During embryogenesis, the bones of the axial and appendicular skeleton and most of the facial bones develop by endochondral ossification. During this process, mesenchymal cells condense and differentiate into chondrocytes forming the cartilage anlagen of the future bones. The chondrocytes undergo a series of maturation steps, from low proliferating distal cells into high proliferating columnar, prehypertrophic, hypertrophic, and terminal hypertrophic chondrocytes. Subsequently, terminal hypertrophic chondrocytes are replaced by bone and bone marrow to form the ossified skeleton (Kronenberg,2003; Goldring et al.,2006). Disruption of any of these steps leads to a wide spectrum of skeletal abnormalities ranging from mild alteration in the shape and size of single bones to severe malformations of the entire skeleton (Zelzer and Olsen,2003).
Several human skeletal disorders and mouse mutants have been associated with mutations in chondroitin sulfate (CS) and/or heparan sulfate (HS) carrying proteins of the proteoglycan network, such as aggrecan, glypican 3, and perlecan. Other mutations affect genes required for the synthesis and modification of the CS chains like chondroitin 6-O-sulfotransferase 1, and the HS chains like the glycosyltransferases exostosin 1 (Ext1) and Ext2, or the HS 2-O-sulfotransferase (Hs2st), demonstrating the importance of proteoglycans for the differentiation of skeletal structures (Schwartz and Domowicz,2002; Thiele et al.,2004; Farach-Carson et al.,2005).
HS proteoglycans consist of core proteins carrying at least one covalently linked HS chain. The linear polysaccharide backbone of the HS chains consists of up to 400 alternating units of N-acetylglucosamine and glucuronic acid synthesized in the Golgi apparatus by heteromeric complexes of Ext1 and Ext2. Subsequently, the HS chains are partially modified along their length by several enzymes, which catalyze N-deacetylation, C5-epimerization, and sulfation at N-, 3-O-, and 6-O-positions of glucosamine residues and at 2-O-positions of iduronic acid residues. The expression of distinct HS modifying enzymes in specific cell types and tissues generates a wide spectrum of structurally diverse HS chains (Ledin et al.,2004; Lamanna et al.,2007). HS proteoglycans have been shown to control the distribution and receptor binding of several growth factors such as bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs), hedgehogs, and Wnts (Farach-Carson et al.,2005; Gorsi and Stringer,2007). For several members of the FGF-family, binding preferences for distinct sulfation patterns of the HS chain have been identified (Ashikari-Hada et al.,2004; Noti et al.,2006), whereas binding modalities of other growth factors are less well characterized.
The HS sulfation pattern of a given cell is mainly determined during the process of HS synthesis in the Golgi apparatus. In addition, it can be modified extracellularly by the secreted 6-O-endosulfatases Sulfatase 1 (Sulf1) and its closely related homolog Sulfatase 2 (Sulf2). Both enzymes catalyze the hydrolysis of sulfate ester bonds from 6-O-positions of glucosamine residues of HS (Lamanna et al.,2007). QSulf1, the quail ortholog of Sulf1, has been shown to regulate multiple signaling pathways either positively or negatively. For example, 6-O-desulfation of HS by QSulf1 releases HS-bound Noggin, an inhibitor of BMP signaling, from the surface of CHO cells, thereby restoring responsiveness of these cells to BMPs (Viviano et al.,2004). QSulf1, Sulf1, and Sulf2 have been shown to promote Wnt signaling in mammalian cell cultures (Ai et al.,2003; Dhoot et al.,2001; Nawroth et al.,2007) and overexpression of QSulf1 in the spinal cord of chicken embryos enhances the range of Sonic hedgehog (Shh) signaling (Danesin et al.,2006). In contrast, overexpression of QSulf1 in Xenopus and chicken embryos inhibits FGF-signaling likely by disrupting the formation of the ternary FGF2-HS-FGFR1 complex (Wang et al.,2004). Several investigations revealed that Sulf1 and Sulf2 are widely expressed in the developing embryo (Ohto et al.,2002; Nagamine et al.,2005; Zhao et al.,2006; Holst et al.,2007; Lum et al.,2007). It is thus likely that secreted 6-O-endosulfatases regulate the binding modalities of HS and different growth factors in a time- and organ-specific manner.
We have previously shown that reduced amounts of HS in Ext1 mutant mice play a critical role in regulating Indian hedgehog (Ihh) signaling in the developing growth plates of endochondral bones (Koziel et al.,2004). In this study, we investigated the role of Sulf1 and Sulf2 during skeletogenesis. We have analyzed the expression patterns of both genes in skeletal tissues of the developing mouse embryo and identified overlapping but distinct expression domains. Similar to other studies, we found both sulfatases expressed in the joint region and identified novel sites of Sulf1 expression such as the hypertrophic chondrocytes, the apical ectodermal ridge (AER), and the branchial arch ectoderm. To analyze the role of the 6-O-endosulfatases in vivo, we investigated mice carrying loss-of-function alleles for both genes. Although homozygous deletion of either Sulf1 or Sulf2 only results in mild skeletal alterations, double homozygous Sulf1;Sulf2 mutants are characterized by reduced body weight and a reduced size of the skeleton. Moreover, they display distinct malformations in the sternum, the lumbar and tail vertebrae, and an advance in bone differentiation, indicating that Sulf1 and Sulf2 play redundant roles in modulating skeletal development.
Expression of Sulf1 and Sulf2 in Skeletal Tissues
To gain insights into the potential role of secreted 6-O-endosulfatases during skeletogenesis, we examined the expression pattern of Sulf1 and Sulf2 during embryonic development. Similar to previous studies (Ohto et al.,2002; Nagamine et al.,2005; Holst et al.,2007; Lum et al.,2007), whole mount and section in situ hybridization on E9.5 to E11.5 embryos revealed that Sulf1 and Sulf2 are broadly expressed throughout the embryo, with partially overlapping regions in the telencephalic vesicle, the nasal placodes, the somites, the tip of the tail (Fig. 1A–F), and the floor plate (Fig. 1H,I). In addition, our study revealed a dynamic expression pattern for both sulfatases during endochondral ossification and branchial arch- and limb outgrowth. At E9.5, Sulf1 is expressed in the rostral surface ectoderm of the first branchial arch (BA1) and at E10.5 in the oral ectoderm and in the distal mesenchyme of the mandibular primordia (Fig. 1A,B,H). In the limb bud, Sulf1 is expressed in the AER at E10.5 and E11.5 (Fig. 1B,C). The identity of BA1 ectoderm and AER was confirmed by hybridization of Fgf8, which is specifically expressed in both ectodermal tissues (Fig. 1G,J). Sulf1 is further expressed in the distal mesenchyme of the limb at E11.5 and E12.5 (Fig. 1C,L,M) but was not detected in the proximal mesenchyme of the limb or the anlagen of the femur at E12.5 (Fig. 1O). In contrast, Sulf2 expression was not detected in the surface ectoderm of BA1 or in the AER of the limb bud (Fig. 1I,Q). Instead, Sulf2 is highly expressed throughout the underlying mesenchyme of both tissues at E10.5 (Fig. 1E,I,Q). At E11.5 and E12.5, Sulf2 expression is down-regulated in the condensing limb mesenchyme (Fig. 1R–T) demarcated by the expression of type II collagen (Col2a1) (Fig. 1N) but maintained in the developing synovial joints of the limb, where it is co-expressed with Sulf1 at E12.5 (Fig. 1O,T) and E14.5 (Figs. 1P,U, 2A,B). Both sulfatases are also co-expressed in fibrocartilaginous joints between individual sternebrae and between the sternum and ribs (sternocostal) at E14.5 (Fig. 2E,F).
At E16.5, expression of Sulf1 is down-regulated in most joints of the body except for the digits, whereas the expression of Sulf2 persists in all joints examined (Fig. 2C,D,G,H). Although Sulf2 expression is down-regulated in the early condensing mesenchyme, it is re-expressed in proliferating chondrocytes of the skeletal elements starting at E13.5 (Fig. 2B,D,F,H and data not shown). Sulf1 is expressed in hypertrophic chondrocytes at E14.5 and E16.5 (Fig. 2A,C,G). Weak expression of Sulf1 and Sulf2 could also be detected in the developing bone at E16.5 (Fig. 2C,D,K,L). We did not observe Sulf1 or Sulf2 expression in the perichondrium of the skeletal elements of the limbs, but the neighboring connective tissue expressed both genes (insets in Fig. 2C,D).
In addition to the developing bones, Sulf1 is expressed surrounding the sternal bands (Fig. 2E) and Sulf2 in the developing tendons at E14.5 (data not shown). Both sulfatases are also expressed in non-overlapping patterns during tooth formation. At E14.5, Sulf1 expression can be detected at low levels in the dental epithelium, whereas Sulf2 is strongly expressed in the underlying mesenchyme (Fig. 2I,J). At E16.5, Sulf1 expression can be found in ameloblasts and Sulf2 is highly expressed in the pulp of the teeth (Fig. 2K,L).
Generation of Sulf1gt- and Sulf2gt-Deficient Mice
To investigate the role of secreted 6-O-endosulfatases, we generated mouse lines deficient for Sulf1 and Sulf2 using BayGenomics gene trap (gt) ES-cell lines. Mutant alleles are referred to as Sulf1gt and Sulf2gt. To characterize these alleles, we determined the exact integration site of the gt-vectors by PCR of genomic DNA using a set of primers covering the respective introns identified by 5′RACE (BayGenomics database, http://baygenomics.ucsf.edu). In the Sulf1gt allele, the gt-vector integrated in intron 1B, 4,105 bp upstream of the non-coding exon 2 (Fig. 3A). Analysis of β-galactosidase activity revealed a weak expression in the digits of E13.5 Sulf1gt/gt mutants (Fig. 3C), which express high levels of Sulf1 mRNA in wild-type embryos (Fig. 1P). As the gt-allele lacks the Sulf1 translation initiation codon located in exon 4, the low amounts of β-galactosidase expression suggests the use of a cryptic translation initiation codon. In the Sulf2gt allele, the gt-vector integrated 1,117 bp downstream of exon 3 (Fig. 3B). β-galactosidase expression from the mutant allele (Fig. 3F) recapitulates the expression pattern of Sulf2 in wild-type embryos (Fig. 1U). Splicing of the mutant allele results in a protein with the N-terminal 138 amino acids of Sulf2 fused to the β-geo reporter. As the active site of Sulf2 is located in exon 4 and exon 5, the Sulf2-β-geo hybrid protein is not expected to be catalytically active.
To investigate the deletion efficiency of the gt-insertions, we measured wild-type and mutant transcript levels of Sulf1 and Sulf2 by quantitative RT-PCR with primers located downstream of the gt-insertion site. The mRNA level of Sulf1 was reduced to approximately 3% in E14.5 Sulf1gt/gt limbs compared to wild-type littermates, indicating that the Sulf1gtmutation generates a hypomorphic allele of Sulf1 (see Supplemental Fig. 1A, which can be viewed at www.interscience.wiley.com/jpages/1058-8388/suppmat). Analysis of expressed sequence tags (ESTs) in the NCBI mouse EST database identified different splice forms in the 5′ UTR of Sulf1. In most of these ESTs, exon 1A (63/83) or exon 1B (6/83) are spliced to exon 2, whereas in some ESTs exon 1A (13/83) or exon 1B (1/83) were directly spliced to exon 3 (Fig. 3A). Since the Sulf1 gt-vector integrated upstream of exon 2, the alternative splicing of this region might facilitate the expression of the endogenous allele. For the Sulf2gt allele, we detected approximately 0.4% of wild-type transcripts in E14.5 Sulf2gt/gt limbs, indicating that it also represents a hypomorphic allele (see Suppl. Fig. 1B). To analyze if homozygous loss of either Sulf1 or Sulf2 affects the transcription level of the remaining sulfatase alleles, we analyzed E14.5 compound mutants by quantitative RT-PCR and did not detect compensatory changes in the expression levels (Suppl. Fig. 1C,D).
In accordance with the data obtained by quantitative RT-PCR, we detected a decrease of wild-type transcripts in homozygous gt-mutants also by in situ hybridization with antisense riboprobes specific for the 3′ UTRs of Sulf1 and Sulf2. For both probes, strong signals were detected in E14.5 wild-type limbs (Fig. 3D,G), whereas Sulf1 is weakly expressed in the joints of Sulf1gt/gt mutants (arrowhead in Fig. 3E) and Sulf2 expression is not detected in Sulf2gt/gt mutants (Fig. 3H).
Phenotypic Characterization of Double Homozygous Sulf1;Sulf2 Mutants
Since single homozygous mutants for either Sulf1gt/gt or Sulf2gt/gt show no obvious phenotype, we generated compound mutants. Double heterozygous Sulf1+/gt;Sulf2+/gt animals are fertile and do not display any obvious phenotype. Crossing double heterozygous with double heterozygous or double homozygous carriers produced normal litter sizes and offspring in Mendelian ratios at embryonic stages E16.5–E18.5 and postnatally at 3 weeks of age (Suppl. Table 1). However, between weaning and the 6th month of life, 11% (5/47) of Sulf1gt/gt;Sulf2gt/gt mutants died, whereas mortality of Sulf1+/gt;Sulf2+/gt animals was 1% (1/80). As the relatively mild phenotype (see below) of the Sulf1gt/gt;Sulf2gt/gt mutants might be due to the hypomorphic nature of the gt-alleles, we also investigated mice carrying targeted deletions of Sulf1 (Sulf1−) and Sulf2 (Sulf2−) (Lamanna et al.,2006). In contrast to Sulf1gt/gt;Sulf2gt/gt mutants, litter sizes of Sulf1−/−;Sulf2−/− mutants are reduced and the offspring displays a high mortality rate of 54% during the first 6 weeks after birth (Lamanna et al.,2007).
Analysis of body weight of Sulf1gt/gt;Sulf2gt/gt animals (6.4 g) at three weeks of age (P21) revealed a significant reduction to 70% (P < 0.0001) compared to Sulf1+/gt;Sulf2+/gt animals and to 77% compared to mutants lacking three alleles of Sulf1 and Sulf2 (Table 1). Occasionally, we observed strongly dwarfed animals in the Sulf1gt;Sulf2gt mouse colony, with a body weight below 5 g at P21. Those animals displayed a starved appearance, were mostly (3/5) homozygous for both gt-alleles (Sulf1gt/gt;Sulf2gt/gt), and rarely retained one functional allele of either Sulf1 (1/5) or Sulf2 (1/5). At later stages of development (5–6 months), the body weight of Sulf1−/−;Sulf2−/− mutants was reduced to ∼60% of that of wild-type controls. In contrast, Sulf1gt/gt;Sulf2gt/gt mutants were not significantly smaller compared to Sulf1+/gt;Sulf2+/gt animals, indicating that surviving gene trap mutants catch up in growth during postnatal development. Additionally, at P21 some (5/21) Sulf1gt/gt;Sulf2gt/gt mutants developed misaligned, overgrown incisors most likely causing feeding problems (Fig. 4B). Similar tooth malformations were also observed in Sulf1−/−;Sulf2−/− mutants. Except for strongly dwarfed individuals, Sulf1−/−; Sulf2−/−, Sulf1gt/gt;Sulf2gt/gt, Sulf1+/gt;Sulf2gt/gt, and Sulf1gt/gt;Sulf2+/gt mutants were fertile and were used for breeding. Recently, a neuronal innervation defect of Sulf1−/−;Sulf2−/− esophagi has been reported likely contributing to the postnatal growth retardation in Sulf1;Sulf2 compound mutants (Ai et al.,2007).
Table 1. Reduced Body Weight of Newborn (P0) and Postnatal (P21) Sulf1gt/gt;Sulf2gt/gt Mutantsa
Mean body weight [(g) ± SD]
% of control
t-test P value
N, number of investigated animals; NS, not significant; SD, standard deviation.
1.17 ± 0.05
1.18 ± 0.08
1.17 ± 0.09
1.06 ± 0.08
9.21 ± 1.81
8.43 ± 1.87
8.32 ± 1.72
6.43 ± 1.35
To exclude the influence of feeding problems on postnatal growth, we investigated animals at perinatal stages. We found the body weight of P0 Sulf1gt/gt;Sulf2gt/gt animals significantly reduced to 91% (P = 0.0016) compared to double heterozygous animals or to mutants carrying one functional Sulf1 or Sulf2 allele, indicating a growth defect independent of postnatal nutrition (Table 1).
Skeletal Defects in Sulf1;Sulf2 Compound Mutants
In addition to their reduced size, Sulf1gt/gt;Sulf2gt/gt animals were frequently characterized by kinky tails indicating defects in skeletal development. In accordance with the reduced body weight, alcian blue and alizarin red staining at P21 revealed smaller skeletons (Fig. 4B). The rib cage was especially reduced in size with a shortened and broadened sternum (Fig. 4F,G). Quantification of radius and sternum length revealed a reduction to 91% (P < 0.0001) and 87% (P < 0.0001), respectively (Table 2).
Table 2. Radius and Sternum Length Are Reduced in Sulf1;Sulf2 Compound Mutantsa
Mean length [(mm) ± SD]
Mean length [(mm) ± SD]
t-test P value
Length was measured on skeletal preparations (E18.5, P0, P21) and on longitudinal paraffin sections of E16.5 radii. Sternum length was measured from the cranial end of the manubrium to the cranial side of the ossified xiphoid, omitting the variable length of the caudal xiphoid. P0 Sulf1−/−;Sulf2−/− mutants were compared to age-matched wild-type animals from independent breedings. gt, gene trap;N, number of investigated animals; ND, not determined; NS, not significant; SD, standard deviation.
The seven P21 control animals consist of 4x Sulf1+/gt;Sulf2+/gt; 1x Sulf1+/+;Sulf2+/gt 1x Sulf1+/gt;Sulf2+/+, and 1x Sulf1+/+;Sulf2+/+ animals.
Similarly, E18.5 Sulf1gt/gt;Sulf2gt/gt mutants were reduced in size displaying a shortening of the radius to 93% (P = 0.0013) and the sternum to 94% (P = 0.046) compared to Sulf1+/gt;Sulf2+/gt littermates. At this stage, sternum length was also reduced to 93% (P = 0.014) in Sulf1gt/gt mutants (Table 2). Measurements of P0 Sulf1−/−; Sulf2−/− animals compared to age-matched wild-type controls revealed a significant reduction in length of radius (to 91%; P = 0.0006) and sternum (to 84%; P < 0.0001) (Table 2). Newborn homozygous Sulf1−/−;Sulf2−/− mice were of similar size as age-matched Sulf1gt/gt;Sulf2gt/gt animals (Table 2).
Although the patterning of the skeletal elements appears grossly normal, we detected four distinct defects in the axial skeleton that occur with high penetrance in Sulf1;Sulf2 compound mutants (Tables 3, 4). Overall penetrance and severity of the skeletal alterations is increased with increasing numbers of mutant alleles and was highest in Sulf1−/−;Sulf2−/− mutants.
Table 3. Frequency of Skeletal Phenotypes of Postnatal (≥P21) Sulf1;Sulf2 Compound Mutantsa
1In wild-type mice, the sternum consists of six ossified elements from cranial to caudal referred to as manubrium, sternebrae 1 to 4 (st1 to st4) and xiphoid. These elements are separated by fibrocartilaginous joints in all wild-type animals investigated (Table 3). In single homozygous Sulf1gt/gt and Sulf2gt/gt animals, we occasionally detected partial fusions of st3 and st4 at postnatal stages (Table 3 and data not shown). This phenotype showed nearly full penetrance (94%) and increased severity in Sulf1gt/gt;Sulf2gt/gt compound mutants. In 65% of the animals, st3 and st4 are completely fused and in 18% st2 and st3 are partially fused in addition (Fig. 4G; Table 3). Moreover, all Sulf1−/−;Sulf2−/− animals analyzed, displayed complete fusion of st3 and st4 and 40% showed additional partial fusion of st2 and st3 (Table 3).Careful examination of perinatal skeletons revealed small bony bridges to broad connections between two or three sternebrae (Fig. 4D,E) in Sulf1gt/gt;Sulf2gt/gt mutants already at E18.5 (Table 4). At P0/P1, 85% of Sulf1gt/gt;Sulf2gt/gt animals investigated showed different degrees of sternebrae fusions, but fusions of more than 3 sternebrae were not observed. In contrast, all of the Sulf1−/−;Sulf2−/− mutants displayed sternebrae fusions, including 29% with fusions between manubrium and 3 or 4 sternebrae (Fig. 4E; Table 4). Animals with asymmetrically attached ribs, which can lead to secondary joint fusions, were occasionally observed in perinatal wild-type (6%, 2/33) and Sulf1gt/gt;Sulf2gt/gt (3%, 1/29) animals and were excluded from the analysis.
2During vertebrae development, ossification is initiated in two lateral and one central ossification centers, which subsequently fuse to form the ossified vertebrae. At E18.5, these ossification centers are still separated in sacral vertebra 2 (S2), whereas thin ossified connections were observed in 39% of wild-type animals at P0/P1 (Table 4). More anterior located vertebrae, such as sacral vertebra 1, and lumbar and thoracic vertebrae, show no sign of fusions at P0/P1. However, E18.5 Sulf2gt/gt mutants displayed fusions of S2 in 86% and fusion of more anterior located vertebrae in 42% of the embryos. Moreover, 92% of E18.5 Sulf1gt/gt;Sulf2gt/gt mutants displayed connections in 1 to 5 vertebrae anterior of S2 (Table 4). In Sulf1gt/gt;Sulf2gt/gt mutants, fusions of the ossification centers are restricted to the sacral and lumbar vertebrae at P0/P1, whereas 50% of the Sulf1−/−;Sulf2−/− mutants display such fusions in up to 12 vertebrae, some of them located in the thoracic region (Table 4). Neither Sulf1gt/gt nor Sulf1gt/gt;Sulf2+/gt animals displayed accelerated ossification of vertebrae, indicating that Sulf2-dependent HS desulfation is the main modulator of this process (Fig. 5A,B; Table 4 and data not shown).
3Later in vertebrae development, the left and right arches of each vertebra grow around the spinal cord and fuse dorsally. In postnatal animals (P21 and older), single open vertebrae, in most cases thoracic vertebra 10 (T10), were found with low penetrance (∼20%) in all genotypes examined (Table 3). The frequency of at least one dorsally split vertebra was increased to 71% in Sulf1gt/gt mutants and to 88% in Sulf1gt/gt;Sulf2gt/gt mutants. More severe phenotypes affecting 2 to 6 vertebrae were observed in 14% of Sulf1gt/gt mutants, 47% of Sulf1gt/gt; Sulf2gt/gt double mutants, and 80% of Sulf1−/−;Sulf2−/− mutants (Fig. 5C–E; Table 3). In contrast to the ossification process, dorsal vertebrae fusion seems to depend more on Sulf1 function, indicating slightly distinct roles of the homologs.
4Calcified fusions of up to 7 distal tail vertebrae, which were never seen in wild-type controls, were identified in 94% of the ≥P21 Sulf1gt/gt;Sulf2gt/gt and 80% of the Sulf1−/−;Sulf2−/−animals, likely leading to the kinky tails observed in postnatal animals (Fig. 5H–J; Table 3).
In addition to the described axial defects, perinatal Sulf1gt/gt;Sulf2gt/gt mutants displayed incomplete closure of the basisphenoid bone of the skull base (5/22) and dorsally malformed second cervical vertebra (C2) (9/29) with low penetrance (Fig. 5F,G,K,L). In 3 of these cases, C2 seemed to be homeotically transformed into the first cervical vertebrae as indicated by the occurrence of a ventral tubercle (Fig. 5N,O). In Sulf1−/−;Sulf2−/− animals, the occurrence of dorsally misshaped second cervical vertebra (11/20) and incomplete closure of the basisphenoid bone (10/10) was increased compared to Sulf1gt/gt;Sulf2gt/gt mutants (Fig. 5M,P; Table 4; data not shown). No patterning defects could be identified in the appendicular skeleton.
Disturbed Chondrocyte Differentiation in Sulf1gt/gt;Sulf2gt/gt Mutants
As we observed a reduced size in skeletons of P0 Sulf1gt/gt;Sulf2gt/gt and Sulf1−/−;Sulf2−/− mutants, we analyzed chondrocyte differentiation in longitudinal sections of E16.5 Sulf1gt/gt;Sulf2gt/gt radii. We found a reduction in total length to 89% (P = 0.0004) compared to Sulf1+/gt;Sulf2+/gt mutants. Interestingly, neither the length of the ossified region (84%) nor the zone of hypertrophic chondrocytes (98%) was significantly altered (Fig. 6I,J; Table 2), whereas the zone of proliferating chondrocytes was reduced to 90% (P < 0.0001) (Table 2). Analyzing chondrocyte proliferation by phospho-Histone H3 staining did not detect significant differences in the proliferation rate of distal (zone I) or columnar (zone II) chondrocytes in Sulf1gt/gt;Sulf2gt/gt mutant animals. Similarly, no significant difference in cell density could be detected in these mutants (Fig. 6O–S). As the alteration in the region of proliferating chondrocytes is small, changes in proliferation rate or cell density might be masked by the phenotypic variation between litters. Alternatively, an accelerated transition from proliferating into hypertrophic chondrocytes might cause the observed reduced region of proliferating cells.
As individual sternebrae are fused postnatally in Sulf1gt/gt;Sulf2gt/gt mutants, we investigated sternum development in more detail at embryonic stages. The sternum develops from two rod-like mesodermal condensations (sternal bands), which fuse in the ventral midline. Signals from the ribs are required to induce and maintain the fibrocartilaginous joints separating the individual sternebrae, which undergo endochondral ossification (Chen,1952,1953; Storm and Kingsley,1996). Defects in sternum segmentation might thus originate in disturbed joint initiation or disturbed maintenance of the joint chondrocytes in an undifferentiated stage.
In longitudinal sections of Sulf1gt/gt;Sulf2gt/gt mutants at E15.5, fusions of single sternebrae were not detectable indicating that the induction of joints was not obviously disturbed. Molecular analysis of chondrocyte differentiation at E16.5 revealed an enlarged zone of Col10a1 expressing hypertrophic chondrocytes in the sternebrae (Fig. 6A–D). Next we analyzed the expression of Patched1 (Ptch1), which is strongly expressed in high proliferating chondrocytes flanking the hypertrophic region, but only weakly in less differentiated chondrocytes of the fibrocartilaginous joint. Double homozygous mutants display a reduced region of less differentiated chondrocytes between the individual Ptch1 domains at E16.5 and E18.5 (Fig. 6E–H,M,N). Furthermore, no undifferentiated chondrocytes were found between st3 and st4 in severely affected mutants at E18.5 (Fig. 6K–N). Accelerated chondrocyte differentiation is thus likely to contribute to the sternebrae fusion observed in postnatal skeletons.
Disturbed Mandible Development in Shh;Sulf1;Sulf2 Mutants
As Sulf1 and Sulf2 are expressed in the first branchial arch at E10.5 (Fig. 1H,I), we carefully analyzed mandible formation in Sulf1gt/gt;Sulf2gt/gt mutants. Indeed, 2 out of 29 perinatal Sulf1gt/gt;Sulf2gt/gt mutants displayed fused anterior tips of the left and right mandible (Fig. 7A–C), which was never observed in embryos carrying at least one wild-type allele of either Sulf1 or Sulf2. Amongst others, Shh signaling plays a critical role in lower jaw development (Brito et al.,2006) and overexpression of QSulf1 increases the range of Shh signaling (Danesin et al.,2006). Loss of Shh signaling leads to holoprosencephaly and cyclopia in Shh−/− mutants, whereas for heterozygous Shh+/− animals, no mandible phenotype has been described (Chiang et al.,1996). To test if loss of 6-O-endosulfatases might decrease Shh signaling below the heterozygous level, we analyzed mice lacking one allele of Shh in the Sulf1gt;Sulf2gt mutant background. Out of 126 embryos (E14.5 to E18.5) heterozygous for Shh+/− and hetero- or homozygous for Sulf1gt and Sulf2gt, three individuals possessed no mandible (agnathia) and two showed severely shortened mandibles (micrognathia) (Fig. 7F,G). In addition to the mandible defect, the agnathic embryos were characterized by small or absent eyes and a narrow craniofacial region (Fig. 7F,G). In contrast, none of the 100 control embryos carrying the wild-type Shh allele and different combinations of Sulf1gt and Sulf2gt alleles displayed a shortened mandible. As one micrognathic individual carried only one Sulf2gt and the second individual one Sulf1gt and one Sulf2gt allele, we cannot completely rule out a sulfatase-independent effect of Shh. Nevertheless, agnathia was only detected in double homozygous Sulf1gt/gt;Sulf2gt/gt mutants, which lacked one Shh allele (3 out of 27) supporting a role for HS 6-O sulfation in modulating Shh signaling during BA1 outgrowth. To investigate if mandible development was also affected in Shh+/−;Sulf1gt/gt;Sulf2gt/gt animals with no obvious lack of mandible structure, we analyzed the size of 5 E18.5 mandibles after alcian blue alizarin red staining. No difference could be detected if compared to Shh+/+;Sulf1gt/gt;Sulf2gt/gt (n = 3) animals. Further analysis on stronger alleles will thus be required to understand the role of sulfatases in Shh signaling.
Skeletal Phenotypes of Sulf1;Sulf2 Mutants
In this study, we have investigated mice carrying loss of function alleles for the secreted HS 6-O-endosulfatases, Sulf1 and Sulf2. Analysis of mice carrying either hypomorphic gene trap or targeted null alleles identified a highly redundant function of both genes. Single homozygous gt-mutants display no obvious phenotype except for mild forms of dorsally split vertebrae in Sulf1gt/gt mutants and premature vertebrae ossification in Sulf2gt/gt mutants. With loss of increasing numbers of alleles, penetrance and severity of the skeletal malformations is increased and double homozygous mutants are additionally characterized by reduced body weight and size of the skeleton, fusions of individual sternebrae and of tail vertebrae, incomplete closure of the basisphenoid, and dorsally misshaped second cervical vertebra.
Both sulfatases are expressed in overlapping but distinct domains in skeletal tissues of the developing embryo. At midgestation, Sulf1 but not Sulf2, are expressed in the AER and the BA ectoderm and both genes are overlappingly expressed in limb and BA mesoderm. Nevertheless, we did not detect alterations in limb patterning. In contrast, fused mandibles, likely due to disturbed outgrowth of the branchial arches, were detected with low frequencies and loss of one allele of Shh increased the severity of the phenotype resulting in agnathic animals in rare cases.
During endochondral bone formation, we confirmed the expression of both sulfatases in the joint region (Ohto et al.,2002; Nagamine et al.,2005; Zhao et al.,2006) and characterized their dynamic expression in fibrocartilaginous and synovial joints. Both genes are co-expressed between E12.5 to E14.5 in all joints examined. At E16.5, the expression of Sulf1 is strongly reduced, whereas Sulf2 remains highly expressed. In spite of their strong expression in the joint region, we did not detect joint fusion in the limbs of Sulf1;Sulf2 mutants, but observed fusion of sternebrae and tail vertebrae, indicating a different dependence of distinct joints on sulfatase function. In accord with Lum et al. (2007), we detected Sulf2 expression in proliferating chondrocytes after E13.5. In addition, we further demonstrate that hypertrophic chondrocytes express Sulf1 but not Sulf2 and that the newly formed bone expresses both sulfatases. Together the expression pattern strongly supports a role for HS 6-O-sulfation in regulating endochondral ossification.
Regulation of endochondral ossification by HS.
Classical experiments on mouse sternum organ cultures have shown that surgical removal of a rib results in complete fusions of adjacent sternebrae, whereas lateral displacement of the tip of the rib generates partially fused sternebrae (Chen,1953). It is thus likely that one or more factors secreted by the ribs are required to either induce or maintain the fibrocartilaginous joints (Chen,1953; Storm and Kingsley,1996). Both Sulf1 and Sulf2 are expressed in the tip of the rib and in fibrocartilaginous joints and might, therefore, play a role in regulating such signals. As we did not observe morphological differences during early sternum development, sulfatase function seems to modulate maintaining rather than inducing signals of the joint structure.
Analysis of chondrocyte differentiation at a molecular level revealed a reduced region of proliferating chondrocytes between individual sternebrae indicating an accelerated initiation of hypertrophy. Similarly, the zone of proliferating chondrocytes was reduced in the radius of E16.5 Sulf1gt/gt;Sulf2gt/gt mutants. An acceleration of chondrocyte differentiation is likely to result in premature endochondral ossification and, subsequently, in the observed fusion of sternebrae and vertebrae.
FGF signaling is known to be an important regulator of endochondral bone formation. Fgf-receptor 3 (Fgfr3) is expressed in proliferating chondrocytes (Minina et al.,2005) and Fgf18 in the perichondrium surrounding the proliferating cells (Liu et al.,2002). Loss of Fgfr3 or Fgf18 leads to an increased region of proliferating chondrocytes (Liu et al.,2002), whereas constitutive activation of Fgfr3 reduces the region of proliferating cells and accelerates the replacement of hypertrophic chondrocytes by bone (Naski et al.,1998; Minina et al.,2002). Signaling of FGFs to their receptors (FGFR) requires the formation of an FGF-HS-FGFR ternary complex. Previous experiments demonstrated that overexpression of QSulf1 impairs the formation of this complex, thereby decreasing FGF signaling (Wang et al.,2004). Furthermore, cells deficient for Sulf1 and/or Sulf2 are hyper-sensitive to FGF2 stimulation (Lamanna et al.,2006; Holst et al.,2007). An increased sensitivity of chondrocytes for FGF signals in Sulf1;Sulf2 mutants might thus contribute to the observed skeletal phenotypes.
Interestingly, a direct role for HS in regulating endochondral ossification has previously been identified in mouse mutants carrying a hypomorphic allele of Ext1. These mutants produce about 20% of HS and are characterized by an increased region of proliferating chondrocytes due to an increased range of Ihh signaling. Vice versa, treatment of limb explant cultures with HS restricts the range of Ihh signaling and reduces the region of proliferating cells (Koziel et al.,2004). A similar phenotype has been described in mice mutant for the N-deacetylase/N-sulfotransferase 1 (Ndst1−/−) gene. The decreased HS sulfation levels of these mutants (Ledin et al.,2004) leads to an increased zone of proliferating chondrocytes, which has been attributed to either activated BMP-receptor- or Parathyroid hormone related peptide (PTHrP) signaling (Hu et al.,2007). In this respect, it is interesting to note that the expression of Sulf1 and Sulf2 is increased towards the joint region where Ihh regulates the expression of PTHrP, the main negative regulator of the zone of proliferating chondrocytes. Overexpression of QSulf1 has been shown to increase the signaling range of Shh (Danesin et al.,2006), which is structurally highly related to the Ihh protein. The reduced zone of proliferating chondrocytes in Sulf1;Sulf2 mutants might thus be the result of a restricted range of Ihh signaling leading to a reduced expression of PTHrP and/or other yet unknown target genes.
In summary, Sulf1;Sulf2 mutants bear similarities to mice with activated FGF or reduced Ihh signaling indicating a potential dependence of these signaling systems on HS 6-O-sulfation levels. We can, however, not exclude a role of these sulfatases in modulating other signaling systems like that of the BMPs or Wnts, which are important regulators of endochondral bone formation.
HS in patterning the skeleton.
In addition to disturbed endochondral ossification, patterning defects of the axial skeleton might contribute to the vertebrae phenotypes. Due to the complex interactions of many signaling systems during axial patterning, it is difficult to attribute the observed relatively mild defects to a specific signal. It is interesting to note that mutants affecting the synthesis and modification of HS bear similarities in aspects of their skeletal alterations. Comparing such mutants might ultimately help to understand the complex interactions between distinct growth factors and HS. Loss of one allele of Shh in a Sulf1gt/gt;Sulf2gt/gt background for example leads to agnathia with low penetrance, indicating that Shh signaling during mandible outgrowth depends on the HS sulfation pattern. This hypothesis is supported by previous experiments revealing an increased incidence of severe skull defects including agnathia observed in NDST1−/−;Shh+/− mutants compared to single NDST−/− mice (Grobe et al.,2005).
The skeletal alteration of the Sulf1−/−;Sulf2−/− mutant, like small rib cage and short, broadened sternum with fused sternebrae, bear similarities to those seen in Hs2st−/− mutants (Bullock et al.,1998). Analysis of highly sulfated S-domains of HS isolated from Sulf1−/−;Sulf2−/− mutant cells revealed the expected increase in HS 6-O-sulfation and a slight decrease in 2-O-sulfation (Lamanna et al.,2006). In contrast, Hs2st−/− mice compensate loss of 2-O-sulfation by a proportional increase in N- and 6-O-sulfation (Merry et al.,2001). The common aspects of the phenotypes seen in Hs2st−/− and Sulf1−/−;Sulf2−/− mutants might thus result from increased 6-O-sulfation and/or decreased 2-O-sulfation of HS. A detailed analysis of the sulfation pattern of the HS chain will, however, be required to identify common and distinct motives between different mutants. Linking those to specific signaling pathways will ultimately give insight into the molecular origin of the individual phenotypes.
Sulf1 and Sulf2 loss-of-function alleles.
In addition to the hypomorphic gt-alleles analyzed in this study, independent gt-insertions at different positions of both genes have recently been described. The Sulf1VICTR48 allele carries a gt-vector insertion in intron 2 (Holst et al.,2007) downstream of the alternatively spliced exon 2, likely representing a stronger deletion allele as Sulf1gt. For Sulf2, two other genetrap alleles, Sulf2VICTR37 and Sulf2XST155 with integration sites in intron 2 and intron 5, respectively, have been described (Lum et al., 2006; Holst et al.,2007). Both gt-alleles are predicted to translate into a catalytically inactive sulfatase. These alleles seem to differ in the amount of transcribed wild-type RNA, as Sulf2VICTR37 expresses 0.9–9.1% of the wild-type transcript depending on the tissues examined, whereas for the Sulf2XST155 allele no wild-type RNA has been detected in tissue of adult mice (Lum et al., 2006; Holst et al.,2007). In accordance with our study, neither of the single mutations resulted in dramatic phenotypes. Reduced weight and runted animals were observed in Sulf2XST155/XST155 animals, likely due to the stronger deletion efficiency. A high neonatal lethality (18% survivors at P10) has been reported for hypomorphic Sulf1VICTR48;Sulf2VICTR37 mutants (Holst et al.,2007). This was not observed in the hypomorphic Sulf1gt/gt;Sulf2gt/gt mutants used in this study and even Sulf1−/−;Sulf2−/− null mutants showed a higher survival rate (46% at 6 weeks). The high variability in postnatal development might reflect differences in the genetic background in addition to the altered expression levels of the different alleles.
Double homozygous Sulf1VICTR48;Sulf2VICTR37 mutants display skeletal defects similar to those observed in this study including a reduced body weight and skeletal size, fusions of individual sternebrae, a split basisphenoid bone, and a reduced zone of proliferating chondrocytes in the limb (Holst et al.,2007). In addition, we found dorsally split vertebrae, accelerated ossification of the sacral and lumbar vertebrae, and calcified fusion of the tail vertebrae in Sulf1gt/gt;Sulf2gt/gt and Sulf1−/−;Sulf2−/− mutants. We have also started to characterize endochondral ossification at a molecular level and found an accelerated chondrocyte differentiation at least in the sternum. As bone density is not significantly altered in E18.5 Sulf1VICTR48;Sulf2VICTR37 mutants (Holst et al.,2007), the observed alterations might reflect a transient modulation of bone development. Bone density studies have to be carried out at postnatal stages in Sulf1−/−;Sulf2−/− mutants to better understand the role of secreted sulfatases during bone mineralization.
In addition to the intracellularly defined sulfation pattern, our study supports a role for extracellular modifications of HS 6-O-sulfation in organizing growth factor signaling during skeletogenesis. Our investigation of single homozygous mutants revealed subtle differences in the skeletons. Nevertheless, loss of increasing numbers of sulfatase alleles enhances severity and penetrance of the skeletal alterations and leads to malformations at additional sites. Gene dosage seems thus to be more critical for differentiation of the skeleton than the identity of the individual genes expressed. All phenotypic alterations were observed in the axial but not in the appendicular skeleton, indicating different sensitivity for loss of sulfatase function in distinct regions of the skeleton. Interestingly, the skeletal phenotype of Sulf1−/−;Sulf2−/− null mutants is still comparatively mild. Growth factor signaling seems thus not to depend on but rather be modulated by extracellular modifications of HS 6-O-sulfation during endochondral ossification.
Generation Sulf1 and Sulf2 deficient mice
Representative ESTs for the 5′UTR splice variants of Sulf1 (CD348948, BU703701, BY310862, BY321800) and Sulf2 (BM461108, BB645393) as depicted in Figure 3 were downloaded from the NCBI database. 129/Ola derived embryonic stem (ES) cell lines XM190 (Sulf1Gt(pGT0Lxf)XM190Byg) and PST111 (Sulf2Gt(pGT1TMpfs)PST111Byg) containing gene trap (gt) vector insertions in the Sulf1 and Sulf2 locus, respectively, were obtained from BayGenomics (Stryke et al.,2003). Both ES cell lines were injected into C57BL/6 blastocysts. Chimeric founder males were backcrossed into the C57BL/6 background. Analysis was carried out on offspring from N3 to N6 crosses. Mutant mice were genotyped by PCR with primers flanking the respective 5′ integration sites of the gt-vector: for Sulf1 primers Sulf1-wt-F: 5′-GTGGACATGGGTGCACAGATACAC, Sulf1-wt-R: 5′-CTTCATTTCAGGAGTGTGTGGTAGG, and XM190-R: 5′-GACAGTATCGGCCTCAGGAAGATC generate PCR products of 338 bp and 539 bp for the wild-type and mutant allele, respectively. For Sulf2 primers Sulf2-wt-F: 5′-CGCGTGTTAGAGGTCTCGTAGG, Sulf2-wt-R: 5′-GCTGTTCAGGGTGCTTGAGCCA, and PST111-R: 5′-CAGAAGCAGGCCACCCAACTG generate PCR products of 211 bp and 576 bp for the wild-type and mutant allele, respectively. PCR conditions were 94°C for 30 sec, 63°C for 30 sec, and 72°C for 50 sec for a total of 34 cycles. For time pregnancies, noon on the day of the vaginal plug was defined as embryonic (E) day 0.5. Sulf1−/−;Sulf2−/− mutants were maintained on a mixed genetic background (129/Ola;C57BL/6) and genotyped as described (Lamanna et al.,2006). Mice heterozygous for a targeted deletion of Shh (Chiang et al.,1996) were maintained on C57BL/10 background. For genotyping PCR with primers Shh-geno-F2: 5′-CGTTGGCTACCCGTGATATT and Shh-geno-R2: 5′-TAGGAGACAGCCTCGGAAGA amplified a ∼900-bp DNA fragment specific for the mutant allele. PCR conditions were 94°C for 40 sec, 59°C for 40 sec, and 72°C for 50 sec for a total of 35 cycles.
Total RNA was isolated from E14.5 forelimbs and transcribed into cDNA using random hexamers and TaqMan Reverse Transcription Reagents (Applied Biosystems). Quantitative RT-PCR was performed using SYBR Green on a 7900HT Fast Real-Time PCR System with 15 sec 95°C and 1 min 60°C for 40 cycles as recommended by Applied Biosystems. Primers were as follows: RT-Sulf1-Ex12-F: 5′-GATACCAGACAGCCTGTGAGC; RT-Sulf1-Ex13-R: 5′-GAGTAGAGGTTGCGTGCATTC; RT-Sulf2-Ex11-F: 5′-TGGACGCCGTAAGCTCTTTA; RT-Sulf2-Ex12-R: 5′-GGCTGAGGCACAGTATCCAA, RT-Gapdh-F: 5′-GGGAAGCCCATCACCATCTT; RT-Gapdh-R: 5′-CGGCCTCACCCCATTTG.
In Situ Hybridization, Immunohistochemistry, Histology, and Skeletal Preparations
Embryos were fixed overnight in 4% paraformaldehyde, dehydrated and embedded in paraffin. Seven-micrometer sections were stained with alcian blue, hematoxylin and eosin, or used for in situ hybridization. Probes were previously described: Col2a1 and Col10a1 (Minina et al.,2002), and Ptch1 (Koziel et al.,2004) and Fgf8 (Minina et al.,2005). Probes for Sulf1 were PCR amplified from IMAGE clone 4500954 with primers Sulf1-3utr-F: 5′-GACAGTTATGGGATGGATG and T7: 5′-TAATACGACTCACTATAGGGA, and for Sulf2 from IMAGE clone 3155559 with primers Sulf2-3utr-F: 5′-CCAGAAATGAAGAGACCTTC and Sp6: CTATTTAGGTGACACTATAG. IMAGE clones were obtained from RZPD (Berlin). Digoxygenin-UTP labeled riboprobes were generated with T7 or Sp6 RNA polymerases (Roche). Hybridization was carried out as described (Brent et al.,2003). For β-galactosidase staining, embryos were fixed on ice in 1% formaldehyde, 0.2% glutaraldehyde, 0.02% NP40 in PBS for 30 min, and stained at 37°C in 1 mM X-gal in 2 mM MgCl2, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide in PBS. Proliferation was measured on E16.5 radii sections as described (Stricker et al.,2006). The anti-phospho-Histone H3 antibody (Upstate) was applied 1:200 and the Alexa Fluor 546 coupled secondary anti-rabbit antibody (Molecular Probes) 1:500. For cell density calculation, the total number of DAPI (4′,6-Diamidine-2′-phenylindole dihydrochloride, Roche) stained nuclei of two 0.01 mm2 squares were counted in zone I and zone II each (Fig. 6O,P). For cell proliferation, phospho-Histone H3–positive cells in zone I and zone II were counted, the total area measured using ImageJ software, and H3+ cells per 0.01 mm2 calculated (Fig. 6.Q,R). Alcian blue and alizarin red skeletal staining was performed by a modified protocol from McLeod (1980). All measurements were carried out on at least 3 samples as indicated in the tables. Unpaired Student's t-test with a P < 0.05 was considered as statistically significant.
We thank Harald Ehlen, Sigmar Stricker and Manuela Wülling for critical discussion of the manuscript and Martin Vingron for continuous support.