|Home | About | Journals | Submit | Contact Us | Français|
The Additional sex combs (Asx) gene of Drosophila behaves genetically as an enhancer of trithorax and Polycomb (ETP) in displaying bidirectional homeotic phenotypes, suggesting that is required for maintenance of both activation and silencing of Hox genes. There are 3 murine homologs of Asx called Additional sex combs-like1, 2, and-3. Asxl1 is required for normal adult hematopoiesis; however its embryonic function is unknown. We used a targeted mouse mutant line Asxl1tm1Bc to determine if Asxl1 is required to silence and activate Hox genes in mice during axial patterning. The mutant embryos exhibit simultaneous anterior and posterior transformations of the axial skeleton, consistent with a role for Asxl1 in activation and silencing of Hox genes. Transformations of the axial skeleton are enhanced in compound mutant embryos for the Polycomb group gene M33/Cbx2. Hox a4, a7, and c8 are derepressed in Asxl1tm1Bc mutants in the antero-posterior axis, but Hox c8 expression is reduced in the brain of mutants, consistent with Asxl1 being required both for activation and repression of Hox genes. We discuss the genetic and molecular definition of ETPs, and suggest that the function of Asxl1 depends on its cellular context.
To maintain the determined state, cells must pass on their gene expression patterns to their daughter cells in the absence of the transient signal or asymmetric localization that initiated the expression pattern. This process is epigenetic, because the change in gene expression is not accompanied by a change in DNA sequence (Cavalli, 2002). Two groups of genes are required to maintain gene expression patterns in higher eukaryotes. Polycomb group (PcG) proteins silence repressed genes, whereas trithorax group (trxG) proteins maintain activation (Brock and Fisher, 2005). These maintenance proteins (MPs) were originally discovered in Drosophila, and each of these has mammalian homologs (Brock and Fisher, 2005) (Schwartz and Pirrotta, 2008). The best known targets of MPs are the Hox genes that are required for axial patterning (Wellik, 2007). Genome-wide binding assays show that hundreds of other loci important for development, cell differentiation, signaling pathways, and some PcG genes themselves, are also targets (Bernstein et al., 2006; Boyer et al., 2006; Lee et al., 2006; Negre et al., 2006; Schwartz et al., 2006; Tolhuis et al., 2006). Mutations in MP genes have pleiotropic functions in axial patterning, stem cell differentiation, chromosomal position-effects, X-inactivation and imprinting, sex determination, chromosome segregation, RNA interference, organogenesis, hematopoiesis, and oncogenesis (Sparmann and van Lohuizen, 2006).
PcG and trxG proteins associate in complexes that alter histone modification, nucleosome structure, or nuclear structure (Muller and Verrijzer, 2009; Simon and Kingston, 2009; Brock and Fisher, 2005). PcG and trxG proteins may act by poising transcription (Mendenhall and Bernstein, 2008). Classically, animals bearing mutations of PcG genes exhibit posterior transformations in the antero-posterior (AP) axis that are enhanced in double mutants with other PcG genes. Similarly, trxG genes exhibit anterior transformations that are enhanced in animals mutant for two trxG genes.
However, this tidy division of maintenance proteins into two categories has been confounded by the discovery of mutations in genes encoding maintenance proteins that exhibit simultaneous anterior and posterior transformations and that enhance the homeotic transformations of both trxG and PcG mutations. These genes have been termed Enhancers of Trithorax and Polycomb (ETPs) (Brock and Lohuizen, 2001; Gildea et al., 2000). The phenotypes of ETPs imply that these genes function in maintenance of both activation and silencing of Hox genes. It has been proposed that the terms PcG and trxG be reserved for proteins that are members of complexes with demonstrated activity on chromatin, and that other proteins be classified either as proteins that recruit these complexes to their targets, or as trxG and PcG cofactors (Grimaud et al., 2006). Accordingly, ETPs could be subunits in both trxG and PcG complexes, recruit both trxG and PcG complexes, or be cofactors for both trxG and PcG complexes.
Additional sex combs (Asx) is an ETP in Drosophila that exhibits simultaneous anterior and posterior homeotic transformations (Breen and Duncan, 1986; Sinclair et al., 1992), and enhances homeotic phenotypes of both PcG (Sinclair et al., 1992) and trxG mutations (Milne et al., 1999). Mammals have three Asx homologs, termed Additional sex combs-like 1 (Asxl1), Asxl2, and Asxl3 (Fisher et al., 2003; Fisher et al., 2006; Katoh, 2003; Katoh, 2004). Asx-like proteins contain only two features conserved with Asx: an amino terminal region termed ASXH/ASXM containing two putative nuclear receptor co-regulator binding (NR box) motifs; and a carboxy-terminal PHD domain (Fisher et al., 2003);(Katoh, 2003). Additional sequence features are conserved within mammalian but not Drosophila Asx homologs, including three additional carboxy-terminal NR boxes (Fisher et al., 2006; Katoh, 2003; Katoh, 2004).
We recently generated a loss-of-function Asxl1 mutant mouse model, Asxl1tm1Bc, and used it to demonstrate that mouse Asxl1 is required for normal adult hematopoiesis (Fisher et al., submitted). Here we employ the Asxl1tm1Bc mouse mutant to determine if Asxl1 demonstrates conserved ETP function. As predicted, mutant embryos exhibit simultaneous anterior and posterior homeotic transformations of the AP axis, consistent with our observations of Hox gene misregulation along the AP axis. We show that Asxl1 is required for activation and repression of Hox genes. We discuss the definition of ETP in mammals, and propose that Asx-like protein function depends on cellular context.
The null Asxl1tm1Bc mutant allele was generated using a replacement targeting vector which inserted a PGK promoter-driven neomycin resistance cassette into exon 5 of Asxl1 (Fisher et al., submitted), which is upstream of the conserved ASXH and PHD domains in Asxl1 (Fisher et al., 2006). Genotyping was routinely performed on DNA from adult tail tips, liver from newborns, or yolk sac from embryos, by PCR as described (Fisher et al., submitted). All experiments with animals were performed in accordance with the regulations established by the Canadian Council on Animal Care, and all protocols were reviewed by the UBC Animal Care Committee.
Cbx2tm1Ykf (M33Cterm) mice (Katoh-Fukui et al., 1998b) on a C57BL/6J background for at least Nll generations were crossed to mice heterozygous for Asxl1 to generate double heterozygous mice. Asxl1+/tm1Bc; Cbx2+/tm1Ykf mice were interbred to generate other mutant allele combinations. Genotyping was routinely performed on DNA from adult tail tips, liver from newborns, or yolk sac from embryos, by triple primer PCR as follows. Primer M33A (5’ GTAGCCAAGCCAGAGCTGAA 3’) and antisense primer M33B (5’ AGAGGCCTCTTTGGTGTGG 3’) amplify a 200 bp fragment of the wild-type allele, and primers PGKPr (5’ CCGCTTCCATTGCTCAGCGGT 3’) and M33B amplify a 325 bp fragment of the Cbx2tm1Ykf mutant allele (data not shown).
Whole mount skeletons of newborns were prepared and stained with Alizarin Red and Alcian Blue 8GX as described (van der Lugt et al., 1994).
Experimental genotype ratios were compared to expected ratios using the Chi-squared test. Comparison of means between groups was done using the Student’s t-test, and significance was defined as p values <0.05.
In situ hybridization to E10.5 mouse embryos was carried out as previously described (Belo et al., 1997). The Hoxa4 riboprobe was obtained from Yoshihiro Takihara (Takihara et al., 1997). The Hoxa7 riboprobe was reverse transcribed from clone 7514185 obtained from IMAGE consortium. The Hoxc8 riboprobe has been described previously (Hanson et al., 1999).
Mice heterozygous for Asxl1tm1Bc exhibited Mendelian segregation ratios up to and just following birth. However, by postnatal day 21 (P21), only 28% of the expected number of Asxl1tm1Bc/ tm1Bc mice was observed (Table 1). The majority of deaths occur within 1–3 days after birth. Most ill or dead Asxl1tm1Bc/ tm1Bc newborns and young pups had suckled and appear outwardly normal. The cause of death remains unclear as the mice are not anaemic (Fisher et al. submitted), and have normal lung, kidney, liver, and heart histology (data not shown). Adult Asxl1tm1Bc/ tm1Bc mice showed a significant reduction in body weight, splenomegaly, as well as a reduction in thymus and testis weight compared to wild-type mice (Fig. 1), but no change in weight of the liver, lungs, heart, or kidneys (data not shown). Four of 6 males and 3/3 female Asxl1tm1Bc/ tm1Bc mice tested were fertile. We occasionally noticed absence of one ovary in Asxl1tm1Bc/ tm1Bc mutant females (data not shown).
We examined Asxl1tm1Bc newborns (P1) for skeletal defects (Fig. 2, Table 2). The anterior arch is normally associated with only the C1 vertebra via a cartilaginous connection (Fig. 2A). In 13/14 of Asxl1tm1Bc/tm1Bc mice the anterior arch was larger than in wild-type mice, and is additionally associated with the C2 vertebra via an ectopic cartilaginous process (black arrow in Fig. 2B), consistent with a C2 to C1 anterior homeotic transformation. In 2/14 of Asxl1tm1Bc/tm1Bc mice the first cervical (C1) vertebrae was split laterally, similar to the break in ossification observed between the exoccipital and supraoccipital bones of the skull (Fig. 2B, grey arrow). This is interpreted as a C1 to occipital bone anterior transformation. An ectopic complete rib (2/14) or partial rib bud (3/14) on the C7 vertebra was seen in Asxl1tm1Bc/tm1Bc (Fig. 2E), but not in wild-type mice (Fig. 4D), interpreted as a C7 to T1 posterior homeotic transformation. We observed small holes in the xiphoid process in 7/14 Asxl1tm1Bc/tm1Bc mice, but not in wild type mice (data not shown). Asxl1tm1Bc/tm1Bc mice (9/14) lacked the 13th thoracic (T13) rib on one or both sides, or had rib buds instead of complete ribs on one or both of the T13 vertebrae (Fig. 2G), indicating posterior transformation of thoracic T13 towards a lumbar (L1) vertebral identity. We did not observe sacral region transformations or defects in patterning of the appendicular skeleton in Asxl1tm1Bc/tm1Bc mutants.
In Drosophila, ETP mutations enhance phenotypes of PcG mutations. Therefore we investigated the effects of a mutation in the PcG gene Cbx2/M33 on Asxl1tm1Bc phenotypes (Fig. 2). At day P21 the frequency of surviving Asxl1+/+; Cbx2+/ tm1Ykf mice (63%) agreed with that of Cbx2+/ tm1Ykf (68%) mutants found previously (Katoh-Fukui et al., 1998b). Similarly, survivorship frequencies of Asxl1+/ tm1Bc;Cbx2+/+ mice at P1 and P21 were consistent with the frequency of single Asxl1+/ tm1Bc mutants (Table 1 and Table 3). Therefore background effects in the compound mutant were not significantly altering survivorship rates due to the Asxl1tm1Bc mutation. However, by day P21, the genetic background does appear to affect viability of the Cbx2 tm1Ykf / tm1Ykf mutants, as we found this genotype at 6% frequency, versus 60% seen in the original study (Katoh-Fukui et al., 1998a).
Asxl1+/ tm1Bc; Cbx2+/ tm1Ykf trans-heterozygotes showed enhanced lethality compared to single mutants at P21 (Table 3). Strikingly, no Asxl1 tm1Bc /tm1Bc; Cbx2 tm1Ykf / tm1Ykf double homozygous mutant mice were seen at P21. Nor were double homozygous mutants found alive at P1 or E10.5–11.0, showing that the lethal phase occurs prior to E10.5–11.0. Consistent with this, four resorbed decidua were observed at E10.5–11.
We examined skeletal defects in Asxl1 tm1Bc; Cbx2 tm1Ykf compound mutant mice at P1 to detect enhancement of homeotic phenotypes (Fig. 2, and Table 2). In the Asxl1+/tm1Bc; Cbx2+/ tm1Ykf trans-heterozygotes (n=11), 82% exhibited a C2 to C1 transformation (Fig. 4C), compared to 32% of Asxl1+/tm1Bc mutants (n=19) and 67% of Cbx2+/ tm1Ykf mutants (n=6). The two Asxl1tm1Bc/tm1Bc; Cbx2+/ tm1Ykf mice analyzed both showed a C2 to C1 transformation. As well, 64% of Asxl1+/tm1Bc; Cbx2+/ tm1Ykf trans-heterozygotes exhibited sternal defects, including an abnormal xiphoid process and defective ossification in the 5th sternebra compared to only 33% of Cbx2+/ tm1Ykf mutants and 5% of Asxl1+/tm1Bc mutants. The two Asxl1tm1Bc/tm1Bc; Cbx2+/ tm1Ykf mice exhibited an offset joining of ribs to the sternum referred to as a ‘crankshaft’ sternum, and one of the mice had only 6 vertebrosternal ribs attached to the sternum on the left side instead of 7 as occurs in a wild type mouse (Fig. 2H, I). These two mice also had reduced or absent ossification centers (vertebral bodies) in some of the cervical vertebrae on the ventral side (not shown). A similar phenotype was observed in an Asxl1+/tm1Bc; Cbx2+/ tm1Ykf mouse found dead shortly after birth. In addition, this mouse also had a split scapula, and a fusion between the atlas and the exoccipital bones of the skull (Fig. 2C). Similar sternal phenotypes have been observed in Cbx2tm1Cim/tm1Cim (van der Lugt et al., 1996) and Cbx2tm1Ykf/ tm1Ykf mice (Katoh-Fukui et al., 1998b) but not in Cbx2 or Asxl1 heterozygous mutant mice, indicating enhancement of these phenotypes in the Asxl1;Cbx2 mutant trans-heterozygotes. The penetrance of T13 to L1 transformations was increased in the Asxl1+/tm1Bc; Cbx2+/ tm1Ykf trans-heterozygotes (45%) compared to 26% in Asxl1+/ tm1Bc and 17% in Cbx2+/ tm1Ykf mutants. Overall, our results show an enhancement of bidirectional skeletal homeotic transformations in compound Asxl1;Cbx2 mutants compared to single Asxl1 and Cbx2 mutants.
The skeletal transformations observed in Asxl1tm1Bc mutants are consistent with mis-expression of Hox genes. Therefore we analyzed embryos for changes in the pattern of mRNA expression of Hoxa4, Hoxa7 and Hoxc8 in mutant and wild-type E10.5 embryos by whole mount in situ hybridization (Fig. 3). These Hox genes show ectopic expression in rae28, bmi-1, mel-18 or M33 mutants, and reduced expression in Mll mutants, and thus provide a basis for comparison to the results obtained in Asxl1tm1Bc mutants. In wild-type embryos, the anterior boundary of Hoxa4 ectodermal expression is detected at the upper cervical spinal cord and spinal ganglia, and expression levels decrease caudally until it is undetectable at the thoracic level. In mesoderm, Hoxa4 is expressed anterior to prevertebra 1, and is reduced posterior to prevertebra 6 (Toth et al., 1987). In 5/7 homozygous Asxl1tm1Bc embryos, the anterior boundaries of both ectodermal and mesodermal expression are shifted anteriorly by 1 or 2 somites relative to wild-type (Fig. 3; blue and red arrows respectively). Two of these 5 mutants also show Hoxa4 expression in caudal mesoderm where Hoxa4 expression is never detected in wild type.
In wild-type mice, ectodermal Hoxa7 expression begins at the C5 somite and ends at the 4th sacral (S4) somite, whereas mesodermal expression begins at T3 and ends at T13 (Dressler and Gruss, 1989), (Puschel et al., 1991). Anterior boundaries of Hoxa7 in ectoderm and mesoderm are shifted rostrally in 8/8 homozygous Asxl1tm1Bc mutants by 1 somite to C4 and T2 respectively (Fig. 3). These results show that Hoxa4 and Hoxa7 are derepressed in Asxl1tm1Bc mutants.
The effect of Asxl1tm1Bc mutation on Hoxc8 expression in embryos is more complex. Hoxc8 expression is shifted anteriorly in ectoderm and mesoderm of 8/8 homozygous Asxl1tm1Bc mutants by 1 somite (Fig. 3). The overall intensity of staining in the ectoderm and mesoderm of mutant embryos is much greater than in wild-type embryos, and expression of Hoxc8 can be detected more caudally than in wild type, consistent with a role for Asxl1 in repressing Hoxc8 in ectoderm and mesoderm. However, in the developing brain, Hoxc8 expression is consistently reduced in mutant embryos compared to wild type embryos where it is strongly expressed (Kwon et al., 2005), suggesting that Asxl1 is required for activation of Hoxc8 in the brain. The mutant embryos cannot simultaneously be overstained for expression in somites and spinal cord, and understained in the brain, so the differences between the Asxl1tm1Bc mutant and wild-type embryos do not reflect technical problems with staining.
The homeotic transformation phenotypes of the Asxl1tm1Bc mutants are mild relative to that of most other PcG mutants in mice. The hybrid genetic background deceases the severity of the Asxl1tm1Bc phenotypes reported here, as further backcrossing to C57BL/6J has resulted in complete embryonic lethality of homozygous mutants (C.F, unpublished). Similar effects have been seen in mice mutant for Mel18/Pcgf2 upon further backcrossing to a consistent background (Suzuki et al., 2002). Murine Asxl2Gt(AQ0356) gene-trap mutants on a mixed genetic background also exhibit incompletely penetrant perinatal lethality (Baskind et al., 2009). As Asxl1 and Asxl2 have similar, broad expression patterns in mouse embryos and stem cells (Fisher et al., 2006), redundancy of Asxl2 may mask early functions of Asxl1. Embryonic stem cell expression of other MPs correlates with early developmental function (Ayton et al., 2001; Donohoe et al., 1999; O'Carroll et al., 2001; Voncken et al., 2003). The generation of compound Asxl1; Asxl2 mutant mice will allow testing of potential redundancy, and of early developmental roles for Asxl genes.
Anterior transformations are usually associated with loss of function Hox mutations, and thus the C2 to C1 and C1 to occipital bone transformations observed in Asxl1tm1Bc mice suggest that Asxl1 is required for Hox activation. Anterior homeotic transformations in the C1 and C2 cervical vertebrae are also seen in Hoxb4, Hoxd3, and Hoxd4 mutant mice (Condie and Capecchi, 1994; Horan et al., 1995). However, in some cases the complexity of the Hox code means that loss of function Hox mutations can cause posterior transformations, as in the case of C7 to T1 thoracic vertebral transformations observed in Hoxa4 (Horan et al., 1994), Hoxa5 (Jeannotte et al., 1993), or Hoxa6 knockout mice (Kostic and Capecchi, 1994). Therefore the C7 to T1 transformation in Asxl1tm1Bc mutants could also be consistent with an activating role for Asxl1.
In most cases, posterior homeotic transformations are associated with derepression of Hox expression (Pollock et al., 1995). This is the case for the transformations of C7 to T1, and T13 to L1 vertebrae observed in Asxl1tm1Bc mutant embryos. The observations that Hoxa4, Hoxa7 and Hoxc8 are derepressed, and that Asxl1tm1Bc mutations enhance the phenotypes of Cbx2 mutations are also consistent with a role for Asxl1 in silencing of Hox genes. Together, the anterior and posterior transformations of Asxl1tm1Bc mutants suggest that Asxl1 is an ETP. Consistent with this conclusion, we observed a role for Asxl1 in both silencing and activation of Hoxc8.
It may be that ETP phenotypes of Asxl1 are indirect consequences of Asxl1-mediated gene expression. The homeotic transformations we observe could arise because Asxl1 is a co-regulator of RAR in the cervical vertebrae. Mice in which retinoid receptor function is impaired by gene targeting (Mark et al., 2006) or by dominant negative mutations (Yamaguchi et al., 1998), exhibit anterior transformations of cervical vertebrae. This role for Asxl1 is also consistent with the enhancement of anterior phenotypes observed in Asxl1tm1Bc;Cbx2tm1Ykf double mutants because Cbx2 antagonizes the RA pathway and functions in establishing the early spatiotemporal sequence of activation of a subset of Hox genes (Bel-Vialar et al., 2000). However, posterior derepression of Hox genes, and posterior transformations are inconsistent with a direct consequence of Asxl1 acting through the RAR, because defects in RAR signaling cause anterior transformations (Daftary and Taylor, 2006).
Indirect phenotypes could also arise because Asx and some Drosophila PcG genes positively regulate each other’s expression. Cbx2 is a known member of the mammalian Polycomb group Repressing Complex 1 family (Simon and Kingston, 2009) yet mutants exhibit both anterior and posterior transformations (Core et al., 1997; Katoh-Fukui et al., 1998b). In addition, the Sex combs extra homologs Ring1/Ring1a and Rnf2/Ring1b show opposing loss-of-function phenotypes in mutant mice despite both gene products being subunits of PRC1. Homozygous Ring1tm1Mvi mutants exhibit anterior transformations (del Mar Lorente et al., 2000), whereas Rnf2tm1Hko(Ring1bred) hypomorphic mutants exhibit posterior transformations (Suzuki et al., 2002). These observations suggest the need for caution when inferring molecular function from phenotypic analysis (Grimaud et al., 2006).
An alternative explanation for the ETP phenotypes of Asxl1 mutations is that Asxl1 function is context-dependent. In this model, Asxl1 is required in the trunk to silence Hoxa4, a7, and c8, but is required in the brain to activate Hoxc8. In Drosophila, Asx is unusual among PcG genes because Asx mutants exhibit tissue-specific effects (Soto et al., 1995) even though expression is ubiquitous (Sinclair et al., 1998), which would be consistent with context-dependent function. In support of this interpretation, n human HeLa cells, mouse Asxl1 tethered to GAL4 binding sites activates a heterologous thymidine kinase promoter driving a luciferase reporter, and also increases RAR-dependent-activation of the reporter in transient transfection assays; however, in other mammalian cell types, Asxl1 acts as a RAR corepressor (Cho et al., 2006). In addition, Asxl1 is a member of a repressive complex containing Histone H1.2 (Kim et al., 2008).
Interestingly, Asxl2Gt(AQ0356) mutants exhibit anterior transformations of thoracic vertebrae at low penetrance, and highly penetrant posterior transformations of sacral vertebrae (Baskind et al., 2009). These mutant phenotypes are consistent with expectations for an ETP although as they are based upon a single gene-trap mutant model they should be confirmed in an independent Asxl2 null mutant line. Since sacral region transformations were never seen in Asxl1tm1Bc mice, this suggests that a functional division has occurred whereby Asxl1 predominantly regulates more anterior cluster Hox genes, while Asxl2 function is more focused on the posterior cluster Hox genes, albeit with substantial overlap in regulation of mid-cluster Hox genes. In addition, there is a pattern in both Asxl1 and Asxl2 mutants whereby the more rostrally located axial vertebrae undergo anterior transformation, while the more caudally located vertebrae demonstrate posterior transformations. This could imply a mechanistic role of Asx-like proteins in the activation of more rostral Hox genes, but in repression of more caudally located Hox genes, within the functional domain of each respective Asx-like family member. We propose that Asxl1 function is determined by its interacting partners or chromatin environment to bring about changes in gene regulation. Identifying the target genes and interaction partners of Asxl1 and other Asx-like proteins, and how they bring about changes in gene regulation remains an important challenge for the future.
This research was supported by grants from the Canadian Institutes of Health Research and the N.I.H (USA) to H.W.B., and from the Terry Fox Foundation to R.K.H. J.L.H. was supported by the N.I.H. (R01-CA-0078815) and a SCOR grant from the Leukemia and Lymphoma Society of America. C.L.F. was supported by a Medical Research Council of Canada studentship. C.D.H. was supported by a CIHR New Investigator award and is a MSFHR Scholar.
We thank Toru Higashinakagawa for sending Cbx2tm1Ykf mutant mice used in these studies. Patty Rosten provided technical advice, and the genotyping assay for Cbx2tm1Ykf mutant mice. We thank Rewa Grewal and Pol Gomez for technical assistance.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.