This article was accepted for inclusion in Developmental Dynamics 233#2, June 2005–Special Focus on Limb Development.
Transcriptome analysis of the murine forelimb and hindlimb autopod†
Article first published online: 29 JUL 2005
Copyright © 2005 Wiley-Liss, Inc.
Volume 234, Issue 1, pages 74–89, September 2005
How to Cite
Shou, S., Scott, V., Reed, C., Hitzemann, R. and Stadler, H. S. (2005), Transcriptome analysis of the murine forelimb and hindlimb autopod. Dev. Dyn., 234: 74–89. doi: 10.1002/dvdy.20514
- Issue published online: 16 AUG 2005
- Article first published online: 29 JUL 2005
- Manuscript Accepted: 8 JUN 2005
- Manuscript Revised: 22 APR 2005
- Manuscript Received: 3 JAN 2005
- Shriners Hospital for Children Research (HSS)
- National Institutes of Health (HSS)
- Biostatistics Shared Resource of the OHSU Cancer Institute. Grant Number: P30 CA69533
- distal limb development;
- gene expression;
To gain insight into the coordination of gene expression profiles during forelimb and hindlimb differentiation, a transcriptome analysis of mouse embryonic autopod tissues was performed using Affymetrix Murine Gene Chips (MOE-430). Forty-four transcripts with expression differences higher than 2-fold (T test, P ≤0.05) were detected between forelimb and hindlimb tissues including 38 new transcripts such as Rdh10,Frzb, Tbx18, and Hip that exhibit differential limb expression. A comparison of gene expression profiles in the forelimb, hindlimb, and brain revealed 24 limb-signature genes whose expression was significantly enriched in limb autopod versus brain tissue (fold change >2, P ≤ 0.05). Interestingly, the genes exhibiting enrichment in the developing autopod also segregated into significant fore- and hindlimb-specific clusters (P ≤ 0.05) suggesting that by E 12.5, unique gene combinations are being used during the differentiation of each autopod type. Developmental Dynamics 234:74–89, 2005. © 2005 Wiley-Liss, Inc.
The analysis of limb development has provided remarkable insights into molecular mechanisms required to form complex three-dimensional structures. In mice, limb bud initiation occurs at embryonic day (E) 9.0 as a small outgrowth of the lateral plate mesoderm at the junction of the caudal and thoracic somites (Tickle et al., 1976; Solursh et al., 1990; Kaufman and Bard, 1999). Following limb bud initiation, the growth and development of the forelimb and hindlimb follows a remarkably similar process of axial patterning, using signals such as fibroblast growth factors (FGF) to maintain proximal-distal proliferation, sonic hedgehog to establish anterior-posterior polarizing activity, and Wnt7a and En1 to specify dorsal-ventral polarity in the developing limb (Riddle et al., 1993; Parr and McMahon, 1995; Chiang et al., 1996; Crossley et al., 1996; Ohuchi et al., 1997; Loomis et al., 1998; Min et al., 1998; Sekine et al., 1999; Lewandoski et al., 2000; Moon and Capecchi, 2000). Interestingly, while many genes required for limb development are expressed in both the fore- and hindlimb, surprisingly little is known about how these genes are coordinated to direct the morphological diversification of these two structures (see reviews by Niswander, 2003; Mariani and Martin, 2003; Logan, 2003; Gurrieri et al., 2002; Tickle, 2003). Gene knockout and ectopic expression studies suggest that Tbx5, Tbx4, and Pitx1 play essential roles in controlling limb development and identity (Logan et al., 1998; Rallis et al., 2003; Naiche and Papaioannou, 2003; Agarwal et al., 2003, Minguillon et al., 2005). However, the regulatory targets of Tbx5, Tbx4, and Pitx1 required for limb differentiation remain to be discovered (Logan, 2003).
To date, studies examining gene expression profiles during forelimb and hindlimb development have been limited to small numbers of genes in cDNA microarrays or a serial analysis of gene expression (SAGE), which lacked the statistical sensitivity to detect many of the genes known to be expressed in a differential manner including: Tbx5, HoxC10, HoxD11, and HoxD9 (Qin et al., 2002; Margulies et al., 2001). To advance our understanding of the molecular mechanisms participating in the formation of the fore- and hindlimb, a genome-wide microarray approach was used to identify genes and gene combinations functioning during post-specification limb development. Using Affymetrix Murine GeneChips (MOE-430-A and B), the genome-wide gene expression profiles of developing fore- and hindlimb autopods were compared at embryonic day 12.5. From this analysis, significant differences (P ≤ 0.05) were detected for 44 genes among which 38 were previously unreported for differential expression in the developing limbs. Transcriptome analysis of members of the Fgf, Wnt, Hedgehog, Aldh, Hox, Bmp, and Tbx gene families revealed significant differences in individual and clustered gene expression profiles, indicating that each autopod type expresses a unique combination of genes during differentiation. Together, these findings provide a fundamental resource to identify new candidates for genetically associated limb malformations as well as a means to understand how gene expression is coordinated during limb development.
Gene Expression Profiles: Similarities and Differences Among Forelimb, Hindlimb, and Brain Transcriptomes
Transcriptomes of mouse embryonic day-12.5 autopod and adult brain were generated using three independent hybridizations of Affymetrix MOE430 GeneChips. In sum, 26,179 transcripts representing 57.9% of the probe sets on the MOE430 GeneChip were detected in the limb autopods (Table 1A). A comparison of fore- and hindlimb-enriched transcripts revealed a remarkable degree of conservation between the two limb types with a Pearson Correlation Coefficient of 0.997 (Fig. 1A). Similarly, false discovery rate (FDR) analysis of the forelimb and hindlimb transcriptomes confirmed that the majority of the detected transcript signals are statistically identical between the fore- and hindlimb samples (FDR = 0.01; q-value ≥ 0.05, data not shown), suggesting that limb-type-specific differentiation may result from the differential utilization of commonly expressed genes, as well as the relatively small number of genes that are expressed in a specific limb type. In contrast, a comparison of limb versus brain transcriptomes revealed significantly different sets of transcripts being expressed in the two tissue types, producing a Pearson Correlation Coefficient of 0.674 (Fig. 1B) and FDR q-value <0.05. Analysis of representative fore- and hindlimb sections confirmed the site of dissection included only autopod tissues (Fig 1C).
Next, to test the fidelity of the limb transcriptome analysis to detect differential gene expression, we examined the expression of Tbx5, Tbx4, Hoxc10, and Pitx1, which are differentially expressed during fore- and hindlimb development (Logan and Tabin; 1999, Nelson et al., 1996). Expression analysis of these genes revealed greater than 2-fold differences between the forelimb and hindlimb levels (P ≤ 0.05) (Table 1B), confirming the sensitivity of the microarray transcriptome analysis to detect differential gene expression, while providing a quantitative measurement of Tbx5, Tbx4, Hoxc10, Pitx1 expression in the developing autopod. Interestingly, transcriptome analysis also detected differential expression of Hoxd9 and Hoxd11, which exhibited changes greater than 1.2-fold between autopod types (P ≤ 0.05) (Table 1B, Fig. 2).
Pair-wise analysis of forelimb versus hindlimb transcript signals revealed 44 transcripts exhibiting more than a 2-fold difference (P < 0.05), whereas at fold changes less than two, 816 transcripts were detected as differentially expressed between the fore- and hindlimb autopods (P < 0.05) (Tables 2 and 3). Among the 44 transcripts exhibiting differential expression were 18 expressed sequence tags (ESTs). BLAST (NCBI) analysis of these ESTs revealed no functional annotation for any of the expressed sequences, suggesting that several additional genes may be differentially expressed during forelimb and hindlimb differentiation (data not shown).
|Criteria||Probe set number*||Percent of total probe sets**|
|P < 0.01, >1.2 fold†||145||0.523958|
|P < 0.01, >2 fold||13||0.046976|
|P < 0.05, >1.2 fold||816||2.948616|
|P < 0.05, > 2 fold||44||0.158994|
|Probe seta||Gene||Ontology, protein domain||Accession||Forelimb mean||Hindlimb mean||Fold change||P value||F P call (%)||H P call (%)|
|1425694_atfl||T-box 5||Transcription factor, T-box domain||AF140427||351.9||7.3||−48.21||0.002685||100||0|
|1425695_atfl||T-box 5||Transcription factor, T-box domain||AI425735||898.93||54.33||−16.54||0.000328||100||0|
|1437933_atfl||Hedgehog-interacting protein||EGF-like domain||BB040396||136.57||37.73||−3.62||0.02164||66||0|
|1421426_atfl||Hedgehog-Interacting protein||Signal transduction, EGF-like domain||NM_020259||159.4||48.07||−3.32||0.049692||100||100|
|1422667_atfl||Keratin complex 1, acidic, gene 15||Cytoskeleton organization,||NM_008469||205.63||64.3||−3.2||0.036873||100||0|
|1438083_atfl||Hedgehog-interacting protein||Signal transduction, EGF-like domain||BB773386||76.5||24.4||−3.14||0.043136||100||33|
|1451440_atfl||Chondrolectin||Binding activity, C-type lectin||AF311699||79.9||25.9||−3.08||0.032254||100||66|
|1430788_at||RIKEN cDNA 4930441F12 gene||Unknown||AK019605||15.87||6||−2.64||0.034777||100||0|
|1436987_atfl||RIKEN cDNA 5430433G21 gene||Unknown||AV293123||35.63||14.23||−2.5||0.010624||100||33|
|1458000_at||desmoglein 1||Cell adhesion molecule activity||BB151286||11.97||4.83||−2.48||0.028501||66||0|
|1456312_x_at||Gelsolin||Cytoskeletal protein-binding activity||AV224521||302.97||124.13||−2.44||0.017268||100||100|
|1459101_at||ESTs, weakly similar to L1 repeat||Unknown||BG066299||17.83||7.43||−2.4||0.036299||66||0|
|1425100_a_at||Phosphodiesterase 6G, cGMP-specific||Enzyme activity||BC014723||72.2||30.5||−2.37||0.027132||66||0|
|1436363_a_at||Nuclear factor I/X||Transcription factor, CTF/NF-I family domain||AW049660||768.97||330.9||−2.32||0.003996||100||100|
|1435802_at||Mus musculus 15 days embryo male testis cDNA||Unknown||BE134048||177.33||77.1||−2.3||0.036933||100||100|
|1440766_atfl||ESTs, weakly similar to RIKEN cDNA 5730493B19||Unknown||AV001099||190.63||87.5||−2.18||0.026365||100||100|
|1423493_a_at||Nuclear factor I/X||Transcription factor, CTF/NF-I family domain||BB315728||370.1||171.03||−2.16||0.009765||100||100|
|1449417_at||Ameloblastin||Extracellular matrix organization and biogenesis||NM_009664||97.9||45.27||−2.16||0.007133||100||100|
|1433600_at||Adrenergic receptor, alpha 2a||G-protein coupled receptors and signal transducer||BB333433||55.2||26.6||−2.08||0.001507||100||66|
|1426968_a_atfl||RIKEN cDNA 3110069K09 gene/Rdh10||Enzyme activity, glucose/ribitol dehydrogenase||BG073496||802.6||391.37||−2.05||0.023219||100||100|
|1422370_at||Olfactory receptor 49||Receptors and signal transducer||NM_010991||66.37||32.57||−2.04||0.014414||100||66|
|1436364_x_at||Nuclear factor I/X||Transcription factor, CTF/NF-I family domain||AW049660||426.13||208.87||−2.04||0.004621||100||100|
|1418203_athl||Phorbol-12-myristate-13-acetate-induced protein 1||Apoptosis||NM_021451||140.33||281.33||2||0.026408||100||100|
|1433434_at||Hypothetical protein 3222402N16||35.57||75.3||2.12||0.042872||66||100|
|1416658_at||Frizzled-related protein||Signal transduction, Frizzled CRD region||NM_011356||188.6||406.2||2.15||0.032246||100||100|
|1439798_athl||HOX-C10 (HOX-3.6) homolog||Transcription factor, homeodomain||BB779859||46.33||100||2.16||0.00852||0||66|
|1423473_athl||Septin 2||Cell division/GTP binding protein||AV304911||11.7||26.8||2.29||0.021837||33||100|
|1459344_at||RIKEN cDNA 9630019E01 gene||Unknown||BB456107||10.97||26||2.37||0.015668||0||66|
|1426621_a_at||RIKEN cDNA 2900026H06 gene||G-protein beta WD-40 repeat||BB560759||14.6||36.9||2.53||0.014955||0||100|
|1422606_athl||Collagenous repeat-containing sequence||Extracellular, collagen triple helix repeat||NM_030888||13.53||37.07||2.74||0.000589||100||100|
|1426321_athl||T-box 4||Transcription factor, T-box domain||AY075134||122.5||346.43||2.83||0.034627||100||100|
|1429404_athl||RIKEN cDNA 2010317E24 gene||Unknown||AK008577||14.47||49.67||3.43||0.008288||0||66|
|1438810_athl||DNA segment, Chr 10, ERATO Doi 755, expressed||Unknown||BG069546||23.97||82.67||3.45||0.009855||100||100|
|1458349_s_at||Mouse 10 days neonate skin cDNA/BB278163||Unknown||BB278163||3.17||11.43||3.61||0.02489||0||66|
|1456033_athl||T-box 4||Transcription factor, T-box domain||AV226212||167.9||684.17||4.07||0.007643||100||100|
|1419514_athl||Paired-like homeodomain transcription factor 1||Transcription factor, paired-like homeodomain||U54499||52.33||223.67||4.27||0.046938||0||100|
|1449488_athl||Paired-like homeodomain transcription factor 1||Transcription factor, paired-like homeodomain||NM_011097||84.27||386.83||4.59||0.014211||0||100|
A comparison of brain versus forelimb, hindlimb, and pooled limb transcriptomes revealed that only 24 genes showed significant enrichment in the individual limb autopod types, whereas greater than 5,000 genes were detected as enriched in the brain (fold change >2; P ≤ 0.05) (Table 4). Relaxing the criteria for limb-enrichment (fold change > 1.2, P ≤ 0.05) increased the number of limb-enriched genes to 186 in the forelimb and 205 in the hindlimb (Table 4).
|Tissue||1.2-fold enriched probe sets||2-fold enriched probe sets|
|Probe set number||Percent of total probe setsa||Probe set number||Percent of total probe setsa|
Analysis of Key Developmental Genes During Autopod Differentiation
To discern the developmental programs operating in each autopod-type, we examined whether members of key developmental gene families were differentially expressed in the forelimb or hindlimb autopod including members of the HoxA, HoxB, HoxC, HoxD, T-box, Bone morphogenetic protein (BMP), FGF, Wnt, and hedgehog gene families. For the Hox genes, significant differences in forelimb versus hindlimb expression were detected for Hoxc10, Hoxd9, and Hoxd11, whereas no significant differences in limb-type expression were detected for members of the HoxA and HoxB gene families (P ≤ 0.05) (Fig. 2 and data not shown). Hoxc11 and Hoxc12 were not included in this analysis as neither gene was annotated on the MOE 430-A and -B Gene Chips.
Analysis of the T-box gene family revealed Tbx4 to be the only family member to exhibit significant enrichment in the developing hindlimb (P ≤ 0.05), a finding consistent with previous reports of Tbx4 expression and loss of function phenotype (Fig. 2) (Saito et al., 2002; Naiche and Papaioannou, 2003). In the forelimb, Tbx18 and Tbx20 also exhibited differential expression (P ≤ 0.05), joining Tbx5 as forelimb-enriched T-box genes (Fig. 2).
Analysis of the hedgehog and FGF gene families identified differential expression of hedgehog interacting protein (Hip) and Indian hedgehog (Ihh), whereas no significant differences were detected between fore- and hindlimb expression for sonic hedgehog (Shh) or any FGF family members (P ≤ 0.05) (Fig. 3). A fourth hedgehog family member, Desert hedgehog (Dhh), did not exhibit detectable levels of expression in either the fore- or hindlimb arrays (data not shown). For the BMP family and its receptors, only Bmp7 and Bmp5 exhibited significant enrichment in the forelimb or hindlimb autopods, although several genes did exhibit a generalized enrichment in either limb or brain tissues including: Bmp2, Bmp4, Bmp5, Bmp7, and BmprIa in the limb and Bmp6 in the brain (fold change greater than 2, P ≤ 0.05) (Fig. 4). Genes belonging to the Wnt family did not exhibit significant differences in expression in either autopod type; however, Wnt5a, Wnt5b, Wnt6, and Wnt11 were significantly enriched in the limb versus brain tissues (fold change greater than 2, P ≤ 0.05) (Fig. 4).
Recognizing that the development of several embryonic regions including those forming the eye, kidney, limb, and brain can be compromised by perturbations in retinoic acid or bio-aldehyde metabolism, we examined whether the genes whose products control the processing of these compounds exhibited autopod-specific expression (Choudhary et al., 2005, Waclaw et al., 2004, Niederreither et al., 2002; Abu-Abed et al., 2002; Hayes and Morriss-Kay, 2001; Berggren et al., 2001, Vasilou et al., 2000). From this analysis, we identified differential limb expression for Rdh10, a novel retinol dehydrogenase previously thought to function exclusively in the retina (Wu et al., 2002) (Fig. 5A) (RIKEN cDNA 3110069K09; gb:BG073696). Prior to this investigation, the GeneChip probe set corresponding to Rdh10 (1426968_a_at) was not well annotated. Querying the sequence by BLAST (NCBI), we found it had 98% identity with Rdh10 (NM_133832), suggesting an additional role for retinol-retinal conversion during fore- and hindlimb differentiation. Analysis of other genes functioning during retinoic acid metabolism revealed Rdh5 expression was significantly higher in brain versus limb tissues, whereas Rdh11 was enriched in the autopods independent of limb type (P ≤ 0.05) (Fig. 5A). Interestingly, the genes whose products control retinoic acid catabolism, Cyp26a1 and Cyp26b1, also appeared to be enriched in the limbs. However, no significant differences in their distribution between autopod limb types were detected (Fig. 5A).
Genes required for bio-aldehyde processing also exhibited differential tissue-specific expression with Aldh1a1 being almost exclusively expressed in the brain, whereas Aldh1a2 expression was most notable in the developing limbs (Fig. 5A). Finally, no significant differences between autopod type or brain expression were detected for Aldh1a3, Aldh1a7, Aldh1b1, Aldh2, Aldh7a1, Rdh1, Rdh6, Rdh7, and Aldh9a1 expression (Fig. 5).
Analysis of retinoic acid receptor expression in the fore- and hindlimb autopod also revealed no significant differences in forelimb versus hindlimb expression for RXRα, RXRβ, RARα, RARβ, RARγ-a, and RARγ (P ≤ 0.05) (Fig 5B). However, several retinoic acid receptors were significantly enriched in a limb- or brain-specific manner including: RARα, RARy. RARγ-a in the limb and RXRβ in the brain (P ≤ 0.05) (Fig. 5B).
Hierarchical Clustering Analysis
Although significant differences in individual gene expression were detected in the forelimb, hindlimb, and brain, we also hypothesized that different gene clusters may be functioning in an autopod-type-specific manner. To test this hypothesis, hierarchical clustering analysis was used to detect probe set clusters exhibiting consistent enrichment in each autopod type (>2-fold enrichment, P ≤ 0.05; Fig. 6). Restricting the clustering analysis to limb-enriched probe sets (no brain enrichment) revealed significant clustering for several probe sets present on the MOE 430-A (Fig. 6A) and MOE 430-B (Fig. 6B) GeneChips including Tbx5, Chondrolectin, hedgehog interacting protein(Hip), Ameloblastin, and Keratin Complex 1 in the forelimb (Fig. 6A and B) and Septin2, Pitx1, Tbx4, Phorbol-12-myristate-13-acetate-induced protein 1 (Pmaip1), Cors26, and Hoxc10 in the hindlimb (Fig. 6B). Including the probe sets present both in the brain and limb increased the number of probe sets contributing to the forelimb and hindlimb clusters, adding nuclear factor I/X (Nfix), desmoglein, adrenergic receptor alpha 2A, olfactory receptor 49, phosphodiesterase 6G, and gelsolin to the forelimb cluster and frizzled related protein-b1(frzb) to the hindlimb cluster (Fig. 6C and D).
Validation of the Microarray Data
To confirm the differential expression of forelimb, hindlimb, and brain-enriched genes, semi-quantitative RT-PCR, quantitative real-time PCR (QRT RT-PCR), and in situ hybridization were used to assess the GeneChip results. Using semi-quantitative RT-PCR, 13 genes exhibiting differential expression by GeneChip were randomly selected and assessed for significant enrichment in the forelimb, hindlimb, or brain tissues. Although the magnitude of fold change differences varied between the GeneChip and RT-PCR methods, the overall trends were the same (Table 5). For example, genes that exhibited little or no expression in the autopod or brain by GeneChip analysis (absent: P call <60%) were also not detected by RT-PCR in the corresponding tissues, including: Hoxa11 and hemoglobin Y beta-like embryonic chain(Hbb-y) in the brain and myelin basic protein (Mbp) in the developing limbs (Fig. 7, Table 5). More importantly, genes determined by GeneChip to be differentially expressed in the fore- or hindlimb, consistently demonstrated the same differential expression by RT-PCR (Tables 5 and 6, Fig. 6). For Rdh10, both the GeneChip and RT-PCR methods detected differential expression favoring the forelimb (2.05- and 1.58-fold, P ≤ 0.05). Finally, using QRT RT-PCR, we assessed whether genes exhibiting small but significant differences in autopod type expression (fold change < 2, P ≤ 0.05) consistently demonstrated the same differential expression. In all cases examined, the small changes but significant changes detected by GeneChip were confirmed by QRT RT-PCR including the differential expression of Hoxd9, Hoxd11, Tbx18, Tbx20, Bmp5, and Bmp7 (Table 7).
|Gene name||RTPCR transcripts intensitya||Array transcripts intensityb||Fold changec|
|Hip||81.96||28.16||33.60||159.40||48.07||48.97||F/H = 2.91||F/H = 3.32|
|Rdh10||45.89||29.02||6.86||802.60||391.37||212.93||F/H = 1.58||F/H = 2.05|
|Nfix||140.44||127.11||150.34||768.97||330.90||3,004.00||F/H = 1.1||F/H = 2.3|
|Gsn||91.06||58.00||97.37||302.97||124.13||783.73||F/H = 1.57||F/H = 2.44|
|Chodl||61.82||23.20||9.50||79.90||25.90||32.07||F/H = 2.66||F/H = 3.08|
|Kirt1-15||97.98||74.95||8.98||205.63||64.30||9.73||F/H = 1.31||F/H = 3.19|
|Sept2||139.93||151.13||121.54||11.70||26.8||12.10||H/F = 1.08||H/F = 2.29|
|Pmaip1||35.63||48.82||13.85||140.33||281.33||16.90||H/F = 1.37||H/F = 2.0|
|Frzb||57.74||122.86||65.90||188.60||406.20||280.36||H/F = 2.13||H/F = 2.15|
|Hoxa11||32.84||33.45||0.00||248.6||338.93||98.80||H/Be||H/B = 3.43|
|Hbb-y||116.20||121.83||3.06||2,827.47||3,279.57||4.20||H/B = 39||H/B = 780|
|Cbln1||4.00||3.37||8.11||205.87||413.33||1,580.13||B/H = 2.4||B/H = 3.82|
|Mbp||12.97||11.56||179.63||8.50||3.67||10,885.07||B/H = 15.5||B/H = 2965|
|Probe sets||Genes||Forwared primer 5′-3′||Reverse primer 5-3′||Product length (bp)|
|Gene||Accession||Probe set||Primer1||Primer2||Forelimb/hindlimb fold change|
Next, using in situ hybridization, we examined whether qualitative differences in gene expression patterns could aid in validating the quantitative GeneChip and RT-PCR assessments. Interestingly, while transcripts corresponding to the differentially expressed probe sets were present in the developing autopods, the in situ hybridization method could not be used to assess gene expression levels as many of the genes exhibited diffuse expression throughout the autopod types including Rdh10, Frzb, and Septin2 (Figs. 8 and 9, 10). Hip expression did appear to be differentially expressed in the forelimb autopod, however, this difference could also be attributed to heterochronic differences in forelimb versus hindlimb initiation.
Ontology of Genes Expressed in Developing Limb Autopods
To test whether functional differences exist among forelimb, hindlimb, and brain gene clusters, we examined the ontology of the clustered transcripts exhibiting ≥2-fold differential expression. Interestingly, the most significant ontology groups detected as enriched in the limb were “Transcription factor, T-box” and “Transcription factor activity,” which included Tbx4, Tbx5, Pitx1, and Hoxc10 (Table 3) (P < 0.001, hyper-geometric distribution; Li and Wong, 2001a,b), suggesting that a high degree of transcriptional regulation is required during autopod differentiation.
The limb is one of the most intensely studied structures in developmental biology, yet few investigations have examined how genome function is coordinated during limb development. The E12.5 autopod was selected as the tissue source as many structures exhibiting morphological diversification are actively differentiating in this discrete limb domain. While it is arguable that microarray analysis of younger (E10–11) limb stages would provide insights into the transcriptome required for limb-type specification, it is equally important to discern how these initial specification signals are translated into morphologically distinct structures. Clearly, additional stages of limb development must be analyzed to fully define the developmental processes required for limb specification and differentiation. Using this analysis as a template, investigators can focus their studies on individual limb stages, minimizing individual costs while producing a unified and statistically sound assessment of limb developmental transcriptomes.
The Impact of Heterochrony
Because forelimb initiation precedes hindlimb initiation by roughly one gestational day, it is arguable that some differences in limb-type gene expression reflect the developmental heterochrony of the two limb types rather than functional differences in gene expression. This heterochronic effect would most likely manifest in earlier limb stages, particularly between the times of fore- and hindlimb initiation, and be represented in the GeneChip analysis by large numbers of differentially expressed genes. However, in the E12.5 autopod, the heterochronic effect appears to be reduced, as relatively few genes exhibited differential expression, including Pitx1, which appears to be one of the few master regulators of limb-type identity (Minguillon et al., 2005). Indeed, for the majority of the transcripts detected in the E12.5 forelimb and hindlimb autopods, most were found to be expressed at similar levels (Fig. 1A). Thus, although we cannot exclude the possibility that some differentially expressed probe sets reflect a heterochronic effect, it is more likely that many of these expression differences reflect gene functions operating in a limb-type-specific manner.
Key Criteria for Microarray Analysis Of Developmental Tissues
In general, the identification and characterization of transcriptomes in a developmental tissue requires both the precise excision of the source RNA tissue as well as large quantities of intact RNA to accurately assess the gene expression profiles (Ko, 2001). Furthermore, recent studies suggest that the same microarray manufacturer should be used during the replicate analysis as a high degree of signal discordance exists among the major chipset manufacturers when probed with the same RNA samples (Tan et al., 2003). To meet these criteria, we focused our analysis on the E12.5 autopod using the expression profile of Hoxa13 to define the tissue domain for RNA isolation (Stadler et al., 2001). This approach facilitated the accurate dissection of wild type autopod tissues, producing, for the first time, a statistically identical series of limb expression arrays (Pearson correlation coefficient, 0.997; P < 0.01). While the total number of transcripts detected in the E12.5 forelimb and hindlimb are similar to those detected at E11.5 by SAGE (Margulies et al., 2001), this investigation affords a higher level of sensitivity and statistical relevance as three separate microarray analyses were performed for each autopod type using independently derived tissue and RNA samples. The benefits of this increased sensitivity are illustrated by our detection and biological validation of unique gene clusters functioning during limb-type differentiation. Without replication, most microarray analyses are unable to detect significant differences in gene expression below the 2-fold level (Tanaka, 2000). However, using replicate analyses, significant fold-change differences as low as 1.2 can be detected (Tanaka, 2000). In the developing limb, this increase in sensitivity may be necessary to define the regionalized gene expression profiles controlling key morphological events such as cell adhesion, programmed cell death, cell proliferation, and cell migration (Zou and Niswander, 1996; Zou et al., 1997; Stadler et al., 2001; Knosp et al., 2004).
The number of transcripts functioning in the limb may be higher than reported here, as several unknown ESTs were identified to be functioning in a limb-specific manner. Furthermore, because the stringency of our analysis was quite high (P call>60%), several genes may have been excluded from the identified transcriptomes. Similar studies by Li and Wong (2001a,b) and Irizarry et al. (2003) support this conclusion since stringent “present call” criteria, as well as low gene expression levels, can result in the MAS5.0 algorithm assigning an “absent call” to functionally important expressed sequences.
Limb-Type-Specific Gene Clusters
Genes whose products function within the same transcriptional or developmental pathway can be detected by cluster analysis (Eisen et al., 1998). Our detection of limb-specific clusters containing Tbx4 and Pitx1 (hindlimb) or Tbx5 (forelimb) are consistent with the limb-specific functions of these genes (Logan et al., 1998; Logan and Tabin, 1999; Agarwal et al., 2003; Naiche and Papaioannou, 2003; Rallis et al., 2003; Minguillon et al., 2005). The consistency of the GeneChip analysis to detect genes with known limb-type-specific functions raises the possibility that genes clustering with Pitx1, Tbx4, and Tbx5 may also have limb-specific functions. Here future investigations aimed at determining the function of Septin2, Pmaip1, and Cors26, chondrolectin, Hip, and ameloblastin in the limbs as well as the transcriptional potential of Pitx1 or Tbx-family members to regulate these candidate genes may provide additional insights into the transcriptomes operating during limb development.
Evidence for Limb-Type-Specific Retinoic Acid Metabolism
The biological and functional annotation of genes on the MOE430 array is more inclusive with the recent advances in bioinformatics (Cheng et al., 2004). This increase in functional annotation allowed detection of several new genes within the retinoic acid (RA) signaling family facilitating, for the first time, a limb-type-specific cluster analysis. RA is known to play an important role in limb patterning and development (Yashiro et al., 2004; Tabin, 1995), yet the genes whose products produce or suppress the biologically active forms of RA have been poorly characterized with respect to their coordinated expression in discrete tissues (Zou and Niswander, 1996; Kedishvili et al., 2002; Wu et al., 2002; Yamamoto et al., 1999). Our detection of significant differences in the expression of Aldh1a2, Rdh5, Rdh11, Rdh10, RARα, and RARy in the forelimb versus hindlimb or brain suggests that members of the RA signaling pathway are differentially deployed in the embryo and possibly the limb. The differential response of Aldh1a2 homozygous mutant forelimbs and hindlimbs to RA supplementation supports this conclusion, although further work is required to determine whether differential RA processing exists in each limb type (Niederreither et al., 2002).
Using an advanced Affymetrix MOE430 GeneChip technique, we detected all previously reported limb-type specific genes, confirming the efficacy of high-density gene microarrays to detect, in an unbiased manner, biologically relevant differences in gene expression. More importantly, we also identified several new genes with specific limb-type expression. Together, these findings suggest that the forelimb and hindlimb autopods express unique combinations of genes, providing a correlative insight into the mechanisms required for limb-type identity and coordinated genome deployment.
Wild type mice were derived from a Hoxa13 mouse colony maintained on a mixed 129/C57BL/6 background as described (Stadler et al., 2001). Wild type C57BL/6 mice were used as the tissue source for the brain microarray analysis. All experiments were carried out in accordance with an approved mouse handling protocol.
Tissue Collection and RNA Isolation
Embryonic day (E) 12.5 embryos were collected from pregnant females maintained in a temperature-regulated specific pathogen-free mouse facility operated on a 12-hr light/dark cycle. Embryonic day 0.5 was established at noon the day a vaginal plug was detected. Limb tissues were matched for developmental similarity using the murine limb staging method described by Wanek et al. (1989). Harvested embryos were immediately placed into ice-cold DEPC-treated phosphate buffered saline (PBS), which was replaced with RNA later (Ambion) for microscopic dissection of the limb autopod. Harvested forelimb and hindlimb autopods were stored individually in 1.5-ml centrifuge tubes at −80°C before RNA isolation. Three independent microarray analyses were performed for the forelimb and hindlimb autopod tissues,. Total RNA used for each analysis represented an independent sample derived from eight forelimb or hindlimb autopods. Total RNA samples for the RT-PCR analysis were independently collected from dissected autopods following the same procedure as the microarray samples. Total RNA was isolated using RNA STAT-60 (Tel-Test Inc.) or Trizol (Invitrogen) as described by the manufacturer. Brain total RNA was isolated from homogenized tissue adult mouse brain using the Trizol reagent (Invitrogen) as described by Hitzemann et al. (2003).
Total RNA (0.5 μg) isolated from forelimbs, hindlimbs, or brain tissue was processed into cDNA using the Superscript II as described by the manufacturer (Invitrogen). cDNA produced by reverse transcription was amplified by PCR using primers specific for each candidate gene. A list of PCR primers, amplicon size, and gene annotation is provided in Table 6. PCR controls consisted of identical amplifications using total RNA that was not reverse transcribed (No-RT) as well as separate reactions substituting water for the cDNA template (Negative control). Candidate gene amplicons were produced using 25–30 cycles of amplification before the onset of plateau phase amplification. No amplicons were detected in PCR reactions using the No-RT samples or negative control samples (data not shown). The PCR amplicons were fractionated on 2% agarose-TAE gels and stained with ethidium bromide. Gel images were captured using an ImageMaster VDS system and LISCAP software (Pharmacia Biotech). DNA band intensity was quantified using Image Quant 5.1 software as described by the manufacturer (Molecular Dynamics, Eugene, OR). Control amplifications of GAPDH were assessed for each sample to normalize PCR product intensity for each tissue sample.
Quantitative Real Time RT-PCR
Relative quantitation of gene expression was performed using ABI PRISM 7700 system (ABI Biosystems) and SYBR Green reagent method according to the manufacturer's instructions (PE-Applied Biosytems). Each PCR reaction was 15 μl and the comparative CT method was used to assess fold-change differences following the manufacturer's instructions (PE-Applied Biosystems).
Primers and gene information are provided in Table 7.
In Situ Hybridization
E 12.5 and E 13.5 embryos were processed for in situ hybridization as described by Manley and Capecchi (1995). Riboprobes were produced as described by Morgan et al. (2003) using the primers and gene regions depicted in Table 8.
|Probe sets||Genes||Forward Primer 5′-3′||Reverse Primer 5′-3′||Product length (bp)||Position on (cDNA)|
*Blast analysis of the probe set EST/BG073696 revealed 98% identity with Rdh10. (NM_133832), PCR Primers were designed based on the NM_133832/Rdh10 sequence.
Microarray Experiment and Data Analysis
Limb autopod RNA samples were hybridized to gene microarrays manufactured by Affymetrix (Mouse 430 A and B chips) at the Microarray Core Facility of Oregon Health and Science University using protocols described by the manufacturer (Affymetrix Inc., 2003). Prior to array hybridization, RNA sample quality was confirmed using a bioanalyzer (Agilent) to ensure the integrity of the 18s and 28s rRNA species as described by the manufacturer (Agilent Technologies). Three separate hybridizations consisting of a matched pair of Mouse 430 A and B chipsets were performed for each tissue type. For each hybridization, an independently collected total RNA sample was used. Array experiments and their data quality met Affymetrix array QC criteria (Affymetrix Inc., 2002b, 2003). MAS5.0 software (Affymetrix) was used to obtain and to normalize the gene expression signals. The signals were normalized using global scaling to target signals of 350 and 70 for 430A and 430B gene chips, respectively. Normalization of each array data set was performed by the OHSU Microarray Core Facility. Microarray data was analyzed using MAS5.0, dChip1.3, SPSS-11.0, Microsoft Excel and Access (Affymetrix Inc., 2002a) (Li and Wong, 2001a,b). SPSS and Excel were used to complete general data statistics and graphs. MAS 5.0 expression signals of all our arrays were organized in Excel and transferred into dChip using the “reading external data” function. dChip was employed to perform transcript ontology analysis and clustering analysis with Euclidean distance (centroid linkage). To statistically compare transcript differences between different tissues, a t-test with fold change, P value, and “present” call criteria of the dChip program was used. The transcript was called “present” in a tissue when the average P call of the three replicate arrays was larger than 60% (P call >60%). Transcript signal intensities detected in the developing forelimb were compared to corresponding signals in both brain and hindlimb to determine which genes or expressed sequences exhibited statistically significant enrichment in forelimb, hindlimb, or brain-specific tissues. The Affymetrix MOE 430 A and B GeneChips utilize oligonucleotide probe sets to detect and designate known genes, ESTs, and RIKEN cDNAs. These probe sets represent the complementary sequences to transcripts isolated from the autopod tissues. To simplify the nomenclature used in text as well as to facilitate the autopod type comparisons, we elected to use current gene names or the term “transcript” interchangeably with the corresponding probe sets whenever possible. The autopod GeneChip data sets have been formatted to meet the MIAME standard, and have been submitted for public access using the Gene Expression Omnibus: GEO Series Accession number: GSE2560.
The authors thank Diane Sexton and Emily Morgan for expert technical assistance. This study was supported in part by a Shriners Hospital Postdoctoral Research Fellowship (S. Shou) and a grant from Shriners Hospital for Children Research (H.S.S.).
- 2002. Differential expression of the retinoic acid-metabolizing enzymes CYP26A1 and CYP26B1 during murine organogenesis. Mech Dev 110: 173–177. , , , , , .
- 2002a. Microarray Suite 5.0, User's documents. .
- 2002b. GeneChip Expression analysis data analysis fundamentals. .
- 2003. GeneChip Expression analysis technical manual.
- 2003. Tbx5 is essential for forelimb bud initiation following patterning of the limb field in the mouse embryo. Development 130: 623–633. , , , , , , , .
- 2001. Expression and regulation of the retinoic acid synthetic enzyme RALDH-2 in the embryonic chicken wing. Dev Dyn 222: 1–16. , , , .
- 2004. NetAffx Gene Ontology Mining Tool: a visual approach for microarray data analysis. Bioinformatics 20: 1462–1463. , , , , , , , , , , , .
- 1996. Cyclopia and defective axial patterning in mice lacking Sonic hedgehog gene function. Nature 383: 407–413. , , , , , , .
- 2005. Role of aldehyde dehydrogenase isozymes in the defense of rat lens and human lens epithelial cells against oxidative stress. Invest Ophthalmol Vis Sci 46: 259–267. , , , , , , .
- 1996. Roles for FGF8 in the induction, initiation, and maintenance of chick limb development. Cell 84: 127–136. , , , .
- 1998. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 95: 14863–14868. , , , .
- 2002. Limb anomalies: Developmental and evolutionary aspects. Am J Med Genet 115: 231–244. , , , .
- 2001. Retinoic acid specifically downregulates Fgf4 and inhibits posterior cell proliferation in the developing mouse autopod. J Anat 198: 561–568. , .
- 2003. A strategy for the integration of QTL, gene expression, and sequence analyses. Mamm Genome 14: 733–747. , , , , , , , , , , , , , , , , .
- 2003. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 31: e15. , , , , , .
- 1999. The anatomical basis of mouse development. New York, NY: Academic Press. 291 p. ,
- 2002. Evidence that the human gene for prostate short-chain dehydrogenase/reductase (PSDR1) encodes a novel retinal reductase (RalR1). J Biol Chem 277: 28909–28915. , , , , , , , .
- 2004. HOXA13 regulates the expression of bone morphogenetic proteins 2 and 7 to control distal limb morphogenesis. Development 131: 4581–4592. , , , .
- 2001. Embryogenomics: developmental biology meets genomics. Trends Biotechnol 19: 511–518. .
- 2000. Fgf8 signalling from the AER is essential for normal limb development. Nat Genet 26: 460–463. , , .
- 2001a. Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application. Genome Biol 2: RESEARCH0032. , .
- 2001b. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc Natl Acad Sci USA 98: 31–36. , .
- 2004. A mixture model for estimating the local false discovery rate in DNA microarray analysis. Bioinformatics 20: 2694–2701. , , , .
- 2003. Finger or toe: the molecular basis of limb identity. Development 130: 6401–6410. .
- 1999. Role of Pitx1 upstream of Tbx4 in specification of hindlimb identity. Science 283: 1736–1739. , .
- 1998. Differential regulation of T-box and homeobox transcription factors suggests roles in controlling chick limb-type identity. Development 125: 2825–2835. , , .
- 1998. Analysis of the genetic pathway leading to formation of ectopic apical ectodermal ridges in mouse Engrailed-1 mutant limbs. Development 125: 1137–1148. , , , , .
- 1995. The role of Hoxa-3 in mouse thymus and thyroid development. Development 121: 1989–2003. , .
- 2001. A comparative molecular analysis of developing mouse forelimbs and hindlimbs using serial analysis of gene expression (SAGE). Genome Res 11: 1686–1698. , , .
- 2003. Deciphering skeletal patterning: clues from the limb. Nature 423: 319–325. , .
- 1998. Fgf-10 is required for both limb and lung development and exhibits striking functional similarity to Drosophila branchless. Genes Dev 12: 3156–3161. , , , , , , , .
- 2005. Tbx5 and Tbx4 are not sufficient to determine limb-specific morphologies but have common roles in initiating limb outgrowth. Dev Cell 8: 75–84. , , .
- 2000. Fgf8 is required for outgrowth and patterning of the limbs. Nat Genet 26: 455–459. , .
- 2003. Loss of Bmp7 and Fgf8 signaling in Hoxa13-mutant mice causes hypospadia. Development 130: 3095–3109. , , , .
- 2003. Loss of Tbx4 blocks hindlimb development and affects vascularization and fusion of the allantois. Development 130: 2681–2693. , .
- 1996. Analysis of Hox gene expression in the chick limb bud. Development 122: 1449–1466. , , , , , , , , , .
- 2002. Embryonic retinoic acid synthesis is required for forelimb growth and anteroposterior patterning in the mouse. Development 129: 3563–3574. , , , , .
- 2003. Pattern formation: old models out on a limb. Nat Rev Genet 4: 133–143. .
- 1997. The mesenchymal factor, FGF10, initiates and maintains the outgrowth of the chick limb bud through interaction with FGF8, an apical ectodermal factor. Development 124: 2235–2244. , , , , , , , , , , , .
- 1995. Dorsalizing signal Wnt-7a required for normal polarity of D-V and A-P axes of mouse limb. Nature 374: 350–353. , .
- 2003. Statistical implications of pooling RNA samples for microarray experiments. BMC Bioinformat 4: 26. , , , , , .
- 2002. PBX, MEIS, and IGF-I are potential mediators of retinoic acid-induced proximodistal limb reduction defects. Teratology 66: 224–234. , , , , .
- 2003. Tbx5 is required for forelimb bud formation and continued outgrowth. Development 130: 2741–2751. , , , , , , , .
- 1993. Sonic hedgehog mediates the polarizing activity of the ZPA. Cell 75: 1401–1416. , , , .
- 2002. Specification and determination of limb identity: evidence for inhibitory regulation of Tbx gene expression. Development 129: 211–220. , , , , .
- 1999. Fgf10 is essential for limb and lung formation. Nat Genet 21: 138–141. , , , , , , , , , , .
- 1990. Transient expression of a cell surface heparan sulfate proteoglycan (syndecan) during limb development. Dev Biol 140: 83–92. , , , , .
- 2001. Loss of Eph-receptor expression correlates with loss of cell adhesion and chondrogenic capacity in Hoxa13 mutant limbs. Development 128: 4177–4188. , , .
- 2003. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A 100: 9440–9445. , .
- 1995. The initiation of the limb bud: growth factors, Hox genes, and retinoids. Cell 80: 671–674. .
- 2003. Evaluation of gene expression measurements from commercial microarray platforms. Nucleic Acids Res 31: 5676–5684. , , , , , , , , .
- 2000. [Transcriptome analysis and pharmacogenomics]. Nippon Yakurigaku Zasshi 116: 241–246. .
- 2003. Patterning systems: from one end of the limb to the other. Dev Cell 4: 449–458. .
- 1976. Positional signalling by mouse limb polarising region in the chick wing bud. Nature 259: 396–397. , , , .
- 2004. Role of human aldehyde dehydrogenases in endobiotic and xenobiotic metabolism. Drug Metab Rev 36: 279–299. , , .
- 2004. The homeobox gene Gsh2 is required for retinoid production in the embryonic mouse telencephalon. Development 131: 4013–4020. , , .
- 1989. A staging system for mouse limb development. J Exp Zool 249: 41–49. , , , , .
- 2002. Cloning and characterization of a novel all-trans retinol short-chain dehydrogenase/reductase from the RPE. Invest Ophthalmol Vis Sci 43: 3365–3372. , , , , , , .
- 1999. Mutations in the gene encoding 11-cis retinol dehydrogenase cause delayed dark adaptation and fundus albipunctatus. Nat Genet 22: 188–191. , , , , , .
- 2004. Regulation of retinoic acid distribution is required for proximodistal patterning and outgrowth of the developing mouse limb. Dev Cell 6: 411–422. , , , , , , , , .
- 1996. Requirement for BMP signaling in interdigital apoptosis and scale formation. Science 272: 738–741. , .
- 1997. Distinct roles of type I bone morphogenetic protein receptors in the formation and differentiation of cartilage. Genes Dev 11: 2191–2203. , , , .