DNAJC17 is localized in nuclear speckles and interacts with splicing machinery components

DNAJC17 is a heat shock protein (HSP40) family member, identified in mouse as susceptibility gene for congenital hypothyroidism. DNAJC17 knockout mouse embryos die prior to implantation. In humans, germline homozygous mutations in DNAJC17 have been found in syndromic retinal dystrophy patients, while heterozygous mutations represent candidate pathogenic events for myeloproliferative disorders. Despite widespread expression and involvement in human diseases, DNAJC17 function is still poorly understood. Herein, we have investigated its function through high-throughput transcriptomic and proteomic approaches. DNAJC17-depleted cells transcriptome highlighted genes involved in general functional categories, mainly related to gene expression. Conversely, DNAJC17 interactome can be classified in very specific functional networks, with the most enriched one including proteins involved in splicing. Furthermore, several splicing-related interactors, were independently validated by co-immunoprecipitation and in vivo co-localization. Accordingly, co-localization of DNAJC17 with SC35, a marker of nuclear speckles, further supported its interaction with spliceosomal components. Lastly, DNAJC17 up-regulation enhanced splicing efficiency of minigene reporter in live cells, while its knockdown induced perturbations of splicing efficiency at whole genome level, as demonstrated by specific analysis of RNAseq data. In conclusion, our study strongly suggests a role of DNAJC17 in splicing-related processes and provides support to its recognized essential function in early development.

SCIENTIFIC RepoRts | (2018) 8:7794 | DOI: 10.1038/s41598-018-26093-1 to differential alternative splicing in blood cells from Autism Spectrum Disorder (ASD) 7 . Thus, although the specific function(s) of DNAJC17 is still unknown, it is likely that corresponding gene alterations result in detrimental effects.
Dnajc17 is the vertebrate ortholog of Saccharomyces Cerevisiae Cwc23, an essential protein involved in pre-mRNA splicing 8,9 . However, there is no functional data about the implication of DNAJC17 in this process. Notably, DNAJC17 contains a RRM motif, which represents the most common RNA binding motif in vertebrates and characterizes many splicing factors, like serine-arginine rich (SR) proteins and heterogeneous ribonucleoproteins (hnRNPs) 10 . In this study, we have investigated the functional role of human DNAJC17 by combined transcriptomic/proteomic approaches. Additional protein localization experiments provided information on the predominant nuclear localization of the protein, where it co-localizes with splicing speckles. Minigene splicing assays, together with a splicing-specific analysis of RNAseq data, suggested DNAJC17 direct role in cell mechanisms related to pre-mRNA splicing. By introducing novel functional informations, this work complements preliminary data on its ortholog in yeast 9 , which were obtained with gene deletion mutants, and confirmed the involvement of molecular chaperones in the splicing machinery at splicing speckles.

Results
Gene expression profiling highlights general biological processes influenced by DNAJC17 expression. To shed light on DNAJC17 function, we analyzed gene expression profiles induced by its depletion. HeLa cells were transfected with either a DNAJC17-specific siRNA (DNAJC17 kd ) or a scramble siRNA, as a control (DNAJC17 wt ). Western blot analysis performed at 96 h after siRNA transfection confirmed a strong reduction of DNAJC17 levels in DNAJC17 kd cells as compared to control DNAJC17 wt cells (Fig. 1A). DNAJC17 silencing effect was then investigated by comparing DNAJC17 kd cells transcriptome with that of DNAJC17 wt cells through high-throughput sequencing of total RNA (RNA-Seq). Three replicates for both conditions were sequenced producing about 7 × 10 7 reads for sample, that were mapped to reference genome (Homo sapiens, GRCh37) using TopHat aligner 11 . Correctly mapped reads were then processed by Cufflinks 12 to assemble genes and transcripts and to calculate their relative expression levels. A total of 20825 assembled genes were selected for further analysis to highlight differential gene expression. Using absolute fold change ≥1.5 with corrected p-value ≤ 0.05 as cut-off, we identified 884 genes (4,24% of expressed genes) as differentially expressed in DNAJC17 kd cells (Table S1). Among them, 360 (40%) were up-regulated and 524 (60%) down-regulated in response to DNAJC17 depletion ( Fig. 1B and Table S1). Gene Ontology (GO) terms enrichment analysis using R package clusterProfiler 13 revealed that differentially regulated genes are enriched in functional categories representing ten Biological Processes (q-value ≤ 0.05) and seven Molecular Functions (q-value ≤ 0.05) (Fig. 1C). In both groups, enriched terms are related to protein synthesis-related processes, mainly regarding amino acid metabolism and aminoacyil-tRNA synthesis. These results suggest a pleiotropic effect of DNAJC17 on several general cellular functions, which may underlie its essential role during early embryogenesis 3 .
Proteomic analysis of DNAJC17 interactome. As DNAJ proteins generally function as co-factors in larger protein complexes 14 , we further characterized DNAJC17 function using a proteomic approach to analyze its protein interactome upon expression of a GFP-tagged protein. To generate a GFP-DNAJC17-expressing cell line, the DNAJC17 cDNA from Sv129 mouse strain was cloned in Flp-In T-REx expression vector in frame with GFP ( Fig. 2A). Such expression vector was transfected into HeLa cells to obtain a cell line exhibiting tetracycline-inducible expression of GFP-DNAJC17 (GFP-DNAJC17 cells). As a control, we also generated stable HeLa cell line with tetracycline (Tet)-inducible expression of GFP (GFP cells). After transfection of each vector, HeLa cells were antibiotic-selected and the pools of colonies obtained were used for further analyses.
Western blot analysis of whole protein extract from control HeLa cells, GFP-DNAJC17 and GFP cells confirmed the expression of either GFP-DNAJC17 or GFP in a tetracycline-dependent manner at both 8 and 16 h of treatment (Fig. 2B). Immunofluorescence analysis at the same time points showed that GFP-DNAJC17 exhibits nuclear localization, similarly to the endogenous DNAJC17 in thyroid cells 3 , whereas GFP alone is detectable both in the nucleus and in the cytoplasm (Fig. 2C). For all further experiments, we used 16 h of Tet treatment.
Total extract from GFP-DNAC17 cells were used to identify DNAJC17 interacting proteins by Co-immunoprecipitation (Co-IP) experiments followed by mass spectrometry analysis. GFP cells were used as negative control in parallel experiments. Total protein extracts from GFP-DNAJC17 and GFP cells were immunoprecipitated by using GFP trap beads, digested with trypsin and analyzed by nanoLC-ESI-LIT-MS/MS. Thus, mass spectrometry analysis identified 70 proteins uniquely observed in the GFP-DNAJC17 sample, when compared with the GFP counterpart (Table S2). These results allowed describing the Interactome Network of DNAJC17, which is shown in Fig. 3. Classification of the interacting proteins by the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg/) identified two general groups of cellular functions related to RNA metabolism: the one involving components related to the spliceosome machinery and the one including species involved in ribosome structure/function.

Validation and in vivo interaction of DNAJC17 with spliceosomal proteins.
To validate the reliability of the ascertained DNAJC17 interactomic data, we selected PRP19, PLRG1, SNRNP200 and XAB2 because their knockout in mouse causes early embryo lethality [15][16][17] (International Mouse Phenotyping Consortium IMPC MP: 0013292 www.mousephenotype.org) like that of DNAJC17 3 . Interestingly, all four selected proteins play a role in pre-mRNA splicing (Table 1). We also selected CDC5L protein, whose knockout is not reported in literature, but it is known to interact with PRP19 e PLRG-1 18 . Relevant informations about the selected interactors are reported in Table 1. Total protein extracts from GFP-DNAC17 cells were immunoprecipitated by using GFP trap beads to capture GFP-DNAJC17-associated protein complexes (for details see Materials and Methods). Immunoprecipitates were then resolved by western blot confirming that DNAJC17 interacts with PRP19, PLRG1, CDC5L, SNRNP200 and XAB2 proteins (Fig. 4).
To test if such interactions take place also in vivo, subcellular co-localization of DNAJC17 with its candidate interactors was then analyzed by immunofluorescence in fixed cells. As it is shown in Fig. 5, co-localization of DNAJC17 with SNRNP200 and PRP19 was more evident respect to that with PRLG-1 and CDC5L. To measure subcellular co-localization, confocal images were analyzed with the Image J software using the JACoP plugin calculating the Pearson's coefficient and following the Coste's approach 19 . Immunofluorescence assay indicated that DNAJC17 co-localizes with PRP19 (mean Pearson's coefficient: 0.805 ± 0.057; Coste's randomization P value: 100%), PLRG1 (mean Pearson's coefficient: 0.739 ± 0.050; Coste's randomization P value: 100%), CDC5L (mean Pearson's coefficient: 0.571 ± 0.037; Coste's randomization P value: 100%), SNRNP200 (mean Pearson's coefficient: 0.651 ± 0.057; Coste's randomization P value: 100%). Due to technical problems, we were unable to confirm the XAB2 co-localization. To further confirm the co-localization of DNAJC17 with two candidate interactors (PLRG-1 and SNRNP200), we performed double immunofluorescence also in parental HeLa cells by detecting the endogenous protein with a specific antibody 3 ( Figure S1). These data confirmed that DNAJC17 shares subcellular compartment with spliceosomal proteins, thus suggesting a possible role of DNAJC17 in modulating spliceosomal activity. DNAJC17 localizes at nuclear speckles. Based on its interaction with PRP19, PLRG1, CDC5L and SNRNP200 and the speckled pattern of nuclear localization (Fig. 5), we tested if DNAJC17 interaction with its partners occurs at nuclear speckles, where all these proteins reside. To this purpose, we checked DNAJC17 co-localization with a specific marker of such compartment, namely SC35, a splicing factor of the serine-rich (SR) family of proteins generally used to label splicing speckles [20][21][22] . Immunofluorescence analysis showed that the nuclear dots of DNAJC17 clearly overlapped with SC35 staining (Fig. 6A), thus demonstrating that DNAJC17 localizes at splicing speckles (mean Pearson's coefficient: 0.648 ± 0.035; Coste's randomization P value: 100%). As PRP19 and CDC5L interact within speckles, we also verified whether DNAJC17 co-localizes with such complex in the same compartment. Double immunofluorescence analysis of GFP-DNAJC17 cells with PRP19 and SC35 antibodies indicated a high co-localization index for the three proteins (mean Pearson's coefficient: 0.722 ± 0.052; Coste's randomization P value: 100%) (Fig. 6B).
DNAJC17 modulates splicing efficiency. The presence of a RRM motif in its structure, as well as its interaction/localization with splicing factors in nuclear speckles, suggested a possible role of DNAJC17 in the modulation of the splicing activity. To test whether DNAJC17 modulates alternative splicing, GFP-DNAJC17, GFP and parental HeLa cells were used to perform in vivo splicing assays using the E1A minigene, a commonly used splicing target containing several 5′ and 3′ alternative splice sites (Fig. 7A) 23 . In GFP-DNAJC17 cells transfected with E1A for 42 h and treated with Tet for 6 h, we observed a perturbation of E1A minigene splicing (Fig. 7B,C). In particular, we observed a reduction of the unspliced pre-mRNA and a concomitant increase in splicing efficiency of the 13 S variant (Fig. 7B,C), which utilizes the proximal 5′ splice site. In contrast, no significant changes of the 12 S and 9 S alternative splice variants were observed (Fig. 7C). These results indicated that DNAJC17 is able to modulate E1A minigene splicing.
To further corroborate the hypothesis of a role of DNAJC17 in splicing, we re-analyzed RNAseq data obtained in DNAJC17 kd cells using a recently published pipeline designed for determination of pre-mRNA splicing efficiency in yeast 24 . This workflow calculates splicing efficiency values separately for the 5′ and 3′ splice junctions of each intron, as the ratio between the number of exon-exon junction-spanning reads (transreads) and the number  of reads covering the first base of 5′ intron end (5′ efficiency) or the last base of 3′ intron end (5′ efficiency) (Table S3). We considered only splicing efficiency changes whose absolute values of ratios were ≥1 (Fig. 8A,B). While the majority of introns show unaltered splicing in DNAJC17 kd cells (Fig. 8A,B grey dots), there are specific genes displaying increased or decreased intron retention (Fig. 8A,B green or red dots, respectively). It is noteworthy that the number of transcripts with increased intron retention (in green) is higher respect to that with decreased retention (in red), and also their extent of variation showed higher values (Table S3). To validate the analysis, we measured by RT-qPCR intron retention ratio of five genes (DONSON, FUBP1, two different introns of SARS, SARS.a and SARS.b, SLC7 and RPS9) showing various degrees of splicing impairment in DNAJC17 kd cells (Table S4). Splicing efficiencies measured by RT-qPCR show similar trends respect to those calculated using RNAseq datasets (Fig. 8C).

Discussion
Notwithstanding previous studies that identified Dnajc17 as a susceptibility gene for congenital hypothyroidism 2 and demonstrated its wide expression in mouse since early embryogenesis, in which it plays an essential role in pre-implantation stages 3 , the molecular function of DNAJC17 protein is still elusive. In this study, a multi-omic approach was undertaken to address this issue, which was complemented with additional localization and functional experiments. In particular, transcriptomic analysis of DNAJC17-depleted cells showed that hampering of Dnajc17 expression affected the level of structural genes involved in gene expression processes, mainly related to tRNA and amino acid modifications. These results suggest that DNAJC17 may be involved in one (or more) gene expression-related cellular function(s). On the other hand, proteomic experiments aimed at defining the DNAJC17 protein interactome identified 70 proteins as selectively interacting with DNAJC17. Most of these interactors fall into two protein functional groups: i) spliceosome components/interactors; ii) ribosome structural proteins. Some of these candidates were validated by independent co-IP/western blotting experiments. Both identified protein networks are involved in RNA metabolism, suggesting once again that DNAJC17 may be implicated in gene expression. The presence in the DNAJC17 protein interactome of a huge number of factors involved in splicing-related processes suggested that DNAJC17, similarly to its ortholog Cwc23 in yeast 9 , may play a role in pre-mRNA splicing. Interestingly, several candidate interactors reported in this study have already been demonstrated to play an essential role in mouse development, similarly to DNAJC17 3 . This is the case of 4 DNAJC17-binding partners, namely PRLG1, PRP19, SNRNP200 and XAB2. Among that, PRLG1 and PRP19, together with another protein identified as a DNAJC17 interactor, namely CDC5L, have been demonstrated forming a complex that plays a key role in catalytic activation of the spliceosome 18,25 . Furthermore, knock-out mice for Prlg1 and Prp19 showed embryo lethality at the first cell division and blastocyst stage, respectively 15,16 . The interaction and co-localization of DNAJC17 with CDC5L, PRLG1 and PRP19 was confirmed by both co-IP and immunofluorescence assays. Such interaction with the core of the PRP19/CDC5L complex suggested a possible involvement of DNAJC17 in the correct formation and/or function of the latter molecular machinery. A similar function was also hypotesized for SNRNP200, a U5 small nuclear ribonucleoprotein (snRNP) helicase that is required for spliceosome assembly, and whose depletion blocks the second step of splicing 26 . On the other hand, it was observed that the homozygous knockout of SNRNP200 causes embryonic lethality before implantation (International Mouse Phenotyping Consortium IMPC MP: 0013292 www.mousephenotype.org).
Based on the ascertained interaction of DNAJC17 with several proteins involved in splicing, we further hypotesized that such interaction could take place within nuclear speckles. These are indeed discrete nuclear domains where pre-messenger RNA, splicing factors, snRNP particles, spliceosome subunits and non-snRNP protein splicing factors accumulate. In particular, it is well accepted that nuclear speckles act as storage compartments able supplying splicing factors to active transcription sites [20][21][22] . In this study, we were able to demonstrate that DNAJC17 co-localizes with SC35, a well-established marker of localization in nuclear speckles, thus demonstrating that this chaperone is present in these subnuclear structures.
Localization in nuclear speckles, together with the association with many splicing proteins, was strongly suggestive for a role of DNAJC17 in splicing, likely to its yeast ortholog Cwc23 9 , whose function was hypothesized based on experiments with gene deletion mutants. In agreement with this hypothesis, E1A minigene assays demonstrated that DNAJC17 overexpression in HeLa cells was able to modulate splicing efficiency. Moreover, by re-analyzing DNAJC17 kd RNAseq data trough a workflow specifically designed for splicing efficiency analysis, we highlighted few hundreds of introns differentially retained in DNAJC17 kd respect to DNAJC17 wt cells. Hence, our data strongly suggest that DNAJC17 plays a modulatory role in splicing mechanisms, which hence appears to be evolutionary conserved moving from yeast to human. Molecular chaperones have already been shown to control splicing processes, both in yeast 9 and in human cells, although in these latters only chaperones controlling cytoplasmic steps of spliceosome assembly have been identified 27 . In this paper we report the first evidence of interaction of a HSP40 protein with spliceosomal core components in the nucleus.
It is worth noting that homozygous truncating mutations of DNAJC17 in humans segregate with a novel syndromic form of retinitis pigmentosa (RP), which is associated with hypergammaglobulinemia 4 . RP is a genetically heterogeneous disorder, however recent evidences have pointed out that splicing defects due to mutations of several different genes are at the basis of the pathogenesis of such disease 28 .
Taken together, our data represent the first functional characterization of human DNAJC17; they strongly suggest that this protein is involved in splicing events, although its exact role in this process remains to be defined. The identification of numerous splicing-related interactors will allow to test if DNAJC17 is able to modulate their activity by acting as a chaperone. Thus, further studies are required to establish whether DNAJC17 plays a role in general or specific alternative splicing events. Nevertheless, its involvement in such a crucial phenomenon should hypothetically explain the embryo lethality in DNAJC17 knockout mouse as well as the association of different mutations in Dnajc17 to different complex disorders in humans 4,5,7 .

Methods
Generation of Stable HeLa Cell Lines. DNAJC17 cDNA from Sv129 mouse strain previously cloned 3 was amplified by PCR with primers containing BamHI and XhoI restriction sites (DnajC17 Bamh1Fw: 5′gagctcggatc-catggcggtgaccaaagagctctt3′; DnajC17Xho1Rev: 5′ctctagactcgagctacgtgggccgcccctc3′). The resulting PCR product was digested and cloned into a modified pCDNA5/FRT/TO (Invitrogen, BioSource International, USA) in which the GFP coding sequence was cloned upstream the multiple cloning sites 29 . The construct was sequenced to verify the absence of mutations. pCDNA5/FRT/TO containing GFP or chimeric GFP-DNAJC17 was transfected in Flp-In T-REx HeLa cells (Invitrogen, BioSource International, USA) by using the Effectene Transfection Reagent (Qiagen, Milan, Italy) and the positive cells were selected according to the manufacturer's instructions (Invitrogen K6500-01). After the selection, GFP or GFP-DNAJC17 expression was induced with tetracycline, and fluorescent cells were sorted using a MoFlo sorter (Beckman Coulter, Milano, Italy). GFP or GFP-DNAJC17 fusion protein expression was tested by western blot before and after tetracycline induction. RNA sequencing. Differential gene expression was analysed in DNAJC17-interfered HeLa cells by a RNA sequencing (RNA-Seq) approach. RNA was extracted from 3 independent samples for both knockdown and wild type conditions; it was sequenced on HiSeq. 1500 (Illumina, San Diego, CA, USA) according to manufacturer's protocols. About 7 × 10 7 high quality sequence reads were produced for sample with a mean quality score ≥30 (99.9% base quality accuracy). The human Ensembl GRCh37 genome was used as reference for read mapping, which was performed by TopHat aligner 11 . Properly mapped reads (about 5 × 10 7 ) were further processed using Cufflinks package 12 . Selected reads were assembled into transcripts, which were compared among the six samples and with known annotations to generate a unique final transcriptome. Resulting 20825 genes were then tested for differential expression between DNAJC17 kd vs DNAJC17 wt and filtered for corrected p-value (FDR) ≤0.05 and absolute fold change ≥1.5 (884 significantly deregulated genes; 360 up-regulated and 524 down-regulated genes).
RNAseq aligned reads were further analyzed to calculate the splicing efficiency in the two conditions, using the method described by Prevorovsky et al. 24 . Briefly, the splice sites were predicted through the analysis of the reads spanning the exons junctions (setting the minimum and maximum intron length to 20 and 11000 bases, respectively). Then, we calculated the splicing efficiency for each single intron and separately for the 5′ and 3′ splice site, as the ratio between the number of spanning reads and the number of reads covering the first base of 5′ intron end (5′ efficiency) or the last base of 3′ intron end (5′ efficiency).
Immunoblotting. Whole-cell lysates of HeLa, GFP-DNAJC17 and GFP cells were prepared in sample buffer (1% w/v Triton X-100, 0.1% w/v SDS, 0.50% w/v sodium deoxycholate, 50 mM TrisHCl pH 8.5X mM MgCl 2 , 150 mM NaCl, 1 mM DTT, 0.5 PMSF, X w/v Protease inhibitor cocktail, 50 mM sodium fluoride, 0.5 mM sodium pyrophosphate and 0.5 mM sodium orthovanadate) and normalized for protein concentration by Pierce BCA Protein assay kit (Thermo Scientific, Milano, Italy). A 20 μg amount of protein samples was resolved on a precast NuPAGE 4-12% Bis-Tris Gel (Life Technologies, Monza, Italy) and transferred on a polyvinylidene difluoride (PVDF) membrane (Millipore, Milano, Italy). Nonspecific binding sites were blocked by incubation with 5% w/v non-fat dry milk in TBS (20 mM TrisHCl pH 7.6, 140 mM NaCl) containing 0.1% w/v Tween 20. Immunodetection was performed by using one of the following primary antibodies: DNAJC17 home-made rabbit polyclonal antibodies 3 ; GFP monoclonal antibody (Chromotek, Planegg-Martinsried, Germany); PRP19 polyclonal antibody (LifeSpan BioSciences, Seattle, WA, USA); CDC5L monoclonal antibody (Santa Cruz Biotechnology, Dallas, Usa); PLRG1 polyclonal antibody (Novus Biologicals, Littleton, CO, USA); HELIC2  4) were untreated or treated with Tet for 6 h. Splicing products are indicated on the right of the panel. (C) Quantification of the major E1A mRNA variants: white bars and black bars represent the percentage of each isoform in GFP and GFP-DNAJC17 cells, respectively, either untreated or treated with Tet for 6 h. Data represent the mean ± SD of three independent experiments. Figure 8. Splicing efficiency in DNAJC17 kd . The scatter plots represents for each intron 5′ (A) and 3′ (B) splicing efficiency in DNAJC17 kd (x axis) and in the control DNAJC17 wt (y axis), as the mean across the three replicates. Considering a cutoff of absolute values ≥1, introns whose splicing efficiency is reduced following DNAJC17 silencing are labeled in green; instead red indicates increased efficiency compared with control. (C) Comparison of relative splicing efficiencies at the 5′ and 3′ ends of selected introns calculated from the RNA-seq data with corresponding splicing efficiencies as determined by RT-qPCR (means of 3 independent RT-qPCR experiments ± SD in triplicates). For each intron, ratio between exon-exon junction transreads and intron 5′ or 3′ ends in DNAJC17 kd cells are reported relative to DNAJC17 wt control. of RT-PCR reactions were run on agarose gel and digital images were captured and analyzed by ChemiDoc XRS with ImageLab software (Bio-Rad, Hercules, CA, USA).
Data availability statement. The datasets and reagents generated and/or analysed during the current study are available from the corresponding authors on reasonable request.