TY - JOUR
AU - Hales, Barbara F.
AB - Abstract Bioactive retinoids are potent limb teratogens, upregulating apoptosis, decreasing chondrogenesis, and producing limb-reduction defects. To target the origins of these effects, we examined gene expression changes in the developing murine limb after 3 h of culture with teratogenic concentrations of vitamin A. Embryonic day 12 CD-1 limbs were cultured in the absence or presence of vitamin A (retinol acetate) at 1.25 and 62.5μM (n = 5). Total RNA was used to probe Atlas 1.2 cDNA arrays. Eighty-one genes were significantly upregulated by retinol exposure; among these were key limb development signaling molecules, extracellular matrix and adhesion proteins, oncogenes, and a large number of transcriptional regulators, including Eya2, Id3, Snail, and Hes1. To relate these expression changes to teratogenic outcome, the response of these four genes was assessed after culture with vitamin A and retinoid receptor antagonists that are able to rescue retinoid-induced malformations; expression levels were correlated with limb malformations. Lastly, pathways analysis revealed that a large number of the genes significantly affected by retinoid treatment are functionally linked through direct interactions. Several regulatory gene cascades emerged relevant to morphogenesis, cell-fate, and chondrogenesis; moreover, members of these cascades crosstalk with one other. These results indicate that retinoids act in a coordinated fashion to disrupt development at multiple levels. In sum, this work proposes several unifying mechanisms for retinoid-induced limb malformations, identifies novel retinoid targets, and highlights Eya2, Id3, Snail, and Hes1 as potential key teratogenic effectors. limb development, apoptosis, chondrogenesis, retinoid receptor antagonists, gene expression, teratogenicity Vitamin A (retinol) and its metabolites, particularly the major bioactive metabolite, retinoic acid, are essential endogenous regulators of morphogenesis at physiological levels, but in excess are teratogenic to practically every developing organ system (Geelen, 1979). Due to their diverse application, the use of retinoids and their derivatives is blossoming; exposure remains high in spite of cautionary measures and pregnancy prevention programs (Honein et al., 2001). Exposed fetuses have a 25-fold increased risk of congenital malformation. In both animals and humans, the pattern of anomalies induced depends on the developmental stage at the time of exposure (Lammer et al., 1985; Shenefelt, 1972). The limb is a useful experimental model for studying how teratogens disturb the pathways of normal embryogenesis. During limb development, the undifferentiated cells of the limb bud mesenchyme are imparted with a specific three-dimensional positional identity, subsequently differentiating into tendon, cartilage, and muscle in the proper pattern. All the while the limb is elongating and limb shape is refined by selective programmed cell death by apoptosis (for review see, Francis-West and Tickle, 1996). Retinoids are crucial mediators of normal limb development at several stages (Mic et al., 2004; Niederreither et al., 2002), but in excess they interfere with the signals orchestrating these processes in time and space (for review see, Lee et al., 2004). In animal studies, exposure to excessive levels of retinoids during mid-organogenesis results in reductive limb defects (Kochhar, 1973). Numerous studies have shown that retinoids cause changes in cell proliferation (Kochhar, 1985) and cell adhesion (Kwasigroch and Kochhar, 1980), upregulate apoptosis (Zakeri and Ahuja, 1994), and decrease chondrogenesis in the developing limb bud (Kwasigroch et al., 1986), but the molecular mediators underpinning these alterations have yet to be clearly defined. Retinoids mediate most of their biological activity by binding to a heterodimer consisting of two subclasses of nuclear receptors, the retinoic acid receptors (RARs) and the retinoid X receptors (RXRs) (Mangelsdorf et al., 1995). Receptor heterodimers exist bound to DNA retinoic acid response elements (RAREs) present in target gene regulatory regions; ligand-binding switches the heterodimer from a transcriptional repressor mode to transactivation, by releasing corepressors and recruiting coactivators and basic transcriptional machinery (for review see, Bastien and Rochette-Egly, 2004). More than 500 genes have been identified as retinoid-regulated in some way (Balmer and Blomhoff, 2002); recent studies have begun to link retinoid-inducible gene expression with specific teratogenic outcomes in the embryo (Qin et al., 2002; Williams et al., 2004). Retinol induces dose-dependent decreases in limb outgrowth and chondrogenesis, and upregulates apoptosis in susceptible regions of the limb during organogenesis (Ali-Khan and Hales, 2006). To investigate how retinoids mediate their teratogenic effects during limb development, we have analyzed gene expression changes after vitamin A (all-trans retinol acetate; retinol) treatment of the mid-organogenesis mouse limb in vitro. To target early changes, exposure was limited to 3 h. In order to relate these changes in gene expression to teratogenic outcomes, we cultured limbs with retinoid receptor antagonists that have been previously shown to rescue limb bud apoptosis and morphology (Ali-Khan and Hales, 2006), as well as retinol, and assessed the expression of four transcriptional regulators, Id3, Snai1, Hes1, and Eya2 by real time qRT-PCR. MATERIALS AND METHODS Limb bud cultures and drug treatments. Limb bud cultures were done as previously described (Ali-Khan and Hales, 2003). Briefly, embryos were dissected from timed-pregnant ED12 CD1 mice, forelimbs were excised and cultured. All-trans retinol acetate (Sigma, St Louis, MI) dissolved in 100% ethanol was added to the culture medium. For full investigation of the teratogenicity of retinol, we selected 1.25μM (1 IU) and 62.5μM (50 IU) for use in this study. Limb malformations are relatively mild at 1.25μM (1 IU); after 6 days in culture, treated limbs show some decrease in chondrogenesis, reduced limb outgrowth, and dysmorphogenic skeletal elements (shortened ulna and radius, and some missing, fused or indistinct phalanges and carpalia). In contrast, after culture for 6 days with 62.5μM (50 IU) retinol, treated limbs show severe defects; limb outgrowth and chondrogenesis are seriously retarded (Ali-Khan and Hales, 2003). The total volume of ethanol in vitro did not exceed 0.5% vol/vol. BMS 189453 (BMS 453), a competitive pan-RAR antagonist, a gift from Bristol-Myers Squibb (Wallingford, CT) and HX603, a pan-RXR antagonist, a gift from Dr H. Kagechika, University of Tokyo, were dissolved in DMSO and added to the designated cultures with a final percentage of DMSO of 0.05 or less. Stock solutions of BMS 453 and HX603 were stored, protected from light, at − 20°C between usages. A control study examining the effects of DMSO and ethanol, together and separately, on limb development showed that ethanol at the concentrations used in this work had no significant effect on limb development (assessed using the Limb Morphogenetic Differentiation Scoring System developed by Neubert and Barrach, 1977), or quantification of limb area or percent cartilage. DMSO produced small but significant reductions in limb score and limb area of 3% (p = 0.04) and 2% (p = 0.02), respectively; there was no statistically significant interaction between the two vehicles when they were added to cultures together (for details see, Ali-Khan and Hales, 2006). All animal studies complied with the guidelines set by the Canadian Council on Animal Care. Five separate replicates were completed for cDNA Atlas Array and another five separate replicates for real-time qRT-PCR. RNA extraction. Limbs were removed after 3 h of culture, rinsed with PBS, and stored at − 20°C in RNAlater RNA Stabilization Reagent (Qiagen, Mississauga, ON, Canada). Total RNA from the homogenized limbs was extracted using the RNeasy Mini Kit (Qiagen) for cDNA Atlas Arrays, or using the RNeasy Micro Kit (Qiagen) for real-time qRT-PCR, following the manufacturer's guidelines. The extracted total RNAs were assessed for quality and quantity by formaldehyde agarose gel electrophoresis and spectrophotometric ultraviolet absorbance at 260/280 nm, respectively. Probe preparation and hybridization to cDNA Atlas Arrays. RNA samples were used to probe cDNA arrays (Atlas Mouse 1.2 Array, Clontech Laboratories, Inc, Palo Alto, CA) according to the manufacturer's instructions. RNA (5 μg) was reverse transcribed using Moloney-murine-leukemia virus reverse transcriptase and radiolabeled with [α-32P]dATP (Amersham Pharmacia Biotech, Baie d'Urfé, PQ, Canada; 10 μCi/μl). Labeled cDNAs were purified from unincorporated 32P nucleotides by filtering through CHROMA SPIN-200 DEPC-H2O columns (Clontech). After prehybridization, the probe was added to nylon cDNA arrays spotted with 1176 cDNAs; hybridization was allowed to occur overnight with continuous agitation at 68°C. After 18 h, the arrays were washed three times for 30 min with washing solution 1 (2× sodium chloride/sodium citrate (SSC), 1% SDS) and once for 30 min with washing solution 2 (0.1× SSC, 0.5% SDS) with continuous agitation at 68°C. After a final wash with 2× SSC at room temperature, the membranes were wrapped and exposed to a phosphorimager plate for 24 h. Five arrays, each the result of a separate independent culture experiment, were completed for each treatment. Analysis of gene expression. Arrays were visualized after scanning with a phosphorimager (Storm; Molecular Dynamics, Sunnyvale, CA), and images were imported to AtlasImage software (version 2.0; Clontech) for quantification. The raw data for each gene (intensity minus the background) generated by AtlasImage software were imported into GeneSpring (version 7.2; Agilent Technologies, Palo Alto, CA) for further analysis. For each individual array, the gene detection threshold was set at a raw signal intensity of 2× the background intensity on that array. To minimize experimental variation, expression values were normalized to the median of all measurements on that individual array (standard experiment-to-experiment normalization). Statistical comparisons were made of retinol-treated versus control limbs by Student's t-test (p ≤ 0.05) for genes that had passed a ≥ 1.5-fold change filter in at least three out of five replicates. For K-means cluster analysis, “gene-to-gene normalization” in addition to “experiment-to-experiment normalization” was applied to the data; the gene expression value for each gene was normalized relative to the median of all values taken for that gene in each treatment. This allowed for clearer visualization of cluster patterns with all genes being closer on the vertical axis . Genes were only included in the group of genes upregulated from below threshold (Table 5), if they were significantly upregulated by 1.5-fold or more from their relative intensity value in control (0μM) (p ≤ 0.05); if the control value for these was below the cutoff threshold for expression, the p value is not shown in Table 5. All other genes not expressed in control but expressed after exposure to 1.25 or 62.5μM retinol are listed in the supplementary data. Volcano plots were generated using Genespring GX 7.3. Pathways Assist 3.0 software (Stratagene, La Jolla, CA) was used to visualize direct relationships between genes significantly affected by retinol exposure. Objects were limited to proteins and pathways. Controls were limited to promoter binding, direct regulation, expression, and regulation only. The starting group for analysis was all genes significantly upregulated by retinol (i.e., Tables 2–5345). Real-time qRT-PCR. RNA was diluted to working concentrations of 10 ng/ml and QuantiTect One-Step SYBR Green RT-PCR (Qiagen) was completed using the Roche LightCycler (Roche Diagnostics, Laval, QC, Canada) according to the manufacturer's instructions. PCR thermal cycling parameters were: 95°C for 15 min (1 cycle), 94°C for 15 s, 55°C for 30 s, and 72°C for 20 s (for 35–45 cycles, depending on the primers). Embryonic hindlimb tissues were used to make 1, 25, 50, and 100 ng/ml RNA stocks for standard curves for quantification. RT-PCR primers (Table 1) were designed with Primer3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) and synthesized at the Sheldon Biotechnology Centre (McGill University, Montréal, QC, Canada); expression was normalized against 18S rRNA, and melting curve analyses were done following each PCR to determine the output and detection quality (i.e., formation of primer-dimers). Each treatment consisted of RNA from five separate culture experiments and each RNA sample was measured in duplicate. TABLE 1 RT-PCR Primers Gene name  Symbol  Accession no.  L primer  R primer  DNA-binding protein Id-3  Id3  M60523  ACTCAGCTTAGCCAGGTGGA  GTCAGTGGCAAAAGCTCCTC  Eyes absent homolog 2  Eya2  U71208  TACAGTGCGACCAAGACAGG  CCTCCAGAAAGGCATGTTGT  Snail protein homolog  Snai1  M95604  CTTGTGTCTGCACGACCTGT  CTTCACATCCGAGTGGGTTT  HLH factor hairy and enhancer of split 1  Hes1  D16464  ACACCGGACAAACCAAAGAC  CGCCTCTTCTCCATGATAGG  18S RNA  Rn18s  X00686  AAACGGCTACCACATCCAAG  CCTCCAATGGATCCTCGTTA  Gene name  Symbol  Accession no.  L primer  R primer  DNA-binding protein Id-3  Id3  M60523  ACTCAGCTTAGCCAGGTGGA  GTCAGTGGCAAAAGCTCCTC  Eyes absent homolog 2  Eya2  U71208  TACAGTGCGACCAAGACAGG  CCTCCAGAAAGGCATGTTGT  Snail protein homolog  Snai1  M95604  CTTGTGTCTGCACGACCTGT  CTTCACATCCGAGTGGGTTT  HLH factor hairy and enhancer of split 1  Hes1  D16464  ACACCGGACAAACCAAAGAC  CGCCTCTTCTCCATGATAGG  18S RNA  Rn18s  X00686  AAACGGCTACCACATCCAAG  CCTCCAATGGATCCTCGTTA  View Large Statistical analysis for qRT-PCR. Expression data were analyzed using two-way ANOVA (group 1 = retinol, group 2 = BMS 453 or HX603) and the post hoc Holm-Sidak multiple comparison test (Sigmastat Statistical Software, SPSS Inc, Chicago, IL). Data are expressed as means ± SEM. The minimum level of significance is p < 0.05 for all tests. RESULTS Global Gene Expression Changes Induced by Retinol in the Developing Limb In control limbs (0μM, no exogenously added retinol), 478 (40%) of the 1176 genes present on the array were expressed, attesting to the transcriptionally active nature of this developing tissue (Fig. 1). In limbs exposed to 1.25 or 62.5μM retinol, 31 and 38% of the genes were expressed, respectively. The majority of detected genes were expressed in all three treatment groups (352); however, some genes were exclusively expressed in only one treatment or in two treatments. The list of all the genes expressed in control limbs (Supplementary Data), and those that were exclusively expressed in one or two treatment groups (Supplementary Data and Supplementary Data), can be found in the supplementary data. FIG. 1 View largeDownload slide Global gene expression; number of genes expressed in each different treatment group, 0, 1.25 and 62.5μM retinol, exclusively expressed and expressed in common between treatment groups. FIG. 1 View largeDownload slide Global gene expression; number of genes expressed in each different treatment group, 0, 1.25 and 62.5μM retinol, exclusively expressed and expressed in common between treatment groups. TABLE 2 Genes Significantly Upregulated between 0 and 1.25μM Retinol         Relative intensity      Functional classification  Symbol  Genbank accession  p value  0μM  1.25μM  Fold change  Set  Basic transcription factors                    Eyes absent homolog 2  Eya2  U71208  0.003  2.91  5.74  1.9  4          Relative intensity      Functional classification  Symbol  Genbank accession  p value  0μM  1.25μM  Fold change  Set  Basic transcription factors                    Eyes absent homolog 2  Eya2  U71208  0.003  2.91  5.74  1.9  4  View Large TABLE 3 Genes Significantly Upregulated between 0 and 62.5μM Retinol         Relative intensity      Functional classification  Symbol  Genbank accession  p value  0μM  62.5μM  Fold change  Set  Basic transcription factors                    HLH factor hairy and enhancer of split 1  Hes1  D16464  0.05  1.52  3.02  2.0  3      DNA-binding protein ID3  Id3  M60523  < 0.01  0.94  2.99  3.2  4      Snail protein homolog  Snai1  M95604  0.02  0.83  1.90  2.3  4      TBX2 protein—T-box protein 2  Tbx2  U15566  0.03  1.33  2.73  2.1  3      Homolog of chicken slug zinc finger protein  Snai2  U79550  0.02  0.89  1.69  1.9  3  Transcription activators and repressors                    Caudal type homeobox 2  Cdx2  S74520  0.04  0.76  1.55  2.1  3      Sim transcription factor  Sim2  U42554  0.01  0.65  1.21  1.9  3      Homeobox protein 8  Msx2  X59252  0.02  0.61  1.37  2.3  3      Transcription factor E2F3  E2f3  AF015948  0.05  1.01  1.68  1.7  3  Cyclins                    G1/S-specific cyclin D1  Ccnd1  S78355  0.03  3.07  6.46  2.1  3  Extracellular transporters and carrier proteins                    Apolipoprotein E precursor  Apoe  M12414  0.04  4.74  7.62  1.6  4  Oncogenes and tumor suppressors                    JunD  Jund1  J05205  0.04  0.85  1.39  1.6  4      N-myc proto-oncogene protein  Nmyc1  X03919  < 0.01  0.90  2.12  2.3  3  (Other) receptors                    D-factor/LIF receptor  Lifr  D26177  0.04  0.87  1.48  1.7  3      Lymphotoxin receptor (TNFR family)  Ltbr  U29173  0.01  0.49  0.89  1.8  2  Growth factors, cytokines, and chemokines                    Bone morphogenetic protein 4 precursor  Bmp4  X56848  0.04  0.50  0.91  1.8  4      Bone morphogenetic protein 7  Bmp7  X56906  < 0.01  0.62  1.55  2.5  4      WNT 7B protein precursor  Wnt7b  M89802  < 0.01  0.44  1.08  2.5  3      WNT 11 protein precursor  Wnt11  X70800  0.03  2.87  5.47  1.9  3      Secreted frizzled-related sequence protein 3  Sfrp3  U68058  0.05  0.66  1.09  1.7  2  Intracellular transducers, effectors, and modulators                    Neurogenic locus notch homolog 2  Notch2  D32210  0.02  2.77  5.43  2.0  4      Fibroblast growth factor 3 precursor  Fgfr3  M81342  0.03  0.81  1.73  2.1  4      retinoid × receptor-alpha  Rxra  M84817  0.04  4.39  7.64  1.7  4      Neurogenic locus notch homolog 1 precursor  Notch1  Z11886  0.04  1.10  2.17  2.0  4      syndecan1  Sdc1  Z22532  < 0.01  1.41  2.87  2.0  4      Activin receptor IIB precursor  Acvr2b  M84120  0.01  0.50  0.86  1.7  3      Protein-tyrosine kinase transmembrane receptor  Ror1  AB010383  0.04  0.37  0.62  1.7  4  Intracellular adaptors and receptor-associated proteins                    Crk adaptor protein  Crk  S72408  0.01  2.59  5.37  2.1  3  Intracellular kinase network members                    Mitogen-activated protein kinase p38  Mapk14  U10871  0.04  4.21  7.77  1.8  3      MAPKAPK-2; MAP kinase-activated kinase  Mapkapk2  X76850  0.04  0.57  1.04  1.8  3      MAP kinase kinase 3  Map3k3  U43187  0.04  0.63  1.24  2.0  4      Casein kinase II alpha 1–related sequence 4  Csnk2a1  U51866  0.01  3.25  7.46  2.3  4  Phospholipases and phosphoinositol kinases                    Phosphatidylinositol 3-kinase regulatory alpha subunit  Pik3r1  M60651  0.04  1.00  2.15  2.2  4  Calcium-binding proteins                    Calmodulin  Calm1  X61432  0.04  1.85  3.44  1.9  3  Kinase activators and inhibitors                    58-kDa inhibitor of RNA-activated protein kinase  Dnajc3  U28423  0.01  3.47  6.80  2.0  4  Proteases and protease inhibitors                    Cathepsin D  Ctsd  X53337  0.02  1.61  3.38  2.1  4      72-kDa type IV collagenase type  Mmp2  M84324  0.02  1.74  3.77  2.2  4      Cathepsin B1  Ctsb  M14222  0.04  0.93  2.04  2.2  4      Tissue inhibitor of metalloproteinases 3  Timp3  L19622  0.02  0.95  2.32  2.4  4  Apoptosis-related proteins                    Growth arrest and DNA-damage–induced protein 45  Gadd45a  L28177  0.018  0.46  0.75  1.6  3  Cytoskeleton and motility proteins                    Nidogen precursor  Nid1  X14194  0.01  1.95  4.30  2.2  4      Endoglin precursor  Eng  S69407  0.01  0.98  2.20  2.2  4      Lama5 subunit precursor  Lama5  U37501  < 0.01  0.41  0.73  1.8  3      Kinesin heavy chain member 1B  Kif1b  D17577  0.02  0.32  0.66  2.0  3  DNA damage repair/stress response proteins                    Rad50; DNA repair protein  Rad50  U66887  0.01  0.56  1.05  1.9  3  Functionally unclassified proteins                    Frataxin  Fxn  U95736  0.01  0.58  1.07  1.8  3      Klotho protein  Kl  AB005141  0.02  0.64  1.18  1.8  3  Housekeeping genes                    Phospholipase A2  Ywhaz  D78647  0.02  5.54  22.65  4.1  4      Glyceraldehyde-3-phosphate dehydrogenase  Gadph  M32599  0.04  17.60  40.95  2.3  4      Myosin I alpha  Myo1b  L00923  0.02  0.55  1.24  2.3  3          Relative intensity      Functional classification  Symbol  Genbank accession  p value  0μM  62.5μM  Fold change  Set  Basic transcription factors                    HLH factor hairy and enhancer of split 1  Hes1  D16464  0.05  1.52  3.02  2.0  3      DNA-binding protein ID3  Id3  M60523  < 0.01  0.94  2.99  3.2  4      Snail protein homolog  Snai1  M95604  0.02  0.83  1.90  2.3  4      TBX2 protein—T-box protein 2  Tbx2  U15566  0.03  1.33  2.73  2.1  3      Homolog of chicken slug zinc finger protein  Snai2  U79550  0.02  0.89  1.69  1.9  3  Transcription activators and repressors                    Caudal type homeobox 2  Cdx2  S74520  0.04  0.76  1.55  2.1  3      Sim transcription factor  Sim2  U42554  0.01  0.65  1.21  1.9  3      Homeobox protein 8  Msx2  X59252  0.02  0.61  1.37  2.3  3      Transcription factor E2F3  E2f3  AF015948  0.05  1.01  1.68  1.7  3  Cyclins                    G1/S-specific cyclin D1  Ccnd1  S78355  0.03  3.07  6.46  2.1  3  Extracellular transporters and carrier proteins                    Apolipoprotein E precursor  Apoe  M12414  0.04  4.74  7.62  1.6  4  Oncogenes and tumor suppressors                    JunD  Jund1  J05205  0.04  0.85  1.39  1.6  4      N-myc proto-oncogene protein  Nmyc1  X03919  < 0.01  0.90  2.12  2.3  3  (Other) receptors                    D-factor/LIF receptor  Lifr  D26177  0.04  0.87  1.48  1.7  3      Lymphotoxin receptor (TNFR family)  Ltbr  U29173  0.01  0.49  0.89  1.8  2  Growth factors, cytokines, and chemokines                    Bone morphogenetic protein 4 precursor  Bmp4  X56848  0.04  0.50  0.91  1.8  4      Bone morphogenetic protein 7  Bmp7  X56906  < 0.01  0.62  1.55  2.5  4      WNT 7B protein precursor  Wnt7b  M89802  < 0.01  0.44  1.08  2.5  3      WNT 11 protein precursor  Wnt11  X70800  0.03  2.87  5.47  1.9  3      Secreted frizzled-related sequence protein 3  Sfrp3  U68058  0.05  0.66  1.09  1.7  2  Intracellular transducers, effectors, and modulators                    Neurogenic locus notch homolog 2  Notch2  D32210  0.02  2.77  5.43  2.0  4      Fibroblast growth factor 3 precursor  Fgfr3  M81342  0.03  0.81  1.73  2.1  4      retinoid × receptor-alpha  Rxra  M84817  0.04  4.39  7.64  1.7  4      Neurogenic locus notch homolog 1 precursor  Notch1  Z11886  0.04  1.10  2.17  2.0  4      syndecan1  Sdc1  Z22532  < 0.01  1.41  2.87  2.0  4      Activin receptor IIB precursor  Acvr2b  M84120  0.01  0.50  0.86  1.7  3      Protein-tyrosine kinase transmembrane receptor  Ror1  AB010383  0.04  0.37  0.62  1.7  4  Intracellular adaptors and receptor-associated proteins                    Crk adaptor protein  Crk  S72408  0.01  2.59  5.37  2.1  3  Intracellular kinase network members                    Mitogen-activated protein kinase p38  Mapk14  U10871  0.04  4.21  7.77  1.8  3      MAPKAPK-2; MAP kinase-activated kinase  Mapkapk2  X76850  0.04  0.57  1.04  1.8  3      MAP kinase kinase 3  Map3k3  U43187  0.04  0.63  1.24  2.0  4      Casein kinase II alpha 1–related sequence 4  Csnk2a1  U51866  0.01  3.25  7.46  2.3  4  Phospholipases and phosphoinositol kinases                    Phosphatidylinositol 3-kinase regulatory alpha subunit  Pik3r1  M60651  0.04  1.00  2.15  2.2  4  Calcium-binding proteins                    Calmodulin  Calm1  X61432  0.04  1.85  3.44  1.9  3  Kinase activators and inhibitors                    58-kDa inhibitor of RNA-activated protein kinase  Dnajc3  U28423  0.01  3.47  6.80  2.0  4  Proteases and protease inhibitors                    Cathepsin D  Ctsd  X53337  0.02  1.61  3.38  2.1  4      72-kDa type IV collagenase type  Mmp2  M84324  0.02  1.74  3.77  2.2  4      Cathepsin B1  Ctsb  M14222  0.04  0.93  2.04  2.2  4      Tissue inhibitor of metalloproteinases 3  Timp3  L19622  0.02  0.95  2.32  2.4  4  Apoptosis-related proteins                    Growth arrest and DNA-damage–induced protein 45  Gadd45a  L28177  0.018  0.46  0.75  1.6  3  Cytoskeleton and motility proteins                    Nidogen precursor  Nid1  X14194  0.01  1.95  4.30  2.2  4      Endoglin precursor  Eng  S69407  0.01  0.98  2.20  2.2  4      Lama5 subunit precursor  Lama5  U37501  < 0.01  0.41  0.73  1.8  3      Kinesin heavy chain member 1B  Kif1b  D17577  0.02  0.32  0.66  2.0  3  DNA damage repair/stress response proteins                    Rad50; DNA repair protein  Rad50  U66887  0.01  0.56  1.05  1.9  3  Functionally unclassified proteins                    Frataxin  Fxn  U95736  0.01  0.58  1.07  1.8  3      Klotho protein  Kl  AB005141  0.02  0.64  1.18  1.8  3  Housekeeping genes                    Phospholipase A2  Ywhaz  D78647  0.02  5.54  22.65  4.1  4      Glyceraldehyde-3-phosphate dehydrogenase  Gadph  M32599  0.04  17.60  40.95  2.3  4      Myosin I alpha  Myo1b  L00923  0.02  0.55  1.24  2.3  3  View Large K-Means Cluster Analysis after Retinol Treatment To further characterize the response of the developing limbs to retinoid exposure, K-means cluster analysis was used to visualize the overall trends in gene expression. The starting group for analysis was all detected genes (518). Four distinct expression profiles emerged. Figures 2A, 2C, 2E, and 2G depict the average expression profile for the genes within each set, while Figures 2B, 2D, 2F, and 2H show the functional breakdown of each set. The genes in this study were categorized according to the functional classifications supplied with the Clontech Atlas Mouse 1.2 Arrays. FIG. 2 View large Download slide Gene expression profiles obtained by K-means clustering after exposure to 1.25 or 62.5μM retinol. Patterns of gene expression fell into four main groups. Graphs show average expression profile for the genes in each set; Set 1 (A), progressive decreased expression between 0 and 62.5μM retinol. Set 2 (C), expression decreased between 0 and 1.25μM retinol, and rose between 1.25 and 62.5μM retinol. Set 3 (E), slightly altered expression between 0 and 1.25μM retinol but increased expression between 1.25 and 62.5μM retinol. Set 4 (G), progressive increased expression between 0 and 62.5μM retinol. The number of genes in each set is indicated in brackets. The functional classification of each K-means cluster set is illustrated in pies (B, D, F, and H). Genes have been classified by their primary classification from the Clontech Atlas Mouse 1.2 array gene list. See Supplementary Data for gene names, accession numbers, and symbols. The number of genes in each set is indicated in brackets. FIG. 2 View large Download slide Gene expression profiles obtained by K-means clustering after exposure to 1.25 or 62.5μM retinol. Patterns of gene expression fell into four main groups. Graphs show average expression profile for the genes in each set; Set 1 (A), progressive decreased expression between 0 and 62.5μM retinol. Set 2 (C), expression decreased between 0 and 1.25μM retinol, and rose between 1.25 and 62.5μM retinol. Set 3 (E), slightly altered expression between 0 and 1.25μM retinol but increased expression between 1.25 and 62.5μM retinol. Set 4 (G), progressive increased expression between 0 and 62.5μM retinol. The number of genes in each set is indicated in brackets. The functional classification of each K-means cluster set is illustrated in pies (B, D, F, and H). Genes have been classified by their primary classification from the Clontech Atlas Mouse 1.2 array gene list. See Supplementary Data for gene names, accession numbers, and symbols. The number of genes in each set is indicated in brackets. Set 1 (53 genes) included genes that show a dose-dependent decline in expression from control (0μM) to 62.5μM retinol (Figs. 2A and 2B). These may represent transcripts that are involved in the normal functioning of the developing limb or whose decline contributes to dysmorphogenesis. Interestingly, Set 2 (126 genes) included genes that were downregulated at 1.25μM retinol, but were increased in expression at 62.5μM retinol (Figs. 2C and 2D). This group may represent genes that respond to a relatively smaller retinoid surge, 1.25μM, by decreasing in expression, but then under more severe stress are upregulated. Set 3 (220 genes) (Figs. 2E and 2F) and Set 4 (119 genes) (Figs. 2G and 2H) both included genes that were upregulated by retinoid exposure. The former showed relatively smaller changes in expression between control (0μM) and 1.25μM, but striking induction between 1.25 and 62.5μM retinol, while genes in Set 4 showed more progressive incline from control (0μM) to 62.5μM retinol. Some of the genes in Sets 3 and 4, which showed a dose-dependent response, may be directly regulated by retinoids through RAREs. Strikingly, most of the genes affected by retinol exposure belonged to the same functional groups in all four cluster analysis sets (Figs. 2B, 2D, 2F, and 2H). Gene lists, including accession numbers and the primary functional classifications for each of the four sets, are provided in the Supplementary Data. “Transcription factors,” “transcriptional regulators,” “growth factors, cytokines, chemokines and their receptors, ” “intracellular signaling,” and “cell-cycle-related” transcripts dominated each cluster set. “Oncogenes and tumor suppressors” were also highly featured in the sets that were generally upregulated by retinol, Sets 2, 3, and 4. Statistical Analysis of Gene Expression Changes after Retinol Treatment To select for genes that are more important in initiating limb malformations and are potentially directly regulated by retinoids, the data were filtered for statistically significant expression changes of 1.5-fold or more between treatment groups (Figs. 3A and 3B). In Figure 3, volcano plots show fold change plotted against statistical significance. Comparison of the fold change between control (0μM) and 1.25μM retinol (Fig. 3A) and control (0μM) and 62.5μM retinol (Fig. 3B) clearly shows that although genes were both up and downregulated at 1.25μM retinol, the overriding response to 62.5μM retinol was upregulation. None of the genes on the array were downregulated with statistical significance by the retinol treatments. Only one gene was significantly upregulated between control (0μM) and 1.25μM retinol; Eya2, a transcription factor involved in eye development (1.9-fold) (Fig. 3A) (Table 2), but 50 genes were significantly upregulated between control (0μM) and 62.5μM retinol (Fig. 3B) (Table 3) and 24 genes were significantly upregulated between 1.25 and 62.5μM retinol (Table 4). Fifteen genes were induced from below threshold (Table 5). The majority of these upregulated genes fell into Set 3 or Set 4 of the K-means cluster analysis sets (see final column in Tables 2–434). FIG. 3 View largeDownload slide Volcano plots showing gene expression changes after exposure to (A) 1.25μM and (B) 62.5μM retinol. Statistical significance (−log10[P]) plotted against fold change (log2[fold change]). FIG. 3 View largeDownload slide Volcano plots showing gene expression changes after exposure to (A) 1.25μM and (B) 62.5μM retinol. Statistical significance (−log10[P]) plotted against fold change (log2[fold change]). TABLE 4 Genes Significantly Upregulated between 1.25 and 62.5μM Retinol         Relative intensity      Functional classification  Symbol  Genbank accession  p value  0μM  62.5μM  Fold change  Set  Basic trancsription factors                    DNA-binding protein ID3  Id3  M60523  < 0.01  1.48  2.89  2.0  3  Transcriptional activators and repressors                    Caudal type homeobox 2  Cdx2  S74520  0.04  0.52  1.55  3.0  3      Sim transcription factor  Sim2  U42554  0.02  0.63  1.21  1.9  3      Homeobox protein 8  Msx2  X59252  0.03  0.58  1.37  2.4  3      D-3-phosphoglycerate dehydrogenase  Phgdh  L21027  0.03  0.72  1.47  2.0  3  Cyclins                    G2/M-specific cyclin B2  Ccnb2  X66032  0.04  0.84  1.98  2.3  2      G1/S-specific cyclin D1  Ccnd1  S78355  0.05  2.74  6.46  2.4  3      G1/S-specific cyclin E1  Ccne1  X75888  0.04  0.62  1.36  2.2  3  Cell cycle regulating kinases and CDK inhibitors                    Cyclin-dependent kinases regulatory subunit 2  Cks2  AA289122  0.02  1.09  2.76  2.5  2      Special AT-rich sequence-binding protein 1  Satb1  U05252  < 0.01  0.68  1.72  2.5  2  Oncogenes and tumor supressors                    N-myc proto-oncogene protein  Nmyc1  X03919  0.03  1.03  2.12  2.1  3      Tyro3 precursor  Tyro3  U18342  0.04  0.63  1.90  3.0  3  Heat shock proteins and stress response                    105-kDa heat shock protein  Hsp105  L40406  0.04  1.23  2.42  2.0  3      Etoposide-induced p53 responsive mRNA  Ei24  U41751  0.03  0.84  1.55  1.8  3  Intracellular transducers, effectors, and modulators                    Frizzled-3; disheveled receptor  Fzd3  U43205  0.02  3.24  5.46  1.7  3      retinoid x receptor alpha  Rxra  M84817  0.05  4.89  7.64  1.6  4      Frizzled homolog 6  Fzd6  U43319  0.02  1.35  2.45  1.8  3  Apoptosis-associated proteins                    Survival of motor neuron  Smn  U77714  < 0.01  1.01  2.35  2.3  3      Presenilin 2  Psn  U57324  0.04  0.72  1.55  2.2  3  Intracellular adaptors and receptor-associated                    Crk adaptor protein  Crk  S72408  0.04  2.32  5.37  2.3  3  Intracellular kinase network members                    MAPKAPK-2  Mapkapk2  X76850  0.04  0.53  1.04  2.0  3  Cytoskeletal and motility proteins                    Bone proteoglycan precursor II  Dcn  X53929  0.04  0.62  1.34  2.2  3  Functionally unclassified proteins                    Frataxin  Fxn  U95736  0.04  0.55  1.07  2.0  3          Relative intensity      Functional classification  Symbol  Genbank accession  p value  0μM  62.5μM  Fold change  Set  Basic trancsription factors                    DNA-binding protein ID3  Id3  M60523  < 0.01  1.48  2.89  2.0  3  Transcriptional activators and repressors                    Caudal type homeobox 2  Cdx2  S74520  0.04  0.52  1.55  3.0  3      Sim transcription factor  Sim2  U42554  0.02  0.63  1.21  1.9  3      Homeobox protein 8  Msx2  X59252  0.03  0.58  1.37  2.4  3      D-3-phosphoglycerate dehydrogenase  Phgdh  L21027  0.03  0.72  1.47  2.0  3  Cyclins                    G2/M-specific cyclin B2  Ccnb2  X66032  0.04  0.84  1.98  2.3  2      G1/S-specific cyclin D1  Ccnd1  S78355  0.05  2.74  6.46  2.4  3      G1/S-specific cyclin E1  Ccne1  X75888  0.04  0.62  1.36  2.2  3  Cell cycle regulating kinases and CDK inhibitors                    Cyclin-dependent kinases regulatory subunit 2  Cks2  AA289122  0.02  1.09  2.76  2.5  2      Special AT-rich sequence-binding protein 1  Satb1  U05252  < 0.01  0.68  1.72  2.5  2  Oncogenes and tumor supressors                    N-myc proto-oncogene protein  Nmyc1  X03919  0.03  1.03  2.12  2.1  3      Tyro3 precursor  Tyro3  U18342  0.04  0.63  1.90  3.0  3  Heat shock proteins and stress response                    105-kDa heat shock protein  Hsp105  L40406  0.04  1.23  2.42  2.0  3      Etoposide-induced p53 responsive mRNA  Ei24  U41751  0.03  0.84  1.55  1.8  3  Intracellular transducers, effectors, and modulators                    Frizzled-3; disheveled receptor  Fzd3  U43205  0.02  3.24  5.46  1.7  3      retinoid x receptor alpha  Rxra  M84817  0.05  4.89  7.64  1.6  4      Frizzled homolog 6  Fzd6  U43319  0.02  1.35  2.45  1.8  3  Apoptosis-associated proteins                    Survival of motor neuron  Smn  U77714  < 0.01  1.01  2.35  2.3  3      Presenilin 2  Psn  U57324  0.04  0.72  1.55  2.2  3  Intracellular adaptors and receptor-associated                    Crk adaptor protein  Crk  S72408  0.04  2.32  5.37  2.3  3  Intracellular kinase network members                    MAPKAPK-2  Mapkapk2  X76850  0.04  0.53  1.04  2.0  3  Cytoskeletal and motility proteins                    Bone proteoglycan precursor II  Dcn  X53929  0.04  0.62  1.34  2.2  3  Functionally unclassified proteins                    Frataxin  Fxn  U95736  0.04  0.55  1.07  2.0  3  View Large TABLE 5 Genes Upregulated by Retinol Treatment from below the Threshold of Expression in Any Treatment Group       Relative intensity    Functional classification  Symbol  Genbank accession  0μM  1.25μM  62.5μM  Set  Basic transcription factors                  Homeobox protein 7.1  Msx1  X14759  n/a  n/a  0.80  3      Distaless homeobox protein 3  Dlx3  S81932  n/a  n/a  0.48  3  Transcriptional activators and repressors                  c-Ets2  Ets2  J04103  0.52  n/a  0.91  3      Brain factor 1  Foxg1  U36760  0.59  n/a  0.76  2  Oncogenes and tumor supressors                  c-Cbl proto-oncogene  Cbl  X57111  n/a  n/a  0.79  3      Met proto-oncogene  Met  Y00671  0.56  n/a  0.63  2  Heat shock proteins                  Osp94 osmotic stress protein; hsp70 related  Hspa41  D49482  n/a  n/a  0.59  3  Other intracellular transducers, effectors, and modulators                  Phosphoprotein enriched in astrocytes 15  Pea15  X86694  n/a  0.37  0.61  4      I-kappa B alpha  Nfkbia  U36277  n/a  n/a  0.76  3  Intracellular protein phosphatases                  Protein tyrosine phsophatase  Ptpri  D83966  n/a  n/a  0.94  2  GTPase activity-modulators                  Ran GTPase activating protein 1  Rangap1  U08110  n/a  n/a  0.78  4  Kinase activators and inhibitors                  Cas; Crk-associated substrate (Bcar1)  Crkas  U48853  n/a  n/a  0.70  2  Other stress-response proteins                  Oxidative stress-induced protein mRNA  Sqstm1  U40930  0.46  n/a  0.75  2  Membrane channels and transporters                  Golgi 4-transmembrane spanning transporter  Laptm4a  U34259  0.34  n/a  0.58  2  Housekeeping genes                  Hypoxantine-guanine phosphoribosyltransferase  Hprt1  J00423  n/a  n/a  1.03  4        Relative intensity    Functional classification  Symbol  Genbank accession  0μM  1.25μM  62.5μM  Set  Basic transcription factors                  Homeobox protein 7.1  Msx1  X14759  n/a  n/a  0.80  3      Distaless homeobox protein 3  Dlx3  S81932  n/a  n/a  0.48  3  Transcriptional activators and repressors                  c-Ets2  Ets2  J04103  0.52  n/a  0.91  3      Brain factor 1  Foxg1  U36760  0.59  n/a  0.76  2  Oncogenes and tumor supressors                  c-Cbl proto-oncogene  Cbl  X57111  n/a  n/a  0.79  3      Met proto-oncogene  Met  Y00671  0.56  n/a  0.63  2  Heat shock proteins                  Osp94 osmotic stress protein; hsp70 related  Hspa41  D49482  n/a  n/a  0.59  3  Other intracellular transducers, effectors, and modulators                  Phosphoprotein enriched in astrocytes 15  Pea15  X86694  n/a  0.37  0.61  4      I-kappa B alpha  Nfkbia  U36277  n/a  n/a  0.76  3  Intracellular protein phosphatases                  Protein tyrosine phsophatase  Ptpri  D83966  n/a  n/a  0.94  2  GTPase activity-modulators                  Ran GTPase activating protein 1  Rangap1  U08110  n/a  n/a  0.78  4  Kinase activators and inhibitors                  Cas; Crk-associated substrate (Bcar1)  Crkas  U48853  n/a  n/a  0.70  2  Other stress-response proteins                  Oxidative stress-induced protein mRNA  Sqstm1  U40930  0.46  n/a  0.75  2  Membrane channels and transporters                  Golgi 4-transmembrane spanning transporter  Laptm4a  U34259  0.34  n/a  0.58  2  Housekeeping genes                  Hypoxantine-guanine phosphoribosyltransferase  Hprt1  J00423  n/a  n/a  1.03  4  View Large Significant Upregulation between Control (0μM) and 62.5μM Retinol Among the genes upregulated between control (0μM) and 62.5μM retinol (Table 3) are several well-characterized retinoid targets, including the oncogene Nmyc1, the RARE-regulated T-box transcription factor Tbx2 (Boskovic and Niles, 2004; Wada et al., 1992), and several genes that are involved in the retinoid system/nuclear receptor signaling: RXRα (Rxra), the LIF receptor (Lifr) (Lane et al., 1999), and calmodulin (Calm1), which is able to regulate the activity of the nuclear receptor coactivator CBP (Chawla et al., 1998). Many other genes of interest, some of which have been previously linked to apoptosis or chondrogenesis in the developing limb, were over-expressed as a consequence of retinol exposure. Members of fundamental limb development signaling pathways were a key affected group; multiple genes from the Wnt, BMP, Notch, and FGF families were upregulated, as was as the cell surface receptor syndecan 1 (Sdc1)(see Fig. 4). FIG. 4 View largeDownload slide Probable direct functional linkages between genes significantly upregulated by retinol: all the genes significantly upregulated by retinol in this study were used as the starting group for analysis. Only those that have been directly linked to one another via a primary interaction in published literature to date are shown. Arrows indicate the direction of interactions, + indicates positive influence, T junctions indicate inhibition, color and shape of the linkages indicate the type of interaction between the two gene products; expression; regulation, promoter binding or direct regulation. FIG. 4 View largeDownload slide Probable direct functional linkages between genes significantly upregulated by retinol: all the genes significantly upregulated by retinol in this study were used as the starting group for analysis. Only those that have been directly linked to one another via a primary interaction in published literature to date are shown. Arrows indicate the direction of interactions, + indicates positive influence, T junctions indicate inhibition, color and shape of the linkages indicate the type of interaction between the two gene products; expression; regulation, promoter binding or direct regulation. A number of the most highly upregulated transcripts on the array were transcription-related genes. Many of these are developmentally important; basic Helix-Loop-Helix and homeobox transcription factors, as well as repressors, were highly represented. Also related to morphogenesis were genes involved in protein dynamics and cell-extracellular matrix (ECM) attachment. Proteases (Ctsd, Ctsb, Mmp2) and the inhibitor Timp3 (see Fig. 4), as well as the major ECM components, glycoproteins laminin α5 (Lama5), nidogen (Nid), and the integral membrane protein endoglin (Eng) were all affected. Stress-response and oncogenes related to the regulation of cell-fate decision-making and apoptosis were also affected by retinol exposure in the limbs; these included Jund1, Nmyc1, Rad50, and Dnajc3, as well as cyclin D1 (Ccnd1) and the transcription factor E2F3 (E2f3) (see Fig. 4). A number of intracellular signaling mediators, in addition to the members of major signaling pathways mentioned above were affected, including three Map kinases (see Fig. 4), and the phosphoinositol kinase Pik3r1. Finally, several functionally unclassified and housekeeping genes were upregulated, including phospholipase A2 (Ywhaz), the most highly upregulated gene on the array (4.1-fold) (see Fig. 4). Significant Upregulation between 1.25 and 62.5μM Retinol A number of genes in this group (Table 4) were also present in the previous list (Table 3), while the remainder were part of the cluster analysis Set 2 that were downregulated at 1.25μM. Some may be mediators of stress-management, cell cycle exit and apoptosis, and do not necessarily represent a retinoid-specific response. Accordingly, several cyclins and cell cycle regulating kinases were represented; three were members of Set 2. The apoptosis-inducers survival of motor neuron (Smn) and presenilin (Pse) (Psen2) (see Fig. 4) as well as two stress response genes (Hsp105 and Ei24) were upregulated. Interestingly, presenilin is also involved in Notch and Wnt family signal transduction; two other noncanonical Wnt family members were also highly upregulated (Fzd3 and Fzd6). Genes Induced from below the Detection Threshold Genes in this group (Table 5) were either downregulated to below the threshold of detection at 1.25μM retinol, before being upregulated at 62.5μM (Set 2) or induced de novo by the retinol treatments (Sets 3 and 4). It is tempting to speculate that these latter genes may be especially important in triggering or executing retinoid-mediated apoptosis and teratogenesis. There were many genes of interest in this category. The developmental transcription factor distal-less 3 (Dlx3 [Set 3]) and transcriptional regulator c-Ets2 (Ets2 [Set 3]) (see Fig. 4) were highly induced. Stress-response proteins were affected (Hspa14, Sqstm1 [both Set 2]), as was the death receptor intracellular transducer Pea15 (Set 4). Pathways Analysis Next we employed Pathways Assist software and manual searches to visualize known regulatory relationships between the genes that were significantly affected by retinol in our study. The linkages mapped in Figure 4 show only primary interactions between the depicted genes. This analysis highlighted the fact that a large number of these genes are functionally linked, directly influencing either the expression or regulation of one another. Several regulatory cascades and cohorts of genes emerged that are pertinent to morphogenesis, cell fate decision-making, patterning, and chondrogenesis; moreover, these cohorts crosstalk with one other. For example, the bone morphogenetic protein Bmp4 appears as a major node of activity. Regulating and being regulated by it are the developmental signaling family members Bmp7, Fgfr3, Sdc1, Acvr2b, the homeobox transcription factors Msx1, Msx2, and Mapk14, all of which have been reported to be integral to the regulation of apoptosis and/or chondrogenesis in the limb bud and/or in developing craniofacial structures (Albertson et al., 2005; Dupe et al., 1999; Rodriguez-Leon et al., 1999; Zuzarte-Luis et al., 2004). Through not only Msx1 but also Mapk14, this cohort influences a cell cycle and proliferation-regulating group made up of the cyclins Ccnd1, Ccne1, and the transcription factors Jund1, Ets1, and E2f3. Going a step further, E2f3 regulates Nmyc1, which upregulates Id3. We hypothesize that this latter group and another that feeds back to Ccnd1, the Notch1, Notch2, and Hes1 group, may regulate chondrogenic differentiation in the limb (Light et al., 2005; Vasiliauskas et al., 2003; Watanabe et al., 2003); Notch1 and Notch2 are also mediators of limb bud apoptosis (Pan et al., 2005). Another cohort, Crk, Crkas (Bcar1), Cbl, and the adjacent Timp3 and Mmp2, are involved with cell adhesion and ECM remodeling, crucial aspects of morphogenesis (Abraham et al., 2005; Huang et al., 2002). Taken together, these results lead to the hypothesis that retinoid excess acts in a coordinated fashion, targeting genes at multiple levels of signaling networks. Crosstalk between signaling cascades may propagate these effects, disturbing cellular processes that are crucial to proper morphogenesis. Array Verification by qRT-PCR A search for key molecular mediators of retinoid-induced malformation must link changes in gene expression to abnormal phenotype. To move toward this goal, we selected four genes of particular interest from those significantly upregulated by the teratogenic retinol treatments. Two were involved in our pathways analysis, the transcriptional repressors Id3 and Hes1. Additionally, we included the repressor Snail and the transcription factor Eya2, and verified the retinoid responsiveness of all four by qRT-PCR. Echoing the microarray analysis (Fig. 5), all four genes showed dose-dependent upregulation with retinol treatment (Fig. 6); this upregulation was significant between control (0μM) and 62.5μM retinol. Previous work shows that the pan-RAR and pan-RXR antagonists, BMS 453, and HX603, respectively, revert vitamin A–induced limb defects (including decreased chondrogenesis and upregulated apoptosis), back to the control phenotype. Optimal limb bud rescue in the presence of 1.25μM retinol occurred with 0.5μM BMS 453 or 12μM HX603. These antagonists produced more modest improvements at 62.5μM retinol (Ali-Khan and Hales, 2006). Likewise, in the present work, over-expression of Id3, Snai1, Hes1, and Eya2 was reduced to close to control values at 0.5μM BMS 453 (Figs. 6A, 6C, 6E, and 6G) or 12μM HX603 (Figs. 6B, 6D, 6F, and 6H). Both BMS 453 and HX603 were able to reduce retinol-induced over-expression of Id3, Snai1, Hes1, and Eya2 with similar success, indicating that these four genes are indeed regulated by retinoid signaling through the RAR-RXR heterodimer at some level; whether this is direct, mediated through an RARE, or indirect, mediated through intermediates, remains to be determined. In sum, gene expression analysis in the presence of the antagonists verifies that Id3, Snai1, Hes1, and Eya2 are retinoid-responsive and are upregulated as a prequel to aberrant retinol-induced apoptosis and limb malformations. Their return to baseline expression levels is correlated with reductions in ectopic apoptosis and improvements in limb outgrowth, and chondrogenesis, suggesting that Id3, Snai1, Hes1, and Eya2 are indeed molecular mediators of retinol-induced teratogenesis. FIG. 5 View largeDownload slide Microarray analysis of (A) Id3, (B) Snai1, (C) Hes1, and (D) Eya2 expression in whole limb RNA extracts after 3 h culture with varying concentrations of retinol. Statistical analysis using GeneSpring Student's t-test (p ≤ 0.05); significant change versus control (0μM retinol) is represented by (*). FIG. 5 View largeDownload slide Microarray analysis of (A) Id3, (B) Snai1, (C) Hes1, and (D) Eya2 expression in whole limb RNA extracts after 3 h culture with varying concentrations of retinol. Statistical analysis using GeneSpring Student's t-test (p ≤ 0.05); significant change versus control (0μM retinol) is represented by (*). FIG. 6 View largeDownload slide qRT-PCR analysis of (A) and (B) Id3, (C) and (D) Snai1, (E) and (F) Hes1, (G) and (H) Eya2 expression in whole limb total RNA extracts after 3 h culture with varying concentrations of retinol and BMS 453 (A, C, E, and G) or retinol and HX603 (B, D, E, and H). Statistical analysis with two-way ANOVA and Holm-Sidak multiple comparison test; significant change versus control (0μM retinol), within retinol-treated limbs is represented by (*), p < 0.001; significant change versus 62.5μM retinol within BMS 453- and HX603-treated limbs is represented by (#), p < 0.001. FIG. 6 View largeDownload slide qRT-PCR analysis of (A) and (B) Id3, (C) and (D) Snai1, (E) and (F) Hes1, (G) and (H) Eya2 expression in whole limb total RNA extracts after 3 h culture with varying concentrations of retinol and BMS 453 (A, C, E, and G) or retinol and HX603 (B, D, E, and H). Statistical analysis with two-way ANOVA and Holm-Sidak multiple comparison test; significant change versus control (0μM retinol), within retinol-treated limbs is represented by (*), p < 0.001; significant change versus 62.5μM retinol within BMS 453- and HX603-treated limbs is represented by (#), p < 0.001. DISCUSSION The goal of this study was to uncover potential molecular mechanisms of retinol-induced limb malformation. Exposure to teratogenic concentrations of retinol during mid-organogenesis, normally a period of tight transcriptional control, shifts the expression profiles of large numbers of genes in the developing limb. Despite the relatively short exposure period, 3 h, and the dilution effect of using whole limb homogenates in this study, many of these gene expression changes were statistically significant. The majority of the responder genes in this study did so in a dose-dependent fashion (Set 4 and some genes in Set 3; Tables 2–5345; Fig. 5). These were the ones of most interest to us; many can be linked to the discrete endpoints that are obvious after retinol treatment: decreased chondrogenesis and outgrowth, and increased apoptosis. Most were expressed in control limbs (0μM retinol), implying roles in normal development, where endogenous retinol levels should contribute to their basal expression. Interestingly, a large number of the gene expression changes we noted impinged on apoptosis and chondrogenesis, morphogenesis and cell motility, often through an ability to alter cell-cell and matrix-cell adhesion, or to respond to or transduce such changes. Impaired Condensation Multiple groups have pointed to disturbances in the regulation of chondrogenesis as a key pathological event leading to limb-reduction defects after retinoid exposure (Cho et al., 2003; Kochhar, 1973; Weston et al., 2003). Our results highlight several stages in this process that may be targeted by retinol. Cell condensation is a crucial early step in morphogenesis and tissue formation, including establishment of the limb cartilage anlagen. Cell attachment to other cells, and importantly to the ECM, organizes developing tissues, regulates differentiation and survival, and influences growth factor signaling (for review see, Lonai, 2003). Retinoid exposure altered levels of the key ECM/cell adhesion genes, Lama5, Nid, and Eng, and also Sdc1, which play an important role in epithelial-mesenchymal interactions and chondrogenesis, and facilitate fibroblast growth factor signaling by presenting these molecules to their receptors (Hall and Miyake, 2000; Mundhenke et al., 2002). We hypothesize that changes in these gene products may not only alter tissue architecture and hinder chondrogenic differentiation but also deprive developing tissues of key growth or survival factors. Many other genes involved in cell adhesion were significantly affected by retinol exposure. Alterations in cell surface attachment can activate downstream mediators including kinases, phosphatases, and GTPases (Shum et al., 2003); several of these were upregulated by retinol, some from below threshold (Rangap1 and Ptpri; Table 5). The adaptor Cbl and the transducers Crk and Crkas are known to regulate cell adhesion and motility (Huang et al., 2002; Scaife et al., 2003), while Calm1, Msx1, Msx2, and Snail family Zn finger repressors (Snai1 and Snai2) can all mediate adhesion and cell-sorting through the regulation of cadherins (Lincecum et al., 1998). The latter five have defined roles, not only in chondrogenesis (for review see, Hall and Miyake, 2000; Mototani et al., 2005; Seki et al., 2003) but also in apoptosis in the developing limb and/or craniofacial regions (Dowd et al., 1991; Montero et al., 2001; Zuzarte-Luis et al., 2004). The NFκB inhibitor Nfkbia, in addition to other mechanisms involved in inducing limb malformation, also affects both skeletal and craniofacial development through the regulation of epidermal-mesodermal interactions (Sil et al., 2004). Taken together, the abnormal expression of these gene products could disrupt cell-cell and cell-matrix attachment at multiple levels providing several routes for alterations in limb morphology, chondrogenesis, and apoptosis (see Fig. 4). Impaired Differentiation: Id3, Hes1, and Snai1 Cell adhesion is intimately related to the timing of differentiation. The transcriptional repressors we have highlighted in this study may interfere with chondrogenesis at this switch point; Id3, Hes1, and Snai1 (as well as Snai2) are markers of undifferentiated states. Snai1 and Snai2, expressed in condensed precartilage mesenchyme but not in chondrocytes (Nieto et al., 1992), repress expression of the key cartilage ECM components, collagen type2A and aggreccan (Seki et al., 2003). Similarly, Id3 expression remains high in both the amphibian limb (Shimizu-Nishikawa et al., 1999) and neural crest cells under the influence of Nmyc1 (see Fig. 4), maintaining multipotency until the appropriate time for differentiation (Light et al., 2005). A similar regulatory mechanism might exist in the limb. In humans, Id3 has been functionally linked to proliferation and differentiation in cartilage (Asp et al., 1998). Similarly, the Notch effector Hes1 (see Fig. 4) may determine skeletal element length by regulating the differentiation rate of proliferating chondrocytes; misexpression of the Hes1 homolog in the developing chick results in shortened limbs (Vasiliauskas et al., 2003). Thus, we hypothesize that retinol-mediated over-expression of Snai1, Id3, and Hes1 may perturb the differentiation program of pre-chondrocytes, providing another mechanism that could underlie the decreased cartilage deposition seen in treated limbs. Upregulated Apoptosis Dramatic upregulation of apoptosis is another key event preceding retinoid-induced limb malformations (Ali-Khan and Hales, 2006; Alles and Sulik, 1989; Zakeri and Ahuja, 1994). While this effect is most probably multigenic, Eya2 is an excellent candidate for a key mediator decisive in triggering this process. Recent work confirms that Eya2 is retinoid-responsive and mediates apoptosis in the eye (Matt et al., 2005). Interestingly, when over-expressed, it was able to initiate cytochrome-c efflux from the mitochondria, a key cell survival decision point (Clark et al., 2002). Cytochrome-c is also released during retinoid-induced apoptosis in the limb (Ali-Khan and Hales, 2003). Hence, in our system, we suggest that Eya2 may mediate both normal developmental apoptosis, as well as the aberrant induction seen after over-expression by retinol. Verification of the functional roles for the genes in this study will be hastened by localization of their transcripts in the developing limb. Gene Regulation Transcriptional regulators, secreted growth factors, and signaling transducers, all families of genes capable of propagating widespread downstream change, were highly affected by the retinol exposures. Expression profiles in response to 1.25 and 62.5μM retinol were markedly different, while they were both up and downregulated by 1.25μM retinol (about 10-fold endogenous levels [Horton and Maden, 1995]), suggesting a more specific regulatory response, the vast majority of genes were over-expressed in response to 62.5μM retinol. There were no genes that were downregulated with statistical significance in this study. Timing, as well as different regulatory mechanisms may account for these observations. Several genes can be directly repressed by retinoids through liganded receptors (Kirfel et al., 1997), but for the most part repression is achieved through the induction of intermediates (Balmer and Blomhoff, 2002), most of which have not been identified; the transcriptional repressors we have highlighted in this work are candidates. Alternatively, for some genes, decreased expression may result from antagonism of other transcription factors (Caelles et al., 1997), or squelching of mutually utilized cofactors (Kamei et al., 1996). Similarly, induction of the genes on our lists may occur through upregulation of intermediates (or upstream cascade members; see Fig. 4), through direct RARE-mediated induction of the gene itself, or through less-specific mechanisms. Some of the genes that were induced at 62.5μM retinol, a concentration that induced widespread apoptosis by 24 h, are likely to be nonspecific players, responding to the high concentrations of exogenous retinol. All these influences, however, must factor into retinoid-induced teratogenesis. Teratogenicity of Retinoids Interestingly, some of the gene expression changes we noted after retinoid treatment of the developing limb were similar to those reported after exposure of cranial neural crest cells (CNCCs), to excess retinoic acid (Williams et al., 2004). In that report, isolated CNCCs in culture were treated for 6–48 h. We target more immediate effects with a tighter 3 h exposure window and offer a dose-response perspective on expression changes. Additionally, the variety of cell-types present in the developing limb allowed us to detect potential alterations in chondroprogenitors, and in undifferentiated mesoderm susceptible to retinoid-induced apoptosis. Teratogen-mediated apoptosis and altered chondrogenesis are common endpoints after exposure to developmental toxicants; elucidation of the pathways mediating such events should contribute to our understanding of normal and abnormal development in the limb and other structures. In summary, these studies provide a first critical step in an integrated approach to defining the molecular basis for the teratogenic action of retinoids in the developing limb. Eighty-one retinoid-responsive genes were identified including novel retinoid targets, a number of which have defined functions in morphogenesis, apoptosis, and chondrogenesis. This is the first demonstration that the transcriptional regulators Snai1, Id3, Hes1, and Eya2 are induced during vitamin A–mediated limb malformations. We propose several potential mechanisms for retinoid-induced limb defects, as well as opening new avenues for the elucidation of retinoid function in the developing embryo and in other systems. This work was supported by the Canadian Institutes for Health Research. We are grateful to Amar V. Singh and the Knudsen Laboratory at the Birth Defects Centre, Systems Analysis Laboratory, University of Louisville, Kentucky, for the Pathways Assist analysis, and to Dr Bernard Robaire for critical assessment of the manuscript. We thank Bristol Myers Squibb and Dr Hiroyuki Kagechika for their gifts of BMS 189453 and HX603, respectively. References Abraham D Ponticos M Nagase H Connective tissue remodeling: Cross-talk between endothelins and matrix metalloproteinases Curr. Vasc. Pharmacol.  2005 3 369 379 Google Scholar CrossRef Search ADS PubMed  Albertson RC Payne-Ferreira TL Postlethwait J Yelick PC Zebrafish acvr2a and acvr2b exhibit distinct roles in craniofacial development Dev. Dyn.  2005 233 1405 1418 Google Scholar CrossRef Search ADS PubMed  Ali-Khan SE Hales BF Caspase-3 mediates retinoid-induced apoptosis in the organogenesis-stage mouse limb Birth Defects Res. A Clin. Mol. Teratol.  2003 67 848 860 Google Scholar CrossRef Search ADS PubMed  Ali-Khan SE Hales BF Retinoid receptor antagonists alter the pattern of apoptosis in organogenesis stage mouse limbs Toxicol. Sci.  2006 90 208 220 Google Scholar CrossRef Search ADS PubMed  Alles AJ Sulik KK Retinoic-acid-induced limb-reduction defects: Perturbation of zones of programmed cell death as a pathogenetic mechanism Teratology  1989 40 163 171 Google Scholar CrossRef Search ADS PubMed  Asp J Thornemo M Inerot S Lindahl A The helix-loop-helix transcription factors Id1 and Id3 have a functional role in control of cell division in human normal and neoplastic chondrocytes FEBS Lett.  1998 438 85 90 Google Scholar CrossRef Search ADS PubMed  Balmer JE Blomhoff R Gene expression regulation by retinoic acid J. Lipid Res.  2002 43 1773 1808 Google Scholar CrossRef Search ADS PubMed  Bastien J Rochette-Egly C Nuclear retinoid receptors and the transcription of retinoid-target genes Gene  2004 328 1 16 Google Scholar CrossRef Search ADS PubMed  Boskovic G Niles RM T-box binding protein type two (TBX2) is an immediate early gene target in retinoic-acid-treated B16 murine melanoma cells Exp. Cell Res.  2004 295 281 289 Google Scholar CrossRef Search ADS PubMed  Caelles C Gonzalez-Sancho JM Munoz A Nuclear hormone receptor antagonism with AP-1 by inhibition of the JNK pathway Genes Dev.  1997 11 3351 3364 Google Scholar CrossRef Search ADS PubMed  Chawla S Hardingham GE Quinn DR Bading H CBP: A signal-regulated transcriptional coactivator controlled by nuclear calcium and CaM kinase IV Science  1998 281 1505 1509 Google Scholar CrossRef Search ADS PubMed  Cho SH Oh CD Kim SJ Kim IC Chun JS Retinoic acid inhibits chondrogenesis of mesenchymal cells by sustaining expression of N-cadherin and its associated proteins J. Cell. Biochem.  2003 89 837 847 Google Scholar CrossRef Search ADS PubMed  Clark SW Fee BE Cleveland JL Misexpression of the eyes absent family triggers the apoptotic program J. Biol. Chem.  2002 277 3560 3567 Google Scholar CrossRef Search ADS PubMed  Dowd DR MacDonald PN Komm BS Haussler MR Miesfeld R Evidence for early induction of calmodulin gene expression in lymphocytes undergoing glucocorticoid-mediated apoptosis J. Biol. Chem.  1991 266 18423 18426 Google Scholar PubMed  Dupe V Ghyselinck NB Thomazy V Nagy L Davies PJ Chambon P Mark M Essential roles of retinoic acid signaling in interdigital apoptosis and control of BMP-7 expression in mouse autopods Dev. Biol.  1999 208 30 43 Google Scholar CrossRef Search ADS PubMed  Francis-West P Tickle C Limb development Curr. Top. Microbiol. Immunol.  1996 212 239 259 Google Scholar PubMed  Geelen JA Hypervitaminosis A induced teratogenesis CRC Crit. Rev. Toxicol.  1979 6 351 375 Google Scholar CrossRef Search ADS PubMed  Hall BK Miyake T All for one and one for all: Condensations and the initiation of skeletal development Bioessays  2000 22 138 147 Google Scholar CrossRef Search ADS PubMed  Honein MA Paulozzi LJ Erickson JD Continued occurrence of accutane-exposed pregnancies Teratology  2001 64 142 147 Google Scholar CrossRef Search ADS PubMed  Horton C Maden M Endogenous distribution of retinoids during normal development and teratogenesis in the mouse embryo Dev. Dyn.  1995 202 312 323 Google Scholar CrossRef Search ADS PubMed  Huang J Hamasaki H Nakamoto T Honda H Hirai H Saito M Takato T Sakai R Differential regulation of cell migration, actin stress fiber organization, and cell transformation by functional domains of Crk-associated substrate J. Biol. Chem.  2002 277 27265 27272 Google Scholar CrossRef Search ADS PubMed  Kamei Y Xu L Heinzel T Torchia J Kurokawa R Gloss B Lin SC Heyman RA Rose DW Glass CK et al.   A CBP integrator complex mediates transcriptional activation and AP-1 inhibition by nuclear receptors Cell  1996 85 403 414 Google Scholar CrossRef Search ADS PubMed  Kirfel J Kelter M Cancela LM Price PA Schule R Identification of a novel negative retinoic acid responsive element in the promoter of the human matrix Gla protein gene Proc. Natl. Acad. Sci. U.S.A.  1997 94 2227 2232 Google Scholar CrossRef Search ADS PubMed  Kochhar DM Limb development in mouse embryos. I. Analysis of teratogenic effects of retinoic acid Teratology  1973 7 289 298 Google Scholar CrossRef Search ADS   Kochhar DM Skeletal morphogenesis: Comparative effects of a mutant gene and a teratogen Prog. Clin. Biol. Res.  1985 171 267 281 Google Scholar PubMed  Kwasigroch TE Kochhar DM Vitamin A alters the internal viscosity of fragments of limb-bud mesenchyme J. Embryol. Exp. Morphol.  1980 59 325 339 Google Scholar PubMed  Kwasigroch TE Vannoy JF Church JK Skalko RG Retinoic acid enhances and depresses in vitro development of cartilaginous bone anlagen in embryonic mouse limbs In Vitro Cell. Dev. Biol.  1986 22 150 156 Google Scholar CrossRef Search ADS PubMed  Lammer EJ Chen DT Hoar RM Agnish ND Benke PJ Braun JT Curry CJ Fernhoff PM Grix AW Jr Lott IT et al.   Retinoic acid embryopathy N. Engl. J. Med.  1985 313 837 841 Google Scholar CrossRef Search ADS PubMed  Lane MA Chen AC Roman SD Derguini F Gudas LJ Removal of LIF (leukemia inhibitory factor) results in increased vitamin A (retinol) metabolism to 4-oxoretinol in embryonic stem cells Proc. Natl. Acad. Sci. U.S.A.  1999 96 13524 13529 Google Scholar CrossRef Search ADS PubMed  Lee GS Kochhar DM Collins MD Retinoid-induced limb malformations Curr. Pharm. Des.  2004 10 2657 2699 Google Scholar CrossRef Search ADS PubMed  Light W Vernon AE Lasorella A Iavarone A LaBonne C Xenopus Id3 is required downstream of Myc for the formation of multipotent neural crest progenitor cells Development  2005 132 1831 1841 Google Scholar CrossRef Search ADS PubMed  Lincecum JM Fannon A Song K Wang Y Sassoon DA Msh homeobox genes regulate cadherin-mediated cell adhesion and cell-cell sorting J. Cell. Biochem.  1998 70 22 28 Google Scholar CrossRef Search ADS PubMed  Lonai P Epithelial mesenchymal interactions, the ECM and limb development J. Anat.  2003 202 43 50 Google Scholar CrossRef Search ADS PubMed  Mangelsdorf DJ Thummel C Beato M Herrlich P Schutz G Umesono K Blumberg B Kastner P Mark M Chambon P et al.   The nuclear receptor superfamily: The second decade Cell  1995 83 835 839 Google Scholar CrossRef Search ADS PubMed  Matt N Dupe V Garnier JM Dennefeld C Chambon P Mark M Ghyselinck NB Retinoic acid-dependent eye morphogenesis is orchestrated by neural crest cells Development  2005 132 4789 4800 Google Scholar CrossRef Search ADS PubMed  Mic FA Sirbu IO Duester G Retinoic acid synthesis controlled by Raldh2 is required early for limb bud initiation and then later as a proximodistal signal during apical ectodermal ridge formation J. Biol. Chem.  2004 279 26698 26706 Google Scholar CrossRef Search ADS PubMed  Montero JA Ganan Y Macias D Rodriguez-Leon J Sanz-Ezquerro JJ Merino R Chimal-Monroy J Nieto MA Hurle JM Role of FGFs in the control of programmed cell death during limb development Development  2001 128 2075 2084 Google Scholar PubMed  Mototani H Mabuchi A Saito S Fujioka M Iida A Takatori Y Kotani A Kubo T Nakamura K Sekine A et al.   A functional single nucleotide polymorphism in the core promoter region of CALM1 is associated with hip osteoarthritis in Japanese Hum. Mol. Genet.  2005 14 1009 1017 Google Scholar CrossRef Search ADS PubMed  Mundhenke C Meyer K Drew S Friedl A Heparan sulfate proteoglycans as regulators of fibroblast growth factor-2 receptor binding in breast carcinomas Am. J. Pathol.  2002 160 185 194 Google Scholar CrossRef Search ADS PubMed  Neubert D Barrach H-J Neubert H-JMD Kwasigroch TE Techniques applicable to study morphogenetic differentiation of limb buds in organ culture Methods in Pre-Natal Toxicology; Evaluation of Embryotoxic Effects in Experimental Animals. Teratology Workshop April 1977 Berlin  1977 Stuttgart Georg Thieme Publishers 241 251 Niederreither K Vermot J Schuhbaur B Chambon P Dolle P Embryonic retinoic acid synthesis is required for forelimb growth and anteroposterior patterning in the mouse Development  2002 129 3563 3574 Google Scholar PubMed  Nieto MA Bennett MF Sargent MG Wilkinson DG Cloning and developmental expression of Sna, a murine homologue of the Drosophila snail gene Development  1992 116 227 237 Google Scholar PubMed  Pan Y Liu Z Shen J Kopan R Notch1 and 2 cooperate in limb ectoderm to receive an early Jagged2 signal regulating interdigital apoptosis Dev. Biol.  2005 286 472 482 Google Scholar CrossRef Search ADS PubMed  Qin P Cimildoro R Kochhar DM Soprano KJ Soprano DR PBX, MEIS, and IGF-I are potential mediators of retinoic acid-induced proximodistal limb reduction defects Teratology  2002 66 224 234 Google Scholar CrossRef Search ADS PubMed  Rodriguez-Leon J Merino R Macias D Ganan Y Santesteban E Hurle JM Retinoic acid regulates programmed cell death through BMP signalling Nat. Cell Biol.  1999 1 125 126 Google Scholar CrossRef Search ADS PubMed  Scaife RM Courtneidge SA Langdon WY The multi-adaptor proto-oncoprotein Cbl is a key regulator of Rac and actin assembly J. Cell Sci.  2003 116 463 473 Google Scholar CrossRef Search ADS PubMed  Seki K Fujimori T Savagner P Hata A Aikawa T Ogata N Nabeshima Y Kaechoong L Mouse Snail family transcription repressors regulate chondrocyte, extracellular matrix, type II collagen, and aggrecan J. Biol. Chem.  2003 278 41862 41870 Google Scholar CrossRef Search ADS PubMed  Shenefelt RE Morphogenesis of malformations in hamsters caused by retinoic acid: Relation to dose and stage at treatment Teratology  1972 5 103 118 Google Scholar CrossRef Search ADS PubMed  Shimizu-Nishikawa K Tazawa I Uchiyama K Yoshizato K Expression of helix-loop-helix type negative regulators of differentiation during limb regeneration in urodeles and anurans Dev. Growth Differ.  1999 41 731 743 Google Scholar CrossRef Search ADS PubMed  Shum L Coleman CM Hatakeyama Y Tuan RS Morphogenesis and dysmorphogenesis of the appendicular skeleton Birth Defects Res. C Embryo Today  2003 69 102 122 Google Scholar CrossRef Search ADS PubMed  Sil AK Maeda S Sano Y Roop DR Karin M IkappaB kinase-alpha acts in the epidermis to control skeletal and craniofacial morphogenesis Nature  2004 428 660 664 Google Scholar CrossRef Search ADS PubMed  Vasiliauskas D Laufer E Stern CD A role for hairy1 in regulating chick limb bud growth Dev. Biol.  2003 262 94 106 Google Scholar CrossRef Search ADS PubMed  Wada RK Seeger RC Reynolds CP Alloggiamento T Yamashiro JM Ruland C Black AC Rosenblatt JD Cell type-specific expression and negative regulation by retinoic acid of the human N-myc promoter in neuroblastoma cells Oncogene  1992 7 711 717 Google Scholar PubMed  Watanabe N Tezuka Y Matsuno K Miyatani S Morimura N Yasuda M Fujimaki R Kuroda K Hiraki Y Hozumi N et al.   Suppression of differentiation and proliferation of early chondrogenic cells by Notch J. Bone Miner. Metab.  2003 21 344 352 Google Scholar CrossRef Search ADS PubMed  Weston AD Hoffman LM Underhill TM Revisiting the role of retinoid signaling in skeletal development Birth Defects Res. C Embryo Today  2003 69 156 173 Google Scholar CrossRef Search ADS PubMed  Williams SS Mear JP Liang HC Potter SS Aronow BJ Colbert MC Large-scale reprogramming of cranial neural crest gene expression by retinoic acid exposure Physiol. Genomics  2004 19 184 197 Google Scholar CrossRef Search ADS PubMed  Zakeri ZF Ahuja HS Apoptotic cell death in the limb and its relationship to pattern formation Biochem. Cell Biol.  1994 72 603 613 Google Scholar CrossRef Search ADS PubMed  Zuzarte-Luis V Montero JA Rodriguez-Leon J Merino R Rodriguez-Rey JC Hurle JM A new role for BMP5 during limb development acting through the synergic activation of Smad and MAPK pathways Dev. Biol.  2004 272 39 52 Google Scholar CrossRef Search ADS PubMed  © The Author 2006. Published by Oxford University Press on behalf of the Society of Toxicology. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org
TI - Novel Retinoid Targets in the Mouse Limb during Organogenesis
JF - Toxicological Sciences
DO - 10.1093/toxsci/kfl037
DA - 2006-06-13
UR - https://www.deepdyve.com/lp/oxford-university-press/novel-retinoid-targets-in-the-mouse-limb-during-organogenesis-VfQBSbbzwj
SP - 139
EP - 152
VL - 94
IS - 1
DP - DeepDyve
ER -