To achieve new insights into the coordinate regulation of gene expression during osteoblast differentiation we utilized an approach involving global analysis of gene expression to obtain the identities of messenger RNAs (mRNAs) expressed using an established in vitro model of bone development. MC3T3-E1 osteoblast-like cells were induced to differentiate by the addition of β-glycerophosphate (β-GP) and ascorbic acid. RNA samples derived from induced and uninduced control MC3T3-E1 cells were used to prepare complementary DNA (cDNA) for serial analysis of gene expression (SAGE). A preliminary SAGE database was produced and used to prepare a hybridization array to further facilitate the characterization of changes in the expression levels of 92 of the SAGE-mRNA assignments after induction of osteoblast differentiation, specifically after 6 days and 14 days of ascorbate treatment. SAGE-array hybridization analysis revealed coordinate induction of a number of mRNAs including Rab24, calponin, and calcyclin. Levels of MSY-1, SH3P2, fibronectin, α-collagen, procollagen, and LAMP1 mRNAs, present at day 6 postinduction, were markedly reduced by day 14 postinduction. A number of unanticipated and potentially important developmental genes were identified including the transforming growth factor β (TGF-β) superfamily member Lefty-1. Lefty-1 transcript and translation product were found to be induced during the course of MC3T3-E1 cell differentiation. We present evidence, using transient transfection and antibody neutralization approaches, that Lefty-1 modulates the induction of alkaline phosphatase (ALP) after treatment of MC3T3-E1 cells with ascorbate and β-GP. These data should provide useful new information for future analysis of transcriptional events in osteoblast differentiation and mineralization.
The osteoblast-like cell line MC3T3-E1, established from newborn mouse calvaria, undergoes an ordered, time-dependent developmental sequence leading to the formation of multilayered bone nodules by 30–35 days. This developmental pattern is characterized by the replication of preosteoblasts (days 1–10), followed by growth arrest and the expression of mature osteoblastic characteristics such as matrix maturation (days 10–21) and the eventual mineralization of the extracellular matrix after day 25. Specifically, this cell line has been shown to form bone nodules after treatment with β-glycerophosphate (β-GP) and ascorbic acid concomitant with coordinate expression of many different osteoblast-associated messenger RNAs (mRNAs) as well as production of collagenous extracellular matrix and subsequent mineralization.(1–5) Detailed analysis of the induction of the mineralized phenotype has shown that ascorbate transport is Na+ ion-dependent,(6) promotes osteoblast differentiation after the accumulation of type 1 collagen,(7) and is synergistic with bone morphogenetic protein (BMP)-2.(8) Thus, the induction of differentiation of MC3T3-E1 cells, initiated by the combined action of ascorbic acid and β-GP, represents a well-established and versatile model system for the study of coordinately controlled gene expression.
Many separate studies of patterns of gene expression during osteoblast differentiation have been described.(9) Such studies include the detailed analysis of the patterns of expression of a number of bone-related mRNAs including osteocalcin, osteonectin, or secreted protein acidic and rich cysteine (SPARC); fibronectin and the collagens(10); the insulin-like growth factors and their binding proteins(11); and the BMPs.(12) These and other individual mRNA species have shown a reproducible and coordinated pattern of expression as the osteoblast assumes a more differentiated phenotype.(9)
The importance of determining the global patterns of gene expression occurring during bone formation,(13) as attested to by the growing interest in molecular phenotypic analysis of osteoblast cells(14) and clinical correlates of bone disease,(15) cannot be overstated. However, in spite of extensive efforts expended by many investigators, only a small portion of the estimated 50,000 mRNAs present in a given cell type, representing even fewer gene families, have been characterized within the context of osteoblast differentiation. We reasoned that, given the power of serial analysis of gene expression (SAGE), important insights would result from a higher throughput, unbiased global sampling and analysis of gene expression during the osteoblast differentiation process.
SAGE(16–18) is a powerful analytical process by which the identities and levels of hundreds or thousands of mRNA species within a particular cell type may be determined. Short sequence tags (approximately 10 base pairs [bp]) are prepared from virtually all mRNA transcripts within a given cell. Concatenation and subsequent cloning of multiple independently derived SAGE tags allow analysis of many tags per clone in a serial manner by DNA sequence analysis. Collectively, analysis of gene expression is achieved through computer-assisted compilation and comparison of the mRNA-derived tags with databases such as GenBank and the GenBank expressed sequence tag database (dbEST). Therefore, SAGE provides a potentially important tool for the elucidation of gene expression patterns that contribute to a developmental process. SAGE has not, until now, been used to study the multifactorial changes in transcriptional activity that occurs on initiation of the complex transcriptional program of osteoblast differentiation. Thus, investigation of the process of osteoblast differentiation using SAGE should prove to be of enormous informational value for the generation of hypotheses regarding osteoblast differentiation in general and the processes leading to bone development and tissue calcification in particular. These insights should facilitate the understanding of normal bone development as well as associated diseases such as osteoporosis and osteogenic malignancies that afflict many individuals including those suffering from advanced breast and prostate cancers.
Here, we show that although expression of many genes known to play a role in osteoblast biology is revealed, many other genes and uncharacterized ESTs have been identified that are both surprising and of potential functional significance. We chose one of these, Lefty-1, to further explore the significance of this distant transforming growth factor β (TGF-β) relative to osteoblast biology.
MATERIALS AND METHODS
The SAGE method was performed essentially as described.(17) PolyA+ RNA was prepared by two passages over oligo-dT cellulose (MessageMaker kit; Gibco-BRL, Gaithersburg, MD, U.S.A.) and subsequently converted to double-stranded complementary DNA (cDNA) with a cDNA synthesis kit (Gibco-BRL) using the manufacturer's protocol and reversed-phase high-performance liquid chromatography (HPLC) purified 5′-biotin-dT18 (Integrated DNA Technologies, Inc., Coralville, IA, U.S.A.). The cDNA was subsequently cleaved with Nla-III, the 3′ biotinylated fragments captured on streptavidin-coated magnetic beads (Dynal, Inc., Oslo, Norway), the bound cDNA divided into two pools, and linkers containing recognition sites for BsmFl and bearing Nla-III (New England BioLabs, Inc., Beverly, MA, U.S.A.) complementary termini ligated to each pool. SAGE tags were released with BsmFl, the tag ends were filled in with DNA polymerase (Klenow fragment; Gibco-BRL), and the tags from each of the two pools were ligated using T4 DNA ligase (Gibco-BRL) overnight at 16°C. The ligation products were amplified, the polymerase chain reaction (PCR) products were fractionated by polyacrylamide gel electrophoresis (PAGE), and the 104-bp PCR product containing the ligated ditags was excised from the gel and isolated. The purified ditags were released by cleavage with Nla-III. Gel purified ditags were self-ligated to produce ditag concatemers(16) and, after gel purification, the concatemers were cloned into the SphI site of pZero (Invitrogen, Carlsbad, CA, U.S.A.). Colonies were screened for concatemer size by PCR with M13 forward and M13 reverse primers and selected clones were sequenced using the ABI dye terminator method and the M13 forward primer.
Sequence files were analyzed by means of software provided by the SAGE program group,(17) which identifies the anchoring enzyme site with the proper spacing, extracts the two intervening tags, and records them. The putative identity of each tag was established by its presence in GenBank or dbEST databases (release 105). Additional software was developed (C.P.H. Vary, unpublished data) to import the SAGE data into a Microsoft Access database supporting automated linkage of the SAGE access database to the NCBI GenBank, dbEST, and PubMed databases.
MC3T3-E1 cells were maintained in α-modified essential medium (α-MEM) containing 10% fetal bovine serum supplemented with antibiotic antimycotic solution (Sigma-Aldrich, St. Louis, MO, U.S.A.). Cells were seeded at a density of 2.5 × 104 cells/cm2 and were subcultured two to three times weekly (1:10 ratio). For the induction of osteoblast phenotype, cells at 80% confluence were treated with α-MEM supplemented with 10 mM β-GP and 50 μg/ml ascorbic acid.
Total cytoplasmic RNA was extracted using the guanidinum thiocyanate method.(19) Northern blot analysis was performed as described.(19–21) Complex cDNA probes were labeled with [3H]deoxycytosine triphosphate (dCTP) by random hexamer priming. Hybridizations were performed overnight at 50°C, and filters were washed at 65°C as described.(21) Northern blots were hybridized to probes derived from gel purified cloned fragments corresponding to bone receptor kinase 1 (Brk-1)/BMPR1a, SPARC, OSF-2/CBFA1, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNAs.
Tags assigned to murine ESTs were used to extract IMAGE consortium clone identification numbers using the NCBI Entrez database. These clone numbers were used to direct the construction of arrays of bacterial colonies bearing the IMAGE cDNA clones (Research Genetics, Inc., Huntsville, AL, U.S.A.). After overnight growth, colonies were lysed, fixed by UV irradiation, and the arrays were washed. Total RNA was prepared as described above, and polyA+ RNA was derived from this preparation using oligo-dT cellulose affinity chromatography (MessageMaker kit; Gibco-BRL). After mRNA isolation, cDNA was prepared using oligo-dT priming followed by random hexamer-primed labeling of the cDNA with [3H]dCTP. IMAGE clone concentrations at each position in the array were estimated by hybridization with a common probe for the ampr region of the parent vector. Four control array positions were included for a blank control, an ampr positive control, pBluescript, and pUC. All IMAGE clone identities were confirmed by DNA sequence analysis.
cDNA was synthesized by reverse transcription in a 20-μl reaction containing 40 ng mRNA, 2.5μM oligo-dT (Perkin Elmer, Foster City, CA, U.S.A.), 0.5 mM of each deoxynucleoside triphosphate (dNTP; Perkin Elmer), 6.5 U RNase inhibitor (Perkin Elmer), and 25 U MMuLV reverse transcriptase (Perkin Elmer) in the presence of 50 mM Tris-HCl (pH 8.3), 8 mM MgCl2, and 10 mM DTT. The reverse transcription reactions were incubated at 37°C for 1 h and then boiled for 2 minutes. The cDNA was amplified by PCR using the following oligonucleotide primers: forward OSF-2/periostin primer, AAAGTAAAAGTTGGCCTTAGCGACC and reverse osteoblast-specific factor (OSF)-2/periostin primer, CAGAAGCTCCCTTTCTT-CGCTAGT (353 bp); forward Iroquois-3 primer, GAGGAAGAGAGCAAACGCGAG and reverse Iroquois-3 primer, TCGCTAGTTTTGCAGTCCGAA (220 bp); forward Lefty-1 primer, CACCTGCTAGTGTTCGGAATGG and reverse Lefty-1 primer, TGGACACGAGCCTAGAATCGA (219 bp); forward OSF-2/CBFA1 primer, GCAAGAAGGCTCTGGCGTTTA and reverse OSF-2/CBFA1 primer GCC-CACAAATCTCAGATCGTTG (724 bp); forward alkaline phosphatase (ALP) primer, GAT-CGGGACTGGTACTCGGATAA and reverse ALP primer CACAT-CAGTTCTGTTCTTCGGGTAC (155 bp); forward GAPDH primer, GGAGATTGT-TGCCATCAACGA and reverse GAPDH primer, GAAGACACCAGTAGACTCCA-CGACA (222 bp). The reverse transcription reaction (1 μl) was used as the template in a 25-μl PCR volume with 0.5 μM of each primer, 2.5 mM MgCl2, 0.2 mM of each dNTP, and 0.5 U Taq polymerase (Promega Corp., Madison, WI, U.S.A.) in the presence of 50 mM KCl, 10 mM Tris-HCl (pH 9.0), and 1% Triton X-100. Cycling was performed in a Perkin Elmer 9600 thermal cycler, with an initial denaturation of 1.5 minutes at 95°C, followed by 35 cycles of 20 s at 95°C, 45 s at 53°C, and 45 s at 72°C, and then a final extension of 10 minutes at 72°C. PCR products were resolved by electrophoresis on 1.6% agarose gels and were visualized by ethidium bromide staining.
Western blot analysis
MC3T3-E1 cells were harvested by gentle trypsinization, and then lysed in Triton X-100 extraction buffer (20 mM Tris-HCl, pH 7.5, 0.15 M NaCl, and 1% Triton X-100). The protein concentrations were determined by the Bradford assay. Equal amounts of total cellular protein were used for Western blot analysis after resolution of proteins by 8% sodium dodecyl sulfate (SDS)-PAGE. The proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad, Hercules, CA, U.S.A.), blocked, and probed with a 1:1000 dilution of rabbit polyclonal anti-Lefty antibody that had been immunopurified using a GST-Lefty fusion protein followed by glutathione-sepharose affinity chromatography. Subsequently, membranes were incubated with monoclonal mouse anti-rabbit antibody conjugated to ALP. Bound antibody was visualized by incubation of the membrane with the ALP substrate CDP-Star (Tropix, Bedford, MA, U.S.A.) and exposure to X-ray film (Kodak, Rochester, NY, U.S.A.).
Cloning and cell transfection
Full length Lefty-1/Stra3 cDNA was digested with EcoRI, and the resulting restriction fragment containing the entire Lefty-1 coding region was purified by agarose gel electrophoresis, and ligated into the vector pCS2+. MC3T3-E1 cells were transfected with the sense PCS2 + Lefty-1 construct using standard protocols. Briefly, MC3T3-E1 cells at 60–70% confluence were washed once with α-optimem-I (Gibco-BRL, Life Technologies, Grand Island, NY, U.S.A.) and a mixture of 10 μg plasmid, α-optimem and LipofectAMINE (Gibco-BRL, Life Technologies) was added to the cells in a total volume of 50 μl per well in a 96-well white, clear-bottom microtiter plate (Costar, Inc., Cambridge, MA, U.S.A.). Transfection was allowed to proceed for 5 h and was terminated by removal of the transfection medium and addition of 10 ml of α-MEM containing 10% fetal calf serum (FCS).
Chemiluminescent detection of ALP
MC3T3-E1 cells were removed from a single 10-cm plate, washed, and diluted 3-fold in α-MEM with 10% FCS. One hundred microliters of this preparation was seeded into each of 96-microtiter wells. This generally provided cells that were 50–70% confluent on the following day. Sixteen to twenty-four hours after plating, cells were incubated with α-MEM with 10% FCS, with or without ascorbate/β-GP, supplemented with the appropriate cytokine or antibody. TGF-β was obtained from R & D Systems (Minneapolis, MN, U.S.A.), and anti-HA antibody was obtained from Santa Cruz, Inc. (Santa Cruz, NM, U.S.A.). All wells in the 96-well microtiter plate were washed twice with 0.10 M sodium bicarbonate. Wells were then fixed for 5 minutes with methanol and then subsequently washed twice with 0.10 M sodium bicarbonate. After removal of the bicarbonate, each well received 100 μl of 1% CSPD (Tropix, Inc.) in 0.1 M diethanolamine buffer containing 5 mM MgCl2 and 10% Sapphire II (Tropix Inc.). Chemiluminescence was measured at 15-minute intervals in a microtiter plate chemiluminometer (Zylux, Inc., Maryville, TN, U.S.A.) over a period of 4–6 h. Data from replicate wells (minimum, n = 8) were used to calculate SDs and 95% CIs by standard methods.
Kinetics of expression of selected SAGE identified mRNAs
The initiation of induction of the differentiated phenotype, at the molecular level, was assessed by Northern blot analysis of RNA species corresponding to SAGE-identified as well as known osteoblast markers, including the murine BMP type 1 receptor (Brk-1/BMPR1a), SPARC, and OSF-2/CBFA1. Shown in Fig. 1A, the ascorbate-induced cells showed a Brk-1/BMPR1a hybridization signal occurring at day 2 after induction. SPARC mRNA hybridization, shown in Fig. 1B, was not detected at day 1 after ascorbate induction, but appeared by days 2–3, whereas untreated cells showed lower levels of SPARC transcript. Long-term Northern blot analysis indicated that in ascorbate-treated cells these and other mRNA species were induced between days 2 and 4, and their levels were maintained through days 6–14 (data not shown). In Fig. 1C, a control for osteoblast differentiation, OSF-2/CBFA1,(22) is shown to be induced with a pattern similar to those seen above. Northern blot hybridization of GAPDH also is shown in Fig. 1C as a control for lane loading levels. Other markers of differentiation, including induction of ALP as measured by enzymatic activity (below), also were consistent with cessation of proliferation and induction of differentiation. Osteocalcin, a late marker of induction of the osteoblast phenotype, was assayed by both reverse-transcription PCR (RT-PCR) and Northern blot, and was expressed at significant levels only after about 14 days of induction (data not shown). These results establish a pattern of transcription activity consistent with induction of osteoblast differentiation.
Partial SAGE profile of MC3T3-E1 cells
The MC3T3 SAGE database comprised 1385 ascorbate-induced and 1089 -uninduced sequence tags representing 967 and 725 separate gene products, respectively. Of these 2474 SAGE tags corresponding to about 1692 different mRNA species, approximately 11% could be associated with known murine gene products in the GenBank database. An additional 40% could be assigned to murine ESTs in the dbEST. The remainder could not be associated with currently known gene products.
Among the tags occurring with the highest frequencies, in either the induced or uninduced MC3T3-E1 cells, were tags frequently associated with osteoblast transcripts. Shown in Table 1, these included tags associated with SPARC, procollagen α-2(α) and α-1 type I collagen. Tags associated with other known osteoblast markers, included in Table 2, were seen at lower frequency and included Brk-1/BMPR1a,(23) fibronectin receptor,(24,25) and 14 kDa lectin.(9) As anticipated, a much larger number of genes not as commonly associated with osteoblast function or development were observed. These gene products, grouped by known or proposed function, are shown in Table 2. These mRNA species include calponin,(26) a protein more commonly associated with vascular smooth muscle(27) and calcification of atherosclerotic plaques,(28) the murine p21 translationally controlled peptide, and the protein phosphatase pp2A.(29) Uncharacterized ESTs corresponding to GenBank entries AA000418, AA419754, and thymosin β-10 were prominently found in the uninduced SAGE tag population, as shown in Table 1. Thymosin β-10 has not been observed previously in this context though thymosin β-4, observed in this study as an EST assigned SAGE tag (not shown), has been identified previously as a gene in which expression is associated with the osteoblast like phenotype.(30)
Table Table 1.. Most Frequently Observed SAGE Tags
Table Table 2.. GenBank Entries Observed in the SAGE Database Corresponding to Ascorbate/β-Glycerophosphate-Induced MC3T3 Genes
Expression of MC3T3-E1 RNA species by SAGE-based array hybridization and RT-PCR
A hybridization array containing 92 IMAGE clones identified by SAGE was used to further define the levels of these mRNA species at two temporal points along the ascorbate-induced developmental pathway. Table 2 lists the genes corresponding to the IMAGE clone selections used in construction of this array. Table 2 also provides the GenBank accession number and the phosphorimager-derived difference in hybridization signals, for day 14 relative to day 6 postascorbate induction, for all IMAGE clones on the hybridization array.
Those portions of the array representing several of the most induced and repressed mRNA hybridization signals observed for the period from 6 to 14 days postinduction are presented in Fig. 2. We found that multiple hybridization of the same array membrane was a useful approach in that normalization of hybridization signals could be avoided facilitating a more direct comparison. Thus, Figs. 2A and 2B are composed of sections of autoradiographic exposures of the same array membrane after successive hybridizations with labeled cDNA derived from cells grown for either 6 days or 14 days of ascorbate induction, respectively. The blank controls for all hybridizations were negative as shown in the insets below Figs 2A and 2B. The differences in the hybridization signals at day 14 compared with day 6 after ascorbate induction (as estimated by phosphorimager analysis) are presented in Fig. 2C. Increases from day 6 to 14 postinduction were most evident for Rab24, Fau, α-actin, and H2-calponin. Decreases from day 6 to 14 postinduction were most apparent for Lamp1, fibronectin, procollagen, and MSY-1 mRNAs. These results establish quantitative patterns of coordinate regulation of these gene products.
Several mRNA species were further studied by amplification of reverse transcriptase reactions by PCR. Figure 3 shows that, consistent with the data presented in Table 2, Lefty-1 and OSF-2/periostin appear in the ascorbate induced samples, as compared with a GAPDH standard reaction shown at the bottom of Fig. 3. Iroquois-3, detected by SAGE at day 14 postinduction, also is detected only in the induced day 14 RNA sample by RT-PCR, as shown in the third panel of Fig. 3. RT-PCR for OSF-2/CBFA1 and ALP mRNA are included in Fig. 3, panels 4 and 5, respectively, as confirmatory markers of the differentiated osteoblast phenotype,(22) and are shown to be up-regulated.
DNA Sequence analysis, used to confirm the identities of all RT-PCR products, was necessary in particular to establish the identity of the Lefty RT-PCR product. Lefty-1 and Lefty-2 mRNAs are nearly identical at the level of nucleotide sequence, and primers that reliably distinguish Lefty-1 and −2 were not readily available. However, comparison of the sequence of the Lefty-1 RT-PCR product with the sequence of Lefty-2 cDNA confirmed that the RT-PCR product was indeed Lefty-1. Subsequent attempts to detect Lefty-2 by RT-PCR failed on all attempts suggesting that it is not expressed in these cells (data not shown).
In addition, many tags representing uncharacterized or unknown mRNA species have been identified and have been assigned for further study. Approximately 450 ESTs with unknown functions will be further assessed by array and Northern blot hybridization analysis. Finally, SAGE tags corresponding to several hundred unidentified murine gene products, 10% of which occur with an appreciable frequency in the induced or uninduced MC3T3-E1 cells, require further study after isolation of the corresponding cDNAs from cDNA libraries.
Induction of Lefty-1 mRNA and its translation product during MC3T3-E1 differentiation
SAGE identification of Lefty-1 was of special interest because it is a member of the TGF-β family of cytokines, potentially important in the context of bone development and differentiation as a possible modulator of BMP/TGF signaling.(31) Therefore, we sought to address the possible functional consequences of Lefty-1 expression during induction of bone differentiation by ascorbate and β-GP. Parallel analysis of samples of induced and uninduced MC3T3-E1 cells by Western blot and RT-PCR, shown in Fig. 4, confirmed that Lefty-1 protein as well as mRNA was present in MC3T3-E1 cells and up-regulated during induction of differentiation by ascorbate and β-GP. Demonstration of increasing levels of Lefty protein along with Lefty-1 mRNA underscored the likely functional significance of Lefty-1 during the induction of the osteoblast phenotype.
Lefty-1 modulation of MC3T3-E1 ALP activity
To further establish the functional consequences of Lefty-1 expression in MC3T3-E1 cells, we investigated the effects of transient Lefty-1 expression on the induction of osteoblast ALP activity. The rationale for this approach is based on the observation that Lefty-1 may negatively modulate the effects of BMP during embryogenesis(31–33) and that ALP induction is known to be regulated positively by BMP-2 and regulated negatively by TGF-β,(34–36) thus making it a likely function to be altered by TGF-β-related regulatory pathways, as well as being an established and easily measured marker of progression of the differentiated osteoblast phenotype. MC3T3-E1 cells were transfected with Lefty-1 cDNA, in the sense orientation, in the expression vector pCS2+. Transfection efficiencies with MC3T3-E1 cells were in the range of 20–30% as judged by cotransfection with green fluorescent protein-containing vectors (data not shown). For these experiments measurement of ALP was conducted at day 10 postinduction because preliminary experiments suggested that acceptable statistical significance, indicated by the 95% CI bars in Fig. 5, is achieved at or before this time. Transient Lefty-1 expression, shown in Fig. 5A, resulted in depression of the relative levels of ALP induction at day 10 postconfluence, as compared with ascorbate induction alone or in the insertless-vector controls. This effect was most apparent when cells were induced with ascorbate and β-GP. This result suggests that transient expression of Lefty-1, which produces elevated levels of Lefty-1 in the conditioned medium as well as transfected cells,(31) slows the appearance of ALP activity in MC3T3-E1 cells.
To confirm the modulation of ALP activity by Lefty-1, we attempted to block the effects of endogenous Lefty-1 in the conditioned culture medium using anti-Lefty antibody. The anti-Lefty antibody used is a rabbit polyclonal antibody that has been immune-purified using a GST-Lefty-1 fusion protein construct.(31) MC3T3-E1 cells were incubated with ascorbate and β-GP alone and with either anti-HA antibody, anti-Lefty antibody, soluble receptor for TGF-β, or TGF-β. The soluble receptor for TGF-β has been shown previously to abrogate response to exogenous and endogenous TGF-β in vitro and in vivo.(37) For this series, measurement of ALP was conducted at day 6 postinduction in order to conserve anti-Lefty antibody and because preliminary experiments indicated that acceptable statistical significance was achieved by this time. Figure 5B shows that after 6 days of treatment, MC3T3-E1 cells treated with anti-Lefty antibody, but not ascorbate alone or anti-HA antibody, appeared to elevate the levels of ALP. Under the conditions of the assay, Fig. 5B also shows that TGF-β decreased ALP levels as expected. Finally, as a control for the specificity of the TGF-β result, MC3T3-E1 cells incubated with the soluble receptor for TGF-β increased levels of ALP. Under the conditions of the experiment no significant changes in cell density or total protein between experimental treatments were evident (data not shown). As measured by chemiluminescence, ALP activity is rising rapidly in induced cells throughout the period of days 6–10 as reflected in the differences in the y axes of Figs. 5A and 5B. These results, taken together, support the hypothesis that Lefty-1 expression and inhibition can alter the appearance of induced ALP activity in opposing directions.
Bone formation in vivo is a complex phenomenon whereby recruitment and replication of mesenchymal precursors of osteoblasts, their differentiation into preosteoblasts, osteoblasts, and mature osteoblasts ultimately results in the accumulation and mineralization of the extracellular matrix. The MC3T3-E1 clonal osteoblastic cell line was derived from mouse calvaria and undergoes an ordered, time-dependent developmental sequence leading to formation of multilayered bone nodules over a 30- to 35-day period. This developmental pattern is characterized by the replication of preosteoblasts followed by growth arrest and the expression of mature osteoblastic characteristics such as matrix maturation, and eventual formation of multilayered nodules with a mineralized extracellular matrix.(2) This in vitro osteoblast culture system is biologically relevant in that the temporal sequence of expressed genes encoding osteoblast phenotype markers follows the specific pattern of gene expression and cell-type distribution observed in neonatal long bones and during fetal calvarial development in vivo.(2,38)
The ascorbate/β-GP-mediated induction of the osteoblast phenotype in MC3T3-E1 cells is a well-established model system. However, it is important to note that a limitation of this study is that it reflects transcriptional processes characteristic of the transformed MC3T3-E1 cell line. Recent studies of global gene expression in colon cancer tissue, normal colonic epithelium, and related transformed cell lines addressed this issue. SAGE data from these cell and tissue sources suggest that transformed cell lines provided reliable models for their in vivo counterparts, though caution was emphasized in the extension of individual expression data to all tissues studied.(18) It will therefore be of considerable interest to test the hypotheses implicit in the MC3T3-derived SAGE data in normal osteoblasts from different sources such as calvarial and long bone. Another limitation of the present study centers on the combined use of ascorbate and β-GP, because individual effects of ascorbate or β-GP are known. Indeed, β-GP has been reported to have effects on osteoblast and chondrocyte gene expression in vitro.(39,40) Therefore, an important application of the SAGE-derived gene arrays described here will be to permit rapid analysis of these issues, in terms of the levels of the specific mRNAs documented in Table 2, and in the contexts of primary osteoblast cells treated with ascorbate, β-GP, and other cytokines.
This preliminary SAGE data set represents about 3–5% of the total RNA species in the cell and about 35% of the number of tags (7202) recently sampled to define the transcriptional effects of a single protein, p53.(41) It is important to note that this level of SAGE will identify both high copy number mRNAs and lower copy mRNAs, the latter albeit with reduced efficiency. Given the large number of transcriptional changes attendant on execution of a cellular differentiation program, SAGE identification of many known and novel mRNAs will result from sequencing efforts that fall within the range of capabilities of most laboratories. As shown here, such efforts provide important information on the mRNA population changes resulting from complex differentiation programs such as osteoblastogenesis.
The detection of Brk-1/BMPR1a or BMP receptor type 1,(42) procollagens,(43) SPARC,(44) OSF-1,(45) OSF-2/periostin,(46–48) fibronectin, the fibronectin receptor,(24,25) the fibroblast growth factor (FGF) receptor,(49) and minopontin/osteopontin (OPN)(50,51) (Tables 1 and 2) agrees well with the previously reported expression of these markers in osteoblast-related cell types. The relative decreases observed in the hybridization signals for the α-1(I) procollagen and α-1(VI) collagen mRNAs, from day 6 to 14 postinduction (Fig. 2), are also consistent with the transient kinetics observed by others for collagen synthesis after ascorbate/β-GP induction in MC3T3-E1 cells.(2,52) In this regard, we note that SAGE detected a similar differentiation-related function for the Notch receptor ligand Jagged-1 in vitro. SAGE of NIH3T3 cells expressing a soluble form of Jagged-1 (sJ-1) revealed repression of the pro-α-2(I) collagen transcript.(53) These data, taken together, suggest a mechanism by which Jagged/Notch signaling in osteoblasts may modulate extracellular matrix composition. In a related sense, increased expression of the type IV collagenase transcript (Table 2) during osteoblast induction of differentiation, also seen in sJ-1 transfected NIH3T3 cells,(53) may be relevant to the process of differentiation because proteolytic modification of collagen is a property of osteoblast differentiation(54,55) that is repressed by cytokine inhibitors of osteoblast differentiation including TGF-β.(56) Finally, the coordinate regulation of collagens and SPARC agrees with the reported ascorbate inducibility of these gene products(43,44); however, the relatively high levels of SPARC transcript indicated by SAGE (Table 1) is remarkable and suggests that the osteoblast differentiation process involves large amounts of SPARC, possibly reflecting a critical need for SPARC counteradhesive properties, extracellular matrix turnover, and/or modulation of cell migratory behavior.(57,58)
A number of genes were detected that correspond to important transcription factors and cytokines. Members of the homeobox class of transcription factors have been previously implicated in osteoblast differentiation. Msx-2 has been detected in phenotypically mature, but not immature rat osteoblasts and may modulate the osteocalcin promoter,(59) along with Msx-1 and Dlx-5.(60) Iroquois-3 (Irx-3) is potentially significant in that it is a previously unrecognized member of the set of homeobox transcription factors expressed in MC3T3-E1 cells. The mouse Irx-3 gene is expressed early in the prospective neural plate in a subset of neural precursor cells(61) during early mouse embryogenesis.(62) Its role in osteoblast development remains to be determined, but it may participate in osteoblast gene regulation in a fashion that overlaps with but is distinct from the other homeobox genes previously identified in osteoblasts.
Among the many less-characterized mRNAs revealed in the SAGE data set are mRNAs encoding important regulatory proteins that may have significant but as yet undescribed roles in osteoblast metabolism or differentiation. It is these gene products, possessing important functions in other known cellular contexts, that will likely provide the most interesting insights into novel processes involved in osteoblast function. For example, the presence of a Rab5/Rab6 binding protein, previously discovered by a yeast two-hybrid screen,(63) points to a previously unrecognized role for this protein as well as the particular interacting partners, Rab family members, in osteoblast metabolism. Indeed, Rab24, observed as being strongly up-regulated here (Fig. 2C), has been associated with endocytic function in other contexts(64) and is thought to be a potent regulator of events in the endocytic pathway.(65) Calcyclin is an S100 protein homologue that is inducible on differentiation of neuroblastoma cells(66) and is thought to be a calcium-dependent switch.(67) This information implicates calcyclin in the differentiation-dependent metabolism of calcium in MC3T3-E1 cells(68) and suggests further study along those lines. EBI1 is a seven-transmembrane G-protein coupled receptor thought to be specific to lymphoid cells.(69) Recent data indicates that EBI1, also known as chemokine receptor 7, binds the ligand ELC/MIP-3β(70) and induces a rise of intracellular free calcium concentrations in, as well as the directional migration of, human monocyte-derived mature dendritic cells. Thus EBI1 may function to modulate free calcium levels in osteoblasts, independent of, or in concert with other cation-dependent G-protein coupled receptors.(71,72)
As noted above, it is well established that several members of the TGF-β superfamily of cytokines including TGF-β, the BMPs,(8,73) activins,(74) and their receptors(75) are critical determinants of osteoblast regulation and differentiation. Lefty-1 is a distant TGF-β family member related to the Drosophila protein nodal.(33) Lefty-1 has been proposed as a regulator of Lefty-2 and nodal(76) and PitX2.(32,77) PitX2 has been linked to craniofacial abnormalities associated with Rieger's syndrome,(78) suggesting that part of the Lefty-1/PitX2 regulatory axis may be relevant to induction of osteoblast differentiation. Lefty-1, detected here by SAGE and inducible by ascorbate, is also known to be retinoic acid inducible during mouse development.(31) Lefty-1 may act independently of or modulate the activities of other TGF-β related cytokines. For instance, Lefty-1 may act by interfering with members of the BMP family of TGF-β-related cytokines resulting in modulation of differentiation. Indeed, interference with BMP2/4 has been proposed as a function for Lefty-1 during mouse embryogenesis.(31) Alternatively, because TGF-β can inhibit differentiation(79) of osteoblasts, Lefty-1 could mimic or potentiate the effects of TGF-β. Transient expression of Lefty-1 inhibited ALP induction, and immune-purified antibody to Lefty-1 stimulated ALP induction. These effects don't distinguish the above alternatives but are consistent with a role for Lefty-1 that opposes the effects of BMP in these cells and is consistent with its proposed role in development.(31) It is important to note that the consequences of Lefty-1 expression may depend critically on many variables including the cell-culture model system used or the time or state of differentiation because the effects of the related BMPs and TGF-β are also complex and time dependent. Other experiments (not shown) indicate that PitX2, but not nodal or Lefty-2 (above), are expressed in MC3T3-E1 cells. Therefore, it will be important to extend these preliminary results to a full analysis of the signal pathways involved upstream and downstream of Lefty-1/expression in osteoblasts.
In conclusion, SAGE has provided identification of many novel genes expressed during osteoblast differentiation as defined by the ascorbate-induced MC3T3-E1 model system. Combining SAGE with cDNA array hybridization facilitates rapid and efficient assessment of coordinate gene expression. SAGE and SAGE-directed cDNA array hybridization, therefore, may provide an approach to assess the expression and variability of potentially hundreds or thousands of unknown as well as known mRNA species that may participate in osteoblast differentiation. Furthermore, SAGE-based arrays provide tools to extend rapidly the SAGE data to samples and tissues not easily accessible to SAGE. The data that has accumulated from application of this powerful methodology in an osteoblast-like system, providing novel insights implicating numerous regulatory genes previously unreported in osteoblastogenic processes. These genes can now serve as targets to focus on in investigating specific disease processes such as osteoporosis, osteomalacia, and osteogenic malignancies.
The authors thank Dr. V. Velculescu and Dr. K. Kinzler for helpful discussions and Dr. T. Maciag, Dr. R. Friesel, B. Conley, and J.D. Smith for many helpful suggestions. The authors also thank Professor P. Chambon for the generous gifts of Lefty/Stra3 antibody and the Lefty-1/Stra3 cDNA clone, Dr. H. Hamada for the sequence for the mouse Lefty-2 cDNA, and Dr. V. Koteliansky and Dr. P. Gotwals, Biogen, Inc., for soluble TGF-β receptor. This work was supported in part by a group grant from the MRC (Canada) and by grant 1P01CA78582 from the National Institutes of Health (NIH; A.S.).