Transcriptional profiling of in vitro smooth muscle cell differentiation identifies specific patterns of gene and pathway activation

Joshua M. Spin, Shriram Nallamshetty, Raymond Tabibiazar, Euan A. Ashley, Jennifer Y. King, Mary Chen, Phillip S. Tsao, Thomas Quertermous

Abstract

Mesodermal and epidermal precursor cells undergo phenotypic changes during differentiation to the smooth muscle cell (SMC) lineage that are relevant to pathophysiological processes in the adult. Molecular mechanisms that underlie lineage determination and terminal differentiation of this cell type have received much attention, but the genetic program that regulates these processes has not been fully defined. Study of SMC differentiation has been facilitated by development of the P19-derived A404 embryonal cell line, which differentiates toward this lineage in the presence of retinoic acid and allows selection for cells adopting a SMC fate through a differentiation-specific drug marker. We sought to define global alterations in gene expression by studying A404 cells during SMC differentiation with oligonucleotide microarray transcriptional profiling. Using an in situ 60-mer array platform with more than 20,000 mouse genes derived from the National Institute on Aging clone set, we identified 2,739 genes that were significantly upregulated after differentiation was completed (false-detection ratio <1). These genes encode numerous markers known to characterize differentiated SMC, as well as many unknown factors. We further characterized the sequential patterns of gene expression during the differentiation time course, particularly for known transcription factor families, providing new insights into the regulation of the differentiation process. Changes in genes associated with specific biological ontology-based pathways were evaluated, and temporal trends were identified for functional pathways. In addition to confirming the utility of the A404 model, our data provide a large-scale perspective of gene regulation during SMC differentiation.

  • gene expression
  • microarray

vascular smooth muscle cells (SMCs) respond to injury by modulating their phenotype from a differentiated, contractile state to a synthetic and proliferative phenotype (32). This dedifferentiation is a critical aspect of many pathological processes, including hypertension, atherosclerosis, and postangioplasty restenosis (35). Despite considerable efforts, the details of SMC differentiation and dedifferentiation in development and disease have remained elusive. Until recently, one of the primary barriers to the study of SMC differentiation has been the absence of a robust in vitro model. Primary SMC cultures undergo rapid dedifferentiation after a few passages, with accompanying transcriptional changes of many cellular markers, such as smooth muscle (SM) α-actin, transgelin (SM22α), and smooth muscle myosin heavy chain (SM-MHC) (32). Within the last few years, the development of several in vitro culture models of SMC differentiation has advanced significantly this area of research. These include the Monc-1 neural crest cell line, mouse embryonic 10T1/2 cells, chick proepicardial cells, mouse embryonic stem cells, and a clonal line of P19 mouse embryonal carcinoma cells, designated A404 (22, 24, 27, 30, 46). The A404 cell line, developed by Owens and colleagues (5), was derived from P19 mouse embryonal carcinoma cells transfected with a SM α-actin promoter/puromycin-N-acetyltransferase (SMA-PAC). They showed that A404 cells treated with all-trans retinoic acid (RA) followed by puromycin differentiate rapidly and efficiently into SMCs, as demonstrated by upregulation of SM-MHC and other SMC-specific marker genes (30).

Although the A404 line and related models have provided some insights into the genetics of SMC differentiation, no large-scale transcriptional analysis of the process has yet been performed. Microarray-based profiling provides simultaneous expression information on thousands of genes, permitting large-scale analysis of biological pathways, elucidation of new roles for genes involved in other processes, and characterization of unknown genes. High-throughput methods have been used to study several aspects of cardiovascular disease, including atherosclerosis, arterial restenosis, myocardial infarction, cardiac hypertrophy, and heart failure (see review, Ref. 21). They have also been applied to cardiac chamber-specific gene expression (40). Additionally, microarray profiling and serial analysis of gene expression (SAGE) have examined cultured vascular SMCs (4, 33).

In this study we used a 60-mer in situ synthesized oligonucleotide mouse microarray platform, representing over 20,000 known mouse genes, to explore large-scale gene expression in differentiating A404 cells. Performing rigorous statistical analysis, we identified numerous genes expressed in differentiating SMCs and were able to detect patterns of gene expression and pathway activation. These data suggest new roles for known genes and reveal unknown genes that may play a role in SMC differentiation. Future studies examining these genes in detail will contribute additional insights into the role of SMCs in development and disease.

MATERIALS AND METHODS

Cell culture and RNA isolation.

Cryopreserved, undifferentiated P19-A404 cells were the generous gift of Gary K. Owens (University of Virginia). Cells were plated in 10-cm dishes and maintained in basal medium: α-minimum essential medium (α-MEM, catalog no. M0644; Sigma, St. Louis, MO) supplemented as previously described with 7.5% fetal bovine serum (FBS; Summit Biotechnology, Ft. Collins, CO), 200 μg/ml l-glutamine, and penicillin/streptomycin (Invitrogen, Carlsbad, CA) (30).

A set of dishes was harvested at time 0 (control group). Medium in another set of plates was changed at time 0 to basal medium supplemented with 1 μmol/l all trans-RA (Sigma), and cells were maintained in this environment for either 48 h (RA48 group) or 96 h (RA96 group) before harvesting. A final set of plates were treated with RA for 96 h and then passaged into basal medium containing 0.5 μg/ml puromycin (Clontech, Palo Alto, CA). These plates were maintained in puromycin for 48 h, and then harvested (Puro group). The entire time course was performed twice. Cells were harvested at 50–90% confluence in pooled groups of 5–6 dishes to create between 3 and 6 biological replicates at each time point.

During harvesting, medium was removed, and 1 ml of TRIzol was added to each dish. Cells were removed with a plastic cell lifter and placed in RNase-free microcentrifuge tubes. Chloroform (200 μl) was added to each tube, and samples were vortexed and then incubated at room temperature for 10 min. Samples were then centrifuged at 16,000 g for 15 min. The aqueous phase was pipetted into a new tube, then concentrated at medium temperature (Savant SpeedVac Plus SC110A; Center for Applied Genetic Technologies, Athens, GA) to a volume of ∼200 μl. Sets of five samples were pooled, and RNA was purified and isolated using the Qiagen RNeasy Midi Kit Protocol as previously described (40). Final total RNA concentrations determined by spectrophotometry varied from 0.447 μg/μl to 3.096 μg/μl in RNase-free water. Samples were stored at −80°C prior to hybridization.

Immunocytochemistry.

Control and differentiating P19-A404 cells were examined with indirect immunofluorescence using well-established methods. Labeling was performed using primary monoclonal antibodies against smooth muscle markers calponin (1:8,000) and SM α-actin (1:600) (Sigma) and the neuronal markers peripherin (1:200) and glial fibrillary acidic protein (GFAP; 1:200) (Santa Cruz Biotechnology, Santa Cruz, CA) as per established methodology.

Microarray hybridization and data acquisition.

RNA hybridizations were performed using the Agilent Mouse (Development) Oligo Microarray G4120A platform, consisting of 20,371 60-mer oligonucleotides representing over 20,000 known mouse genes and derived largely from sequences from the National Institute on Aging cDNA 7.4K and 15K mouse clone sets (9). A common reference consisting of RNA derived from whole 17.5-day mouse embryos was utilized as previously described (40).

Briefly, 10 μg of total RNA were primed with 2 μl of 100 μM T16N2 DNA primer at 70°C for 10 min, then reversed transcribed at 42°C for 1 h in the presence of 400 U SuperScript II RTase (Invitrogen), and 100 μM each dATP, dTTP, dGTP, with 25 μM dCTP, 25 μM Cy3- or Cy5-labeled dCTP (NEN Life Science, Boston, MA), and RNase inhibitor (Invitrogen). RNA was then degraded with RNase A, and labeled cDNAs were purified using QIAquick PCR columns (Qiagen). Oligoarray control targets and hybridization buffer (Agilent In Situ Hybridization Kit Plus) were added, and samples were applied to microarrays enclosed in Agilent SureHyb-enabled hybridization chambers. After hybridization, slides were washed sequentially with 6× SSC/0.005% Triton X-102 and 0.1× SSC/0.005% Triton X-102 before scanning. Slides were hybridized for 17 h at 60°C in a rotating oven, washed, and then scanned on an Agilent G2565AA scanner. Images were quantified using Agilent Feature Extraction Software (version A.6.1.1).

Microarray quality control analysis.

Summary box plots portraying the expression value distribution of each of the biological replicates, and log-log plots comparing sets of biological replicates were generated using Genedata Expressionist software (http://www.genedata.com/index.php). After examining these plots, samples that showed a high degree of variability from other similarly treated biological replicates were removed from the data set and excluded from further analysis. The final analysis groups consisted of six control replicates, five RA48 replicates, six RA96 replicates, and six puromycin replicates.

Data analysis.

After excluding arrays that displayed excessive differences in gene expression from similarly treated replicates, the remaining 5–6 biological replicates for each time point from both time course iterations were grouped together. Unsupervised hierarchical clustering of the replicates was performed (Genedata, South San Francisco, CA). Furthermore, significance analysis of microarrays (SAM) was performed in a two-class, unpaired fashion using K-nearest neighbor imputation with 10 neighbors, a random seed, and 500 permutations to identify genes that were significantly up- or downregulated between adjacent time-points, using a false-detection ratio (FDR) cutoff of <1 (P < 0.01) (41). Intersections of gene lists identified in this fashion were obtained, providing expression patterns over the time course. These pattern lists were examined with principal components analysis (PCA) using a covariance matrix (Genedata) (34). Genes were ordered by loading within the first eigenvector.

Using GenBank accession numbers, UniGene cluster numbers, and LocusLink gene symbols [National Center for Biotechnology Information (NCBI), National Institutes of Health], the microarray was annotated as fully as possible using Gene Ontology (GO) annotation terms (3). The abundance of these terms for each expression pattern was studied. Heat maps were created using HeatMap Builder 1.0 (E. A. Ashley, J. M. Spin, and C. Watt: Stanford University, http://quertermous.stanford.edu/heatmap.htm). All microarray data were submitted to the Gene Expression Omnibus (GEO) database at the NCBI (GSE1506; http://www.ncbi.nlm.nih.gov/geo/).

Quantitative reverse-transcription polymerase chain reaction.

Verification of microarray results and SMC marker expression during differentiation was performed employing real-time reverse-transcription polymerase chain reaction (RT-PCR) amplification for SM α-actin and transgelin (SM22α). Two representative samples from each time point were used. The specificity of the primer sequences (found in Supplemental Table S4, available online at http://mozart.stanford.edu/TQLab/lvad/Josh/spin.htm and at the Physiological Genomics web site)1 was verified previously (30).

After DNase treatment, synthesis of cDNA was performed from 3 μg of RNA using MMLV reverse transcriptase (SuperScript II kit, Invitrogen). DNA samples were combined with a reaction mixture containing QuantiTect SYBR Green PCR Master Mix (Qiagen) per standard protocol. Amplification was carried out in triplicate at 50°C for 2 min, and 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Quantity was expressed relative to 18S endogenous control (Ambion, Austin, TX). Fold differences were calculated by dividing the treated samples (RA48, RA96, and Puro samples) by the averaged control samples. Negative controls were included for all samples and revealed no DNA contamination.

RESULTS

Validation of the A404 in vitro differentiation model.

The differentiation protocol composed of 96 h of all-trans RA, followed by 48 h of puromycin treatment converted A404 cells to SMCs with high efficiency in a fashion similar to that described previously (30). Immunocytochemistry of control cells showed no detectable staining for SM α-actin or calponin, while <1% stained positive for GFAP and peripherin. After 96 h of RA treatment, ∼80–85% of cells stained strongly for SM α-actin, ∼60–70% were calponin positive, and ∼5–10% were GFAP positive and peripherin positive. After 48 h of puromycin treatment, >90% of cells stained positively for SM α-actin and calponin, whereas GFAP and peripherin staining were undetectable (data not shown).

Differential gene expression with SMC differentiation.

The Agilent Mouse (Development) Oligo Microarray as currently annotated carries 20,280 gene features. This platform was used to assess gene expression in differentiating A404 cells. Expression value box plots of all 27 replicates revealed 26 with excellent data normalization, consistent variability, and no significant over- or undersaturation. Comparative log-log plot matrices of the expression data for the 26 remaining replicates yielded 23 arrays with high correlation within each treatment group (correlation coefficients ranged from 0.71 to 0.98 overall). Unsupervised hierarchical analysis of the final 23 replicate sets, containing 5–6 replicates per treatment group, showed tight clustering of the control group. The treatment replicates clustered appropriately as well, although there was some division by membership into the first and second experimental repeats. See Supplemental Figs. S1–S3 detailing these analyses, at http://mozart.stanford.edu/TQLab/lvad/Josh/spin.htm and at the Physiological Genomics web site.

Two-class, unpaired SAM analysis comparing A404 Puro group cells with the control group cells revealed 2,739 genes that were significantly upregulated after differentiating with RA followed by puromycin (FDR < 1). A set of 2,227 genes were downregulated after puromycin treatment (FDR < 1) (Supplemental Figs. S4–S5). In comparison, somewhat fewer genes were significantly upregulated in the RA48 group over control (1,181), while a larger number of genes showed consistently increased expression in the RA96 group over control (3,278). Puro and RA96 groups were more similar, with only 488 genes significantly higher in puromycin and 274 higher in RA96 when the two were compared directly.

As expected, many marker genes for differentiated SMCs, such as the genes for SM α-actin, smoothelin, calponin, caldesmon, transgelin (SM22α), aortic preferentially expressed gene 1 (Apeg1), and α-1 tropomyosin showed significantly increased expression in the Puro group compared with control. The gene formerly known as mouse α-1 actinin and now known as striamin (UniGene, NCBI) was also found in this group. SM α-actin gene expression, required for A404 cell survival during puromycin treatment, was upregulated an average of 147-fold over control. However, it was also significantly upregulated in the RA48 and RA96 groups (22-fold and 101-fold, respectively) over control (FDR < 1). Complete lists of the genes up- and downregulated at various time points, ordered by SAM (d) score, are provided online in the Supplemental Material (at http://mozart.stanford.edu/TQLab/lvad/Josh/spin.htm and at the Physiological Genomics web site).

The expression levels of neuronal markers Neurod1 (neurogenic differentiation 1) and Mtap2 (microtubule-associated protein 2) were not significantly changed during the treatment course. Cardiomyocyte homeobox gene Nkx2.5 also showed no significant expression change from control A404 cells. In contrast to the results of Manabe and Owens (30), we found that Acta1 (actin, α1, skeletal muscle) was increased 1.98-fold in the RA96 group over control, a level maintained in the Puro group, with no significant change in expression from control to RA48 cells.

Quantitative real-time PCR.

We performed quantitative RT-PCR using SYBR Green on a representative pair of samples from each time point to obtain relative expression changes for two SMC markers [SM α-actin and transgelin (SM22α)]. The quantitative RT-PCR results for SM α-actin expression correlated well with the microarray studies, while transgelin expression closely resembled the microarray results for transgelin 2 (a close homolog). Fold-change data are available online (Supplemental Table S5).

Gene expression patterns associated with in vitro differentiation.

The timing of changes in gene expression during SMC differentiation can reveal functional patterns. Accordingly, we examined groups of genes that exhibited uniform behavior throughout the treatment course, consistently rising, falling, or remaining the same from group to group. Rising and falling were defined as significantly up- or downregulated by unpaired SAM (FDR < 1), whereas unchanging genes were those that were neither up- or downregulated using the same criteria. Four patterns were felt to have the most biological significance (Fig. 1). Pattern 1 represents genes that were unchanged from control to RA48 and then rose significantly at each subsequent treatment point. Pattern 2 represents genes that rose significantly from control to RA48 and were subsequently unchanged. Pattern 3 represents those genes that rose from control to RA48, rose again from RA48 to RA96, and then were unchanged, whereas pattern 4 represents genes that showed increasing expression at each treatment point.

Fig. 1.

Generation of differentiation patterns. All four treatment groups were compared in all possible two-way analyses using significance analysis of microarrays (SAM), cutoff false-detection ratio (FDR) <1. Four of the identified differentiation patterns are shown here, with gene totals. Genes were required to satisfy all criteria at each step for inclusion, using Venn intersections of array feature numbers. See Figs. 25 for symbolic depictions of the differentiation patterns. Control, control group; RA48, 48 h of retinoic acid; RA96, 96 h of retinoic acid; Puro, 96 h of retinoic acid and then 48 h of puromycin.

Genes within each group were ranked by the degree to which their expression adhered to the chosen pattern. PCA of each group identified one dominant eigenvector, the profile of which closely matched the expression pattern. The first eigenvectors represented 45.2%, 34.4%, 48.2%, and 59% of the gene expression variance for patterns 1–4, respectively. Genes were ranked by relative loading within the first eigenvector. The top 50 (or in the case of pattern 4, all 62) genes for each pattern are presented in rank order in row-normalized heat maps (Figs. 2  5). A symbolic representation of the expression pattern is also shown. The genes in each pattern clearly reflect the expected trends in expression. The PCA results show proper clustering and separation of all experimental replicates, and demonstrate close adherence to the expected patterns in eigenspace. Full lists of the genes for patterns 1–4 are provided in Supplemental Table S1.

Fig. 2.

Genes identified in differentiation patterns were subjected to principal components analysis (PCA) and ordered by relative loading within the largest eigenvector. A: row-averaged heat map of top 50 genes exhibiting expression following pattern 1. GenBank accession number and gene name are shown. B: symbolic representation of gene expression profile, and heat map legend showing expression range. C: PCA. Summary 3-dimensional graph of experimental replicates for all treatment groups. Axes represent relative loading of each replicate for the top 3 eigenvectors (Eig 1, 2, and 3, respectively).

Fig. 3.

Genes identified in differentiation pattern 2. See legend to Fig. 2 description.

Fig. 4.

Genes identified in differentiation pattern 3. See legend to Fig. 2 for description.

Fig. 5.

Genes identified in differentiation pattern 4. Description is that same as for Fig. 2, except top 62 genes are shown.

Patterns 2–4 encompass genes that showed early increases in expression with RA treatment, whereas pattern 1 genes had a delayed response to RA. Patterns 1 and 4 include genes that were already rising with RA but were further upregulated by puromycin treatment. Genes in patterns 2 and 3 were upregulated with RA, but showed no significant change in expression with puromycin. Of the various smooth muscle markers mentioned above, Apeg1 and α-1 tropomyosin expression followed pattern 2, the two calponin features displayed patterns 2 and 3, transgelin followed pattern 3, striamin and caldesmon followed patterns 3 and 4, and SM α-actin followed pattern 4.

Of 651 unique genes in patterns 1–4, 97 were unknowns/ESTs, and 115 were RIKEN genes. Many more uncharacterized genes were up- or downregulated in Puro group vs. control. An example from pattern 3 is 1600019O04Rik, a gene with 89.5% homology to human FAD104. Expression of this transcript has been identified in aorta and mesenchymal stem cells (UniGene, NCBI). Its protein product has nine fibronectin type 3 domains and a transmembrane domain and is thought to relate to adipocyte differentiation (Nucleotide, NCBI). It has domain similarity to the ephrin receptor family but also to a group of receptors that mediate netrin-dependent axon guidance. Another example is D4Ertd89e from pattern 1, which has been GO annotated (DNA binding, nucleosome assembly/chromosome organization and biogenesis). It has 78.7% homology to the human gene C9orf87. The protein has a large domain of unknown function, and a small histone H4 domain.

Functional annotation analysis.

To more completely characterize the differentially regulated genes, we examined each gene on the microarray for available GO annotation and then chose biologically significant terms. Of the 20,280 features, 8,655 were GO annotated.

Although some evolution of the gene lists in each chosen annotation category occurred over the treatment course, most genes that were upregulated in RA48 over control remained upregulated throughout. Examining treated vs. control, we found the number of genes in each annotation category as a percentage of the total number of annotated genes remained largely unchanged (Fig. 6). Transcriptional/DNA binding genes made up a slightly larger proportion of the annotated genes at the RA96 and puromycin time points. Compared with the treated vs. control lists, when puromycin-treated cells were evaluated against RA96, more genes were upregulated in the adhesion, angiogenesis, growth factor/chemokine/cytokine, and matrix groups, whereas fewer cytoskeletal, development, and differentiation-related genes were upregulated.

Fig. 6.

Gene Ontology (GO) annotation terms for selected pairwise SAM comparisons (FDR < 1) of treatment groups were obtained, using gene lists with unique names. A subset of 10 annotation terms are shown as a percentage of the total number of these terms for each comparison.

The relative percentage annotation makeup of differentiation patterns 1–4 are shown in Fig. 7. In contrast to the two-time-point comparisons, all annotation group proportions showed clear differences between patterns. For example, pattern 2 contains a much higher proportion of signal transduction/signaling, differentiation, and development genes than the other patterns and also contains the lowest proportion of adhesion, angiogenesis, and matrix genes. See Supplemental Table S2 for gene totals in each annotation category.

Fig. 7.

GO annotation terms for unique genes exhibiting differentiation patterns 1–4 were examined. Ten annotation terms are shown as a percentage of the total number of these terms for each pattern.

DISCUSSION

In this study we explored large-scale gene expression in A404 cells differentiating toward the SMC lineage. Previous investigations of differential gene expression employing in vitro models of SMC development have focused primarily on known SMC markers (22, 24, 27, 30, 46). However, two studies using Monc-1 cells identified novel SMC differentiation factors [Arid5b/Mrf2, Usf1, and Usf2 (13, 45)], while a study using 10T1/2 cells and proepicardial cells examined the cysteine-rich LIM-only proteins (Csrp1, Csrp2, and Csrp3) (10). The two isoforms of Arid5b/Mrf2 are AT-rich interaction domain transcription factors that induce SMC marker genes in 3T3 cells (45), and Csrp1/Crp2 and Csrp2/Crp2 act as transcription cofactors in complexes with serum response factor (SRF) and Gata6 proteins and may promote differentiation of pluripotent cells into SMCs (10). We found that Arid5b/Mrf2 expression followed pattern 4 and was upregulated 6.7-fold in the Puro group over control, supporting its role as an early regulator of SMC differentiation. Whereas Csrp2/Crp2 and Csrp3/Crp3 expression were unchanged during the treatment course, Csrp1/Crp1 was upregulated early (pattern 2) and was 5.9-fold increased in Puro over control.

Several previous studies have focused on the role of activin A, a homodimer of inhibin βA, in SMC differentiation and atherosclerosis. Activin A is a member of the TGF-β superfamily and modulates the proliferation and differentiation of various target cells, including endothelial cells, macrophages, SMCs, and osteoclasts. Activin A enhances osteoclastogenesis by stimulating IκB-α (39). It is expressed in human atherosclerotic lesions, promotes the SMC contractile phenotype, and suppresses SMC proliferation (17). We found that inhibin βA gene expression was strongly upregulated (31.8-fold) in puromycin-treated A404 cells over control cells and was the gene found to adhere most closely to pattern 1, implying a late response to RA and further upregulation when puromycin was added. Other inhibin (inhibin α and inhibin βB) gene expression was not significantly changed. IκB-α gene expression was also significantly higher in the Puro group compared with control (3.9-fold), and it followed pattern 3, increasing throughout RA stimulation and leveling off with puromycin.

The transcription factor Ankrd1/CARP is thought to play a role in cardiac and skeletal muscle differentiation, and has also been identified in vascular SMCs. Notably, activin A induces Ankrd1/CARP expression in cultured SMCs (15). Injured arteries and cultured vascular SMCs treated with TGF-β show increased Ankrd1/CARP expression (25). In our study Ankrd1/CARP was increased 8.2-fold in the Puro group vs. control group, and its expression followed pattern 3. Taken together these results support the notion that activin A signaling plays a crucial role in actively differentiating SMCs, likely through enhancement of IκB-α and Ankrd1/CARP expression. However, the patterns of the expression increases in the A404 model imply a yet earlier signal, perhaps TGF-β. The expression of Tgfb1 was increased 2.0-fold in puromycin-treated A404 cells vs. control but followed expression pattern 1, rising late in response to RA treatment. Tgfb2, however, was increased 26.7-fold and followed expression pattern 3, suggesting that this factor may contribute earlier in the differentiation process. Expression of a TGF-β signaling molecule known to have a functional role in regulating SMC gene expression, Smad3/Madh3, was increased in puromycin-treated cells over control (1.9-fold), as was Smad1/Madh1 (1.7- to 1.9-fold).

Chromatin remodeling and acetylation are important for SMC marker gene expression (26). We identified four factors with known roles in chromatin modification that were upregulated with A404 differentiation, none of which have previously been connected to SMC differentiation: Smyd1, Baz1a, and Mecp2 (all pattern 2) and Mbd2 (pattern 3). Smyd1, also known as Bop, is an HDAC-dependent transcriptional repressor found in cardiomyocyte and skeletal muscle precursors that regulates Hand2 and is required for cardiac differentiation and morphogenesis (20). Baz1a is an ATP-dependent chromatin remodeling protein closely related to Williams syndrome transcription factor (6). Of note, Williams syndrome is a developmental disorder that includes significant cardiovascular anomalies among its manifestations. Mecp2 has been found in the central nervous system where it is thought to act as a gene silencer through DNA methylation (12). Its role in the A404 model might be suppression of neuronal gene expression. Mbd2 is a component of the large protein complex MeCP1 (also including HDAC1, HDAC2, and RbAp46/48), which represses transcription from densely methylated genes. Mbd2 exists in two isoforms, one of which enhances CREB-dependent gene expression by interacting with RNA helicase A (19).

To link patterns of gene expression changes to gene function, we defined biologically relevant patterns and then ranked the genes within resulting groups by their adherence to the pattern. The genes in pattern 1 exhibited a delayed response to RA, suggesting a need for prior activation of other factors, or perhaps the presence of elaborated matrix or other environmental changes prior to upregulation. In contrast, genes in patterns 2–4 were all upregulated within the first 48 h of RA treatment. Pattern 2 genes reached maximal expression levels in the RA48 group and were unchanged thereafter, possibly reflecting the achievement of a biological steady state. Genes in pattern 3 rose with RA treatment but were not significantly changed after puromycin, perhaps responding to withdrawal of the differentiation stimulus. Pattern 4 genes, led by SM α-actin, increased throughout differentiation.

We employed GO annotation analysis to gain clues to biological processes from the gene differentiation patterns. Pattern 2 showed a large proportion of genes devoted to differentiation, development, and signal transduction, with relatively few matrix, adhesion, and angiogenesis genes and a relatively high proportion of transcription factors. This is consistent biologically with the observed early rise in expression followed by stable transcription levels. Pattern 1 contained proportionately more angiogenesis genes and growth factors/cytokines/chemokines, with fewer transcription factors and cell cycle genes, as would be expected from a later response to RA. Pattern 3 had the highest percentage of matrix genes, and the second highest percentage of genes involved in adhesion, but the smallest percentage of cytoskeletal genes and growth factors. Transcription factors and genes involved in adhesion, cell cycling and the cytoskeleton were predominant in pattern 4, with few or no development and differentiation genes found and the smallest percentage of signaling factors.

A total of 52 genes with unique symbols/gene names identified as transcription factors, cofactors or transcriptional regulators by GO annotation were present in the four A404 differentiation patterns described above. There were 4 transcription-related genes in pattern 1, 29 in pattern 2, 12 in pattern 3, and 7 in pattern 4. See Supplemental Table S3 for full gene lists. Some of these factors were identified previously as important in SMC differentiation, including Gata6 (pattern 2), Klf5 (pattern 3), and both Klf4 and Arid5b/Desrt/Mrf2 (pattern 4) (1, 28, 42, 45). Several of the patterned transcriptional genes are believed to be involved in neuronal differentiation but might play a role in SMC differentiation in this model, such as Emx2 (pattern 1), Atbf1, Bach2, Foxk1, Gli2, Gli3, Hoxb8, Hoxb9, Mecp2, Nr2f1, Pbx1 (all pattern 2), Bhlhb2, Hic1 (both in pattern 3), and Mtpn (pattern 4). Interestingly, some of the upregulated genes are thought to suppress SM α-actin (Klf4, Purb, Rbms1: patterns 3 and 4), while others suppress neuronal growth (Hes1, Maged1, Ndn: all pattern 2).

The SRF-CArG box [CC(A/T)6GG]-dependent pathway plays a central role in SMCs, as SRF dimerizes to activate SMC marker genes via its MADS box domain in conjunction with cofactors, a process enhanced by TGF-β (8, 26, 31). SRF also binds to the early response gene Fos, permitting serum inducibility of growth (37). Recent studies have highlighted myocardin, a coactivator of SRF discovered in cardiac tissue that also serves as a muscle-specific transcriptional regulator of CArG-dependent SMC marker genes, but does not activate Fos (11, 16, 43, 44, 47). In contrast, the ternary complex factors (TCFs) of the ETS domain family (Elk1, Elk3, Elk4) may act as SRF coactivators for growth and proliferation signaling in SMC (7). Myocardin is expressed in undifferentiated A404 cells (but not P19 cells), and transcription increases with A404 differentiation (47). Neither the myocardin gene Myocd nor Elk1 were present on the array. However, Elk3 [implicated in angiogenesis and vasculogenesis (7)] followed pattern 4 and was upregulated 4.6-fold in puromycin-treated cells over control (Elk4 was unchanged). Although Elk3 acts as a transcriptional repressor, activation of ERK signaling by Ras expression converts it to an activator (7). One of the targets of Elk3 (also a target of TGF-β), Egr1, usually thought of as a growth/dedifferentiation factor, was upregulated 7.0-fold in Puro over control (pattern 3) (29).

Another process active in SMC differentiation involves homo- and heterodimers of basic helix-loop-helix (bHLH) factors binding to E-boxes in SMC marker gene promoters. To date, no bHLH factor as crucial or specific as is MyoD for skeletal myogenesis has been identified in SMCs (26). We identified differential regulation of several bHLH genes. Idb2, Idb3 (pattern 2), and Twist1, were all upregulated with A404 differentiation, whereas Tcf21/capsulin and Idb1 were unchanged. Hes1 (pattern 2) is a Notch target gene and inhibits neuronal differentiation (23). Bhlhb2 (pattern 3) (aliases DEC1 and Stra13), is associated with chondrocyte differentiation and adipogenesis, among other functions (36, 48). It is a target of TGF-β, is upregulated by RA stimulation of P19 cells during neuronal differentiation, and represses E-box elements through HDAC-dependent and independent mechanisms (38).

Although AP-1 complex genes Fos and Jun are often assumed to be associated with cell growth and proliferation, AP-1 may vary from proliferative to differentiation-inducing depending on its makeup, and complexes containing Fosl2/Fra2 are more differentiation specific (2). Complexes comprised of Jun, JunD and Fosl2/Fra2 are required both for proliferation and cardiomyocyte differentiation of P19 cells, whereas AP-1 in SMCs is largely comprised of JunB (18). In our study, Jun (pattern 2), Fos, Fosl2/Fra2, and Junb (all pattern 3) were all upregulated during the treatment course. We also found that Ifi16 was upregulated 2.1-fold in Puro over control and followed pattern 1, rising only after the AP-1 complex genes. This transcriptional regulator is in a family (HIN-200) whose members play roles in myoblast differentiation, has itself been linked to myeloid differentiation, and requires AP-1 activity for expression (14). These data suggests a positive role for AP-1 members in SMC differentiation and support evidence that SMC differentiation and proliferation may occur together (32).

It is possible that some of the genes found to be significantly upregulated during A404 differentiation were participating in neuronal differentiation. However, it seems unlikely that such genes would remain elevated after puromycin treatment, when no cells with neuronal markers were detectable by immunocytochemistry. Although we were able to distinguish a large number of genes thought to be crucial for SMC differentiation and development, it is unknown how closely our results mimic the in vivo process. We have focused our discussion on upregulated genes. However, an almost equally large set of genes were downregulated in the Puro group vs. control. The withdrawal of many of these genes might also be crucial for SMC differentiation.

Although our results must be interpreted within the context of possible limitations of the in vitro A404 system, they nevertheless represent the first comprehensive genomic evaluation of a defined SMC lineage. This provides a valuable foundation of knowledge and many candidate genes and advances our understanding of the complex mechanisms that regulate SMC differentiation and phenotypic switching in disease.

GRANTS

This work was supported by the Donald W. Reynolds Cardiovascular Clinical Research Center at Stanford University.

Acknowledgments

We thank Gary Owens for thoughtful insights and review of the manuscript and for providing the A404 cells.

Footnotes

  • 1 The Supplemental Material for this article (Supplemental Tables S1–S5 and Figs. S1–S5, as well as sets of sample SAM analyses and a Supplemental methods and results section) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00148.2004/DC1.

  • Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

    Address for reprint requests and other correspondence: T. Quertermous, Stanford Medical School, Division of Cardiovascular Medicine, 300 Pasteur Drive, Falk CVRC, Stanford, CA 94305 (E-mail: tomq1{at}stanford.edu).

REFERENCES

View Abstract