Autism-associated mutation in Hevin/Sparcl1 induces endoplasmic reticulum stress through structural instability

Hevin is a secreted extracellular matrix protein that is encoded by the SPARCL1 gene. Recent studies have shown that Hevin plays an important role in regulating synaptogenesis and synaptic plasticity. Mutations in the SPARCL1 gene increase the risk of autism spectrum disorder (ASD). However, the molecular basis of how mutations in SPARCL1 increase the risk of ASD is not been fully understood. In this study, we show that one of the SPARCL1 mutations associated with ASD impairs normal Hevin secretion. We identified Hevin mutants lacking the EF-hand motif through analyzing ASD-related mice with vulnerable spliceosome functions. Hevin deletion mutants accumulate in the endoplasmic reticulum (ER), leading to the activation of unfolded protein responses. We also found that a single amino acid substitution of Trp647 with Arg in the EF-hand motif associated with a familial case of ASD causes a similar phenotype in the EF-hand deletion mutant. Importantly, molecular dynamics (MD) simulation revealed that this single amino acid substitution triggers exposure of a hydrophobic amino acid to the surface, increasing the binding of Hevin with molecular chaperons, BIP. Taken together, these data suggest that the integrity of the EF-hand motif in Hevin is crucial for proper folding and that ASD-related mutations impair the export of Hevin from the ER. Our data provide a novel mechanism linking a point mutation in the SPARCL1 gene to the molecular and cellular characteristics involved in ASD.


Results
Hevin mutants lacking the EF-hand motif are generated in the Usp15-deficient brain. Previously, we have reported that USP15 influences the spliceosome cascade and UPR pathway despite the unclear relevance 11 (Fig. 1A). Thus, we looked for ASD-associated abnormal variants that changed splicing upon Usp15 deficiency and are involved in the UPR pathway. In this context, we checked the expression profile dataset obtained from mice brains 11 and found that the 3'-region of the hevin transcript tends to be lacking in the Usp15-deficient brains (Fig. 1B, Supplementary Table S1). Thus, we speculated that the 3'-region of the hevin transcript is susceptible to abnormalities in the spliceosome cascade regulated by USP15. We determined which base sequence of the hevin transcript was altered in the Usp15-deficient brain. To do this, we conducted a 3'-rapid amplification of cDNA ends (3'-RACE) and identified two 3'-fragments lacking the EF-hand motif. One was the EF-hand deletion mutant and the other was the trans-splicing variant, which fused with a fragment of NADH: ubiquinone oxidoreductase subunit A11 (Ndufa11) (Fig. 1C-G). Because the EF-hand motif is pivotal for regulating protein structure and functions 24 , we hypothesized that EF-hand deletion mutants show abnormal structure, resulting in a decrease of secretion efficiency.
Hevin mutants lacking the EF-hand motif activate the UPR signaling caused by abnormal trafficking. To investigate whether these mutants are secreted to the extracellular space, we constructed the expression plasmids of full-length Hevin mutants and transfected them into culture cells. Although Hevin WT was efficiently secreted to the extracellular space, the deletion mutant of EF-hand 2 (Hevin ΔEF) and the fusion protein between Hevin lacking the EF-hand and Ndufa11 C-terminal region  showed reduced secretion efficiency compared to Hevin WT (Fig. 2A). These data suggest that the integrity of the EF-hand motifs is crucial for the secretion of Hevin. We next examined whether Hevin mutants accumulate in the ER. To do this, we introduced each plasmid into HeLa cells and stained using an anti-Hevin antibody. Consequently, Hevin mutants were found to be prone to accumulate in the ER compared to Hevin WT (Fig. 2B,C). These data demonstrate that an abnormal EF-hand region suppresses a Hevin export from the ER. Next, we examined whether the accumulation of Hevin mutants activates the UPR pathway. The expression of both Hevin ΔEF and Hevin-N11 increased the mRNA of the UPR markers, Bip and Chop (Fig. 2D,E) www.nature.com/scientificreports/ BIP, which is the ER chaperone protein, was increased when Hevin mutants were expressed. Importantly, the BIP expression level was similar to that of thapsigargin-stimulated cells (Fig. 2F), suggesting that Hevin EF-hand deletion mutants accumulate in the ER and activate the UPR pathway.

ASD-associated W647R mutation of Hevin activates the UPR signaling. Both SPARCL1 and
Usp15 mutations are associated with ASD 2,5,9,10 . Importantly, previous transcriptome analyses have identified single amino acid substitutions in SPARCL1/Hevin that are linked to ASD 5 . One of these mutants in which  www.nature.com/scientificreports/ Trp 647 is replaced with Arg is in the EF-hand motif that is also perturbed in Usp15 knockouts 5 (Fig. 3A, Supplementary Fig. 1). Strikingly, this amino acid is highly conserved among aves, rodents, and primates (Fig. 3B). Therefore, we postulated that this mutant [human W647R (hW647R)] could exhibit a similar cellular phenotype to the truncated Hevin mutants that we observed in Usp15 KO mice in Fig. 2. First, to investigate whether this amino acid substitution in the EF-hand motif affects Hevin trafficking, we constructed an expression plasmid using the mouse Hevin sequence. Since human Trp 647 corresponds to mouse Trp 633 (Fig. 3B), we replaced mouse Trp 633 with Arg [mouse W633R (mW633R)]. As we expected, the secretion efficiency in Hevin mW633R was attenuated compared to that in Hevin WT (Fig. 3C). In addition, the Hevin mW633R mutant was mostly localized in the ER, as well as Hevin ΔEF and Hevin-N11 ( Fig. 3D and E) and was not transported to the Golgi apparatus efficiently ( Fig. 3F and G). Furthermore, the expression of Hevin mW633R increased the mRNA level of the UPR markers, Bip and Chop, and the protein level of BIP ( Fig. 3H-J), suggesting that a single amino acid substitution in the EF-hand causes Hevin accumulation in the ER, followed by activation of the UPR signaling. Next, we investigated the mechanisms through with the Hevin mutant increases ER stress. Since unfolded proteins promote dissociation of BIP from the ER stress sensors such as IRE1, leading to interaction with freed BIP, we examined whether Hevin mW633R exhibits unfolded protein-like behavior. We found that the binding affinity of Hevin mW633R to BIP is higher than that of WT ( Fig. 3K and L). Importantly, Ca 2+ existence has little  www.nature.com/scientificreports/ effect on this binding. These data suggest that the Hevin mW633R mutant induces improper folding regardless of Ca 2+ existence and supports our idea that the Hevin mutant exhibits a different structure compared to WT.
Hevin W647R mutant shows improper folding of the hydrophobic core. Because BIP tends to recognize the hydrophobic sites of misfolded proteins 25,26 , we speculate that the Hevin WR mutant exhibits structural instability. To verify this hypothesis, a set of μs-order all-atom molecular dynamics (MD) simulations was performed for the EC domain (515-664 amino acid) of WT and hW647R mutant (Fig. 4A). As the first analysis of both trajectories, the root-mean-square deviation (RMSD) for each initial structure during each MD simulation was calculated. As shown in the RMSD profiles, the hW647R system structurally fluctuated compared to the WT system since the hW647R RMSD value increased during the MD simulations, indicating that a replacement of Trp 647 with Arg causes an structural instability (Supplementary Fig. 2A-D). To address the origin of this structural fluctuation, the solvent-accessible surface area (SASA) was measured in both systems. It is reasonable to measure the SASA values to address the mutational effect since the structural fluctuation might be caused by a protein surface deformation, in which a part of the protein surface of the WT system might be exposed to the external solvent upon this amino acid substitution. For all of the residues included within 8 Å around the mutation site, their SASA distributions significantly changed in shape, and the SASA values increased in the hW647R system compared to the WT system ( Fig. 4B and C). These analyses indicate that this amino acid substitution causes structural fluctuation of the hW647R system. As the second analysis, the radial distribution function (RDF) of water was defined for all of the residues included within 8 Å around the mutation site. The RDFs around each residue allow for identifications of the essential residues that increase SASA. Among all of the calculated distributions, the RDFs of water around V520 showed characteristic profiles (Fig. 4D), where the RDF around V520 for the hW647R system significantly increased compared to that for the WT system. This increase in RDF indicates that the water density around V520 became higher due to the point mutation. Based on the difference in these RDFs, the difference in the density of water was compared among the corresponding residues as follows: From the decomposed accumulation, we focused on the top ten residues with the especially large RDF differences (Fig. 4E). Based on the structural observation in the WT system, the hydrophobic residues included in the top ten residues had showed characteristic properties, that is, they tightly contacted each other and constructed a core site. Notably, this core formed in the WT system was not formed in the hW647R system ( Fig. 4F and G). To evaluate the structural changes of the hydrophobic core quantitively, the radius of gyration (R g ) of seven hydrophobic residues was measured. The R g value of these hydrophobic residues characterizes the size of each hydrophobic core. The distributions of R g differed between the two systems ( Fig. 4H,I). Specifically, the median value of R g measured in the WT system was 8.847 Å while that measured in the hW647R system was 9.675 Å, indicating that the hydrophobic core of the hW647R system was changed. Therefore, it is likely that Hevin hW647R has improper folding around the hydrophobic residues, resulting in structural instability.

Discussions
In this study, we identified that the defects in the Hevin EF-hand motif decrease secretion efficiency and accumulate in the ER. In addition, the ASD-associated Hevin mutant also shows impaired trafficking and hampered an export from the ER. Indeed, the Hevin mutant exposes a hydrophobic region, which is normally hidden inside, to the surface, inhibiting the export of Hevin from the ER and activating the UPR pathway ( Fig. 5). Thus, our findings provide a potential mechanism that links an ASD mutation in the SPARCL1 gene to UPR.
Here, we found that this single amino acid substitution in the EF-hand has two cellular phenotypes. First, the expression of Hevin WR activates the UPR pathway. Second, the mutation attenuates the secretion efficiency. As such, we presume that the activation of the UPR pathway and/or the decrease of secretion efficiency are possible mechanisms that connect the Hevin dysfunctions to ASD. Previously, ER stress was proposed as a cellular mechanism related to the onset of ASD 27 . In fact, the UPR-related transcripts are dysregulated in the frontal cortices of ASD patients 28 . In most cases of regulating the UPR, three distinct pathways, the inositol-requiring enzyme 1 (IRE1)-X-box binding protein 1 (XBP1) pathway, protein kinase R-like ER kinase (PERK)-eukaryotic initiation factor 2α (eIF2α) pathway, and activating transcription factor 6 (ATF6) pathway, are upregulated by the ER stress, such as abnormal Ca 2+ homeostasis in the ER and perturbed ER-Golgi trafficking. Because Hevin is predominantly expressed in astrocytes, it is likely that the perturbed the UPR pathways in astrocytes could be an important mechanism underlying ASD pathogenesis. A key question for future research is the biological relevance of how the ER stress response in astrocytes contributes to neural circuits and brain homeostasis. In Caenorhabditis elegans, overexpression of Xbp1 in astrocyte-like glial cells extends the life span via changing neuropeptide secretion machinery 29 . This observation demonstrates that upregulation of the UPR in astrocytelike glial cells is pivotal for regulating ER stress resistance and longevity, exhibiting a positive effect in nematodes. Interestingly, a mutation in Xbp1 is responsible for the onset of bipolar disorder 30 . Hence, these results raise the interesting hypothesis that dysfunctions of the Xbp1 axis could also regulate both neurodevelopmental and psychiatric disorders. In contrast, overactivation of PERK-eIF2α signaling in astrocytes causes neuronal loss and neurodegeneration in mice 31 . Strikingly, impaired UPR signaling markedly changes their secretory factors from astrocytes, suggesting that each UPR signaling in astrocytes regulates the surrounding cells in a non-cell-autonomous manner. Thus, it is plausible that the upregulation of ER stress induced by Hevin WR may change the global secretome, resulting in circuit dysfunction that is causal to ASD. Another possibility is that UPR signaling controls the timing of neural and glial differentiation. For instance, OASIS, a subfamily of ATF6,  33 , supporting the idea that ER stress is related to ASD. Hevin promotes synaptogenesis through stabilizing the interaction between NRX1α and NL1B 20 . The structure of Hevin is divided into three main parts, the acidic domain, FS domain, and EC domain 22 . The FS domain binds to NRX1α and NL1B concurrently, and the EC domain, of Hevin, including the EF-hand motif, is predicted to be essential for interacting with the collagen in a Ca 2+ -dependent manner 22,23 . We found that only a small amount of the Hevin mW633R mutant can be secreted to the extracellular space even in a mutation of the EF-hand. These data imply that exposure to the hydrophobic region of the C-terminus may result in the acquiring of novel functions that disturb innate functions in the extracellular space even in the presence of Ca 2+ . One hypothesis is that the binding affinity between Hevin and collagen proteins is changed and affects the functions of synaptogenesis. Ehlers-Danlos syndromes, which exhibit autism-like phenotypes, arise from mutation in Col1A and Col5A genes 34 . Therefore, the Hevin WR mutant could also impact the conditions of the extracellular matrix and brain microenvironment. Previous studies have reported that Sparc, a homolog of Hevin, suppresses synaptogenesis. Intriguingly, the deletion mutant lacking an acidic domain (SLF) exhibits a similar phenotype as Sparc in which SLF suppresses synaptogenesis 35 . Hevin is also cleaved at the center region (approximately 350 amino acids) by a metalloprotease, ADAMTS4, and produces the fragment containing both FS and EC domains 36 . Therefore, it is likely that Hevin hW647R mutant acquires a distinct function, like Sparc and SLF. However, further studies will be needed to clarify this possibility in the future.
Previous studies have reported that other mutations in the SPARCL1 gene associated with ASD have been revealed, and most mutations reside in the FS domain and EC domain 5 (see Fig. 3A). Mutations in which Thr 516 and Met 587 are replaced with Met and Ile, respectively, are next to the residue (Asp 517 and Tyr 588 ) with large RDF differences. Thus, it is likely that the native hydrophobic core is unfolded when these amino acids are substituted. ASD-associated amino acid substitutions around the EF-hand motif may change the native hydrophobic core, resulting in a fluctuated structure of Hevin. Notably, Ca 2+ deprivation has less effect on the binding between BIP and the Hevin WR mutant. Thus, it is plausible that exposure of the hydrophobic region covered not only by Trp 647 but also by Thr 516 and Met 587 impairs Hevin secretion and synaptogenesis, regardless of Ca 2+ .
In summary, here we found that the EF-hand motif in Hevin plays an important role for proper trafficking to the extracellular spaces. A loss of this domain causes protein accumulation in the ER and activates the UPR pathway. Furthermore, we found that ASD-associated mutation in which Trp 647 is replaced with Arg exhibits a similar phenotype in cells. Importantly, in the hW647R mutant Hevin the hydrophobic core of the EF hand is exposed to the surface likely causing structural instability. Taken together, our findings provide multiple molecular mechanisms linking an ASD point mutation in the SPARCL1 gene with cellular phenotypes underlying the onset of ASD.

Materials and methods
Plasmids. Mouse Hevin cDNA was amplified from pMD-mSPARCL1 (Sino Biological Inc. MG50544-M) and cloned into pCRblunt. The deletion mutant of EF-hand (Hevin ΔEF) was constructed using the inverse plymerase chain reaction (PCR) method. Briefly, Hevin ΔEF cDNA was amplified from pCRblunt-Hevin, followed by conducting a T4 polynucleotide kinase reaction (TAKARA). After phosphorylation, this fragment was self-ligated and constructed a circular plasmid as a pCRb-Hevin ΔEF. The Hevin-Ndufa11 (Hevin-N11) frag- www.nature.com/scientificreports/ ment was amplified from Hevin cDNA and Ndufa11 cDNA was obtained from the 3'-RACE method; next, these fragments were ligated using an In-Fusion system (TAKARA). The Hevin-N11 fragment was inserted into the pCR-blunt using the In-Fusion system. The gene encoding Hevin mW633R was generated by a quick-change method using KOD polymerase. The GST-Hevin proteins were generated by the SLiCE method 37 . Briefly, the full-length Hevin cDNA was amplified using KOD plus DNA polymerase (TOYOBO) and inserted into the BglII sites of pCS4 vector. The GST-tagged Hevin and Hevin mW633R mutant were generated using the SLiCE reaction. The GST fragment was subcloned from pGEX-6P-1 plasmid into the site between the Hevin signal-peptide and another part using the SLiCE reaction. The primers used were as follows:

3′-rapid amplification of cDNA ends (3′-RACE). Total RNAs from WT and USP15-deficient brains
were isolated by ISOGEN II (NIPPON GENE) according to the manufacturer's instructions. The cDNAs were synthesized by reverse transcriptase and 100 units of ReverTra Ace (TOYOBO) together with 25 pmol random hexamer primer (TOYOBO), 20 nmol dNTPs, and total RNAs. The 3′-region of the hevin transcript was determined using the 3′-RACE method. First-strand synthesis was conducted using the Hevin outside primer and Hevin adaptor primer. Then, the second round of PCR was conducted using the Hevin inside primer and Hevin amplification primer. The 3′-fragment amplified by 3′-RACE was inserted into the pCR-blunt and the DNA sequences were analyzed. The primers used were as follows: Hevin outside primer, 5′-aaatgctgaaccttcagatgagggc-3′; Hevin adaptor primer, 5′-ggccacgcgtcgactagtacttttttttttttttttt-3′; Hevin inside primer, 5′-agagactcttggctggagaccatc-3′; Hevin amplification primer, 5′-ggccacgcgtcgactagtac-3′; Antibodies. Cell culture and transfection. HEK293T 38,39 cells and HeLa cells 38,39 were cultured in Dulbecco's modified Eagle's medium (high glucose) (Wako) containing 5% fetal bovine serum (FBS), 100 units penicillin, and 100 mg streptomycin (P/S Thermo Fisher Scientific). Cells were transfected using polyethyleneimine MAX (PEI Max) (Polyscience). The amounts of the plasmids and PEI Max were optimized in proportion to the relative www.nature.com/scientificreports/ surface area and number of cells. Cells were plated at 1.5 to 5.0 × 10 5 cells in 4 ml per 60 mm dish and incubated at 5% CO 2 and 37 °C for 1 day. The plasmids (8.0-16.0 μg) were mixed with 500 μl of Opti-MEM (Thermo Fisher); 1.0 μg/ μl of PEI Max (32.0-64.0 μl) was mixed with 500 μl of Opti-MEM in another tube. Both solutions were combined and incubated for 20 min at room temperature, followed by adding these mixtures to cells. After incubation for 1 h, the media were changed to fresh culture medium. Neuro-2a cells were cultured in Eagle's minimum essential medium (Wako) containing 10% FBS and P/S. Cells were plated at 1.5 × 10 5 cells on a 6-well plate and incubated at 5% CO 2 and 37 °C for 1 day. The plasmids ( Immunoblot analysis. HEK293T cells or Neuro-2a cells were plated at 0.8 or 1.5 × 10 5 cells on a 6-well plate and incubated at 37 °C with 5% CO 2 for 1 d. Cells were transfected and then incubated for 3 days. The cells were collected with lysis buffer [20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 0.5% NP-40, and 1 mM DTT]. Cell lysates were centrifuged at 14,000 rpm for 5 min. The supernatant was run on SDS-PAGE for protein separation, followed by electrophoretic transfer to a polyvinylidene difluoride membrane (Pall). After 1 h blocking by 5% skim milk at room temperature, membranes were incubated with primary antibodies overnight at 4 °C. The proteins on the membranes were then detected with HRP-conjugated secondary antibodies and chemiluminescence reagents (Chemi-Lumi One Super, Nacalai Tesque). All data were reproduced in at least two independent experiments.  22 . To model the W647R system, this amino acid substitution was induced using the Pymol program (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.). In each system, the missing amino acid residues were added using the SWISS-MODEL program 41 . The modeled structures were solvated with the TIP3P water model 42 in an orthorhombic box. The LINCS 43 and SETTLE 44 algorithms were adopted to extend the MD time step to 2 fs. The modified Berendsen thermostat 45 and Parrinello-Rahman method 46,47 controlled the temperatures and pressures of all systems, respectively. The particle mesh Ewald method 48 evaluated electrostatic interactions using a real-space cutoff value of 10 Å. The cutoff value for the van der Waals interactions was set to 10 Å. All of the MD simulations were performed with the GPU version of the GROMACS 2020 package 49 using the Amber 14SBonlysc force field 50 . Before conformational sampling, energy minimizations on the initially modeled systems removed the steric crashes of atoms. Subsequently, 100-ps NVT (T = 300 K) and 100-ps NPT (T = 300 K and P = 1 bar) MD simulations relaxed each system. Lastly, the final snapshots of the NPT simulations on each system were specified as the starting configurations of their production runs. To generate statistically reliable MD trajectories, multiple 1-μs MD simulations (1-μs × 3 runs for each system) were independently started from each relaxed configuration by regenerating their initial velocities. Finally, the first 0.4-μs trajectories were considered as the equilibration phase, and the remaining 0.6-μs trajectories were used for analyses in each system. Statistical analysis. Prism ver.8.4.3 software (GraphPad Software, Inc.) was used for all statistical analyses.

Quantitative real-time PCR (qPCR
Statistical significance was analyzed using Student's t-test and analysis of variance (ANOVA) followed by Dunnett's test.

Data availability
The GEO accession number for the array data set is GSE145385 11 . The PDB code of the Hevin crystal structure is 7KBU 22 . The authors declare no competing interests. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
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/.