Novel Regulators of Fgf23 Expression and Mineralization in Hyp Bone

Abstract

We used gene array analysis of cortical bone to identify Phex-dependent gene transcripts associated with abnormal Fgf23 production and mineralization in Hyp mice. We found evidence that elevation of Fgf23 expression in osteocytes is associated with increments in Fgf1, Fgf7, and Egr2 and decrements in Sost, an inhibitor in the Wnt-signaling pathway, were observed in Hyp bone. β-Catenin levels were increased in Hyp cortical bone, and TOPflash luciferase reporter assay showed increased transcriptional activity in Hyp-derived osteoblasts, consistent with Wnt activation. Moreover, activation of Fgf and Wnt-signaling stimulated Fgf23 promoter activity in osteoblasts. We also observed reductions in Bmp1, a metalloproteinase that metabolizes the extracellular matrix protein Dmp1. Alterations were also found in enzymes regulating the posttranslational processing and stability of Fgf23, including decrements in the glycosyltransferase Galnt3 and the proprotein convertase Pcsk5. In addition, we found that the Pcsk5 and the glycosyltransferase Galnt3 were decreased in Hyp bone, suggesting that reduced posttranslational processing of FGF23 may also contribute to increased Fgf23 levels in Hyp mice. With regard to mineralization, we identified additional candidates to explain the intrinsic mineralization defect in Hyp osteoblasts, including increases in the mineralization inhibitors Mgp and Thbs4, as well as increases in local pH-altering factors, carbonic anhydrase 12 (Car12) and 3 (Car3) and the sodium-dependent citrate transporter (Slc13a5). These studies demonstrate the complexity of gene expression alterations in bone that accompanies inactivating Phex mutations and identify novel pathways that may coordinate Fgf23 expression and mineralization of extracellular matrix in Hyp bone.

Gene expression array analysis of Hyp bone identifies novel Phex-dependent pathways associated with elevated Fgf23 expression, altered proteolysis and abnormal matrix mineralization.

X-linked hypophosphatemic rickets is a hereditary disorder characterized by renal phosphate wasting, aberrant vitamin D metabolism, and rickets/osteomalacia that is due to inactivating mutations of PHEX, or phosphate-regulating gene with homologies to endopeptidases on the X chromosome, a cell surface endopeptidase predominately located in osteoblasts and osteocytes (1,2). Inactivating mutations of PHEX lead to increased production of fibroblast growth factor (FGF)23, principally by osteocytes, as well as rickets and osteomalacia due, in part, to an intrinsic defective in the mineralization of extracellular matrix (3). FGF23 principally functions as a phosphaturic factor (4,5,6) and counterregulatory hormone for 1,25-(OH)₂D production (3) via binding to Klotho/FGF receptor (FGFR) complexes in the kidney (7,8).

The study of the Hyp mouse homolog of X-linked hypophosphatemic rickets (6) and mice with conditional deletion of the Phex gene in the osteoblast lineage (9) have identified several bone abnormalities associated with inactivation of Phex. First, inactivation of Phex in osteocytes leads to increased expression of Fgf23 in these cells due to an unknown matrix-derived Fgf23-stimulatory factor, which appears to be released or activated by osteocytes when Phex is inhibited or mutated (6,10). Evidence for this intrinsic factor includes the finding that explantation of bone from Hyp mice into wild-type (WT) mice fails to correct the high expression of Fgf23 in osteocytes (10). In addition, the absence of Phex is necessary but not sufficient to stimulate FGF23, because a functional Phex is missing in osteoblasts of Hyp mice, but osteoblasts do not up-regulate Fgf23 expression until they differentiate into osteocytes embedded in bone matrix, which likely contains Fgf23-stimulatory factors (6,10).

A second abnormality in Hyp bone is the phosphate-independent, intrinsic mineralization defect. This appears to be due to the secretion by Phex mutant osteoblasts of a putative factor(s) that inhibits mineralization of extracellular matrix, referred to as Minhibin (10,11). Screening of substrate phage libraries identified the ability of Phex to cleave small peptides, such as the ASARM (acidic serine aspartate-rich MEPE-associated motif) peptide derived from matrix extracellular phosphoglycoprotein (MEPE) (12), which can inhibit bone mineralization, but other physiologically relevant mineralization inhibitors are likely to be present, because deletion of Mepe fails to rescue the Hyp phenotype (13). In addition, mutations or ablation of Dmp1, an extracellular matrix protein that promotes mineralization, leads to increased osteocyte expression of FGF23 and rickets/osteomalacia (14,15). Regardless, the specific intermediate steps and potential interrelationships linking Phex and Dmp1 to regulation of FGF23 and mineralization of extracellular matrix remain poorly defined.

Inactivation of the Phex gene is also associated with increases in proteolytic activity of bone (16). Recent studies indicate that inhibition of proteolytic activity in Hyp bone by CA074 and pepstatin improves the mineralization defect without correcting the hypophosphatemia (17). There are important gaps in our understanding of how deficiency in Phex regulates FGF23 expression in osteocytes, impairs bone mineralization, and activates proteolytic pathways.

Microarray technology has been largely limited to the identification of differentially expressed genes in cultured osteoblasts (18,19,20,21); however, genome-wide expression analysis of intact whole-bone samples has recently been successfully employed to identify novel pathways in bone development in WT and mutant Runx2 mouse models (22). The success of this analysis may be due to the osteoblast-lineage-specific role of Runx2 in bone tissues, a feature shared with Phex, which is mainly localized to osteocytes in cortical bone. Finally, cortical bone analysis may have greater predictive value for assessing gene expression profiles, because isolation of osteocytes is technically challenging, confounded by cell heterogeneity, and subject to alterations in the ex vivo experimental conditions.

In the current study, to gain insights into how inactivating Phex mutations result in increased Fgf23 expression in osteocytes as well as phosphate-independent defects in bone mineralization, we have used microarray expression profiling to identify genes that are specifically expressed in osteocyte-rich cortical bone from WT and Hyp mice. We have used these data to generate new hypotheses regarding the regulation of Fgf23, bone mineralization, and protease activity in Hyp bone.

Results

Microarray analysis

To determine the genes in cortical bone the expression of which is modulated by inactivation of Phex in bone on a genome-wide basis, we performed microarray expression profiling on cortical bone derived from Hyp mice and WT littermates as described in Materials and Methods. Based on a P value cut-off of 0.05 and a 2-fold change of gene expression, we found that 118 transcripts out of the total of about 34,000 present on the array were significantly modulated in Phex mutant Hyp mice, with 41 showing higher expression (up-regulated transcripts) (Table 1) and 77 genes showing lower expression (down-regulated transcripts) (Table 2). Of these 77 down-regulated transcripts, 20 were decreased by more than 2.5-fold (Table 2), and an additional 57 genes were reduced between 2- and 2.5-fold in Hyp bone (supplemental Table 1 published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org).

Table 1.

Up-regulated genes (>2-fold) in Hyp cortical bone

Probe set identification	Fold-change	UniGene	Gene symbol	Gene title
1422176_at	48.8	Mm.347933	Fgf23a	Fibroblast growth factor 23
1449388_at	7.7	Mm.20865	Thbs4a	Thrombospondin 4
1422253_at	5.9	Mm.443177	Col10a1	Collagen
1416579_a_at	4.1	Mm.4259	Tacstd1	Tumor-associated calcium signal transducer 1
1418509_at	4.1	Mm.21454	Cbr2a	Carbonyl reductase 2
1420855_at	3.7	Mm.275320	Eln	Elastin
1449340_at	3.3	Mm.43375	Sostdc1a	Sclerostin domain containing 1
1422632_at	3.2	Mm.113590	Ctsw	Cathepsin W
1460187_at	3.2	Mm.281691	Sfrp1a	Secreted frizzled-related protein 1
1439947_at	3.1	Mm.302865	Cyp11a1	Cytochrome P450
1419483_at	3.1	Mm.2408	C3ar1	Complement component 3a receptor 1
1422542_at	3.1	Mm.391232	Gpr34a	G protein-coupled receptor 34
1448891_at	3.0	Mm.45173	Msr2a	Macrophage scavenger receptor 2
1435162_at	3.0	Mm.263002	Prkg2a	Protein kinase
1449456_a_at	2.9	Mm.1252	Cma1	Chymase 1
1421262_at	2.9	Mm.299647	Lipg	Lipase
1425951_a_at	2.8	Mm.271782	Clec4na	C-type lectin domain family 4
1451031_at	2.7	Mm.42095	Sfrp4a	Secreted frizzled-related protein 4
1419015_at	2.5	Mm.13828	Wisp2a	WNT1-inducible signaling pathway protein 2
1427683_at	2.4	Mm.290421	Egr2a	Early growth response 2
1448730_at	2.4	Mm.1135	Cpa3a	Carboxypeptidase A3
1426063_a_at	2.4	Mm.247486	Gem	GTP-binding protein (gene overexpressed in skeletal muscle)
1421227_at	2.4	Mm.14424	Gzme	Granzyme E
1420249_s_at	2.3	Mm.137	Ccl6a	Chemokine (C-C motif) ligand 6
1419561_at	2.2	Mm.1282	Ccl3	Chemokine (C-C motif) ligand 3
1450869_at	2.2	Mm.241282	Fgf1a	Fibroblast growth factor 1
1448416_at	2.2	Mm.243085	Mgpa	Matrix Gla protein
1456736_x_at	2.2	Mm.440194	5230400G24Rika	RIKEN cDNA 5230400G24 gene
1425681_a_at	2.2	Mm.180750	Prnd	Prion protein dublet
1450020_at	2.1	Mm.44065	Cx3cr1	Chemokine (C-X3-C) receptor 1
1417408_at	2.1	Mm.273188	F3	Coagulation factor III
1448929_at	2.1	Mm.235105	F13a1a	Coagulation factor XIII
1450199_a_at	2.1	Mm.220821	Stab1a	Stabilin 1
1420338_at	2.1	Mm.4584	Alox15a	Arachidonate 15-lipoxygenase
1426663_s_at	2.0	Mm.200307	Slc45a3a	Solute carrier family 45
1438405_at	2.1	Mm.330557	Fgf7	Fibroblast growth factor 7
1424754_at	2.1	Mm.193094	Ms4a7	Membrane-spanning 4-domains
1434342_at	2.1	Mm.235998	S100ba	S100 protein
1425214_at	2.1	Mm.235193	P2ry6a	Pyrimidinergic receptor P2Y
1449401_at	2.1	Mm.439732	C1qca	Complement component 1
1418440_at	2.1	Mm.130388	Col8a1a	Collagen

^aGenes known to be expressed in cultured osteoblasts and/or osteocytes. 

Table 2.

Down-regulated genes (>2.5-fold) in Hyp cortical bone

Probe set identification	Fold-change	UniGene	Gene symbol	Gene title
1421979_at	317.0	Mm.2529	Phexa	Phosphate regulating gene with homologies to endopeptidases on the X chromosome
1421245_at	3.8	Mm.265602	Sosta	Sclerostin
1428485_at	3.8	Mm.277921	Car12	Carbonic anhydrase 12
1417262_at	3.3	Mm.292547	Ptgs2a	PG-endoperoxide synthase 2
1434628_a_at	3.2	Mm.286600	Rhpn2a	Rhophilin
1418910_at	2.9	Mm.595	Bmp7a	Bone morphogenetic protein 7
1451236_at	2.9	Mm.46233	Rerga	RAS-like
1431362_a_at	2.8	Mm.30162	Smoc2a	SPARC-related modular calcium binding 2
1418723_at	2.8	Mm.155520	Edg7a	LPA receptor
1429459_at	2.8	Mm.89313	Sema3d	Sema domain
1424528_at	2.8	Mm.45127	Cgref1a	Cell growth regulator with EF hand domain 1
1443412_s_at	2.8	Mm.204820	Mmp16a	Matrix metallopeptidase 16
1450770_at	2.7	Mm.318710	3632451O06Rik	Similar to Leishmaniaproteophosphoglycan
1429280_at	2.6	Mm.322500	Col22a1a	Collagen
1434049_at	2.6	Mm.76648	Entpd3	Ectonucleoside triphosphate diphosphohydrolase 3
1459729_at	2.6	Mm.340778	Slc13a5	Solute carrier family 13 (sodium-dependent citrate transporter)
1426251_at	2.6	Mm.108557	Cpza	Carboxypeptidase Z
1453588_at	2.6	Mm.300	Car3a	Carbonic anhydrase 3
1451406_a_at	2.5	Mm.3401	Pcsk5a	Proprotein convertase subtilisin/kexin type 5
1460604_at	2.5	Mm.45435	Cybrd1	Cytochrome b reductase 1

^aGenes known to be expressed in cultured osteoblasts and/or osteocytes. 

Although cortical bone devoid of bone marrow and growth plates consists predominately of cells in the osteoblast lineage, to gain insights into alterations due to changes in gene expression within the osteoblast/osteocyte lineage, we identified differentially regulated genes in Hyp bone known to be expressed in osteoblasts/osteocytes. This was accomplished by cross referencing our microarray data sets with the Gene Expression omnibus (GEO) profiles of osteoblasts and by searching public databases for genes reported to be expressed in the osteoblasts lineage. Of the 41 mRNAs up-regulated in Hyp bone, at least 25 are known to be expressed in the osteoblast/osteocyte lineage. Similarly, of the 20 genes down-regulated, greater than 2.5-fold, 15 are reported to be expressed in the osteoblast lineage.

The relevance of this microarray analysis of cortical bone was further substantiated by confirming the principal gene expression abnormalities in Hyp-derived osteoblasts, which are known to have increased Fgf23 expression and absence of Phex function (6). Indeed, Fgf23 was the gene most markedly up-regulated in Hyp bone (Table 1), whereas Phex was below the detection limits (Table 2). In addition, hierarchical cluster analysis of the complete set of transcripts expressed in from the microarray data found that the gene expression patterns from Hyp bone were distinctively different from WT bone (data not shown), which indicates that Hyp bone has a different genetic profile. A clustering of genes that were up-regulated or down-regulated by more than 2-fold (Fig. 1) also demonstrated a clear separation between Hyp and WT bone.

Figure 1.

Hierarchical clustering of WT and Hyp cortical bone. Each column represents expression on one specific array; each row indicates expression level of a gene. Blue represents low expression; red represents high expression. Only genes that were up-regulated or down-regulated more than 2-fold were extracted for clustering. The dendrogram shows that the clustering (euclidean distances) separates controls (WT) from Hyp bones (Hyp).

Gene Ontology (GO) analysis

GO analysis of biological processes, molecular function, and cellular components was carried out on the 588 Phex-modulated transcripts (fold change ≥1.5; P < 0.05) in Hyp bone without discriminating between up-regulation and down-regulation. Unique genes (352) had an annotation within the main category of biological processes. Of these, 163 unique genes contributed to significant enrichment of one or more GO-terms, with at least 46% of the genes regulated by Phex belonging to specifically enriched GO-terms (P < 0.01). GO-terms that are specifically enriched in Hyp cortical bone include negative regulation of bone morphogenetic protein (BMP) signaling pathway, regulation of bone mineralization and ossification processes, cell adhesion, and phosphate transport (Fig. 2A). Other terms include hyaluronic acid, heparin, l-ascorbic acid, calcium ion, IGF binding, and extracellular matrix structural constituent in molecular functions (Fig. 2B).

Figure 2.

GO analysis of gene expression alterations in Hyp cortical bone. Microarray data generated from WT and Hyp cortical bone from 12-d-old mice were analyzed based on the GO as described in Materials and Methods. A, Biological processes. B, Molecular functions. Genes with 1.5-fold or more changes in gene expressions between WT and Hyp mice were used in the analysis. The statistical significance of overrepresented GO terms is highlighted through a color scale ranging from yellow (P = 0.05) to dark orange (P < 5 × 10⁻⁷).

GenMAPP analysis

Next, we used GenMAPP analysis to identify signaling pathways, metabolic pathways, and other functional groups altered in Hyp bone. We found that growth factor activity was significantly modified by mutant Phex, with the following transcripts (fold change) affected: Fgf23 (48.8), Bmp1 (0.6), Bmp6 (1.6), Bmp7 (0.4), Csf1 (1.7), Fgf1 (2.0) Fgf13 (0.7), Fgf7 (2.1), and Pdgfd (0.6) (supplemental Fig. S1A). Metallopeptidase activity was also dramatically affected in Hyp bone, including Phex (undetectable), Cpa3 (2.4), Dpep1 (1.6), Enpep (1.5), Naalad2 (1.7), Adam 12 (0.7), Mmp16 (0.4), Mmp23 (0.6), Bmp1 (0.6), and NM_153107 (Cpz, 0.4) (supplemental Fig. S1B). In addition, prostaglandin (PG) and leukotriene metabolism was altered in Hyp bone, including Pla2g4a (0.7), Alox15 (2.1), Ptgs2 (0.3), and Cbr2 (4.1) (supplemental Fig. S1C). Other signaling pathways altered by mutant Phex include cell-cell signaling (supplemental Fig. S1D), TGF-β signaling (supplemental Fig. S1E), and Wnt signaling (supplemental Fig. S1F).

Real-time RT-PCR confirmation of microarray gene expression

Using real-time PCR we confirmed the fold change in a selected group of genes detected by microarray in RNA isolated from cortical bone of Hyp mice (Table 3). The fold change in Sfrp1, Sfrp4, Wisp2, Egr2, FGF1, MGP, and FGF7 were all increased in both the microarray and real-time RT-PCR analysis by a similar magnitude. Whereas both Fgf23 and Sostdc1 were also increased by both analyses, the fold changes for the expression in microarray vs. real-time RT-PCR were 48.8 vs. 105.9 and 3.3 vs. 12 in FGF23 and Sostdc1, respectively. The fold decrease of Sost, Car12, Ptgs2, Rhpn2, Bmp7, and Pcsk5 were similar by both microarray analysis and real-time RT-PCR.

Table 3.

Fold changes of selected genes in cortical bone and TMOb between WT and Hyp

		Real-time RT-PCR
Gene symbol	Microarray cortical bone (n = 4)	Cortical bone (n = 4)	TMOb (n = 3)
Fgf23	48.8	105.9 ± 11.4a	22.1 ± 9.6a
Thbs4	7.7	7.5 ± 1.6a	8.9 ± 1.0a
Sostdc1	3.3	12.0 ± 0.8a	−7.1 ± 0.1a
Sfrp1	3.2	2.7 ± 0.2a	−1.6 ± 0.1a
Gpr34	3.1	2.9 ± 0.3a	−2.1 ± 0.2a
Sfrp4	2.7	3.3 ± 0.4a	−1.4 ± 0.3
Wisp2	2.5	2.8 ± 0.2a	ND
Egr2	2.4	2.9 ± 0.3a	−1.4 ± 0.2
Fgf1	2.2	2.6 ± 0.5a	−1.9 ± 0.2a
Mgp	2.2	2.3 ± 0.2a	1.9 ± 0.1a
Fgf7	2.1	4.0 ± 0.3a	−1.4 ± 0.1
Sost	−3.8	−4.1 ± 0.2a	−16.0 ± 6.3a
Car12	−3.8	−4.7 ± 0.1a	−1.4 ± 0.2
Ptgs2	−3.3	−4.0 ± 0.1a	−1.8 ± 0.2a
Rhpn2	−3.2	−2.2 ± 0.1a	−1.4 ± 0.1a
Bmp7	−3.0	−3.3 ± 0.1a	1.5 ± 0.2
Pcsk5	−3.8	−2.9 ± 0.1a	−1.6 ± 0.1a

Values shown in the table are mean fold change from microarray data or mean ± sem of the fold changes of gene expression assessed by real-time PCR in cortical bone and TMOb cells from Hyp compared with WT mice. Student’s t test was used for statistic analysis. ND, Not determined. 

^aSignificantly different compared with WT (P < 0.05). 

The relevance of this microarray analysis of cortical bone was further investigated by confirming the principal gene expression abnormalities in Hyp-derived osteoblasts (6). Transgenic mice (TMOBs) Hyp osteoblasts recapitulate the osteoblast developmental stage-dependent increase in Fgf23 and abnormal mineralization observed in Hyp bone in vivo (11). Indeed, Fgf23 was the gene most markedly up-regulated in Hyp bone [22-fold in Hyp compared with WT osteoblasts (Table 1)] whereas Phex was below the detection limits (Table 2) (11). However, of the remainder of the genes differentially expressed in bone in vivo, only Thbs4 and Mgp were increased in TMOB-Hyp compared with TMOB-WT osteoblasts. Thus, Phex-mutant osteoblasts do not possess all of the features observed in cortical bone from Hyp mice (Table 3). There are multiple potential explanations for the differences between cell culture and intact bone, including loss of some of the features of the in vivo Hyp phenotype ex vivo or to detection of environmental effects of the Hyp milieu (such as hypophosphatemia) or confounding effects of greater cell heterogeneity in the in vivo assessment of cortical bone.

We have previously demonstrated that explantation of Hyp bone into WT mice results in persistent increments in FGF23 expression in the Phex-mutant-transplated bone (10). To gain further insights into the changes in gene expression that are intrinsic to the Phex mutation in Hyp bone, we analyzed gene expression profiles on Hyp and WT bones after transplantation into WT mice. Of the up-regulated genes, only Thbs4 (2.9-fold), Sostdc1 (9.7-fold), Gpr34 (2.1-fold), Egr2 (6.2-fold), and FGF1 (4.7-fold) were increased in Hyp bone explanted into WT mice. Of the down-regulated genes, none had significant reductions in the Hyp bone explanted into WT mice.

Analysis of signal transduction pathways and FGF23 promoter activity in vitro

Next, we assessed the effects of various signal pathways identified by microarray analysis to regulate FGF23 promoter activity in ROS17/2.8 osteoblasts transfected with an Fgf23 promoter-luciferase reporter construct. Compared with vehicle treatments, both the addition of FGF1 and FGF2 resulted in a dose-dependent stimulation of luciferase activity in ROS17/2.8 cells transfected with p7980FGF23-luc (Fig. 3, A and B). The stimulation of Fgf23 promoter by FGF2 was completely blocked by the FGFR inhibitor PD173074 (23) at a concentration of 50 nm (Fig. 3C). We also found that LiCl, which activates canonical Wnt signaling by inhibiting glycogen synthase kinase-3β, also stimulated FGF23 promoter activity in a dose-dependent manner (Fig. 3D). In contrast, we also found that TGF-β1 inhibited FGF23 promoter activity in ROS 17/2.8 osteoblasts in a dose-dependent manner (Fig. 3E).

Figure 3.

Factors stimulating Fgf23 promoter activity. Fgf23 promoter activity was assessed in ROS17/2.8 osteoblasts transiently transfected with p7980Fgf23-luc and the pRL-TK construct (internal control for transfection efficiency) as described in Materials and Methods. A and B, Dose-dependent effects of FGF1 (A) and FGF2 (B) on Fgf23 promoter activity. The control represents vehicle containing 10 μm heparin alone. C, Effect of FGFR inhibitor PD173074 (0–100 nm) on FGF2-stimulated Fgf23 promoter activity. PD173074 resulted in a dose-dependent inhibition of FGF2 stimulation of Fgf23 promoter activity. D, LiCl stimulates Fgf23 promoter activity. E, TGFβ-1 suppresses Fgf23 promoter activity. F, Effects of cotransfection of Phex and Dmp1 on Fgf23 promoter activity. Data represent relative luciferase activity expressed as the mean ± sem of triplicates. Each experiment was repeated at least three times. G, 1,25-(OH)₂ vitamin D (calcitriol) stimulates Fgf23 promoter activity used as positive control. *, Values that are significantly different from vehicle (P < 0.05).

To confirm the Wnt signaling change in Hyp bone, we isolated nuclear extract from long bones of WT and Hyp mice at 4 wk of age and examined total β-catenin and phosphorylated β-catenin by Western blot analysis. We observed that total β-catenin in the nuclear extract was increased. In contrast, phosphorylated β-catenin was not altered (Fig. 4A). We also transfected TOPflash (TCL-reporter construct) into osteoblasts derived from calvariae of WT and Hyp mice and found that the relative luciferase activity was about 4.5-fold increased in TmOb-Hyp compared with TmOb-WT osteoblasts. Collectively, these findings support increased Wnt/ β-catenin signaling in Hyp osteoblasts.

Figure 4.

Wnt/β-catenin signaling pathway in long bone and osteoblasts from WT and Hyp mice. A, β-Catenin expression in Hyp bone. Nuclear extract from long bone of WT and Hyp mice at 4 wk of age was immunoblotted with β-catenin and phospho-β-catenin antibodies. β-Actin was used as an internal control for equal protein loading. B, Increased TOPflash promoter activity in Hyp-derived osteoblasts. TMOb-WT and TMOb-Hyp osteoblasts were cotransfected with 1.5 μg of TOPflash and 0.015 μg of pRL-TK plasmids. The cells were lysed 24 h after transfection, and the ratio of firefly luciferase to Renilla luciferase signal was quantified. Fold changes of luciferase activity relative to TMOb-WT are shown. Values represent fold changes expressed as the mean ± sem of triplicates. *, P < 0.05 (t test) compared with TMOb-WT.

Loss of function Dmp1 mutations also cause increased expression of Fgf23 in osteocytes through unknown mechanisms (14,15). However, Dmp1 expression in Hyp bone was not decreased in either microarray or real-time PCR (Dmp1 expression was increased 1.2-fold by microarray analysis and 1.4-fold by real-time PCR). To evaluate whether Phex and Dmp1 directly regulate FGF23 promoter activity in osteoblasts, we cotransfected Phex and Dmp1 expression constructs with p7980FGF23-luc in ROS17/2.8 cells and assessed their effects on FGF23 promoter activity. We failed to observe any effects of Phex, Dmp1, or combined Phex and Dmp1 overexpression on FGF23 promoter activity (Fig. 3F). In addition, we tested the effects of indomethacin (1–10 μm), PGD2 (1–50 ng/ml), PGE2 (1–50 ng/ml), PGF2a (1–50 ng/ml), phosphate (1–10 mm), pyrophosphate (10–500 μm), IGF-I (1–50 ng/ml), IGF-II (1–50 ng/ml), BMP2 (1–50 ng/ml), BMP7 (1–50 ng/ml), calcitonin (1–500 ng/ml), FGF7 (1–50 ng/ml), and FGF23 (1–50 ng/ml) on FGF23 promoter activity, and they had no effects on FGF23 promoter activity in ROS17/2.8 cells (supplemental Fig. S2). As a positive control for these studies, we confirmed that calcitriol stimulated FGF23 promoter-reporter activity as previously reported (3) (supplemental Fig. S2).

Discussion

A major challenge in understanding the pathogenesis of the Hyp phenotype is the identification of pathways linking inactivating Phex mutations with intrinsic bone abnormalities leading to increased Fgf23 gene expression in osteocytes and defective mineralization of bone. From the microarray analysis of genes up-regulated or down-regulated in Hyp cortical bone (Tables 1 and 2, and supplemental Table 2), we identified several biological processes and pathways known to be abnormal in Hyp mouse bone. In this regard, we confirmed up-regulation of Fgf23 and undetectable Phex transcripts in cortical bone from Hyp mice, indicating that our microarray analysis reflects known abnormalities in gene expression in Phex-deficient osteoblasts/osteocytes in vivo and in vitro. In addition, GO analysis identified biological processes and molecular functions consistent with the abnormalities in mineralization, growth factor signaling, and extracellular matrix metabolism in Hyp bone (Fig. 2). Indeed, genome-wide expression analysis of intact whole-bone samples of Hyp mice identified novel pathways associated with the principal abnormalities in Hyp, including increased Fgf23 expression, increased proteolytic processing, and impaired mineralization of extracellular matrix.

With regard to Fgf23 regulation, we identified FGF receptor and Wnt/β-catenin activation as potentially new molecular pathways stimulated by inactivation of Phex. Microarray analysis showed that Fgf1 (2.2-fold), Fgf7 (2.1-fold), and Egr2 (2.4-fold), a down stream signal for FGF-dependent pathways (8), were elevated in Hyp bone. Moreover, we found that both FGF1 and FGF2 stimulate FGF23 promoter activity in Ros17.2.8 osteoblasts, and this effect is inhibited by an FGFR antagonist (Fig. 3). Recent studies also have shown that αvβ3 integrins binding to FGF1 leads to FGFR activation (24). Wnt1-inducible signaling pathway protein 2 (Wisp2), also known as CCN5, which links integrins to extracellular matrix-associated proteins (25), was increased in Hyp bone. Fibroblastic growth factor 7 (Fgf7) also has been reported to have phosphaturic effects in vivo (26,27,28), and activating mutations of FGFR1 are associated with increased FGF23 levels (29). In addition, excess FGF2 is associated with impaired osteoblast-mediated mineralization in vitro (30) and in vivo (31). These associations are consistent with the hypothesis that inactivating Phex mutations leads to alterations in FGFs in the extracellular matrix, which regulate both FGF23 expression in osteocytes and osteoblast-mediated mineralization, thereby providing coordination between systemic phosphate homeostasis and the local mineralization process (24,29,32). The mechanisms whereby Phex mutations increase FGFR signaling and whether this pathway is an important regulator of Fgf23 expression in vivo will require additional studies.

We also found evidence for increased Wnt signaling pathways in Hyp bone. Hyp bone displayed a significant decrease in the Wnt antagonist Sclerostin (Sost), a secreted protein that inhibits Wnt signaling by binding to Wnt coreceptors, low-density lipoprotein receptor-related protein 5 and 6 (38). Sost is coexpressed with Phex and Fgf23 in osteocytes, which adds support for the possibility that osteocytes are central to the pathogenesis of Hyp phenotype. A decrease in Sost would be expected to enhance Wnt signaling, leading to increased bone mass, which is consistent with the observation of severe osteosclerosis in the metaphyseal region of Hyp bone after the hypophosphatemia has been corrected (10). In addition, we confirmed increased Wnt signaling in Hyp as evidenced by increased β-catenin expression in Hyp bone and increased TOPflash luciferase activity in Hyp-derived osteoblasts (Fig. 4). Additional studies found that activation of the Wnt-pathway with LiCl stimulated Fgf23 promoter activity in cultured osteoblasts (Fig. 3). However, other results suggest that the regulation of Wnt signaling in Hyp is complex. In this regard, the Wnt antagonist Sfrp1 that has a nonredundant role to suppress bone formation was increased in Hyp bone. Sfrp1^−/− mice have increased bone mineral density and bone formation associated with increased osteoblastic proliferation and differentiation (35). In addition, carboxypeptidase Z (Cpz), which removes carboxyl-terminal basic amino acid residues from Wnt-4 and activates Wnt/β-catenin signaling (36), is reduced in Hyp bone (37). The increase in Sfrp1 (and Sfrp4) along with the decrease in Cpz would be expected to diminish Wnt-signal in Hyp bone. In addition, a limited number of Wnt-related pathways were indentified in the GenMAPP analysis, where only Prkch (1.79), Ctnnb1 (0.8), and Ccnd2 (1.27) and Ccnd3 (1.23) were affected (data not shown). Because of the multiple changes in both positive and negative regulators of Wnt signaling, the role of Wnt in regulation of Fgf23 in Hyp remains uncertain. Nevertheless, the coordination between Sost and Fgf23 expression in osteocytes provides another potential mechanism linking osteoblast-mediated bone formation and systemic phosphate homeostasis, a possibility that will also require additional studies and in vivo validation.

We also found evidence for alterations in proteolytic pathways in Hyp bone. Although the major defect in Hyp bone is increased Fgf23 production, the possibility of a concomitant abnormality of Fgf23 degradation due to altered expression of a subtilisin-like proprotein convertases that cleaves Fgf23 at its conserved RXXR site has not been excluded. In this regard, we found that proprotein convertase subtilisin/kexin type 5 (Pcsk5) was significantly decreased in Hyp bone. Ablation of Pcsk5 in mice results in severe bone hypoplasia, retarded ossification, and absence of kidneys (39,40), but the effects of Pcsk5 on Fgf23 metabolism have not been investigated. In addition, Galnt3 is decreased approximately 1.9-fold in Hyp bone, which might contribute to decreased secretion and/or increased degradation of Fgf23. Galnt3 encodes a glycosyltransferase that mediates O-glycosylation of Fgf23, which appears to be required for secretion and protein stability as evidenced by the observation that mutations in Galnt3 result in low circulating levels of intact Fgf23 and a syndrome of tumoral calcinosis (41,42,43). Further studies are warranted to determine whether decrement in Pcsk5 impairs the degradation of Fgf23 and contributes to increased Fgf23 levels in Hyp mice.

Secreted related frizzled protein 4 (Sfrp4), which can function as an inhibitor to tolloid proteinases (44), was increased, whereas Bmp1, a tolloid proteinase that metabolizes Dmp1 into functional N-and C-terminal fragments (45), was decreased in Hyp bone. These findings raise the possibility that phosphaturic effects of Sfrp4 in vivo (26,27) might be indirectly related to altered metabolism of Dmp1 leading to increased FGF23 expression. However, alterations in Dmp1 processing have not been reported in Hyp bone. Moreover, neither Dmp1 nor Phex overexpression regulates Fgf23 promoter activity in ROS17/2.8 osteoblasts (Fig. 3), suggesting that Phex mutations increase Fgf23 transcription through secondary effects either downstream from, or independent of, Dmp1. Indeed, we observed no changes in Dmp1 expression in Hyp-derived bone.

Phex-deficient osteoblasts derived from Hyp mice also have an intrinsic defect in the mineralization of extracellular matrix that is independent of phosphate and mediated by a putative mineralization inhibitor, called Minhibin (11). In addition, increased proteolysis may be involved in the pathogenesis of impaired mineralization in Hyp (17). Many factors are known to regulate the mineralization of bone matrix, including extracellular matrix proteins that either inhibit (such as Mepe) or induce mineralization (such as Dmp1) and the local ratio of ionic inhibitors and inducers of mineralization, such as pyrophosphates and phosphate, respectively, the local concentrations in bone of which regulated by phosphatases and pyrophosphate transporters (Enpp1 and Ank) and Tnap (46,47). We found that Mepe message levels was increased 1.8-fold but found no significant changes in Dmp1, Enpp1, and Tnap in Hyp bone. Ank (1.7) and Enpp6 (1.7) were increased and decreased, respectively, in Hyp bone. More importantly, we identified several previously unrecognized abnormalities in factors regulating mineralization in Hyp bone. For example, we found that the expression of matrix Gla-protein (Mgp), an inhibitor of mineralization (48,49,50), was increased in Hyp bone. Thrombospondin 4 (Thbs4) was also found markedly increased in Hyp bone as well as Hyp-derived cultured osteoblasts. Thbs4 is a member of the group B thromospondin family of extracellular matrix glycoproteins containing heparin- and calcium-binding sites, as well as binding sites for collagens I, II, III, and V, laminin-1, fibronectin, and matrilin-2. Thbs4 is known to be expressed in osteogenic tissues (51,52), but its physiological function in bone has not been established. However, the observations that ablation of the related Thsbs3 results in accelerated endochondrial ossification (53) and overexpression of Thbs1 inhibits mineralization in osteoblast cultures (54), suggest that increased Tbhs4 in Hyp osteoblasts may play a role in the intrinsic mineralization defect. We found that carbonic anhydrase 12 (Car12) and 3 (Car3) as well as the sodium-dependent citrate transporter (Slc13a5), were decreased in Hyp bone. Car12 and Car 3 are members of the family of zinc metalloenzymes that catalyze the reversible hydration of carbon dioxide to bicarbonate and thereby have the potential to influence mineralization and osteoclast function by regulating the pH in the local milieu. Although little is known about the membrane-bound Car12 in bone, Car2, a cytosolic member of this family, causes autosomal recessive osteopetrosis via disruption of osteoclast function (55). The reduction of Slc13a5 might lead to local increases in citrate levels, which could impair mineralization, as well as deprive the cell of citrate necessary for energy metabolism in Hyp bone. Recent studies suggest that Fgf23 may also have direct effects on bone to inhibit mineralization (56). We failed to detect Klotho expression in bone on microarray analysis, consistent with the failure to detect Klotho transcripts in bone by RT-PCR (57). Fgf23 also failed to stimulate Fgf23 promoter activity in cultured osteoblasts (data not shown). Consequently, it is unlikely that putative direct actions of Fgf23 are mediated by FGFR-Klotho-dependent mechanisms.

The expression of many other genes were also altered in Hyp bone, but for most of these we currently lack evidence to support their role in the Hyp bone phenotype (see supplemental data). These include down-regulation of BMP pathways that regulate osteoblast function, skeletal development and adpipogenesis (58,59,60), complex alteration in arachidonic acid metabolism and lipoxygenase pathways (61,62), and up-regulation of purinergic (P2Y) G protein-coupled receptor family, Gpr34 and P2ry6, which could have a role in mineralization (63,64,65). Increases in Cyp11a1, which catalyzes the metabolism vitamin D₃ to form two hydroxylated products, 20-hydroxyvitamin D₃ and 20,22-dihydroxyvitamin D₃ (66), was also observed in Hyp bone. Other genes differentially regulated in Hyp bone are discussed in supplemental data.

There are several limitations to our studies. Although we identified multiple potential intermediates linking Phex to increments in Fgf23 and defects in mineralization, our analysis failed to identify putative substrates for Phex. Thus, there remains a gap in the precise mechanism whereby Phex mutations lead to the observed changes in gene expression. Our studies also highlight the difficulty in choosing the optimal model system to study gene expression profiles in complex systems, such as bone. We chose to focus on whole-bone analysis because this tissue contains all information that could potentially impact upon Fgf23 and mineralization, including multiple cell types and environmental factors for regulating gene expression in the appropriate physiological/pathological context. In contrast, evaluation of osteoblasts ex vivo, although having the advantage of greater specificity for the osteoblast lineage, are limited by cell dedifferentiation leading to cellular heterogeneity, loss of important in vivo hormonal/environmental factors, and the low abundance of osteocytes, where Phex appears to have specific functions. Indeed, the failure to identify many of the changes observed in analysis of Hyp cortical bone in immortalized osteoblasts derived from Hyp bone, highlights the advantages and disadvantages of analysis of cell lines and bone tissue. Assessment of bone encompasses the effects of loss of Phex, effects of other cells in bone that may be affected by Fgf23, and alterations in the systemic milieu derived from Fgf23 effects on the kidney (67).

We recently proposed a model whereby inactivating mutations of Phex, Dmp , and other genes impairing the proteolytic cleavage of Dmp1 might result in increased Fgf23 expression in osteocytes (Fig. 5) (32). In this model inactivating mutations of Phex, and commensurate impairment of Dmp1 proteolytic cleavage, were postulated to stimulate Fgf23 expression in osteocytes via impaired competency of extracellular matrix to mineralize and accumulation of putative matrix-derived Fgf23-stimulating factors. The current microarray analysis of Hyp bone supports many aspects of this model and also identifies Fgf and Wnt-signaling pathways as both potential regulators of Fgf23 production and coordinators of Fgf23, mineralization, and osteoblast-mediated bone formation. In addition, we define new candidates for matrix factors that mediate the intrinsic defect in mineralization in Hyp bone, including the extracellular matrix glycoprotein Tbhs4 and pH-altering transporters Car12, Car3, and Slc13a5. Further in vivo studies will be necessary to confirm the functional significance of the pathways identified by microarray analysis in the pathogenesis of the Hyp phenotype. In addition, the proximate factors linking Phex to alterations in osteocyte functions and extracellular matrix composition remain unknown.

Figure 5.

Model of gene expression pathways in Hyp bone. Inactivating mutations of Phex, Dmp1, and Fgf1 lead to increased Fgf23 production by osteocytes, and mutations of the RXXR site in Fgf23 prevents the degradation of this phosphaturic factor. Dmp1 and Phex mutations indirectly regulate Fgf23 promoter activity through the accumulation in the extracellular matrix of unknown Fgf23-stimulating factors. Wnt, Fgf, and integrin-dependent pathways are potential autocrine/paracrine factors that stimulate Fgf23 production. Dmp1 also has arginine-glycine-aspartic acid and ASARM motifs that may permit binding to integrins and Phex and has conserved sites for proteolytic cleavage by Bmp1. Increments in Sfrp4 and Pcsk5 and decrements in Bmp1 may create a functional Dmp1-deficient state by interfering with Dmp1 processing. Dmp1 and Phex mutations also result in intrinsic mineralization defects. Microarray analysis identified increments in known (Mepe, and Mgp) and potential (Thbs4) mineralization inhibitors as well as reduction in Car12, Car3, and Slc13a5 that could acidify the local milieu leading to mineral dissolution. Also shown are PG and leukotriene pathways consisting of the RAGE ligand, S100b, which can induce Ptgs2 expression and Alox15 that inhibits osteoblastogenesis; the purinergic (P2Y) G-protein coupled receptors, Gpr34 and P2ry6, that potentially sense pyrophosphate and Cyp11a1, which regulates vitamin D metabolism (see GPCR-dependent pathways in supplemental data).

Materials and Methods

Animals and tissue collection

Hyp mice, which have 3′-deletion of Phex gene on a C57BL/6 genetic background, were purchased from The Jackson Laboratory (Bar Harbor, ME) and maintained in the vivarium at University of Kansas Medical Center under standard diet purchased from Harlan Talked (Madison, WI), which contains 0.6% Ca, 0.54% Pi, and 2200 IU vitamin D. All the mice were used in accordance with the recommendations in the “Guide for Care and Use of Laboratory Animals” and the guidelines established by the University of Kansas Medical Center Institutional Animal Care and Committee. Hyp mice were genotyped by PCR as previously described (6). Because we cannot genotype Hyp female mice by PCR, we only used male mice (e.g. WT and Hyp littermates) in our study. Long bones were collected from 12-d-old WT and Hyp mice after death by cervical dislocation. Both tibia and femurs were excised and cleaned from connective tissue and muscles. Cortical bone was isolated because of its high expression of FGF23 in osteocytes of Hyp mice (10). To accomplish this, both ends of the long bones were clipped, and the remaining cortical fragments were inserted into a pipette tip and spun in a centrifuge tube at 8000 × g for 2 min to remove the bone marrow elements. The resulting cortical bone samples were snap frozen in liquid nitrogen and stored in −80 C for future RNA isolation. Cortical bones from two mice were combined for RNA isolation, and a total of four replicates per group (WT and Hyp) were used in the microarray experiment (see below). We also isolated, from WT and Hyp, long bone that had been explanted into WT mice as previously described (10).

RNA isolation

To isolate total RNA from long bone for microarray, we ground snap frozen long bone samples with a mortar and pestle and then extracted RNA using TRI REAGENT (Molecular Research Center, Inc., Cincinnati, OH) as previously described (68). RNA was isolated from pooled cortical bone samples obtained from a tibia and femur pair obtained from two mice of the same genotype. RNA integrity was confirmed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Palo Alto, CA) before microarray analysis. For confirmatory real-time PCR analysis, total RNA was extracted from long bone and cells using the TRI REAGENT protocol and treated with ribonuclease-free deoxyribonuclease (QIAGEN, Valencia, CA) to remove any residual DNA.

Microarray analysis

A total of four replicates per group (WT and Hyp) were used in the microarray experiment. Total RNA (5 μg) from each pooled sample (n = 4 mice in each group) was used to synthesize cDNA using the SuperScript Double Stranded cDNA synthesis kit (Invitrogen, Carlsbad, CA). The reverse transcriptase is driven by the annealing of an oligo dT primer coupled with a T7 promoter sequence. Then, the cDNA product was used to make the biotinylated cRNA according to the one-cycle target-labeling protocol from GeneChip Expression 3 prime-Amplification Labeling Kit (Affymetrix, Inc., Santa Clara, CA). The Affymetrix GeneChip Mouse Genome 430 2.0 Arrays (Affymetrix, Inc.) were hybridized for 16 h at 45 C with fragmented biotin-labeled cRNA at a concentration of 50 ng/μl using GeneChip Hybridization oven 640. GeneChips were washed, stained, and scanned according to the manufacturer’s instructions using Affymetrix GeneChip Scanner 3000 7G with Autoloader at University of Kansas Medical Center.

The current experimental design and analysis are compliant with current MIAME guideline (69), and the data are available in the Gene Expression Omnibus (GEO) http://www.ncbi.mlm.nih.gov/projects/geo/NCBI database (accession number pending).

Real-time quantitative RT-PCR

Real-time quantitative RT-PCR was performed as previously described (68). Briefly, first-strand cDNA was synthesized using iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Inc., Hercules, CA). Total RNA (1 μg) was used in each 20-μl reverse transcriptase reaction and 2 μl of cDNA was used in each PCR. The iCycler iQ Real-Time PCR Detection System and iQ SYBR Green Supermix (Bio-Rad) were used for real-time quantitative PCR analysis. A no-reverse transcriptase control was also amplified to confirm the absence of DNA contamination. The fold change in genes relative to the cyclophilin A endogenous control was determined by the 2^−ΔΔCt method (70). Sequences of primers used for regular and real-time quantitative RT-PCR were listed in supplemental Table 2.

TMOb cell culture and differentiation

TMOb-WT and TMOb-Hyp immortalized cells lines derived from normal and Hyp mice calvariae were maintained in α-MEM containing 10% fetal bovine serum and induced for differentiation toward osteoblasts in the differentiation medium (α-MEM containing 10% fetal bovine serum, 5 mm β-glycerophosphate, and 25 μg/ml ascorbic acid) to induce osteoblastic differentiation as reported previously (71). During the culture period, the medium was changed every 3 d. Both TMOb-WT and TMOb-Hyp cells were harvested for RNA extraction for a culture periods of up to 18 d.

TOPflash activity assay

Both TMOb-WT and TMOb-Hyp cells were seeded in six-well plates at density of 2 × 10⁵ cells per well. TOPflash Reporter Plasmid (1.5 μg/well) (Upstate Biotechnology, Inc., Lake Placid, NY) and 15 ng/well pRL-TK Renilla plasmid (Promega Corp., Madison, WI) were transfected into the cultured cells with Lipofectamine 2000 (Invitrogen) following the protocol from the manufacturer. After an additional 24-h culture period, cells were harvested and the luciferase activities were measured with Dual-Luciferase Reporter Assay kit (Promega). The relative luciferase activity was calculated by firefly activity (TOPflash)/Renilla activity.

FGF23 promoter studies

A 7980-bp FGF23 5′-flanking region from −7980 to −1 relative to the translation start site ATG was amplified from genomic DNA extracted from C57Blk6/J mouse using PCR with forward primer: GGCTAGCTCAGGTCACTGGTTCTGAAG and reverse primer: CCTCGAGCACAGCACTGAGTGGCTAATGC. The PCR product was cloned into a pGL3-Basic vector (Promega Corp.) between NheI and XhoI restriction sites to generate a Fgf23 promoter/firefly luciferase reporter-construct (p7980FGF23-luc). An RL-TK construct (Promega) was used as an internal control for transfection efficiency. ROS17/2.8 osteoblasts were grown and transfected with promoter constructs using the FuGENE 6 Transfection Reagent (Roche Applied Science, Indianapolis, IN) (3). Briefly, ROS17/2.8 osteoblasts were plated in 24-well plates. After overnight incubation, the cells were transfected with FGF23-promoter constructs and pRL-TK using the FuGENE 6 Transfection Reagent for 16 h and replaced with quiescent medium (DMEM/F-12 medium containing 0.1% BSA), after overnight incubation, the quiescent medium was replaced with the medium containing different growth factors indicated in Results. In the FGF treatment studies, FGF1 and FGF2 were prepared in the quiescent medium containing 10 μg/ml of heparin. After a 24-h treatment period, the treated cells were lysed in 100 μl Passive Lysis Buffer (Promega), and 20 μl of the sample were assayed for Firefly and Renilla luciferase activities using a Dual-Luciferase Reporter Assay System (Promega). Human recombinant FGF1 and human TGF-β were purchased from Sigma Chemical Co. (St. Louis, MO). Mouse recombinant FGF2 was purchased form PROSPEC protein specialists (Rehovot, Israel). FGFR inhibitor PD173074 was purchased from EMD Chemicals, Inc. (San Diego, CA).

To evaluate effects of Phex and Dmp1 on FGF23 promoter activity, we used previously generated Phex expression construct in pcDNA3.1/V5-His-TOPO vector (71) and generated a Dmp1 expression construct with similar strategy using total RNA from long bone and RT-PCR method with primer Dmp1 F (5′-CGCATCCCAATATGAAGACTG-3′) and Dmp1 R (5′-GTAGCCGTCCTGACAGTCAT-3′). We cotransfected Phex/pcDNA3.1/V5-His-TOPO, Dmp1/pcDNA3.1/V5-His-TOPO or both with the FGF23 promoter/reporter construct (p7980FGF23-luc) in ROS17/2.8 cells, quiesced the cells for 2 d and then measured the Firefly and Renilla luciferase activity.

Nuclear protein extraction and Western blot

Cell nuclear fraction protein was obtained following the previously reported method (72) with some modification. Long bones from 4-wk-old mice were collected and cleared from soft tissues. Long bones (50–100 mg) without marrow were homologized with Ultra-Turrax (IKA-T25 digital) in 1 ml of solution A (10 mm HEPES at pH 7.9, 10 mm KCl, 1.5 mm MgCl₂, 0.34 m sucrose, 10% glycerol, 1 mm dithiothreitol, 10 mm NaF, 1 mm Na₂VO₃, and protease inhibitors cocktail, 0.1% Triton X-100) and then incubated on ice for 5 min. Cytoplasmic proteins were separated from nuclei by centrifuge at 1300 × g for 4 min at 4 C. The isolated nuclei pellets were washed twice with solution A and lysed in 1 ml of high-salt NETN lysis buffer (50 mm Tris/HCl, pH 7.5; 420 mm NaCl; 1 mm EDTA; 1% Nonidet P40; 10 mm NaF; and 1 mm Na₂VO₃, 1 mm dithiothreitol, and protease inhibitors cocktail), incubated on ice for 20 min, and then centrifuged at 14,000 × g for 20 min at 4 C. The supernatant was transferred to a fresh tube and used as nuclear extract for Western blot analysis. Twenty to 40 μg protein were separated on a 10% SDS-PAGE gel (Invitrogen) and transferred to a nitrocellulose membrane. The following antibodies were used in Western blot analysis: anti-β-catenin antibody (1:1000 dilution; Cell Signaling Technology, Inc., Beverly, MA), phospho-β-catenin antibody (Ser33/37/Thr41; 1:1000, Cell Signaling Technology, Inc.), and anti β-actin antibody (1:1000; Sigma-Aldrich). Horseradish peroxidase-conjugated antimouse or antirabbit IgG was used as a secondary antibody (1:4000 and 1:8000, respectively; Sigma-Aldrich), and the ECL Plus Western Blotting Detection Kit (Amersham Biosciences, Piscataway, NJ) was used for detection.

Statistical analysis

Microarray data were analyzed using GeneSpring GX 9 (Silicon Genetics). The Robust Multichip Averaging probe summarization algorithm (73) was used to perform background correction, normalization, and probe summarization. The baseline transformation to median of all samples was applied. Genes were filtered to include only those that had a raw score about 20.0 percentile in at least one of eight sample values. The significance analysis was performed using standard t test with Benjamini-Hockberg multiple testing correction. To minimize the false positive/negative error rate in our MicroArray, we used a combination of confidence (P < 0.05) and fold change (i.e. ≥ 2.0-fold) to restrict our gene lists. To analyze these expression data in terms of processes disrupted in the Hyp mice, we performed a functional analysis based on the Gene Ontology (GO) classification. Each gene on the microarray corresponds to one or more specific GO terms according to the function of its proteins. In the analysis, we compared the number of modulated genes with the total number of genes on microarray in each GO term, with restriction to the main GO class Biological Process.

For GO analysis we used the BiNGO (74), a plug-in for Cytoscape (28), an open source bioinformatics software platform. Real-time RT-PCR data were analyzed using Student’s t test. A significant difference was determined at P ≤ 0.05. Biological pathways were identified using GenMAPP (http://www. genmapp.org/), a recently developed tool for visualizing expression data in the context of biological pathways.