Fig. S2a). We then differentiated them in cortical neurons using the protocol from Schuster et al. 2020 (Fig. S2b and S2c). These iPSC-derived neurons do not show any significant change in CAPRIN1 levels at neuronal maturation (D36), nor any overt morphological alteration of the cell soma or the neurites (Fig. 4a).
In particular, both iPSCs and iPSC-derived neurons harboring the CAPRIN1 P512L mutation did not display any protein aggregates, even upon proteasomal inhibition (Fig.  S3a-b and S4a). Their absence, however, is reported for many other iPSC-derived neuronal cells lines that harbor mutations associated with protein aggregation in tissue sections from individuals suffering of ataxia, Parkinson's disease or Huntington's disease (HD) [56][57][58].
Since in several disease models electrophysiological changes in neurons precede neuronal loss [59], we recorded the spontaneous neuronal activity using a microelectrode array system. Interestingly, while CAPRIN1 WT/WT and CAPRIN1 WT/P512L neurons increased their firing rate upon maturation, CAPRIN1 P512L/P512L neurons showed a clearly reduced spike rate and almost no bursting throughout the whole recording period (Fig. 4b-e). On the other hand, after an initial overlap, also the activity of CAPRIN1 WT/P512L neurons progressively decreased (Fig. 4b-e).

CAPRIN1 P512L iPSC-derived neurons show impaired stress granules dynamics
Since CAPRIN1 represents one of the main components of SGs [6,11], and disease-linked mutations in TARDBP, FUS or C9ORF72 cause an increase of cells presenting SGs upon stress [60][61][62], we hypothesized that the CAPRIN1 P512L mutation could alter their dynamics. Therefore, we treated the iPSC-derived neurons with sodium arsenite (SA), a common SG inducer [6], and studied their resolution at different time points. Intriguingly, upon SA treatment, a higher fraction of CAPRIN1 WT/P512L neurons showed SGs in comparison to both CAPRIN1 WT/WT and CAPRIN1 P512L/P512L neurons ( Fig. 5a and b). Moreover, in CAPRIN1 WT/P512L neurons the resolution of the SGs occurred slower than in the other cell lines, resulting in the persistence of SG for a longer time after stress removal. Strikingly, this difference could not be observed in CAPRIN1 P512L/P512L neurons, where the SGs resolution tended to be even faster than in the CAPRIN1 WT/WT neurons, suggesting a more complex scenario where the CAPRIN1 properties and interactions might play a major role.

CAPRIN1 P512L adopts an extended conformation
To investigate whether the P512L mutation influences CAPRIN1 structure, we characterized recombinantly produced and purified mGFP-CAPRIN1 and mGFP-CAPRIN1 P512L (Fig. 6a). We used nano-differential scanning fluorimetry (nanoDSF) to monitor the tertiary structure and unfolding transitions of CAPRIN1. This revealed significant differences in the fluorescence ratio (F350/F330) at 20 °C and increased stability of mGFP-CAPRIN1 P512L in comparison to mGFP-CAPRIN1 (Fig. 6b,  Table 2). These data suggest that the mutation does not cause a substantial destabilization of the protein and that the two proteins have a similar tertiary structure. Dynamic light scattering (DLS) and fluorescence correlation spectroscopy (FCS) measurements showed that mGFP-CAPRIN1 P512L exhibits an increased hydrodynamic radius compared to that of mGFP-CAPRIN1 (Fig. 6c, Table 2). Taken together, the data suggest that CAPRIN1 P512L adopts an extended yet near-native conformation.

The P512L mutation does not impair CAPRIN1 dimerization
Given the differences in hydrodynamic radius between CAPRIN1 and CAPRIN1 P512L and the ability of CAPRIN1 to form dimers [12], we used FCS to test for changes in CAPRIN1 oligomerization. We measured the brightness of mGFP-CAPRIN1 and mGFP-CAPRIN1 P512L and compared it to the brightness of free GFP. Both proteins were shown to associate into dimers in solution even at concentrations as low as 50 nM (Fig. 6d). Consistent with the formation of CAPRIN1 dimers, the GFP brightness decreased when mGFP-CAPRIN1 was mixed with an excess of unlabeled CAPRIN1, demonstrating the formation of spectroscopic heterodimers (Fig. S5a). In accordance with the distance between the mutated residue and the annotated dimerization domain (residues 132-251) [12], our data demonstrate that the P512L mutation does not affect dimerization, but rather results in an expanded conformation of the protein.

CAPRIN1 P512L aggregation is enhanced by RNA
Since CAPRIN1 is an RBP, we examined whether this conformational change would alter its affinity for RNA. To this end, we incubated CAPRIN1 with ATTO590labelled single-stranded RNA (ssRNA). CAPRIN1 P512L showed reduced RNA affinity (K D CAPRIN1 : ~ 506 ± 223 nM; K D CAPRIN1−P512L : ~ 947 ± 239 nM; Fig. 6e). We then tested the reversibility of the CAPRIN1-RNA interaction by adding unlabeled long homopolymeric polyA RNA as a competitor. In accordance with the previous data, CAPRIN1 binds ssRNA ~ twofold tighter than CAPRIN1 P512L (K I CAPRIN1 : 60.5 ± 0.7 ng/µl; K I CAPRIN1−P512L : 33.5 ± 12 ng/µl; Fig. 6f). To further investigate CAPRIN1 P512L properties, we observed mGFP-CAPRIN1 and mGFP-CAPRIN1 P512L by fluorescence microscopy. While mGFP-CAPRIN1 displayed a diffuse signal, mGFP-CAPRIN1 P512L formed Page 15 of 20 526 small agglomerates, confirming the increased aggregation propensity seen in our cell models (Fig. 6g). Since CAPRIN1 is an RBP and recent studies demonstrated the pivotal role of RNA in the modulation of protein aggregation [63,64], we next incubated the purified proteins with RNA. Strikingly, while mGFP-CAPRIN1 remained soluble, mGFP-CAPRIN1 P512L formed large, microscopically visible aggregates (Fig. 6g). This effect was independent of the RNA type, and all RNA types tested caused aggregation of mGFP-CAPRIN1 P512L (Fig. S5b). Since the association of RBPs with nucleic acids is often driven and stabilized through electrostatic interactions, we increased the salinity after complex formation to distinguish weaker (reversible) from stronger (irreversible, indicative of aggregates) interactions. Increasing the salinity reduced the degree of aggregation only to some extent, and the addition of RNase A did not dissolve the aggregates (Fig. 6g). This suggests that CAPRIN1 P512L misfolding might be triggered by RNA, but that RNA is not necessary for aggregate persistence. Consistent with this, our FISH analysis in CAPRIN1 P512L transfected cells showed that the formed aggregates do not contain polyA RNA (Fig. S5c).
Taken together, these data indicate that the Pro512Leu mutation alters the dynamics of binding to RNA which might influence the aggregation propensity of CAPRIN1.

Discussion
To date, CAPRIN1 has been associated with two conditions: increased CAPRIN1 expression has been connected to certain cancers [65], while its reduction has been linked with autism spectrum disorders and speech delay [14][15][16][17]. In contrast, we identify the recurrent de novo CAPRIN1 P512L mutation in two independent individuals with early onset progressive ataxia and intellectual disability, which increases the protein propensity to aggregate and causes electrophysiological alterations in iPSC-derived neurons.
Strikingly, both affected individuals carry the identical de novo c.1535C > T variant, an event which is per se highly unlikely to occur by chance (Note S2). This variant is not reported in gnomAD, where CAPRIN1 constraint metrics indicate that the gene has a reduced tolerance for missense mutations (Z = 1.69; o/e: 0.76 (95% CI 0.69-0.84)) and missense SNVs in the ± 100 bp range from the variant have all very rare frequencies (< 0.0001).
The P512L substitution affects the highly conserved proline residue exchanging a secondary structure breaker for a non-polar, aliphatic leucine ( Fig. 1d and e) [65]. Although this amino acid substitution affects a residue close to CAPRIN1 PrLD (residues 537-709) and its surrounding region is enriched for residues found in prion-like domains, such as serine and glutamine, the proline-surrounding sequence is highly conserved. This suggests that this region is not disordered but adopts a specific fold that is most probably disrupted by the introduction of the leucine residue. In particular, the fact that the P512L substitution lead to an increase in the hydrodynamic radius of the protein, suggests that this proline is in the cis configuration in the wild-type protein and generates a kink in the polypeptide chain [66]. We hypothesize that this kink is absent in CAPRIN1 P512L , causing the observed extended conformation. Moreover, due to the distance of this residue from potential post-translational modifications sites, this substitution is unlikely to influence them [67].
Our data suggest that the leucine substitution renders the protein prone to misfolding and aggregation, which is accelerated by the presence of RNA. Therefore, it is highly likely that the P512L acts in a gain-of-function manner. Furthermore, the gain-of-function model is in accordance with the increasing evidence that, conversely, CAPRIN1 reduction (e.g. haploinsufficiency) causes a different phenotype, characterized by language impairment, attention-deficit/ hyperactivity disorder. deficit hyperactivity disorder (ADHD) and ASD [14][15][16][17]. This suggests an intriguing parallel between CAPRIN1 and FMR1: FMR1 loss-of-function mutations are linked to Fragile X syndrome, a neurodevelopmental disorder characterized by intellectual disability due to hypermethylation of long CGG repeats (> 200) [68], while FMR1 gain-of-function mutations are linked to FXTAS, a neurodegenerative disorder where shorter CGG expansions  lead to increase in FMR1 expression [39,69]. An important difference between the CAPRIN1 neurodegenerative disorder and FXTAS is the different disease onset: while the former Fig. 6 Dynamics of CAPRIN1 P512L and RNA. a Purified mGFP-CAPRIN1 and mGFP-CAPRIN1 P512L stained with Coomassie (CBB) on SDS-PAGE. b CAPRIN1 P512L adopts a more extended conformation than CAPRIN1. The fluorescence ratio (F350/F330) at 20 °C of the two proteins is shown (mGFP-CAPRIN1: 0.845 ± 0.006; mGFP-CAPRIN1 P512L : 0.864 ± 0.009; n = 9; unpaired t test: ***p < 0.001). c CAPRIN1 P512L adopts a more extended conformation than CAPRIN1. Hydrodynamic radii (rH) were calculated using FCS (mGFP-CAPRIN1 P512L : 4.8 ± 0.5 nm; mGFP-CAPRIN1 P512L : 5.8 ± 0.2 nm; n = 3; unpaired t test: p < 0.05). d CAPRIN1 P512L dimerization is not impaired. The GFP brightness comparison of 50 nM mGFP-CAPRIN1 and mGFP-CAPRIN1 P512L to mGFP indicate that both proteins form dimers. e CAPRIN1 P512L has a reduced RNA affinity. Fluorescence anisotropy measured the binding affinity of CAPRIN1 and CAPRIN1 P512L using ATTO590 ssRNA. : ~ 947 ± 238.6 nM; n = 3). f CAPRIN1 P512L has a reduced RNA affinity. An increasing amount of unlabeled polyA was added to ATTO590 ssRNA-bound mGFP-CAPRIN1 or mGFP-CAPRIN1 P512L and changes in anisotropy were measured (K I CAPRIN1 : 60.5 ± 0.7 ng/µl; K I CAPRIN1−P512L : 33.5 ± 12 ng/µl; n = 2). g CAPRIN1 P512L aggregation is enhanced by RNA incubation. Upon RNA addition, mGFP-CAPRIN1 P512L aggregates while mGFP-CAPRIN1 remains soluble. KCL or RNase A were added to check the reversibility of the interaction (Scale bar: 10 µm) ◂ occurs during childhood, the latter affects individuals older than 50 years of age [39]. This suggests that the protein quality control machinery is unable to control the aggregation properties of CAPRIN1 P512L and it could be related to the importance of the other proteins sequestered in the aggregates.
Intriguingly, we observed that CAPRIN1 P512L inclusions contain ATXN2 and GEMIN5. This is interesting because ATXN2 polyQ repeats cause autosomal dominant SCA2 (≥ 33 repeats) or increase ALS risk (31-32 repeats) through neuronal ATXN2 aggregation [10] and the affected individuals show both progressive ataxia and muscle weakness and atrophy. On the other side, biallelic mutations in GEMIN5 have recently been linked to early-onset neurodevelopmental delay and ataxia [38]. This suggests that sequestration of ATXN2 and GEMIN5 in CAPRIN1 P512L inclusions could at least in part be responsible for the observed gainof-function phenotypes. In fact, CAPRIN1 and many other stress granule proteins are embedded in a dense network of protein-protein interactions even before they assemble into stress granules [70][71][72]. The fact that mutant CAPRIN1 is in a near-native state, suggests that many of these interactions will remain intact during misfolding and aggregation. Accordingly, CAPRIN1 aggregation may inactivate many associated proteins, providing an explanation for the severity of the disease phenotype.
Given the importance of the cell-specific environment in investigating a phenotype [25], we prioritized a neuronal model over patient-derived cell lines. Therefore, we generated the heterozygous CAPRIN1 WT/P512L and the homozygous CAPRIN1 P512L/P512L isogenic iPSC lines using CRISPR/Cas9 technology and differentiated them in cortical neurons, since both affected individuals suffered also from intellectual disability and even showed cortical atrophy and to avoid the considerable limitations of iPSC-derived cerebellar neurons, such as low cellular yield, need of coculture with mice-derived progenitors or long differentiation duration of the available protocols [73]. At neuronal maturation (D36), we did not observe marked differences in differentiation efficiency (Fig. S2b and S2c), which would indicate a correct neuronal maturation and confirm the neurodegenerative nature of the disorder. We did not observe the formation of CAPRIN1 aggregates, even after proteasomal inhibition in both iPSCs and iPSC-derived neurons (Fig.  S3a, S3b and S4a). One possible justification is a caveat of the disease model per se: iPSC-derived neurons usually lack the maturity of postnatally differentiated neurons and this acquires particular relevance when investigating a neurodegenerative disorder, where neuronal ageing plays a crucial role in the development of a phenotype [58]. In fact, iPSC-derived neurons used to study diseases characterized by protein aggregation like Alzheimer's disease, Parkinson's disease or Huntington's disease mostly failed to detect amyloid beta, α-synuclein or HTT aggregation, respectively [56][57][58]. Based on the characterization of the iPSC-derived neurons conducted in the protocol's original paper [25], it is reasonable to expect that the iPSC-derived neurons used in this study are not mature enough. Indeed, studies have shown that bypassing the iPSC state through direct reprogramming of somatic cells maintained aging hallmarks and recapitulated some disease phenotypes that are not present in iPSC-derived neurons [74]. This alternative approach was not pursued because of reproducibility concerns. However, it is important to note that iPSC-derived neurons could exhibit disease-specific alterations even in the absence of overt protein aggregation [58]. Interestingly, the activity of the iPSC-derived neurons decreases with the number of mutated CAPRIN1 alleles: while the spike rate and burst number of CAPRIN1 WT/P512L neurons is initially comparable to the one of CAPRIN1 WT/WT neurons, that of CAPRIN1 P512L/P512L neurons is always low: this suggests that pathophysiological changes in the neurons precede the formation of aggregates. This is in agreement with many NDs characterized by protein aggregation, where alterations in several cellular processes are found in neurons without inclusions, even in animal models [59,75]. In CAPRIN1 WT/P512L neurons, the mutation causes an increase in the SG formation and a reduction in their resolution, in line with disease-linked mutations in other ND-related genes (TARDBP, FUS or C9ORF72) [60][61][62]. However, the homozygous mutation does not cause an increase in their formation, but even a slightly faster resolution. A possible explanation of this phenomenon is that the increased aggregation propensity of CAPRIN1 P512L might be counterbalanced by its reduced affinity for RNA. In homozygous CAPRIN1 P512L neurons, the protein's low affinity for RNA tends to hinder the formation of SGs, compensating its increased aggregation. By contrast, in heterozygous neurons, the aggregation-prone but low-RNA-binding CAPRIN1 P512L can still form dimers with the wild-type CAPRIN1, and its aggregation propensity might increase SGs formation. These data suggest differences in the assembly of SGs that could be due altered RNA binding affinity, but whether SGs promote the formation of pathological CAPRIN aggregates or whether CAPRIN1 aggregates independently of SGs remains to be investigated.
Indeed, the pivotal role of protein-RNA interactions in aggregation is becoming increasingly clear: while for some mutations of ALS-related genes, such as TARDBP, FUS, HNRNPA2B1 and HNRNPA1, the purified mutated proteins alone were often sufficient to enhance its aggregation propensity [20,[40][41][42], recent studies demonstrated the RNA ability to hinder or promote aggregation [63,64,76]. For example, RNA is able to antagonize TARDBP aggregation and disease-associated mutations would promote aberrant phase transitions of RNA-deficient TARDBP proteins [77]. Conversely, RNA increases the aggregation propensity of CAPRIN1 P512L and RNA removal does not revert the conformational change: this indicates that the misfolding might be irreversible. Interestingly, polyG RNA is able to trigger not only CAPRIN1 P512L aggregation but the one of CAPRIN1 too. A possible explanation for this phenomenon could be the intrinsic ability of polyG RNA to undergo phase separation due to the formation of G-quadruplex structures [78].
In conclusion, we identify with the P512L mutation a highly critical domain in CAPRIN1, the alteration of which associates with early-onset ataxia and intellectual disability, thereby associating another PrLD-containing protein to a novel neurodegenerative disorder. Moreover, we provide further evidence for the pivotal role of protein-RNA interactions in the assembly of aggregates.

Data availability
The exome data of family A are stored in the EGA database under the access numbers EGAN00001366922, EGAN00001366923, EGAN00001366924. Request should be addressed to the NeurOmics data sharing committee. The exome data for family B can be made available upon reasonable request.

Conflict of interest
The authors have no relevant financial or non-financial interests to disclose.

Ethical approval and informed consent
This study conformed to standards outlined in the Declaration of Helsinki and was approved by the Ethics Committee of the University of Cologne (Reference Number 13-022) and Ospedale Pediatrico Bambino Gesù (Reference Number 1702_OPBG_2018). Informed written consent for the collection of human material, for the participation in the study and for publication purposes was obtained from the respective subjects or their legal guardians following the regulations of the Ethics Committees of the University of Cologne and Ospedale Pediatrico Bambino Gesù.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.