Author Correction: De novo design of RNA-binding proteins with a prion-like domain related to ALS/FTD proteinopathies

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

degeneration 15 . Given these commonalities, cytoplasmic aggregation of RNA-binding proteins could constitute a critical process in the development of the ALS/FTD pathological cascade linked to the RNA quality control system.
An emerging question involving these ALS/FTD-related RNA-binding proteins is whether their associated cytotoxicity is dependent upon specific physiological functions related to target RNAs or their robust and non-specific assembly with ribonucleoproteins. To address this, we generated artificial RNA-binding proteins harboring repeats of the G/S-Y-G/S-motif-enriched low-complexity sequence with and without enriched glutamine residues and signal sequences for nuclear import/export. To minimize superfluous functions in mammalian cells, a well-characterized plant-based RNA-recognition domain [poly(A)-binding protein (PABPs)] 16 was selected as a standard RNA-recognition motif (RRM). These proteins exhibited features similar to those of ALS-linked molecules, including insoluble aggregation, formation of cytoplasmic IBs and RNA-granule components, and instigation of cytotoxicity. These findings indicated that cytoplasmic RNA-binding proteins containing PrLDs can initiate neurotoxicity independent of their specific physiological functions. We propose that these artificial RNA-binding proteins effectively replicated features of ALS-linked molecules and constitute novel experimental molecules that enable enhanced understanding of the molecular basis of ALS/FTD-related cytotoxicity.

Characterization of SYG and SYGQ-nuclear-export signal (NES)-green fluorescent protein (GFP) expression in cultured cells. To reveal cytotoxicity induced by RNA-binding proteins containing
PrLDs, we generated synthetic cDNAs encoding 67 repeats of the G/S-Y-G/S motif, 50 repeats of G/S-Y-G/S + Q or G/S-Y-G/S + Q/N motifs containing the PABP RRM, signal sequences for nuclear import/export, and GFP tags ( Fig. 1a and Supplementary information). We assembled the base pairs necessary to encode the G/S-Y-G/S motif and minimize repeats of the nucleotide sequence to avoid formation of aberrant RNA structures. Additionally, we used the well-characterized PABP RRM from Citrus sinensis, which shares 43% identity with the human PABPN1 16 , to minimize the possibility of functional redundancy in mammalian cells. Furthermore, we added a NLS (MPKKKRKVGG) or NES (MLPPLERLTLDGG) to investigate compartment-specific cytotoxicity.
We inserted these cDNAs into a mammalian expression vector, followed by transfection into various cell lines and analysis by western blot. As shown in Fig. 1b, we detected full-length forms of the SYGQ-NES [predicted size: ~68 kDa; estimated isoelectric point (pI): 5.76; http://web.expasy.org/compute_pi/] and SYG-NES proteins (predicted size: ~68 kDa; estimated pI: 5.76) at ~73 kDa according to western blot results. Both artificial proteins were also detected within the stacking gel, indicating that both were prone to aggregation and formation of high-molecular-weight complexes. A series of deletion experiments revealed that the SYG ± Q PrLD was essential for the formation of these high-molecular-weight complexes (Fig. 1c). However, we observed low levels of the SYGQ/N-NES protein (predicted size: ~68 kDa; estimated pI: 5.76) in 293 T cells, with nearly undetectable formation of high-molecular-weight complexes (Fig. 1d).
Furthermore, soluble and insoluble fractionation showed that the SYG and SYGQ-NES proteins were present in the RIPA-insoluble and urea-soluble fractions, indicating that both formed strongly insoluble aggregates similar to candidate proteins found in the brains of ALS/FTD patients (Fig. 1e). Therefore, we focused on the SYG and SYGQ-NES proteins in subsequent experiments.

Characterization of IBs formed in SYGQ-NES-expressing cells. A characteristic feature of ALS/FTD
is the presence of IBs consisting of ALS-related RNA-binding proteins and immunoreactive to p62/sequestosome 1 throughout the affected regions. To examine subcellular localization of the artificial proteins, we performed immunocytochemistry in 293 T cells (Fig. 2a). Our results indicated that the SYGQ-NES protein formed characteristic p62-positive cytosolic IBs [mean percentage of cells with IBs ± standard deviation: 7.28 ± 0.28 (n = 3); IBs formed by molecules resulting from deletion experiments were rarely observed ( Fig. 2 and Supplementary  Fig. 2). Additionally, we also generated other artificial Q/N-rich prion-like proteins, synthetic prion-forming domains (sPFD) designed by a prion-aggregation-prediction algorithm and control prion-forming domains (cPFD), which exhibit low prion propensity and used sPFD as a control (Supplementary information) 17 . Unexpectedly, western blot showed that cPFD, and not sPFD, formed high-molecular-weight complexes in the stacking-gel for unknown reasons ( Supplementary Fig. 2). All mutant constructs, including those encoding the sPFD protein, showed rare IB formation (Fig. 2), indicating that the SYGQ-repeat PrLD, RRM, and cytoplasmic localization domains were critical for IB formation.
Previous studies demonstrated that IBs involving TDP-43 and FUS in ALS/FTD-patient brain tissue recruit other RNA-binding proteins, such as components of SGs 8,10,13,[18][19][20] . As shown in Fig. 3a Furthermore, administration of the inhibitor of translational elongation, emetine, stabilizes polysomes and suppresses SG formation 21 . As shown in Fig. 3b, treatment with this drug dispersed IBs [(mean percentage of cells containing IBs formed under normal conditions in the presence and absence of emetine: 11.1 ± 0.6% vs. 5.3 ± 1.7%, respectively (P < 0.05)] in 293 T cells expressing SYGQ-NES and also mutant FUS 10 , but not aggresomes containing mutant forms of superoxide dismutase 1. These findings strongly suggested that SYGQ-NES IBs exhibited characteristics associated with SGs. Furthermore, ALS-causative mutations in FUS abnormally interact with SMN and lead to loss of nuclear Gemini of coiled bodies (GEM), and mutation of TDP-43 or FUS disrupt GEM assembly or stabilization [22][23][24][25][26] .
Here, we observed that expression of SYG/SYGQ-NES proteins significantly reduced nuclear GEMs consistent with results observed in the presence of other ALS-linked RNA-binding protein (Fig. 4).   [8][9][10][11]18 . To examine whether SYGQ-NES proteins are recruited to SGs, we performed double immunofluorescence staining in HeLa cells, which is commonly used for examining the characters of SG 6-8 after treatment with arsenite, a well-established SG inducer (38). As shown in Fig. 5, both artificial proteins, predominantly SYGQ-NES, were recruited to SGs following arsenite treatment. A series of deletion experiments revealed that SYG/SYGQ repeats, cytoplasmic localization domains, and partial Q-rich domains and RRMs were critical for SG localization. Furthermore, analysis of the number of HeLa cells expressing SYG/SYGQ-NES proteins with arsenite-induced SGs showed that  (Fig. 6). These findings indicated that SYGQ-NES expression dysregulated SG formation similar to results observed in the presence of ALS-associated mutants [8][9][10][11] .

SYG/SYGQ-NES cytotoxicity.
To determine whether SYG/SYGQ-NES expression leads to neurotoxicity, we measured neurite length on differentiated Neuro2a cells, a well-established model of neurite outgrowth [28][29][30] . SYG/SYGQ-NES-expressing cells become rounded and shrink, and showed shorter neurites compared with those observed on control cells (Fig. 7). We also observed shorter neurites in cells containing ALS-related TDP-43 (Q343R) and wild type FUS constructs. Additionally, deletion experiments revealed that SYG/SYGQ repeats, cytoplasmic localization domains and RRMs were critical for the disturbance of neurite outgrowth (Fig. 8a). Graphs indicate the ratio of the mean intensity of GFP in SGs to that observed in the cytoplasm. The fluorescent distributions in SGs from cells expressing SYG/SYGQ-NES and ΔRRM were higher than those observed in cells expressing GFP alone. *P < 0.003, vs. GFP; # P < 0.0002; vs. SYGQ-NES (one-way ANOVA). Scale bar: 10 μm.
We then assessed the induction of cell death following transfection of the artificial molecules as determined by propidium iodide-positive (PI+) cells in GFP-positive Neuro2a cells and cell sorting. As shown in Fig. 8b, we observed increased levels of PI + cells expressing both artificial proteins, but higher levels in the presence of SYG-NES, indicating that expression of either SYG or SYGQ-NES resulted in cytotoxicity.

Discussion
Increasing evidence supports hypotheses involving gain of toxic functions resulting from cytoplasmic accumulation of RNA-binding proteins involved in ALS/FTD pathogenesis. Overexpression of human TDP-43 and FUS in mice causes progressive motor-neuron degeneration similar to ALS phenotypes, including cytoplasmic aggregation [31][32][33][34][35] . However, several FUS-knockout mutant models do not exhibit phenotypes associated with motor-neuron dysfunction 36,37 , although expression of mutant FUS can also directly recapitulate ALS-like motor-neuron abnormalities in mice independent of reductions in FUS activity 38,39 . Haploinsufficiency mutations involving FUS and that lead to nonsense-mutation-mediated mRNA decay were identified in an essential-tremor-affected family that did not reveal any signs of ALS 40 . Therefore, these findings suggested that mislocated RNA-binding proteins are critical for neurodegeneration development in a gain of toxic function manner.  Our findings provided several insights into the cellular-biochemical features of ALS/FTD-linked RNA-binding proteins. Artificial RNA-binding proteins containing SYG/SYGQ repeats initiated aggregation and assembly with SG components and caused neuronal toxicity. Low-complexity sequences referred to as G/S-Y-G/S motifs, cytoplasmic localization and RRMs are essential for inhibition of neurite outgrowth (Fig. 8). SG localization requires G/S-Y-G/S motifs, cytoplasmic localization domains, partial Q-rich domains, and RRMs (Fig. 5), and IB formation also requires Q-enriched G/S-Y-G/S motifs, cytoplasmic localization domains, and RRMs (Fig. 2). Notably, the SYG-NES protein showed formation of high-molecular-weight complexes according to western blot analysis (Fig. 1), but we rarely observed IB formation in cells expressing SYG-NES proteins (Fig. 2). Although it might be possible that cells expressing SYG-NES proteins were unhealthy (rounded, shrunken, and floating) prior to the maturation of visible IBs, IB formation might be required for additional cellular process(es) beyond characteristics required formation of high-molecular-weight complexes. These characteristics were distinct from those observed in a protein harboring a de novo design of sPFDs 17 (Figs 2,5, and 8) and other ALS-linked aggregated proteins, such as SOD-1 (Fig. 3b), indicating that the artificial SYG/SYGQ-NES proteins lacking specific physiological functions associated with target RNAs effectively mimicked specific pathological features associated with ALS/FTD-linked RNA-binding proteins.
Although the detail pathological mechanism of the SYG/SYGQ-NES protein remains unknown and requires further study, recent evidence suggests that ALS-linked RNA-binding proteins, such as TDP-43 and FUS, are mislocated to the cytoplasm and might affect target RNAs and other RNA-binding proteins, resulting in impairment of RNA metabolism and RNA quality control mechanisms, such as those associated with SGs, during neurodegeneration 1,41 . The SYG-NES protein is strongly cytotoxic in the absence of IBs, indicating that IB formation is unnecessary for cytotoxicity (Figs 2 and 8); however, high-molecular-weight complexes of low-complexity sequences, including G/S-Y-G/S motifs, might interfere with other RNA-binding proteins, including TDP-43 and other SG components, resulting in cell toxicity as a result of RNA dysregulation.
Several points should be kept in mind during the exploration of ALS/FTD pathogenesis using artificial SYG/ SYGQ-NES proteins, and alternative interpretations should be considered. First, although most ALS/FTD-linked RNA-binding proteins are ubiquitously expressed in all tissues, ALS/FTD is a degenerative disease that predominantly affects the frontal lobe containing primary and secondary motor neurons from the spinal cord, which are anatomically related. Therefore, future studies need to confirm motor-dominant neurodegeneration following SYG/SYGQ-NES expression under the control of a ubiquitous promoter in transgenic animals.
Second, abundant molecular and cellular evidence suggests that the misfolding of specific proteins, such as α-synuclein, β-amyloid, and tau, can act as seeds for aggregation and can be transmitted intercellularly, thereby constituting a manner of self-dissemination similar to that observed with prions 42 . Recent neuropathological studies of postmortem brains from ALS patients described a pattern of phospho-TDP-43 immunoreactivity that agrees with a systematic spreading of TDP-43-related pathology 43,44 . Cell-biological studies also showed TDP-43 soma-to-soma transmission between cells and propagation within neuroanatomical systems across axon terminals 45 . Therefore, future studies should examine whether artificial proteins harboring low-complexity sequence(s) in PrLD(s) also exhibit similar mechanisms of seeded aggregation and propagation.
Third, affected regions in most ALS patients are characterized by TDP-43-positive IBs in the cytoplasm and also by the depletion of TDP-43 protein in the nucleus 46,47 . Analyses by cross-linking and immunoprecipitation suggested that there might be thousands of RNA molecules targeting TDP-43 and FUS during the mRNA-maturation and gene-regulation processes [48][49][50] . A study using a conditional mouse TDP-43 gene-targeting approach in spinal cord motor neurons reported progressive development of ALS-related phenotypes 51 . Therefore, theories supporting a loss-of-function phenotype associated with TDP-43 proteinopathy and leading directly to RNA dysregulation and development of neurodegeneration remain plausible 46,47,52 . In this study, we were unable to directly examine loss-of-function phenotypes associated with ALS pathogenesis using artificial proteins.
Forth, analysis by cross-linking and immunoprecipitation of sequences suggested that approximately 30% mRNA of the murine transcriptome interacted with TDP-43 and FUS [48][49][50] . Therefore, FUS and TDP 43-proteinopathies possibly cause gross RNA dysregulation rather than specific mRNA disruptions in ALS/FTD development. Therefore, an RRM of PABP from Citrus sinensis, which represents an authentic ortholog of the animal PABPs, was used for these artificial proteins to avoid specific binding to RNA in mammalian cells. We speculated that the overexpressed plant-based RRM interacted widely and non-specifically with mRNAs and caused gross RNA dysregulation; however, this does not exclude the possibility that off-target tight-binding to unexpected RNAs might affect the interpretations in studies using de novo proteins. Future studies should define comprehensive artificial SYG/SYGQ-NES proteins -RNA interaction and compare it with other ALS-linked RNA binding protein interactions.
Finally, most neurodegeneration-associated aggregated proteins, such as tau and TDP-43, are easily targeted by protein degradation systems and are multiply fragmented in the affected brain 47,53,54 . As shown in Fig. 1, many low molecular weight degradation products of SYG/SYGQ-NES were also detected using immunoblot analysis. Despite the presence of a prominent ~27 kDa fragment, which is predicted to be naïve GFP from the molecular weight, we cannot exclude the possibility of degradation products having unexpected effects and resulting in modified phenotypes. Here, the instability index for SYGQ-NES calculated by ProtParam (http://web.expasy. org/protparam/) was high (49.22), indicating that this must be improved to ensure stable expression in future studies. However, our findings suggested that the use of de novo proteins, specifically SYG/SYGQ-NES, effectively allowed the study of mechanisms associated with gain of toxic functions related to ALS/FTD pathogenesis. These results support the acquisition of an enhanced understanding of RNA dysregulation mediated by aggregated cytoplasmic RNA-binding proteins and promote the development of new strategies for the treatment of ALS/ FTD proteinopathies.
Immunofluorescence with cultured cells. At 48 h after transfection, cells were fixed with 4% paraformaldehyde on ice for 10 min and then permeabilized in 0.2% Triton X-100 at room temperature for 10 min. Post-fixation with 4% paraformaldehyde on ice for 5 min followed. After blocking of nonspecific binding, coverslips were incubated with primary antibodies and then incubated with Alexa 488 and/or Alexa 586-conjugated secondary antibodies (Invitrogen). Immunofluorescent staining was examined using a Leica TCS SP5 confocal microscope (Leica, Wetzlar, Germany).
For cytotoxicity analysis in differentiated Neuro2a cells, neurites in more than 30 transfected cells from independent experiments (n = 3) were evaluated using ImageJ, and cells containing neurites of at least two cell-body-diameter length were scored as neurite-bearing cells 28,29 . Statistical analysis. Statistical analysis of the data was performed by one-way ANOVA with Fisher's protected least squares difference test or Student's t-test using the Statview 5.0 system (Statview, Berkley, CA, USA).