Selective Distal Enhancer Control of the Mmp13 Gene Identified through Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR) Genomic Deletions*

Background: Mmp13 is vital to bone homeostasis and controlled by a plethora of stimuli. Results: Mmp13 is modulated by distinct distal enhancers for basal (−30 kb) and vitamin D regulation (−10 kb). Conclusion: Enhancer deletions lead to altered transcription factor occupancy and expression for Mmp13. Significance: Specific CRISPR deletions reveal the repressive secondary effect of VDR, coordinated multi-enhancer gene control, and distal basal regulation of Mmp13. Matrix metalloproteinase 13 (Mmp13, collagenase-3) plays an essential role in bone metabolism and mineral homeostasis. It is regulated by numerous factors, including BMP-2, parathyroid hormone, and 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3), through transcription factors such as Runt-related transcription factor 2 (RUNX2), CCAAT/enhancer-binding protein β (C/EBPβ), OSX, and vitamin D receptor (VDR). During osteoblast maturation, the basal expression of Mmp13 and its sensitivity to 1,25(OH)2D3 are strikingly increased. In this report, ChIP-sequencing analysis in mouse preosteoblasts revealed that the Mmp13 gene was probably regulated by three major enhancers located −10, −20, and −30 kb upstream of the gene promoter, occupied by activated VDR and prebound C/EBPβ and RUNX2, respectively. Initially, bacterial artificial chromosome clone recombineering and traditional mutagenesis defined binding sites for VDR and RUNX2. We then employed a CRISPR/Cas9 gene editing approach to delete the −10 and −30 kb Mmp13 enhancers, a region proximal to the promoter, and VDR or RUNX2. VDR-mediated up-regulation of Mmp13 transcription was completely abrogated upon removal of the −10 kb enhancer, resulting in a 1,25(OH)2D3-directed repression of Mmp13. Deletion of either the −30 kb enhancer or RUNX2 resulted in a complete loss of basal transcript activity and a ChIP-identified destabilization of the chromatin enhancer environment and factor binding. Whereas enhancer deletions only affected Mmp13 expression, the RUNX2 deletion led to changes in gene expression, a reduction in cellular proliferation, and an inability to differentiate. We conclude that the Mmp13 gene is regulated via at least three specific distal enhancers that display independent activities yet are able to integrate response from multiple signaling pathways in a model of activation and suppression.

The primary function of Mmp13 (collagenase-3, Mmp- 13) in osteoblasts and chondrocytes is to facilitate the breakdown of extracellular proteins, such as collagen during bone remodeling (1). However, Mmp13 also plays non-skeletal roles and has been implicated in the invasiveness and progression of different types of cancers, including breast, lung, esophageal, multiple myeloma, and chondrosarcomas (2)(3)(4). The Mmp13 gene was first cloned from a human breast tumor (5) and was found to be part of a larger family of matrix metalloproteases. The Mmp13 knock-out mouse displays profound defects in growth plate cartilage and endochondral ossification and exhibits increased interstitial collagen accumulation (6). Studies also link the loss of Mmp13 in an ApoE null background to an increase in atherosclerosis due to this same collagen accumulation (7). Importantly, Mmp13 can be regulated by a litany of different regulatory factors and cellular stimuli. Early studies found that Mmp13 was regulated by BMP2 (8), PTH 2 (9), estrogens (10), and heparin (11). It was also shown to be responsive in human chondrosarcoma cells to FGF-2 (fibroblast growth factor 2) (2). Mmp13 expression is similarly increased by cytokines, such as IL-1␤ and TNF␣, activated by cFOS through an AP-1 site located near the gene promoter (12), yet inhibited in osteoblast cultures by IGF-1 (13). In studies of PTH action, Mmp13 was found to be regulated by Runt-related transcription factor-2 (RUNX2) through a Runt binding domain DNA sequence located near an AP-1 site adjacent to the Mmp13 promoter (9). Additional studies revealed that both PKC and PKA signaling pathways are active at the AP-1 and RUNX2 binding sites, respectively (14,15). This ongoing work has focused much of the molecular investigation of Mmp13 regulation by PTH and other hormones on these specific promoter proximal sites in bone cells.
RUNX2 is an essential transcription factor for the generation of osteoblasts and for the formation of chondrocytes because mice lacking this transcription factor are unable to complete these required tasks for normal bone formation (1,16,17). RUNX2 and its family members, RUNX1 and RUNX3, bind to DNA using a common heterodimer partner CBF␤ to coordinate the expression of numerous genes (18). RUNX2 is known to drive the expression of genes responsible for the osteoblast phenotype, such as osteopontin (Spp1), osteocalcin (Bglap), vitamin D receptor (Vdr), osterix (Sp7), and Runx2 itself (19). Similar to Mmp13, aberrant expression of RUNX2 manifests in numerous cancerous cells (20,21). The Mmp13 gene was also found to be regulated by the VDR-activating ligand 1,25(OH) 2 D 3 in rats as well as in mouse MC3T3-E1 cells (22,23). This latter study suggested that the induction of Mmp13 by 1,25(OH) 2 D 3 was due to an interaction at an AP-1 site located near the transcriptional start site (TSS) but not at an adjacent RUNX2 site. However, no VDR DNA binding response element (VDRE) was identified near Mmp13 in that investigation.
Recent ChIP-seq studies suggest that many if not most genes are regulated through multiple elements located distal to promoters, often within introns and in surrounding intergenic regions many kilobases from the promoters of the genes that they regulate (24,25). We have recently cataloged genome-wide binding patterns for RUNX2 and C/EBP␤ as a function of osteoblast differentiation and have shown that the binding sites for these transcription factors are predominantly located tens if not hundreds of kilobases from their target genes and that the cistromes (collection of all cisacting transcriptional elements) for both of these master regulators change in a dynamic fashion during the differentiation process (19,24). Interestingly, the binding of RUNX2 and C/EBP␤ can occur either to a single enhancer of a target gene or to separate regulatory enhancers that contribute to the modulation of the same gene but in both cases integrate the transcriptional output. We also discovered that these two factors were often bound to genes that were in turn regulated by 1,25(OH) 2 D 3 and that the VDR was also frequently present at these distal enhancer sites as well (19,24), suggesting the presence of a general "osteoblast enhancer complex." These observations, together with the absence of an unbiased analysis of the Mmp13 landscape, raise the possibility that in addition to the promoter proximal elements investigated over the past decade, this gene may also be controlled by more distal regulatory regions.
In this report, we identified three specific distal enhancers located Ϫ10, Ϫ20, and Ϫ30 kb upstream of the Mmp13 TSS using ChIP-seq data obtained from previous studies (26). Subsequent analyses revealed that the VDR bound almost exclusively to the Ϫ10 kb enhancer, and that this enhancer mediated the actions of 1,25(OH) 2 D 3 at the Mmp13 gene, strongly supported by evidence from traditional mutagenesis, bacterial artificial chromosome (BAC) clone recombineered reporters, and CRISPR/Cas9 genomic deletion. The basal level of expression of the Mmp13 gene was highly regulated by RUNX2; this control, however, was exerted via transcription factor binding at the enhancer located Ϫ30 kb from the promoter proximal region. Both deletion of this enhancer and genetic knockout of RUNX2 in the cell line abrogated virtually all basal transcrip-tional activity of Mmp13. Finally, ChIP analysis in daughter cells containing the specific enhancer deletions revealed that the loss of the Ϫ30 kb region or deletion of RUNX2 expression not only eliminated RUNX2 binding but disrupted transcription factor interaction at both the Ϫ10 kb enhancer and the promoter proximal region. Here we describe a complex mechanism wherein the interplay between RUNX2, C/EBP␤, and the VDR is evident at individual enhancers that control Mmp13 expression. This interplay is capable of mediating both induction and repression of the gene, mechanisms that may have broad implications for 1,25(OH) 2 D 3 -mediated regulation in additional target cells.

EXPERIMENTAL PROCEDURES
Reagents-1,25(OH) 2 D 3 was obtained from SAFC Global (Madison, WI). All ChIP antibodies were purchased from Santa Cruz Biotechnology, Inc., Millipore (Billerica, MA), or Abcam (Cambridge, MA), as reported previously (24). All quantitative real-time PCR (qPCR) reagents (Fast Start SYBR Green Master Mix (with rox)) were obtained from Roche Applied Science, and TaqMan gene expression assays were obtained from Life Technologies, Inc. (Applied Biosystems, Foster City, CA). All qPCR was conducted on a StepOnePlus from Applied Biosystems. Primers were obtained from IDT (Coralville, IA), and the sequences are available upon request and on our laboratory's website.
Cell Culture and Differentiation-MC3T3-E1 cells (preosteoblast; POB) (early passage line from Sudo et al. (27)) were cultured in minimum Eagle's medium ␣ (MEM␣) modification supplemented with 10% heat-inactivated fetal bovine serum from Hyclone (Logan, UT) and 1% penicillin-streptomycin from Invitrogen. UAMS-32 (28) and UAMS-32PB (UAMS-PB, blasticidin-resistant PTH receptor expressing UAMS) (29) (gift from Charles O'Brien, University of Arkansas Medical School) were cultured in MEM␣ with 10% fetal bovine serum and 1% penicillin-streptomycin. For differentiation, cells were grown to confluence and then transferred to differentiation medium (10 mM ␤-glycerophosphate and 50 g/ml ascorbic acid) for the indicated lengths, replenishing the medium every 2-3 days until the assay. Cells were stained for calcium content with Alizarin Red as reported previously (19). Calcium content was quantitatively measured by decalcifying the cells with 0.6 N HCl at 4 ºC for 24 h, followed by a triplicate analysis using the o-cresolphthalein complexone method (Sigma-Aldrich) normalized to total protein content of the calcium-stripped cells.
Gene Expression Analysis-Cells were treated for 24 h with the indicated factors, and 1 g of isolated total RNA was reverse transcribed using the High Capacity cDNA Kit (Applied Biosystems) and then diluted to 100 l with RNase/DNase I-free water. qPCR was performed using primers specific to a select set of genes by TaqMan analyses.
Molecular Cloning-The pCH110-␤ galactosidase reporter plasmid was described previously (30). All gene-specific pTK plasmids were constructed by cloning the appropriate mouse DNA fragments obtained through DNA amplification of genomic MC3T3-E1 DNA into the pTK-luc vector using BamHI, SalI, and/or HindIII restriction sites. Mutations were introduced using the QuikChange site-directed muta-genesis kit (Stratagene, La Jolla, CA) as per the manufacturer's protocol and recommendations. BAC luciferase reporter constructs were created via recombineering using the BAC clone RP23-355F15 and methods established previously (31,32). All CRISPR plasmids were constructed by cloning into the BbsI site as described previously (33). RUNX2 lentiviral overexpression plasmid (pLeGO-RUNX2-2A-GFP) was created through Gibson isothermal assembly (34) of PCR cDNA for mouse RUNX2 derived from pCMV-RUNX2 (35) and pLeGO-G (36) (Addgene 27347). The RUNX2 cDNA was inserted together with a P2A sequence (GGATCCGGAGCCAC-GAACTTCTCTCTGTTAAAGCAAGCAGGAGACGTGGAA-GAAAACCCCGGTCCT) prior to the GFP of the LeGO-G plasmid. RUNX2 was PCR-amplified using the primers available on our laboratory's Web site, gel-purified, and combined through Gibson assembly with an alkaline phosphatase-pretreated, BamHI-digested LeGO-G plasmid. The resulting clones were sequenced for verification of the appropriate RUNX2-2A insertion and confirmed by Western blot for size (slight shift due to additional P2A sequence 3Ј) and appropriate expression.
Analysis of Enhancer Activity Using Reporter Genes-MC3T3-E1 cells were seeded into 24-well plates in MEM␣ containing 10% FBS at a concentration of 5.0 ϫ 10 4 cells/well and transfected 24 h later with Lipofectamine PLUS (Invitrogen) in serum-and antibiotic-free medium. Individual wells were cotransfected with 250 ng of a luciferase reporter vector and 50 ng of pCH110-␤gal. After transfection, the cells were cultured in medium supplemented with 20% FBS and 100 nM 1,25(OH) 2 D 3 or ethanol vehicle. Cells were harvested 18 h after treatment, and the lysates were assayed for luciferase and ␤-galactosidase activities in triplicate Ϯ S.E. as described previously (37). Luciferase activity was normalized to ␤-galactosidase activity in all cases.
Lentiviral Expression and Cell Proliferation-The pLeGO-RUNX2-2A-GFP or parental pLeGO-GFP plasmids were cotransfected with packaging plasmids pMD2.6 (Addgene 12259) and psPAX (Addgene 12260) into 293FT cells (Invitrogen) using FuGene HD (Promega) in MEM␣. After 24 h, medium containing virus was removed from transfected 293FT cells, filtered (Millex 0.33-m filter, Millipore), and placed on the UAMS-PB or target cells using 4 g/ml Polybrene (Millipore) for 24 h. Infection of UAMS-PB or target cells was carried out for 72 h, replenishing with 293FT viral medium every 24 h. Cells were passaged a minimum of 7 times prior to FACS sorting for GFP-positive cells in standard non-viral medium. GFP-positive cells were collected and propagated for Western blot, gene expression, or cell proliferation assay. All cells were plated at 5,000 cells/well in a 24-well plate for the cell proliferation assay. Triplicate plated cells were trypsinized and counted with a standard hemocytometer over 5 days.
ChIP-seq-Chromatin immunoprecipitation was performed as described previously (38). Briefly, UAMS-PB or CRISPR-deleted clonal cells were treated for 3 h with vehicle or 100 nM 1,25(OH) 2 D 3 as reported previously (39). Kidney and intestine samples were collected after a 1-h 10 g/kg dose of 1,25(OH) 2 D 3 into C57bl/6 mice as described previously (40). Samples were subjected to immunoprecipitation using either a control IgG antibody or the indicated experimental antibody (19,24). All ChIP and ChIP-seq methodologies, including statistical information and data processing, were performed as reported recently (19).

Mmp13 Is Regulated by Three Distinct Distal Enhancers
Located Upstream of the TSS-Mmp13 transcription is up-regulated during early precursor cell differentiation into mature, mineralizing osteoblasts (14) and modulated by a myriad of regulatory factors that include 1,25(OH) 2 D 3 (23). These conclusions are confirmed by the observations in Fig. 1, which demonstrate that basal expression of Mmp13 in preosteoblastic MC3T3-E1 cells (POBs) was indeed increased substantially in differentiated osteoblasts (OBs) following a 15-day exposure to osteogenic medium and that Mmp13 expression was up-regulated by 1,25(OH) 2 D 3 in both states of differentiation. Interestingly, however, the absolute level of Mmp13 up-regulation was much more robust in mature cells, as reported previously for a cohort of additional osteoblastic genes (24). To determine the molecular basis for this complex regulation, we searched for active binding sites for RUNX2, C/EBP␤, and VDR at the Mmp13 gene locus as documented in Fig. 2 using ChIP-seq data sets that we had obtained previously (19,24). These data sets also included analyses of the levels of key histone marks across the genome that represent signatures of both enhancer structure and activity (histone H3 Lys-4 monomethylation and histone H4 Lys-5 acetylation) and transcriptional output (histone H3 Lys-36 trimethylation). The ChIP-seq data tracks for these features at the Mmp13 locus in either POBs or OBs are depicted in Fig. 2A, where each track is composed of a normalized tag density data stream in the presence of either vehicle (yellow) or 1,25(OH) 2 D 3 (blue) with overlapping occupancy appearing green. As can be seen, three major upstream regions of activity in POBs were apparent for VDR, C/EBP␤, and RUNX2 centered at Ϫ10, Ϫ20, and Ϫ30 kb, respectively. C/EBP␤ was at the Ϫ30 kb region as well, and a small amount of RUNX2 binding was similarly apparent at the Ϫ10 kb region and particularly near the promoter but largely restricted to the differentiated OBs. Whereas both C/EBP␤ and RUNX2 were prebound to these sites (green), DNA binding of the VDR required the presence of 1,25(OH) 2 D 3 (blue). Of major significance, while histone H3 Lys-36 trimethylation strongly decorated the Mmp13 gene body as was expected of this mark, each of the three factor binding regions was reproducibly marked by elevated levels of histone H3 Lys-4 monomethylation and histone H4 Lys-5 acetylation, modifications that often specify the location of regulatory enhancers. VDR binding decreased at the Ϫ10 kb region in OBs, due to changes in the epigenetic landscape but also to a decrease in VDR expression in OBs (24). This finding raises the possibility that, like other gene targets in this cell type, the more robust response of Mmp13 to 1,25(OH) 2 D 3 was probably facilitated through increased cooperative interactions among the three enhancers. Finally, the results documented in Fig. 2B revealed that although VDR binding across the Mmp13 locus was similar in other osteoblast lineage cells (mouse mesenchymal stem cells and osteoblastic UAMS-PB cells), receptor binding at the Ϫ10 kb region was specific for bone cells and absent in murine intestine and kidney tissues, where Mmp13 is expressed at extremely low levels. These analyses suggest that at least three distal enhancers are likely to play a significant role in both the basal and 1,25(OH) 2 D 3 -regulated expression of the Mmp13 gene.
VDR-mediated Up-regulation of Mmp13 Expression Is Controlled through the Ϫ10 kb Region-Although the VDR was found predominantly bound to the Ϫ10 kb region, modest amounts of the VDR were also localized to the more distal regions of the gene (Ϫ20 and Ϫ30 kb) as well as to the promoter proximal region. As a consequence, we cloned small DNA segments marked by each of these peaks of activity (500 -1,000 bp) into a luciferase reporter vector, transfected the plasmids into POBs, and assessed their relative contribution to the up-regulation of Mmp13 expression in response to 1,25(OH) 2 D 3 . In Fig.  3A, although the Ϫ10 kb region conferred a striking dose-dependent transcriptional up-regulation of the reporter, none of the other constructs exhibited such independent regulatory capability. These data suggest that this enhancer probably plays a dominant role in 1,25(OH) 2 D 3 activation of Mmp13 and that VDR binding to the regions outside the Ϫ10 kb location are incapable of mediating such regulation and could be due to chromatin cross-linking artifacts. A position weight matrixbased analysis via MatInspector (Genomatix) was used to identify repeated DR3-type VDREs in the Ϫ10 kb region (41,42). Two candidates of the sequence, AGGTGAgcgGGTTCA (VDRE1) and GGTTCAaacGGAATT (VDRE2), that were separated by ϳ400 bp were found located directly under the ChIPseq peak maxima (Fig. 3B). To confirm the functionality of these potential binding sites, we introduced separate and combinatorial mutations (AGG to TTT in VDRE1 and GGT to AAA in VDRE2) into the cloned Ϫ10 kb segment and contrasted the ability of these mutant segments with that of the unmodified parental vector to mediate response to 1,25(OH) 2 D 3 . In Fig. 3C, mutations in either VDRE1 or VDRE2 strongly reduced response to the hormone, and the combination of mutations in both VDREs completely abrogated the ability of 1,25(OH) 2 D 3 to induce an up-regulation of the TK reporter construct. These results suggest that two VDREs located in the Ϫ10 kb region are the exclusive mediators of 1,25(OH) 2 D 3 action in our constructs.
RUNX2 Activity Is Mediated through the Ϫ30 kb Region of the Mmp13 Gene-A similar approach was used to identify response elements for the RUNX2 transcription factor (RUNX2-REs). Accordingly, a pair of highly ranked RUNX2-REs identified by position weight matrix analysis was discovered 10 bp apart in the Ϫ30 kb enhancer directly under the peak maxima, as illustrated in Fig. 3B. We tested the activity of the parental construct as well as a version in which the two RUNX2-REs located at the center of the Ϫ30 kb region had been mutated (GTGG to TTTT in RUNX2-RE-1 and CCGC to CTTT in RUNX2-RE-2) by cotransfecting these reporter plasmids together with a pCMV-RUNX2 expression construct (or pCMV-empty control) into MC3T3-E1 cells. In Fig. 3D, although increased reporter activity was evident in the parental construct when exogenous RUNX2 was present, this activity was significantly reduced following the introduction of mutations at the RUNX2-REs. These experiments localize functional regulatory elements for RUNX2 to the Ϫ30 kb region and indicate that at this site, RUNX2 is fully capable of activity independent of its actions at other sites. The absence of RUNX2-inducible activity at the promoter proximal region of Mmp13 was surprising in light of the minor appearance of RUNX2 at this region and of previously reported studies as well (9). The majority of RUNX2 binding at this site and at the Ϫ10 kb region is largely observed only in mature OBs. Thus, it is possible that RUNX2 regulatory activity manifests itself at these sites only after differentiation because RUNX2 binding is indeed dynamic.
The Ϫ10 kb Enhancer Mediates 1,25(OH) 2 D 3 Action in the Context of the Mmp13 Gene Locus-The above approach assesses enhancer activity that was manifested independently of both the Mmp13 gene transcription unit and its natural chromatin environment. To address this final issue, we obtained a 182-kb BAC clone spanning the entire Mmp13 transcription unit and its associated upstream regulatory segments (chr9: 7,171,215-7,353,797), inserted a luciferase reporter cassette (P2A-luciferase-TK-neomycin) into the 3Ј-untranslated region of the gene, and stably introduced this construct into MC3T3-E1 cells. Mutant versions of this construct containing either deletions of the Ϫ10, Ϫ20, and Ϫ30 kb enhancers or a specific mutation altering the two VDREs were also stably integrated and examined for 1,25(OH) 2 D 3 response as well. As can be seen in Fig. 3E, although the recombineered parental BAC together with versions containing deletions at Ϫ20 and Ϫ30 kb were fully capable of mediating response to 1,25(OH) 2 D 3 , neither the version containing the Ϫ10 kb deletion nor that containing mutations of the VDREs was active. These results strongly suggest that the enhancer identified by ChIP-seq analysis Ϫ10 kb upstream of the Mmp13 gene and the VDREs localized within mediate the actions of 1,25(OH) 2 D 3 and that the sequences within the Mmp13 promoter proximal region do not appear to be involved in this regulation. CRISPR Genomic Editing Confirms the Mechanism for Mmp13 Regulation by 1,25(OH) 2 D 3 -The above experiments involving the VDR were made possible because of the ligandinducible nature of the receptor's action at the gene, where -fold change in activity can be analyzed independent of the construct's basal activity. The activity of transcription factors like RUNX2 contributes to the basal expression of genes such as Mmp13, however, and are not ligand-activated per se. Thus, their contribution to the basal activity of stable BAC clones is difficult to interpret given the random integration of individual clones into the genome, the uncertain influence of the environment at these sites of integration, and the potential for the presence of multiple copies. To overcome these pitfalls, the direct genome editing technique of the CRISPR/Cas9 system was employed (33) to target and excise specific enhancer regions of the Mmp13 gene using separate CRISPR guide RNAs or to delete the expression of factors such as the VDR or RUNX2, schematically outlined in Fig. 4A. Interestingly, CRISPR-targeted RNAs that bound directly at the two separate VDREs in the Ϫ10 kb region could be used because the NGG PAM sequence requirement for guide RNAs was fulfilled by the sequence of the canonical VDRE half-site, AGGTCA. We turned to UAMS-PB cells for this purpose because this line was similar to MC3T3-E1 cells with respect to differentiation (24) and because it displayed similar VDR ChIP-seq patterns (Fig.  2B). UAMS-PB cells offer several additional distinct advantages; they represent a clonally isolated cell line, stably express the PTH receptor, and, perhaps most importantly, appear to be more efficient at CRISPR non-homologous end joining recombination, thereby facilitating recovery of homozygous daughter cell lines. CRISPR Ϫ10 kb guide plasmids 1 and 2 were transfected into the UAMS-PB cells, GFP-positive cells were isolated through FACS sorting into 96-well plates, and the genotypes of the isolated clones were screened via PCR and confirmed via sequencing as described under "Experimental Procedures." The frequency of desired homozygous non-homologous end joining enhancer deletions was ϳ1.5-2%, and 2-4 genetically modified clones, as well as at least one intact experimental wild type clone (E/WT), were recovered and tested for expression of Mmp13 and other genes. In Fig. 4B, both the UAMS-PB wild type (WT) and an E/WT clone retained temporal responsiveness to 1,25(OH) 2 D 3 (100 nM) and response to treatment with FGF2 (50 ng/ml) but not BMP2 (50 ng/ml), compounds known to modulate Mmp13 activity (2,8). Mmp13 was strongly inhibited by FGF2 and does not appear on the scale. The Ϫ10 kb enhancer-deleted daughter cell line, on the other hand, was no longer inducible by 1,25(OH) 2 D 3 , as was anticipated, although it retained repressive response to FGF2. These effects were confirmed in a second clone containing a larger Ϫ10 kb enhancer deletion as well as in replicate clones (data not shown). In contrast, however, virtually all additional target genes examined were unaffected by the deletion, including Sp7 and Runx2 as well as others. Surprisingly (and to be explored below), we also found that in the absence of the Ϫ10 kb enhancer, Mmp13 expression was now potently down-regulated in a time-dependent manner by 1,25(OH) 2 D 3 . These data at the endogenous Mmp13 gene locus confirm previous observations that VDRmediated activity is contained within the Ϫ10 kb enhancer while raising the possibility of a secondary independent activity of 1,25(OH) 2 D 3 to suppress Mmp13 expression. Elimination of the Ϫ30 kb Region or RUNX2 Expression Disrupts Basal Mmp13 Transcription-Based upon the initial success with the CRISPR approach, we created an additional series of UAMS-PB daughter cell lines, as outlined in Fig. 5A. First, a region located proximal to yet distinct from the Mmp13 pro-moter (Ϫ98 to Ϫ454 bp, Pro) was deleted in the UAMS-PB cell line. This region contained several highly ranked RUNX2-REs identified through in silico analysis that were also highly conserved in the rat Mmp13 promoter (43,44). Second, a daughter cell line was created that contained two deletions, one at the Ϫ10 kb enhancer (Ϫ10k KO) and the other proximal to the Mmp13 promoter (Pro, Ϫ98 to Ϫ454 bp). Third, we deleted the region comprising the Ϫ30 kb enhancer (Ϫ30k KO). Finally, we also created several additional deletions to eliminate VDR expression (VDR KO), either in the parent cell line or in combination with the Ϫ10 kb deleted daughter cell line (Ϫ10k/ VDR KO) and to eliminate RUNX2 expression (RUNX2 KO). These latter gene knockouts were obtained by targeting exon 3 of both the VDR (Fig. 5B) and the RUNX2 genes (Fig. 5C), which resulted in frameshift mutations within the genes due to small ϳ50-bp deletions that ensured the loss of not only VDR and RUNX2 protein expression but of all putative splice variants as well.

Mmp13
W T led not only to a reduction in the basal expression of Mmp13 but to a surprising 1,25(OH) 2 D 3 -mediated suppression of Mmp13 transcripts relative to the induction seen in the wild type and experimental control cell lines. This suppression was dose-dependent (data not shown). No changes, as seen in Fig.  5E, were again observed in the expression of control RNAs for Vdr, Runx2, Sp7, or Spp1. In contrast, however, genetic deletion of the VDR either from the wild type UAMS-PB parental cell line or from the Ϫ10 kb enhancer daughter cell line (Ϫ10k/ VDR KO), resulted in a complete abrogation of the regulatory effects of 1,25(OH) 2 D 3 at all target genes, including the ability of the hormone to induce Mmp13 in the parental cell line and to suppress the gene in the absence of the Ϫ10 kb enhancer. Surprisingly, the loss of VDR in the wild type UAMS-PB cells resulted in a more pronounced suppression of the basal level of Mmp13 transcripts than that seen in the absence of the Ϫ10 kb enhancer, although it was clear in each case that both the VDR and the enhancer itself contribute either directly or indirectly to the basal expression of Mmp13. It is also worth noting that despite the CRISPR deletion and frameshift mutation of the exon 3 region, small amounts of Vdr mRNA were still detectable in the VDR KO line via RT-PCR analysis, although VDR protein was not produced. The same was true of the Runx2 mRNA in the RUNX2 KO line. Interestingly, deletion of the promoter proximal segment also caused a minor reduction in basal expression of Mmp13, although the gene's inducibility by 1,25(OH) 2 D 3 still remained evident. Curiously, however, although elimination of both the promoter proximal and the Ϫ10 kb region led to a striking loss of basal Mmp13 expression, repression by 1,25(OH) 2 D 3 was still apparent. This result lends credence to the idea that the promoter proximal region contains an important basal regulatory component for the Mmp13 gene. Additional evidence, however, suggests that this region may not be essential for controlling the expression of Mmp13: 1) elimination of the Ϫ30 kb enhancer caused a much more profound reduction in basal Mmp13 activity, as seen in Fig. 5D, and 2) loss of RUNX2 expression was equally debilitating to baseline Mmp13 transcription. Importantly, however, despite a loss in basal activity, the daughter cell lines containing either a deletion of the Ϫ30 kb enhancer or elimination of RUNX2 expression retained Mmp13 inducibility by 1,25(OH) 2 D 3 , demonstrating that the Ϫ10 kb region remained functional despite the change in basal expression of Mmp13 due either to loss of the Ϫ30 kb enhancer or loss of its RUNX2-dependent activity. As expected, the loss of RUNX2 expression also influenced to variable degrees the basal expression of other osteoblastic genes, including Runx2, Vdr, Sp7, and Spp1. It was also notable that the loss of RUNX2 expression has a dramatic impact on Sp7 expression, a dependence that has been described previously (45). These CRISPR data revealed that Mmp13 regulation by RUNX2 and VDR is mediated through the Ϫ30 and Ϫ10 kb regions, respectively. They also show that the effects of 1,25(OH) 2 D 3 on Mmp13 expression are complex. Thus, whereas 1,25(OH) 2 D 3 can induce Mmp13 expression through direct effects mediated by VDR DNA binding, the hormone also appears able to exert secondary effects on Mmp13 gene activity by down-regulating Runx2 and Sp7 expression and, in so doing, reduces basal Mmp13 expression, as summarized in the speculative model documented in Fig. 5F. It is also possible that the VDR may interact directly with RUNX2 itself, as has been shown (46). In either case, CRISPR-mediated enhancer deletion analysis has unveiled what appears to be a secondary effect of 1,25(OH) 2 D 3 on Mmp13.
Loss of RUNX2, Not Mmp13, Adversely Affects Osteoblast Differentiation-To assess the impact that elimination of either RUNX2, VDR, or MMP13 has on the osteoblast differentiation, parental UAMS-PB cells (WT, E/WT), or the Ϫ10k KO, Ϫ30k KO, VDR KO, and RUNX2 KO cell lines were examined for their expression of these factors by Western blot and assessed for their ability to differentiate after 15 days in osteogenic medium. As can be seen in Fig. 6A, the basal level of expression of the Mmp13 protein was generally consistent with the transcript levels that were seen in the cell lines examined in Fig. 5D, with lower MMP13 levels evident in the Ϫ30k KO and RUNX2 KO lines. As seen in Fig. 6B, however, although the parental line exposed to osteogenic medium differentiated fully into mineralized osteoblasts, the presence or absence of MMP13 protein in several of the mutant daughter cell lines, in which RUNX2 expression was unaffected, had no impact on the differentiation process. In contrast, deletion of RUNX2 but not deletion of the Ϫ30 kb enhancer completely prevented osteoblast differentiation and mineralization, as documented in Fig. 6C. As seen in Fig. 6D, loss of RUNX2 resulted in a dramatic reduction in cellular proliferation, which was fully rescued when RUNX2 expression was restored. The results of this experiment also show that re-expression of RUNX2 via lentivirus expression is capable of rescuing the differentiation process and mineralization as well, although this effect was not fully restored despite a rather dramatic overexpression of the RUNX2 protein (Fig. 6E). The expressions of genes, including Sp7, were partially rescued as well (Fig. 6F). We conclude that the dramatic down-regulation of Mmp13 as a result of enhancer deletion had no effect on cellular growth or differentiation, although detailed features of cell function were not examined. RUNX2 deletion, on the other hand, uncovered a dramatic consequence of the protein on osteoblast proliferation, differentiation, and mineralization, effects that are consistent with previous studies of the loss of RUNX2 expression both in vitro and in vivo (16).
Genomic Organization of the Mmp13 Gene Relies upon Ϫ30 kb Enhancer and the Presence of RUNX2-Loss of either the Ϫ30 or Ϫ10 kb enhancer had a profound effect on basal Mmp13 transcription (Fig. 5D), which was probably the consequence of direct genome binding alterations. To examine this possibility, a ChIP-qPCR analysis of the Mmp13 locus at Ϫ30 kb, Ϫ20 kb, Ϫ10 kb, Ϫ2 kb, and the promoter proximal region was performed for the UAMS-PB parental WT and E/WT cells and the CRISPR daughter cell lines Ϫ10k KO, Ϫ30k KO, and RUNX2 KO. Cells were treated with either vehicle or 1,25(OH) 2 D 3 for 3 h and then subjected to ChIP analysis using antibodies to the VDR (blue), RUNX2 (red), C/EBP␤ (green), and IgG (black) (Fig. 7A). qPCR was performed using primers located within the individual peak regions, as defined by ChIP-seq analysis; therefore, regions deleted by CRISPR were not detected (N.D. in Fig. 7). As documented in Fig. 7A, ChIP analysis of RUNX2, VDR, and C/EBP␤ in both the parental line (WT) and the experimental clone (E/WT) confirmed observations made by ChIP-seq analysis, which demonstrated that the highest level of RUNX2 binding was present at the Ϫ30 kb region, C/EBP␤ was predominantly bound to both the Ϫ20 and Ϫ30 kb regions, and 1,25(OH) 2 D 3 -inducible VDR binding was localized to the Ϫ10 kb region. Very modest but measureable levels of VDR, RUNX2, and C/EBP␤ binding were also observed near the promoter, consistent with data obtained by ChIP-seq analysis. Importantly, although the recruitment of VDR and C/EBP␤ at their respective sites across the gene (Ϫ30 kb, Ϫ20 kb, and promoter proximal) was generally unaffected by deletion of the Ϫ10 kb enhancer as compared with the control cell lines, the removal of this enhancer had a modest inhibitory effect on RUNX2 binding at both the Ϫ30 kb and the promoter proximal regions. This effect on RUNX2 binding appears to be exaggerated when the overall expression of the VDR was eliminated. This suggests that genomic organization and transcription factor binding activity across the Mmp13 locus are affected as a result of loss of the Ϫ10 kb segment, but only nominally. In contrast, however, loss of the Ϫ30 kb enhancer had a significant effect on transcription factor occupancy at other sites across the Mmp13 gene locus. Accordingly, C/EBP␤, VDR, and RUNX2 binding at the Ϫ20 kb, the Ϫ10 kb, and the promoter proximal regions, respectively, were all reduced from their residual or inducible levels in the parent cell line. Almost iden-tical effects at these enhancer regions were obtained when RUNX2 expression was eliminated, thus lowering recruitment of VDR and C/EBP␤ to the Ϫ10 kb and Ϫ20 kb enhancer regions. The specificity of these data were confirmed through analysis of the control gene MyoD family inhibitor (Mdfi, I-mfa), which is known to be regulated by 1,25(OH) 2 D 3 in osteoblasts (24,47). At this gene, strong VDR and RUNX2 binding as well as weaker C/EBP␤ binding was observed ϳ8 kb upstream of the Mdfi gene TSS in MC3T3-E1 cells by ChIP-seq analysis (Fig. 7B) and confirmed by ChIP-qPCR in the parental UAMS-PB cells (Fig. 7C). RUNX2 binding was compromised only in the RUNX2 KO line, whereas the VDR and C/EBP␤ binding remained unchanged. These data suggest that RUNX2 is essential for the overall expression of Mmp13 and that its binding to the Ϫ30 kb region is required for the interaction of VDR, C/EBP␤, and RUNX2 at the Ϫ10, Ϫ20, Ϫ30 kb, and promoter proximal regions, which are key for appropriate transcriptional regulation at the Mmp13 gene locus. Similar effects of VDR binding at the Ϫ10 kb region appear to selectively impact RUNX2 but not C/EBP␤ binding at the Ϫ30 kb enhancer and the promoter proximal regions. These potential interactions are illustrated in the model shown in Fig. 7D.

DISCUSSION
Collagenase-3, the product of the Mmp13 gene, is required for proper formation of cartilage and bone but also serves a multitude of additional physiological roles in non-skeletal tissues and pathological roles in metastatic cancer. Accordingly, its expression is regulated by a diverse set of local as well as endocrine stimuli, although the molecular mechanisms that underlie these regulatory events are not well understood and have been limited to traditional studies focused on elements located near the Mmp13 promoter (9 -15). In the present studies using unbiased ChIP-seq analyses, we challenged the idea that regulation of Mmp13 gene expression in bone cells is mediated through promoter proximal elements and show that this gene is regulated primarily through the actions of three distal enhancers located Ϫ10, Ϫ20, and Ϫ30 kb upstream of the Mmp13 promoter that enable the activity of the VDR, C/EBP␤, and RUNX2. Analysis of Mmp13 transcripts revealed promoter-independent roles for the distal Ϫ10 and Ϫ30 kb enhancers in mediating the activities of VDR and RUNX2, respectively, while supporting a minor involvement of the regulatory region proximal to the promoter in basal Mmp13 expression. Nevertheless, both the loss of the Ϫ30 kb enhancer and the elimination of RUNX2 expression had similar devastating effects on basal Mmp13 transcript production, seemingly diminishing the importance of RUNX2 binding at the promoter relative to its interaction at the distal Ϫ30 kb enhancer site.
The studies described made use of CRISPR/Cas9 gene editing to prepare daughter UAMS-PB cell lines in which one or more enhancers were removed from the Mmp13 gene as well as removal of either RUNX2 or VDR. Preparation of these daughter cell lines required the transfection of multiple plasmids, target cell enrichment, and extensive genotype screening using PCR to identify homozygous clones. As indicated, ϳ1.5-2% of the cell lines examined contained homozygous enhancer deletions, although an extremely high percentage of cells contained either a single cut or the expected deletion at a single allele. Based upon this and reported literature, we determined that the low efficiency of obtaining homozygous cell lines was due to the need for two sets of guide RNAs to obtain deletions of the desired size (300 -500 bp), whereas the gene deletions (VDR/ RUNX2 KO at 50 bp) occurred at a higher rate (ϳ10%). Targeting of the deletions was not observed to be perfect either, probably due to fill back by the DNA repair machinery. Indeed, many of the deletions extended several nucleotides beyond the cut site in either direction; however, because the enhancer deletions were arbitrary, based upon the general span covered by the peak obtained through ChIP-seq analysis, this minor imprecision was not problematic. Surprisingly, initial attempts using the MC3T3-E1 cell line were unsuccessful. We speculate that this may to be due to an aberrantly polyploid genome shared by many cultured cell lines, an overactive DNA repair machinery, or simply an increased efficiency of CRISPR-mediated non-homologous end joining recombination in UAMS-PB cells. Finally, we routinely obtained and examined multiple daughter cell lines for each deletion to ensure that the general basal properties of each line were similar to those of the parental UAMS-PB line and that control genes were unaffected by the Mmp13 enhancer deletions. Despite the clonal isolation of the UAMS-PB line, selection and regrowth of single cells frequently altered the properties of some of the daughter cell lines, making it imperative that each line be thoroughly examined for proliferation and differentiation capabilities as well as gene expression profiles.
The results of the enhancer deletion analyses suggest that the three enhancers and the promoter proximal regulatory region are highly interactive and do not function in isolation. For example, deletion of the Ϫ30 kb enhancer and thus loss of RUNX2 regulation exclusively at this site exerted a major negative impact on VDR binding and regulatory activity at the Ϫ10 kb enhancer, loss of C/EBP␤ binding at the Ϫ20 and Ϫ30 kb enhancers, and loss of a small amount of RUNX2 binding near the promoter. However, loss of cellular RUNX2 expression also strongly affected both C/EBP␤ and VDR binding at these same sites. These observations lead to the conclusion that it is not the loss of the Ϫ30 kb enhancer per se but rather the presence of RUNX2 at the site that is key to the three-dimensional interaction. Similarly, deletion of the Ϫ10 kb enhancer and loss of VDR regulation at the site also reduced RUNX2 binding at the Ϫ30 kb enhancer and at the promoter proximal region. Although this effect on RUNX2 binding was modest, it was exaggerated in cells in which VDR expression had been eliminated. In contrast, C/EBP␤ binding at its cognate sites was unaffected by the either the loss of the Ϫ10 kb enhancer or the loss of VDR expression. Nevertheless, similar conclusions can be drawn as well regarding the role of VDR in facilitating enhancer interaction. It is possible that small amounts of factor binding seen at Ϫ30 kb, Ϫ20 kb, and the promoter proximal region by ChIP-seq analysis, as exemplified by the VDR, may arise as a result of the cross-linking of proteins from complex formation rather than direct DNA binding to each enhancer. This idea, in the case of the VDR, appears to be reinforced by the fact that aside from the Ϫ10 kb segment, none of the remaining regions were capable of mediating 1,25(OH) 2 D 3 activity when examined in isolation, although alternative explanations are also possible. Overall, however, these studies support the concept of a striking looping model as seen in Fig. 7D for the Mmp13 gene, wherein protein complexes at each of the distal enhancers are brought into proximity of the Mmp13 promoter region via protein-protein interaction.
Observations at the Mmp13 gene also provide additional support for the idea that changes in the binding and activity of transcription factors, such as RUNX2 and C/EBP␤, and the modifications to the epigenome that occur during differentiation leading to an altered transcriptome may cause dynamic transitions in both transcription factor binding and response to secondary factors, such as 1,25(OH) 2 D 3 . This change is seen at many genes but is exemplified by the enhanced response of the Mmp13 gene to 1,25(OH) 2 D 3 following differentiation as seen in Fig. 1. The observation that loss of a dominant regulatory enhancer (Ϫ10 kb) for the VDR unveils a secondary effect of 1,25(OH) 2 D 3 to suppress Mmp13 expression is also informative because it highlights an underlying mechanism through which 1,25(OH) 2 D 3 can induce expression of a gene under one condition and suppress the same gene under another, as was seen for Bmp2 regulation (24). We speculate that this effect may be due to the ability of the VDR to suppress RUNX2 expression (Figs. 4 and 5), although a direct protein-protein interaction between RUNX2 and the VDR is also possible (46). In fact, the results in our control gene Mdfi suggest that VDR is required for RUNX2 action at that gene but perhaps not vice versa. This provides one mechanism for gene suppression by 1,25(OH) 2 D 3 for which little is known. Interestingly, this mechanism in association with RUNX2 activity is not limited to secondary suppression because RUNX2 can be a suppressor of the expression of many genes (48). Accordingly, the down-regulation of RUNX2 expression or activity at a particular gene could result in increased responsiveness to 1,25(OH) 2 D 3 . Further studies of these interactions at genes regulated by RUNX2 will be illuminating with regard to this mechanism.
Our previous studies using ChIP-seq analysis defined an osteoblast enhancer complex composed of a single enhancer to which RUNX2, C/EBP␤, and the VDR bind in an organized fashion to control the expression of genes integral to the osteoblast phenotype (24). The present studies support a more complex version of the osteoblast enhancer complex in which linearly separated enhancers for a single target gene can bind these same three transcription factors independently yet interact to form a complex at the Mmp13 promoter that is capable of integrating the activities of the signaling pathways that regulate these factors. This finding has considerable implications for the mechanisms through which a gene can be regulated but also points to the complexity of identifying activity that is inherent to a particular enhancer when deletion of that enhancer affects indirectly the activity of another. Unfortunately, the contribution of C/EBP␤ to the osteoblast enhancer complex or Mmp13 expression is even less well understood. However, C/EBP␤ is known to play a role in osteoblast function in vivo, has been shown to interact directly with RUNX2, and may serve as a structural scaffold for basal activity of Mmp13 expression (49 -51). Because the elimination of RUNX2 and the deletion of the Ϫ30 kb enhancer yielded the same results despite a large amount of C/EBP␤ binding, it is difficult to determine what contribution C/EBP␤ has, much less the contribution by the Ϫ20 kb enhancer, which may be completely structural for enhancer looping.
Previous studies have suggested that Mmp13 expression is regulated by PTH via the binding of RUNX2 to a site(s) near the Mmp13 promoter (9). Further studies have indicated that 1,25(OH) 2 D 3 also regulates Mmp13 expression via elements similarly located (23) as well as by the estrogen receptor (10) at the promoter proximal region. The focus of virtually all of these studies was limited to the Mmp13 promoter, however, utilizing traditional approaches that included transient transfection analysis and mutagenesis to explore the cis-acting components within this regulatory region and more recently direct PCR ChIP analysis to detect the presence of specific proteins. Our studies using unbiased alternative approaches indicate that more upstream components are certainly involved in the regulation of Mmp13 expression in osteoblastic cells. These approaches involved the identification of transcription factor binding sites, the dissection of fundamental regulatory regions located within the endogenous gene via CRISPR deletion, and the analysis of transcriptional activity through direct measurement of the output of authentic Mmp13 transcripts from the endogenous gene. They also involved correlative deletion of both RUNX2 and VDR expression. These studies suggest that some activity of RUNX2 may be present at the promoter proximal region, although it is clear that this activity (either with or without our deletion studies) is neither exclusive to the regulation nor dominant when compared with the activity of RUNX2 at sites distal to the Mmp13 promoter. Similarly, we speculate that our understanding of human MMP13 gene regulation is probably incomplete as well. There have been several studies focused on the gene's promoter proximal region for induction, repression, or inhibition (52,53). Furthermore, upstream segments of both mouse and human MMP13 share segments of homologous sequence (data not shown). However, the roles of the conserved regions are likely to remain uncertain until we are able to examine the human MMP13 gene in detail via ChIPseq and CRISPR in relevant cell lines. These findings highlight the need for re-examination of genes as exemplified through Mmp13 to ensure that the understanding of their mechanisms of regulation is contemporary.
Taken together, the regulation of Mmp13 expression appears to be mediated via a complex arrangement of enhancers that interact with VDR, RUNX2, and C/EBP␤. Most strikingly, the 1,25(OH) 2 D 3 -mediated regulation is present at the Ϫ10 kb enhancer, whereas most basal activity is attributed to the Ϫ30 kb enhancer. This mechanism of regulation appears to be present in mouse cells of mesenchymal origin, and we speculate that subsets of these regulatory regions may participate in the control of Mmp13 expression in other tissues, including metastatic cancer cells, where both RUNX2 and other isoforms of RUNX are expressed. The mechanisms of distal regulation studied here open new avenues for investigation toward a more complete understanding of Mmp13 regulation.