Defects of the spliceosomal gene SNRPB affect osteo‐ and chondro‐differentiation

Although gene splicing occurs throughout the body, the phenotype of spliceosomal defects is largely limited to specific tissues. Cerebro‐costo‐mandibular syndrome (CCMS) is one such spliceosomal disease, which presents as congenital skeletal dysmorphism and is caused by mutations of SNRPB gene encoding Small Nuclear Ribonucleoprotein Polypeptides B/B′ (SmB/B′). This study employed in vitro cell cultures to monitor osteo‐ and chondro‐differentiation and examined the role of SmB/B′ in the differentiation process. We found that low levels of SmB/B′ by knockdown or mutations of SNRPB led to suppressed osteodifferentiation in Saos‐2 osteoprogenitor‐like cells, which was accompanied by affected splicing of Dlx5. On the other hand, low SmB/B′ led to promoted chondrogenesis in HEPM mesenchymal stem cells. Consistent with other reports, osteogenesis was promoted by the Wnt/β‐catenin pathway activator and suppressed by Wnt and BMP blockers, whereas chondrogenesis was promoted by Wnt inhibitors. Suppressed osteogenic markers by SNRPB knockdown were partly rescued by Wnt/β‐catenin pathway activation. Reporter analysis revealed that suppression of SNRPB results in attenuated Wnt pathway and/or enhanced BMP pathway activities. SNRPB knockdown altered splicing of TCF7L2 which impacts Wnt/β‐catenin pathway activities. This work helps unravel the mechanism underlying CCMS whereby reduced expression of spliceosomal proteins causes skeletal phenotypes.


Introduction
Cerebro-costo-mandibular syndrome (CCMS) is a congenital skeletal dysmorphism comprising micrognathia, cleft palate, glossoptosis, rib gaps, scoliosis, a narrow chest and subsequent upper airway obstruction [1][2][3].The genetic origin of CCMS was identified as mutations within the SNRPB gene, encoding Small Nuclear Ribonucleoprotein Polypeptides B and B 0 (SmB/B 0 ), which lead to reduced expression levels of functional SmB/B 0 proteins [4,5].SmB/B 0 is one of the core components of the major spliceosome particles [6].
splicing cofactors to regulate splicing [6].Most human genes undergo splicing, including alternative splicing which produces multiple isoforms [7].Precise and context-dependent splicing is crucial for normal embryonic development as well as in adult life [8][9][10].
Despite the ubiquitous requirement of splicing in the human body, the phenotype of spliceosomal defects is very specific and highly localised [11][12][13][14][15].Besides CCMS, other examples of spliceosomal defects include Nager syndrome, mandibulofacial dysostosis with microcephaly (MFDM; also categorised as Guion-Almeida type, MFDGA) and Burn-McKeown syndrome, which are caused by mutations of SF3B4 (encoding SAP49, a component of U2), EFTUD2 (encoding SNRP116, a component of U5) and TXNL4A (encoding one of U5 snRNP proteins), respectively [16][17][18], and some cases of retinitis pigmentosa caused by mutations in various snRNPs [13].Elucidating the mechanisms for the regional specificity is of medical significance because it will empower our understanding of the complexities of gene transcription and phenotype presentation.
As the spliceosome is a large complex of snRNPs and associated factors, defects in a single component do not appear to disrupt all splicing events.Rather, they result in changes in the efficacy of splicing of only a limited set of genes.Analysis of SNRPB-knockdown in HeLa cells revealed a reduction in the inclusion of alternative exons in certain genes, while the splicing of constitutive exons and alternative splicing of other genes remained unaffected [19].Such biased alteration of splicing may therefore be the key to understanding the mechanism of the tissue-specific phenotype of spliceosomal defects.
Here we explored the impact of reduced SNRPB expression on osteo-and chondro-differentiation in vitro by monitoring the expression of differentiation genes.We report that reduced SNRPB suppresses osteogenesis and promotes chondrogenesis in vitro.We identified altered splicing in Dlx5 and TCF7L2 as a part of the possible cause of the cell differentiation defect.Furthermore, reduced SNRPB causes suppressed sensitivity to Wnt signals and facilitated BMP pathway activation.These results help us understand how spliceosomal defects lead to skeletal defects.

Osteoblast differentiation is affected by SNRPBreduced Saos-2 cells
We first examined the effect of SNRPB-knockdown on osteogenesis, using a human osteogenic progenitor-like cell line Saos-2.Saos-2 cells differentiate into osteoblasts following induction with ascorbic acid and bglycerophosphate [42].Differentiation accompanies the increased expression of ALP [43], eventually resulting in calcium deposition upon achievement of the terminal differentiation state [44].Following induction, control siRNA transfected cells presented a rapid increase of ALP expression and, on day 15, successful calcium deposition, whereas SNRPB-knockdown caused a slow and modest increase of ALP and attenuated calcium deposition (Fig. 1A-E; Fig. S1).A repeat of siRNA transfection on day 0 (the day of induction commencement), day 5 and day 10, effectively reduced ALP secretion and calcium deposition (Fig. 1A).Applying siRNA only on day 5 and/or 10 was not as effective (Fig. 1B,C), whereas transfection of siRNA only on day 0 sufficiently suppressed the ALP increase and calcium deposition similarly to that on day 0 and day 7 (Fig. 1D,E), suggesting that SNRPB expression is most required at the initial stage of osteodifferentiation in Saos-2 cells.
The effect of reduced SmB/B 0 was confirmed by CRISPR-mediated genomic mutation, in which the first exon-intron boundary of SNRPB was targeted for splicing disruption, resulting in the retention of intron 1 including a premature termination codon and subsequent nonsense-mediated mRNA decay (NMD) (Fig. S2A-D).Thus, both siRNA-mediated knockdown and CRISPR-mediated mutation cause reduced SmB/B 0 protein expression, either transiently or stably like CCSM cases, respectively.Such SNRPB-mutated Saos-2 cells showed reduced ALP secretion and calcium deposition (Fig. 1F).From day 11 onwards, when the level of ALP began to decline in control cells, SNRPB-mutated cells demonstrated slightly higher ALP secretion than the control (Fig. 1F).This may reflect a delayed peak of ALP induction in SNRPB-mutated cells (Fig. S3C, see below).
Osteoprogenitors cease proliferation by exiting the cell cycle when differentiating into osteoblasts [45].As the cell proliferation kinetics would affect the secretion of ALP, the viable cell counting assay was conducted in SNRPB knockdown-and mutated-Saos-2 cells and compared with controls during the differentiation (Fig. 1G,H).SNRPB-deficient groups showed fewer viable cells compared to the control at the early stage of differentiation (5-8 days after siRNA transfection and 3-5 days in SNRPB-mutated cells after induction for differentiation).Together with the result of low ALP secretion (Fig. 1A-F), it was speculated that low levels of SNRPB expression impact the growth of Saos-2 cells.On the other hand, at the late differentiation stage (15 days after siRNA transfection and 8 days in SNRPB-mutated cells after induction for differentiation), there were more viable cells in SNRPB-deficient groups compared to controls (Fig. 1G,H).This suggests that SNRPB-deficient cells failed to exit the cell cycle and retained the cell proliferating ability longer than the control cells, suggesting a defect in osteodifferentiation.
We next investigated the expression of osteogenic genes in Saos-2 cells; lineage commitment genes Dlx5, Runx2 and Msx2 [20,21,46] and functional differentiation markers ALP and Osterix [24,25] (Fig. 2; Fig. S3).In both SNRPB-knockdown and mutated Saos-2 cells, Dlx5 transcription was initially higher than the control on day 3, yet was lower on the following days, whereas Runx2 was low from day 3 onwards (Fig. 2A,C).A noticeable difference between the transient knockdown by siRNA and stable mutation was a high ALP on day 8 in stably mutated cells (Fig. 2C).Normalising all data to day 3's control samples revealed the time-course change in gene expression (Fig. S3), in which control cells showed the peak of ALP transcription on day 3, whereas SNRPB-mutated cells presented a delayed peak on day 8.The shift of the transcript peak likely reflects higher ALP secretion in SNRPB-mutated cells compared to the control on day 11 (Fig. 1F).Another difference between the transient knockdown and stable mutation of SNRPB was on Msx2, in which downregulation on day 5 either recovered or persisted, respectively, by unknown reasons (Fig. 2A,C).The differentiation marker Osterix, which is induced in the second half of the differentiation phase, showed suppressed expression in both knockdown and mutated cells (Fig. 2A,C), consistent with the low calcium deposition seen in Fig. 1.Western blotting confirmed reduced Dlx5 by SNRPB knockdown on day 3 (Fig. 2B), whereas in SNRPB mutated cells downregulation of Dlx5 was more prominent on day 8 (Fig. 2D).Reduction of Runx2 was most prominent in SNRPB mutated cells on day 3. Reduced SmB/B 0 was also confirmed on day 3 after siRNA transfection and not on day 8, and consistent low expression in mutated Saos-2 cells (Fig. 2B,D).On qPCR, the suppressed expression of Dlx5 and Runx2 by SNRPB siRNA was partly rescued on day 8 by SNRPB DNA transfection (Fig. 2E,F).A similar result was obtained in western blots, where reduced Dlx5 by SNRPB siRNA was rescued by SNRPB DNA transfection (Fig. 2H).
The result that the initial increase of Dlx5 transcripts on day 3 was not accompanied by any increase in Runx2 in both SNRPB knockdown and mutant cells (Fig. 2A,C) suggested that the endogenously expressed Dlx5 was defective in SNRPB-deficient cells, as Dlx5 would normally upregulate Rnux2 [20,23] (see also control siRNA in Fig. 2G) and establish positive feedback loop [22] to induce ALP [24].Indeed, exogenous transfection of Dlx5 rescued the expression level of Runx2 in SNRPB-knockdown cells (Fig. 2G).We hence examined whether the Dlx5 transcript was affected in SNRPB-knockdown cells.It has been reported that Dlx5 may be alternatively spliced with retention of the 2nd intron, which leads to NMD (GenBank accession KAI2546789).qPCR analysis showed higher levels of  (A-E) Knockdown of SNRPB by siRNA transfection in Saos-2 cells.The left graphs show alkaline phosphatase (ALP) activity in the conditioned medium during Saos-2 cell differentiation following transfection of control (black) or SNRPB (orange) siRNA.The relevant siRNA was transfected in sextuplicate on the day(s) indicated by the circle on the xaxis.Every time siRNA was transfected, the conditioned medium was changed to the fresh medium.In (D), the medium was changed on day 7 after the measurement without transfection.On day 15, all samples were stained with Alizarin red (AZ) to reveal calcium deposition.The right bar charts show the quantification of AZ staining.(F) CRISPR-mediated SNRPB mutation in Saos-2 cells, analysed for ALP activity and calcium deposition as above.The medium was changed every time after sampling on days 3, 5, 8 and 11.Mutation of SNRPB and downregulation of SmB/B 0 expression are shown in Fig. S2.The scale bar for the AZ staining is 500 lm.(G, H) Viable cell counting assay following siRNA transfection (G) and CRISPR-mutated Saos-2 cells (H).Graphs show the mean AE SD; n = 6 for ALP from sextuplicate samples and n = 12 for Alizarin red as two non-overlapping images were taken from each well in (A-F); n = 6 for the cell viability assay in (G, H).In AZ quantification and cell viability assays, data were normalised to the average of control samples.*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test, by comparison with the relevant control on each day.
the intron-retained form of Dlx5 in both SNRPBknockdown-and CRISPR-mutated-cells when compared to the control groups (Fig. S4A,B).Non-quantitative PCR using primers spanning exons 2 and 3 showed retention of the 2nd intron in SNRPBknockdown cells (Fig. S4C), supporting the intron retention.On western blots, however, the truncated form was undetectable by antibodies targeting the 1st exon-encoding domain (Fig. S4D,E), suggesting that the intron-retained form undergoes NMD or remains below the detection level.The result suggests that impaired splicing of Dlx5, at least in part, accounts for the attenuated osteo-differentiation in reduced SmB/B 0 cells.

Accelerated chondrogenesis in SNRPB-reduced HEPM cells
Next, the effect of reduced SmB/B 0 was examined on chondrogenesis.HEPM cells undergo chondrogenesis through a similar mechanism to other mesenchymal stem cells in high-density three-dimensional (3D) aggregate cultures with TGFb3 and dexamethasone [47] via induction of transcription factors Dlx5/6 and Sox9 and structural proteins Collagen type 2A1 (Col2A1) and Aggrecan [29,48].Dlx5 and Dlx6 are functionally redundant in chondrogenesis [26] and work independently of Sox9 [28].These markers were assessed in HEPM cell aggregates transfected with either siRNA or sgRNA targeting SNRPB.To note, a stable HEPM cell line with CRISPR-mediated SNRPB mutation was unattainable, as mutated cells ceased proliferation after several passages.This might be due to the mechanism similar to the heterozygous lethality of SNRPB knock-out mouse embryos [49].Therefore, HEMP cells were transfected with sgRNA and Cas9 1 day before induction for differentiation.The mutation was confirmed by detecting functional and defective SNRPB transcripts (Fig. S2E-I).
As shown in Fig. 3, chondrogenic markers Dlx5/6, Col2A1 and Aggrecan tended to be expressed at higher levels in HEPM aggregates transfected with SNRPB siRNA or sgRNA (Fig. 3A,B,D,E).Dlx5 showed increased expression from day 2 and day 4 in sgRNA and siRNA transfected aggregates, respectively, whereas Col2A1 and Aggrecan showed higher levels from day 4 onwards.Control aggregates showed the expression peak of both Col2A1 and Aggrecan on day 8, whereas those with SNRPB siRNA or sgRNA showed a continued increase of Aggrecan until day 11 (Fig. 3E).Increased Collagen II expression was confirmed immunohistochemically (Fig. 3F,G).In contrast, the difference in Sox9 expression between control and SmB/ B 0 -deficient aggregates was not consistent or significant (Fig. 3C).Taken together, during the chondrogenic differentiation in HEPM cell aggregates, reduced expression of SmB/B 0 causes promoted and possibly prolonged chondrogenesis with elevated Dlx5/6 expression followed by high levels of Col2A1 and Aggrecan, while the changes in Sox9 was not consistent.
We next examined the impact of the Wnt pathway in chondrogenesis in HEMP cells using two chemical Wnt pathway blockers with different modes of action; IWR1 as mentioned above [61] and PNU-74654 (PNU) which interferes with b-catenin-dependent transcription [63].As reported in other chondrogenic conditions [57,58,64,65], the Wnt pathway blockers generally promoted chondrogenesis in the HEPM aggregate assay (Fig. 4I-O).Application of the Wnt p activator BIO, on the other hand, blocked the aggregation formation in the chondrogenic medium, hence the tissue samples were not collectable.These results were consistent with previous reports that Wnt pathway activation disfavours chondrogenesis in mesenchymal stem/progenitor cells [64,[66][67][68][69].

Low response to Wnt signals in SNRPBknockdown cells
Next, the effect of reduced SmB/B 0 on Wnt and BMP pathway activities was examined using well-established reporters; the Wnt pathway reporter is driven by the b-catenin and TCF/LEF complex which binds the conserved TCF/LEF-binding sequence upon pathway activation [70], whereas the BMP pathway reporter is driven by nuclear-translocated Smad4 [71,72].
Saos-2 cells with knockdown or mutation of SNRPB showed reduced responses to Wnt pathway activators Wnt3a and BIO, while reporter activation by b-catenin transfection did not change significantly (Fig. 5A,C).With regard to the BMP pathway reporter, enhanced BMP pathway activities were observed in SNRPB-mutated Saos-2 cells (Fig. 5D) in response to BMP ligands or constitutively active BMP receptor IA [73] but no significant changes were observed in siRNA-mediated knockdown (Fig. 5B).When BMP4 protein was provided for the immediate activation instead of plasmid transfection, an enhanced response was seen, although it was not statistically significant (Fig. 5B).HEPM cells were refractory to exogenously transfected Wnt3a, b-catenin or a constitutively active form of LRP6 (LRP6DN) [74], nonetheless, they were responsive to BIO and another GSK3 inhibitor CHIR-99021 (CHIR) [75] and showed attenuated responses following SNRPB-knockdown similarly to Saos-2 cells (Fig. 5E).The effect of SNRPBknockdown on the BMP pathway was not significant like Saos-2, however, the application of BMP4 protein caused a significant enhancement (Fig. 5F).
These results demonstrate that reduced SmB/B 0 potentially leads to suppressed Wnt pathway activities in Saos-2 and HEPM cells, likely at the level downstream of GSK3b, and enhanced BMP pathway activity in Saos-2 cells where SNRPB is genetically mutated.
Changes in the splice pattern of TCF/LEF by SNRPB-knockdown SNRPB-knockdown is known to cause reduced efficiency in the inclusion of certain alternative exons in specific genes [19].As SNRPB-knockdown is phenocopied by Wnt pathway blockers in Saos-2 and HEPM cells (Fig. 4) and the pathway suppression appears at the level downstream of GSK3b (Fig. 5), a defect was suspected at the transcriptional level of the Wnt pathway.We therefore examined TCF/LEF transcriptional factors that splicing patterns vary depending on the cell type, developmental stage and pathological condition [76][77][78][79].In the absence of b-catenin, TCF/LEF forms a complex with other co-repressors such as Groucho and works as a transcriptional repressor.Upon the Wnt pathway activation, TCF/LEF works as a transcriptional activator by recruiting b-catenin at their N 0 terminus [76].TCF/LEF isoforms created by a combination of alternative splicing are the inclusion or exclusion of; exon 4 and LVPQ/SxxSS motifs in TCF7L2, the context-dependent regulatory domain (CRD) of TCF7 and LEF1, and the C-terminal binding protein (CtBP) binding sites on the C-terminus (C 0 ) tail of TCF7 and TCF7L2 (Fig. S5), that serve to recruit co-repressors [76][77][78][79][80][81][82][83][84][85].Besides, exons 1-3 are either included or excluded by the alternative promoter activity, leading to N+ (containing the b-catenin-binding domain) and NÀ (lacking it) forms, respectively.Based on these, LEF1, TCF7 and TCF7L2 isoforms were analysed by PCR following SNRPB-knockdown in Saos-2 and HEPM cells (Fig. 6).
In Saos-2 cells, SNRPB knockdown resulted in reduced expression of LEF1 and TCF7 in all tested isoforms including the total transcripts (Fig. 6A), suggesting that specific splicing modulation was not likely by SNRPB knockdown.As LEF1 and TCF7 are both targets of the Wnt pathway [86,87], the reduced transcription may be a secondary effect of SNRPB knockdown affecting the endogenous Wnt pathway.In contrast, in TCF7L2, the reduction of the exon 4 included form (Ex 4+) was detected by qPCR without a significant change in the total TCF7L2 transcripts (Fig. 6C).Although, non-quantitative PCR using primers across the exons 1 to 5 did not show obvious change and rather revealed that Ex4+ was the minority (Fig. 6B left).The splicing at C 0 (splicing patterns and the detecting primers P1R to P4R are explained in Fig. S5) [82,88] did not show a significant change (Fig. 6B right, C).
In HEPM chondrogenic aggregates, no significant changes in LEF1 and TCF7 were seen by qPCR (Fig. 6D), and TCF7 N+ form was not detected.For TCF7L2, changes were not detected in Exon 4 (Fig. 6E left, F), but were found in the LVPQ-SxxSS motifs on qPCR (Fig. 6F), suggestive of weak transcriptional activity [84,85].Regarding the C 0 , two primer sets, P3R and P4R showed significant differences (Fig. 6F).Although, P3R detects both short C 0 (DC) and full C 0 transcripts, that are 327 and 254 bp in the non-quantitative PCR (Fig. 6E right), but the 254 bp band is a mix of a DC and full transcripts (Fig. S5), therefore, it was not possible to identify the resultant transcript changes.Nonetheless, the primer P4R detected a reduction in full C 0 forms (Fig. 6F).The P2R was not detected by qPCR.As such, the splicing pattern of TCF/LEF as well as the effect of SNRPB knockdown varies between Saos-2 and HEPM cells.
To further examine the cell-type difference in the effect of SNRPB knockdown, HEK293 cells were employed based on the high sensitivity to Wnt and BMP signals.In reporter assays using HEK293 cells, SNRPB knockdown showed suppressed Wnt pathway activation by Wnt3A and BIO, similar to Saos-2 and HEPM cells, however, exogenous b-catenin resulted in further activation of the reporter (Fig. S6A,C).Regarding the BMP pathway, there was no effect of SNRPB knockdown 1 day after the activator transfection, however, significant enhancement was seen 2 days after (Fig. S6B,D).This was reminiscent of Saos-2 cells in Fig. 5, where siRNA-mediated knockdown showed no effect whereas genetic mutation had an enhancing effect.The splicing pattern of TCF/ LEF was different from Saos-2 and HEPM cells (Fig. S6E-G), among which, noticeable was the reduction of TCF7L2 Ex4+ detectable in both non-qPCR and qPCR (Fig. S6F left, G).Splicing changes at the C 0 terminus of TCFL2 by SNRPB knockdown were also in favour of acting as a strong activator (Fig. S6F right, G), which explains the enhanced reporter activity by exogenous b-catenin (Fig. S6A,C).The data demonstrate that down-regulation of SNRPB results in changes in the splicing pattern of TCF7L2 depending on the cell type, which may contribute to altering the Wnt pathway at the transcription level.
Overall, our results suggest that reduced SNRPB expression causes the altered splicing pattern of specific genes, such as TCF7L2 resulting in attenuated Wnt pathway activities, and together with other affected genes such as Dlx5, leads to affected osteoand chondro-differentiation.

Discussion
Both osteogenesis and chondrogenesis are tightly regulated by exogenous signals and transcriptional regulation at multiple steps.Our results demonstrate that the SNRPB-knockdown impacts the differentiation process, which is, at least in part, explained by the affected Wnt pathway and other genes' altered splicing including Dlx5.There are limitations in the study of cell differentiation using in vitro models: Saos-2 cells are osteoprogenitor-like, therefore, they are already lineage-restricted.HEPM cells have a wider differentiation capacity [89], however, while they are responsive to Wnt inhibitors and readily upregulate chondrogenic markers (Fig. 4I-O), they are relatively refractory to additional Wnt signals (Fig. 5E), hence, the differentiation capacity might be biased, i.e., a weak suppression caused by SNRPB knockdown might have given a large impact on chondrodifferentiation.Besides, the splicing pattern of TCF7L2 following SNRPB knockdown was considerably different between Saos-2, HEPM and HEK293 cells (Fig. 6; Fig. S6E-G).In addition, the isoforms of TCF7L2 are very complex [80,82,90]; for example, the impact of including the SxxSS motif resulting in reduced transcriptional activity is much more prominent in the DC form than the full-C form, whereas the impact of excluding exon 4 is little in the presence of LVPQ-SxxSS motifs [82].The expression of co-repressors would further add complexity and cell-type differences.Hence, the effect of splice changes in TCF7L2 by reduced SNRPB expression varies considerably depending on the cell type, and the mechanism of reduced SNRPB expression in suppressing osteogenesis and accelerating chondrogenesis is yet to be fully clarified.
In CCMS patients, the skeletal defects are highly region-specific such as the mandible and the proximal part of the ribs, as such, osteo/chondrogenesis is largely unaffected in most bones.This suggests that the effect of reduced SmB/B 0 expression can be compensated in most regions during embryogenesis.We suspect that the phenotype is manifested only when multiple developmental steps are affected in the same cell lineage.One major phenotype of CCMS is dysmorphism in the maxilla and mandible.The craniofacial morphogenesis requires precisely regulated Wnt and BMP signals prior to the onset of skeletal differentiation, such as the balanced cell proliferation, cell type commitment and differentiation in pharyngeal arches [34,91,92].As such, although this study has focused on the differentiation of chondrogenesis and osteogenesis lineages, defects in SNRPB expression likely affect other developmental processes, either via affected Wnt/BMP signals or altering other genes' splicing.
The formation of rib gaps, another major phenotype of CCMS, is unique to the syndrome.Postmortem studies of CCMS cases revealed that the rib gap space was filled by fibrovascular tissues with or without cartilage tissues [93][94][95][96], which appears to suggest that a failure is possible in both chondrogenesis and osteogenesis.The localised defect within the ribs further suggests the primary defect at much earlier stages of development, for example during somite patterning in which progenitors of vertebrae and ribs arise by balanced signals including Wnts and BMPs [97].Indeed, local application of BMP4 protein or Wnt inhibitors in developing somites affects somite patterning and results in proximal rib defects resembling CCMS phenotype in chick embryos [98,99].It should be noted that the expression of SmB/B 0 is ubiquitous at the early (gastrula) stages of embryogenesis and becomes region-specific later as tissues and organs differentiate [100].Cartilage primordia express SmB/B 0 in the whole body without clear regional differences [100], once again suggesting additional mechanisms involved in the phenotype presentation.Interestingly, knockout of Dlx5 in mouse embryos results in skeletal defects in the maxilla and mandible [101].Furthermore, double-knockout of Dlx5 and Dlx6 causes proximal rib defects at the position similar to CMMS rib gaps [28], suggesting that the regions affected in CCMS patients are susceptible to the shortage of Dlx5.This may partly explain why the entire skeletal structures are not affected in CCMS patients.Recent transcriptome analysis of conditional deletion of SNRPB in mouse neural crest cells identified differentially spliced transcripts required for craniofacial development [49].These studies support the different degrees of SNRPB requirements in different genes and cell types.
Another issue about tissue-specific phenotypes is the selectivity of affected genes among spliceosomal syndromes.In the analysis of Burn-McKeown Syndrome, it was reported that the inclusion of exon 4 in TCF7L2 transcripts is increased in the patient-derived induced neural crest cells [102].Although in our SNRPB knockdown, Ex4+ was either decreased (Saos-2, HEK293) or no significant change (HEPM) in the tested cell lines, it seems that the splicing of exon 4 of TCF7L2 is susceptible to spliceosomal defects.The resulting phenotype may also depend on the cellular context, that is, how crucially TCF7L2 is responsible for the b-catenin-dependent transcription, i.e., how much other TCF/LEF are available in the cell.It is worth noting that TCF7L2 is predominantly expressed in pharyngeal arches [103], which might make TCF7L2 crucial for the craniofacial phenotype.In the study of another spliceosomal syndrome, Nager syndrome, the patient's cartilage sample expressed lower levels of Dlx5/6 and Sox9 [104].Although in our studies the changes in Sox9 expression in HEPM cells were not as prominent, Dlx5/6 appears susceptible commonly in SNRPB and SF3B4 spliceosomal defects.As such, emerging results show that certain specific genes are more susceptible to spliceosomal defects than others, which may be one of the reasons why common craniofacial and skeletal phenotypes are seen among different spliceosomal syndromes [14].It is noteworthy that mutations of several different spliceosomal genes lead to retinitis pigmentosa in common [13], which is another example of gene specificity and tissue selectivity in the susceptibility to spliceosomal defects.Together these findings help elucidate the mechanism whereby the impact of spliceosomal defects is genespecific and cell-type dependent, which causes localised and specific phenotypes.
For osteodifferentiation analyses in Saos-2 cells, cells were prepared in sextuplicate in 24-well plates, and a 15 lL of the medium was taken out of 500 lL from each well (n = 6).ALP activity was measured using paranitrophenylphosphate (N2770; Sigma-Aldrich) following the manufacturer's protocol, using the standard curve made on the day by serial dilutions of ALP (Promega, Madison, WI, USA) in DMEM.Three readings were averaged from each well's medium (technical triplicates).Data were analysed by unpaired equal variance t-test.On day 15, cells were fixed with 4% paraformaldehyde in PBS and stained with 0.5% Alizarin red (Merck).The staining intensity was quantified using IMAGEJ (National Institute of Health, Bethesda, MD, USA) at the fixed colour threshold.Two photos were taken at the non-overlapping area of each of the sextuplicate wells (n = 12) and the average of 12 was normalised to counteracting control groups.For viable cell counting, Cell Counting Kit-8 (96992; Merck) was used.The same number of control and SNRPB-deficient cells (either siRNA transfected or mutated) were seeded in sextuplicate and incubated with the reagent for 2 h at 37 °C following the supplier's protocol.After the incubation, the absorbance was measured at 450 nm and the control value of wells without cells (average of 6 wells) was subtracted.
HEK293 were obtained from ATCC and cultured in DMEM (Merck), 10% FBS (Merck) and 1% GlutaMax TM (Thermo-Fisher Scientific).Transfection of siRNA and DNA was conducted using the transfection reagent (sc-29528; Santa Cruz) and Polyfect (Santa Cruz), respectively.All cell lines were authenticated and mycoplasma-free.

CRISPR editing
SNRPB sgRNA (Horizon Discovery) targets the first exonintron junction including the translation initiation codon (CCATGGTAAGGAGGGCACAG).Saos-2 and HEPM cells were transfected with sgRNA and Cas9 (Thermo Fisher Scientific) using CRISPRMAX transfection reagent (Thermo Fisher Scientific).For Saos-2 cells, clonal selections by seeding cells in < 1 cell/well dilution resulted in several clones, that were subject to another round of cloning by seeding cells in < 1 cell/well dilution.The sequence of the clone used in this study is shown in Fig. S2C.For HEPM cells, two clones were obtained, however, the cells did not proliferate after several passages, therefore, HEPM cells were transfected with sgRNA and Cas9 1 day before each experiment.The CRISPR-mediated mutation was confirmed by detecting the retained intron by qPCR (Fig. S2E-I).

PCR
Total RNA was collected from relevant cells using SV Total RNA isolation system (Promega) or Trizol (Invitrogen, Waltham, MA, USA) and cDNA was synthesised using reverse transcriptase (Superscript IV; Invitrogen) and oligo dT primers.Primers used for PCR are shown in Table S1.Non-quantitative PCR was performed to detect splice variants and analysed on 2.0% agarose or 12% acrylamide gels.qPCR reaction was performed using SYBR Green (Qiagen, Hilden, Germany) and real-time PCR instruments (7900HT; Applied Biosystems, Waltham, MA, USA or Mx3005P; Agilent Technologies, Santa Clara, CA, USA).The primers were first confirmed for the efficiency to be > 98% and a single peak in the dissociation curve.Fold changes were calculated based on DC t against b-actin for Saos-2 and HEK293 or b-2 microglobulin for HEPM, and DDC t against control siRNA transfected cells, with technical triplicates.Error bars were calculated as a standard deviation of DDC t of biological triplicates unless stated otherwise.Statistical analysis was performed using unpaired equal variance t-test of DDC t values.In the post hoc analysis, only comparable pairs were analysed and displayed in the graph.For example, in Fig. 2C-E, the comparison between [control siRNA without SNRPB DNA transfection] and [SNRPB siRNA with SNRPB DNA transfection] was not considered comparable.

Western blotting
For western blot analyses, Saos-2 cells were collected in RIPA buffer [150 mM NaCl, 20 mM Tris-HCl (pH 7.6), 1% NP40, 1% Sodium deoxycholate, 0.1% SDS, 1 mM Fig. 6.Changes in splicing isoforms of LEF/TCF following SNRPB knockdown.LEF1, TCF7 and TCF7L2 isoforms were analysed in Saos-2 cells induced for osteogenesis for 5 days (A-C) or in HEPM cell aggregates induced for chondrogenesis for 4 days (D-F), either by qPCR (A, C, D, F) or non-quantitative PCR (B, E). (A, D) qPCR analysis of LEF1 and TCF7 isoforms, detecting all LEF1 or TCF7 transcripts (all), the Cysteine-rich domain included form (CRD) and N-terminal b-catenin-binding domain included form (N). The N forms are produced by alternative promoter choice, not by splicing.In HEPM aggregates, TCF7 N form was not detected by qPCR.(B, E) Non-quantitative PCR analysis of TCF7L2 splicing.In (B), Saos-2 cells were transfected with either control siRNA (C) or SNRPB siRNA (S) and induced for osteodifferentiation without or with CHIR.In (E), HEPM cells were transfected with control siRNA (Csi); SNRPB siRNA (Ssi) or SNRPB sgRNA with Cas 9 (Ssg) and induced for chondrodifferentiation in 3D aggregates.In the left panels, primers spanning exons 1-5 were used to detect the inclusion and exclusion of Exon 4 (amplicons 353 and 284 bp, respectively).In the right panels, primers spanning exons 12-17 were used to detect various C-terminal forms as shown in Fig. S5.The C-terminal truncated forms are detected as 130, 181 and 327 bp fragments, whereas the C-terminal domains-including forms are detected as 203 and 254 bp fragments.Size markers are shown on the right.(C, F) qPCR analysis of TCF7L2 isoforms, detecting all transcripts (all) at the high mobility group (HMG) domain and various isoforms as indicated.The size of the fragment (bp) indicates the corresponding fragment detected in the non-quantitative PCR in (B, E), not the size of the qPCR amplicon.In HEPM aggregates, P2R amplicon was not detected in qPCR.The graphs show the mean AE SD of three biological samples (n = 3), with each of them being the average of technical triplicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test.Asterisks indicate the comparison with control siRNA-transfected samples.

Reporter assay
Relevant cells were seeded on 24-well plates and transfected with control or SNRPB siRNA as described above as required (day 0) or CRISPR-mutated cells were used.On the following day (day 1), cells were transfected with firefly luciferase reporter plasmids; the Wnt pathway-specific reporter (TOPflash; Sigma-Aldrich) or BMP-responseelement reporter [72] (kind gift from P. ten Dijke), along with Renilla luciferase reporter with constitutive thymidine kinase promoter (E6921; Promega) for normalisation, along with various pathway activator constructs; Wnt3a (kind gift from A. P. McMahon), constitutively active LRP6 (kind gift from C. Niehrs) [74], b-catenin [105], BMP4/7 [106] or constitutively active BMP receptor IA (kind gift from K. Cho) [73] or control plasmid [107] using Polyfect (Qiagen).On day 2, chemical compounds such as 8 lM of BIO (361550; Merck), 2 lM of CHIR (SML1046; Merck) and 10 ngÁmL À1 BMP4 recombinant protein (120-05; Pepro-Tech) were added to the medium as required.Cells were lysed on day 3 (48 h after DNA transfection) unless stated otherwise for the result shown in Fig. S6C,D, and processed for dual-luciferase reporter assay (E1910; Promega).Luminescence was quantified by GloMaxÒ 20/20 Luminometer (Promega).The firefly luciferase reporter activity was normalised by Renilla luciferase activity, and further normalised against no-activation control, which is defined as relative luciferase unit (RLU).All assays were done in biological triplicates.Statistical analysis was performed by one-way ANOVA followed by unpaired equal variance ttest of biologically triplicated samples.

Statistical analysis
All experiments were technically triplicated.The method of statistical analysis of each experiment is described in the relevant Figure legend.

Fig. 1 .
Fig. 1.Affected osteoblast differentiation in SNRPB reduced cells.(A-E)Knockdown of SNRPB by siRNA transfection in Saos-2 cells.The left graphs show alkaline phosphatase (ALP) activity in the conditioned medium during Saos-2 cell differentiation following transfection of control (black) or SNRPB (orange) siRNA.The relevant siRNA was transfected in sextuplicate on the day(s) indicated by the circle on the xaxis.Every time siRNA was transfected, the conditioned medium was changed to the fresh medium.In (D), the medium was changed on day 7 after the measurement without transfection.On day 15, all samples were stained with Alizarin red (AZ) to reveal calcium deposition.The right bar charts show the quantification of AZ staining.(F) CRISPR-mediated SNRPB mutation in Saos-2 cells, analysed for ALP activity and calcium deposition as above.The medium was changed every time after sampling on days 3, 5, 8 and 11.Mutation of SNRPB and downregulation of SmB/B 0 expression are shown in Fig.S2.The scale bar for the AZ staining is 500 lm.(G, H) Viable cell counting assay following siRNA transfection (G) and CRISPR-mutated Saos-2 cells (H).Graphs show the mean AE SD; n = 6 for ALP from sextuplicate samples and n = 12 for Alizarin red as two non-overlapping images were taken from each well in (A-F); n = 6 for the cell viability assay in (G, H).In AZ quantification and cell viability assays, data were normalised to the average of control samples.*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test, by comparison with the relevant control on each day.

Fig. 2 .
Fig. 2. Suppressed expression of osteodifferentiation genes in SNRPB reduced cells.(A) RT-qPCR analysis of osteodifferentiation genes in Saos-2 cells, following transfection with either control or SNRPB siRNA and induced for osteodifferentiation for 3, 5, 8 and 11 days as indicated.y-Axis shows fold expression (DDC t ) in SNRPB siRNA-transfected cells normalised to control siRNA-transfected cells.y-Axis value 1 means that the SNRPB siRNA transfected cells show the same level of expression as control cells.(B) Western blot of control siRNA (Csi) or SNRPB siRNA (Ssi) transfected Saos-2 cells on days 3 and 8. GAPDH is a loading control.A representative of three biological repeats is shown.The molecular size marker (kDa) is shown on the right.(C) RT-qPCR analysis similar to (A), using CRISPR-mediated SNRPB-mutated Saos-2 cells.For the time-course changes in the expression level (normalised all data to day 3 control samples), see Fig. S3.(D) Western blot of control (C) or SNRPB-mutated (S) Saos-2 cells on days 3 and 8. (E-G) Rescue experiment of SNRPB siRNA by SNRPB (E, F) or Dlx5 (G) DNA transfection and subsequent 8 (E, F) or 5 (G) days of incubation.Grey bars show control siRNA transfected cells, whereas orange bars show SNRPB-siRNA transfected cells.SNRPB DNA transfection fully or partly rescues the expression of full-length encoding Dlx5 (full) (E) and Runx2 (F).Dlx5 transfection rescues the expression of Runx2 in SNRPB-knock-down cells (G).In (A, C, E-G), the graphs show the mean AE SD of three biological samples (n = 3), with each of them being the average of technical triplicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test.In (A, C), the comparison was made between control and SNRPB-deficient groups on each day.In (E-G), data were compared with control siRNA + control DNA, except that SNRPB siRNA + SNRPB DNA was compared with SNRPB siRNA + control DNA.(H) Western blot of Saos-2 cells transfected with control siRNA (Csi) or SNRPB siRNA (Ssi) together with control (C), SNRPB (S) or Dlx5 (D) DNA.

Fig. 3 .
Fig. 3. Enhanced expression of chondrodifferentiation genes in SNRPB reduced cells.(A-E) RT-qPCR analysis for indicated chondrodifferentiation genes, transfected with control siRNA (Ci), SNRPB siRNA (Si) or SNRPB sgRNA and Cas9 (Sg).The graph is presented as the fold change compared to the day 2 control siRNA transfected samples.Dlx5 primers used here detect the transcript encoding functional fulllength Dlx5.The graphs show the mean AE SD of three biological triplicates (n = 3), with each of them being the average of technical triplicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test, by comparison with the control siRNA transfected samples collected on day 2. (F, G) Immunostaining of high-density 3D culture aggregates of HEPM cells incubated for 21 days without (no induction) or with cartilage induction supplements, transfected with either control siRNA, SNRPB siRNA or SNRPB sgRNA as indicated, stained with anti-Collagen II antibodies (pink) and DAPI (blue).The scale bar is 200 lm.The bar chart (G) is the quantification of Collagen II staining intensity in control siRNA (Ci), SNRPB siRNA (Si) or SNRPB sgRNA (Sg) aggregates, shown as the mean AE SD of four biological replicates (n = 4); *P ≤ 0.05, ***P ≤ 0.001 in t-test.

Fig. 4 .
Fig. 4. The effect of Wnt and BMP pathway modulators on differentiation.(A-C) qPCR analysis of osteodifferentiation genes (Dlx5, Runx2 and Msx2) in Saos-2 cells, 2 days after induction for differentiation in the presence of DMSO (control), BIO (Wnt pathway activator), IWR1 (Wnt pathway inhibitor) or K02288 (BMP pathway inhibitor).(D) Western blot showing the effect of Wnt and BMP pathway inhibitors on Dlx5 expression.(E-G) qPCR analysis of rescue of SNRPB-knockdown by BIO for osteodifferentiation genes.Saos-2 cells were transfected with control (grey bars) or SNRPB (orange bars) siRNA and cultured without or with BIO for 5 days.Asterisks indicate comparisons with control siRNA without BIO, except that SNRPB siRNA + BIO was compared with SNRPB siRNA without BIO.(H) Western blot showing the effect of BIO and siRNA transfection on Dlx5 expression.(I-M) qPCR analysis of chondrodifferentiation genes on day 21 in HEPM cell aggregates.Cells were subject to high-density cultures, induced for differentiation and cultured with DMSO as control or Wnt pathway inhibitors PNU-74654 (PNU) or IWR1.The qPCR graphs in (A-C, E, F, I-M) show the mean AE SD of three biological samples (n = 3), with each of them being the average of technical triplicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test, compared to control samples.(N, O) Immunostaining of high-density 3D aggregate culture of HEPM cells incubated for 21 days without (no induction) or with cartilage induction supplements and either DMSO, PNU-74654 (PNU) or IWR1 as indicated, stained with anti-Collagen II antibodies (pink) and DAPI (blue).The scale bar is 200 lm.The bar chart (O) is the quantification of Collagen II staining intensity shown as the mean AE SD of four biological replicates (n = 4); *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 in t-test.