Whole transcriptome expression profiling of mouse limb tendon development by using RNA-seq

Abstract

Tendons are fibrous connective tissues that transmit force between muscle and bone. Whereas the molecular and cellular mechanisms of bone and muscle development have been well studied, that of tendon development is poorly understood. Using the Scx-GFP transgenic mice, we isolated GFP⁺ cells from the developing mouse limbs at E11.5, E13.5, and E15.5, respectively, and carried out whole transcriptome RNA-seq analysis. Comparing the gene expression profiles of GFP⁺ and GFP⁻ cells in the E13.5 limb isolated over 1500 genes that exhibited enrichment of mRNA expression by at least 1.5-fold in the GFP⁺ cells. Of these, 778 genes showed expression up-regulated by more than 1.5-fold from E11.5 to E13.5 and 516 genes showed expression up-regulated by more than 1.5-fold from E13.5 to E15.5 in the GFP⁺ cell population. Interestingly, over 30 genes encoding transcription factors are among the early-activated genes in the GFP⁺ cells. Whole mount and section in situ hybridization analyses showed that many of these transcription factor genes have distinct patterns of expression during limb development and identified Foxf2 expression as a specific marker for differentiated dorsal limb tendon cells. Together, these data provide a valuable resource for further investigation of the molecular mechanisms regulating tendon development.

Introduction

Tendons and ligaments provide physical and mechanical connections between muscle and bone, and contribute to the complex architectures of the musculoskeletal system. Mechanical forces generated by muscle contractions are transmitted through tendons to drive movement of the skeletal system. During force transmission, tendons and ligaments withstand stress and tension and, thus, are prone to acute injuries. In addition, chronic stress overload and aging related tissue degeneration cause the common tendon disorder known as tendinopathy. Most acute tendon injuries and some chronic tendon diseases require surgical tissue repair, which is clinically challenging since tendon healing is slow and complete recovery is rarely achieved. To improve tendon treatment strategies, recent research has started to focus on understanding the basic mechanisms of tendon development and tendon cell differentiation.

Tendon development initiates with the specification of tendon progenitor cells within the somite, limb mesoderm, and craniofacial mesenchyme, respectively. Tendon progenitors condense and reside between osteoblasts and myoblasts, where they undergo rapid proliferation and differentiate into tenocytes. Tenocytes synthesize various extracellular matrix molecules, including predominantly type I collagen, type III collagen, as well as the small leucine-rich proteoglycans decorin, biglycan, fibromodulin, and lumican, and organize them to form tendon fibrils¹. Highly oriented tendon fibrils are organized into more complex hierarchical structures, tendon fibers, which are surrounded by connective tissue sheath, called epitenon^2–4. Although the anatomy of tendons has been relatively well characterized, mechanistic studies of tendon development have been hampered by the lack of tendon-specific molecular markers at distinct stages of tendon cell differentiation. The discovery of tendon-specific expression of the helix-loop-helix class transcription factor Scx in 2001 significantly facilitated molecular and genetic studies of tendon development⁵. Pryce et al. (2007) demonstrated that an 11-kb genomic DNA containing the Scx gene is sufficient to drive transgenic reporter expression in all tendon cell lineages, including craniofacial tendon cells derived from the neural crest, axial tendon cells derived from the somites, and limb tendon cells derived from lateral plate mesoderm⁶. Although Scx is expressed in the early tendon progenitor cells and its expression persists throughout tendon development, it is not required for tendon cell specification as mice lacking Scx exhibit defects in development of specific subpopulations of tendons, i.e., the force transmitting tendons⁷. Two other transcription factors, Mohawk homeobox (Mkx) and early growth response 1 (Egr1), have since been shown to play critical roles in tendon development as well. Mkx is expressed in differentiating tendon cells and regulates tendon fibril growth and organization postnatally^{8, 9}. Egr1isexpressed in differentiated tendon cells and plays an important role in the tendon healing process in damaged tendon tissues¹⁰.

To gain better understanding of the molecular mechanisms of tendon development, we have carried out whole transcriptome RNA-seq analysis of gene expression profiles of limb tendon cells in mice. We report here the RNA-seq datasets and identification of additional transcription factors with distinct patterns of expression during limb development.

Materials and Methods

Mice

The Scx-GFP transgenic mice have been described previously⁶ and generously provided by Dr. Ronen Schweitzer (Shriners Hospital for Children, Portland, Oregon). The transgenic mice are maintained by crossing to wildtype CD1 mice. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Cincinnati Children’s Hospital Medical Center.

Fluorescence-activating cell sorting (FACS)

The limbs of E11.5, E13.5 and E15.5 Scx-GFP transgenic embryos were harvested and the forelimbs and hindlimbs, respectively, were pooled from multiple embryos (nine E11.5 embryos, six E13.5 embryos, six E15.5 embryos). The freshly dissected embryonic limbs were digested with the trypsin-EDTA solution (Invitrogen) at 37°C for 4 minutes. After inactivation of trypsin with DMEM containing 10% FBS, cells were dissociated by pipetting. The dissociated limb cells were suspended in PBS with 2% FBS and 10 mM EDTA, and filtered through a 40 μm nylon cell strainer (BD Falcon, 352340). GFP+ cells from E11.5, E13.5, and E15.5 hindlimbs were isolated using BD FACSAria II. In addition, both GFP⁺ and GFP⁻ cells from the pooled E13.5 forelimbs were isolated using BD FACSAria II.

RNA-seq and data analysis

Total RNAs were extracted from pooled FACS sorted cells of three limbs for each stage studied, using the Qiagen RNeasy Micro Kit (Qiagen catalog# 74004). Sequencing libraries were generated by using Illumina Nextera DNA Sample Prep Kit and sequenced using Illumina Hiseq 2000. Sequenced reads were mapped to the reference mouse genome (mm9) using Bowtie version 0.12.7. Single-end reads were aligned using Tophat version 1.4.1 for Illumina¹¹.RNAseq data were analyzed using Avadis NGS software, with the fragments per kilobase exon per million mapped sequences (FPKM) value calculated for each RefSeq gene for relative gene expression. For analyses of differential expression, the fold change cut-off was set at 1.5-fold or higher and P-values <0.01 from the AudicClaverie test were considered to be statistically significant, with Benjamini Hochberg FDR multiple testing correction¹¹. Functional annotations analysis was carried out using online tools at https://toppgene.cchmc.org, as previously described^{12, 13}.

In situ hybridization assays

The plasmid containing the Scx cDNA for synthesis of cRNA probes for in situ hybridization has been described previously⁷ and generously provided by Dr. Ronen Schweitzer (Shriners Hospital for Children, Research Division, Portland, Oregon). Other cDNA templates were amplified by PCR from E11.5 mouse embryonic limb cDNAs using gene specific primers (Supplementary Table 1) and subcloned in the pBluescript II KS+ plasmid vector. All cRNA probes were tested for specificity by whole mount in situ hybridization of E12.5 mouse embryos and cDNA templates that did not result in tissue specific expression patterns were either validated with a second, non-overlapping, template or replaced by a more specific template.

For whole mount in situ hybridization, embryos were harvested from wildtype CD1 mice at predetermined gestational stages and fixed in 4% paraformaldehyde in PBS overnight at 4°C, dehydrated through methanol series and directly used for experiments. For section in situ hybridization, embryos were fixed in 4% paraformaldehyde in PBS overnight at 4 °C, dehydrated through ethanol series and embedded in paraffin. Serial sections of 7-μm thickness were used for experiments. In situ hybridization assays were performed as previously described¹⁴.

Results

Characterization of whole transcriptome expression profiles of limb tendon development by using RNA-seq

The Scx-GFP transgenic mice have been shown exhibit strong GFP expression throughout tendon development⁶. However, it has also been reported that the Scx-GFP transgene is much more broadly expressed than the endogenous Scx mRNA in the early developing limbs, with the difference been most pronounced in E10.5 and E12.5⁶. As limb tendon progenitor cells start to differentiate from E12.5 to E13.5¹⁵, we first compared directly the patterns of GFP expression in Scx-GFP transgenic embryos with that of endogenous Scx mRNA expression in wildtype embryos at E13.5. We found that the pattern of Scx-GFP is highly similar to that of Scx mRNA expression in the autopod region while GFP expression domain is broader than the endogenous Scx mRNA expression in the proximal limb regions at this stage (Fig. 1A–H). We next performed quantitative real-time RT-PCR analysis of expression of select markers for tendon and other cell types in GFP⁺ and GFP⁻ cells isolated from microdissected E13.5 forelimbs by using fluorescence activated cell sorting and found that expression of tendon marker genes, including Scx, tenomodulin (Tnmd), and Col1a1, were highly enriched, whereas expression of the chondrocyte marker Sox9 and of the osteoblast marker Runx2 were not, in the GFP⁺ cells in comparison with the GFP- cells in the E13.5 forelimb (Fig. 1I). These results suggest that comparative gene expression profiling analysis of the GFP⁺ versus GFP⁻ cells in the E13.5 limb tissues of Scx-GFP transgenic embryos is a valid approach to identify early tendon developmental regulators.

Fig. 1. Analysis of gene expression profiles of developing limb tendons at E13.

(A–D) In situ hybridization detection of Scx mRNA expression in the wildtype fore- (A, B) and hind-limbs (C, D). The Scx mRNA signals are detected in purple color. (E–H) The patterns of green fluorescent protein (GFP) expression in the Scx-GFP transgenic mouse embryos recapitulate the endogenous Scx mRNA expression patterns in both fore- (E, F) and hind-limbs (G, H). (I) Data from Quantitative real-time RT-PCR assay of select markers in the FACS-isolated GFP⁺ and GFP⁻ cells from E13 forelimbs.

Total RNAs were prepared from FACS-isolated E13.5 forelimb cells and processed for whole-transcriptome RNA-seq. The resulting data were analyzed using the Avadis NGS software, with the fragments per kilobase exon per million mapped sequences (FPKM) value calculated for each RefSeq gene for relative levels of gene expression. With the FPKM cut-off threshold at 5.0, transcripts corresponding to 11035 genes were included for analysis of differential expression in the GFP⁺ versus GFP⁻cells. Of these, transcripts of 1555 genes exhibited more than 1.5 fold higher levels of expression in the GFP⁺ than the GFP⁻ cells. Gene ontology (GO) analysis of these 1555 genes showed that the GO terms related to tendon, including “skeletal system development”, “connective tissue development”, “cartilage development”, “collagen binding”, “collagen fibril organization”, “proteoglycan binding”, “abnormal joint morphology”, “abnormal tendon morphology”, are highly represented (Supplementary Table 2).

Although expression of many tendon marker genes is significantly enriched in the GFP⁺ cells in the E13.5 forelimb of Scx-GFP transgenic embryos (Supplementary Table 3), many of the GFP⁺ cells are not tendon cells because the Scx-GFP transgene is expressed more strongly and in broader domains than the endogenous Scx mRNAs⁶. To further facilitate identification of early tendon developmental regulators, we performed RNA-seq of FACS-isolated GFP⁺ cells from E11.5, E13.5, and E15.5 hindlimb tissues, respectively, of Scx-GFP transgenic embryos. At E11.5, the limb tendon progenitor cells are beginning to commit to the tendon cell fate and both the Scx-GFP transgene and the endogenous Scx mRNAs are expressed in broad domains in the developing limb buds. By E13.5, Scx-expressing tendon tissues are morphologically distinguishable from surrounding tissues in the autopod and zeugopod regions in both fore- and hind-limbs¹⁶. By E15.5, most limb tendon progenitor cells have differentiated into tenocytes and expression of the Scx-GFP transgenic reporter becomes highly restricted to the tendon tissues^{6, 16}. Thus, we reasoned that important positive regulators of tendon development should be up-regulated from E11.5 to E13.5 and some positive regulators as well as markers of tendon cell differentiation should be up-regulated from E13.5 to E15.5. Indeed, the levels of expression of several known marker genes for tendon differentiation, including Tnmd, lumican (Lum), fibromodulin (Fmod), decorin (Dcn), Col1a1, Col1a2, Col3a1, Col6a1 and Col6a2, were significantly increased from E11.5 to E13.5 and E15.5, even though the relative abundance of endogenous Scx mRNAs decreased slightly from E13.5 to E15.5, in our GFP+ RNA-seq datasets (Fig. 2A), consistent with enrichment of the tendon cell lineage in the GFP+ cell population.

Fig. 2. Differential gene expression analysis of developing limb tendon by using RNA-seq.

(A) Comparison of levels of expression (FPKM) of Scx, Tnmd, Lum, Fmod, Dcn, Col1a1, Col1a2, Col3a1, Col6a1, Col6a2 mRNAs in the RNA-seq datasets from FACS-isolated E11.5, E13.5, and E15.5 hindlimb GFP+ cells. (B) Venn diagrams showing partial overlap of enriched gene sets from differential gene expression analysis of RNA-seq data sets. 1555 genes were found to exhibit at least 1.5-fold increase in expression in GFP⁺ cells versus GFP⁻ cells from forelimbs of E13.5 Scx-GFP transgenic mouse embryos (Red circle). These were compared with the genes whose expression was increased by more than 1.5-fold from E11.5 to E13.5 (blue circle) and from E13.5 to E15.5 (green circle) in the GFP⁺ hindlimb cell populations.

Differential expression analysis of the RNA-seq datasets showed that over 1700 genes exhibited increased levels of mRNA expression by more than 1.5-fold in the GFP+ populations from E11.5 to E13.5 (Fig. 2B). Of these, 778 were in the GFP+ cell enriched dataset from the E13.5 forelimb tissues, suggesting that they were activated in tendon progenitor cells during early tendon development. Expression of 379 of these genes were further up-regulated by more than 1.5-fold from E13.5 to E15.5, indicating that they are good tendon differentiation markers. GO analysis showed that many of these genes encode extracellular matrix components or proteins interacting with extracellular matrix (Table 1), whereas many of the other 399 genes in the 778 gene-set are involved in regulation of cell development or differentiation (Table 2). The top 20 enriched genes in each pair-wise comparison of the RNA-seq datasets from different stages are included in Supplementary Tables 4 – 6.

Table 1.

GO analysis of 379 genes enriched in Scx-GFP⁺ limb cells at E13.5 and continuously up-regulated from E11.5 to E15.5

GO category	GO ID	GO term	p value	# of Genes
Molecular Function	GO:0005201	Extracellular matrix structural constituent	4.268E-24	24
	GO:0005518	Collagen binding	1.512E-21	22
	GO:0019838	Growth factor binding	1.663E-15	22
	GO:0005539	Glycosaminoglycan binding	1.325E-13	25
Biological Process	GO:0030198	Extracellular matrix organization	2.002E-45	68
	GO:0030199	Collagen fibril organization	6.771E-24	20
	GO:0007155	Cell adhesion	5.741E-19	69
	GO:0001501	Skeletal system development	4.840E-16	41
	GO:0061148	Connective tissue development	2.315E-10	23
	GO:0051216	Cartilage development	3.280E-9	19
	GO:0000904	Cell morphology involved in differentiation	8.627E-9	43
Human Phenotype	HP:0001388	Joint laxity	1.841E-16	21
	HP:0002758	Osteoarthritis	2.113E-11	13
	HP:0001367	Abnormal joint morphology	2.407E-9	46
	HP:0002813	Abnormality of limb bone morphology	5.840E-9	54
	HP:0003549	Abnormality of connective tissue	8.320E-9	39
Mouse Phenotype	MP:0005503	Abnormal tendon morphology	1.487E-13	13
	MP:0002108	Abnormal muscle morphology	1.925E-13	59
	MP:0005508	Abnormal skeleton morphology	5.163E-10	74
	MP:0003098	Decreases tendon stiffness	6.642E-7	4
	MP:0000163	Abnormal cartilage morphology	3.094E-6	24

Table 2.

GO analysis of 399 genes enriched in Scx-GFP⁺ cells at E13.5 and up-regulated from E11.5 to E13.5 only

GO category	GO ID	GO term	p value	# of Genes
Molecular Function	GO:0005509	Calcium ion binding	6.803E-7	34
	GO:0045499	Chemorepellent activity	4.576E-6	4
	GO:0008307	Structural constituent of muscle	1.265E-4	6
	GO:0043394	Proteoglycan binding	2.120E-4	5
Biological Process	GO:0048468	Cell development	2.444E-13	87
	GO:0009887	Organ morphogenesis	2.122E-11	51
	GO:0001501	Skeletal system development	4.561E-11	34
	GO:0000904	Cell morphogenesis involved in differentiation	1.272E-10	47
	GO:0045595	Regulation of cell differentiation	2.024E-10	64
	GO:0061448	Connective tissue development	4.591E-8	20
	GO:0051216	Cartilage development	7.191E-7	16
Human Phenotype	HP:0009803	Short phalanx of finger	3.354E-6	12
	HP:0009115	Aplasia/hypoplasia involving the skeleton	6.477E-5	27
	HP:0009810	Abnormality of the joints of the upper limbs	1.582E-4	17
Mouse Phenotype	MP:0002113	Abnormal skeleton development	8.617E-7	29
	MP:0000166	Abnormal chondrocyte morphology	1.109E-6	12
	MP:0005076	Abnormal cell differentiation	5.647E-4	29

Characterization of expression of transcription factors in early tendon development

One problem hindering tendon research is that only three transcription factors, Scx, Mkx, and Egr1, have been shown to play important roles in tendon cell differentiation. Thus, we focused on identifying transcription factors that are up-regulated during early tendon development. We found 31 transcription factor genes, including Mkx, whose expression levels were enriched in the GFP⁺ forelimb cells at E13.5 and also significantly up-regulated, by more than 1.5-fold, from E11.5 to E13.5 in the FACS-isolated GFP⁺ hindlimb cells (Table 3). Of these, 13 were further up-regulated by more than 1.5-fold from E13.5 to E15.5 in the FACS-isolated GFP⁺ hindlimb cells. To further characterize their expression patterns during tendon development, we performed both whole mount and section in situ hybridization analyses. These analyses revealed that many of these transcription factor genes were broadly expressed during limb development (Supplementary Fig. 1), with only a few showing tendon-specific expression. For example, three members of the early B-cell factor (Ebf) family are among the 31 transcription factor genes significantly enriched in the Scx-GFP⁺ forelimb cells at E13.5 and significantly up-regulated from E11.5 to E13.5 in the GFP⁺ hindlimb cells. Upon examination by whole mount in situ hybridization analysis, we found that, although all three Ebf genes were expressed in the developing limb mesenchyme, they were not specific for tendon (Fig. 3A–U, and Supplementary Fig. 2). Comparison of their expression with that of Scx by using section in situ hybridization on adjacent cross sections of the E13.5 forelimbs showed that expression of Ebf1 and Ebf3 was concentrated in the connective tissues surrounding the digit cartilage but absent from the dorsal and ventral limb tendons (Fig. 3J–U). Expression of Ebf2 was partially overlapping with that of Scx mRNAs and was mostly in the connective tissues in between the cartilage and the tendon tissues located dorsally and ventrally (Fig. 3L, P, T, compare to Scx expression in Fig. 3J, N, and R, respectively). These results are consistent with a recent report of expression patterns of the Ebf family genes during mouse limb development¹⁸. On the other hand, we found that expression of Six2 and Foxf2 are highly specific in limb tendon development (Fig. 4, and Supplementary Fig. 3). Although a previous study suggested that both Six1 and Six2 are expressed during limb tendon development¹⁹, our differential expression analysis of the RNA-seq datasets did not pick up Six1. Direct comparison of the expression patterns of Six1 and Six2 with that of Scx and Tnmd confirmed that Six1 is primarily expressed in the perichondrium and skeletal muscle in the developing limb whereas Six2 is highly expressed in the differentiating limb tendon cells by E13.5 (Fig. 4A – Ab, and Supplementary Fig. 3), with much stronger expression in the ventral than the dorsal limb tendons (Fig. 4, K, O, S, W, Aa). Our in situ hybridization results showing lack of Six1 expression in the developing mouse limb tendon are consistent with another study that showed non-overlapping patterns of expression of Six1 and Scx mRNAs in the developing mouse limbs²⁰. Remarkably, we found that Foxf2 expression is absent in the early developing limb prior to E12.5 and is only activated in the differentiating dorsal limb tendon cells, but not in ventral limb tendon, at E13.5 (Fig. 4D, H, L, P, T, X, Ab). Consistent with the RNA-seq data showing continued up-regulation of Foxf2 mRNA expression from E13.5 to E15.5 (Table 3), in situ hybridization analysis detected a graded proximal-distal pattern of Foxf2 mRNA expression in the E13.5 limbs, with strongest mRNA signals in in the digit region (Fig. 4, T, X, Ab). These data identify Foxf2 expression as a specific marker for differentiated dorsal limb tendon.

Table 3.

Transcription factors activated during early tendon development

Gene Symbol	Gene Description	FPKM values
		E11(GFP+)	E13(GFP+)	E15(GFP+)
Mkxa	Mohawk homeobox	2.92371	29.2162	20.8919
Bnc2a	basonuclin 2	9.19681	39.9182	31.2196
Creb3l1b	cAMP responsive element binding protein 3-like 1	0.95854	37.0428	118.088
Creb3l2a	cAMP responsive element binding protein 3-like 2	56.5512	114.06	123.322
Creb5a	cAMP responsive element binding protein 5	8.42502	37.2559	29.2659
Dmrt2a	doublesex and mab-3 related transcription factor 2	3.55406	14.0685	10.5184
Ebf1a	early B cell factor 1	50.7322	91.7113	73.8705
Ebf2a	early B cell factor 2	8.74121	42.0117	35.7571
Ebf3a	early B cell factor 3	30.2354	94.1107	68.8637
Foxf2b	Forkhead box F2	1.29488	4.47666	7.19177
Foxp1a	Forkhead box P1	12.2231	46.6546	34.5489
Hic1a	hypermethylated in cancer 1	5.59986	41.599	40.1295
Hivep2b	human immunodeficiency virus type I enhancer binding protein 2	0.7615	7.2643	19.2601
Klf14b	Kruppel-like factor 14	1.03155	6.09617	29.6916
Meox1a	mesenchyme homeobox 1	17.3961	35.7012	53.9161
Myod1b	myogenic differentiation 1	3.00055	5.56645	15.6337
Myogb	Myogenin	3.323	8.62464	35.9339
Nfibb	nuclear factor I/B	7.72681	59.2753	90.4447
Nficb	nuclear factor I/C	11.0855	32.8171	79.9389
Nfixb	nuclear factor I/X	4.60583	55.5602	131.575
Osr1a	Odd-skipped related 1 (Drosophila)	29.585	56.2089	50.3268
Plagl1b	pleiomorphic adenoma gene-like 1	18.2221	103.312	457.855
Sim2b	single-minded homolog 2 (Drosophila)	3.21879	5.816	9.59725
Six2a	sine oculis-related homeobox 2 homolog (Drosophila)	5.53047	29.5461	15.2109
Tshz2a	teashirt zinc finger family member 2	16.9844	43.3862	53.111
Zcchc5b	zinc finger, CCHC domain containing 5	2.1941	25.7706	90.3176
Zfp354ca	zinc finger protein 354C	19.2049	37.4625	37.6418
Zfp580a	zinc finger protein 580	12.3642	21.3423	19.844
Zfp9a	zinc finger protein 9	10.8411	32.3694	30.8232
Zhx3a	zinc fingers and homeobox-3	8.66091	13.9502	16.6142
Zim1b	zinc finger, imprinted 1	7.57369	16.3261	59.0092

^a18 genes up-regulated from E11 to E13;

^b13 genes up-regulated from E11 to E13 and from E13 to E15.

Fig. 3. Expression of the Ebf family genes during early limb tendon development.

(A – C) At E11.5, the three Ebf genes were expressed in partially overlapping domains in the forelimb mesenchyme. (D – F) At E12.5, Ebf1 and Ebf2 mRNAs were mainly expressed in the autopod, around the developing digits (D, E), while Ebf3 mRNAs were also expressed in the zeugopod (F). (G – U) At E13.5, Ebf1 and Ebf3 were expressed in cells surrounding the developing tendon, cartilage and bones, but not in the dorsal or ventral tendons. Expression of Ebf2 mRNAs partially overlapped with Scx mRNAs in the dorsal and ventral tendons in the autopod (L, P, T, compared with J, N, R, respectively). Cross sections (J – U) were from mid-phalangeal (J – M), metacarpal (N – Q) and Wrist (R – U) regions. All sections are shown dorsal side up.

Fig. 4. Foxf2 and Six2 exhibit differential expression in dorsal and ventral limb tendons.

Expression patterns of Six1, Six2, and Foxf2 mRNAs in E11.5, E12.5, and E13.5 forelimbs were compared with that of Scx mRNAs by whole mount in situ hybridization (A – P) and with that of Tnmd by section in situ hybridization (Q – Ab). Cross sections (Q – Ab) are from mid-phalangeal (Q – T), metacarpal (U – X) and Wrist (Y – Ab) regions. All sections are shown dorsal side up. Six1 is strongly expressed in the proximal limb mesenchyme from E11.5 to E12.5 (B, F, J, N) and its expression in the developing limb becomes restricted to the perichondrium and muscle cells by E13.5 (R, V, Z). Six2 mRNA expression (C, G, K, O, S, W, Aa) exhibited significant overlap with that of Scx (A, E I, M) and Tnmd (Q, U, Y), but is stronger in the ventral limb tendon than the dorsal limb tendon (K, O, S, W). Foxf2 mRNA expression was not detected in the limbs at E11.5 and E12.5 (D, H). At E13.5, Foxf2 mRNA expression was specifically detected in the dorsal limb tendon within the autopod (T, X, Ab).

Discussion

The discovery of Scx as a tendon lineage marker in 2001⁵ and subsequent generation of the Scx-GFP transgenic mice⁶ have enabled unprecedented characterization of tendon tissues at the cellular and molecular levels in mice. Havis et al. (2014) reported recently the first transcriptomic analysis of developing mouse limb tendon cells using microarray hybridization analysis of FACS-isolated GFP⁺ cells from E11.5, E12.5, and E14.5 forelimb tissues of Scx-GFP transgenic mouse embryos¹⁷. Although microarrays provide a powerful approach for high-throughput analysis at the genomic scale, the hybridization-based approach is known to suffer from cross-hybridization noise and limited dynamic range of detection due to signal saturation. Thus, our RNA-seq datasets generated in this study complements the microarray datasets from Havis et al. (2014) and provide a valuable resource for the tendon research field¹⁷. We have deposited these RNA-seq datasets into the Gene Expression Omnibus database at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/geo, accession number GSE65180) so that other researchers will be able to independently analyze these datasets to incorporate into their own research on tendon biology.

In contrast to bone and muscle development in which the transcriptional regulatory mechanisms have been extensively studied, only three transcription factors have been shown to play important roles in tendon development. Mice lacking Scx exhibit the earliest and most severe tendon differentiation defects thus far characterized⁷. However, many tendons and ligaments still form and function in the Scx^−/− mutant mice⁷. Mice lacking either Mkx or Egr1 exhibit postnatal tendon maturation defects^{8, 9, 21}. To improve the understanding of the molecular regulation of tendon development, we focused on identification of other transcription factors specifically expressed during tendon development. However, although we isolated over 30 transcription factor genes from differential expression analyses of the early limb tendon RNA-seq datasets, in situ hybridization assays showed that most of them did not exhibit restricted expression in the developing tendon cells. On the other hand, GO analysis of the genes whose expression was significantly enriched in the Scx-GFP⁺ cells showed a clear association with connective tissue development and many of the top 20 enriched genes in each pair-wise comparison encode known markers of tendon cell differentiation. Thus, the Scx-GFP⁺ cell populations were clearly enriched with tendon cells although each of the FACS-isolated population must also contain non-tendon cells, consistent with the initial report of the Scx-GFP transgenic mice in which the Scx-GFP was shown expressed in much broader domains than that of the endogenous Scx mRNAs, particularly during early stages of tendon development⁶. It is possible that the cell dissociation and FACS isolation procedures might alter the RNA profiles of the isolated cells, which, combined with the ectopic Scx-GFP expression during early limb tendon development and the fact that genes encoding transcription factors are usually expressed at low abundance compared with genes encoding extracellular matrices or other structural proteins, limited our success in identifying tendon-specific transcription factors. In the future, it will be helpful to combine the Scx-GFP reporter with in situ cell isolation procedures, such as laser capture microdissection, for more accurate transcriptome expression profiling of the developing tendon.

The combined RNA-seq and in situ hybridization analyses allowed us to identify Foxf2 expression as a specific marker of differentiated dorsal limb tendon. Foxf2 is a member of the large Forkhead family of transcription factors^{22, 23}. It is strongly expressed in the developing facial mesenchyme and mice lacking Foxf2 die perinatally with cleft palate²⁴. The specific expression of Foxf2 in the differentiated dorsal limb tendon suggests that distinct molecular pathways are employed in regulating differentiation of different tendons. Consistent with this hypothesis, the levels of Six2 mRNAs appear much higher in the ventral limb tendons than the dorsal limb tendons. Yamamoto-Shiraishi and Kuroiwa (2013) recently showed that Six2 expression in the dorsal autopod, but not in the ventral autopod, depends on Hox13 function²⁵. Further investigation of the relationship between Foxf2 and Six2 as well as their function in limb tendon development will provide novel insights into the molecular mechanisms of musculoskeletal system development and organization.