René St-Arnaud¹, G. Antonio Candeliere¹, and Shoukat Dedhar²

¹Genetics Unit, Shriners Hospital, and Departments of Surgery and Human Genetics, McGill University, Montréal (Québec) Canada H3G1A6

²Division of Cancer Research and Department of Medical Biophysics, University of Toronto, Reichmann Research Building, Sunnybrook Health Science Centre, Toronto (Ontario) Canada M4N 3M5

Received 07/05/96; Accepted 07/12/96; On-line 08/01/96

Association with multiple protein targets appears like a common mechanism through which activators can fully stimulate transcription. The dimerization of the VDR with itself or with RXR, RAR,

and T3R has been described above. This section will review the interaction of the VDR with other partners. These protein-protein interactions are hypothesized to play key regulatory roles in the control of vitamin D-dependent gene transcription.

4.1. Interactions with the basic transcriptional machinery

Recent genetic and biochemical evidence has revealed that the activation of gene expression involves sequence-specific DNA binding transcriptional activators such as the VDR, the basic or general initiation factors TFIIA, B, D, E, F and H, and a third class of molecules that mediate interaction between the activators and the basal factors via direct protein-protein contacts (reviewed in ref. 49). At least one of the protein targets of the VDR within the basic transcriptional machinery has been identified: using the powerful yeast two-hybrid screening methodology (50), MacDonald et al. (51) have shown that the VDR interacts directly with TFIIB. The domains of the two proteins involved in the interaction were identified using deletion mutants both in vitro and in vivo (51, 52). Interestingly, cooperative interactions between VDR and TFIIB were demonstrated using transient transfection assays in embryonic cells; cooperativity was not observed when the same experiments were attempted in fibroblasts (52). A probable interpretation of these results is that the TFIIB-VDR interaction may be modulated by cell-type-specific accessory factor(s) (52). These accessory factors could be functionally equivalent to the transcriptional coactivators that have been shown to be essential for activated gene transcription (49; 53). Cell-specific coactivators have recently been characterized (54, 55). The identification of coactivator molecules interacting with the VDR will undoubtedly increase our comprehension of the molecular mechanisms implicated in vitamin D-stimulated transcriptional responses.

4.2. Interactions with NF1

Recent work suggests that interactions between different families of transcription factors and steroid hormone receptors may provide another level of regulation of receptor function. One such class of factors is the NF1 gene family.

The nuclear factor 1 (NF-1) family of proteins have been shown to be implicated in the transcriptional activation of many cellular and viral genes (56-58) as well as in the stimulation of adenovirus DNA replication (59). NF-1 family members bind the CCAAT box (hence their original designation as CCAAT-binding transcription factors or CTF), making contact with TGG residues on the opposite strand to activate transcription (59). The name 'CCAAT-binding transcription factors' was deemed confusing considering the fact that a variety of transcription factors, some of which definitely unrelated to NF-1, had been shown to bind the CCAAT motif (examples include C/EBP and CBF/NF-Y). Thus, NF-1 was selected as the 'family name' and the various gene family members received letter designations based either on their sequence of publication or some connection with previous names. Thus CTF/NF-1 has been replaced by the name NF1-C. The NF-1 proteins consist of a family of related polypeptides generated by the alternative splicing of RNA (60). To date, four family members encoded by multiple genes have been cloned (61).

Mutation of the NF-1 binding site in the mouse mammary tumor virus (MMTV) long terminal repeat abolishes the glucocorticoid responsiveness of this promoter (62). Similarly, mutating the NF-1 site within the vitellogenin promoter inhibits estrogen inducibility (63). Thus functional interactions between nuclear hormone receptors and NF-1 family members appear to influence transcriptional activation by steroid hormones. Such synergism has also been described between the VDR and NF1-C, as the VDR transactivates 10 times more strongly when a NF-1 binding site is added to a reporter plasmid containing a DR-3 type VDRE (64).

As previously mentioned, we identified VDR and RXR as components of the complex that bound the c-fos VDRE (48). However, our results also showed that a putative NF-1 family member bound the response element in conjunction with the nuclear hormone receptors. Interestingly, the study of the promoter elements implicated in the control of the expression of the human c-fos gene has revealed the importance of an element centered at a position corresponding to the murine c-fos VDRE and hypothesized to bind members of the NF-1 family (65).

Three lines of evidence support the involvement of a NF-1 family member as a component of the 1,25 (OH)₂D₃ -responsive complex binding to the c-fos VDRE. First, the core sequence of the response element (5'-GCCAAG-3') responsible for binding specificity (48) is a high affinity NF-1 binding site (59). Second, a canonical NF-1 binding site (from the adenovirus E5 promoter) efficiently competed for the binding of the vitamin D-responsive complex to the c-fos VDRE (not shown). Finally, antibodies directed against NF-1 recognized the VDRE-bound complex in gel retardation assays (48). However, our results suggest that previously identified members of the NF-1 family (NF1-C1, NF1-C2, and NF1-C3) cannot interact with VDR and RXRalpha to bind the c-fos VDRE: mixing of recombinant VDR and RXRalpha proteins with nuclear extracts from HeLa cells, which express high levels of functional NF1-C1, 2, and 3 binding activity (60), failed to reconstitute the binding activity to the c-fos VDRE (48). Moreover, we have expressed the cDNAs for NF1-C1, NF1-C2, and NF1-C3 using in vitro transcription and translation systems and have failed to demonstrate interactions with the recombinant receptors (not shown). Moreover, the vitamin D-responsive complex binding the element could only be detected in nuclear extracts from bone cells (48). This supports the notion of a bone-specific NF-1 family member involved in the regulation of target genes in osteoblasts.

Our results suggest a model of trimeric binding where the NF-1 family member anchors the complex to the response element, and the VDR and RXR proteins are then tethered to the site by a combination of protein-protein and protein-DNA interactions. This model raises the intriguing possibility that interactions with the nuclear hormone receptors would endow a ligand-independent transcriptional activator, the NF-1 family member, with dual ligand-switching capabilities, i.e. the capacity to respond both to 1,25 (OH)₂D₃ and 9-cis retinoic acid.

The characterization and molecular cloning of the osteoblast-specific NF-1 factor interacting with the VDR, an ongoing effort in our laboratory, should further our understanding of osteoblastic differentiation and provide new insight into the control of bone-specific gene expression and the molecular mechanism of action of vitamin D.

4.3. Interactions with calreticulin

A number of local and systemic factors have been shown to influence gene expression in osteoblasts. Of particular importance are the effect of attachment of preosteoblasts to various components of the extracellular matrix as well as the modulation of gene expression mediated by members of the nuclear hormone receptor family.

Binding to extracellular matrix components mediated through the integrin family of cell surface receptors modulates gene expression and differentiation of bone cells. For example, differentiation of canalicular cell processes was observed following contact of osteoblastic cells with laminin (66). At the molecular level, both constitutive and retinoic acid (RA)-induced gene expression is affected when preosteoblastic cells are plated on various substrates (67). These observations suggest that factors that can influence substratum attachment via integrin receptors as well as factors that modulate the function of nuclear hormone receptors may have dramatic effects on gene expression, differentiation and function of osteoblasts. We have shown that the multifunctional protein calreticulin can regulate the affinity state of integrins (68, 69) and inhibit DNA binding by nuclear hormone receptors (70). Interestingly, our most recent results also show that calreticulin plays an important role in osteoblast function and bone formation (71).

Calreticulin has been considered to be a major calcium binding protein of the endoplasmic reticulum of non-muscle cells (72-74). The functions of calreticulin, however, still remain obscure, and as described below, the protein has all the hallmarks of a multifunctional protein. The primary sequence of calreticulin contains putative recognition sequences for phosphorylation by protein kinase C (in the N terminal domain), casein kinase II, and tyrosine kinase (74). In addition calreticulin also appears to have a sequence with marked similarity to the active site of protein kinase C. In fact, calreticulin has recently been described as a phospho-protein with possible autokinase activity (75). Other functional motifs in calreticulin include a nuclear targeting signal, a proline-rich region, and a concentration of acidic residues in the C-terminal of the protein, ending with the ER retention signal, KDEL. Consistent with these putative signals, calreticulin has been localized in the ER and recently also in the nucleus (76). Thus, these interesting features of calreticulin suggest that it may have multiple functions resulting from covalent modifications, e.g. by phosphorylation or calcium binding, and depending on its intracellular localization.

Calreticulin has previously been demonstrated by us to function as an integrin binding protein. Specifically, calreticulin binds to the highly conserved KXGFFKR motif present in the cytoplasmic domains of all integrin -subunits (77). A computer analysis of protein data banks for the presence of the KXGFFKR sequence motif in other proteins revealed that a highly homologous sequence, KGFFKR, was present in the DNA binding domain of all known members of the steroid receptor family (78). In some of the receptors there is a conservative substitution of the G (glycine) to a V (valine) or an A (alanine). Also of interest is the observation that the second amino acid in the integrin motif KXGFFKR which is the most variant in this family, is not present in the steroid receptors. This motif appears in the DNA binding domain of steroid receptors between the two Zn²⁺ fingers, but adjacent to the first Zn²⁺ finger. Site-directed mutagenesis of amino acids in this motif results in either the alteration of DNA sequence specificity, or the complete abrogation of the binding of the receptor to its DNA element (79). Thus, this sequence motif is a DNA binding motif involved in the binding of steroid receptors to DNA during the transcriptional regulation of target genes by these receptors. By analogy with the interaction of calreticulin with the KXGFFKR sequence in the cytoplasmic domain of integrin -subunits, it was felt that calreticulin may also bind to steroid receptors via this motif and thus modulate gene expression. Thus some aspects of cellular differentiation and function may be regulated by a calreticulin-dependent signal transduction pathway linking the binding of extracellular matrix components by the integrin family of cell surface receptors to the coordinate regulation of gene expression in the nucleus (Fig.3) (80).

Fig. 3: Schematic representation of the calreticulin-binding motif (KXGFFKR) in the cytoplasmic domain of integrin alpha-subunits and the DNA-binding domain of nuclear receptors (see text). The diagram also illustrates that occupation of integrin KXGFFKR by calreticulin locks the integrins into a 'high-affinity' state that allows ligand binding. Interaction of calreticulin with the KXFF[K/R]R motif of nuclear hormone receptors inhibits receptor activity and receptor-dependent gene transcription.

Indeed, we and others have demonstrated that calreticulin can interact with members of the steroid hormone and nuclear receptor family and inhibit their binding to DNA response elements (70, 71; 81-83). This inhibition results in an inhibition of transcription via these receptors. Calreticulin has now been demonstrated to interact in a functional manner with the androgen receptor (70), the GR (81), the RAR (70), the peroxisome proliferator-activated receptor (83) and the VDR (71; 82). Purified calreticulin inhibited the binding of the VDR to characterized VDREs in gel retardation assays (71; 82). This inhibition was due to direct protein/protein interactions between the vitamin D receptor and calreticulin (71) and resulted in the abrogation of the vitamin D-dependent transcriptional activation through the VDRE (71; 82). Hence calreticulin may be a major regulator of gene expression via nuclear hormone receptors and as such may have a profound effect on many aspects of cellular physiology, including the response of bone cells to vitamin D.

We have used a gain-of-function strategy to examine this putative role of calreticulin in the regulation of vitamin D-dependent transcriptional activation and mineralization in osteoblastic cells. Northern blot assays revealed that expression of calreticulin transcripts declined during the differentiation of MC3T3-E1 osteoblastic cells (71). Interestingly, preliminary data from our laboratory suggests expression of calreticulin transcripts in preosteoblastic cells from embryonic bone (not shown), supporting the results obtained with the in vitro model of osteoblastic differentiation.

Osteoblastic cells were transfected with calreticulin expression vectors; stable transfected cell lines overexpressing recombinant calreticulin were established and assayed for vitamin D-induced gene expression and the capacity to mineralize. Constitutive calreticulin expression inhibited basal and vitamin D-induced expression of the osteocalcin gene, whereas osteopontin gene expression was unaffected. This pattern mimicked the gene expression pattern observed in parental cells prior to down-regulation of endogenous calreticulin expression. Our results have demonstrated that calreticulin inhibits VDR binding to both the osteopontin and the osteocalcin VDRE (71). However, the two genes responded dissimilarly when the calreticulin-expressing clones were challenged with vitamin D (71). Despite differences between the sequences of the two VDREs (22; 39; 84) that could affect VDR binding affinity to each response element (20), we were unable to detect any difference in the capacity of calreticulin to inhibit VDR binding to each VDRE. Thus the structure of the response element cannot seem to account for the variation in transcription. It is more likely that the expression of each gene is dictated by the structure of the entire promoter region, and not just a single element. Indeed, developmental and physiologic responsiveness of the osteocalcin gene to 1,25 (OH)₂D₃ has been shown to involve, in addition to specific ligand-receptor interactions and binding to DNA response elements, a complex set of event including enhanced receptor gene expression, critical receptor phosphorylation, formation of multiple receptor-protein complexes, as well as overlapping DNA elements at the VDRE locus (85 and references therein). It is likely that calreticulin influences only one aspect of this complex regulatory pathway, namely the interaction of the ligand-bound receptor to its response element, as we have demonstrated in vitro.

Aberrant calreticulin expression in bone cells perturbs the differentiation and function of these cells. In long-term cultures of parental or vector-transfected cells, 1,25 (OH)₂D₃ induced a 2- to 3-fold stimulation of ⁴⁵Ca accumulation into the matrix layer (17; 71). Constitutive expression of calreticulin inhibited the 1,25 (OH)₂D₃ -induced ⁴⁵Ca accumulation (71). This result correlated with the complete absence of mineralization nodules in long-term cultures of calreticulin-transfected cells (71).

As previously mentioned, calreticulin binds to a-integrin subunits (68) and recent results suggest that this interaction can modulate the affinity state of integrins (69). Moreover, calreticulin can modulate nuclear hormone receptor-dependent gene expression (70, 71; 81-83). Taken together, these observations support the existence of a calreticulin-modulated signal transduction pathway linking substratum attachment via integrin receptors to the control of gene expression. Our results further support an important role for this pathway in the regulation of osteoblastic differentiation and function.