The vertebrate limb is formed from a limb bud, a protrusion from the main body, which elongates and undergoes morphogenesis to construct a stereotypical limb structure. The development of the limb requires coordinated cell proliferation and differentiation, which depend on highly regulated actions of signaling pathways, such as those initiated by Wnt, Shh, Bone morphogenetic protein (BMP), and Fibroblast growth factor (FGF) (Johnson and Tabin, 1997; Capdevila and Izpisua Belmonte, 2001; Niswander, 2003; Tamura et al., 2008; Duboc and Logan, 2009; Towers and Tickle, 2009; Zeller et al., 2009).
Wnt signaling in particular is known to have multiple roles in the early process of vertebrate limb development, including driving downstream genetic programs and coordinating the functions of signaling centers (Yang, 2003; Grigoryan et al., 2008). Wnt/β-catenin signaling triggers Fgf10 expression in the lateral plate mesoderm during limb initiation in chick and zebrafish (Kawakami et al., 2001; Ng et al., 2002). Then, Fgf10-dependent Wnt signaling in the overlying surface ectoderm regulates formation of the apical ectodermal ridge (AER), a specialized epithelial structure required for distal outgrowth and proximal-distal patterning (Mariani et al., 2008) and for Fgf8 expression in the AER (Kengaku et al., 1998; Kawakami et al., 2001; Ng et al., 2002; Barrow et al., 2003; Soshnikova et al., 2003). Wnt7a from the dorsal ectoderm induces Lmx1b in the underlying mesenchyme and specifies dorsal fate of the limb mesenchyme (Riddle et al., 1995; Vogel et al., 1995; Chen et al., 1998).
Wnt7a from the dorsal ectoderm also plays a role coordinating the function of signaling centers by maintaining Shh expression in the posterior mesenchyme, a region called the zone of polarizing activity (Parr and McMahon, 1995; Yang and Niswander, 1995). Wnt7a−/− mouse limb bud and dorsal ectoderm-removed chick limb bud fail to maintain Shh expression in the posterior mesenchyme. Implanting Wnt7a-expressing cells into dorsal ectoderm-removed limb bud rescued Shh expression. A previous study in the chick limb bud suggested that the Wnt7a-maintenance of Shh is mediated by a specific frizzled receptor (Kawakami et al., 2000). Shh has multiple roles during limb development (Zeller et al., 2009); Shh instructs anterior-posterior digit pattern formation (Riddle et al., 1993) and promotes proliferation of digit progenitors (Towers et al., 2008; Zhu et al., 2008), and is essential for normal development of the autopod (Chiang et al., 1996). Since Wnt signaling is involved in such a variety of processes during early processes of limb development, its activity needs to be under tight control, and, thus, alterations in Wnt signaling disrupts normal development of the limb (Grigoryan et al., 2008). However, it is not well understood how each tissue elicits a proper level of Wnt signaling in order to regulate the development of the limb.
We have recently demonstrated that HMGB2, a chromatin factor, facilitates Wnt/β-catenin signaling both in vitro and in adult articular cartilage (Taniguchi et al., 2009a). HMGB2 is a member of the high mobility group box (HMG) family, which are nuclear proteins, characterized by two basic HMG-box domains. Among four HMGB proteins, HMGB1, 2, and 3 have an additional long acidic C-terminal tail (Bustin, 1999). Biochemical studies have been performed mainly with HMGB1 and HMGB2, and demonstrated that HMGB1 and HMGB2 bind to DNA without sequence specificity (Bianchi and Agresti, 2005; Grasser et al., 2007; Stros et al., 2007). HMGB1 and HMGB2 have a role in formation of nucleoprotein complexes through altering chromatin structures that allow for binding of other factors (Pil and Lippard, 1992; Paull et al., 1993) and facilitating diverse DNA modifications (Agresti and Bianchi, 2003; Stros, 2010). They are also known to regulate various activities, such as transcription, replication, and DNA repair (Bianchi and Agresti, 2005).
Given that Hmgb genes are involved in these fundamental cellular processes, it was expected that they play critical roles in animal development and physiology. The in vivo function of Hmgb genes in a mammalian model has been studied by expression analysis and mutant mouse phenotype analysis. During embryonic development, Hmgb1 is reported to be ubiquitously and highly expressed (Pauken et al., 1994), and Hmgb2 is also known to be highly and widely expressed by Northern blot analysis (Ronfani et al., 2001). Contrary to this, expression of Hmgb3 is only demonstrated by reverse-transcription-polymerase chain reaction of whole embryos (Vaccari et al., 1998), and expression of Hmgb4 has not been studied. In adult tissues, Hmgb1 is reported to have ubiquitous and high level expression by whole tissue extract assays (Mosevitsky et al., 1989; Muller et al., 2004); however, other Hmgb genes show restricted expression. Hmgb2 is highly expressed in thymus and testis (Ronfani et al., 2001) as well as articular cartilage surface, which contains a progenitor cell population (Taniguchi et al., 2009b). Hmgb3 is highly expressed in hematopoietic stem cells in the bone marrow (Nemeth et al., 2003), and Hmgb4 expression is limited to testis (Catena et al., 2009). These expression analyses suggest that Hmgb1 may have roles broadly in embryonic development and in adult tissues, and other Hmgb genes may have redundant functions depending on the stages of embryos and adults, as well as specific tissues.
Functional studies of Hmgb genes have been carried out with targeted mutations in mice. Hmgb1−/− mice are born without significant morphological defects, but die within a day by hypoglycemia (Calogero et al., 1999), which is caused by insufficient activation of a glucocorticoid receptor. Hmgb2−/− mice show defects in spermatogenesis (Ronfani et al., 2001) as well as failure to maintain articular cartilage homeostasis in adults (Taniguchi et al., 2009a, b). Hmgb3−/− mice exhibit erythrocythemia (Nemeth et al., 2005). These studies clarify that Hmgb1, 2, 3 are not required for embryonic development, except that Hmgb1−/− limb long bones show delay in endochondral ossification (Taniguchi et al., 2007). Normal embryonic development in these mutant mice is likely due to functional compensation based on their structural similarity. This is particularly likely the case for Hmgb1 and Hmgb2, given their high expression in embryos and shared biochemical characteristics.
In order to address the functional redundancy of Hmgb1 and Hmgb2 during embryonic development and to test whether Hmgb genes modulate Wnt/β-catenin signaling and/or other signaling pathways in the developing limb, we generated double mutants of Hmgb1; Hmgb2. Consistent with the hypothesis of functional redundancy, Hmgb1; Hmgb2 double null mutants arrested at E9.5. Further analysis suggests that the functions of Hmgb1 and Hmgb2 are not completely equivalent in the development of the autopod. Analysis of the phenotype in the developing limb revealed a lack of the most posterior digit, digit 5, in the Hmgb1−/−; Hmgb2+/− forelimb, and this is associated with downregulation of Shh expression and Wnt and BMP target genes. Further analysis with zebrafish embryos also supports enhancement of Wnt/β-catenin signaling by hmgb genes. Thus, our data suggest that Hmgb1 and Hmgb2 are required for maintaining a proper level of Shh expression in the limb through enhancing Wnt/β-catenin signaling, and further suggest that Shh, Wnt, and BMP signaling at the posterior region regulate the development of digit 5 in the forelimb.
Combined Activities of Hmgb1 and Hmgb2 Are Required for the Development of Digit 5 in the Mouse Forelimb
Previous studies have shown that Hmgb1−/− embryos show delayed endochondral ossification in long bones during limb development without other morphological defects (Taniguchi et al., 2007). Hmgb2−/− mice did not exhibit any defects in limb development (Ronfani et al., 2001). Based on their structural conservation as well as ubiquitous and high level expression in the developing limb bud (see Supp. Fig. S1), we hypothesized that functional redundancy between Hmgb1 and Hmgb2 might allow for development of normal morphology in Hmgb1−/− and Hmgb2−/− limbs. In order to address this possibility, we generated Hmgb1; Hmgb2 double mutant embryos. At E10.5, we obtained 7 Hmgb1−/−; Hmgb2−/− embryos out of 257 embryos (2.7%); however, they were arrested at E9.5 (Table 1). At E11.5 and E12.5, we did not obtain any Hmgb1−/−; Hmgb2−/− embryos out of 66 embryos collected. This suggests that Hmgb1 and Hmgb2 have redundant functions and are required for development of the mouse embryo beyond E9.5.
Table 1. Summary of Number of Hmgb1; Hmgb2 Embryos Collected at E10.5a
KO, Het, and WT indicate -/-, +/-, and +/+, respectively.
In order to address their function in limb development, we examined limb skeletons of an allelic series of Hmgb1; Hmgb2 at E13.5–16.5 (Fig. 1). We found that Hmgb1−/−; Hmgb2+/− embryos showed a defect in the development of the most posterior digit, digit 5, in the forelimb (Fig.1D), while we did not observe significant defects in other areas of the body. Based on the morphology of remaining elements, the lost skeletal elements are the phalanges and metacarpal for digit 5, while other elements appeared to be normal (Fig.1H). This phenotype was observed only in the forelimb but not in the hindlimb (Fig.1I–P). Other allelic combinations either showed no skeletal phenotype or showed the same defect at very low frequencies (Table 2). This result suggests that Hmgb1 and Hmgb2 are functionally redundant for limb development, and that the function of Hmgb1 is more important than that of Hmgb2 for the development of digit 5 in the forelimb.
Table 2. Summary of Skeletal Phenotype at E13.5–16.5a
Number of embryos
Number with the phenotype
Phenotype is defined as lack of digit 5 in the forelimb.
KO, Het, and WT indicates -/-, +/-, and +/+, respectively.
Early Onset of the Digit 5 Defect in the Hmgb1−/−; Hmgb2+/− Mutant Forelimb Bud
To further characterize the loss of digit 5 in the Hmgb1−/−; Hmgb2+/− mutant embryo, we examined cartilage formation at earlier stages as well as region-specific molecular markers in the developing forelimb bud. At E13.5, the loss of digit 5 was evident by Alcian Blue staining (Fig. 2A, B). This defect was detectable even at an earlier stage, E11.5. Chondrogenic condensation for digits 1–4 was visible by detecting Sox9, while the posterior region including the digit 5 primordia was lacking in the Hmgb1−/−; Hmgb2+/− forelimb bud (Fig. 2C, D). Correlating with normal development of the anterior skeletal elements shown in Figure1, Pax9 expression, which marks the anterior region of the limb bud, was normal in the Hmgb1−/−; Hmgb2+/− forelimb bud (Fig. 2E, F). Hoxd13, which marks the autopod region, is also detected, suggesting normal specification of the autopod region. A slight expansion in the anterior region where Hoxd13 is not expressed, as well as the disappearance of the most posterior Hoxd13-expressing structure, suggests a slight effect on the anterior-posterior patterning by this stage (Fig. 2G, H). These results demonstrate that the posterior autopod region is specifically affected in the Hmgb1−/−; Hmgb2+/− forelimb bud by E11.5, and suggest that molecular alterations that cause the phenotype have taken place in earlier stages.
Late Onset and Reduced Level of Shh Signaling in the Hmgb1−/−; Hmgb2+/− Limb Bud
The lack of digit 5 is reminiscent of downregulation of Shh signaling (Zhu et al., 2008). Thus, we first examined the expression of Shh. Normal Shh expression in the posterior mesenchyme of the forelimb bud starts at E9.75 (Echelard et al., 1993; Charite et al., 2000). At E10.0 (28 somite stage), we did not detect Shh expression in the Hmgb1−/−; Hmgb2+/− forelimb bud, while Shh is expressed in the control limb bud (Fig. 3A, B). Consistent with this, the expression of Gli1, a target of Shh signaling in the Hmgb1−/−; Hmgb2+/− forelimb bud at 27 and 28 somite-stage embryos was significantly low compared to control embryos (Fig. 3C, D). These results indicate that the onset of Shh expression is delayed in the Hmgb1−/−; Hmgb2+/− forelimb bud.
At E10.5, we detected Shh expression in the posterior mesenchyme in the Hmgb1−/−; Hmgb2+/− limb bud, though it was clearly lower than the control limb bud (Fig. 3E, F). Consistent with this, Patched1 (Ptc1), another target of Shh signaling (Ingham and McMahon, 2001), and Gremlin1, whose maintenance requires Shh signaling (Nissim et al., 2006; Benazet et al., 2009) are downregulated (Fig. 3G–J). Furthermore, CyclinD1, which is known to be regulated by both Shh signaling and Wnt/β-catenin signaling (Shtutman et al., 1999; Tetsu and McCormick, 1999; Towers et al., 2008), is also expressed at a lower level (Fig. 3K, L). These results demonstrate that Shh signaling is initiated with a delay and is not activated to the normal level in the Hmgb1−/−; Hmgb2+/− forelimb bud.
Normal Activity of AER-FGF and Wnt7a
Signaling pathways in the developing limb bud form feedback loops to maintain the activity of each other. Shh expression in the posterior mesenchyme is maintained by FGFs emanating from the AER (Laufer et al., 1994; Niswander et al., 1994; Lewandoski et al., 2000; Moon and Capecchi, 2000) and Wnt7a emanating from the dorsal ectoderm (Parr and McMahon, 1995; Yang and Niswander, 1995). In order to address if downregulation of Shh expression at E10.5 is caused by downregulation of AER-FGF signal and/or Wnt7a from the dorsal ectoderm, we examined their activity in the Hmgb1−/−; Hmgb2+/− forelimb bud.
We first examined Fgf8 expression in the AER because Fgf8 is the central factor for limb development among four Fgf genes expressed in the AER (Mariani et al., 2008). In the Hmgb1−/−; Hmgb2+/− forelimb bud, Fgf8 is expressed similar to control limb bud, which contrasts with the lower level of Shh expression (Fig. 4A–D). Since the sum of AER-FGF function is important for normal limb development, we next monitored if AER-FGF activity is altered by examining expression of Fgf10 and Mkp3. It has been demonstrated that expression of these genes depends on FGF activity from the AER (Ohuchi et al., 1997; Eblaghie et al., 2003; Kawakami et al., 2003; Mariani et al., 2008). We detected expression of both genes comparable to control limb bud (Fig. 4E–H). To further evaluate AER-FGF activity, we examined phosphorylated ERK1/2 (pERK1/2) on sectioned samples. A previous study has demonstrated that pERK1/2 in developing mouse limb buds depends on FGF signaling (Corson et al., 2003). We detected a normal level of pERK1/2 in the control and the Hmgb1−/−; Hmgb2+/− forelimb bud (Fig. 4I, J). The pERK1/2-positive cell layers were 10.4 ± 0.89 (n=5) in control and 10.2 ± 0.82 (n=5) in the Hmgb1−/−; Hmgb2+/− limb bud. These data suggests that AER-FGF activity is not significantly affected in the Hmgb1−/−; Hmgb2+/− forelimb bud.
Dorsal ectoderm produces WNT7a, which has been shown to be required for maintenance of Shh expression (Parr and McMahon, 1995; Yang and Niswander, 1995). To examine if reduced Shh expression in the Hmgb1−/−; Hmgb2+/− forelimb bud at E10.5 is due to lowered Wnt7a activity, we monitored expression of Lmx1b, a gene regulated by WNT7a from the dorsal ectoderm (Riddle et al., 1995; Vogel et al., 1995). We detected normal expression of Lmx1b (Fig. 4K–N), suggesting that reduced Shh expression in the Hmgb1−/−; Hmgb2+/− limb is unlikely due to reduction of Wnt7a function from the dorsal ectoderm.
Taken together, these results suggest that the lower level of Shh expression in the Hmgb1−/−; Hmgb2+/− forelimb bud is not caused by alteration in AER-FGF or WNT7a from the dorsal ectoderm.
Reduced Wnt/β-Catenin Signaling and BMP Signaling in the Posterior Region of the Hmgb1−/−; Hmgb2+/− Forelimb Bud
Our recent analysis has demonstrated that Hmgb2 enhances Wnt/β-catenin signaling (Taniguchi et al., 2009a). The apparently normal expression of Lmx1b in the Hmgb1−/−; Hmgb2+/− forelimb bud suggests normal Wnt7a signaling in the dorsal mesenchyme (Fig. 4K–N). There remains the possibility that uncharacterized Wnt7a signaling via β-catenin in the posterior mesenchyme is affected in the Hmgb1−/−; Hmgb2+/− forelimb bud, while Wnt7a signaling through a β-catenin-independent pathway maintains normal expression of Lmx1b (Kengaku et al., 1998). To address this, we examined expression of Msx2. While Msx2 is a known target of BMP signaling in the limb (Pizette and Niswander, 1999; Khokha et al., 2003; Michos et al., 2004), studies have further demonstrated that in the presence of BMP signaling, Wnt/β-catenin signaling directly regulates the Msx2 promoter via the Lef1-β-catenin pathway and facilitates Msx2 expression (Willert et al., 2002; Hussein et al., 2003). In the Hmgb1−/−; Hmgb2+/− forelimb bud, expression of Msx2 in the posterior mesenchyme is significantly downregulated, while anterior expression is not altered (Fig. 5A–D), suggesting that Wnt/β-catenin signaling and/or BMP signaling is downregulated in the Hmgb1−/−; Hmgb2+/− forelimb bud. To confirm the effect on BMP signaling, we examined expression of Msx1, another well-recognized target of BMP signaling (Khokha et al., 2003; Michos et al., 2004). The expression of Msx1 is also downregulated in the posterior mesenchyme, while anterior expression remained unaffected (Fig. 5E, F). These results suggest that BMP signaling is downregulated in the posterior mesenchyme in the Hmgb1−/−; Hmgb2+/− forelimb bud.
We next examined expression of known targets of Wnt/β-catenin signaling in the limb bud. Axin2, a negative feedback regulator, is a target of Wnt/β-catenin signaling (Jho et al., 2002; Leung et al., 2002; Lustig et al., 2002). Expression of Axin2 was strongly detected in the AER of the control embryo (Fig. 5G), however, its expression domain in the posterior region was significantly smaller in the Hmgb1−/−; Hmgb2+/− forelimb bud (Fig. 5H). Since we detected normal expression of Fgf8 (Fig. 4A–D), the reduction in Axin2 expression is probably not due to a loss of posterior AER structure. Rather, this result suggests that Wnt/β-catenin signaling is low in the posterior AER. DKK1 is a negative regulator of Wnt/β-catenin signaling, and is also a target of Lef1/TCF-β-catenin signaling (Gonzalez-Sancho et al., 2004; Niida et al., 2004; Chamorro et al., 2005). As previously shown (Mukhopadhyay et al., 2001), Dkk1 expression is detected in the anterior and posterior margin of the limb mesenchyme (Fig. 5I). In the Hmgb1−/−; Hmgb2+/− forelimb bud, Dkk1 expression domain in the posterior region was reduced (Fig. 5J). Together with the smaller expression domain of Cyclin D1 (Fig. 3K, L), a target of Shh and Wnt/β-catenin signaling, our results suggest that β-catenin pathway activity is lower in the posterior region of the Hmgb1−/−; Hmgb2+/− forelimb bud.
Global Downregulation of Canonical Wnt Signaling in hmgb1-hmgb2 Knockdown Zebrafish
HMGB is a well-conserved nuclear protein found in multicellular organisms (Sessa and Bianchi, 2007), suggesting conserved functions in animal species. To further examine the function of hmgb genes in the enhancement of Wnt/β-catenin pathway during embryonic development, we used Top-dGFP zebrafish embryos (Dorsky et al., 2002), in which the expression of dGFP is dependent on TCF/Lef1-β-catenin activity. We knocked down hmgb1 and hmgb2 by injecting morpholinos (MOs) and monitored GFP expression by in situ hybridization. As previously shown (Dorsky et al., 2002), GFP expression is detected in known tissues with high Wnt/β-catenin activities at the 18 somite-stage, such as forebrain, midbrain, and tail bud (Fig. 6A). Both hmgb1-morphants and hmgb2-morphants show reduced GFP signal in all these tissues in a dose-dependent manner (Fig. 6B, C, Table 3) with more downregulation by hmgb1-MO. Furthermore, downregulating both hmgb1 and hmgb2 resulted in more severe downregulation of GFP expression at a higher frequency (Fig. 6D, Table 3). This synergistic activity was also confirmed with a second set of morpholinos (Supp. Table S1). While injection of synthetic mRNA for Hmgb1 and Hmgb2 did not affect Top-GFP expression, co-injection of mouse Hmgb1 mRNA and Hmgb2 mRNA could partially rescue the expression of Top-GFP (Fig. 6E, Table 4). These results demonstrate that enhancement of Wnt/β-catenin signaling by hmgb1 and hmgb2 is a general feature in vertebrate embryos, and that this enhancement operates not only in the limb bud but also in other sites where Wnt/β-catenin signaling is active.
Table 3. Dose-Dependency and Synergistic Function of hmgb1 Morpholino and hmgb2 Morpholino for the Expression of Top-GFP in the Zebrafish Embryo
Affected embryos (%)
Control MO (2 ng)
hmgb1 MO (0.5 ng)
hmgb1 MO (1 ng)
hmgb2 MO (0.5 ng)
hmgb2 MO (1 ng)
hmgb1 MO (0.5 ng) + hmgb2 MO (0.5 ng)
hmgb1 MO (1 ng) + hmgb2 MO (1 ng)
Table 4. Rescue of Top-GFP Expression in hmgb1/2-Double Morphants by hmgb1/2 mRNA at 18-Somite Stage
Weaker than normal expression but clearly stronger than expression with severe downregulation.
Control MO (2 ng)
hmgb1 MO (1 ng) + hmgb2 MO (1 ng)
Hmgb1 mRNA (100 pg) + Hmgb2 mRNA (100 pg)
hmgb1 MO (1 ng) + hmgb2 MO (1 ng) + Hmgb1 mRNA (100 pg)+ Hmgb2 mRNA (100 pg)
At a later stage (36 hr post-fertilization; hpf), hmgb1-morphants show a smaller head, an undulating notochord in small populations, and ventrally bent tail structures (Fig. 6G), similar to wnt8 knockdown embryos (Lekven et al., 2001; Shimizu et al., 2005). This phenotype was milder in hmgb2-morphants (Fig. 6H), but became more severe and frequent in hmgb1-hmgb2 double morphants (Fig. 6I, Table 5). This phenotype was also partially rescued by co-injecting Hmgb1 mRNA and Hmgb2 mRNA (Fig. 6J, Table 5). The correlation of the gross morphology phenotype and downregulation of Top-dGFP signal suggests that, similar to mouse mutant limb, hmgb1 function is more important than hmgb2 function in enhancing β-catenin activity in the zebrafish embryo.
Table 5. Synergistic Function of hmgb1 and hmgb2 Morpholinos and Rescue by mRNA Injection of the 36-hpf Body Morphology
Short, curly tail (%)
Control MO (2 ng)
Hmgb1 MO (1 ng)
Hmgb2 MO (1 ng)
hmgb1 MO (1 ng) + hmgb2 MO (1 ng)
Hmgb1 mRNA (100 pg) +Hmgb2 mRNA (100 pg)
hmgb1 MO (1 ng) + hmgb2 MO (1 ng) + Hmgb1 mRNA (100 pg) + Hmgb2 mRNA (100 pg)
Downregulation of hmgb1 and hmgb2 Causes Downregulation of shh Expression in the Pectoral Fin
The pectoral fin of zebrafish is derived from the lateral plate mesoderm and is homologous to the forelimb in tetrapod animals (Grandel and Schulte-Merker, 1998; Freitas et al., 2006). Remarkable similarities of developmental and genetic mechanisms between pectoral fin buds and forelimb buds have been found (Mercader, 2007). Thus, we examined the functional conservation of hmgb1 and hmgb2 during zebrafish pectoral fin development. The development of the pectoral fin bud in hmgb1-hmgb2 double morphants is not significantly affected at 36 hpf as visualized by prx1 expression (Fig. 7A, B, Supp. Table S2). However, the activation status of Wnt signaling, visualized by the expression of Top-dGFP, is significantly downregulated in hmgb1-hmgb2-knockdown fin buds (Fig. 7C, D, Table 6). Similar to the mutant mouse forelimb bud, expression of shh and ptc1 is significantly downregulated in hmgb1-hmgb2-knockdown fin buds (Fig. 7E–J, Table 6). These results demonstrate a conserved requirement of hmgb1 and hmgb2 for the enhancement of Wnt/β-catenin signaling and expression of shh in the vertebrate forelimb bud.
Table 6. Gene Expression in the Pectoral Fin Bud of Embryos Injected With hmgb1 MO and hmgb2 MOa
Weak signal, small domain
Affected embryos (%)
Injection amount is 2 ng for control MO, 1 ng for hmgb1 MO, and 0.5 ng for hmgb2 MO.
hmgb1 MO + hmgb2 MO
hmgb1 MO + hmgb2 MO
hmgb1 MO + hmgb2 MO
Functional analyses of the Hmgb genes have identified their roles in adult animals, such as regulation of immune response, cancer, spermatogenesis, and hematopoietic stem cell self-renewal (Ronfani et al., 2001; Dumitriu et al., 2005; Nemeth et al., 2006; Tang et al., 2010); however, their function in embryonic development was not well understood. In this report, we examined the limb phenotype of Hmgb1 and Hmgb2 double mutants. Previous studies with individual mutant mice did not show any defects in the limb, except for the delay of endochondral ossification in late-stage Hmgb1−/− mice (Taniguchi et al., 2007). Our attempt to obtain double null embryos revealed that Hmgb1−/−; Hmgb2−/− embryos arrested at E9.5 (Table 1), suggesting that the combined function of Hmgb1 and Hmgb2 is essential for the development of mouse embryos. Furthermore, our approach of double knockout of Hmgb1 and Hmgb2 found that these genes are genetically redundant, and that their combined function is required for proper forelimb development.
We have recently demonstrated Hmgb2-mediated enhancement of Wnt/β-catenin signaling in adult tissue. Reduction of Hmgb2 expression during aging correlates with reduction of Wnt/β-catenin signaling in the adult diarthrodial joint, and HMGB2 enhances expression of Wnt/β-catenin target genes in vitro (Taniguchi et al., 2009a, b). Moreover, the HMG domain within HMGB2 is responsible for interaction with Lef-1, suggesting that HMGB1 is also involved in enhancing Wnt/β-catenin signaling (Taniguchi et al., 2009a). Our present data show that hmgb2-dependent enhancement of the Wnt/β-catenin signaling is also present in a variety of embryonic tissues in developing zebrafish embryos (Fig. 6). Furthermore, hmgb1 functions to enhance the Wnt/β-catenin signaling in zebrafish embryos. The Wnt/β-catenin reporter expression in zebrafish embryos demonstrates that hmgb1 and hmgb2 are functionally redundant, and suggests that the function of hmgb1 is more important than that of hmgb2 (Fig. 6). While exact mechanisms by which Hmgb genes function during embryonic development are still not well understood, enhancing Wnt/β-catenin signaling is, at least in part, a mechanism for Hmgb1 and Hmgb2 to control organogenesis. Interestingly, Hmgb3 is also known to regulate Wnt/β-catenin signaling activity. In contrast to Hmgb1 and Hmgb2, a study has demonstrated that abrogating Hmgb3 results in constitutive activation of Wnt/β-catenin signaling in hematopoietic stem cells (Nemeth et al., 2006). Thus, although it has a high degree of structural similarity with Hmgb1 and Hmgb2, Hmgb3 seems to have an opposite function, and the enhancement of Wnt/β-catenin signaling might be a specific function of Hmgb1 and Hmgb2.
Our analysis strongly suggested that the function of Hmgb1 and Hmgb2 affects not only Wnt/β-catenin signaling but also other signaling pathways in developing limb buds. The skeletal phenotype of the Hmgb1−/−; Hmgb2+/− forelimb is the selective loss of digit 5, which is reminiscent of early downregulation of Shh signaling, while other elements were intact. Because the development of digit 5 requires a high level of Shh signaling (Harfe et al., 2004), a disruption of normal Shh signaling would contribute to the loss of digit 5 in the Hmgb1−/−; Hmgb2+/− forelimb bud. Indeed, a late onset and lower activation of Shh expression as well as reduced level of Shh target genes (Fig. 3) support this idea. However, reduction of Shh signaling alone does not simply account for the phenotype. For instance, a study utilizing an inducible Shh knockout has demonstrated that expansion of digit 3 primordia is the most sensitive to lowered Shh signaling, and that digit 3 is the first digit to be lost upon eliminating Shh during the digit progenitor expansion period in the developing limb bud (Zhu et al., 2008). This is in contrast to the loss of digit 5 in the Hmgb1−/−; Hmgb2+/− forelimb bud. Furthermore, the anterior domain where Hoxd13 is not expressed, which is correlated with Gli3 processing (Wang et al., 2007a, b), showed a slight expansion (Fig. 2G, H). Gli3 processing is counteracted by Shh signaling (Litingtung et al., 2002; te Welscher et al., 2002), suggesting that Shh-dependent counteraction of Gli3 processing is impaired, but not as severely in Shh mutant limb buds. Taken together, we favor the idea that Shh signaling is reduced in the Hmgb1−/−; Hmgb2+/− forelimb bud. However, the disruption of Shh signaling alone is not sufficient to cause the loss of digit 5.
The late onset and reduction of Shh expression (Fig. 3) suggest that Hox function is reduced. By eliminating multiple Hoxa and Hoxd genes in mice, it has been shown that 5′ Hox genes induce Shh expression in a gene-specific and dosage-dependent manner (Tarchini et al., 2006). Interestingly, HMGB1 is shown to interact with Hoxd proteins and enhance their transcriptional activity (Zappavigna et al., 1996). Thus, 5′-Hox gene function could be reduced in the Hmgb1−/−; Hmgb2+/− forelimb bud, which would contribute to the late onset and lower activation of Shh expression. Among four Hox clusters, the Hoxd genes are expressed with a posteriorly biased manner and have redundant function in the developing limb (Tarchini et al., 2006; Zakany and Duboule, 2007). Therefore, Hoxd activity might also be reduced in the Hmgb1−/−; Hmgb2+/− forelimb bud, which may also contribute to the loss of digit 5 phenotype in the Hmgb1−/−; Hmgb2+/− forelimb bud.
In contrast to Shh signaling, Wnt/β-catenin signaling seems to be affected both in the ectoderm and mesoderm in the forelimb bud. Downregulation of Axin2 in the AER and Dkk1 and Msx2 in the mesenchyme suggest downregulation of Wnt/β-catenin signaling in the posterior region of the AER and mesenchyme (Fig. 5). Several Wnt genes are known to function in mouse limb development, including Wnt3, Wnt5a, Wnt7a, and Wnt9a (Parr et al., 1993; Parr and McMahon, 1995; Yamaguchi et al., 1999; Barrow et al., 2003; Spater et al., 2006). Among them, normal expression of Lmx1b suggests that Wnt7a function is not affected in the Hmgb1−/−; Hmgb2+/− forelimb bud. Lmx1b is shown to be downregulated in a limb-mesenchyme-specific knockout of the Ctnnb1 gene (encoding β-catenin) (Hill et al., 2006). Our data do not contradict this report. β-catenin signaling is clearly downregulated in hmgb1/2-knockdown zebrafish embryos, but not completely abolished (Fig. 6). Moreover, in the Hmgb1−/−; Hmgb2+/− forelimb bud, one copy of the Hmgb2 gene is still present. This condition might be sufficient for maintaining Lmx1b expression in the Hmgb1−/−; Hmgb2+/− forelimb bud. Similar to Shh signaling, we favor that Wnt/β-catenin signaling is downregulated but it is not as severe as causing the skeletal phenotype by itself alone.
Downregulation of both Msx1 and Msx2 in the posterior mesenchyme suggests that BMP signaling is impaired in the Hmgb1−/−; Hmgb2+/− forelimb bud (Fig. 5A–F). However, compared to Wnt/β-catenin signaling and Shh signaling, it is less clear how Hmgb1 and Hmgb2 regulate BMP signaling. Since HMGB proteins bind chromatin without sequence specificity and alter conformation to facilitate binding of other transcription factors (Pil and Lippard, 1992; Paull et al., 1993), it is conceivable to speculate the possibility that binding of HMGB proteins may facilitate Smad-dependent BMP signaling. Such mechanisms are left to be open for HMGB study in the future.
Our analysis of the Hmgb1−/−; Hmgb2+/− forelimb revealed a redundant function of Hmgb1 and Hmgb2 in mouse embryonic limb development. Gene expression analysis suggests that the combined function of Hmgb1 and Hmgb2 integrates Wnt/β-catenin signaling, Shh signaling, BMP signaling, and Hox activity (Fig. 8). Combined alteration of these signaling pathways at the posterior region in the Hmgb1−/−; Hmgb2+/− forelimb bud may have lead to the loss of digit 5. Furthermore, our study revealed redundant functions of hmgb1 and hmgb2 in the enhancement of Wnt signaling in a variety of embryonic structures in zebrafish. Since Hmgb1 and Hmgb2 are widely co-expressed in early stage mouse embryos (Supp. Fig. S1), their combined function might be widely important for development of other organs. The data that double null embryos arrest at E9.5 supports importance of their function beyond this stage. Further study with a conditional gene inactivation approach will identify combined function of Hmgb1 and Hmgb2 in organogenesis.
Care and experimentation on mice and zebrafish were done in accordance with the Institutional Animal Care and Use Committee of the University of Minnesota, the Scripps Research Institute, and the National Institute of Advanced Industrial Science and Technology, Japan.
Both Hmgb1 mutant (Calogero et al., 1999) and Hmgb2 mutant (Ronfani et al., 2001) have been published previously. We generated Hmgb1+/−; Hmgb2+/−mice by breeding Hmgb1+/− and Hmgb2+/− mice, which were kept on Balb/c and C57BL/6 background, respectively. Mutant embryos were obtained from timed mating of the Hmgb1+/−; Hmgb2+/− mice. Embryo genotyping was done according to previous publications.
Top-dGFP zebrafish line (Dorsky et al., 2002) was used to detect the activation status of Wnt signaling. Morpholino and mRNA injections were done as previously described (Kawakami et al., 2004; Oishi et al., 2006). Morpholinos were designed by GeneTools (Philomath, OR). The sequences of the MOs are
Standard control MO from GeneTools (targeting human beta globin intron mutation 5′-CCTCTTACCT CAGTTACAATTTATA-3′) was used as a control. mRNA of mouse Hmgb1 and Hmgb2 are synthesized by using the mMessage mMachine Kit (Ambion, Austin, TX) according to the manufacturer's instruction.
In Situ Hybridization Analysis and Skeletal Preparation
Whole mount in situ mRNA expression analysis in mouse and zebrafish embryos was done following standard protocols (Kawakami et al., 2004, 2005; Oishi et al., 2006). Skeletal preparation of mouse embryos was done as previously described (Kawakami et al., 2009).
Immunostaining was performed similar to our previous experiments (Bluske et al., 2009). In brief, samples were fixed with 4% PFA/PBS+0.1% Tween20 (PBT) overnight at 4°C, washed with PBT, and dehydrated by serially increasing methanol concentration. After rehydration, samples were treated with 30% sucrose/PBS, embedded in OCT, and sectioned at 14 μm. Sections were washed with PBT, blocked with 10% goat serum/PBT, incubated with anti-phospho ERK1/2 (Cell Signaling, Danvers, MA; 9101, 1:1,000 dilution), washed with PBT, and incubated with Alexa594 anti rabbit IgG (Invitrogen, Carlsbad, CA; 1:1,000 dilution). After washing and counterstaining with DAPI, we detected pERK1/2 signal by using Zeiss LSM 710 Laser Scanning Microscopes. Images were analyzed by ZEN2009 software, and pERK1/2-positive cell layers were manually counted.
We are grateful to investigators for generously sharing materials: Dr. Marco E. Bianchi (Hmgb1 mutant mice and Hmgb2 mutant mice), Dr. Randall Moon (Top-dGFP fish line), Dr. Yasushi Nakagawa (Axin2 probe), and Dr. Juan Carlos Izpisua Belmonte (Shh, Fgf8, and Hoxd13 probes). We are also grateful to Dr. Michael O'Connor for allowing us to use Zeiss LSM710, Thu Quach, Austin Johnson, Emily Hsu, and Miu Yamashita for their excellent technical support and Krista Bluske for editorial assistance. This study is supported by grants from Minnesota Medical Foundation (Y.K.), NIH grants AG007996 and AR056026 (M.L), and the Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (C), 22570197 (I.O.).