Cornelia de Lange syndrome, cohesin, and beyond

Clinical presentations of CdLS

Diagnostic criteria and clinical particularities of CdLS due to NIPBL, SMC1A or SMC3 mutations

Certain diagnostic criteria of CdLS have been proposed previously (36, 37), mainly based on clinical presentations with specific facial features, hand profile and neurodevelopmental characteristics. Recently, Kline et al. have proposed minimal criteria for the diagnosis of CdLS, and a scoring system to evaluate the severity (7). Molecular identification of mutations in NIPBL, SMC1A and SMC3 is considered as the decisive index in that individuals with a pathogenic mutation in any one of the associated genes are defined as having CdLS (7). As defined by Kline et al. (7), in order to diagnose an individual clinically (e.g. without a gene mutation) there are specific criteria that should be met. A severity scoring system based on developmental milestones, malformations (particularly of the upper limb) and hearing loss was created, which was found to correlate well with specific brain malformations, IQ levels and mutations in NIPBL (7).

About 60% of CdLS probands have a heterozygous mutation in NIPBL. Genotype–phenotype correlations amongst a large number of probands indicate that presumably haploinsufficient NIPBL mutations (protein truncating mutations such as nonsense mutations, splice site mutations, and out-of-frame deletions or insertions) usually result in a more severe cognitive and structural phenotype than missense mutations (38). Approximately 5% of probands with a clinical diagnosis of CdLS were found to have missense or small in-frame deletion mutations in SMC1A, and one individual was found to have an in-frame 3 bp deletion in the SMC3 gene (12). The SMC1A and SMC3 cases have mild to moderate mental retardation without significant impairments in growth or structural abnormalities of the limb or other organ systems (12). Notably, probands with SMC1A or SMC3 mutations demonstrated some clinical features that are in contrast to the ‘classical’ form of CdLS (12). This cohort tend to have a more prominent nasal bridge than is typically seen in CdLS (11), and the majority of them had birth weights within normal parameters with normal head circumferences and growth measures later in life as well (Fig. 1). For the most walking and speech are often acquired, and overall, they exhibit a much milder level of cognitive involvement (12).

Approximately 65% of CdLS probands with a confident clinical diagnosis have mutations in one of cohesin-associated genes (NIPBL, SMC1A or SMC3). The molecular etiology of the remaining 35% of probands is unknown at this time. Mutations in the regulatory sequences of these three genes may account for a small percentage of cases and are not routinely screened for in clinical testing. Still additional genes are yet to be identified to account for these molecularly uncharacterized individuals, and with over 20 known genes implicated in the cohesin complex and its regulation (and many more yet to be implicated), there are plenty of potential candidates.

Cohesin biology

Cohesin is a dynamic complex regulated at various cell cycle stages by multiple mechanisms (Fig. 2) (39). The primary biological role identified for cohesin is to control sister chromatid segregation during both mitosis and meiosis. Four evolutionarily conserved subunits form the core structural components of the cohesin complex: two SMC (structural maintenance of chromosomes) proteins SMC1A and SMC3, one kleisin protein RAD21 (also called MCD1 or SCC1), and SA1/SA2 (also called STAG1/STAG2). SMC1B, REC8 and STAG3 are cohesin subunits in meiosis (40). Homologs of the cohesin complex and its regulatory genes have been identified in several eukaryotic model systems such as Saccharomyces cerevisiae, Schizosaccharomyces pombe, Drosophila melanogaster, Xenopus laevis, mouse and humans (41).

The protein structures of SMC1A and SMC3 are very similar, each spanning 1000–1500 amino acids, and both contain globular domains at the N - and C-terminal ends and a globular hinge domain in the middle which separates the alpha helical structures (42). N and C termini fold at the hinge domain to form an antiparallel coiled coil by bringing together the two halves of the alpha helix, which also subsequently form the head domain. Three highly conserved motifs, Walker A, Walker B, and a signature motif, were identified in the ATPase domain in the cohesin head formed by the N and C termini (42, 43). SMC1A and SMC3 dimerize at the hinge domains forming a V-shaped structure most likely through a hydrophobic interaction (44). The N terminus of the RAD21 subunit interacts with the head domain of SMC3 and the C terminus interacts with the head domain of SMC1A, with ATP being required for this association. RAD21 was therefore suggested to be a central regulator of cohesin function because it simultaneously crosslinks the SMC heterodimer, lies in close proximity to the ATPase active sites in cohesin head domains, and binds to the fourth known cohesin subunit, SA1/SA2 (45). The maturely formed cohesin is a four-subunit, ring-like complex with globular hinge and head domains separated by a 45 nm of coiled coil. Several working models have been proposed for the molecular basis of sister chromatid cohesion, with ‘topological’ and ‘handcuff’ being the most representative (44, 46). The topological model proposes no physical contact between chromatin and cohesin. The cohesin ring opens up at the head domain followed by the sister chromatids entering the ring and becoming topologically entrapped when RAD21 re-locks the ring. Alternatively, the handcuff model proposes specific interactions of DNA–protein and protein–protein. One cohesin ring binds one single chromatid after which inter-complex oligomerization occurs. Both hinge domain and coiled-coil domain have been suggested to be able to serve as the tethering surfaces; however, the prominent role of the hinge in cohesin dynamics implies that it could be the target of regulatory factors (46–48). At present, a lot of current biological observations support that cohesin binds to chromatin, and none of the models is able to rule out interactions between cohesin and DNA (46).

Cohesin binds to chromatin during the G1/S phase in budding yeast; however, in vertebrates binding happens during telophase of the preceding cell division (49). Additionally, cohesin binds at G2/M phase when a double-strand DNA break is created. Removal of cohesin from chromosome arms commences in prophase and is completed by anaphase (Fig. 2). In yeast, cohesin-associated regions (CARs) are distributed on average at 15-kb intervals on chromosome arms at transcriptional convergent sites which are usually AT-rich (50, 51). Although no consensus DNA sequence for cohesin binding has been identified, cohesin binding is enriched at heterochromatin (52) and the chromatin surrounding a DNA double-strand break (DSB) (53). Several factors that could affect cohesin–chromatin association have been demonstrated. First, all four subunits of cohesin and a well assembled complex are essential for chromatin binding in vivo (54, 55). Second, normal structure and function of all three motifs in the SMC head domains are indispensable (56, 57). Third, specific residues in the coiled-coil regions that are close to the heads are required. Fourth, the hinge domains of both SMC proteins also have a role in cohesin binding to chromatin (58, 59).

The heterodimer NIPBL/hSCC4 complex is evolutionarily conserved and required for loading of cohesin onto chromatin in both mitosis and meiosis, and to all chromosome regions currently under study such as heterochromatin, CARs, centromeres, and DSBs (53, 60). Our understanding of how NIPBL loads cohesin onto DNA is poor and no such study has been reported yet. The yeast homolog of NIPBL, Scc2, was suggested to be a kinase through sequence homology analysis (61), and mutations in Scc2 have greatly reduced Rad21 phosphorylation (62). NIPBL/SCC4 seems to be involved in all cohesin activities including SMC ATPase activation, hinge dimerization, chromatin binding, and chromatin remodeling (46). In addition to NIPBL/SCC4, certain histone modifications and local chromosome structures are also targeted by cohesin (52, 63, 64). Distinct protein complexes such as MRX, CEN complexes and Rep which are distinct from NIPBL/SCC4 can also recruit cohesin to specific chromosome regions (53, 65, 66). In summary, cells elaborately regulate chromatin binding of cohesin temporally and spatially, but most of the mechanisms are unknown at present. This regulation of the complex enables cohesin to perform diverse biological functions. Mutations in those associated major genes are expected to contribute to multiple human diseases, both known and unknown. Cohesin-independent mechanisms of cohesion have also been suggested in that complete unpairing of sister chromatids cannot be achieved by knocking down cohesin function alone. Condensin complexes (67), origin recognition complexes (68), centromere complexes (69) and DNA catenation (70) have all been reported to have a role in mediating cohesin-independent sister chromatid cohesion.

The removal of cohesin from chromosomes is another field that has been extensively studied. Cohesin begins to disassociate from the chromosome arms during prophase whereas the pericentric cohesin is protected until the onset of anaphase (71–73). In vertebrates, both cohesin subunits Stag2 and Rad21 are substrates of Polo-like kinase 1 (Plk1). A functioning Wapl is also required for prophase removal. At the same time, Shugoshin/MEIS-322/Sgo1 (74) and sororin (75) are able to protect centromere cohesins. An evolutionally conserved protein Pds5/BimD/Spo76 interacting with Wapl, sororin and Eco1 is also involved in maintaining pericentric cohesion in prophase (76). Inactivation of cohesion and the complete dissolution of cohesin at the onset of anaphase enable faithful sister chromatid segregation. Securin and separase were identified to regulate this event. Securin binds and inhibits protease activity of separase (Esp1) before anaphase (77). At the beginning of anaphase, securin is degraded by anaphase-promoting complex (APC) and separase is activated (78). The active separase cleaves Rad21 and the cohesin complex is further degraded (77). Mechanisms other than cohesin cleavage to ensure the complete inactivation of all cohesion have also been proposed (79).

Cohesin function and the etiologies of CdLS

CdLS is a genetically heterogeneous dominant multi-system developmental disorder that can be caused by mutations in one of at least three genes identified to date: the human Scc2 ortholog NIPBL and cohesin subunits SMC1A and SMC3 (9–12). All of these genes encode proteins that have been implicated to play roles in chromosome segregation as well as in DNA repair, gene expression and chromosome conformation (80). Cohesion or cell cycle defects were first speculated to cause developmental deficits in CdLS. Until now, there is no clear answer. One early study in a large cohort of CdLS samples reported that precocious sister chromatid separation was identified in 41% of probands’ lymphoblastoid cell lines vs 9% in healthy controls (81); however, another study with limited specimens found CdLS cells had no obvious cohesion defects (82). More experiments, larger groups of probands with clearly identified gene mutations, as well as additional assays able to sensitively evaluate chromosome segregation are required to clarify this question.

Increasingly, however, mounting evidence suggests that the dysregulation of gene expression that results from mutation in cohesin genes is more likely to represent the underlying pathogenesis of CdLS. The first key evidence of cohesin mediated gene regulation arose from screens in Drosophila for genes that regulate gene expression in a dosage-sensitive manner (83, 84). In Drosophila, the homeobox genes cut and Ultrabithorax (Ubx) control multiple aspects of development and the Drosophila Scc2 homolog, Nipped-B, is required for long-range activation of these two genes. The effect of reducing cohesin dosage on cut expression is contradictory to the effect of reducing Nipped-B leading to the hypothesis that cohesin might function as an insulator factor interfering with long-range communication between the wing margin enhancer and the cut promoter, and that Nipped-B facilitates long-range gene activation by controlling cohesin dynamics (85). Of note, although heterozygous Nipped-B mutant flies had reduced cut and Ubx expression without detectable cohesion defects, the heterozygous Nipped-B null alleles only reduced Nipped-B mRNA by some 25%, and that further reduction to 50% wild-type levels by in vivo RNAi would be lethal; however, cohesion defects were not observed (84). Whole genome mapping of cohesin and Nipped-B chromatin binding sites in Drosophila has revealed that the loading factor Nipped-B co-localize with cohesin, suggesting that Nipped-B and cohesin still stay together on the chromatin after the loading process (86). These data revealed very different cohesin binding patterns in yeast and flies, and possibly very different targeting mechanisms which may be related to different roles for cohesin and Nipped-B either in cohesion formation or in gene regulation (87).

Recently, homozygous Drosophila mutants of Smc1 or Scc3 have shown defective axon pruning in the post-mitotic mushroom body (88, 89). The ecdysone steroid hormone receptor (EcR), an important factor for axon pruning, was found to have reduced protein expression in these mutants (89). Overexpression of EcR in the post-mitotic neurons in those knockout flies could rescue the pruning defect indicating that reduced EcR expression is primarily responsible for the pruning defect (89). The block of pruning could also be induced by knocking out another cohesin subunit Rad21 in the same neurons indicating a complete cohesin complex is essential for pruning event (88). As all of these above observations occurred in post-mitotic (non-dividing) cells, chromosome segregation and cell cycle regulation are clearly not involved. In addition, cholinergic neurons’ lack of Rad21 caused abnormal larva locomotion without obvious effects on mitosis (88). In other model organisms, such as zebrafish, the functions of Rad21 and Smc3 are needed for proper expression of runx1 and runx3 genes in early embryonic development (90). Genome-wide chromatin immunoprecipitation experiments in human and mouse cells have revealed co-localization of cohesin and CTCF, a zinc-finger protein with enhancer blocking/boundary activities (91–93). Knockdown of either cohesin or CTCF would influence genome-wide gene expression levels, and knockdown of cohesin alone resulted in dysregulated gene expression of CTCF targets (91, 92). Genome-wide gene expression and ChIP array studies in lymphoblastoid cell lines have revealed a highly conserved transcriptional profile in CdLS probands which tightly associates to cohesin binding status (94).

CdLS probands with NIPBL missense mutations are usually more mildly affected, whereas severely affected probands are much more likely to carry protein truncating alleles (38). In Drosophila, the expression of cut and Ubx is less affected in mutants with missense Nipped-B alleles similar to some CdLS-causing mutations than those with truncating or null alleles (95). Thus the milder effects of NIPBL missense mutations in humans could reflect milder effects on gene expression. Moreover, CdLS probands who carry heterozygous SMC1A or SMC3 mutations tend to be more mildly affected and have a much lower prevalence of overt structural defects (12). The mutations in SMC1A or SMC3 are all either missense or small in-frame deletions that preserve the protein reading frame, and cell lines from probands with cohesin subunit mutations lack cohesion defects and produce normal levels of SMC1A (4). More interestingly, the level of NIPBL mRNA was only 30% reduced in CdLS probands’ cell lines and heterozygous NIPBL knockout mice as well as in Nipped-B mutant flies (4). This slight impairment in NIPBL expression (<30%) or a single amino acid changes in SMC1A or SMC3 can lead to enormous effects on human development without considerable defects in sister chromatid cohesion. These findings further support the hypothesis that it is likely the cohesin’s role in regulating gene expression that is being impacted in CdLS and causative of the resulting phenotype. Several lines of evidence support this hypothesis. First, none of the CdLS-causing mutations in SMC1A is predicted to inactivate the ATPase activity, interfere with interactions with other cohesin subunits, or disrupt the coiled-coil arm (12). This is consistent with the normal sister chromatid cohesion seen in probands. Second, SMC1A is an X-linked gene and reported to escape X-inactivation (96); however, there are both male and female probands affected with heterozygous missense mutations, and no female proband was found to carry a loss-of-function mutation such as a nonsense or frame shift mutation (12). It is therefore presumed that SMC1A dose reduction is not likely to be the cause of CdLS, but rather the mutations are more likely to result in a slightly altered cohesin dynamic function that interferes with its ability to regulate (94). Third, naturally occurring SMC1A missense mutations in CdLS probands have been shown to have an enhanced DNA binding affinity than wild-type proteins in vitro. This is, to date, the most direct evidence that cohesin dynamics are changed by gene mutations (97).

CTCF is the only well characterized insulator protein in vertebrates. In addition to its enhancer blocking and barrier activities, CTCF can function as a transcription factor at a subset of genomic sites and can directly bind to RNA polymerase II, indicating a potential function in directly regulating gene transcription both positively and negatively (98, 99). Additionally, several known chromatin modifications, including DNA methylation, occur near CTCF binding sites. Formation of higher order chromatin structure was also mediated by CTCF (100). Cohesin and CTCF were found not only to significantly co-localize but also to be functionally associated with each other in the mammalian genome (91, 92, 101). It was therefore suggested that cohesin could regulate gene expression through multiple mechanisms that are quite similar to the regulatory mechanisms of CTCF. Cohesin and CTCF may act either cooperatively or independently (91, 92). As a result of these findings, it will be of interest to investigate if the phenotypic manifestations of CdLS could result from epigenetic alterations in the genome such as loss imprinting at specific loci.

Currently, 35% of clinically diagnosed CdLS probands do not carry an identifiable mutation in one of the three implicated genes. This would indicate that either new disease causing genes remain to be discovered or non-coding regions of the known genes require re-screening. Additional candidate genes would include the other components of the cohesin ring: RAD21 and STAG1/STAG2, or other proteins that interact with or regulate cohesin and modulate its chromatin association dynamics, for example, PDS5, WAPL and Sororin. Knockout animals for one of the two genes encoding a vertebrate Pds5 protein, Pds5b, were recently reported (102, 103). Mice with homozygous null Pds5b alleles die soon after birth and presented some congenital anomalies found in probands of the CdLS.

Cohesin pathway and other human disorders

Future directions

Cohesin’s role in mammalian development was not appreciated until genes involved in structural components and in the regulation of this complex were positionally cloned and found to be causative of multi-system disorders when disrupted. Since the identification of the first cohesin gene to be mutated in CdLS, additional developmental disorders and diseases have been implicated in this pathway (summarized in Fig. 2). The term ‘cohesinopathy’ has been used to describe human conditions due to mutations in cohesin subunits or regulatory proteins of cohesin complex. Pathogenic mutations in new genes involved in cohesinopathies are expected to be identified. Extra effort is needed to understand how cohesin functions in meiosis and human malignancy, and how cohesion pathway is involved in gene expression and epigenetic regulation.