Mice lacking sister chromatid cohesion protein PDS5B exhibit developmental abnormalities reminiscent of Cornelia de Lange syndrome

Sister chromatid cohesion, the linkage of sister chromatids from S phase until onset of anaphase, is important for accurate chromosome segregation during cell division and is mediated by a highly conserved multi-protein complex, called cohesin. The cohesin complex consists of core components,SMC1, SMC3, SCC1 and SCC3, and cohesin activity is regulated by its regulatory factors, PDS5, SCC2, SCC4 and ECO1(Hartman et al., 2000; Nasmyth and Haering, 2005). In addition to chromosomal dynamics, cohesin and its regulatory factors also play important roles in development. For example, Mau-2, a recently identified metazoan homolog of yeast SCC4, not only participates in sister chromatid cohesion, but also guides cell movement and axonal outgrowth during development (Benard et al.,2004; Seitan et al.,2006; Watrin et al.,2006). In Drosophila, mutants of cohesion factors (e.g. SMC1 and NIPPED-B) regulate transcription of cut and other genes encoding developmental regulators through attenuating cohesin-mediated blocking of long-range enhancer and promoter communication(Dorsett et al., 2005; Rollins et al., 2004). This additional developmental function has recently been extended to humans with the discovery that mutations in NIPBL, SMC1A and SMC3 are responsible for ∼50% of cases of Cornelia de Lange syndrome (CdLS). CdLS patients present with a variety of developmental anomalies, including dysmorphic facial features, mental retardation, growth delay, limb reduction,genital anomalies and cardiac defects(Deardorff et al., 2007; Krantz et al., 2004; Musio et al., 2006; Tonkin et al., 2004).

PDS5 is a regulatory component of cohesin and interacts genetically or physically with the cohesion factors SCC2, ECO1, SMC1, SMC3, SCC1 and SCC3(Losada et al., 2005; Tanaka et al., 2001) to ensure proper sister chromatid cohesion during mitosis and meiosis(Hartman et al., 2000; Panizza et al., 2000; van Heemst et al., 1999). The N-terminal region of PDS5 contains several HEAT repeats, a motif that is involved in protein-protein interactions and is present in other cohesin subunits and regulatory factors (i.e. SCC2)(Neuwald and Hirano, 2000; Panizza et al., 2000). In vertebrates, there are two homologs of PDS5, PDS5A and PDS5B, that both contribute to cohesion dynamics(Losada et al., 2005). Interestingly, in addition to its regulation of chromatid cohesion, PDS5 also appears to directly regulate transcription of a number of developmental genes. For example, similar to SCC2/NIPPED-B and SMC1, PDS5 regulates transcription of the Drosophila cut gene during wing development(Dorsett et al., 2005). In humans because PDS5B (also known as APRIN in human and mouse) is located in a region where loss of heterozygosity is commonly detected in tumors (13q12.3), it may act as a tumor suppressor(Geck et al., 2000). Furthermore, PDS5B regulates androgen-induced differentiation of prostate epithelial cells (Geck et al.,2000). Finally, lethality in yeast Pds5 mutants can be rescued by overexpression of topoisomerase II, however, TOPII does not compensate for the cohesion deficit in these mutants, indicating that PDS5 has crucial functions in yeast beyond those related to chromosome cohesion(Aguilar et al., 2005).

To delineate physiological roles of PDS5B in mammals we generated Pds5B-deficient mice using homologous recombination. Loss of Pds5B results in a number of developmental defects, including dysmorphic facies, cleft palate, skeletal patterning and bone development defects, cardiac malformation, distal enteric nervous system (ENS)aganglionosis, abnormal autonomic nervous system formation, and depletion of primordial germ cells. Together, these results indicate that PDS5B and perhaps the cohesin complex itself is a critical regulator of multiple aspects of organogenesis. The association of CdLS with mutations in several different cohesion proteins and the constellation of developmental defects in Pds5B^-/- mice indicate the widespread role of the cohesion factors in normal development, and provide a possible mechanism for their role in the pathogenesis of CdLS and related developmental disorders. As the first mouse model resembling CdLS, the Pds5B-deficient mice will be useful in delineating how PDS5B and cohesin regulate developmental processes.

Pds5B-deficent mice were generated by homologous recombination and genotypes were determined by Southern and PCR analyses.

The Pds5B recombination construct encompasses genomic regions surrounding the first coding exon. A 7.9 kb 5′ arm upstream of the first coding exon of Pds5B containing 20 nt of 5′ UTR and a 6.3 kb 3′ arm downstream of the first coding exon of Pds5B were generated by PCR (See below for primer sequences) and cloned into pRS315 by yeast gap repair (Le et al.,2005). The final targeting construct was generated using yeast gap repair by co-transforming yeast with linearized 7.9 kb 5′ arm in pRS315,6.3 kb 3′ arm in pRS315, the β-gal-PGK-Neo(β-gal-Neo) cassette, and two adapters. Two adapters were generated by annealing adapter forward and reverse primers and filled by Herculase DNA polymerase (Stratagene). For cloning the 5′ arm and 3′ arm by gap repair, pRS315 was digested with SacI and XhoI. For cloning the final target construct, 5′ arm pRS315 and 3′ arm pRS315 were digested with XhoI and MluI,respectively. The final Pds5B recombination construct was confirmed by sequencing analysis to ensure veracity of the Pds5B arm regions.

The Pds5B targeting vector was linearized with KpnI and transfected by electroporation into R1 embryonic stem cells. Homologous recombinants were screened by Southern blot hybridization using a probe downstream of the 3′ arm. Two properly targeted ES cell clones were injected into blastocysts to generate chimeric mice. The chimeras were mated and successful germline transmission of the targeted allele was detected by Southern blotting. Heterozygous mice grew normally and were mated toβ-actin Cre mice to remove the PGK-Neo cassette. The phenotypes of two independent lines (before or after excision of the PGK-Neo cassette) were similar. All animal procedures were performed according to our institutional approved protocols. Experiments were performed on mice from a mixed 129Sv/B6 background. The genotypes of Pds5B mutant mice were determined using Southern blotting or PCR using primers P1, P2, P3 (95°C 30 seconds,58°C 60 seconds, 72°C 90 seconds; 35 cycles). Wild-type and mutant alleles were amplified with P1-P2 (390 nt amplicon) and P1-P3 (450 nt amplicon) primer pairs, respectively. (See below for primer sequences.)

Primers, containing Pds5B genomic and associated pRS315sequences, used for PCR amplification of the 5′ arm and 3′ arm of the targeting construct were as follows:

5′ arm forward primer:TTAATGCGCCGCTACAGGGCGCGTCACGCGTCTGTGCTGGAGCTCACATGACCTGTCCT; 5′ arm reverse primer:ACAGCTATGACCATGATTACGCCAACTCGAGGATGGAAATGTCTGTACCCCTAAAGCAAGAAAA; 3′ arm forward primer: TTAATGCGCCGCTACAGGGCGCGTCACGCGTCAGTGTATGCTGCAGACAGCTGGGCAGTT;3′ arm reverse primer:ACAGCTATGACCATGATTACGCCAACTCGAGCAGCCTGGTTTGCAGAGAGTTCTAGCACA.

Primers used to generate adapters for the final targeting construct using yeast gap repair:

Adapter 1 forward primer: GATCCCCCAAGCTTACT TAG ATCTCGAGCTAGCACGCGTGATGGAAATG; adapter 1 reverse primer:TTTTCTTGCTTTAGGGGTACAGACATTTCCATCACGCGTGCTAGC. Adapter 2 forward primer:ATACATTATACGAAGTTATAC CGGTTAATTAAGGGCTCGAGCTAGCGGGCCC; Adapter 2 reverse primer: CACACAAACTGCCCAGCTGTCTGCAGCATACACTGGGGCCCGCTAGCTCGAGCCCTTAA.

Primers used to genotype Pds5B mutant mice: P1,CCAGACCTGAAGAATTGGTGGAGGA;P2,ACCCTCCTGGAGTCAAGGAAA; P3,ACCCTCCTGGAGTCAAGGAAA.

Rabbit anti-PDS5B polyclonal antibodies were raised using C-terminal peptide immunogen NH2-CATKENDSSEEMDV-COOH (Pacific Immunology, CA, USA). The same peptide was crosslinked to a solid support by using SulfoLink coupling Gel (Pierce Biotechnology, IL, USA) and used for affinity purification of antibodies from serum.

Tissues used in immunohistochemistry were fixed in paraformaldehyde or Bouin's fixative. For antigen retrieval, paraffin-embedded sections were deparaffinized in xylene, hydrated and boiled in 1 mM EDTA solution for 30 minutes prior to immunostaining (Jain et al., 2004; Naughton et al.,2006). Primary antibodies used in this study include sheep tyrosine hydroxylase (TOH; Chemicon), rabbit neuron-specific class IIIβ-tubulin (TuJ1; Covance), rat GCNA1 (George C. Enders, PhD, University of Kansas Medical Center, Kansas City, KS, USA), mouse BrdU (Roche), which were visualized using donkey anti-sheep HRP (Jackson ImmunoResearch), donkey anti-rabbit Alexa Fluor 488 (1:100; Molecular Probes), goat anti-rat Cy3(Jackson ImmunoResearch), or goat anti-mouse Alexa Fluor 488 (Molecular Probes) secondary antibodies. Bisbenzimide (2 μg/ml; Sigma) was used for nuclear staining.

Western blot analysis was performed using standard techniques using proteins derived from embryonic day (E) 14.5 forelimbs. The limbs were lysed in 2× SDS protein lysis buffer (100 mM Tris-HCl, pH 6.8, 4% SDS,1× protease inhibitor cocktail; Roche). The western blots were probed with rabbit anti-PDS5B antibody (1:500 dilution) and PDS5B was visualized using chemiluminescence substrate (Pierce). For a loading control, mouse anti-β-tubulin antibody (DSHB, University of Iowa, IA, USA) was also used at 1:20,000 dilution.

Alizarin Red S and Alcian Blue staining of newborn mice was performed as previously described (Wellik and Capecchi,2003). Briefly, newborn mice were skinned and dehydrated in 95%ethanol. They were then stained in Alcian Blue solution for 3 days, rehydrated and incubated in 1% KOH for 2 days, and stained with Alizarin Red S for 3 days. Embryonic and newborn heads were fixed in 4% paraformaldehyde at 4°C for 16 hours. The fixed heads were properly oriented in paraffin, coronal sections were prepared, sections were stained with Hematoxylin and Eosin(H&E) and examined microscopically for evidence of cleft palate.

The thoraxes of newborn mice was fixed in 4% formaldehyde before the heart was dissected free and embedded in paraffin. Entire neonatal hearts were serially sectioned at 6 μm thickness, stained with H&E, and inspected for defects. Two people (J.M.E. and P.Y.J.) examined the set of sections from every heart under a 5× microscope objective. The two examiners independently made the same diagnosis for every heart in this study.

For analysis of the enteric nervous system, the E12.5 or E18.5 whole gut was dissected from the mouse and fixed with 4% paraformaldehyde at 25°C for 30 minutes. After fixation, explants were washed three times in TBST (1 mM Tris, 150 mM NaCl, 0.2% Triton X-100) and blocked with 4% donkey serum in TBST(1 hour, 25°C) before incubation with primary antibody (4°C,overnight) in TBST. Primary antibody (rabbit anti-Tuj1, 1:100; Covance)staining was visualized using donkey anti-rabbit Alexa Fluor 488 (1:100;Molecular Probes) secondary antibody after washing three times with TBST(Fu et al., 2006). For cross-sectional analysis, post-natal day (P)0 gut was fixed in 4%paraformaldehyde and embedded in paraffin. Paraffin sections were stained with anti-Tuj1 (rabbit; Covance; 1:500) and visualized using goat anti-rabbit Cy3(1:500; Jackson ImmunoResearch). P0 gut staining with acetylcholinesterase and analysis of the percentage of distal colon colonized by ENS precursors was performed as previously described (Jain et al., 2004).

The sympathetic nervous system of embryos and newborns was analyzed using whole-mount TOH immunohistochemistry as described previously(Enomoto et al., 2001).

Embryonic gonads were dissected, fixed in Bouin's solution overnight at 4°C, embedded in paraffin, and 6 μm sections were prepared. Primordial germ cells were examined by H&E staining, alkaline phosphatase histochemistry, or GCNA1 immunohistochemistry. For germ cell identification,anti-GCNA1 antibody (1:200, rat polyclonal) was used, whereas Sertoli cells were identified using anti-GATA4 antibody (1:200, goat Sertoli cells polyclonal; Santa Cruz Biotechnology)(Naughton et al., 2006).

For BrdU and GCNA1 double immunostaining, timed pregnant females were injected with 200 mg/kg body weight BrdU 2 hours prior to sacrifice(Jain et al., 2004). Paraffin-embedded sections of embryonic testis were incubated with BrdU primary antibody (1: 200; Roche) and visualized using goat anti-mouse immunoglobulin labeled with Alexa Fluor 488 (1:500; Molecular Probes)secondary antibody, followed by staining with GCNA1 (1:200) and goat anti-rat immunoglobulin labeled with Cy3 (1:500, Jackson ImmunoResearch). TUNEL was performed as previously described (Jain et al., 2004).

Testes from E16.5 mutant and wild-type mice were transplanted subcutaneously onto the back and/or flank of castrated 4- to 8-week-old male nude mice (Taconic No. NCRNUM, Germantown, NY, USA) as previously described(Honaramooz et al., 2002). The grafted donor testes were harvested and processed for histological evaluation at the indicated time points.

Chromosome analysis was performed in the Chromosome Analysis Laboratory at University of Southern California directed by Chih-Lin Hsieh, PhD. Mouse embryonic fibroblasts from wild-type and Pds5B mutant mice (passage 3) were cultured in DMEM with 10% fetal bovine serum. Chromosomes were harvested the day after plating after growth for 1 hour in colcemid (1μg/ml) using standard hypotonic (0.075 M KCl) treatment and fixed in methanol:acetic acid (3:1). Slides were prepared and chromosomes were analyzed using the Geimsa-trypsin-Wrights (GTW) banding method. Two cells were fully analyzed from each line, and all were found to have normal karyotypes. Twenty-two to 25 cells were examined for precocious sister chromatid separation (PSCS). To avoid selection bias, the first 22-25 metaphases encountered were included in the analysis regardless of the length of the chromosomes or differences in the quality of banding.

Mouse cDNA fragments corresponding to Pds5B (probe 1: nt 2903-3438 and probe 2: nt 4561-5098) were generated by RT-PCR and cloned into the BamHI and HindIII sites of pBluescript II KS+ vector. Sense or antisense ³⁵S-labeled cRNA probes were generated from these Pds5B fragments and in situ hybridization was performed as previously described (Song et al., 2002). Frozen sections (12 μm) were mounted on poly-L-lysine-coated slides, fixed in cold 4% paraformaldehyde in PBS, acetylated and hybridized at 45°C for 4 hours in hybridization buffer containing the ³⁵S-labeled antisense cRNA probes. After hybridization, the sections were treated with RNase A (20 μg/ml) at 37°C for 20 minutes. RNase A-resistant hybrids were detected by autoradiography. Sections hybridized with the sense probes served as negative controls.

RNA was prepared from mouse adult and embryonic tissues using Trizol(Invitrogen) and quantified by the Ribogreen fluorometric assay (Molecular Probes). Reverse transcription was performed using M-MLV reverse transcriptase(Invitrogen) and oligo(dT) and random hexamers as primers. qRT-PCR was performed using a Model 7700 instrument (Applied Biosystems) using Sybr Green I fluorescence (Molecular Probes) as described previously(Svaren et al., 2000). Target genes were analyzed using standard curves to determine relative levels of gene expression. Individual RNA samples were normalized according to the levels of GAPDH or 18S mRNA. (See below for primer sequences.)

Primers used for qRT-PCR analysis: mouse Pds5B,ACCCTCCTGGAGTCAAGGAAA and CAGAGTCCTGGTCCAT GT CCAT; mouse GAPDH,TGCCCCCATGTTTGTGATG and TGTGG TCATGAGCCCTTCC; mouse 18S ribosomal RNA, CGGCTA CCA CATCCAAGGAA and GCTGGAATTACCGCGGCT.

Flag-tagged human PDS5A in pCR3.1 expression vector and the human PDS5B (KIAA0979) cDNA were gifts from Kasid Usha(Georgetown University, DC, USA) and Kazusa DNA Research Institute (Chiba,Japan), respectively. To generate the carboxyl-terminal EGFP-PDS5B fusion protein, KIAA0979 was digested with BamHI and EcoRI, blunted with Klenow, and cloned into the blunted XhoI site in pEGFP-C1(Clontech). The EGFP-FLAG-PDS5A fusion protein was generated by digesting the Flag-PDS5A construct with PmeI and inserting this fragment into the blunted BglII site of pEGFP-C1. HeLa cells were cultured in DMEM with 10% fetal bovine serum. Transient expression of EGFP-tagged PDS5A and PDS5B was achieved using Lipofectamine-mediated transfection of plasmids.

As an initial investigation of PDS5B function we surveyed Pds5BmRNA expression levels in adult and embryonic mouse tissues. In the adult we found the highest levels of Pds5B expression in brain and testis(Fig. 1A). At the cellular level, in situ hybridization experiments revealed high Pds5Bexpression in neurons of adult hippocampus and dentate gyrus(Fig. 1C). In the testis, high PDS5B expression was detected at the periphery of the seminiferous tubules in a pattern consistent with immature germ cells(Fig. 1B). In the embryo, Pds5B expression was comparable in all tissues examined (data not shown), and in situ hybridization revealed that Pds5B was ubiquitously expressed at the cellular level(Fig. 1D).

The expression of PDS5B, a regulatory factor of cohesin, in post-mitotic as well as dividing cells, suggested that it participates in multiple processes in mammals. Therefore, to determine its physiological roles in mice, we generated Pds5B null mice by homologous recombination(Fig. 1E-G). To confirm the absence of Pds5B expression in the homozygous Pds5B mice, we quantified Pds5B mRNA in E14.5 Pds5B^-/- brain and examined PDS5B levels in Pds5B^-/- limbs by western blotting using PDS5B antibodies. We did not detect Pds5B mRNA or protein (Fig. 1H,I; see Fig. S1 in the supplementary material) in embryos homozygous for the mutant Pds5B allele, indicating that it is a true null allele. Pds5B^-/- embryos were present at the expected Mendelian ratio until age E16.5. However, only about 75% of the expected number of Pds5B^-/- fetuses were born alive and none of the mice homozygous for the mutant Pds5B allele survived beyond P1. The newborn animals had signs of respiratory distress, including labored breathing, use of accessory muscles of respiration as indicated by head bobbing and intercostal and abdominal retractions, cyanosis and pallor. They all died shortly after birth, most probably from cardiac and/or respiratory abnormalities (see below).

Wild-type and Pds5B^-/- mutant embryos were of similar size until E16.5, but then the mutants began to show signs of growth retardation. Newborn Pds5B^-/- mice were smaller than their wild-type littermates, and had short stature, short limbs, a small head and facial dysmorphisms (short snout, short low chin and thin upper lip; Fig. 2A-C,K). These abnormalities, which are similar to those reported for CdLS patients, along with the fact that mutations in other cohesion proteins are associated with CdLS, suggested to us that the Pds5B-deficient mice could be a valuable model of this developmental syndrome. To substantiate this idea, we searched for other abnormalities in Pds5B^-/- mice that are commonly found in patients with CdLS.

To test for skeletal abnormalities similar to those in CdLS patients, we stained wild-type and mutant embryos with Alizarin Red S and Alcian Blue to highlight the skeleton, and found that several bones including the mandible,ulna, radius, humerus and scapula were shorter in the Pds5B^-/- mice (Fig. 2D-F). Shortened limbs are associated with mild cases of CdLS(Ireland et al., 1993);however, we did not observe the limb truncations or fused digits (syndactyly)that are observed in more severely affected CdLS patients. All Pds5B^-/- mice exhibited sternal malformations [wild type(WT): 0/24; Pds5B^-/-: 37/37] characterized by either two unfused ossification centers or lack of ossification of the sternebrae(Fig. 2G). In addition, the sternums of Pds5B^-/- mice were shorter than those of wild-type mice (Fig. 2G). We also found other abnormalities in skeletal patterning, with incomplete penetrance and variable expressivity, including delayed ossification of digital bones and vertebrae (data not shown), fusion of extensively ossified rib anlage at C7 and the costal cartilage of the T1 rib (WT: 0/10; Pds5B^-/-: 10/18; Fig. 2H), loss or hypoplasia of the 13th ribs (WT: 0/10; Pds5B^-/-: 20/20), and unfused ossification centers in the vertebrae (WT: 0/10; Pds5B^-/-: 5/18; Fig. 2I) and hyoid bone (data not shown).

Another developmental abnormality observed in some children with CdLS is cleft palate. An examination of this region in Pds5B^-/-mice revealed a complete cleft of the secondary palate(Fig. 2J,K) and failure of the posterior palatal bones to extend fully and meet at the midline(Fig. 2L,M; WT: 0/24; Pds5B^-/-: 25/33). To further characterize the dysmorphogenesis of the palatal shelves in Pds5B mutant embryos, we examined the palatal shelves of wild-type and Pds5B^-/-embryos at E12.5, E14.5 and E18.5. Whereas palatal shelves (PS) in Pds5B^-/- embryos were seemingly normal at E12.5(Fig. 2N,Q), by E14.5 they were clearly abnormal (Fig. 2O,R). In E18.5 wild-type embryos, the medial-edge epithelia of the two PS have normally fused and then degenerated (Fig. 2P), however, the PS of E18.5 Pds5B^-/- embryos were shorter and failed to fuse (Fig. 2S). Using BrdU and TUNEL analysis on E13.5 embryonic palates, we did not observe significant changes in cell proliferation or apoptosis in palatal shelves of Pds5B^-/- embryos compared to WT or Pds5B^+/- embryos (see Fig. S2 in the supplementary material). The cleft palate observed in Pds5B^-/- mice could be the result of abnormal reorientation of palatal shelves or failure of epithelial-mesenchymal transition of medial edge epithelia at a later stage. Clearly, cleft palate is a distinctive feature of Pds5B^-/-mice and could cause aspiration and subsequent asphyxiation, thus contributing to the neonatal lethality of these mice.

Because children with CdLS often have congenital heart defects and Pds5B^-/- mice suffer respiratory distress and perinatal lethality, we searched for cardiac malformations. The incidence of congenital heart defects of any type was significantly higher in Pds5B^-/- compared to wild-type mice (15/24 versus 0/14, P<0.001; Fig. 3A-G). The defects included those that are commonly found in CdLS patients (Jackson et al.,1993). We identified atrial septal defects of the secundum type(ASD, 6/24), ventricular septal defects (VSD, 12/24; ten perimembranous, one muscular, one inlet), and a common atrioventricular canal defect (1/24; Fig. 3C-F) in these mice. Of note, one Pds5B^-/- mouse had a large perimembranous VSD with anterior malalignment of the outlet septum, causing the aorta to shift over the VSD toward the right ventricle(Fig. 3C). This mouse did not have right ventricular outflow tract obstruction and thus had the forme fruste of tetralogy of Fallot, a common malformation in CdLS. The VSDs and common AV canal defects would be expected to cause pulmonary overcirculation and edema from left-to-right shunting of blood at the ventricular level, thus resulting in respiratory distress and potentially contributing to early postnatal lethality of Pds5B^-/- mice.

The high expression of Pds5B in post-mitotic neurons, the role of other cohesion components such as Mau-2 in axonal guidance and neuronal migration, and the profound mental retardation associated with CdLS,encouraged us to examine the nervous system in Pds5B^-/-mice. Consistent with clinical reports showing that CdLS patients have relatively normal brain anatomy, we did not observe any gross or microscopic structural abnormalities in the brain of Pds5B-deficient mice. This suggests that the cohesin complex does not play a major role in CNS development; instead, it may influence neuronal function or connectivity. Unfortunately, the neonatal lethality of Pds5B^-/- mice currently prevents us from further examining this hypothesis.

An examination of the peripheral nervous system revealed that the sympathetic chain ganglia were well developed with neuronal projections appearing normal in Pds5B^-/- mice (see Fig. S3 in the supplementary material). We also examined the superior cervical ganglia (SCG),which provides sympathetic innervation to the ocular muscles. Abnormalities in the SCG are associated with ptosis, a common condition in CdLS. We found a number of abnormalities in the SCG and its projections to target organs in the Pds5B^-/- mice (Fig. 4A). For example, in 40% (5/13) of Pds5B^-/-mice, the SCG was located far caudal of its normal position(Fig. 4B,C). Furthermore, all Pds5B^-/- mice had thin carotid nerves(Fig. 4D,E) and 30% (4/13) of these mice had unilateral or bilateral absence of the carotid nerve(Fig. 4F,G). These SCG defects are similar to those observed in mice deficient for the neurotrophic factor artemin, which commonly manifest unilateral or bilateral ptosis(Honma et al., 2002). These results suggest that SCG innervation defects are the likely cause of the ptosis observed in many CdLS patients(Jackson et al., 1993).

We next examined the enteric nervous system (ENS), a complex network of neurons and glia within the bowel that is derived from migrating neural crest cells (NCCs). The NCCs migrate in a proximal to distal direction along the length of the bowel and normally reach the mid-colon by E12.5(Fig. 4H). By contrast, in Pds5B^-/- mice, ENS precursors failed to migrate past the ileocecal (IC) junction by this age (Fig. 4I). Furthermore, even in regions of the distal bowel colonized by NCCs, the neuronal density was lower in Pds5B^-/- mice compared to wild-type animals (data not shown).

To determine if the enteric neuron defects occur because of delayed migration of NCCs into the colon, we examined the ENS at E18.5 and P0 using the Tuj1 antibody, which recognizes a neuronal form of tubulin, or acetylcholinesterase (AchE) staining (Fig. 4J-M; see Fig. S3 in the supplementary material). As expected,enteric neurons colonize the entire length of the wild-type neonatal bowel(Fig. 4J,L), whereas 70% (8/12)of Pds5B^-/- mice exhibited variable degrees of abnormal innervation and ganglion formation in the distal colon(Fig. 4K,M). By contrast, small bowel innervation appeared grossly normal in Pds5B^-/- mice(data not shown). Thus, the majority of Pds5B^-/- mice had significant defects in the distal enteric nervous system that resemble Hirschsprung disease in humans. This contrasts with CdLS where distal colonic aganglionosis has not been reported, but where other gastrointestinal problems including persistent emesis and feeding problems are common, occurring in 62%and 77%, respectively, of infants with CdLS(Jackson et al., 1993). These observations suggest that defects in the ENS such as hypoganglionosis may underlie the gastrointestinal symptoms of CdLS patients.

The high expression of Pds5B in adult and embryonic testis, the role of PDS5 in mitosis and meiosis in lower organisms, and the high frequency of genital anomalies seen in people with CdLS, encouraged us to examine germ cell development in Pds5B^-/- mice. We found a severely reduced number of germ cells in testes and ovaries of newborn Pds5B^-/- mice (Fig. 5A-D; see Fig. S4 in the supplementary material) that appeared to be more severe in males. We therefore focused on male germ cell development to further investigate mechanisms underlying germ cell depletion. At E16.5, Pds5B^-/- male mice had an 80% reduction in germ cells in the testis (Fig. 5E,F,I). To determine if the remaining germ cells could undergo spermatogenesis, we performed testis transplantation to propagate E16.5 testis of wild-type and Pds5B^-/- mice as explants in nude mice(Naughton et al., 2006). Using GCNA1 immunohistochemistry, we found a complete depletion of spermatogonial stem cells at 2 weeks after transplantation(Fig. 5G,H). By 6 weeks after transplantation, the explanted Pds5B^-/- testes contained testicular cords that only contained Sertoli cells (see Fig. S4 in the supplementary material). The Pds5B^-/- Sertoli cells were morphologically normal as determined by GATA4 immunohistochemistry (data not shown). These studies indicate that the germ cells remaining at E16.5 in Pds5B^-/- mice were incapable of sustaining spermatogenesis.

The embryonic loss of germ cells at E16.5 in Pds5B-deficient mice suggested that PDS5B is important for primordial germ cell (PGC) development. PGCs are derived from a founder population of ∼45 cells in the extraembryonic mesoderm posterior to the primitive streak around E7.25(Ginsburg et al., 1990). These cells migrate to the genital ridge by E11.5(Gomperts et al., 1994), and continue proliferating until E13.5 in males(McLaren, 2000). To identify the mechanism underlying the reduced number of germ cells in male Pds5B^-/- mice, we evaluated the migration, proliferation and apoptosis of these germ cell precursors. At E12.5, a time point when PGCs enter the genital ridge, there were already significantly fewer numbers of germ cells in the Pds5B^-/- mice. We hypothesized that overt migration defects in Pds5B^-/- PGCs would result in failure of these precursors to reach the genital ridge by E11.5. However,alkaline phosphatase staining of E10.5 gonads to identify PGCs revealed no significant differences in PGC location between wild-type and Pds5B^-/- embryos (data not shown), indicating that PGC migration at this early stage must be relatively normal. In addition, TUNEL analysis revealed no changes in the number of apoptotic germ cells in E12.5 or E16.5 mutant embryos (data not shown). However, when we examined germ cell proliferation in E12.5 embryos using BrdU incorporation, we found a 30%reduction in the number of BrdU-positive germ cells in testes of Pds5B^-/- versus wild-type mice(Fig. 5J; see Fig. S4 in the supplementary material). Although it is possible that germ cell depletion could be partially due to apoptosis or delayed migration at stages in development that we did not examine, it is clear that the reduced proliferation is a major factor in the reduced number of germ cells.

PDS5 is crucial for sister chromatid cohesion(Dorsett et al., 2005; Hartman et al., 2000; Losada et al., 2005; Panizza et al., 2000; van Heemst et al., 1999; Wang et al., 2003; Wang et al., 2002). To test for cohesion defects in Pds5B-deficient mammalian cells, we performed GTW metaphase chromosome banding of mouse embryonic fibroblast (MEF)chromosomes. Two MEF lines were obtained from both Pds5B-deficent and wild-type mice. Quantitative RT-PCR and western blot analyses were performed and we found that these cells were indeed devoid of both Pds5B mRNA and protein (see Fig. S1 in the supplementary material). We analyzed a total of 22-25 metaphase cells per line (two WT and two Pds5B^-/-MEF lines; detailed results are available on request). No obvious cohesion defects such as precocious sister chromatid separation (PSCS) or gross chromosomal abnormalities were observed in the Pds5B^-/-MEFs (Fig. 6A,B). The lack of sister chromatid cohesion abnormalities in Pds5B-deficient cells suggests that PDS5B has novel functions other than sister chromatid cohesion that are important for mammalian embryonic development.

The wide ranging deficits in the Pds5B^-/- mice suggested that PDS5B has acquired additional functional domains during evolution. To examine this possibility, we performed motif searching of mouse PDS5B using Blocks and Pfam motif libraries with the default cut-off scores. We identified the previously reported HEAT repeats within the N terminus(Losada et al., 2005; Neuwald and Hirano, 2000; Panizza et al., 2000) and two AT hook domains in the C terminus that closely match the high mobility group protein (HMGY) signature (Block IPB000116A). Further analysis revealed that these AT hook domains are present in PDS5B from fish to human(Fig. 6C), but are not present in the yeast, worm and fly orthologs. The AT hook domain was initially identified in the HMG-I/Y family as a domain that binds to AT tracts in the minor groove of DNA (Maher and Nathans,1996). The identification of AT hook domains in PDS5B is the first recognizable DNA binding motif reported in a sister chromatid cohesion factor,indicating that PDS5B and the cohesin complex may regulate gene transcription through direct cohesin complex-DNA interactions mediated by PDS5B.

There are two PDS5 homologs in vertebrates(Losada et al., 2005). The N-terminal regions of these proteins both contain HEAT repeats and are closely related, but PDS5A lacks the AT hook domains present in the C-terminal region of PDS5B. Despite their similarity, the severe abnormalities caused by PDS5B deficiency clearly show that PDS5A cannot compensate for at least some of the physiologic roles of PDS5B. We investigated whether their distinct cellular functions could stem, at least in part, from differential subcellular localization of these two proteins. We generated EGFP-tagged PDS5A and PDS5B fusion proteins and determined their subcellular localization in interphase cells using fluorescence microscopy. Whereas both PDS5A and PDS5B were found in the nucleus, PDS5B appeared to be more concentrated in the nucleolus than in the rest of the nucleus. By contrast, PDS5A was less abundant in the nucleolus than in the rest of the nucleus(Fig. 6D).

Mutations in four different cohesion proteins (NIPBL, SMC1A, SMC3 and ESCO2) have been found in patients with CdLS and Roberts syndrome/SC phocomelia (RBS/SC), and are thought to be responsible for the broad spectrum of prenatal and postnatal developmental anomalies(Deardorff et al., 2007; Krantz et al., 2004; Musio et al., 2006; Schule et al., 2005; Tonkin et al., 2004; Vega et al., 2005). These observations clearly indicate that sister chromatid cohesion factors are important for early development of multicellular organisms, either through abnormalities in chromosome dynamics or via alternative functions.

In this work, we generated mice deficient in Pds5B, a sister chromatid cohesion factor. These mice manifest a spectrum of developmental defects similar to those found in human CdLS, including abnormal skeletal patterning, cardiac anomalies and cleft palate. Other phenotypes that are associated with CdLS such as mental retardation, hearing loss and gastroesophageal reflex are difficult to evaluate in this animal model particularly since Pds5B^-/- mice are not viable past P0. However, the striking defects in the enteric and autonomic nervous systems of these mice may be related to the frequent occurrence of ptosis and gastrointestinal problems in CdLS patients. Because of significant phenotypic overlap in Pds5B^-/- mice and people with CdLS(Table 1), and the fact that mutations in other cohesion factors have been identified in CdLS patients, we believe that Pds5B^-/- mice represent the first mouse model resembling human Cornelia de Lange syndrome. In fact, the broad spectrum of symptoms with incomplete penetrance and variable expressivity in these mice is also a feature typically observed in CdLS. These studies and the previously known association of mutations in other cohesion factors and CdLS suggest that screening CdLS patients for PDS5B mutations is warranted. PDS5B mutations, unless they produce dominant negative mutants, may be rare because they would be recessive and most CdLS cases are dominantly inherited (Krantz et al.,2004; Tonkin et al.,2004), and because they may result in lethality unless they are hypomorphic.

Although the molecular function of PDS5 is still under investigation,several organisms require diverse function of PDS5 for chromatid cohesion(Losada et al., 2005). For example, it may act as a regulator of the SMC ATPase through the HEAT motifs shared by SCC2 (Kimura and Hirano,2000) and thereby modulate the dynamics of the cohesin ring. Alternatively, PDS5 may act as a scaffolding protein and promote protein-protein interactions between adjacent cohesin complexes(Losada et al., 2005). An examination of the chromosomes from Pds5B-deficient mouse cells did not show any abnormalities in sister chromatid cohesion (i.e. PSCS). This result was in contrast to those obtained in budding yeast, worm and Drosophila, where PDS5 plays an important role in chromosome segregation (Dorsett et al.,2005; Hartman et al.,2000; Wang et al.,2003). It is possible that our assay in which a total of 50 cells were examined may not have detected subtle defects. Indeed, only mild cohesion defects were found in HeLa cells in which PDS5B was knocked down by RNAi (Losada et al., 2005). Furthermore, our results are consistent with reports of the lack of sister chromatid cohesion defects in cells from CdLS patients(Tonkin et al., 2004; Vrouwe et al., 2007) or only mild cohesion defects in 40% of CdLS patients(Kaur et al., 2005). Overall,these results are consistent with our hypothesis that the abnormalities present in patients with CdLS and related diseases probably result from abnormal cohesin-mediated functions other than sister chromatid cohesion(Dorsett, 2007; Krantz et al., 2004; Tonkin et al., 2004). In this regard, it has been hypothesized that the cohesin complex directly binds to regulatory cis-elements and transcriptionally regulates developmental gene expression (Krantz et al.,2004; Rollins et al.,1999; Tonkin et al.,2004). Support for this idea comes from a number of studies in lower organisms. For example, severe cohesion defects have been observed in Drosophila Pds5 homozygous mutants, but not in heterozygotes, where only defects in cut gene expression and wing (limb) development are observed (Dorsett et al.,2005). In this regard, direct interactions of the cohesin complex with multiple sites between the wing margin enhancer and promoter lead to transcriptional regulation of the developmental gene cut(Dorsett et al., 2005). Yet another example of alternative Pds5B function is found in yeast Pds5mutants, in which topoisomerase II can rescue the lethality, but not the cohesion defect of these mutants (Aguilar et al., 2005). Furthermore, the cohesion protein, MAU-2, is highly expressed in the mammalian adult nervous system and is crucial for neuronal migration and axonal guidance (Seitan et al., 2006; Watrin et al.,2006). PDS5B is also expressed at high levels in neurons, yet its function in adult neurons is unlikely to be related to sister chromatid cohesion as these cells are post-mitotic and thus do not undergo cell division. How cohesin complexes could directly regulate transcription has remained elusive because no DNA binding motifs have been identified in proteins that comprise the cohesin complex or its regulatory factors. Our analysis of PDS5B has identified two AT hook domains, the motif that constitutes the DNA binding domain of HMG proteins(Maher and Nathans, 1996). We propose that PDS5B may constitute an important link in cohesin regulation of transcription through its ability to bind AT rich regulatory sequences and bring the cohesin complex in proximity to promoter elements to regulate transcription.

An alternative explanation for the absence of sister chromatid cohesion defects in Pds5B^-/- mice could be redundancy with its homolog PDS5A (Losada et al.,2005). Physiological evidence for this will require generation of Pds5A^-/- animals, but our preliminary studies indicate significant overlap in PDS5A and PDS5B mRNA expression (B.Z. and J.M. unpublished observations). Indeed, we found overlapping expression in a number of organ systems where Pds5B^-/- mice manifest developmental defects, suggesting differential sensitivity to PDS5B loss in various organs or perhaps reflecting differences in subcellular localization(and function) of these two closely related proteins.

Both PDS5A and PDS5B have previously been reported to be localized in the nucleus by immunohistochemistry (Kumar et al., 2004; Maffini et al.,2002). In these studies, we found that PDS5B is localized to the nucleus with higher levels in the nucleolus. Interestingly, PDS5B lacking the C-terminal 142 residues containing the second AT hook motif was localized to the nucleus but was not concentrated in the nucleolus (B.Z. and J.M.,unpublished observation), suggesting that this domain may contribute to its nucleolar localization. Nucleolar localization of PDS5B indicates that PDS5B and/or cohesin complexes containing PDS5B could regulate rRNA metabolism and ribosome biogenesis. Interestingly, other cohesion proteins (e.g. NIPBL and SMC1A) have also been detected in nucleoli(Andersen et al., 2005), and yeast PDS5 can interact with NIP7 and NOP7, both of which are involved in ribosomal RNA (rRNA) biosynthesis(Davierwala et al., 2005; Krogan et al., 2006). Further evidence of cohesion protein involvement in ribosomal biogenesis comes from studies in yeast, where an additional copy of NOG2, which encodes a GTPase required for 60S ribosomal subunit maturation, is able to suppress the defects caused by mutation of the cohesin protein Scc3(Bialkowska and Kurlandzka,2002). In addition, the cohesin complex is recruited by Sir2 to silenced chromatin in the rDNA loci to maintain rDNA repeat copy number(Kobayashi et al., 2004). Finally, normal development can be disrupted by defects in rRNA metabolism, a process largely confined to the nucleolus. For example, Tcof1, which encodes a nucleolar protein, is mutated in Treacher Collins syndrome, an autosomal dominant disorder of craniofacial development(Dixon et al., 2006). Haploinsufficiency of Tcof1 in mice causes a reduction of mature ribosomes that results in deficient neural crest cell migration and proliferation that leads to craniofacial abnormalities, including cleft palate and mandibular hypoplasia (Dixon et al.,2006). Deficits in the nucleolar functions of cohesion proteins could also contribute to the abnormalities observed in patients with CdLS or related diseases.

In summary, we have discovered that the cohesion protein PDS5B is a critical regulator of multiple aspects of organogenesis, probably via roles unrelated to its ancient function in sister chromatid cohesion. This Pds5B-deficient mouse provides the first mammalian model to study the molecular mechanisms of developmental functions of the cohesin complex and the pathogenesis of CdLS and related disorders with mutations in cohesion factors.

We are deeply indebted to Robert H. Baloh and Jason A. Gustin for careful review of the manuscript and/or helpful discussions of the data. We thank Kelli Simburger, Shelly Audrain, Amber Nielson, Jack Shields, Jennifer Armon,Amy Strickland, Amanda Knoten, Tatiana Gorodinsky and Nina Panchenko for technical assistance. We are grateful to George C. Enders for GCNA1 antibody,Kasid Usha for PDS5A plasmid, Kazusa DNA Research Institute for KIAA0979 cDNA, and Mark Johnston for pRS315 plasmid and help with yeast gap repair. We thank Chih-Lin Hsieh for providing the chromosomal analysis. This work was supported by NIH grants CA111966, AG13730 and NS039358(to J.M.), HD047396 (to S.J.), DK57038 and DK6459201 (to R.O.H.), O'BRIEN center for Kidney Disease Research (P30DK079333), and Washington University Cancer Biology Pathway Fellowship (to B.Z.). P.J. is a Scholar of the Child Health Research Center of Excellence in Developmental Biology at Washington University School of Medicine (HD001487).