Deletion of the sclerotome-enriched lncRNA PEAT augments ribosomal protein expression

The Sclerotome Transcriptome.

Our objective was to comprehensively catalog the genes activated during sclerotome development. To isolate embryonic sclerotome precursors, we genetically marked the somite and manipulated this tissue toward the sclerotome fate with Hh agonist and BMP antagonist. The Notch ligand Dll1 is expressed in precursors of the paraxial, intermediate, and lateral plate, as well as restricted populations in the nervous system (11). The regulatory elements directing Dll1 (12) have been used to generate a mouse line expressing Cre (Dll1-msd) (13). We crossed these Dll1-msd Cre animals to R26-stop-EYFP cre reporter mice to fluorescently mark all derivatives of cells bearing Dll1-msd enhancer activity (Fig. 1A). We dissected competent posterior paraxial mesoderm from Dll1-msd Cre;R26-stop-EYFP embryos at 9-d postconception and cultured explants for 12 h with a “midline cocktail” containing sonic agonist (SAG), to mimic Shh from the notochord and floorplate (1), and the BMPR1 inhibitor LDN 193189 (LDN), to mimic notochord-derived Noggin and dermatome derived Gremlin (2⇓–4). Because we isolated explants from the posterior mesoderm before somite patterning, we hypothesized that any effects would include changes in gene expression coincident with or upstream of sclerotome induction. By quantitative PCR (qPCR) we observed a 14-fold relative induction of Pax1 in LDN and SAG-treated samples and a concomitant reduction in markers of other cell types derived from the Dll1 lineage (Fig. 1B), confirming that sclerotome was induced. We collected EYFP⁺ cells from dissociated SAG/LDN-treated and vehicle control cultures and analyzed gene expression by RNA-seq from four biological replicates (Fig. 1C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Activation of Hh and inhibition of BMP directs paraxial mesoderm toward sclerotome formation. (A) Confocal z-stack (magnification: 100×) of E9 Dll1-msd Cre;R26-stop-EYFP mouse embryo. (B) qPCR on mouse embryo explants cultured with or without LDN and SAG: Pax1 marks the sclerotome; Ptch1 indicates Hh pathway activation; Id1 is a direct BMP target; Pax3 marks dermomyotome; Myf5 marks skeletal muscle progenitors; and Sim1 is expressed in lateral dermomyotome. (C) Schematic of methodology.

Using DESeq2, we identified 264 genes with statistically significant (adjusted P < 0.05), greater than twofold differential expression between control and sclerotome-directed cultures, with 98 transcripts elevated in response to SAG and LDN and 166 transcripts reduced by the treatment (Table S1). Known early sclerotome-specific genes featured prominently in the sclerotome-directed dataset. Pax1 and Pax9 expression was elevated in directed samples 6.2- and 2.4-fold, respectively. Expression of the paralog of Bapx/Nkx3-2, Nkx3-1, was 7.2 times that of control samples. Similar to Pax1, Nkx3-1 expression was detected in somites 2 h after formation of the somite (Fig. 2B). Although Nkx3-1 is dispensable for skeletal formation, the severity of the Bapx/Nkx3-2 skeletal phenotype is enhanced in the absence of Nkx3-1 (14). Heyl, a member of the hairy/enhancer-of-split related family of transcription factors, was expressed at 6.0 times the level of controls. Heyl is expressed in the early sclerotome but is prominently associated with cardiovascular development (15); no somite-formation or somite-patterning role has been ascribed. Notably, only eight transcription factors were induced greater than twofold. In addition to Pax1, Pax9, and Nkx3-1, these transcription factors included Heyl, Eya2, Barx2, Olig2, and Bhlhe41. Our analysis also further highlighted the role of BMP inhibition (2⇓–4); among the 10 genes with greatest differential expression were BMP3, an inhibitory BMP (16), and Neuroblastoma1 (Nbl1, also known as Dan), a DAN family BMP antagonist (17).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

The sclerotome transcriptome. (A) Scatter plot of sclerotome-directed RNAseq data. Blue dots indicate transcripts with greater than twofold differential expression (P < 0.05), red dots highlight select genes elevated in directed samples, and orange dots show repressed examples. (B) E9 whole-mount in situ hybridizations (ISHs) for select transcripts elevated following sclerotome induction. (1 and 2) Nkx3-1 appears in the second most recently formed somite (S1; panel 1), whereas Nkx3-2 is detected in somites 6+ (2). Arrowheads indicate S1. (3) Heyl1 is activated in S1. (4) Pax1. (5 and 6) Bmp3 (5) and Nbl1 (6) are extracellular inhibitors of BMP signaling. (7 and 8) The left and right sides, respectively, of embryos probed for C630043F03Rik (C630). (Magnification: 30×.)

Genes that respond to SAG and LDN with diminished expression underscore the importance of BMP inhibition in sclerotome formation. Members of the inhibitor of DNA binding (Id) family control differentiation by interfering with the DNA binding activity of basic helix–loop–helix transcription factors. Our analysis showed Id1, Id2, and Id3 expression to be reduced (0.36-, 0.12-, and 0.16-fold, respectively). Id genes are targets of BMP (18), so their reduced expression in response to LDN was expected. The potential functional activity for Id proteins in restricting axial skeleton development has not been examined. We also observed diminished expression of genes with roles in the specification of other cell types in the Dll1 lineage, including Myf5 for demomyotome (19) and Pax8 lateral plate mesoderm (20).

Long-noncoding RNAs (lncRNAs) have a wide range of regulatory functions that have only begun to be explored (21⇓–23). We identified several uncharacterized, differentially expressed (DE) lncRNAs with expression in somitic tissues. For example, C630043F03Rik (C630), expressed 2.2-fold over control cultures, exhibits dynamic embryonic expression in the head, first pharyngeal arch, base of the right limb bud, and somites. Among the transcripts with greatest relative induction in the SAG- and LDN-treated samples was a noncoding RNA gene AI646519 (MGI), with 6.7-fold greater expression over controls. AI646519 is spliced and polyadenylated and is transcribed unidirectionally from the minus strand 500 bp 5′ to the Pax1 transcription start site (TSS). The proximity of this lncRNA to Pax1, the transcription factor most associated with early sclerotome, was particularly intriguing. We named this gene PEAT. Expression of PEAT is identical to Pax1 (Figs. 2 B, 4, and 3 C and D). Because lncRNAs have been shown to affect transcription of an adjacent gene (24), we hypothesized that PEAT may regulate the early development of the sclerotome through Pax1.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

The lncRNA PEAT. (A) PEAT is transcribed off the minus-strand from ∼1,000 bp 5′ to the Pax1 TSS (chr2, 127.05 Mb). Cas9 and sgRNAs targeting sites flanking the PEAT transcribed unit were used to generate the ΔPEAT line. (B) PCR on genomic DNA from ΔPEAT homozygotes using primers flanking the sgRNA target sites produces a 430-bp amplicon. (C) E9 whole-mount ISHs (Magnification: 30×.) for PEAT on PEAT embryos (heterozygous and homozygous ΔPEAT from left to right). PEAT is expressed identically to Pax1 (compare with Fig. 2 B, 4). (D) E9 transverse sections of wild-type PEAT and Pax1 ISH at level of fifth most recently formed somite. d, dermomyotome; n, notochord; Nt, neural tube; left sclerotomes (s) delineated by dashed lines.

Deletion of the lncRNA PEAT.

The Pax1 gene is located in a 870-kb gene desert between the developmental transcription factors Nkx2-2 and FoxA2 (Fig. 3). This locus has a rich history of genetic analysis. The naturally occurring, semidominant Undulated-Short Tail (Pax1un-s) mutation is a 122-kb deletion that includes all Pax1 coding sequences and presents with severe defects of the axial skeleton (25). However, animals lacking just the Pax1 gene (Pax1null) are viable and fertile and exhibit only tail defects (26). The explanation for this difference came from deletion analysis that demonstrated the Pax1un-s phenotype to be a consequence of ectopic Nkx2-2 expression in the sclerotome as a result of a repositioned Pax1 3′ enhancer element (27). Because no targeted mutation of the Pax1 locus ablated 5′ sequences but retained the Pax1 coding region, the contribution of PEAT is unknown.

To address the role of PEAT in the early development of the sclerotome, we designed single-guide RNAs (sgRNAs) to delete the PEAT-transcribed unit using the CRISPR/Cas9 system. Because PEAT lacks an ORF and consensus TSS, we used the existing annotation (mm10) and our mapped RNA-seq reads to predict the transcribed region. We injected one-cell mouse embryos with Cas9 mRNA and the sgRNAs and transferred these to pseudopregnant females. Seven of the 9 resulting pups carried the designed deletion. The ΔPEAT deletion followed Mendelian inheritance, and homozygous mutant animals are viable and fertile. ΔPEAT homozygous mutant embryos processed for in situ hybridization for PEAT lacked detectable signal (Fig. 3).

We predicted that the ΔPEAT phenotype would be subtle, like that of the Pax1-null (26), so we assessed changes in gene expression in differentiating sclerotome by RNA-seq. We dissected trunk tissue corresponding to somites 1–24 from embryonic day (E) 9 embryos generated by ΔPEAT heterozygote intercrosses. Because committed sclerotome is mesenchymal, we included flanking tissues so as not to lose any of this population. We sequenced libraries from three wild-type and three homozygous mutants. Reads mapping to regions 5′ to Pax1 were absent, confirming that ΔPEAT eliminates transcription in or near the sgRNA targets (Table S2 and Fig. S1). Despite this finding, we observed no significant difference in Pax1 expression between wild-type and ΔPEAT samples. Although alternative analyses might reveal a consistent effect, at this point we have no indication of a role for PEAT in Pax1 expression, nor do we see the subtle phenotype of the Pax1 mutant (26).

In contrast to Pax1, we did find 700 differentially expressed genes (DEGs) (Fig. 4) (adjusted P < 0.05), although none showed greater than twofold difference in expression between conditions. PEAT ablation resulted in more genes with elevated expression (482) than reduced expression (218) (Fig. 4). Strikingly, Gene Ontology (GO) analysis showed the preponderance of these genes encoded ribosomal subunit proteins (Fig. 4), members of a large family of proteins that modulate ribosome biogenesis and translation. Of the 105 expressed ribosomal protein genes with greater than 10 mapped reads in directly sequenced Dll1-lineage samples (Materials and Methods), 55 displayed significantly elevated expression in ΔPEAT; only one exhibited reduced expression (Table S2). We confirmed by qPCR that the five ribosomal proteins with greatest differential activation all trended higher in independently generated cDNA samples from embryos (SI Appendix, Fig. S2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Deletion of PEAT increases ribosomal protein and BMP pathway gene expression. (A) Volcano plot of ΔPEAT RNA-seq data. Blue dots signify significantly DEGs (P < 0.05). Red dots signify ribosomal protein DEGs. (B) Histogram showing distribution of all 700ΔPEAT DEGs. (C) Histogram showing distribution of the 73 ribosomal protein ΔPEAT DEGs. (D) Cellular component (CC) GO analysis on ΔPEAT DEGs. The bubble plot was generated using GO plot 1.0.2 from GO enrichment analysis (Materials and Methods). Circle size is proportional to the number of genes assigned to a GO term. The z-score is number of genes increased minus genes decreased divided by the square root of the number of genes assigned to a GO category. The GO terms specifically indicating ribosomal protein genes are clustered in the upper right (Table S4). (E) The ΔPEAT DEGs exhibit significant enrichment for genes with Smad1/5-bound proximal promoters. (F) Histogram showing expression of the 25 ΔPEAT DEGs bearing Smad1/5 relative to wild-type controls. Bars indicate log fold-change SE.

Several targets of BMP signaling also appeared in the list of genes with significant elevated expression. These genes included Id3 and Id1 (18), which were expressed at 1.4× that of wild-type controls, and JunD (28), expressed 1.3× over controls. Dpysl2, which was reduced in ΔPEAT samples by 14%, is repressed by BMP4 (29). Because BMP signaling interferes with early sclerotome development, these changes could indicate that PEAT helps maintain a permissive environment in the developing trunk favorable to early axial skeleton development. The lack of an overt ΔPEAT phenotype is consistent with this model. Not only are the changes in gene expression modest, but there is precedent for tolerance of modestly increased BMP signal transduction in vivo; mice conditionally heterozygous for the BMP antagonist Noggin exhibit no negative consequences to a small yet measurable increase in expression of the direct BMP target Id3 (SI Appendix, Fig. S3).

We reasoned that if ΔPEAT samples have elevated BMP signaling, expression of the activated genes might be reduced by interference with BMP signal transduction. We exploited our sclerotome-directed dataset to test this, because these conditions use the BMP antagonist LDN. Indeed, we found 16 such reduced targets, including Id1, Id3 (Tables S1 and S2, and SI Materials and Methods). In addition, Csrp2, whose expression increases by 32% in ΔPEAT mutants and is induced by Bmp2 (30), is repressed to 66% of control levels in sclerotome-directed samples. Furthermore, although BMP has not directly been shown to activate the Wnt-inhibitory Tcf3 (induced 17% in ΔPEAT samples), it is known that Wnt inhibition functions with BMP signaling to promote epithelial stem cell differentiation (31), so a positive Tcf3–BMP relationship is likely.

To substantiate a relationship between PEAT and the BMP pathway, we turned to published ChIP-seq data for the BMP signaling effector proteins, Smad1 and Smad5 (Smad1/5) (29). Fei et al. (29) performed their analysis on mouse embryonic stem cells; there is no existing R-Smad ChIP-seq analysis for the paraxial mesoderm. To generate a list more reflective of the midgestation mesoderm, we cross-referenced the list of genes with Smad1/5 bound in the proximal promoter with genes expressed in the Dll1 mesodermal lineage (Materials and Methods and Table S3). Of 24,062 annotated transcripts, 433 had Smad1/5-bound promoters and were expressed in the Dll1 lineage. Twenty-five genes on this list also appeared in our list of 700 DEΔPEAT genes, which indicates a strong likelihood of association (Fisher’s exact test P = 9.8e⁻4; observed odds ratio = 2.1) (Table S3). This analysis strongly supports our hypothesis that ΔPEAT mutants exhibit altered BMP signaling.

The Expression of PEAT-Regulated Genes During Dll1-Lineage Maturation.

Recently, the impact of ribosomal proteins on specific developmental processes has become appreciated, particularly in the developing somite (32). We hypothesized that PEAT may potentiate sclerotome development by regulating genes that impact sclerotome development. To track changes in the expression of PEAT-regulated genes in the Dll1-lineage, we generated a gene-expression catalog for four successively more mature stages of differentiation (Fig. 5A). We isolated trunk segments from E9 embryos, each with a length corresponding to four somite diameters reflecting ∼8 h of development each. Using the first posterior somite as a landmark, segments A included the posterior mesoderm and presomantic mesoderm, as well as the noncommitted most recently formed somite (33). Segments B contained somites 2–5 and correspond to stages during which initial somite pattern formation occurs. We also collected two additional segments, C (somites 6–9) and D (somites 10–13), reflecting somite maturation (Fig. 5A). We restricted this study to embryos with between 16 and 20 somites to maintain consistency in anterior–posterior identity of dissected segments. Samples from multiple embryos were pooled according to anterior–posterior segment type, dissociated for 30 min, and FACS-sorted for GFP⁺ cells. We generated RNA-seq libraries from batches of 2.5 × 10⁵ cells. Principal component analysis showed that replicates from each group formed tight clusters, indicating highly consistent and reproducible gene-expression profiles (Table S6 and Fig. 5B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Analysis of the Dll1-lineage. (A) Schematic describing the four segments, A to D, dissected from E9.0 Dll1-msd Cre;R26-stop-EYFP mouse embryo. (B) PCA for all RNA-seq samples, including single-replicate unsorted head and trunk samples. (C) Heat map showing average expression of select mesoderm-expressed genes for the 32 h following Dll1-msd initiation (Table S5). (D) Histogram showing the difference in average expression of the 10 ΔPEAT DEGs that are also variable between Dll1-lineage samples A and B. Bars indicate 2× SE. (E) Heat map displaying the variation in average expression of the eight ΔPEAT DEGs that are also DE between both Dll1-lineage samples and the control and sclerotome-directed samples.

The Dll1-lineage stage series tracks transcription during the specification and early development of all Dll1-derived somitic cell types, including dermomyotome and myotome and sclerotome. We found that the lineage-expression analysis faithfully tracks normal maturation of trunk lineages (Fig. 5 B and C); the posterior, early-expressed trunk genes are enriched as expected in the youngest, most posterior segment. In more anterior explants, markers of differentiated somite derivatives appear. This analysis also captured expression of anterior-posterior regionalized markers, notably the Hox genes (Table S7).

We then assessed changes in expression of PEAT-regulated genes across the four Dll1-lineage time points. This analysis identified 29 PEAT-regulated genes that exhibit changes in expression during the 32 h of Dll1-development. Because our primary interest is initial pattern formation within the paraxial mesoderm, we examined expression changes of the 10 PEAT DEGs from segments A to B (Fig. 5D). For 9 of the 10 DEGs, average normalized counts fell from A to B, coincident with the onset of PEAT expression. We also queried whether there were any genes that were DE between (i) Dll1-lineage segments, (ii) wild-type control and ΔPEAT embryos, and (iii) control and sclerotome-directed samples. Eight genes satisfied this triple condition (Fig. 5E) and are candidates to have unexplored activities promoting or repressing sclerotome development.

However, the most striking outcome of this Dll1-lineage expression analysis was the absence of DE ribosomal proteins, the most prominent DEG category in PEAT mutants. There was only one gene associated with translation, Eif5a2. This result does not support a model in which these ribosomal proteins function directly in Dll1-lineage tissue specification. We speculate that the changes in ribosomal protein gene expression in response to PEAT deletion are compensatory, and act to maintain cellular conditions required for normal development.