Angulin proteins ILDR1 and ILDR2 regulate alternative pre-mRNA splicing through binding to splicing factors TRA2A, TRA2B, or SRSF1

Angulin proteins are a group of evolutionally conserved type I transmembrane proteins that contain an extracellular Ig-like domain. In mammals, three angulin proteins have been identified, namely immunoglobulin-like domain containing receptor 1 (ILDR1), immunoglobulin-like domain containing receptor 2 (ILDR2), and lipolysis-stimulated lipoprotein receptor (LSR). All three proteins have been shown to localize at tight junctions (TJs) and are important for TJ formation. Mutations in ILDR1 gene have been shown to cause non-syndromic hearing loss (NSHL). In the present work, we show that ILDR1 binds to splicing factors TRA2A, TRA2B, and SRSF1, and translocates into the nuclei when the splicing factors are present. Moreover, ILDR1 affects alternative splicing of Tubulin delta 1 (TUBD1), IQ motif containing B1 (IQCB1), and Protocadherin 19 (Pcdh19). Further investigation show that ILDR2, but not LSR, also binds to the splicing factors and regulates alternative splicing. When endogenous ILDR1 and ILDR2 expression is knockdown with siRNAs in cultured cells, alternative splicing of TUBD1 and IQCB1 is affected. In conclusion, we show here that angulin proteins ILDR1 and ILDR2 are involved in alternative pre-mRNA splicing via binding to splicing factors TRA2A, TRA2B, or SRSF1.


Results
Identification of TRA2A, TRA2B, and SRSF1 as ILDR1-binding partners. In order to identify ILDR1-binding partners in the inner ear, we performed yeast two-hybrid screening of a chicken cochlear cDNA library using the C-terminal intracellular domain of chicken ILDR1 (228-553 aa) as bait. Among the positive clones identified, several clones encode for a group of splicing factors, including TRA2A, TRA2B, and SRSF1 (Table 1). TRA2A, TRA2B, and SRSF1 belong to SR protein family, which share one or two serine/arginine-rich domain (RS domain) as well as RNA recognition motif (RRM) (Fig. 1A). SR proteins have been shown to play important roles in constitutive as well as alternative pre-mRNA splicing 18 .
We then performed co-immunoprecipitation (co-IP) experiments to confirm the interaction between ILDR1 and the splicing factors. ILDR1 is quite conserved in vertebrates. Chicken and mouse ILDR1 share 95% and 60% homology in the extracellular part and the C-terminal end, respectively. In the following work we focus on mouse proteins. The co-IP results showed that EGFP-tagged mouse TRA2A, TRA2B, or SRSF1 is co-immunoprecipitated with Myc-tagged ILDR1 cytoplasmic domain (Fig. 1B). Likewise, EGFP-tagged full-length ILDR1 could be co-immunoprecipitated with Myc-tagged TRA2A, TRA2B, or SRSF1 (Fig. 1C). As a control, another SR protein, SRSF5, was included in the experiment, and the results showed that SRSF5 is not co-immunoprecipitated with ILDR1, confirming the specific interaction between ILDR1 and TRA2A/TRA2B/SRSF1 (Fig. 1C). To further narrow down which domain(s) of TRA2A/TRA2B/SRSF1 is required for the interaction, we performed co-IP experiments with different domains of TRA2A/TRA2B/SRSF1. The results showed that the RS domain is responsible for the interaction with ILDR1 ( Fig. 1D-F).
Three ILDR1 splicing transcriptional variants have been identified, namely ILDR1α, ILDR1α, and ILDR1β 1 . Full-length ILDR1 (ILDR1α) contains a di-leucine motif and a cysteine-rich region in the cytoplasmic part, and is usually simply referred as ILDR1 for convenience. Compared to ILDR1α, ILDR1α' misses the di-leucine motif, whereas ILDR1β misses the transmembrane domain as well as the cysteine-rich region (Fig. S1A). Expression of ILDR1α and ILDR1α', but not ILDR1β, was detected in mouse inner ear by RT-PCR experiment (data not shown). We then amplified the cDNA of ILDR1α' from mouse inner ear and performed co-IP experiment. The result showed that unlike ILDR1α, ILDR1α' cytoplasmic domain was not co-immunoprecipitated with TRA2B, suggesting that the di-leucine motif is necessary for the interaction between ILDR1 and the splicing factors (Fig. S1B). ILDR1 translocates into the nuclei when TRA2A, TRA2B, or SRSF1 is present. Exogenous ILDR1 has been shown to localize in the cytoplasm of HEK293T cells 1 . Consistently, we found that ILDR1-GFP mainly localizes in the cytoplasm in COS-7 cells ( Fig. 2A). In contrast, TRA2A-mCherry, TRA2B-mCherry, and SRSF1-mCherry localize exclusively in the nuclei (Fig. 2B-D). Interestingly, when cotransfected together with TRA2A-mCherry, TRA2B-mCherry, or SRSF1-mCherry, ILDR1-GFP translocates from the cytoplasm into the nuclei ( Fig. 2E-G, Fig. S2). The recruitment of ILDR1 into the nuclei by these splicing factors suggests that ILDR1 might play a role in splicing regulation. Noticeably, ILDR1α'-GFP does not move into the nuclei when TRA2A/ TRA2B/SRSF1 is present (Fig. S1C-F), consistent with the finding that ILDR1α' does not interact with these splicing factors (Fig. S1B).
Tra2a, Tra2b, and Srsf1 are expressed in the mouse inner ear. Transcriptome analysis has suggested that Tra2a, Tra2b, and Srsf1 are expressed in the mouse inner ear (SHIELD; https://shield.hms.harvard.edu) 25,26 . To examine their expression pattern in the cochlea, in situ hybridization was performed using whole-mount for TRA2B (or SRSF1) with or without ILDR1, and RT-PCR results showed that TRA2B and SRSF1 promote the inclusion of exon 4 of TUBD1 and exon 12 of IQCB1, respectively, whereas ILDR1 antagonizes the function of TRA2B/SRSF1 ( Fig. 3A and B). As for Pcdh19, HEK293T cells were transfected with expression vectors for SRSF1 with or without ILDR1 alongside a minigene consisting of exon 2 and flanking exons/introns of mouse Pcdh19 gene. RT-PCR-based evaluation of pre-mRNA splicing demonstrated that SRSF1 promotes the inclusion of Pcdh19 exon 2, whereas ILDR1 antagonizes the function of SRSF1 (Fig. S4A). Taken together, our data suggest that ILDR1 regulates SRSF1-and TRA2B-meidated alternative pre-mRNA splicing. We then used Ildr1 knockout mice to examine whether loss of ILDR1 affects alternative pre-mRNA splicing in the inner ear. Total RNA were extracted from the cochlea of P0 wildtype and Ildr1 knockout mice, and RT-PCR was performed to examine the alternative splicing of Tubd1, Iqcb1, and Pcdh19 genes. Unexpectedly, no difference was observed between wildtype and Ildr1 knockout mice ( Fig. 3C and D, Fig. S4B). To further examine whether SCiEntifiC RepoRts | 7: 7466 | DOI:10.1038/s41598-017-07530-z loss of ILDR1 affects alternative splicing, RNA from P0 cochlea of wildtype and Ildr1 knockout mice were subjected to RNA-seq analysis, which did not reveal any significant differences in alternative gene splicing (data not shown). To verify the RNA-seq result, twenty-four genes that might show different splicing patterns according to RNA-seq result were picked and their alternative splicing in wildtype and Ildr1 knockout mice was examined by RT-PCR. The results showed that the splicing of these genes is indeed not affected by ILDR1 deficiency (Fig. S5).
Ildr2 is upregulated in the inner ear of Ildr1 knockout mice. The fact that alternative splicing is not affected by loss of ILDR1 prompted us to look for possible explanation. It has been shown that sometimes loss of a particular protein could be compensated for by its homologous protein. As mentioned above, ILDR1 belongs to evolutionally conserved angulin protein family, which includes ILDR1, ILDR2, and LSR. We then examined the expression of Ildr2 and Lsr in Ildr1 knockout mouse. RT-PCR and quantitative real-time PCR showed that expression of Ildr2 is greatly increased in the basilar membrane of Ildr1 knockout mice compared with wildtype mice ( Fig. 4A and B). However, the expression of Lsr was not obviously affected by Ildr1 deficiency (Fig. 4A and B). Whole-mount in situ hybridization was performed to examine the expression pattern of Ildr2 and Lsr in the mouse cochlea. The results showed that Ildr2 and Lsr are expressed in both hair cells and supporting cells (Fig. S6). In situ hybridization results also confirmed that Ildr2, but not Lsr, is upregulated in the cochlea in Ildr1 knockout mice ( Fig. 4C-D').
ILDR2 binds TRA2A/TRA2B/SRSF1. We then performed experiments to examine the possibility that ILDR2 and/or LSR might regulate alternative splicing and compensate for loss of ILDR1. The three angulin proteins have similar domain architecture and share high homology between each other (Fig. 5A). First, we examined whether ILDR2 and LSR interact with TRA2A/TRA2B/SRSF1 by performing co-IP experiments. The results showed that both ILDR2 and LSR could be co-IPed together with TRA2A/TRA2B/SRSF1 (Fig. 5B-D). The alternative splicing of Iqcb1 exon 12 was not affected in Ildr1 knockout mice. The relative exon inclusion rate was calculated from three independently performed experiments. The differences between groups were determined by Student's t-test. **P < 0.01; ***P < 0.001; ns, not significant.
Next, we examined the subcellular localization of ILDR2 and LSR in cultured cells. Similar to ILDR1, when expressed alone in cultured COS-7 cells, GFP-tagged ILDR2 and LSR localize in the cytoplasm ( Fig. 5E and G). When mCherry-tagged TRA2B is present, ILDR2 translocates into the nuclei and colocalizes with TRA2B, whereas LSR still remains in the cytoplasm ( Fig. 5F and H). This result suggest that although all three angulin proteins could bind the splicing factors in vitro, only ILDR1 and ILDR2, but not LSR, colocalize with the splicing factor in the nuclei in cultured cells.
ILDR2 affects alternative pre-mRNA splicing. TUBD1, IQCB1, and Pcdh19 genes were used as target genes to examine whether ILDR2 and/or LSR could affect alternative splicing. RT-PCR-based evaluation of pre-mRNA splicing demonstrated that, similar to ILDR1, ILDR2 inhibits SRSF1-or TRA2B-mediated alternative splicing, whereas LSR does not ( Fig. 6A and B, Fig. S7). Taken together, given the fact that Ildr2 is upregulated in Idlr1 knockout mice and that ILDR2 can regulate alternative splicing as ILDR1 does, we hypothesize that ILDR2 might compensate for the loss of ILDR1 in splicing regulation.
To further test this hypothesis, we knockdown the expression of endogenous ILDR1 and/or ILDR2 in cultured cells and examined its effect on alternative splicing. Transiently-transfected siRNA downregulates the expression of ILDR1 or ILDR2 specifically in HEK293T cells without affecting each other ( Fig. S8A and B). This is in sharp contrast to what happens in Ildr1 knockout mice. As a result, alternative splicing of TUBD1 and IQCB1 was affected in ILDR1 or ILDR2 knockdown cells ( Fig. 6C and D). When the expression of both ILDR1 and ILDR2 was downregulated simultaneously with siRNAs, alternative splicing of TUBD1 and IQCB1 was affected to a greater extent ( Fig. 6C and D, Fig. S8C and D). This result strongly supports the role of ILDR1/2 proteins in splicing regulation.

Discussion
Angulin proteins are evolutionally conserved type I transmembrane proteins containing an extracellular Ig-like domain. At present three mammalian angulin proteins have been identified, namely LSR, ILDR1, and ILDR2, which are also known as angulin-1, angulin-2, and angulin-3, respectively 2 . All three angulin proteins have been shown to localize at tight junctions (TJs) and could recruit tricellulin, another important TJ component 2,30 . In the present work we demonstrate that besides TJs regulation, angulin proteins are also involved in alternative pre-mRNA splicing through binding to specific splicing factors.
Our data show that when expressed in cultured cells, ILDR1 and ILDR2 bind to splicing factors TRA2A, TRA2B, or SRSF1, and translocate into the nuclei. Several lines of evidence suggest that the interaction and translocation are not caused by the overexpression of tagged proteins. First, SR family splicing factor SRSF5 was included in the co-IP experiment as a negative control, and the result showed that SRSF5 is not co-IPed with ILDR1 as TRA2A, TRA2B, or SRSF1 does. Presumably the interaction of RS domain with ILDR1 requires specific amino acids context that does not exist in SRSF5. Second, ILDR1α' is an ILDR1 variant that only misses a di-leucine motif compared to the full length ILDR1. However, ILDR1α' does not bind to TRA2A, TRA2B, or SRSF1, and does not translocate to the nuclei when the splicing factors are present. This result suggests that the di-leucine motif is necessary for the interaction. Third, LSR is an angulin protein family member and is homologous to ILDR1 and ILDR2. Although LSR is co-IPed with the splicing factors in vitro, it does not translocate into the nuclei when the splicing factors are present. Taken together, we believe that the interaction between ILDR1/ ILDR2 and the splicing factors is specific. showing that EGFP-tagged SRSF1 was co-immunoprecipitated with Myc-tagged cytoplasmic fragment of ILDR1, ILDR2, or LSR. Expression vectors were transfected into HEK293T cells to express epitope-tagged proteins, and cell lysis were subject to immunoprecipitation. 5% of total protein was loaded as input. IP indicates antibody used for immunoprecipitation and WB indicates antibody used for detection. Uncropped blots are shown in Fig. S10. ILDR2-GFP (E) and LSR-GFP (G) localize in the cytoplasm. However, when TRA2B-mCherry is present, ILDR2-GFP (F) but not LSR-GFP (H) translocates into the nuclei. Expression vectors were transfected into COS-7 cells to express epitope-tagged proteins. Nuclei were stained with DAPI. Scale bar: 10 μm. (D) RT-PCR revealed that the inclusion of exon 12 of IQCB1 in HEK293T cells was enhanced when ILDR1 and/or IDLR2 were knockdown. The relative exon inclusion rate was calculated from three independently performed experiments. The differences between groups were determined by Student's t-test. **P < 0.01; ***P < 0.001; ns, not significant.
SCiEntifiC RepoRts | 7: 7466 | DOI:10.1038/s41598-017-07530-z It has been shown that transmembrane proteins such as receptor tyrosine kinases (RTK) and Notch receptor can translocate into the nuclei after cleavage by proteases. Through sequential cleavage by multiple proteases, a soluble cytoplasmic domain is released and translocates into the nuclei 31, 32 . We examined the molecular weight of ILDR1, ILDR2 and LSR by performing western blot, and found that ILDR1/ILDR2/LSR is not cleaved into smaller fragment when the splicing factors (SRSF1, TRA2A, or TRA2B) are present. This result suggests that the translocation of ILDR1 and ILDR2 into the nuclei is not mediated by protease cleavage.
Another hypothetical explanation suggests that after activation by ligands, the full-length transmembrane proteins are delivered from the cell surface to the endoplasmic reticulum (ER), then extracted into the cytoplasm and finally translocated into the nuclei, although the detailed mechanism remains elusive 31 . Interestingly, our group and others show that when heterogeneously expressed in cultured cells, ILDR1, ILDR2 or LSR mainly localizes in the cytoplasm with an ER-like pattern 1 . In fact, exogenous ILDR2 has been shown to primarily locate in the ER of cultured hepatoma and neuronal cells 33 . In this scenario, the splicing factors might participate in the shuttling of ILDR1/ILDR2 into the nuclei through binding to them. Further investigation is needed to fully understand the detailed mechanism.
Our data show that the interaction with angulin proteins requires the RS domain of TRA2A/TRA2B/SRSF1. RS domain is involved in protein-protein interactions that facilitate recruitment of the spliceosome 34,35 , or directly contact the pre-mRNA to promote spliceosome assembly 36,37 . RS domain was also suggested to act as a nuclear localization signal (NLS) and regulate the nuclear localization of SR proteins through binding to transpotin-SR 38,39 .
Here we show that the interaction of the RS domain with angulin proteins does not affect the nuclear localization of TRA2A/TRA2B/SRSF1. Nevertheless, interaction with the RS domain interferes with TRA2A/TRA2B/ SRSF1-mediated alternative splicing through a yet unknown mechanism.
We do not observe any changes in alternative splicing in Ildr1 knockout mice. There has been evidence suggesting that angulin proteins could compensate for the loss of each other. For example, in Ildr1 knockout mice, compensatory TJ localization of LSR was observed in the organ of Corti, which is believed to be responsible for recruiting tricellulin to TJs in the absence of ILDR1 5 . In the present work, we found that Ildr2 expression is upregulated in the inner ear of Ildr1 knockout mice, and might compensate for ILDR1 deficiency in alternative splicing regulation. Consistently, alternative splicing is affected when endogenous ILDR1 and ILDR2 expression is knockdown in cultured cells, strongly supporting the role of ILDR1/2 proteins in splicing regulation.
In the present work, we show that ILDR1/ILDR2 could regulate the alternative splicing of TUBD1, IQCB1, and Pcdh19. TUBD1 encodes delta-tubulin that is associated with the centrioles 40 . In testis, delta-tubulin localizes at the manchette in the sperm head as well as along the principal piece of sperm flagellum, and is involved in sperm maturation 41,42 . IQCB1 encodes an IQ domain-containing protein nephrocystin 5, and mutation of IQCB1 gene is the most frequent cause of the renal-retinal Senior-Loken syndrome (SLSN) 43 . IQCB1 localizes to the primary cilia of renal epithelial cells and connecting cilia of photoreceptor cells, and is required for the trafficking of membrane cargos to the cilia 44 . PCDH19 encodes a delta-protocadherin, and mutations of PCDH19 are associated with epilepsy and mental retardation [45][46][47] . Transcriptome analysis suggested that Tubd1, Iqcb1, and Pcdh19 are expressed in the mouse inner ear (SHIELD; https://shield.hms.harvard.edu) 25,26 , whereas their exact roles in hearing remain elusive. Further investigation is also needed to identify more target genes other than TUBD1, IQCB1, and Pcdh19 whose alternative splicing is regulated by ILDR1/2.

Materials and Methods
Animals. Generation and characterization of Ildr1 knockout mice have been described previously 6 . All animal experiments were approved by the Ethics Committee of Shandong University and conducted accordingly. All methods were performed in accordance with the relevant guidelines and regulations.
Plasmid construction. The cDNA encoding the cytoplasmic part of chicken ILDR1 (amino acids 228-553) was cloned into vector pBD-GAL4 Cam (Stratagene, La Jolla, CA, USA) to express the bait protein for yeast two-hybrid screening. The coding sequences of mouse Ildr1, Ildr2, and Lsr were cloned into pEGFP-N2. The coding sequences of mouse Tra2a, Tra2b, Srsf1, and Srsf5 were cloned into pmCherry-N1, pEGFP-C2, and pMyc-C2 (modified pEGFP-C2 with EGFP-coding sequence replaced by Myc-coding sequence). The cDNA encoding the cytoplasmic part of mouse ILDR1, ILDR2, and LSR were cloned into pMyc-C2. The cDNA encoding the RS and RRM domains of mouse TRA2A, TRA2B, and SRSF1 were cloned into pEGFP-C2. Mouse Pcdh19 minigene was amplified from mouse genomic DNA and cloned into pcDNA3.1(+). All the constructs were verified by Sanger sequencing.
Yeast two-hybrid screening. Yeast two-hybrid screening was performed as described previously 48,49 .
Briefly, yeast strain AH109 (Clontech, Mountain View, CA, USA) was sequentially transformed with the bait plasmid and a chicken cochlear cDNA library in HybriZAP pAD-GAL4 vector 50 . HIS3 was used as the reporter gene for the screening in presence of 2.5 mM 3-amino-1,2,4-triazole (3-AT). Positive colonies were further tested for activation of two other reporter genes, ADE2 and lacZ. Then the pAD-GAL4 prey vectors in triple-positive colonies were recovered, and cDNA inserts were determined by Sanger sequencing.

RNA extraction and RT-PCR.
Total RNA was isolated from mouse tissues or cells transfected with expression vectors using TRIzol reagent (Ambion, Carlsbad, CA, USA) according to the manufacturer's instructions. Reverse transcription (RT) was carried out using a cDNA synthesis kit (TaKaRa Bio Inc., Dalian, China). Polymerase chain reaction (PCR) was performed using this cDNA as template with the following primers: Ildr1 forward primer, CCGGCGGCTGATGAAGAAAGACTC, reverse primer, AGGGC AGCAACAGCGGGTAGGA (706 bp); Ildr2 forward primer, GGGCTGCTTGCTGATCTCTT, reverse primer, CAAAGTTCTTCCGCGACAGC (745 bp); Lsr forward primer, GCTATGTCAGATGTCCCTGCT, reverse primer, GTCATAGAGGTCATCCCGGC (725 bp); Tra2b forward primer, TTCCCGAAGTGGAAGTGCTC, r e v e r s e p r i m e r, C C T G C G ATA AT C T C G G C T G T ( 2 2 6 b p ) ; S r s f 1 f o r w a r d p r i m e r, GGACCGCCCTTCGCCTTCGTT, reverse primer, ACTCTGTTCTCGGACCGCCTGGAC (212 bp); β-actin forward primer, ACGGCCAGGTCATCACTATTG, reverse primer, AGGGGCCGGACTCATCGTA (372 bp). To achieve the best possible sensitivity and specificity, cycle lengths for different PCR reaction sets were adjusted between 24 and 36 cycles, and annealing temperatures were adjusted between 55 and 62 °C. The PCR products were separated by electrophoresis on agarose gel. Quantitative real-time PCR. Quantitative real-time PCR was carried out using SYBR ® Premix Ex Taq TM system (Perfect Real Time, Takara). The primers and template were the same as that used in RT-PCR. Amplification and detection were run in a Roche 480 Sequence Detection System with an initial cycle of 95 °C for 10 s followed by 40 cycles of 95 °C for 5 s, 62 °C for 10 s and 72 °C for 5 s. All PCR reactions were performed in triplicate.

Co-immunoprecipitation (co-IP) and western blot.
RNA-seq. RNA-seq was carried out by Genesky Biotechnologies Inc, Shanghai, China. Briefly, the basilar membranes were collected from postnatal day 0 (P0) wild type or Ildr1 -/mice, and the RNA sequencing libraries were constructed from the extracted and amplified RNA using the standard Illumina library preparation protocols. RNA-seq was performed on Illumina HiSeq. 2000 platform using 100 bp PE protocol. Raw sequencing reads were evaluated by FastQC, then trimmed by trim_galore to remove the primers and low-quality (Q < 10) sequences. Cleaned reads were aligned to GRCm38/mm10 mouse genome assembly using Tophat with at most 2 mismatches. Alternative splicing (AS) events of samples were extracted and compared using the tool ASprofile 51 .
Mouse Protocadherin 19 (Pcdh19) genomic sequence spanning exon 1 through 3 was PCR amplified from mouse inner ear genomic DNA using forward primer 5′-TGGAAGCTTTCCATTGCGGGCATTCTCTT-3′ c o nt a i n i n g a f l a n k i n g Hi n d I I I r e s t r i c t i o n s i t e ( u n d e r l i n e d ) a n d r e v e r s e p r i m e r 5′-ATTGGTACCGGGAGGAGCAACTGACAACA-3′ containing a flanking Kpn I restriction site (underlined). This fragment was cloned into pcDNA3.1(+) to construct the reporter minigene plasmid, which was then used for transfection into HEK293T cells together with other expression plasmids. The splicing efficiency of Pcdh19 minigene was evaluated by performing RT-PCR with forward primer 5′-TGGCAATCAAATGCAAGCGT-3′ and reverse primer 5′-ATGCCCATAGGAGTACTCAGC-3′.
Statistical analysis. Data were presented as means ± SD from at least 3 independent experiments. The differences between groups were determined by Student's t-test. *P < 0.05, **P < 0.01, ***P < 0.001.