Secreted Sulfatases Sulf1 and Sulf2 Have Overlapping yet Essential Roles in Mouse Neonatal Survival

Background Heparan sulfate proteoglycans (HSPGs) use highly sulfated polysaccharide side-chains to interact with several key growth factors and morphogens, thereby regulating their accessibility and biological activity. Various sulfotransferases and sulfatases with differing specificities control the pattern of HSPG sulfation, which is functionally critical. Among these enzymes in the mouse are two secreted 6-O-endosulfatases, Sulf1 and Sulf2, which modify HSPGs in the extracellular matrix and on the cell surface. The roles of Sulf1 and Sulf2 during normal development are not well understood. Methods/Results To investigate the importance of Sulf1 and Sulf2 for embryonic development, we generated mice genetically deficient in these genes and assessed the phenotypes of the resulting secreted sulfatase-deficient mice. Surprisingly, despite the established crucial role of HSPG interactions during development, neither Sulf1- nor Sulf2-deficient mice showed significant developmental flaws. In contrast, mice deficient in both Sulf1and Sulf2 exhibited highly penetrant neonatal lethality. Loss of viability was associated with multiple, although subtle, developmental defects, including skeletal and renal abnormalities. Conclusions These results show that Sulf1 and Sulf2 play overlapping yet critical roles in mouse development and are redundant and essential for neonatal survival.


INTRODUCTION
Normal developmental processes of metazoans depend upon cellcell communication, which is mediated by a diverse yet highly conserved set of secreted protein factors. Genetic analysis of embryonic development in C. elegans, Drosophila, zebrafish, and mouse has revealed the biological importance of heparan sulfate proteoglycans (HSPGs) in regulating the activity of multiple secreted growth factors and morphogens in various developmental processes [1,2]. Among these factors are the ligands of the fibroblast growth factor (FGF), bone morphogenetic protein (BMP), wingless/WNT, and Hedgehog (Hh) pathways. HSPGs have been shown to regulate cell signaling positively or negatively, depending on pathway and context.
HSPGs such as glypicans, syndecans, agrins, and perlecans are proteins modified with heparan sulfate during their biosynthetic transit through the Golgi apparatus. Following the addition of a tetrasaccharide linker to specified serines in the polypeptide backbone, the exostosin (EXT) enzymes catalyze the addition of unbranched repeats of the glucuronic acid, N-acetylglucosamine disaccharide [3,4]. The uniformity of this simple disaccharide repeat is then diversified by the activity of several other biosynthetic enzymes, including sulfotransferases with differing specificities, N-deacetylase/N-sulfotransferases (NDSTs), and an epimerase, eventually producing a polysaccharide sidechain of astonishing complexity [5,6]. The sulfotransferases catalyze the addition of sulfate to the N, 3-O, and 6-O positions of N-acetylglucosamine and the 2-O position of iduronic acid. The degree and specific pattern of sulfation is determined, in a tissue-specific manner, by the expression of these biosynthetic enzymes [7]. The sulfation pattern of HSPGs is critical for determining the specificity and affinity of binding to growth factors and morphogens [5].
Heparan sulfate modification of proteins is essential for mouse development. Embryos deficient for the Ext1 glycosyltransferase, and hence devoid of nearly all heparan sulfate, die during gastrulation [8]. The relevance of sulfation pattern of HSPGs to normal development has been underscored by the analysis of the neonatal lethality and pleiomorphic phenotypes of Ndst1and Hs2st-deficient mice [9][10][11]. Whereas Ndst1-deficient embryos exhibit cerebral and pulmonary hypoplasia and craniofacial defects, Hs2st-deficient embryos show cerebral and renal hypoplasia accompanied by skeletal and eye defects. Furthermore, inhibition of the different 6-O-sulfotransferase genes has profound effects on muscle and vascular development in zebrafish [12,13].
An additional regulatory level of HSPG sulfation status was recently suggested by the discovery and cloning of a novel class of secreted sulfatase enzymes [14,15]. These proteins, named Sulf1 and Sulf2 in the mouse, can act on HSPGs on the cell surface as well as within the Golgi apparatus, thereby affecting sulfation status ex post facto [15,16]. The hydrolytic activity of these enzymes is specific to the 6-O-sulfate of N-acetylglucosamine [15,17]. Because these sulfatases are found in soluble as well as cell surface-associated forms [14,15], they also may possess non-cell autonomous or paracrine activity.
Given the critical role of regional HSPG sulfation patterns during development, we sought to assess the importance of Sulf1 and Sulf2 during mouse development. To this end, we generated mouse strains with disruption of the Sulf1 or Sulf2 genes. We report here that deficiency in either Sulf1 or Sulf2 alone did not appreciably affect normal developmental or physiological processes. In contrast, disruption of both Sulf1 and Sulf2 simultaneously resulted in high-penetrance perinatal lethality in conjunction with multiple subtle developmental defects. These results indicate that Sulf1 and Sulf2 play overlapping and essential role(s) during development. Our results confirm and expand upon the recently reported phenotypes of sulfatase-deficient mice by Lamanna et al. [17,18], but differ from the phenotypes of the Sulf2-deficient animals described by Lum et al. [19].

MATERIALS AND METHODS
The Institutional Animal Care and Use Committee approved all protocols used in this study.

Sulf1 and Sulf2 targeting
Embryonic stem (ES) cells with retroviral gene trap vectors inserted into the Sulf1 (OST352220) and Sulf2 (OST311938) loci were generated in collaboration with Lexicon Genetics (The Woodlands, TX) using methods described previously [20]. Sulf1 was disrupted by insertion of the VICTR48 construct; this allele is therefore named Sulf1 Gt(VICTR48)352220Lex . Sulf2 was disrupted by the VICTR37 retroviral construct, which included a promoterless b-galactosidase cassette in addition to the Neomycin resistance gene. The Sulf2 allele generated by the insertion mutation is named Sulf2 Gt(VICTR37)311938Lex . The insertion sites were determined by sequence analysis. The Sulf1or Sulf2-targeted 129/ SvEvBrd ES cells were microinjected into blastocysts and implanted into C57BL/6J females. Germline transmission was assessed in F1 pups by PCR analysis of genomic DNA. Heterozygous mice were either intercrossed to assess the phenotypes on a mixed strain background or backcrossed onto the C57BL/6N background. Unless otherwise specified, the results presented herein were performed using the N4 (Sulf2), or N5 or N6 (Sulf1) backcross generations. For simplicity's sake, we refer to the gene trap alleles of the genes as Sulf1 2 and Sulf2 2 and the wildtype alleles as Sulf1 + and Sulf2 + .

Genotyping of Sulf1 and Sulf2 alleles
We utilized a PCR genotyping strategy to detect the wild type and disrupted alleles of Sulf1 and Sulf2. Tail DNA was prepared using the Extract-N-Amp extraction kit following the manufacturer's protocol (Sigma, St. Louis, MO). PCR conditions were as follows: 94uC for 4 min., 1 cycle; 94uC for 60 s, 60uC for 30 s, 72uC for 60 s (30-35 cycles); 72uC for 10 min., 1 cycle. For genotyping of tails from neonate and young adult mice, 30 cycles were sufficient.
We increased the cycle number to 35 when detecting DNA from embryonic tissue.
The retroviral insertion in the Sulf1 locus was detected using primers Sulf1-R1 and LTR, which generated a 334-bp product. We detected the wild type Sulf1 allele using primers Sulf1-F and Sulf1-R1, which produced a 459-bp amplicon. Using a similar strategy, the insertion allele in the Sulf2 locus were detected using primers Sulf2-R and LTR2, producing a 440-bp band; the wildtype Sulf2 allele was amplified with primers Sulf2-F2 and Sulf2-R, making a 759-bp product. The appropriate three-primer sets were used simultaneously without evidence of interference when evaluating both modified genes. Sulf1; Sulf2 mice were genotyped as above, amplifying the Sulf1 and Sulf2 loci in separate reactions.

RNA purification and quantitative RT-PCR
For quantitative real-time RT-PCR (Q-RT-PCR), total RNA was isolated using the RNeasy kit essentially according to the manufacturer's protocol, including the on-column DNase digestion (Qiagen, Valencia, CA). Immediately prior to column purification, frozen tissue was homogenized in RLT buffer containing b-mercaptoethanol using a stator homogenizer followed by a QiaShredder column (Qiagen). To remove all contaminating genomic DNA, we re-purified the eluted RNA over a fresh RNeasy column, including a second DNase treatment. This routinely reduced the amplification of a control reaction omitting reverse transcriptase to below the limit of detection, allowing us to conclude that signal was indeed due to RNA and not contaminating genomic DNA.

Whole-mount embryo RNA in situ hybridization
Embryos were harvested at E8.5-E12.5 (with noon of the day the vaginal plug was observed designated as E0.5) and fixed in freshly prepared 4% paraformaldehyde (PFA) in PBS (pH 7.4) overnight at 4uC. Primers used to amplify the Sulf1 transcript from a mouse embryo cDNA library to TOPO-clone into pCR-II-TOPO (Invitrogen, Carlsbad, CA) were Sulf1-wmISH-S, ACC AGT CAG CCA GAG CGT and Sulf1-wmISH-AS, CTT CCC ATC CAT CCC ATA ACT. Following sequencing to ensure fidelity of the amplification and orientation of insertion, we generated antisense, DIG-labeled probe from linearized plasmid template using the mMESSAGE mMACHINE kit (Ambion, Austin, TX), following the manufacturer's instructions. Hybridization and detection of DIG-labeled probe were done according to standard protocols.

Section RNA in situ hybridization
Following fixation in freshly prepared 4% PFA in PBS (pH 7.4) overnight at 4uC, we equilibrated embryos (E9.5-E12.5) overnight in 30% sucrose in PBS, followed by 1 h in Tissue-Tek O.C.T. (Sakura). Specimens then were embedded in fresh O.C.T. and frozen in an ethanol/dry ice bath. In situ hybridization was carried out on 20-mm transverse sections of wild type embryos. To generate 35 S-radiolabeled RNA probe for in situ hybridization, we used the mRNAlocator kit (Ambion) according to the included instructions. The template for the probe synthesis was a PCR product with a T7 promoter included in the primer, using the following primers: Sulf1-sISH-S, TAT CCG GTG CAA GCA ACA T; Sulf1-sISH-T7-AS, TAA TAC GAC TCA CTA TAG  GGA GGC TGT TCA GAT GCA GGG TTT G; Sulf2-sISH-S,  CCC ACA ACT TCC TCT TCT GC; and Sulf2-sISH-T7-AS,  TAA TAC GAC TCA CTA TAG GGA GGC CCA TAG CTG  TCC CAG TGA T. In situ hybridization using DIG-labeled probes specific to Collagen Type X and Collagen Type II was performed on paraformaldehyde-fixed, paraffin-embedded tissue, as previously described [22].

Skeletal preparations
Near-term (E18.5) fetuses or neonates were killed, flayed, eviscerated, and fixed in 100% ethanol for 1-2 days, followed by an overnight acetone treatment. Specimens were stained in 0.015% alcian blue, 0.005% alizarin red in 5% acetic acid, 60% ethanol for 2-3 days at room temperature; washed briefly in tap water; then cleared in 1% KOH for several days. Cleared specimens were passed through a series of glycerol solutions of increasing concentration prior to visualization on a Zeiss dissecting microscope.

Micro-computed tomography (mCT) image acquisition
The mouse embryos were imaged with a mCT40 (SCANCO Medical, Basserdorf, Switzerland) x-ray micro-computed tomog-raphy system. A sagittal scout image, comparable with a conventional planar x-ray, was obtained to define the start and end point for the axial acquisition of a series of mCT image slices. The location and number of axial images were chosen to provide complete coverage of the embryo. The embryos were imaged with air as the background media. The mCT images were generated by operating the x-ray tube at an energy level of 45 kV, a current of 177 mA and an integration time of 300 milliseconds. Axial images were obtained at an isotropic resolution of 16 mm.

Analysis of mCT images
The relationship between the image intensity values and bone mineral density was assumed to be linear and was obtained by scanning a 97% pure hydroxyapatite (HA) sample (2.91 gHA/ cm 3 ). Three-dimensional (3D) surface renderings were created from the mCT data with the use of Analyze (AnalyzeDirect Inc., Lenexa, KS), an image analysis software package. The bone surface renderings were generated by applying a bone mineral density of threshold of 0.54 gHA/cm 3 . The bones of interest (femur, humerus, supraoccipital, interparietal, parietal, frontal, basisphenoid and basioccipital) were extracted for quantitative analysis by the applying a series of image analysis steps (thresholding, morphological filtering and region growing operations) to the volumetric image data. For the long bones, the distance between the ends of the bone was measured as a straight line and an average was computed using the contralateral bone measurements. The volume (mm 3 ) and mean bone density (gHA/cm 3 ) of the defined bones was measured and an average was obtained for the ipsilateral and contralateral bones for each embryo.

Necropsy and histological preparation of conceptuses and neonates
Fetuses (E15.5 and E18.5) were removed from the uteri of anesthetized dams. The tip of each fetus's tail was harvested for genotyping by PCR, after which the fetus and placenta were fixed by immersion in either 4% PFA, 10% NBF, or Bouin's solution. Some fetuses were bisected longitudinally just to the left of the midline, while others were decapitated so that the heads could be cut coronally. A few neonates (including placentas) were handled in a similar manner. Fixed tissues were processed into paraffin using a conventional protocol. Serial 5-mm-thick coronal sections of each head block were acquired at several steps, each separated by a distance of 250 mm. Serial 5-mm-thick sections of each torso block were acquired at several steps, each separated by a distance of 500 mm. Serial sections at each step were stained with hematoxylin and eosin (H&E) or cresyl violet (to better delineate subcellular characteristics of neural cells). Lesions were scored (with foreknowledge of the fetus's genotype) using a tiered, semiquantitative grading scale: absent, minimal, mild, moderate, marked. Subsequently, a coded (''blinded'') histopathological assessment for a given organ was performed by inverting all slides and then sorting them into groups using criteria established in the uncoded examination.
Mouse embryonic fibroblast isolation E12.5 embryos were dissected from the extraembryonic tissues in PBS, then the liver and head were removed. After mincing with scissors, the remaining tissue was digested in 0.1% trypsin/EDTA in PBS for 20 min at 37uC, with mixing every 5 min. After washing twice with culture medium (DMEM with 10% fetal bovine serum (FBS), 2 mM L-glutamine, and penicillin/streptomycin), the cells from a single embryo were plated in a 25-cm 2 flask. One day later, cells were trypsinized, counted, and either frozen in aliquots or used immediately for signaling experiments.

FGF signaling in embryonic fibroblasts
First-or second-passage fibroblasts were serum-starved overnight in assay medium containing 0.5% FBS at a density of 10 5 cells per well of a 6-well plate. The following day, prewarmed FGF1-or FGF2-containing assay medium (or assay medium lacking FGF) was added to the cells, which were then returned to the 37uC incubator for 10 minutes. Cells were lysed for 15 minutes at 4uC, with rocking, in 16RIPA buffer (Upstate) supplemented with Complete protease inhibitors (Roche), 2 mM sodium orthovanadate, and 5 mM sodium fluoride. Lysates were cleared by centrifugation (10 minutes at 11,000 g), then prepared for electrophoresis using LDS buffer and DTT (Invitrogen). Equal amounts of protein were loaded per lane of a 4-12% Bis-Tris NuPAGE gel and then separated according to the manufacturer's recommended protocol (Invitrogen). After transferring the protein to nitrocellulose membranes, non-specific binding was blocked using 5% milk in Tris-buffered saline solution containing 0.2% Tween-20 (TBST). Primary antibodies were incubated overnight at 4uC in 1% BSA in TBST (Pierce). Peroxidase-conjugated secondary antibodies were used for 1 hour at room temperature. Peroxidase activity was detected using ECL Plus (GE Healthcare). Antibodies and dilutions used: Phospho-FRS2, 1:1000 (Upstate); Total FRS2, 1:1000 (Upstate); anti-rabbit-HRP, 1:10,000 (Amersham). Bands were quantitated using the integrated density function of ImageJ. Data shown are representative of three independent experiments.

Statistics
Histopathologic (ordinal) data from the blinded microscopic evaluation were assessed using the non-parametric chi-square test. Nominal data were analyzed by two-tailed Student's t-test. All calculations were done using commercial JMP statistical software (v. 5.0; SAS Institute, Cary, NC), and significance assigned to pvalues #0.05.

Targeted disruption of Sulf1
An ES cell clone with a retroviral insertion in the intron between exons 2 and 3 of the Sulf1 gene (using the nomenclature and gene structure information provided in Morimoto-Tomito et al.) [15] was identified by screening a gene trap ES cell library (Fig. 1A). Chimeric mice carrying this targeted Sulf1 allele were generated using standard techniques. Upon intercrossing the first F1 generation of Sulf1 +/2 mice in the mixed genetic background, we observed no deviation from expected Mendelian birth ratios (Table 1). Both male and female homozygous mutant (Sulf1 2/2 ) mice were capable of generating offspring, indicating no obvious fertility defect (data not shown). Similar viability and fertility results were obtained with Sulf1 2/2 mice after backcrossing the mutant allele 5 times onto the C57BL/6N strain (N5). Sulf1 2/2 mice were therefore viable and fertile in both a mixed 129/SvEvBrd; C57BL/6 background and a purer C57BL/6 genetic background.
As the retroviral insertion site is located in an intron upstream of the exon containing the ATG start codon of Sulf1 (Fig. 1A), we suspected the allele might be hypomorphic rather than null. To confirm whether the retroviral insertion indeed resulted in a lossor reduction-of-function allele, we performed quantitative RT-PCR (Q-RT-PCR) using primers designed to amplify the 39 UTR of Sulf1. We found that the retroviral insertion resulted in a severe hypomorphic allele in nearly all tissues examined. Whereas heterozygous animals showed the expected ,50% decrease in Sulf1 expression, homozygotes had between 0.2% and 1.7% of Sulf1 mRNA levels relative to wild-type control tissues (Fig. 1C). Interestingly, we observed no substantial decrease in expression of the Sulf1 message in RNA derived from the testis (Fig. 1C). It has been shown previously that testis tissue produces a truncated Sulf1 transcript [23], consistent with the use of an alternative promoter that might be unaffected by the presence of the genetrap retroviral insertion. Finally, we did not observe any significant change in Sulf2 expression in the tissues derived from the Sulf1 2/2 mice (Fig. 1D).
The hypomorphic nature of the Sulf1 2 allele was further verified by immunohistochemistry. Sections of wild-type (WT) and Sulf1 2/2 embryonic tissue were stained with a Sulf1-specific rabbit polyclonal antibody. Compared to WT embryos, Sulf1 2/2 embryos exhibited a profound loss of expression of the Sulf1 protein in the floor plate of the neural tube, a particularly prominent site of Sulf1 expression (Fig.  S1). Along with the quantitative RT-PCR results (Fig. 1C), these data confirm that the Sulf1 genetrap allele is a severe hypomorph, at the mRNA and protein level.

Embryonic and adult Sulf1 expression
We assessed the expression of Sulf1 during development by in situ hybridization (ISH) in whole mounts and tissue sections of embryos from stages E9.5 to E12.5 ( Fig. S2 and data not shown). Reminiscent of the expression pattern of the rat Sulf1 homolog (also known as RSulfFP1) [24], we detected mouse Sulf1 expression in the floor plate of the midbrain, hindbrain, and neural tube, the choroid plexus, the clefts of the branchial arches, dorsal telencephalon, somites, eyes, palate, tongue, nasal pits, the forelimb arm girdle, and the condensing mesenchyme in the distal limb buds (Fig. S2). ISH on sagittal sections of E15 embryos and E17 fetuses revealed continued expression in the choroid plexus and the cartilaginous anlagen of the digits, as well as scattered expression in the lungs, gastrointestinal tract, brown adipose tissue, salivary gland, and the esophageal lining (data not shown).
Section-based ISH of a panel of adult tissues revealed expression in the photoreceptor layer of the retina, the Harderian gland, blood vessels and some tubules in kidney, blood vessels in the lungs, submucosa of the stomach, and intestinal mucosal epithelium (data not shown). Microarray analysis of an extensive panel of normal adult mouse tissues revealed detectable expression of Sulf1 in nearly all sites examined, with uterus, bladder, mesenteric lymph nodes, prostate, esophagus, lung, mammary gland, and bone showing particularly strong expression (Fig. S3).

Targeted disruption of Sulf2
From a random insertion ES cell library, a clone was isolated with an insertion between exon 2 (which contains the ATG start codon) and exon 3 of the Sulf2 gene (Fig. 1B). Heterozygous chimeric mice were isolated and bred to generate Sulf2 mutant animals. In contrast to a recently published report [19], we observed no deviation from the expected Mendelian birth ratios of homozygous mutant animals, neither in the original mixed 129/SvEvBrd; C57BL/6 background nor in mice at the N4 generation of backcrossing into the C57BL/6N strain (Table 1). We also observed that Sulf2 2/2 males and females were fertile (data not shown), indicating that Sulf2 is not necessary for reproduction.
We verified that the retroviral insertion into the Sulf2 locus also resulted in a hypomorphic allele by Q-RT-PCR. Similar to our results in the Sulf1-disrupted mice, we found ,50% reduction in Sulf2 mRNA levels in tissues derived from Sulf2 heterozygotes but a marked decrease (0.9-9.1% of WT levels, depending on the tissue examined) in Sulf2 expression in tissue from Sulf2 2/2 , and c correspond to Sulf2-F2, Sulf2-R, and LTR2, respectively, and were used to genotype the mice. Primers d and e in the 39 untranslated region refer to Sulf2-3364F and Sulf2-3428R, which were used for Q-RT-PCR amplification of the Sulf2 transcript. (C and D) Q-RT-PCR analysis of Sulf1 (C) and Sulf2 (D) in tissues from mice with different Sulf1 genotypes (white bars, Sulf1 +/+ ; grey bars, Sulf1 +/-; black bars, Sulf1 -/-). The Cycle to Threshold (Ct) values for duplicate or triplicate samples were averaged. Each bar represents a single animal. After normalization to GAPDH or RPL19, relative Sulf1 expression levels (compared to an average of the two WT littermate controls analyzed) were calculated based upon the 2 -DDCt method [21]. MG refers to mammary gland. (E and F) Q-RT-PCR analysis of Sulf1 (E) and Sulf2 (F) expression in tissues from Sulf2 +/+ , Sulf2 +/-, or Sulf2 -/adult mice. Analysis was done as described above for the Sulf1 mice. Abbreviation used: MG refers to mammary gland tissue. doi:10.1371/journal.pone.0000575.g001 homozygotes (Fig. 1F). We did not observe a compensatory change of Sulf1 expression in Sulf2 mutants (Fig. 1E).

Embryonic and adult Sulf2 expression
The retroviral gene trap cassette used to disrupt Sulf2 contained a bgalactosidase reporter, which we used to assess the Sulf2 expression pattern during development. When detecting b-galactosidase activity in a litter of embryos from a Sulf2 +/2 intercross, the intensity of staining corresponded to the genotype status as determined by PCR of genomic DNA (Fig. S4). Sulf2 is expressed in a dynamic pattern at all developmental stages tested (E10.5-E12.5; Figs. S4 and S5). Sectioning the stained embryos, we detected b-galactosidase activity in the floor plate of the hindbrain, floorplate and ventral progenitor region of the neural tube, choroid plexus, roof plate of the forebrain and midbrain, the developing retina, mesenchyme of the head and branchial arches, dorsal root ganglia, somites (dermomyotome), forelimb buds (in a posterior-toanterior gradient), hindlimbs (more uniform intensity), the caudal notochord, and the right ventricle of the heart (data not shown).
ISH of transverse sections through the forelimb level of E11.5-E12.5 embryos confirmed the veracity of the b-galactosidase staining, but also revealed additional expression in the condensing sclerotome ventral to the neural tube; mesenchyme surrounding the dorsal aorta, sinus venosus, and left common cardinal vein; the lining of the esophagus; mesenchymal cells surrounding the bronchi in the lung buds; palate and tongue; and condensing mesenchyme surrounding the presumptive digits in the distal limbuds (Fig. S4).
We also detected b-galactosidase activity by gross observation in a panel of normal tissues from an adult Sulf2 2/2 animal. Activity was undetectable in muscle and spleen. Limited expression was detected in the lung (limited to the cells lining the large bronchial airways), stomach (cells on serosa of stomach, particularly in the pyloric region), and uterus (subset of cells in both the myometrium and endometrium). Abundant activity was also revealed in the ovary and testis (data not shown). Consultation of a recently published microarray analysis of an extensive panel of normal mouse tissues [25] revealed detectable expression of Sulf2 in a wide variety of tissues examined, with ovary, umbilical cord, uterus, adipose, bladder, and intestine showing particularly strong expression (Fig.  S5). Sulf2 was also expressed in a variety of brain structures, including the hippocampus, cerebral cortex, and hypothalamus (Fig. S5).
Absence of major abnormality in Sulf1 2/2 or Sulf2 2/2 mice Neither Sulf1 2/2 nor Sulf2 2/2 mice had any defect in long-term survival, with both males and females living to .1.5 years of age. As described above and previously, Sulf1 and Sulf2 are expressed in a variety of embryonic and adult tissues ( Fig. 1; Figs. S1, S2, S3, S4, S5) [19,24,26]. We therefore cast a wide net in searching for phenotype(s) in the Sulf1 2/2 and Sulf2 2/2 mice. We observed no gross differences in appearance or behavior between wild-type, heterozygous and single mutant mice. Aged adult mice were subjected to necropsy, at which time we performed gross and histological analysis on a subset of organs, including lung, liver, stomach, small intestine, colon, brain, eye, spleen, kidney, and heart. Apart from common age-related changes, which were observed in both control and sulfatase-deficient animals, no notable gross or histological changes were found in the single mutants (data not shown). We conclude that neither the Sulf1nor the Sulf2-deficient animals exhibited any obvious gross or histological abnormalities.
Given that the single mutant mice exhibited very mild or no phenotype, we postulated that the two secreted sulfatases perform redundant function(s) in vivo. To examine this hypothesis, we crossed the Sulf1 and Sulf2 mutant mice to generate compound heterozygotes. Sulf1 +/2 Sulf2 +/2 siblings were then intercrossed, with the expectation that 1/16 th of their offspring would have the Sulf1 2/ 2 Sulf2 2/2 genotype. Interestingly, genotyping at postnatal day 9.463.7 (mean6SD), we observed only ,18% of the expected number of double mutants (11 expected, 2 observed; Table 2). This reduction in viability demonstrates that the two secreted sulfatases perform redundant and essential function(s) in vivo.
We sought to determine at what stage (embryonic or post-natal) the double mutants were perishing. We first intercrossed Sulf1 +/2 Sulf2 2/2 and Sulf1 2/2 Sulf2 +/2 mice and isolated embryos at E18.5, which was typically the day of birth in this strain. At E18.5,  we observed as many double-mutant embryos as would have been predicted by Mendelian inheritance (,25%, Table 2). The Sulf1 2/2 Sulf2 2/2 offspring, however, had a statistically significant ,19% decrease in body weight and a ,12% decrease in crown-rump length ( Fig. 2B and data not shown). Neither Sulf1 2/2 Sulf2 +/2 nor Sulf1 +/2 Sulf2 2/2 single mutants showed any deviation in birth weight from control littermates, thus providing direct evidence for a redundant role for the sulfatases during embryonic development. Although the Sulf1 2/2 Sulf2 2/2 embryos were smaller than control littermates, there were no obvious gross changes in morphology. Overt development of the limbs (including digits), head (including palate), tail, and torso all appeared normal. Similar observations of apparently normally developing Sulf1 2/2 Sulf2 2/2 embryos were also made in multiple litters at E10.5, E11.5, E12.5, E13.5, and E15.5. A thorough histological examination of sectioned E15.5 embryos revealed no divergence of Sulf1 2/2 Sulf2 2/2 from control embryos (data not shown). This indicates that embryonic defects in the double mutants arise late in development.
Dissection of E18.5 double mutant embryos showed that all major organs were present and, apart from a slight decrease in size of most organs, appeared normal morphologically. The one clear exception were the kidneys from double mutant neonates, which were decreased in volume by ,45% relative to kidneys from Sulf1 +/2 Sulf2 +/2 control littermates (Fig. 2E). The small kidneys of the Sulf1 2/2 Sulf2 2/2 embryos were either normal histologically or, in a few cases, had a modest decrease in the thicknesses of cortex and medulla (compare Fig. 3A with 3B and 3C). All other internal organs examined appeared histologically normal (data not shown).
We then observed the neonates shortly after birth to determine a potential cause of death. Genotyping at postnatal day 1-2, we saw approximately a 50% reduction in expected viability between birth and P1-P2 (Table 2). Of 16 double mutant neonates genotyped at P1-P3, 7 (44%) died before weaning, with most of those dying within the first week of life (data not shown). Of 17 double mutants that survived to weaning, 4 (24%) died within 10 days of weaning. In summary, we observed full viability of Sulf1 2/2 Sulf2 2/2 double mutants at E18.5, ,48% survival at P1-P2, and approximately 18% survival at weaning (Table 4), with roughly a quarter of those pups that survive to weaning dying shortly thereafter (two due to severe malocclusion, others for unknown reasons). The long-term survival frequency of Sulf1 2/2 Sulf2 2/2 mice is therefore ,15%.
Consistent with the observed low birth weight, those double mutants surviving to P1-2 showed ,25% reduction in body weight (data not shown). Although the Sulf1 2/2 Sulf2 2/2 neonates had a suitable skin color (pink, indicating sufficient oxygenation) and apparently normal motility, they frequently did not have milk in their stomachs, suggesting that the cause of early neonatal death may have been related to undernourishment or dehydration. Histological analysis of placenta, kidney, stomach, intestine, pancreas, liver, and heart from double mutant neonates revealed no obvious abnormalities explaining the inability to suckle (data not shown). Thus, the cause of neonatal mortality remains undetermined. However, a likely premise is that double knockout neonates had neurological problems (suggested by the extensive Sulf1 and Sulf2 gene expression patterns in the embryonic central nervous system). More thorough neuroanatomic and neurobehavioral assays are ongoing to further investigate the cause of neonatal lethality.
Surviving Sulf1 2/2 Sulf2 2/2 adults are small, exhibit subtle skeletal and renal defects, but are otherwise normal in gross and histological structure We were able to analyze a total of 10 adult (5 male and 5 female) Sulf1 2/2 Sulf2 2/2 animals. All animals showed below-average weight, with one representative cohort of ,14-15 week-old double mutant males (n = 5) having an average weight 26% less than agematched littermate control males (n = 10) (Fig. 2D). Another cohort of three Sulf1 2/2 Sulf2 2/2 and two control females (matched at 11 weeks of age) also exhibited a decreased body weight, with double mutants weighing 29% less than controls (13.8 g vs. 19.5 g, respectively; data not shown).
We collected and weighed a full panel of organs from the adult Sulf1 2/2 Sulf2 2/2 and control mice. When corrected for the aforementioned decrease in body weight, all the organs appeared to be of generally normal size and gross appearance except for the kidneys, which were disproportionately small. For instance, in the male cohort described above, the kidneys of double mutant animals showed a 44% decrease in weight relative to controls (Fig. 2F), which is substantially greater than the 26% discrepancy in total body weight. Despite the decreased kidney size, renal function did not appear to be altered in the double mutants, with normal levels of serum and urine markers testing kidney function (data not shown). This is consistent with the observation that the kidneys of double mutant adults were histologically normal (data not shown). Whereas the Sulf1 2/2 single mutant mice were similar in body and kidney weight to age-matched control adult males, Sulf2 2/2 mice, like the Sulf1; Sulf2 double mutants, had decreased body weight (20% decrease) and kidney weight (29% decrease). The degree of decrease in relative kidney and body weight was less in the Sulf2 single mutants than in the Sulf1; Sulf2 double mutants, indicating that although Sulf2 appears to be the more important of the two sulfatase genes in regulating kidney development, Sulf1 also contributes to kidney growth. Therefore, Sulf1 and Sulf2 appear to play redundant roles in the regulation of body and kidney weight.
Histological analysis of a panel of organs collected from the adult double mutant animals revealed all tissues examined to be histologically normal, with two exceptions. First, one 11-week-old double mutant female had interstitial pneumonia with syncytial cell formation localized to the peribronchiolar region, suggesting the possibility of a latent phenotype related to immunological dysfunction. A survey of immunological cell types in the peripheral blood and spleen showed no obvious defect in immune cell development of the double mutant adults (data not shown). However, more specific analyses might yet reveal a role of the sulfatases in immune function. Second, all of the Sulf1 2/2 Sulf2 2/2 adults examined showed retention of cartilage cores in the cortical bone of the tibial metaphysis, sometimes in conjunction with trabecular disorganization (compare Fig. 3N with 3O and 3P). In summary, surviving Sulf1 2/2 Sulf2 2/2 adults are small in stature, but show apparently normal behavior and have grossly and histologically normal tissues, with the lone exception of the subtle changes observed in the bones.

Sulf1 2/2 Sulf2 2/2 embryos show defects in bone development
Given our observation of a bone defect in the Sulf1 2/2 Sulf2 2/2 adults (Fig. 3O, P) and the expression of Sulf1 and Sulf2 at several sites in bone during embryogenesis (Fig. S2 and S4 and data not shown), we analyzed bone development in the Sulf1 2/2 Sulf2 2/2 embryos. To this end, we stained dehydrated E18.5 embryos with alcian blue and alizarin red to assess bone and cartilage formation. In general, the skeletons of double mutants were similar to those of control littermates, though notably smaller in stature (Fig. 4A, B). One highly penetrant difference (12 of 14 (86%) Sulf1 2/2 Sulf2 2/2 E18.5 embryos examined) was the asymmetry and/or fusion of sternebrae ( Fig. 4E-H, Table 4). Only 1 of 19 (,5%) control embryos showed this anomaly. High penetrance of this defect required both Sulf1 and Sulf2 deficiency, in that only one of five Sulf1 2/2 and none of three Sulf2 2/2 embryos exhibited sternebral defects (data not shown).
Histological examination of E18.5 embryos focusing on skeletal development revealed two other subtle but highly penetrant alterations in Sulf1and Sulf2-deficient embryos. These changes included (i) non-fusion of the posterior basisphenoid bone (Figs. 3K, L, and M; Fig. 4D; Table 4), in one case with the pituitary gland extending ventrally into the gap (arrow, Fig. 4L); and (ii) aberrant fusion of the basioccipital and basisphenoid bones indicated by widening and bulging of the cartilage interface between these two bones (Fig. 3I, Table 4). The increased prevalence of each of these lesions in Sulf1 2/2 Sulf2 2/2 embryos as compared to control embryos was statistically significant (p,0.02 by Chi square method) using blinded analysis or, when appropriate, direct slide-by-slide comparison.
Seeking a more quantitative analysis of the skeletal phenotype of the Sulf1 2/2 Sulf2 2/2 embryos, we analyzed E18.5 embryonic skeletons by micro-computed tomography (mCT). Assessing the ossified bone volume of several different bones in the skeleton, we found that the double mutants exhibited a decrease in bone volume (Tables 3 and 4) and length (data not shown) of the humerus and femur. Likewise, the interparietal, basisphenoid, basioccipital, and supraoccipital bones of the skull also had decreased volumes (usually ,15% smaller than bones from control embryos, Table 3), though the decreased volume of the supraoccipital bone was observed with only 50% penetrance (4 of 8 embryos analyzed had a mean volume of less than 2 SD below the mean of age-matched controls). In contrast, the volume of the frontal and parietal bones of the skull showed no difference between double mutants and controls. In summary, all bones examined except the frontal and parietal bones of the skull were ,15-20% smaller in the Sulf1 2/2 Sulf2 2/2 embryos as compared to Sulf1 +/2 Sulf2 +/2 littermate controls (Table 3). Interestingly, none of the bones from Sulf1 2/2 Sulf2 2/2 embryos showed a change in average bone density as compared to control littermates (data not shown), indicating that the ossified regions of the double mutant embryos' bones were smaller but not less dense.

DISCUSSION Sulf1 and Sulf2 perform redundant and essential functions during embryonic and neonatal development
We describe here the phenotype of mice engineered to contain retroviral gene trap alleles inserted into the Sulf1 and/or Sulf2 loci.
We showed that the Sulf1 and Sulf2 gene trap alleles are profoundly hypomorphic, with no compensatory changes in paralog gene expression (Fig. 1). Mice deficient for either Sulf1 or Sulf2 alone were viable, fertile, and normal by nearly all criteria examined (Table 1, Fig. 2). Sulf2-deficient adult mice did show a tendency towards decreased body weight relative to control animals, but this tendency was not accompanied by any obvious functional or structural changes. In contrast to a recent report [19], we observed no deviation from the expected Mendelian birth ratios of homozygous Sulf2 mutant animals, either in our original mixed 129/SvEvBrd; C57BL/6 background or in mice at the N4 generation of backcrossing into the C57BL/6N strain (Table 1). We also did not observe the previously described runting or lung defects ( Fig. 2 and data not shown) [19]. Perhaps the discrepancies between our results and those of Lum et al. [19] were due to the difference in the 129 genetic strain background used (129/ SvEvBrd in this study vs. 129P2/Ola in Lum et al.): the previouslyreported reduction in decreased Sulf2 2/2 viability was only seen in the 129P2/Ola; C57BL/6 mixed background and not in two other more homogeneous backgrounds (C57BL/6 [N3 and N7] and FVB/N [N1 and N7]). Alternatively, the degree of hypomorphism and/or the nature of any aberrant transcripts generated by the different gene trap alleles used in our and the previous study might have affected the phenotypic outcome. In either case, the effect(s) of Sulf2 deficiency are clearly quite subtle.
We also showed that, as remarked upon briefly by Lamanna et al. [17,18], Sulf1 and Sulf2 are functionally redundant genes essential for viability. We report here the first characterization of the penetrance and timing of perinatal lethality in mice deficient for Sulf1 and Sulf2, as well as the phenotype of the double mutant embryos and neonates. Though Sulf1 and Sulf2 are expressed in a wide variety of embryonic and adult tissues ( Fig. 1; Figs. S1, S2, S3, S4, S5), the development of secreted sulfatase-deficient mice proceeds with remarkably normal integrity, with the sole developmental alterations observed to be decreased body size accompanied by subtle kidney and bone defects. The functions of the secreted sulfatases, though obviously important as evidenced by the neonatal lethality in their absence, are likely to be understated, perhaps via refining or restricting the extent of some other essential signaling pathways.
The bone defects observed in the Sulf1 2/2 Sulf2 2/2 embryos are remarkable in several ways. First, there does appear to be a deficit in bone growth in the double mutants, in that bones located in both the axial and appendicular skeleton, which form by the distinct processes of endochondral and intramembranous ossification, showed a decrease in ossified bone volume (Tables 3  and 4). Second, although the secreted sulfatase-deficient skeletons appeared overall to be patterned normally ( Fig. 4A and B), we did observe several highly specific morphological defects at the midline, including sternal fusion and defective basisphenoid patterning. Interestingly, sternal defects such as those exhibited by the Sulf1 2/2 Sulf2 2/2 embryos have been reported not only in the Hs2st-deficient mice (see below), but also in mice with compound mutations in different Bmp family members [33], EphrinB1 mutants [34], and in a mice engineered to contain extra copies of a hypermorphic Fgfr1 allele or transgenic expression of hypermorphic Fgfr3 allele [30,31]. Perhaps this convergence in phenotypes indicates a role for the secreted sulfatases in the regulation of Bmp, Fgf, and/or Ephrin signaling during sternal development. Our demonstration that Sulf1 2/2 Sulf2 2/2 fibroblasts are hypersensitive to FGF signaling (Fig. 5) supports this supposition. Third, whereas nearly all joints examined in the Sulf1 2/2 Sulf2 2/2 double mutants appeared grossly and histologically normal (Fig. 4), two joints in the base of the skull had Frequency defined as the number expected (by standard Mendelian segregation, see Table 2) divided by the number observed; n defined as total number of animals analyzed in a given age range 2 Frequency defined as the percentage of Sulf1 2/2 Sulf2 2/2 embryos that showed the described phenotype; n was the number of Sulf1 2/2 Sulf2 2/2 embryos analyzed 3 Frequency defined as the percentage of Sulf1 2/2 Sulf2 2/2 animals whose mean volume for the kidney or indicated bone was less than the smallest of the control animals' kidney or bone volumes; n as the number of Sulf1  notable, yet distinct defects (a thickening of the basioccipital/ basisphenoid suture and a failure of the posterior basisphenoid bone to fuse at the midline). A similar basisphenoid defect has been reported in Bmp7and Cdon-null embryos [35,36]. This phenocopy congruence supports the possible role of secreted sulfatases in regulating Bmp and Hedgehog signaling. Finally, it is worth noting several phenotypes that were not observed in the Sulf1 2/2 Sulf2 2/2 embryos; in particular, we did not find skull vault suture defects, cleft palate, appendicular joint fusions, or polydactyly. All of these defects have been associated with altered Fgf, Bmp, or Hh signaling. An interesting implication of this discrepancy is that the sulfatases might regulate heparan sulfatedependent growth factor signaling at some but not all developing bone sites. Further work will be required to confirm this possibility. The exact cause of neonatal lethality remains unclear. Even if the observed decrease in kidney size were to result in altered renal function (a prediction not supported by the apparently normal kidney function we measured in the Sulf1 2/2 Sulf2 2/2 adults that had disproportionately small kidneys), one would not expect such a precipitous decline in neonatal survival. The normal structure of thoracic viscera and their branches (data not shown) suggested that altered cardiovascular or pulmonary function-a common etiology for perinatal mortality-was also not the cause of lethality in this case. Likewise, the digestive organs and skeleton did not exhibit substantive anatomic abnormalities (data not shown). The frequent absence of milk spots in double knockout neonates suggested that an absence or reduction in suckling, leading ultimately to dehydration, might have been a cause of mortality in the double mutants. A potentially deficient feeding response might in turn be due to underlying neurological defects, as suggested by examination of other genes expressed in the central nervous system [37,38]. We are currently exploring this possible pathogenesis as an explanation for the lethality observed in Sulf1 2/2 Sulf2 2/2 neonates.
The phenotype of the surviving Sulf1 2/2 Sulf2 2/2 mice is similar to that of Hs2st mutant mice Several recent studies have revealed an unexpected compensatory link between the 6-O and 2-O sulfation states of heparan sulfate, in both vertebrate (mouse) and invertebrate (Drosophila) systems [17,39,40]. Specifically, tissues from mice with a gene trap disruption of the single 2-O-sulfotransferase (Hs2st) gene exhibit not only the expected decrease in 2-O-sulfated iduronic acid, but also an apparently compensatory increase in 6-O sulfation of Nsulfoglucosamine [40]. Lamanna et al. recently reported that embryonic fibroblasts derived from mice devoid of both the Sulf1 and Sulf2 secreted sulfatases had not only the expected increase in 6-O sulfation of glucosamine, but also a decrease in 2-O sulfated iduronic acid [17], the same trends as reported for the Hs2stdeficient mice.
Mice deficient for Hs2st exhibit defective skeletal development, bilateral kidney agenesis, cleft palate, decreased cerebral cortex thickness, polydactyly, and iris coloboma. In addition, Hs2stdeficent mice are perinatal lethal with full penetrance [9,41]. The Sulf1; Sulf2-deficient mice, as reported here, showed phenotypic similarity to the Hs2st-deficient mice, albeit of lesser severity. Specifically, both mutant strains of mice exhibited sternal fusions, defects in kidney development, and perinatal lethality, indicating that Hs2st and the secreted sulfatases might regulate similar developmental processes. In contrast, Sulf1 2/2 Sulf2 2/2 mice did not show cleft palate, cerebral cortex diminution, digit duplications, or eye defects, despite the fact that Sulf1 and Sulf2 are expressed in all of these tissues (Figs. S2 and S4 and data not shown). The differences between the phenotypes of the two mice might indicate (i) that the sulfatases are not expressed at the correct time and/or place to play a role in all of the same developmental processes as Hs2st, (ii) that the sulfation status changes of the sulfotransferase-and sulfatase-deficient tissues differ in some way not revealed by the previous studies, or (iii) that the effect of deficiency of an intracellular biosynthetic enzyme, Hs2st, might be more profound than the deficiency of secreted enzymes, Sulf1 and Sulf2, that act ex post facto.
The subtlety of the Sulf1; Sulf2 double deficient phenotype would not have been predicted based upon previous in vitro experiments Based almost entirely upon over-expression studies in cell lines grown in vitro, the secreted sulfatases have been shown to modulate a wide variety of heparin-binding growth factor signaling pathways, including WNT, FGF, HB-EGF, HGF, and BMP [16,23,[42][43][44][45]. Except for the cases of WNT1 and BMP4, wherein elevated expression of the sulfatases promotes cell signaling, the literature is in agreement that forced (over)expression of Sulf1 strongly inhibits signaling by various heparin-binding growth factors (including FGF2, HB-EGF, and HGF). Furthermore, recombinant Sulf2 enzyme has been shown to release heparan sulfate-and heparin-bound signaling proteins such as VEGF 165 , FGF1, and SDF1 [46], conferring another potential set of non-cell autonomous functions to the secreted sulfatases.
Given that most of the aforementioned signaling pathways are critical for certain developmental processes and patterning events, one might have predicted from the in vitro studies that secreted sulfatase-deficient mice would have exhibited profound developmental defects in a multitude of organ systems. This prediction was not borne out by the present organismal loss-of-function study. On the contrary, we show here that mice with disrupted Sulf1 and Sulf2 genes are by and large normal, with a few subtle patterning and growth defects. As evidenced by the high penetrance neonatal lethality, clearly the secreted sulfatases are essential for some process critical for life. Perhaps these enzymes are involved in finetuning rather than directing the spatial distribution of heparan sulfate-binding factors.

Conclusion
In summary, we have shown that the Sulf1 and Sulf2 secreted 6-Oendosulfatase genes are redundant and essential for post-natal survival in developing mice. Whereas animals deficient for either Sulf1 or Sulf2 alone have normal birth weight, full viability, and histologically normal organs, mice simultaneously deficient for both Sulf1 and Sulf2 exhibit low birth weight, ,50% neonatal lethality, ,80% lethality as of weaning, failure to thrive, and a variety of skeletal and renal defects that arise during late embryogenesis. Figure S1 Loss of Sulf1 protein expression in Sulf1 2/2 embryos. Immunohistochemistry using a Sulf1-specific affinity purified polyclonal antibody was performed on frozen sections of WT (A, C) or Sulf1 2/2 (B, D) embryos. Note the loss of Sulf1 expression in the floorplates of Sulf1 2/2 embryos. Floor plate cells are outlined in yellow to facilitate visual comparison. Found at: doi:10.1371/journal.pone.0000575.s001 (5.07 MB TIF) Figure S2 Sulf1 is expressed in a dynamically shifting pattern throughout mouse embryonic development. Whole mount in situ hybridization (ISH) using a DIG-labeled Sulf1 probe was performed on WT embryos at ages E9.5 (A, F), E10.5 (B, G), and E11.5 (C, H). Embryos in (F-H) are viewed from a rostral position, with dorsal at the top. Section-based ISH using 35 Slabeled Sulf1 probe was performed on 20-mm transverse forelimblevel sections of E10.5 (D, E), E11.5 (I-J), and E12.5 (K-N) embryos. Abbreviations used: ao, aorta; ar, artery; b, bronchus; cp, choroid plexus; cv, cardinal vein; e, esophagus; lb, distal forelimb bud; n, notochord; nt, neural tube; p, palate; sc, condensing sclerotome; t, tongue. Scale bar in (A) corresponds to 1 mm in panels A, B, and C, and 0.67 mm in panels F, G, and H. Scale bar in ( Figure S5 Sulf2 is also expressed in numerous tissues in adult and developing mice. Expression of Sulf2 in tissues derived from duplicate wild-type C57BL/6 adults (upper two panels) and embryonic samples (lower panel), as determined by consulting the microarray analysis submitted to Gene Expression Omnibus (GEO) with accession number GSE1133. Expression values are shown for probeset gnflm29631_a_at on the custom Affymetrix chip [25]. Each point refers to a different sample; the lines represent the mean expression value for each tissue.