Cbfa1 Isoforms Exert Functional Differences in Osteoblast Differentiation*

Cbfa1 is an essential transcription factor for osteoblast differentiation and bone formation. We investigated functional differences among three isoforms of Cbfa1: Type I (originally reported as Pebp2αA by Ogawa et al. (Ogawa, E., Maruyama, M., Kagoshima, H., Inuzuka, M., Lu, J., Satake, M., Shigesada, K., and Ito, Y. (1993) Proc. Natl. Acad. Sci. U. S. A. 90, 6859–6863), Type II (originally reported astil-1 by Stewart et al. (Stewart, M., Terry, A., Hu, M., O’Hara, M., Blyth, K., Baxter, E., Cameron, E., Onions, D. E., and Neil, J. C. (1997) Proc. Natl. Acad. Sci. U. S. A. 94, 8646–8651), and Type III (originally reported asOsf2/Cbfa1 by Ducy et al. (Ducy, P., Zhang, R., Geoffroy, V., Ridall, A. L., and Karsenty, G. (1997)Cell 89, 747–754). A reverse transcriptase-polymerase chain reaction analysis demonstrated that these isoforms were expressed in adult mouse bones. The transient transfection of Type I or Type IICbfa1 in a mouse fibroblastic cell line, C3H10T1/2, induced the expression of alkaline phosphatase (ALP) activity. This induction was synergistically enhanced by the co-introduction ofXenopus BMP-4 cDNA. In contrast, the transient transfection of Type III cDNA induced no ALP activity. In C3H10T1/2 cells stably transfected with each isoform ofCbfa1, the gene expression of ALP was also strongly induced in cells transfected with Type I and Type IICbfa1 but not in cells with Type III Cbfa1. Osteocalcin, osteopontin,and type I collagen gene expressions were induced or up-regulated in all of the cells stably transfected with each isoform of Cbfa1, and Type II transfected cells exhibited the highest expression level ofosteocalcin gene. A luciferase reporter gene assay using a 6XOSE2-SV40 promoter (6 tandem binding elements for Cbfa1 ligated in front of the SV40 promoter sequence), a mouse osteocalcinpromoter, and a mouse osteopontin promoter revealed the differences in the transcriptional induction of target genes by eachCbfa1 isoform with or without its β-subunit. These results suggest that all three of the Cbfa1 isoforms used in the present study are involved in the stimulatory action of osteoblast differentiation, but they exert different functions in the process of osteoblast differentiation.

The gene targeting in mice of Cbfa1 (core-binding factor), originally identified as a T-cell differentiation regulator (1,2), resulted in a complete lack of bone formation due to a maturational arrest of osteoblasts (3,4). Cbfa1 is also the responsible gene for the human genetic disease of cleidocranial dysplasia (5,6). The promoter region of the genes related to osteoblast differentiation such as osteopontin (OPN), 1 osteocalcin (OSC), and bone sialoprotein contains binding sequences of Cbfa1 (7)(8)(9). The transfection of Cbfa1 gene into non-osteogenic cells such as C3H10T1/2 cells and primary skin fibroblasts directed the differentiation pathway of these cells toward the osteoblast lineage (10). These results indicated that Cbfa1 is one of the essential transcription factors that regulate osteoblast differentiation and bone formation (11). In addition, bone morphogenetic proteins (BMPs), one of the most potent stimulatory factors for osteoblast differentiation, induced or stimulated the expression of Cbfa1 mRNA (10,12). This suggests that Cbfa1 is involved in the signaling pathway of BMP action.
Three subtypes of the ␣-subunit of Cbf (Cbfa1, Cbfa2, and Cbfa3) and one subtype of the ␤-subunit (Cbf␤) have been reported (13,14). The ␣-subunits of Cbf family transcription factors acquire enhanced DNA binding activity when they heterodimerize with the ␤-subunit (1,15). In addition, several isoforms of Cbfa1 have been identified by differential promoter usage or differential splicing (16,17). One isoform, originally cloned from ras-transformed NIH3T3 cells, was named Pebp2␣A (referred to as Type I Cbfa1 hereafter). Recently, three groups of investigators independently identified two other isoforms of Cbfa1, from osteoblasts and lymphoblasts (5,10,16); in these isoforms, two translational start sites are suggested: the shorter isoform (referred to as Type II isoform hereafter) and the longer isoform (referred to hereafter as Type III isoform). Although Ducy et al. (10) demonstrated that the transfection of Type III Cbfa1 into non-osteogenic cells induced gene expression related to osteoblast differentiation, functional differences among the isoforms of Cbfa1 have not been clarified.
We investigated the functional differences among three isoforms of Cbfa1 (Type I, II, and III) by the generation of stably transfected cells with each Cbfa1 isoform and a transient transcriptional assay using Cbfa1 target gene promoter-driven luciferase reporter genes. We demonstrate here that these three isoforms of Cbfa1 have different functions in osteoblast differentiation.

MATERIALS AND METHODS
Cell Culture-The mouse embryonic fibroblast cell line, C3H10T1/2, was purchased from Riken Cellbank (Saitama, Japan). This cell line was maintained in BME medium (Life Technologies, Inc.) containing 10% fetal calf serum (Life Technologies, Inc.) and antibiotics.
Detection of Alkaline Phosphatase Activity-Alkaline phosphatase (ALP) activity was detected histochemically using an Alkaline Phosphatase Substrate Kit IV (Vector Laboratories, Burlingame, CA). The ALP activity of the cell lysates was determined using p-nitrophenyl phosphate as a substrate as described previously (29 -32).
Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR)-The poly(A ϩ ) RNA purification, first-strand cDNA synthesis, and PCR were performed as described (18). The PCR conditions were as follows. After 1 min of preincubation at 94°C, amplification was performed for 35 cycles consisting of 20 s of denaturing at 94°C, 1 min of annealing, and extension at 66°C. The primers used for each isoform were as follows (small letters; restriction enzyme site and Kozak sequence).
Plasmids-A reporter plasmid containing 6 repeats of the consensus Cbfa1 binding site (6XOSE2) was constructed to insert a blunt-ended PCR fragment containing AACCACA-based direct repeats (21) into the SmaI site of the pGL3 promoter vector (Promega, Madison, WI). A reporter plasmid containing a mouse OSC promoter (Ϫ147/ϩ13) (21) was constructed to insert a PCR fragment with the NheI-HindIII sites into the cognate site of the pGL3 basic vector (Promega). A reporter plasmid containing a mouse OPN promoter (Ϫ253/ϩ28) (7) was constructed to insert a PCR fragment with the BamHI-HindIII sites into the BglII-HindIII sites of the pGL3 basic vector. A reporter plasmid for a mouse ALP promoter (Ϫ1838/ϩ81) (22) was constructed to insert a PCR fragment with the HindIII sites into the cognate site of the pGL3 basic vector. An expression plasmid of each Cbfa1 isoform was generated to insert the entire coding sequence with the Kozak sequence into the BamHI (Type I) or BglII (Type II and III) site of the mammalian expression vector pSG5 (Stratagene, La Jolla, CA), respectively. Since our expression plasmids for Type II and III Cbfa1 were constructed using Type I as the template, they have one glutamine deletion in the Q-stretch region compared with the originally reported Osf2/Cbfa1 (1, 10), but we verified the sequence of the Q-stretch region in our genomic clone (3). An expression plasmid of mouse Cbf␤/Pebp2␤ cDNA (14) was generated to insert the entire coding sequence with the Kozak sequence into the EcoRI site of the pcDNA3.1(ϩ) vector (Invitrogen, Carlsbad, CA). Xenopus BMP-4 (xBMP-4) cDNA was a kind gift from Dr. N. Ueno (National Institute for Basic Biology, Okazaki, Japan), and the expression plasmid of xBMP-4 cDNA was generated to insert the entire coding sequence with the Kozak sequence into the EcoRI site of pSG5.
Generation of Stable Transformants of Cbfa1-C3H10T1/2 cells grown to 40 -60% confluence in a 9-cm Petri dish were transfected with a total of 25 g of DNA by calcium phosphate co-precipitation (23). Each Cbfa1 expression plasmid or mock pSG5 (24 g/dish) was co-transfected with 1 g of pSV2neo (Life Technologies, Inc.), and the cells were treated with 450 -500 g/ml G418 (Life Technologies, Inc.) from 2 days after the transfection.
Transient Transfection and Luciferase Assay-C3H10T1/2 cells grown to 40 -60% confluence in a 12-well multiplate were transfected with a total of 1 g of DNA, using the transfection reagent LT-1 (Panvera Corp., Madison, WI). The reporter plasmid (0.2 g/well) was co-transfected with the indicated amount of each expression vector for each type of Cbfa1, with or without its ␤-subunit (0.1 or 0.2 g), and 0.3 g of the reference plasmid pCH110 (Amersham Pharmacia Biotech, Uppsala, Sweden). Bluescribe M13ϩ (Stratagene) was used as the carrier to adjust the DNA amount to 1 g. After 48 h, the luciferase activity was measured using a luminometer (ML-3000, Dynatec Laboratories Inc., Chantilly, VA). Relative luciferase activity was calculated after normalizing the transfection efficiency by ␤-galactosidase activity expressed by pCH110 (23).
RNA Isolation and Northern Blots-RNA isolation and Northern hybridization were performed as described (18). After final washing, the membrane was exposed to a BAS imaging plate (Fuji Film, Tokyo, Japan), and the relative signal intensity was calculated. The partiallength cDNAs of rat ALP (24) and rat type I collagen (ColI) (25) were cloned by PCR. Mouse OPN and OSC cDNA were kind gifts from Dr. S. Nomura (Osaka University Medical School, Osaka, Japan) (26,27).

Both Type I and Type II/III Isoforms of Cbfa1 Are Expressed in Adult Mouse
Bone-We first examined whether Cbfa1 isoforms (Type I and Type II/III) are expressed in bone by RT-PCR analysis using specific primers for each isoform (Fig. 1a) because both Type II and Type III Cbfa1 isoforms are suggested to be translated from the same mRNA (28). As shown in Fig. 1b, both types of transcripts were expressed in adult mouse bone. Note that the duplicate signals were detected by Type II/IIIspecific primers because of alternative splicing in this region, as reported by Xiao et al. (29) (insertion of 33 bp compared with the reported Osf2/Cbfa1 sequence, data not shown).
Transient Transfection with Type I and Type II Cbfa1, but Not with Type III Cbfa1, Induced ALP Activity in C3H10T1/2 Cells-Since ALP is one of the early differentiation markers for osteoblasts (30 -33), we investigated ALP activity in C3H10T1/2 cells transiently transfected with each isoform of Cbfa1 and/or xBMP-4. No ALP-positive cells were found in C3H10T1/2 cells without transfection. Six days after transfection with Cbfa1 isoforms, many ALP-positive cells appeared in the cells transfected with Type I or Type II Cbfa1 (Fig. 2, a-1). Transfection with xBMP-4 also induced ALP activity in C3H10T1/2 cells. The co-introduction of Type I Cbfa1 and xBMP-4 synergistically increased the number of ALP-positive cells and activity (Fig. 2, b-1 and b-2), suggesting some functional linkage between Cbfa1 and BMP-4. No ALP-positive cells were induced by transfection with Type III Cbfa1 or by transfection with mock pSG5 (Fig. 2, a-1 and b-1). The effect of the transient transfection of Cbfa1 and/or xBMP4 on ALP mRNA expression was also verified by RT-PCR (Fig. 2, a-2 and b-2).
The Effects of the Stable Transfection of Cbfa1 on the Expression of Osteoblast-related Genes Vary among Isoforms-To further investigate functional differences in the effects of Cbfa1 isoforms on osteoblast differentiation, we examined gene expressions related to osteoblast differentiation using stably transfected C3H10T1/2 cells with the three isoforms (Type I, II, and III) of Cbfa1. The expression of each exogenous Cbfa1 isoform was ensured by Northern hybridization (Fig. 3a). C3H10T1/2 cells transfected with Type I or Type II Cbfa1 exhibited the expression of ALP mRNA, but no ALP mRNA was detected in the cells transfected with Type III Cbfa1 (Fig. 3a). These results were consistent with those observed in the transient transfection experiments. OSC, OPN, and ColI gene expressions were induced or up-regulated in all cell types transfected with respective isoforms of Cbfa1. The highest induction of ALP gene expression was observed in Type I Cbfa1-transfected cells, and the highest induction of OSC gene expression was observed in the Type II Cbfa1-transfected cells, when each expression level was normalized by that of the corresponding transfected isoform of Cbfa1 (Fig. 3b). There were no apparent changes in the expression levels of OPN and ColI among isoforms of Cbfa1 (Fig. 3, a and b).
Cbfa1 Isoforms Induce Different Transcriptional Activity of the Target Genes-Cbfa1 has the ability to enhance the expression of target genes by binding to its target sequence in the promoter and/or enhancer region (7-9, 34 -37). Thus, we next examined whether the difference of the ability to enhance target gene expression (Figs. 2 and 3) is caused at the transcriptional level, using a luciferase reporter gene assay system. When p6XOSE2-luc was used as a reporter plasmid (10,21), the transfection of expression vector for each Cbfa1 isoform efficiently induced reporter gene activity (Fig. 4a). The doseresponse analysis of Cbfa1 plasmid revealed that Type II Cbfa1 induced the highest luciferase activity among the Cbfa1 isoforms (Fig. 5a). The co-introduction of each Cbfa1 isoform and its ␤-subunit induced no synergistic increase in the reporter gene activity, even in the presence of different amounts of their ␤-subunits (Figs. 4a and 5a). When the reporter plasmid used was pOSC(Ϫ147/ϩ13)-luc (10,21), which includes the mouse OSC promoter region with one functional Cbfa1 binding site, each isoform of Cbfa1 induced luciferase activity when cotransfected with its ␤-subunit expression plasmid (Fig. 4b). The dose-response analysis of Cbfa1 revealed that Type I Cbfa1 induced the highest luciferase activity with no exogenous ␤-subunit, and that each isoform induced luciferase activity similarly except for that at the highest amount of Type III Cbfa1 expression plasmid with the co-introduction of its ␤-subunit (Fig. 5b). When pOPN(Ϫ253/ϩ28)-luc (which includes the mouse OPN promoter region having one functional Cbfa1 binding site near the Ets-1 binding site (7,38)) was used as a reporter plasmid, each isoform induced luciferase activity similarly, when no exogenous ␤-subunit was expressed (Figs. 4c and 5c). With the ␤-subunit expression plasmid, Type III Cbfa1 effectively induced luciferase activity especially with the highest amount of Cbfa1 expression plasmid, as was observed with pOSC(Ϫ147/ϩ13)-luc (Fig. 5c). With the use of pALP(Ϫ1838/ ϩ81)-luc, which includes the bone/liver/kidney (B/L/K)-type ALP gene promoter region (22,39) and one putative Cbfa1 binding site, no apparent enhancement of luciferase activity was observed by the transfection of each Cbfa1 expression plasmid alone or in combination with its ␤-subunit (Figs. 4d  and 5d). A dose-response analysis of Cbfa1 also revealed no obvious enhancement of luciferase activity (Fig. 5d).

DISCUSSION
As noted earlier, there are several isoforms of Cbfa1 (1,10,16), but the function of each isoform has not been clarified. We first examined the expression pattern of isoforms of Cbfa1 in mouse bones by RT-PCR using specific primers for the Type I and Type II/III isoforms, and we found that these isoforms were expressed in adult bones (Fig. 1). This result suggested important roles of each isoform in bones. Ducy et al. (10) and Xiao et al. (29) reported that PEBP2a/Cbfa1 (Type I Cbfa1) was not expressed in osteoblasts. In contrast to these previous studies, we detected both Type I and Type II/III transcripts, because we used cDNA-transcribed mRNA from whole bone in our RT-PCR analysis.
We next examined the effects of the transfection of each Cbfa1 isoform on ALP activity in C3H10T1/2 cells, because Cbfa1-deficient mice exhibited extremely low levels of ALP activity in skeletal tissues (3). With the present transient transfection of each isoform into C3H10T1/2 cells, two Cbfa1 isoforms, Type I and Type II, induced ALP activity, but Type III isoform did not (Fig. 2a). This result was confirmed by the RT-PCR analysis (Fig. 2b). The results were also confirmed at the mRNA expression level by experiments using stably transfected C3H10T1/2 cells with each isoform (Fig. 3). These findings suggest that the Type I and Type II isoforms of Cbfa1 are involved in the early differentiation process of osteoblasts, because ALP activity is one of the early markers during osteoblast differentiation. In addition, the induction of ALP activity by transfection with the Type I isoform was synergistically increased by the co-transfection of xBMP-4 (Fig. 2b)  gests that Cbfa1 is involved in a signaling pathway of BMP-4 in the early differentiation process of osteoblasts.
We investigated whether Cbfa1 directly regulates the transcriptional activity of the ALP gene in a luciferase reporter gene assay (Fig. 4d). For this experiment, we cloned the ALP gene and used about 2 kb of the mouse ALP promoter region. Our sequence analysis and the previous report by Terao et al. (22) indicated that our reporter plasmid contains one consensus Cbfa1 binding site and three consensus-like Cbfa1 binding sites (data not shown, (8,22)). Unexpectedly, none of the three Cbfa1 isoforms used in the present study induced any apparent transcriptional activity of the ALP gene (Figs. 4d and 5d). These results raise the possibility that Cbfa1 binding sites might not be well functioning because they exist far away from the putative transcriptional start site, even if Cbfa1 directly binds there. Kobayashi et al. (39) demonstrated that the deletion of these putative Cbfa1 binding sites from a B/L/K-ALP promoter construct did not cause a significant decrease of promoter activity. Banerjee et al. (40) reported that two distal putative Cbfa binding sites did not function well in the rat osteocalcin promoter. Cbf family transcriptional factors are known to function as both negative and positive transcriptional regulators, and it is also known that the context of the binding sequences of transcription factors is essentially important for the regulation of gene expressions (35,41,42). In addition, a recent analysis of the in vivo promoter activity of the B/L/Ktype ALP gene revealed that the essential region for whole skeletal tissue expression existed in the upstream region (Ϫ4.3/ Ϫ2.0 kb) of this ALP(Ϫ1838/ϩ81) construct (43). Thus, it is likely that functional Cbfa1 binding sequences exist in the 5Ј-upstream region of ALP(Ϫ1838/ϩ81), which we used in this study. Alternatively, other functional binding sites for Cbfa1 or Cbfa1-inducible factors may exist upstream or downstream of ALP(Ϫ1838/ϩ81), as shown in the analysis concerning the TCR␣ enhancer region (35,36).
Ducy et al. (10) reported that the transient transfection of Osf2/Cbfa1 (Type III) into non-osteogenic cells such as C3H10T1/2 cells induced or increased the expression levels of mRNAs related to osteoblast differentiation. We also investigated, using stable transformants obtained from C3H10T1/2 cells, whether three Cbfa1 isoforms (Type I, II, and III) have similar activity (Fig. 3). We confirmed that all of these isoforms induced the mRNA expression of OSC and OPN and increased the expression levels of ColI, although the potency regulating the expression of these mRNAs differed among the isoforms. The Type II isoform more effectively induced OSC expression compared with the Type I and Type III isoforms, but the stimulatory effects on OPN and ColI mRNAs were not so different among each isoform. These results may be closely related to those of the luciferase reporter gene assay (Fig. 4), i.e. a higher induction rate of p6XOSE2-luc by the Type II isoform compared with the Type I and Type III isoforms, and similar induction rates of pOPN(Ϫ253/ϩ28)-luc among the three isoforms without an exogenous ␤-subunit. In the transcriptional activation of the OSC gene, the discrepancy between the results of the promoter analysis and those of the stable transfection experiment may be due to the region of the construct we used in luciferase assay, i.e. endogenous OSC gene expression is regulated by many factors and many functional elements in the gene, such as CREB/CRE (44), MSX (45), and GR/GRE (46). In the transcriptional activation of the OPN gene, the results of the promoter analysis and those of the stable transfection experiment correlated very well; this reflected that the important regulatory element in osteoblastic lineage cells exists in the regions of our construct (38,47). In addition, the highest transcriptional activation of the OPN gene was demonstrated with Type III Cbfa1 transfection when its ␤-subunit was cotransfected. Taken together, our findings suggest that Cbfa1 isoforms have different functions in the regulation of osteoblast-related gene expression.
The results of the present reporter gene analysis suggested that at least two kinds of regulatory mechanisms are involved in the Cbfa1-induced transcription of target genes (Fig. 4). First, Cbfa1 itself may be able to activate target genes effectively in the case of the OPN gene and p6XOSE2 construct. Second, Cbfa1 may regulate the transcription of the target genes in collaboration with its ␤-subunit in the case of the OSC and OPN genes. However, Ducy et al. (10) reported that a single transfection of Osf2/Cbfa1 (Type III Cbfa1) induced the transcription activity of the OSC gene, using the same region of the mouse OSC promoter as that used in the present study. The discrepant results between the Ducy study and ours might arise from the different experimental protocols or different cell usage (i.e. different interaction with cell-specific co-factors, as described below), judging from previous observations (15, 34 -37). One of the candidate co-factors is Cbf␤ (also known as heterodimerizing partner of Cbfa), which is reported to be abundantly expressed in various kinds of cells (14), because we detected stronger activation by far of the pOSC(Ϫ147/ϩ13)-luc and pOPN(Ϫ253/ϩ28)-luc in the presence of exogenous Cbf␤ expression.
Different activities of transcription factors often occur among isoforms generated by alternative splicing. For example, AML1c (a human homolog of mouse Pebp2␣B), one of the runt domain family gene transcripts, is 14 amino acids longer than AML1a in the N-terminal region like til-1 (Type II Cbfa1). No functional difference between AML1a and AML1c has as yet been uncovered, although many functional domains concerning DNA binding, heterodimerization, and transactivation in its C-terminal region have been identified (15). As for SREBP1 (sterol regulatory element binding protein), one of the essential transcription factors in fatty acid metabolism, its two isoforms with different N-terminal structures (SREBP1a and SREBP1c) were shown to have different abilities to induce the target gene expression in experimental animals and cultured cells (48), as is also the case for progesterone receptor (A and B), which is one of the ligand-dependent transcription factors and is essential for sex hormone function (49). Different activities of transcription factors among isoforms generated by alternative splicing might result from the different interaction with cofactors such as co-activators and co-repressors (50). Cbfa1 has been reported to interact directly or indirectly with cofactors including not only its ␤-subunit (14, 15) but also C/EBP (34), Ets-1 (35,38), Myb (37), and ALY (36) on the target gene enhancers and/or promoters. There have been no reports concerning cofactors interacting with Cbfa1 in its N-terminal region; the identification of the coupling protein in this region might clarify the reasons for the functional differences among the isoforms of Cbfa1.