Introduction

Craniofacial abnormalities affecting bone formation in the skull and face are the most common birth defects in infants. Proper formation of these structures involves coordinated processes including proliferation, migration, and differentiation of cranial neural crest cells (CNCCs) (Martik and Bronner 2017; Plein, Fantin, and Ruhrberg 2015). CNCCs are a transient population of progenitor cells that populate the first and second pharyngeal arches and give rise to the maxilla, mandible, palates, and other structures of the developing head (Chai et al. 2000; Jiang et al. 2002; Martik and Bronner 2017). Dysregulation of transcription factors, chromatin remodelers, RNA regulatory proteins, and signaling molecules have been implicated in impaired neural crest development that results in craniofacial defects (Bélanger et al. 2018; Martik and Bronner 2017; Plein, Fantin, and Ruhrberg 2015; Strobl-Mazzulla, Marini, and Buzzi 2012). Spliceosomopathies, which are a group of disorders caused by defects in the splicing machinery, frequently affect the craniofacial skeleton and limb. However, why these tissues are more sensitive to spliceosomal defects remain incompletely understood. Splicing factors and small nuclear ribonucleoproteins (snRNPs) are core components of splicing machinery, which control pre-mRNA splicing and mRNA maturation through the process of alternative splicing (AS). AS includes seven types of events, exon skipping, mutually exclusive exons, alternative 5’ and 3’-splice sites, alternative promoters, alternative polyadenylation, and intron retention (Hooper et al. 2020). Intron retention (IR) is a type of AS abundant in plants and fungi, where introns are retained within the mature mRNA sequence (E.T. Wang et al. 2008). Although initially overlooked in mammals due to low occurrence and challenges in detection, recent technical advances have unveiled its significant roles in various aspects, including cell differentiation and stress response, especially in neuronal, immune, and cancer cells (Marquez et al. 2012; Monteuuis et al. 2019; Braunschweig et al. 2014; Wong and Schmitz 2022).

In exploring the splicing mechanisms in CNCCs, we focused on upstream regulators for splicing factors. The activity of splicing factors is tightly controlled by post-translational modifications, among which methylation is the most abundant, surpassing phosphorylation (Thandapani et al. 2013; Ruta, Pagliarini, and Sette 2021). Splicing regulator methylation mostly occurs on arginine motifs, catalyzed by the protein arginine methyltransferase (PRMT) family of enzymes. PRMTs-mediated methylation of splicing factors determines their subcellular localization, expression, and activity, influencing the splicing product (Blanc and Richard 2017; Liu and Dreyfuss 1995; Snijders et al. 2015). PRMT1 is the most abundant PRMT in most cell types and specifically catalyzes methylation of arginine (R) residues within RG/RGG/GAR repeats, which are motifs enriched in RNA-binding proteins, particularly splicing regulators, resulting in asymmetric dimethylation of arginine (ADMA) (Thandapani et al. 2013; Smith et al. 2004). The importance of PRMT1 in craniofacial development is evident from our earlier investigation in which genetic deletion of Prmt1 in CNCCs leads to cleft palate and mandibular hypoplasia (Gou, Li, Wu, et al. 2018; Gou, Li, Jackson-Weaver, et al. 2018). In the present study, we uncovered a previously unrecognized role of PRMT1 in regulating intron retention (IR) in CNCCs. The mandibular primordium of Prmt1-deficient embryos demonstrated augmented retention of introns in mRNAs that encode bone and cartilage matrix components. We further identified SFPQ as a downstream mediator of PRMT1 in the developing craniofacial structures, that depends on PRMT1 for arginine methylation and protein expression. Depletion of SFPQ in CNCCs partially phenocopied PRMT1 deletion, causing elevated retention of introns in mRNA transcripts of matrix, neuronal, and Wnt signaling pathways genes. We further recognized long gene/intron splicing as a common feature among SFPQ-regulated introns and demonstrated these retained introns contain premature termination codons (PTCs) that trigger nonsense-mediate decay (NMD) to reduce mRNA expression levels. Together, these findings demonstrate that the PRMT1-SFPQ pathway modulates CNCC gene expression via control of splicing. Deficiency of this pathway leads to aberrant intron retention, which induces mRNA decay that downregulates matrix expression in CNCCs during craniofacial development.

Results

CNCCs from the embryonic mandibular process exhibit abundant intron retention, which was further elevated by loss of Prmt1

PRMT1 is the most abundant enzyme of the arginine methyltransferase family in cranial neural crest cells (CNCCs), with an expression level 10∼100 folds higher than other PRMTs (Figure-1A). We previously demonstrated critical roles for PRMT1 in CNCCs during craniofacial skeleton formation using genetic deletion, whereby neural crest-specific deletion of Prmt1 resulted in cleft palate and shorter mandibles (Gou, Li, Wu, et al. 2018; Gou, Li, Jackson-Weaver, et al. 2018). Shorter mandibles, or mandibular hypoplasia is a craniofacial phenotype shared by most spliceosomopathies. To investigate the molecular mechanisms underlying PRMT1’s role in mandible development, we conducted a genome-wide analysis of Prmt1 deletion-induced transcriptional and splicing alterations within the embryonic mandibles. For this purpose, CNCCs, the cell lineage that gives rise to the bone and cartilages in the mandible were isolated from mandibular processes of control (Wnt1-Cre; Rosa^Td) or Prmt1 CKO mutant (Wnt1-Cre; Prmt1^fl/fl; Rosa^Td) mice at embryonic day 13.5 (E13.5) for mRNA extraction and sequencing (Figure-1B). The ensuing comparative analysis using rMATS (Y. Wang et al. 2024) unveiled changes across five types of alternative splicing events (Figure-1C), echoing our earlier finding where PRMT1 regulates exon usage during cardiac morphogenesis (Jackson-Weaver et al. 2020). In contrast to embryonic epicardial cells, we noted a higher abundance of altered intron retention (IR) in mandibular CNCCs (Figure-1C), with a predominance of IR increase when compared to control CNCCs (Figure-1Ca’). The IR changes were further exemplified by the elevation of intronic expression in Pex12, Nkx3-2, Nkiras1, 1810026B05Rik, and the reduction of intronic expression in Tbx1 (Figure-1D-F). To analyze the intronic expression comprehensively at a gene-specific level, we employed IRI (intron reads index) analysis, which is an algorithm that quantifies the ratio of normalized read counts between intronic and exonic regions (Ni et al. 2016; Sun et al. 2023; Tian et al. 2020). Subsequent analysis for retained introns demonstrated that Prmt1 deficiency-induced change in intron retention altered an array of biology pathways. We noted protein binding, metal ion binding, and cartilage development among genes that present increased intronic expression, while mitochondrion and metabolic pathways were identified among genes with decreased intronic expression (Supplemental Table S1). These findings identify intron retention as a prevalent phenomenon in CNCCs during craniofacial development, and the deletion of Prmt1 further enhanced intron retention in genes involved in diverse biological processes.

Figure 1.

Neural crest-specific deletion of Prmt1 caused a significant reduction of matrix gene expression in the developing mandibles

The main documented roles of retained intron in mature mRNA is to reduce gene expression via non-sense mediated decay (NMD) (Schmitz et al. 2017). When intron-retaining transcripts are incorporated into the translation machinery, introns that contain premature termination codons (PTCs) within their open reading frames trigger NMD, leading to transcript degradation and reduced expression of genes harboring these introns (Wong et al. 2016). We observed that intron-retaining transcripts illustrated in Figure-1, Pex12, and Nkiras1 demonstrated increased intronic expression and decreased exonic expression, suggesting reduced gene expression. To systematically analyze genes that are differentially downregulated, we compared the transcriptional profile between control and Prmt1 CKO mutant of mandibular CNCCs and revealed downregulation of 303 genes and concurrent upregulation of 160 genes by Prmt1 deletion (Figure-2A, 2B). Upon pathway analysis, glycosaminoglycan (GAG) degradation (35 out of 303 genes) and extracellular matrix (ECM) (33 out of 303 genes) emerged as the top pathways among downregulated genes (Figure-2C). Findings from GAG degradation and ECM genes align with our understanding that these downregulated genes encode matrix proteins pivotal to osteogenic and chondrogenic differentiation and the formation of bone and cartilage matrix. This notion was supported by bioinformatic analysis using Ingenuity Pathway Analysis (IPA) based on downregulated genes, which suggested a disruption in bone, cartilage, and connective tissue development within Prmt1 CKO mutant mandibles (Figure-2D). The data collected from transcriptional analysis underscore the importance of PRMT1 in orchestrating matrix gene expression critical to bone, cartilage, and connective tissue formation during mandible morphogenesis.

Figure 2.

Deletion of Prmt1 in CNCCs elevated intron retention in mRNAs that encode ECM and GAG degradation enzymes

To determine whether the downregulation of matrix genes is regulated by intron retention, we first examined the intronic expression profile of ECM genes. A prominent elevation of intron retention within ECM genes was revealed by scatter plot where most ECM data points landed above the red line that denotes unchanged intron retention (Figure-3A). ECM genes with the most significant increase in intron expression were labeled and highlighted as red dots, and further exemplified by the track view of Bmp7 and Cthrc1 (Figure-3A, 3B). In these ECM genes, the increased intron retention correlated with the reduced exonic expression, indicated by intronic expression and exonic expression respectively (Figure-3C). To validate the reproducibility of elevated intron retention in ECM transcripts, we expanded our analysis to incorporate four additional embryos from each group, isolated mature mRNA using poly(A)+ magnetic beads, and designed primers for RT-PCR analysis of intronic and exonic region of additional ECM genes. In line with our initial observations, intronic expression increased significantly, coupled with decreased exonic expression in Dcn, Tnn, Lox, Loxl1, Matn2, Col14a1, Adamts12, Adamts2, Adam12, and Scara5 of the mutant CNCCs as compared to the control (Figure-3D). These findings demonstrate that ECM transcripts in Prmt1 CKO mandibles exhibited elevated intron retention, which correlates with a reduction in mRNA expression levels. Next, we analyzed the intronic reads profile of GAG degradation genes, which is the top downregulated gene cluster in the Prmt1 CKO group and clinically associated with craniofacial birth defects. Similar to ECM genes, the majority of GAG degradation genes showed elevation of intron retention indicated by scatter plot (Figure-3E). The increased intron retention of St6galnac3 and Galn11, indicated by higher intronic expression, correlated with reduced mRNA expression, indicated by lower mRNA expression (Figure-3F, 3G). Upon further inspection of the retained introns in ECM and GAG degradation transcripts, retained introns in the significantly downregulated genes all contain PTCs within their open reading frames, suggesting PTC-triggered transcript degradation via NMD-mediated RNA surveillance system (Supplemental Table S2) (Wong et al. 2016). Altogether, these data indicate that Prmt1 deletion elevates intron retention in matrix genes encoding ECM proteins and GAG degradation enzymes and downregulated their mRNA expression within CNCCs.

Figure 3.

PRMT1 methylates SFPQ in CNCCs

PRMT1 is a methyltransferase that catalyzes methylation of arginine (R) residues within RG/RGG/GAR repeats, which are motifs highly enriched in RNA-binding proteins, particularly splicing regulators (Thandapani et al. 2013). PRMT1 catalyzes asymmetric dimethylation of arginine (ADMA) residues, which regulates the turnover, subcellular localization, and splicing activity of splicing regulators (Smith et al. 2004). We therefore hypothesized that PRMT1 regulates intron retention via methylation of splicing regulators. Earlier studies by Graham et al. and our team have characterized the landscape of arginine methylation governed by PRMT1 and identified a cohort of RNA processing proteins exhibiting PRMT1-dependent methylation (Hartel et al. 2019). Among these, six splicing regulators are highly expressed in embryonic CNCCs: SFPQ, SRSF1, EWSR1, TAF15, TRA2B, and G3BP1 (Figure-4A). To discern whether any of the splicing regulators are accountable for elevated IR in Prmt1 deficient CNCCs, we conducted an in-silico analysis of the RNA-seq data of mandibular CNCCs from control and Prmt1 CKO embryos utilizing rMATS (multivariant analysis of transcript splicing, MATS), followed by rMAPS2 (RNA map analysis) (Hwang et al. 2020; Y. Wang et al. 2024). This integrated approach detects differential splicing events, and then generates visual maps illustrating the spatial distribution of RNA binding protein (RBP) motifs, and finally ranks RBPs based on motif density score and p-value (Park et al. 2016). Through this in-silico analysis, we identified a subset of RBPs with significantly differential binding in splicing events altered by PRMT1 (p<0.00001) (Supplemental Table S3). To pinpoint the most pertinent RBPs, we cross-referenced this significant RBP list with the previously identified six PRMT1-dependent and CNCC-enriched splicing regulators (Figure-4A). SFPQ emerged as the prime candidate, boasting the highest motif scores and the lowest P-values (Supplemental Figure-S1A-E, Supplemental Table S4).

Figure 4.

Subsequently, we set out to determine whether SFPQ is methylated by PRMT1 in the craniofacial structures. Employing the proximity ligation assay (PLA), a technique designed for the in-situ detection of protein modifications, we probed the E13.5 embryo sections with anti-SFPQ and anti-methyl-arginine ADMA antibodies utilizing the PLA kit, which revealed the presence of methylated SFPQ as distinct green punctate staining. Within the craniofacial complex, a substantial presence of SFPQ methylation was observed in the mandibular processes at E13.5 (Figure-4B, 4Ca-d). In Prmt1 CKO embryos, the signal of methyl-SFPQ showed marked diminishment within the CNCC-derived mesenchyme region (Figure-4Ce-h, 4D). The same analysis was conducted in the maxillary regions, which demonstrated robust SFPQ methylation in the control group and a dramatic reduction of methylation in Prmt1 CKO embryos (Figure-4B, 4Ea-h, 4F). This CNCC-specific deletion of PRMT1 only reduced SFPQ methylation in the CNCCs, as non-CNCC lineages within the epithelium continued to display methyl-SFPQ signals (Figure-4Ga-f, Supplemental Figure-S2). These findings establish SFPQ’s methylation within the craniofacial complex as a PRMT1-dependent phenomenon.

PRMT1-induced methylation controls the activity of splicing regulators by modifying their stability, subcellular distribution, or RNA binding ability (Rho et al. 2007). We then proceeded to examine whether SFPQ protein expression was regulated by PRMT1 in the craniofacial complex. In control embryos, SFPQ exhibited prominent expression in CNCCs and mainly localized in the nucleus (Figure-4Ha-c, 4Ja-c). In Prmt1-deficient CNCCs, the protein’s expression level witnessed a significant reduction within the mandibular and maxillary processes (Figure-4H-4K). A corroborating Western blot (WB) analysis using tissues from the craniofacial structures, including facial and anterior skull regions of embryonic heads, confirmed this decline in SFPQ expression (Figure-4L, 4M). Taken together, these data demonstrate that PRMT1 methylates SFPQ and promotes SFPQ protein expression in the craniofacial complex.

PRMT1-SFPQ pathway regulates the splicing of matrix genes in CNCCs

To investigate the role of SFPQ in intron retention and gene expression within CNCCs, we isolated primary CNCCs from control (Wnt1-Cre; Td) embryos at E13.5 and enriched CNCCs by Td+ signal using cell sorting. Within a day, Td+ CNCCs were transfected with control or two independent SFPQ siRNAs to deplete SFPQ, and then poly(A)+ mRNA was purified followed by sequencing. To determine whether these primary isolated CNCCs retained CNCC characteristics after siRNA-mediated knockdown, the expression of CNCC markers Twist1, Sox10, Msx1, Snai2, and Tfap2a was examined (Achilleos and Trainor 2012; Ishii et al. 2012), which demonstrates that the CNCC marker expression in siRNA-transfected CNCCs after isolation from E13.5 embryos is comparable to that in fresh isolated CNCCs from E13.5 or E15.5 embryos (Supplemental Figure-S3), validating CNCC identity and supporting this method as a feasible approach for mechanistic study of CNCCs using developmental stage-specific embryos. Subsequently, using sequencing data from these siRNA-transfected CNCCs, we conducted bioinformatic analyses of differentially retained introns and differentially expressed genes between control and Sfpq-depleted CNCCs (Figure-5A, 5B). Genes with differential intron retention were assessed by IRI analysis, which revealed elevated intron retention in 397 (siSFPQ #1 vs. siCont) or 388 (siSFPQ #2 vs. siCont) genes and reduced intron retention in 129 (siSFPQ #1 vs. siCont) or 151 (siSFPQ #2 vs. siCont) genes caused by Sfpq depletion (Figure-5A). Genes showing increased IR in Sfpq depleted CNCCs encompassed “proteoglycan metabolic process” and “proteoglycan biosynthesis process”, which included GAG degradation genes, and “connective tissue development”, which includes ECM genes (Figure-5Ca, 5Cb). Sfpq depletion also led to downregulation of 269 (siSFPQ #1 vs. siCont) or 232 (siSFPQ #2 vs. siCont) genes, among which 81 (siSFPQ #1 vs. siCont) and 60 (siSFPQ #2 vs. siCont) genes overlapped with elevated IR, respectively (Figure-5B). To determine intron retention co-regulated by SFPQ and PRMT1, we cross-analyzed genes with increased intronic expression in Sfpq-depleted CNCCs and Prmt1-deficient CNCCs and uncovered a partial overlap of a total of 64 genes (Figure-5D), represented by matrix genes Col4a2, Adam12, Ntn1, App, St6galnac3, Galnt11 Galnt10, and Asph (Figure-5E, 5F). The elevation of intronic retention in these genes also correlated with a reduction of gene expression (Figure-5G, 5H). GO analysis of differentially downregulated genes by both SFPQ-specific siRNAs (cluster 4 within the heatmap) further confirmed extracellular matrix, extracellular structure, and adhesion as the top categories (Figure-5I, 5J). These data indicate that depletion of SFPQ is sufficient to enhance IR and reduce gene expression for a cluster of matrix genes, suggesting loss of SFPQ as a regulatory mechanism for these genes in Prmt1-deficient CNCCs and SFPQ as a downstream effector for PRMT1 in CNCCs.

Figure 5.

SFPQ depletion in CNCC enhanced intron retention in Wnt signaling and neuronal transcripts and reduced their mRNA expression

More intriguingly, Wnt signaling emerged as the top pathway regulated by SFPQ in CNCCs (Figure-5C). Sfpq depletion increased the percentage of intron-retaining mRNAs of genes in this cluster and reduced total mRNA expression (Figure-6A). These Wnt signaling components-regulated by SFPQ encode both positive and negative regulators of Wnt signaling and encompass both canonical and non-canonical Wnt signal regulators. The SFPQ-regulated Wnt signaling components Wwox also exhibited increased IR and decreased expression in Prmt1 CKO CNCCs. Additionally, neuronal genes occupy a big fraction among SFPQ-regulated introns and genes in CNCCs, as “axon guidance”, “axonogenesis”, “neuron projection development”, and “leaning or memory” emerged as the top biological processes among differentially upregulated introns (Figure-5Ca, 5Cb), and “regulation of membrane potential”, “learning or memory” “cognition” and “locomotory behavior” denotes differentially downregulated genes in Sfpq-depleted CNCCs (Figure-5I, 5J). In CNCCs, Sfpq depletion increased the percentage of intron-retaining mRNAs of these neuronal genes and reduced total mRNA expression (Figure-6B). Upon further inspection, “regulation of membrane potential” was also among identified among genes with decreased expression (38 out of 303 genes) in Prmt1 CKO CNCCs. Altogether, these findings demonstrate that, besides matrix genes, SFPQ regulates Wnt signaling components and neuronal gene expression via intron retention.

Figure 6.

Disturbance of SFPQ activity impaired splicing of long genes and long introns

SFPQ predominantly binds to intronic regions and regulates splicing (Hosokawa et al. 2019). We therefore used published CLIP-seq dataset generated from embryonic mouse brain to determine whether these aberrantly retained introns are SFPQ targets (Hosokawa et al. 2019). The retained Ptk7 intron 1 encompassed a much higher density of SFPQ binding peaks when compared to the spliced intron 6 within the same gene (Figure-7A). We further conducted analysis for all the introns at the genomic level comparing SFPQ binding peaks among introns with elevated retention to introns with reduced retention or no changes and demonstrated that introns with elevated retention in Sfpq-depleted CNCCs showed significantly higher enrichment of SFPQ binding peaks (Figure-7B). We further analyzed SFPQ CLIP-seq peaks at the gene level, comparing genes with elevated IR to genes with no IR changes. Our analysis showed that genes containing elevated IR in Sfpq-depleted CNCCs exhibited a trend towards higher presence of SFPQ binding peaks based on this embryonic brain dataset (Figure-7C). These data suggest that introns with aberrant IR elevation in Sfpq-depleted CNCCs are direct splicing targets bound by SFPQ.

Figure 7:

To understand the molecular basis of how SFPQ targets these matrix, Wnt and neuronal genes and introns, we evaluated characteristics shared among SFPQ-regulated genes, and recognized that they are either long genes (>100kd in length) or retain long introns (>10kd in length). Of note, Wnt pathway gene Wwox and matrix genes St6galnac3 are 913kd and 526kd in length, and their retained introns are extremely long, 639kd and 215kd, respectively, in contrast to a median intron size of 0.6 ∼ 2.4kb in the mouse genome (Hong, Scofield, and Lynch 2006). In the human and mouse genome, long genes mostly occur because of long introns (Breschi, Gingeras, and Guigó 2017; Lopes et al. 2021). We then plotted the length of SFPQ-regulated genes against genes across the genome and uncovered that the genes with increased IR in Sfpq depleted CNCCs have a median length of ∼100kb, which is much longer than genes in the control groups, where the medial length is around 10kb (Figure-7D). These findings suggest that SFPQ-regulated splicing facilitates regulation of long gene expression during craniofacial development.

SFPQ depletion reduces matrix and Wnt signaling gene expression via IR-induced NMD

In Sfpq-depleted CNCCs, retained introns contain PTCs within their open reading frames, suggesting PTC-triggered NMD as the mechanism of degradation (Supplemental Table S5). To test intron-regulated mRNA decay, we utilized the ST2 cell line, which is a mesenchymal cell line frequently used in the mechanistic study of ossification. SFPQ-regulated intron splicing and mRNA expression for Col4a2, St6galnac3, and Ptk7 is conserved in ST2 cells where Sfpq depletion caused an increase in the percentage of intron-retaining mRNAs, and a concurrent decrease in their transcript levels (Figure-7E). To determine whether Col4a2, St6galnac3, and Ptk7 mRNA were degraded through intron-induced NMD, we inhibited NMD with a chemical inhibitor, NMDI14, which targets key components of the NMD machinery, Upf1, and Smg5/7 (Martin et al. 2014). NMDI14 induced accumulation of intron 4 of Col4a2, intron 1 of St6galnac3, and intron 1 of Ptk7 (Figure-7F), indicating that mRNA decay of intron-retaining transcripts occurs through the NMD machinery. To assess the impact of NMD on the total level of these mRNAs, we used PCR primers that detect the fully spliced transcripts and observed that inhibition of NMD restored and increased the levels of these mRNAs in Sfpq-depleted ST2 cells (Figure-7G). These data indicate that depletion of Sfpq causes aberrant intron retention that accelerates mRNA decay through NMD.

Discussion

Here we revealed prevalent intron retention within CNCCs during embryonic development and propose intron-regulated expression of long genes as a causal mechanism for craniofacial deformity. This study demonstrated PRMT1-SFPQ pathway-regulated intron retention and subsequent control of matrix and Wnt pathway component gene expression during craniofacial development. Genetic deletion of Prmt1 or depletion of Sfpq in CNCCs caused aberrant intron retention in genes encoding matrix proteins and Wnt pathway components, which triggers NMD to degrade these transcripts. Matrix and Wnt pathway genes regulated by this mechanism share a common feature as long genes with a median length of 100kb, suggesting higher susceptibility of these genes to perturbation of the PRMT1-SFPQ pathway or splicing defects during craniofacial development. IR has been linked to diseases like cancer and autoimmune diseases (Dvinge and Bradley 2015). This study on PRMT1-regulated IR in controlling the expression of matrix genes in CNCCs, to our knowledge, is the first characterization of IR in craniofacial development. Many of the PRMT1-SFPQ pathway regulated matrix and Wnt signaling genes are linked to congenital conditions. For example, in our study, the long gene Wwox exhibited increased retention of long introns (introns 3 and 4) and decreased expression upon Prmt1/Sfpq deficiency. In human patients, pathogenic variants of WWOX with large deletion within long introns have been associated with epileptic encephalopathy syndrome manifesting shared facial phenotype (Dvinge and Bradley 2015). This deletion within these long introns may affect SFPQ binding or splicing that impairs PRMT1-SFPQ pathway-regulated splicing of human WWOX genes. Another SFPQ target, Ptk7 is associated with neural tube defects and scoliosis (Berger, Wodarz, and Borchers 2017). SFPQ itself is linked to neurodegenerative diseases including amyotrophic lateral sclerosis, frontotemporal dementia, and Alzheimer’s disease, and was recently identified by a transcriptional study as a genetic regulator in CNCCs (Feng et al. 2021).

A main feature shared by SFPQ-regulated genes in CNCCs is gene length, with a median length of 100kb in contrast to a median length of ∼10kb across the genome. In each SFPQ-regulate gene, introns exhibiting the highest retention also tend to be longer introns, for example, intron 4 of Col4a2 which displayed the most dramatic increase in intronic reads within Col4a2 boasts a length of 45kb. Increased IR in neuronal and immune cells has been proposed to facilitate a rapid response to external stimuli, allowing a time frame shorter than that required for de novo transcription (Ni et al. 2016; Mauger, Lemoine, and Scheiffele 2016). Thus, during craniofacial morphogenesis, IR represents a developmental process that enables rapid production of proteins when they are required. Long genes over 100kb in length require two to several hours for transcription, which presents a challenge for the fast tempo during embryogenesis. This intron-regulated mechanism allows cells to control the rate of production post-transcriptionally, and to achieve rapid protein translation without waiting for the time period required for transcription. Long genes regulated by this mechanism are highly enriched in matrix pathways, and responsible for the production of building blocks for bone, cartilage, and connective tissues. Since spliceosomopathies preferentially affect the craniofacial skeleton and limb, intron retention-regulated matrix expression provides mechanistic insights for the higher susceptibility of these tissue types to spliceosome dysfunction.

PRMT1 is a critical regulator of RNA splicing. In mammalian cells, it methylates over forty splicing factors to modulate protein stability and subcellular localization, thereby influencing their accessibility to mRNA (Thandapani et al. 2013; Wu et al. 2022). PRMT1-catalyzed arginine methylation also affects their RNA binding affinity and specificity, altering their activity and the splicing product (Fedoriw et al. 2019; Fong et al. 2019). Therefore, it is not surprising that the disruption of this methyltransferase leads to splicing defects. However, the elevation of IR in this study presented a distinct role for PRMT1 in CNCCs. IR has been linked to diseases like cancer and autoimmune diseases (Dvinge and Bradley 2015). This study on PRMT1-regulated IR in controlling the expression of matrix genes in CNCCs, to our knowledge, is the first characterization of IR in craniofacial development, adding developmental insights to our knowledge of the roles of IR in pathophysiology. Manipulation of SFPQ expression (knockdown with 50% efficiency) demonstrated that SFPQ shares around 10% overlap with PRMT1 in IR-regulated neural crest gene expression. Given that PRMT1 methylates over 40 splicing factors in mammals and many of these splicing factors are abundantly expressed in embryonic CNCCs, we expect that PRMT1-governed splicing will be mediated by multiple splicing regulators besides SFPQ. The functional network of splicing regulation during craniofacial morphogenesis awaits further investigation.

We recognized SFPQ as a key PRMT1 substrate in CNCCs based on an unbiased in silico analysis and further validation in the mandibular and maxillary primordium. This aligns with previous work demonstrating that PRMT1 is responsible for the association of SFPQ with mRNA in mRNP complexes in mammalian cells (Snijders et al. 2015). Our findings on altered neuronal genes in Sfpq-depleted CNCCs also echoed previous reports in which SFPQ regulates many neuronal genes involved in axon extension, branching, viability, and synaptogenesis via direct binding to their intronic regions (Ruskin, Zamore, and Green 1988; Cosker et al. 2016) and loss of SFPQ function compromises splicing patterns, especially the accurate splicing of long introns in embryonic brain and AML samples (Luisier et al. 2018; Taylor et al. 2022; Thomas-Jinu et al. 2017). Findings in this study further revealed previously unappreciated roles of SFPQ in the regulation of matrix genes through IR, which bear functional significance in CNCCs and craniofacial skeleton formation.

Altogether, these findings demonstrate that the PRMT1-SFPQ pathway regulates intron retention in CNCCs and suggests intron retention in long introns or long genes as a mechanism that controls matrix and Wnt signaling gene expression during craniofacial morphogenesis.

Material and methods

Animals

Prmt1fl/fl mice were generously provided by Dr. Stephane Richard (McGill University). Wnt1-Cre; Prmt1^fl/fl; R26R ^tdTomato and Wnt1-Cre; R26R^tdTomato mice were obtained by mating Wnt1-Cre, R26R^tdTomato mice (Jax #009107 and #007914) with Prmt1fl/fl mice. Animals were genotyped using established protocols (Chai et al., 2000; Yu et al., 2009) and all care and experiments followed USC’s IACUC protocols.

Proximity Ligation Assay (PLA)

PLA was performed using a Duolink In Situ kit (Millipore Sigma Cat# DUO92101-1KT), SFPQ antibody (Everest Biotech Cat#EB09523), and pan-methyl arginine antibody, Asymmetric Di-Methyl Arginine (ADMA) (Cell Signaling Technology Cat# 13522) to detect methyl-SFPQ. Tissue sections were blocked with the Duolink blocking solution in a humidity chamber for 1 hour at 37°C before incubating with the primary antibodies (diluted 1:50 in Duolink PLA probe diluent) overnight at 4°C. The next day, tissue sections were incubated with the Goat-PLUS and Rabbit-MINUS PLA probes for 1 hour at 37°C, followed by ligase and amplification solution. After the final washes, tissue sections were mounted with Duolink In Situ mounting media with DAPI to counterstain nuclei. Images were visualized under the Confocal microscope. The number of PLA signals that appear as punctuate green signals in the nuclei of the cells and the quantification was performed with CellProfille software.

Tissue Processing and Immunofluorescence Staining

E13.5 control and Wnt1-Cre; Prmt1^fl/fl mice embryos were fixed in 4% PFA overnight, dehydrated, embedded in O.C.T., sectioned at 8 µm thickness in sagittal orientation, and mounted on glass slides. The samples were submitted to antigen retrieval using Antigen Unmasking Solutions (Vector Laboratories Cat# H-3300-250) and washed and incubated in 0.5% Triton X-100 in PBS for 20 minutes for permeabilization. The samples were briefly washed with PBS, blocked with 10% goat serum for 1 hour, and then incubated with primary antibody SFPQ (Cell Signaling Technology Cat#23020, 1:100) at 4°C overnight in a moisture chamber. After this, the samples were blocked with a secondary antibody (1:400, goat anti-rabbit IgG (H+L) Alexa Fluor 488 Cat# A-11008) for 1 hour followed by DAPI diluted 1:1000 in PBS for 10 minutes. Finally, the samples were mounted with mounting media (Electron Microscopy Science Cat# 1798510) and covered with cover glass for imaging using Keyence BZ-X800 and Leica DMI 3000B microscopes and quantification with CellProfille software.

Western Blotting

Heads of E13.5 control and Wnt1-Cre; Prmt1^fl/fl mice were dissected, followed by removal of the brain. The tissue was homogenized with a pestle followed by lysis with RIPA Lysis and Extraction Buffer. The total protein concentration was determined by comparison with BSA standards. Twenty micrograms of total protein from each sample were loaded into each well of a 10% polyacrylamide gel. Western blot analysis was carried out as previously described (Iwata et al., 2012). Antibodies against PRMT1 (Cell Signaling, Cat# 2449; 1:1000), SFPQ (Cell Signaling Technology Cat#23020, 1:1000), and GAPDH (Cell Signaling, Cat# 97166; 1:5000) were used for Western blot. The samples were analyzed three times in independent blots and the bands were quantified by optical densitometry. The analysis was performed using ImageJ digital imaging processing software (ImageJ 1.48v, National Institutes of Health, Bethesda, MD, USA). The expression of each analyzed protein was normalized with GAPDH.

NCC collection and FACS analysis

The heads of E13.5 Wnt1-Cre; R26R^tdTomato and Wnt1-Cre; Prmt1^fl/fl;R26R^tdTomato mice were dissected and placed in a sterile tube containing Hanks’ Balanced Salt Solution (Thermo Fisher Scientific Cat# 14175079). Following centrifugation, the supernatant was discarded, and the tissue was subjected to TrypLE (Thermo Fisher Scientific Cat# 12605010) incubation for 5-15 minutes at 37°C on a rotator. The resulting dissociated tissue was neutralized by the addition of fetal bovine serum (FBS) and then filtered through a 40μm strainer to obtain a single-cell suspension. After centrifugation, cells were resuspended in an appropriate volume of serum-free medium for FACS analysis. Sorting was performed based on tdTomato fluorescence (Excitation/emission: 554/582 nm), and td-positive cells were collected into separate tubes containing DMEM (Genesee Scientific Cat#25-501) supplemented with 20% FBS.

Primary CNCC culture and siRNA transfection

CNCCs were cultured in DMEM supplemented with 20% FBS in an incubator until they attach to the bottom. Then, reverse transfection with control siRNA (Qiagen Cat# 1027310), SFPQ siRNA #1 (Qiagen Cat#SI05783848), and SFPQ siRNA #2 (Qiagen Cat#SI05783876) at 40 nM using Lipofectamine RNAiMax transfection reagent (Invitrogen) for siRNA delivery were performed. 48h later, total RNA was isolated using TRIzol reagent (Invitrogen) following the manufacturer’s protocols followed by mRNA isolation using NEBNext High-Input Poly(A) mRNA Magnetic Isolation Module (NEBNext® Cat#E3370S).

mRNA isolation and sequencing

Poly(A) mRNA isolation was extracted from total RNA by using the NEBNext High-Input Poly(A) mRNA Magnetic Isolation Module (NEBNext® Cat#E3370S). Five sets of primary isolated CNCCs from control (Wnt1-Cre;ROSA^Td) or Prmt1 CKO (Wnt1-Cre; Prmt1^fl/fl; ROSA^Td), and three sets of isolated CNCCs transfected with siControl or siSFPQs were sequenced at 40 million reads sequencing depth and 150 bp paired-end sequencing.

Bioinformatic Analysis of Differentially Regulated Genes (DEG) and differential Intron Retention (IR)

The sequencing quality of RNA-seq libraries was assessed by FastQC v0.11.8 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Low-quality bases and adapters were trimmed using Trim Galore (v0.6.7,https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). The reads were mapped to mouse genome mm10 using hisat2 (v2.1.0) (Kim et al. 2019). The mapped sam files from hisat2 were converted to bam files which were then further turned to sorted bam files by samtools (v1.7) (Li et al. 2009). Mapped reads were then processed by htseq-count (v1.99.2) to calculate the read count of all genes (Anders, Pyl, and Huber 2015). The sorted bam files were then given to IRTools (https://github.com/WeiqunPengLab/IRTools/) to calculate the IR of each gene and the read count of genes’ constitutive intronic regions (CIRs) and constitutive exonic regions (CERs). The expression level of a gene was expressed as TPM (Transcripts Per Kilobase Million) value. EdgeR was used to identify differentially expressed genes (DEGs) by requiring ≥ 1.5-fold expression changes and adjusted p-value < 0.05 (Robinson, McCarthy, and Smyth 2010). The intron retention index (IRI) of a gene is defined as the ratio of the overall read density of its CIRs to the overall read density of its CERs. Genes with 0< IRI <1 and CER expression greater than 1 were selected to identify differential IR genes. Student’s t-test was used to identify differential IR genes by requiring ≥ 1.5-fold log2(IR) changes and p-value < 0.05. Gene Ontology analysis was done by DAVID (https://david.ncifcrf.gov/tools.jsp) and Metascape (https://metascape.org/gp/index.html#/main/step1) (Sherman et al. 2022; Zhou et al. 2019). The RNA-seq data are deposited to the Gene Expression Omnibus (GEO) database with accession number GSE266474.

rMATS and rMAPS Analysis

rMATS (https://github.com/Xinglab/rmats-turbo) and rMAPS2 (http://rmaps.cecsresearch.org/) were used to find potential RNA binding Proteins (RBPs) that contribute to differential splicing events (Hwang et al. 2020; Y. Wang et al. 2024). rMATS was first used to identify differential alternative splicing events between wild type and PRMT1 CKO samples by requiring FDR < 0.05. The output of rMATs were then fed to rMAPS2 to identify significant RBPs contributing to each type of Splicing events.

Quantitative PCR (qPCR)

For both experiments, Control, and Wnt1-Cre; Prmt1fl/fl mandibles, and siRNA-transfected CNCCs, mRNAs were quantified by real-time PCR with IQ Sybr Green Supermix (Bio-Rad) and normalized against GAPDH mRNA levels. Relative changes in expression were calculated using the ΔΔCt method. Primer sequences are listed in Supplemental Table S6.

Statistical analysis

Two-tailed Student’s t-tests or Fisher’s exact tests were applied for statistical analysis. For all graphs, error bars represent standard deviations. A p-value of <0.05 was considered statistically significant.

Data Availability

RNA-seq data is deposited at the NCBI Gene Expression Omnibus (GSE 171630)

Additional information

Author contributions

J.R and N.U. designed and performed experiments, analyzed data, drafted, and revised the manuscript; Q.C. and W.P. performed bioinformatic analysis of RNA-seq data; M.V., Y.C. & A. M. contributed to embryo collection and CNCC isolation; T.R. contributed to RT-PCR analysis, J.X., and W.P. concepted the study, supervised the performance of experiments and data analysis, and critically revised the manuscript.

Competing interests

The authors declare no competing interests.