|Home | About | Journals | Submit | Contact Us | Français|
The Hox family of transcription factors are expressed at different domains along the rostrocaudal (R-C) body axis during development. To examine the function of Hoxc8 and Hoxc9 in specific cell types and at different developmental times, we have generated and characterized loxP flanked (floxed) Hoxc8 and Hoxc8->Hoxc9 replacement alleles of mice, with either GFP or LacZ reporters. Although all four alleles of mice behave like wild-type controls in motor behavioral testing, slight differences in endogenous Hox gene expression were observed among these alleles depending on the type of reporters used and the presence of Hoxc9 cDNA in the targeting constructs. The efficiency of Cre-mediated recombination was evaluated by crossing these mice with the Nestin-cre and Isl1-cre mice, and the loss of Hoxc8 expression with or without Hoxc9 misexpression was confirmed in embryonic spinal cord. In addition, an up-regulation of reporter gene expression was observed after Cre-mediated recombination. These mice will be useful tools to analyze Hox gene function in a cell type-specific manner.
Homeotic/Hox genes play important roles in defining cellular identity along the rostrocaudal (R-C) body axis during development (Krumlauf, 1994). The function of Hox genes in determining neuronal identity in the hindbrain has been well studied (Keynes and Krumlauf, 1994), while much less is known about their roles in spinal cord development. The expression domains of various Hox-c genes have been shown to correlate with the positions of motor neuron (MN) columns and pools (Dasen et al., 2003; Dasen et al., 2005; Liu et al., 2001). In addition, both gain- and loss-of-function experiments performed in chick embryos demonstrate that Hox6 and Hox9 paralog groups play instructive roles in defining MN columnar identity, while the Hox4, Hox5, Hox7, and Hoxc8 groups of genes delineate different motor pools (Dasen et al., 2003; Dasen et al., 2005). Mouse mutants that have lost Hox9 and Hox10 function exhibit locomotion deficits in the hindlimb region (Carpenter et al., 1997; de la Cruz et al., 1999; Lin and Carpenter, 2003; Wu et al., 2008), while loss of Hoxc8 function results in a forelimb prehension-deficiency phenotype (Tiret et al., 1998).
Hox genes are expressed in multiple tissues during development and their expression patterns change with time, and therefore, the motor behavior deficits observed in Hox mutants could be a compound effect of losing Hox function in both neural and mesodermal tissues. Moreover, neural expression of Hox genes is not limited to MNs, as many spinal interneurons required for coordinated locomotion also express various Hox genes. Thus, cell type-specific analyses will be required to further decipher the role of Hox genes in spinal cord development. To generate conditional loss-of-function and gain-of-function alleles of Hox genes in mouse, we first focused on the Hoxc8 locus, and used the forelimb grip-deficiency phenotype as a landmark to evaluate floxed Hoxc8 and Hoxc8->c9 alleles of mice.
To generate the floxed Hoxc8 allele, a loxP site was inserted in the 5´non-coding region of the Hoxc8 gene, and a second loxP site in the same orientation was inserted 3´ to the three known Hoxc8 polyadenylation (pA) signals. The endogenous 5´splicing donor site (5´SD), intron 1, and the endogenous 3´ splicing acceptor site (3´SA) were also inserted downstream of the second loxP site. We preserved the endogenous intron not only because it contains essential regulatory elements (Awgulewitsch et al., 1990), but also because it has the potential to stabilize the primary, and spliced transcript (Huang and Gorman, 1990). A ribosome reentry site (IRES) followed by either an eGFP (Mombaerts et al., 1996, for tracing axons) or nlsLacZ (nuclear localized LacZ, for marking nuclei) reporter cassette was inserted 3´ to the 3´SA to be used as tracers for the Cre-mediated recombination event. The endogenous 3´ untranslated region (3´UTR) was added 3´ to the reporter sequences. The 3´UTR is included to conserve potential miRNA binding sites that regulate the expression of Hoxc8 (Yekta et al., 2004). A neomycin phosphotransferase (Neo) resistance cassette was also included for positive selection of the transfected ES cells (Fig. 1a).
We also generated conditional Hoxc8->Hoxc9 replacement alleles to examine the long-term effects of Hoxc9 misexpression in mouse. To generate the floxed Hoxc8->c9 allele, a DNA fragment containing the Hoxc9 coding region was inserted between the second loxP site and the additional Hoxc8 intron (Fig. 1a). Two floxed Hoxc8 alleles—one with a GFP reporter, the other with a LacZ reporter, and two floxed Hoxc8->Hoxc9 alleles, with either GFP or LacZ reporters were generated using this strategy.
Since the Hox locus is tightly regulated, any alteration at this locus could potentially affect the expression of surrounding genes. We therefore, characterized these floxed alleles prior to Cre-mediated recombination to ascertain that they behave similar to the wild-type (WT) alleles. We first examined Hoxc6-Hoxc9 mRNA expression in e10.5 mouse embryos using whole-mount in situ hybridization. The expression domains of Hoxc6 and Hoxc9 are very similar among embryos carrying different floxed alleles and their WT littermates at e10.5 (data not shown). However, a ~1-segment rostral extension in neural and mesodermal Hoxc8 expression domain was observed in the GFP-tagged Hoxc8->c9F/F (Hoxc8->c9F/F-GFP) embryos (Fig. 2c), while a ~2-segment rostral extension in the neural and a ~3-segment rostral extension in the mesodermal Hoxc8 expression domains were observed in the LacZ-tagged Hoxc8F/F (Hoxc8F/F-LacZ) and Hoxc8->c9F/F (Hoxc8->c9F/F-LacZ) embryos (Fig. 2d, e). The GFP-tagged Hoxc8F/F (Hoxc8F/F-GFP) embryos exhibit no apparent change in Hoxc8 expression domain as compared to the WT controls (Fig. 2a, b).
To examine the phenotypic consequences of these changes in mesodermal tissues, we performed skeletal staining in e18.5 mouse embryos from different alleles prior to Cre-mediated recombination. WT and the Hoxc8F/F-GFP embryos have 7 cervical vertebrae (C1–C7) and their 6th and 7th ribs (R6s and R7s) are attached to the sterna (Fig. 2f, g, k, l). However, extra ribs extending from the C7 and elongated R8s attached to the sterna were observed in Hoxc8F/F-LacZ and Hoxc8->c9F/F-LacZ embryos (Fig. 2i, j, n, o). The F/+ embryos derived from these two LacZ-tagged alleles have a milder phenotype with either a partial rib extending from the C7 or only one of the R8s attached to the sternum (data not shown). The majority of the Hoxc8->c9F/F-GFP embryos have normal C7vertebra, but their R8s are attached to the sterna (Fig. 2m). No obvious homeotic transformation in skeletons was observed in the Hoxc8->c9F/+-GFP embryos (data not shown).
To ascertain that the minor changes observed in the Hoxc8 expression domain did not impair motor function, we examined 2-month old male mice for their forelimb grip strength, a phenotype that was observed in the Hoxc8−/− animals previously (Tiret et al., 1998). We found no deficit in the grip strength in all 4 floxed alleles of mice when compared to their WT littermates (Fig. 2p), indicating that the minor change observed in the Hoxc8 expression domain did not cause impairment of motor function.
Because of our interest in Hox gene function during spinal cord development, we further evaluated the endogenous Hox gene expression in open-book spinal cord preparations from e13.5 embryos by whole mount in situ hybridization prior to Cre-mediated recombination. We focused our examination on the GFP-tagged floxed alleles that exhibit little or no deviation in endogenous Hox gene expression from the WT allele. The R-C levels were estimated from the positions of the dorsal root ganglia (DRG) that remain attached to the spinal cord during this procedure (see Methods). Slight changes in the Hoxc8 mRNA rostral expression boundary were observed in Hoxc8F/F-GFP, Hoxc8->c9F/+-GFP, and Hoxc8->c9F/F-GFP embryos when compared to WT controls (Fig. 3a, c–e), including a caudal shift of expression at the intermediate region and a more uniform expression in the ventral region of the spinal cord. No obvious change in Hoxc8 mRNA expression was detected in the Hoxc8F/+-GFP embryos (Fig. 3b). No significant change in Hoxc6 mRNA expression was observed between various floxed alleles and WT controls (data not shown). In contrast, the rostral boundary of Hoxc9 mRNA expression in the Hoxc8->c9F/F-GFP ventral spinal cord is shifted ~ 1 segment caudally when compared to WT controls (Fig. 3f, j), while no significant change in Hoxc9 mRNA expression was detected in Hoxc8F/+-GFP, Hoxc8F/F-GFP, and Hoxc8->c9F/+-GFP embryos (Fig. 3g–i).
Hoxc6 protein induces lateral motor column (LMC) formation in the brachial level spinal cord while Hoxc9 protein induces the formation of sympathetic preganglionic neurons in the thoracic level spinal cord. Cross-inhibitory actions between these two proteins at the junction between brachial and thoracic levels delineate the brachial vs. thoracic identity of individual cells (Dasen et al., 2003). We therefore examined the mRNA expression of an LMC marker, Raldh2, which is induced by Hoxc6 and suppressed by Hoxc9 proteins. The Raldh2 expression domain extends caudally by ~1 segment in the Hoxc8->c9F/F-GFP embryos compared to the WT (Fig. 3k, o), and this could result from the caudally displaced Hoxc9 rostral expression boundary. No significant change in Raldh2 mRNA expression was observed in Hoxc8F/+-GFP, Hoxc8F/F-GFP, and Hoxc8->c9F/+-GFP spinal cords compared to the WT controls (Fig. 3l–n).
Hoxc8 protein defines motor pool identity in the caudal cervical level spinal cord, and we therefore examined the mRNA expression of a motor pool marker, Pea3, that is a downstream target of Hoxc8 (Vermot et al., 2005). We did not observe significant differences in the Pea3 expression domains in Hoxc8F/+-GFP, Hoxc8F/F-GFP, and Hoxc8->c9F/+-GFP embryos compared to the controls (Fig. 3p–s). However, in the Hoxc8->c9F/F-GFP embryos, the Pea3 expression domain is extended caudally (~1 segment, Fig. 3t), and this could be a secondary effect of the caudally expanded Raldh2 expression domain which also extended the caudal motor pools (Fig. 3o). These results further demonstrate the sensitivity of the Hoxc locus to genomic manipulation—although these four targeting constructs share the same basic design, subtle changes in endogenous Hox gene expression could occur depending on the type of reporters used, and the presence of the Hoxc9 cDNA in individual targeting constructs. Although we are not certain which component of the exogenous DNA caused this difference, the size difference between the GFP and LacZ reporters (~0.8kb vs. ~3kb) and the presence of an additional ~0.8 kb Hoxc9 cDNA sequence could be one of the possibilities.
To determine if Hoxc8 expression can be eliminated or replaced by Hoxc9 expression after Cre-mediated recombination, we crossed the floxed Hoxc8-GFP and Hoxc8->c9-GFP mice with Nestin-cre (Nescre) (Tronche et al., 1999), and Isl1-cre (Isl1cre) (Srinivas et al., 2001) transgenic mice, and examined Hox protein expression using immunohistochemistry on cross-sectioned spinal cord tissue obtained from e12.5 mouse embryos. By e12.5, the majority of cells in the spinal cord of the Hoxc8F/F-GFP;Nescre and Hoxc8->c9F/F-GFP;Nescre embryos had already lost Hoxc8 protein expression (Fig. 4d, f, and data not shown). In the Hoxc8->c9F/F-GFP; Nescre embryos, ectopic Hoxc9 protein expression at the expense of endogenous Hoxc8 protein is clearly visible using antibodies against either Hoxc8 or Hoxc9 (Fig. 4d, e). However, the Hoxc9 protein expression is not as high as that of the endogenous Hoxc8 protein but instead, is more similar to the endogenous Hoxc9 expression observed at the thoracic levels (Fig. 4a, d, and data not shown), this could reflect a difference in the properties between the anti-Hoxc8 and anti-Hoxc9 antibodies used, or a brachial to thoracic fate change induced by Hoxc9 misexpression. A reduction of Hoxc6 protein expression in these cells is also apparent when compared to the controls (Fig. 4b, c, e, f), demonstrating that misexpression of Hoxc9 reduces the expression of endogenous Hoxc6. Essentially none of the Isl1+ neurons co-express Hoxc8 in e12.5 Hoxc8F/F-GFP;Isl1cre and Hoxc8->c9F/F-GFP;Isl1cre embryos, even in the lateral LMC (LMCl) MNs where Isl1 is only expressed transiently (Fig 4g, h, and data not shown), suggesting that the Isl1-cre has a higher recombination efficiency than Nestin-cre. A higher level of GFP expression can be activated in embryos that have undergone the Cre-mediated recombination and can be observed in freshly dissected embryos under a fluorescent stereomicroscope in comparison to the low levels of GFP expression observed prior to the recombination. Unfortunately, fixation with paraformaldehyde diminishes the GFP signal and an anti-GFP antibody is required to visualize GFP expression in cross-sectioned spinal cord tissues (Fig. 4i, and data not shown). Similarly, high levels of β-galactosidase activity can be observed in the spinal cord of e11.5 Hoxc8F/F-LacZ;Isl1cre and Hoxc8->c9F/F-LacZ;Isl1cre embryos (Fig. 4k and data not shown) in contrast to the low-level β-galactosidase activity observed prior to Cre-mediated recombination (Fig. 4j). The presence of low-level reporter gene expression prior to Cre-mediated recombination indicates that a small percentage of Hoxc8 transcripts are not ployadenylated by the 3 known pA signals.
In summary, we have generated conditional Hoxc8 loss-of-function and Hoxc8 ->Hoxc9 replacement alleles of mice with either GFP or LacZ reporters. Although all four alleles are indistinguishable from WT controls in motor behavioral testing, molecular analyses revealed that the floxed Hoxc8-GFP allele behaved most like the WT in the endogenous Hox gene expression. The floxed Hoxc8->Hoxc9-GFP allele displayed a ~1 segment rostral expansion of Hoxc8 mRNA expression at e10.5, and a ~1 segment caudal displacement of the Hoxc9 mRNA rostral expression boundary in e13.5 spinal cord, while the floxed Hoxc8-LacZ and floxed Hoxc8->Hoxc9-LacZ alleles exhibited a ~2-segment rostral extension in the neural and a ~3-segment rostral extension in the mesodermal Hoxc8 mRNA expression domains at e10.5. Upon Cre-mediated recombination, deletion of Hoxc8 and misexpression of Hoxc9 along with up-regulated GFP or LacZ expression were observed, demonstrating that these alleles of mice will be useful tools to study Hox function in a cell type-specific manner.
To generate Hoxc8 conditional targeting constructs, a 65bp fragment containing the LoxP sequence was cloned into a SacII site within exon1, 5´ to the translation start site. The following sequences were inserted into the Ear1 site (3´ to the 3 known polyadenylation sites) of the Hoxc8 locus in tandem: a second LoxP site in the same orientation followed by a ~1.4kb Nco1-Sal1 fragment containing the Hoxc8 intron, an IRES-eGFP or an IRES-nlsLacZ fragment followed by a ~1.5kb BseR1 fragment containing the 3’ UTR of Hoxc8 and the polyadenylation sequence from bovine albumin gene, and a pGKneo cassette for positive selection. The final targeting constructs were assembled in the pMoluc vector (Feng et al., 2002) and contain a ~9.5kb Sph1-Xba1 Hoxc8 genomic region with the above modifications. The homologus sequence at the 5´ and the 3´ arms are ~2.5kb and ~3.4kb respectively. To generate Hoxc8->Hoxc9 conditional replacement targeting constructs, a ~0.8kb fragment containing Hoxc9 coding region was inserted between the 3´LoxP site and the 1.4 kb Hoxc8 intron sequence in the Hoxc8 conditional knock-out constructs. Targeting constructs were digrested with Not1 and electroporated into W9.5 ES cells as described (Dragatsis et al., 2000). The targeting frequency is ~3%–4% (3 or 4 targeted clones from 96 G418-resistent colonies picked for each construct).
ES cells carrying the targeted alleles were verified by Southern analyses (see below) and were injected into C57BL6 blastocysts using standard procedures (Bradley, 1987). Male chimeras were crossed with C57B/6J females to generate heterozygous offspring. Mice carrying floxed alleles are maintained in a mixed 129/Sv and C57BL6 background and are housed in a barrier facility under 12-hr light/ 12-hr dark cycle. All protocols for animal use were approved by the Institutional Animal Care and Use Committee of the University of Virginia and were in accordance with NIH guidelines.
Genotyping was first performed by Southern blot analysis using Xba1-digested genomic DNA probed with a 500bp 5´-flanking HindIII-Sph1 fragment. PCR genotyping was performed with following primer pairs: LoxP1: 5'-CGCTCTACCTA CG CAGTGAG-3', and LoxP3: 5'-GGAAAACAGGGGGTTGACGAA G-3' (for floxed Hoxc8, and floxed Hoxc8->c9 alleles), LoxP1 and mHoxc8I: 5'-GGAGCGCGGGAGACCAAG-3' (for Δfloxed Hoxc8 alleles), LoxP1 and Hoxc95R: 5'-CCTGGACGCTAGGAGGTC-3' (for Δfloxed Hoxc8->c9 alleles).
For whole mount in situ hybridization, mouse embryos were harvested at e10.5 (noon of the day that the vaginal plug was present is considered e0.5), and spinal cords were isolated from e13.5 embryos. Dorsal root ganglia (DRG) were left on the dissected spinal cord through the in situ hybridization procedure, and images were taken with a Nikon SMZ1500 stereoscope before and after DRG removal to estimate the R-C extent of gene expression. Raldh2, Pea3, Sox5, and Hoxc6 probes were provided by Drs. T. Jessell and J. Dasen. Hoxc8 and Hoxc9 probes were provided by Dr. P. Gruss.
For β-galactosidase staining, e11.5 mouse embryos were fixed for 20’ in 4% paraformaldehye in 0.1M PB (phosphate buffer, pH 7.4), then washed 2×15’ with Buffer A (5mM EGTA, 2mM MgCl2 in 0.1M PB), and 2×15’ with Buffer B (0.2% NP40, 0.1% Sodium Desoxycholate, 2mM MgCl2 in 0.1M PB). Color reaction was performed in Buffer C (0.6mg/ml X-Gal, 5mM K3Fe(CN6), 5mM K4Fe(CN6), in Buffer B) at 37°C in dark and images were taken using a Nikon SMZ1500 stereoscope.
Immunohistochemistry was performed on 10 µm sections from e12.5 embryos using the following antibodies as described (Liu, 2006): rabbit anti-GFP (Invitrogen), mouse anti-Hoxc8 (Covance), rabbit and guinea pig anti-Isl1 (provided by Dr. T. Jessell), guinea pig anti-Hoxc6, and rabbit anti-Hoxc9 (Liu et al., 2001). Images were taken with a Nikon C1 confocal microscope.
Mouse embryos were harvested at e18.5 and processed for skeleton staining as described (McLeod, 1980).
Forelimb grip strength testing was performed with three trials using a grip strength apparatus (San Diego Instruments, Inc.). Graphs and statistical analyses were generated using SigmaPlot.
We thank S. Zeitlin for discussion and comments on this manuscript; T.-C. He for providing the pMOLUC vector. This work was supported by NIH grant NS045933 to J.-P. Liu.