Undulated short-tail Deletion Mutation in the Mouse Ablates Pax1 and Leads to Ectopic Activation of Neighboring Nkx2-2 in Domains That Normally Express Pax1

Corresponding author: Kenji Imai, Institute of Developmental Genetics, Ingolstädter Landstr. 1, D-85764 Neuherberg, Germany., imai{at}gsf.de (E-mail)

Communicating editor: C. KOZAK

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Previous studies have indicated that the Undulated short-tail deletion mutation in mouse Pax1 (Pax1^Un-s) not only ablates Pax1, but also disturbs a gene or genes nearby Pax1. However, which gene(s) is involved and how the Pax1^Un-s phenotype is confined to the Pax1-positive tissues remain unknown. In the present study, we determined the Pax1^Un-s deletion interval to be 125 kb and characterized genes around Pax1. We show that the Pax1^Un-s mutation affects four physically linked genes within or near the deletion, including Pax1, Nkx2-2, and their potential antisense genes. Remarkably, Nkx2-2 is ectopically activated in the sclerotome and limb buds of Pax1^Un-s embryos, both of which normally express Pax1. This result suggests that the Pax1^Un-s deletion leads to an illegitimate interaction between remotely located Pax1 enhancers and the Nkx2-2 promoter by disrupting an insulation mechanism between Pax1 and Nkx2-2. Furthermore, we show that expression of Bapx1, a downstream target of Pax1, is more strongly affected in Pax1^Un-s mutants than in Pax1-null mutants, suggesting that the ectopic expression of Nkx2-2 interferes with the Pax1-Bapx1 pathway. Taken together, we propose that a combination of a loss-of-function mutation of Pax1 and a gain-of-function mutation of Nkx2-2 is the molecular basis of the Pax1^Un-s mutation.

AN allelic series of mutations can provide important tools for understanding gene function and regulation. Earlier studies have demonstrated that three spontaneous mouse mutations constitute an allelic series in Pax1: undulated (Pax1^un; WRIGHT 1947 ), undulated-extensive (Pax1^un-ex; WALLACE 1980 ), and Undulated short-tail (Pax1^Un-s; BLANDOVA and EGOROV 1975 ). Pax1 is a member of the Pax gene family, which consists of transcription factors characterized by the presence of a highly conserved DNA-binding domain, the paired box (DAHL et al. 1997 ; MANSOURI et al. 1999 ). Embryonic expression of Pax1 is detected in the sclerotome of somites from embryonic day (E) 8.5 (DEUTSCH et al. 1988 ; WALLIN et al. 1994 ), the pharyngeal pouch endoderm from E9.5 (WALLIN et al. 1996 ), and limb buds from E10.0 (TIMMONS et al. 1994 ). The recessive Pax1^un allele has a point mutation in the paired box (BALLING et al. 1988 ), while the recessive Pax1^un-ex and the semidominant Pax1^Un-s alleles carry deletion mutations (WALLIN et al. 1994 ; DIETRICH and GRUSS 1995 ). In each allele, affected mice display dysmorphologies as developmental defects in the derivatives of embryonic Pax1-positive tissues, including the vertebral column, the shoulder girdle, the sternum (GRUNEBERG 1950 , GRUNEBERG 1954 ), and the thymus (DIETRICH and GRUSS 1995 ; WALLIN et al. 1996 ). Notably, the three alleles differ significantly in the severity of their phenotypes and in the mode of inheritance, raising a question as to how each mutation correlates with the respective phenotypes. To address this question, we have previously created a targeted null allele of Pax1 (Pax1^null) and made a phenotypic comparison between the natural alleles and the Pax1^null allele as a reference (WILM et al. 1998 ). This analysis has demonstrated that Pax1^un is a hypomorphic allele, while Pax1^un-ex can be regarded as a null allele. In contrast, the Pax1^Un-s deletion mutation, which removes all exons of Pax1 (DIETRICH and GRUSS 1995 ; WALLIN et al. 1996 ), displays a significantly stronger phenotype than Pax1^null displays, indicating that the Pax1^Un-s mutation affects not only Pax1, but also additional gene(s) within or nearby the deletion interval. Remarkably, despite the potential involvement of additional gene(s) other than Pax1, only the derivatives of embryonic Pax1-positive tissues are affected in Pax1^Un-s mutant mice. This characteristic feature of Pax1^Un-s allowed us to propose two different, but not mutually exclusive, hypotheses (WILM et al. 1998 ): First, a gene or genes sharing expression domains with Pax1 are ablated by the Pax1^Un-s deletion ("co-deletion model"). Second, a gene or genes flanking the Pax1^Un-s deletion interval come under the influence of regulatory elements of Pax1, thereby being ectopically activated in the natural expression domains of Pax1 ("ectopic activation model").

In the present study, we delineated the molecular basis of the Pax1^Un-s mutation by determining the Pax1^Un-s deletion interval and by analyzing genes around Pax1. Our results led us to a conclusion that the Pax1^Un-s deletion mutation affects physically linked Pax1, Nkx2-2, and their potential antisense genes, leading to a contiguous gene syndrome due to a loss-of-function mutation of Pax1 and a gain-of-function mutation of Nkx2-2.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Animals:
Pax1^Un-s mutant mice were kindly provided by A. M. Malashenko (Krosnogorsk, Russia). Pax1^null mutant mice (referred to as Pax1^tm1neu with Mouse Genome Informatics (MGI) accession ID MGI:185756 in Mouse Genome Database) were generated and genotyped as described (WILM et al. 1998 ). These mutant strains have been maintained as congenic lines on the C57BL/6J background.

Bacterial artificial chromosome (BAC) library screening:
The CitbCj7 mouse BAC library of 129/Sv origin was purchased from Research Genetics (Huntsville, AL). A total of 10 BAC clones were isolated in multiple rounds of screening by hybridization with a 1-kb XbaI fragment from intron 2 of Pax1 (Fig 1) and subsequently with a mixture of two BAC-end fragments (see below) originating from BAC213 (the T7 end, 5' of Pax1) and BAC132 (the T7 end, 3' of Pax1). In addition, 8 Pax1-positive BAC clones including RP23-224L16 and RP23-382B13 were isolated from the RPCI-23 C57BL/6J library (Research Genetics).

View larger version (8K): In this window In a new window Download PPT slide   Figure 1. Genomic organization of mouse Pax1 and its potential antisense gene Aspax1. Exons are numbered (0–5 for Pax1 and 1–2 for Aspax1) and indicated by boxes (open box, UTR; solid box, coding region). The last exons are indicated by the presence of poly(A) signals. Note the alternative usage of exon 1: 1a and 1b (see RESULTS). Transcriptional orientations of Pax1 and Aspax1 are shown by the arrows. The sequenced region (AF285175) is indicated by the solid line. The 1.0-kb XbaI fragment in intron 2 used for BAC library screening is depicted by the shaded bar.

BAC-end cloning:
pBeloBAC11-based clones from the CitbCj7 BAC library were digested with NotI and BglII, and the resulting fragments in mixture were subcloned into NotI-BamHI cleaved pBluescriptII KS (Stratagene, La Jolla, CA). BAC-end subclones were identified by Southern hybridization with two NotI-HindIII vector fragments, each of which contained either the T7 or the SP6 priming site and flanked the HindIII cloning site of pBeloBAC11. Identified BAC-end subclones were sequenced to design PCR primers for sequence-tagged sites (STSs). The primers for the STSs (designated as D2Neu#) are as listed (see online supplemental Table 1 available at http://www.genetics.org/supplemental/).

BAC transgenesis:
pBeloBAC11-based BAC clones were linearized at the cosN site by -terminase (Takara, Berkeley, CA). Linear BAC DNA was separated by pulsed-field gel electrophoresis (PFGE), electroeluted from 1% agarose PFGE gel, and used for the production of transgenic mice. Transgenic mice were generated by pronucleus injection of linearized transgene constructs into fertilized eggs that were subsequently transferred into the oviducts of pseudopregnant fosters. The presence of both ends of the BAC transgenes was confirmed by PCR with two primer sets specific for the T7 end (5'-CAATGGAAGTCCGAGCTC-3' and 5'-GTCGACTCTAGAGGATC-3') and for the SP6 end (5'-CCGCTCACAATTCCACACA-3' and 5'-CCGGCAGTTTCTACACAT-3').

Determination of the Pax1^Un-s deletion interval:
To clone a segment flanking the deletion interval from Pax1^Un-s genomic DNA, a cassette-ligation-mediated PCR method, originally described as a cDNA cloning method by ISEGAWA et al. 1992 , was adapted as follows. All primer sequences used for this procedure are as listed (see online supplemental Table 2 available at http://www.genetics.org/supplemental/). Two nested forward primers, Un-s1 and Un-s2, were derived from a region downstream of the 3' deletion breakpoint (see RESULTS). EcoRI-digested total genomic DNA (500 ng) from Pax1^Un-s homozygous animals was ligated to preannealed EcoRI linkers, linker 1 and linker 2 (4 pmol each), which harbored sequences for nested reverse PCR primers, rev1 and rev2. One-tenth of each reaction was incubated for 10 min at 94° and subsequently subjected to the first PCR amplification with 1.1 mM Mg(OAc)₂, 1 M GC-Melt, 1x reaction buffer, 1/50 Tth polymerase mix (Advantage-GC genomic PCR kit from CLONTECH, Palo Alto, CA), and 10 µM primer mix (Un-s1/rev1) under the following conditions: 4 min 94° (initial denaturing); 35 cycles of 30 sec at 94°, 5 min at 68° (two-step amplification); and 5 min at 68° (final extension). One-fiftieth of the first PCR reaction was diluted 1:10 and subjected to the second PCR amplification using the same conditions with primer combinations (Un-s2/rev2). The resulting PCR fragments were cloned with the TOPO TA cloning kit (Invitrogen, San Diego).

Computational analysis of genome sequences:
Web-based tools at the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov), including BLAST programs and the GenBank databases, were used for gene search on genomic sequences. Repetitive sequences were masked by using RepeatMasker (http://ftp.genome.washington.edu/RM/RepeatMasker.html). Genome annotation data were surveyed at the Ensemble Genome Server (http://www.ensembl.org) and the NCBI.

Whole-mount in situ hybridization:
For genotyping embryos for this analysis, multiplex PCR amplification was performed with two forward primers (Un-s FW: 5'-AGACATGCCACAGTATTCCC-3' and Un-s FM: 5'-ACATCCATCCAGAGACATGC-3') and one common reverse primer (Un-s RC: 5'-ATGTCCTAGAGATCCACAGC-3') in single reactions. This PCR amplification produces two diagnostic products specific for either the Pax1^Un-s (463 bp) or the wild-type (611 bp) allele (see Fig 3D). Mouse Pax1 (NEUBUSER et al. 1995 ), Nkx2-2 (PABST et al. 1998 ), and Bapx1 (LETTICE et al. 2001 ) RNA probes were labeled with digoxigenin and developed with BM purple (Roche). A probe specific for Pax1 exon 0 was generated from the 465-bp PCR fragment using the following primers: 5'-AACATTAGGGTCCTCCATTCACG-3' and 5'-GCAAAGTGTCTCTTCAACTTTCCG-3'. Whole-mount in situ hybridization was performed according to the protocol of SPORLE and SCHUGHART 1998 . Control embryos were hybridized with a sense probe to verify the specificity of the antisense probe signal.

View larger version (33K): In this window In a new window Download PPT slide   Figure 2. (A) BAC contig map encompassing mouse Pax1. The bold horizontal bars represent 10 BAC clones from the CitbCJ7 mouse BAC library. The shaded horizontal bars represent two sequenced BAC clones from the RPCI-23 mouse BAC library. The solid boxes indicate the regions including exons of Pax1, Nkx2-2, or Foxa2. The orientation of the contig is according to the transcriptional orientation of Pax1. Seven STS markers from this study, D2Neu1–D2Neu7, are shown according to their locations and presence on BAC clones (marked by the solid circles). The deletion interval in Pax1Un-s (between D2Neu2 and D2Neu7) is indicated by the bold arrow. SfiI sites within the interval between D2Neu1 and D2Neu7 are included. (B) Genomic organization of mouse Nkx2-2 and its potential antisense gene Asnkx2-2. Exons are numbered (1–2 for Nkx2-2 and 1–6 for Asnkx2-2) and indicated by boxes: solid ones for Nkx2-2 and shaded ones for Asnkx2-2. The last exons are indicated by the presence of poly(A) signals. Transcriptional orientations are indicated by the arrows above the gene names. The Asnkx2-2 exons distribute in a 148-kb genome interval. Approximate sizes of large intervals (indicated by gaps) are given in kilobases. Note that the first exon of Asnkx2-2 overlaps with the second exon of Nkx2-2 in an antisense orientation and that the last exon of Asnkx2-2, located 30 kb upstream of exon 0 of Pax1, is included in the 5' part of the Pax1Un-s deletion interval (bold horizontal bar with an arrow). The STS marker D2Neu2 defines the 5' breakpoint of the Pax1Un-s deletion interval.

View larger version (34K): In this window In a new window Download PPT slide   Figure 3. Analysis and determination of the Pax1Un-s deletion interval. (A) Southern blot analysis of the Pax1Un-s genome with an 800-bp PstI fragment from the 3' end clone of BAC132 as a probe. Note that the 5.5-kb signal is absent in Pax1Un-s homozygote DNA (-/-). (B) Southern blot analysis of the Pax1Un-s genome with a 2.4-kb D2Neu7-positive EcoRI fragment as a probe, which is 5 kb downstream of the 800-bp PstI segment used as a probe in A. A positive signal in Pax1Un-s homozygote DNA is detected at 2.7 kb, which is larger than that in wild-type DNA (2.4 kb). (C) Southern blot analysis of the Pax1Un-s genome with a 2.0-kb D2Neu2-positive HindIII fragment as a probe. Positive signals in Pax1Un-s homozygote DNA are detected at 3.6 kb (HindIII digest) and 2.7 kb (EcoRI digest), which are different in size from those detected in wild-type DNA (2.0 kb and 8.0 kb, respectively). (D) PCR detection of the Pax1Un-s allele. M, size marker; C, negative control with water as template. (E) Deletion breakpoints in the Pax1Un-s mutation. Forty nucleotides upstream and downstream of the deletion breakpoints are shown in comparison with wild-type sequences. Matched nucleotides are indicated by the vertical bars, and regions flanking the deletion interval are marked by shading. The first and the last nucleotides of the deletion intervals are marked by the angled arrows.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Transcription start sites of Pax1 and identification of a novel 5' exon:
Alignment of a 17,615-bp genome sequence, including mouse Pax1 from this study (GenBank accession no. AF285175), with the published cDNA sequence (GenBank accession no. NM_008780) determined the general genomic organization of five exons of mouse Pax1, spanning a 10-kb genome region (Fig 1). Putative transcription start sites of Pax1 were defined by sequencing five Pax1 cDNA clones that were retrieved from the RIKEN full-length enriched cDNA libraries (CARNINCI et al. 2000 ). In four of them, the 5' untranslated regions (UTRs) extended 66 bases upstream from the published 5' end (NM_008780), leading to the determination of the 5' end of exon 1 (exon 1a in Fig 1; GenBank accession no. AB080656). The remaining clone (RIKEN clone ID 5832428N23; GenBank accession no. AB080657) contained a 5' UTR that was 630 bp longer than the published sequence and included a novel exon (referred to hereafter as exon 0). Exon 0 is located 2.5 kb upstream of exon 1 and splices into the inside of exon 1 (exon 1b in Fig 1). Thus this type of Pax1 transcript lacks the 5' part of the exon 1. The splicing between exons 0 and 1b completely follows the GT-AG rule (MOUNT 1982 ). Furthermore, expression of this version of Pax1 transcripts was confirmed by whole-mount in situ hybridization with a probe specific for exon 0, showing a pattern identical to that with a conventional Pax1 probe (data not shown). Thus we conclude that there are two alternative forms of Pax1 transcripts with either of the two transcription start sites: one at the 5' end of exon 0 or the other at the 5' end of exon 1a. Resulting Pax1 transcripts differ in the structure of the 5' UTR, but the coding sequences are identical.

BAC contig map encompassing the Pax1 locus:
We established a BAC contig map of 400 kb encompassing mouse Pax1 (Fig 2A). Ten clones including five Pax1-positive BACs and their overlapping clones were isolated from the CitbCj7 mouse BAC library. Restriction mapping with SfiI digestion and BAC-end STS mapping (see online supplemental Table 1 available at http://www.genetics.org/supplemental/) allowed us to determine the relative order and orientation of the BAC clones. Furthermore, we isolated two additional BAC clones (RP23-224L16 and RP23-382B13) with larger inserts from the RPCI-23 library and added them to this contig map for the subsequent sequence analysis (see below).

Determination of the Pax1^Un-s deletion interval:
In the following descriptions, we refer to the 5' end and the 3' end of a deletion interval as "5' breakpoint" and "3' breakpoint," respectively. We tested several BAC-end fragments as probes for Southern hybridization on Pax1^Un-s and wild-type genomic DNA to determine the extent of the Pax1^Un-s deletion. Remarkably, an 800-bp PstI fragment from the T7-end clone of BAC132 (10 kb), located 60 kb downstream of Pax1 (Fig 2), gave no signal on homozygous Pax1^Un-s genomic DNA (Fig 3A). This indicates that this region is included in the Pax1^Un-s deletion interval. On the other hand, an STS marker, D2Neu7, which is also derived from the same BAC132 T7-end clone and located 5 kb downstream of the PstI fragment, is present in the Pax1^Un-s genome (data not shown). We subcloned a 2.4-kb EcoRI fragment positive for D2Neu7 from the BAC132 T7-end clone. Genomic Southern analysis with this 2.4-kb EcoRI probe demonstrated a rearrangement in the Pax1^Un-s genome (Fig 3B), indicating that the 2.4-kb EcoRI fragment included the 3' breakpoint of Pax1^Un-s. We determined the sequence of the 2.4-kb fragment and designed the primers Un-s1 and Un-s2 for the cassette-ligation-mediated PCR method (see MATERIALS AND METHODS and online supplemental Table 2 available at http://www.genetics.org/supplemental/). This method allowed us to isolate a 2.2-kb genomic fragment including segments that flanked the deletion interval from homozygous Pax1^Un-s genomic DNA. The sequence comparison of the Pax1^Un-s-derived 2.2-kb fragment with the wild-type-derived 2.4-kb EcoRI fragment determined the 3' breakpoint (Fig 3E). For the identification of the 5' breakpoint, we then established the STS marker D2Neu2 (Fig 2A and online supplemental Table 1 available at http://www.genetics.org/supplemental/) from a part of the 2.2-kb fragment flanking the deletion interval on the 5' side. We subcloned a D2Neu2-positive 2-kb HindIII fragment from BAC42. Genomic Southern analysis with the 2-kb HindIII fragment as a probe detected a rearrangement in the Pax1^Un-s genomic DNA (Fig 3C), indicating that the 2-kb HindIII fragment included the 5' breakpoint. Sequence comparison of the Pax1^Un-s-derived 2.2-kb fragment with the wild-type-derived 2-kb HindIII fragment determined the 5' breakpoint (Fig 3E). No further alterations such as insertions or inversions were confirmed in the Pax1^Un-s mutation by sequence comparison and by Southern analysis. Analysis of mouse Pax1-positive BAC sequences (see below) located the 5' breakpoint 50 kb upstream of exon 1 and the 3' breakpoint 65 kb downstream of exon 5. Consequently, the entire length of the Pax1^Un-s deletion interval is 125 kb (Fig 2A). To facilitate the genotyping of the Pax1^Un-s mutants, we established a PCR-based genotyping assay using three specific primers (Fig 3D).

Genes around Pax1:
Two overlapping mouse BAC clones, RP23-224L16 (183 kb) and RP23-382B13 (225 kb), encompassing the Pax1 locus were sequenced through the National Institutes of Health (NIH)-funded Genome Sequencing Network (Trans-NIH Mouse Initiative) upon our proposal (Fig 2A; GenBank accession nos. AC087416 and AC087417). By BLAST search we looked for genes and potential genes represented by expressed sequence tags (ESTs) in a 350-kb genomic region covered by the two BAC sequences. Aside from Pax1 itself, we hit one EST cluster showing significant homology to a segment of the genome sequence, which is located 2.3 kb upstream of Pax1 exon 1. This EST cluster consists of seven ESTs (representatively, GenBank accession nos. AI646519, BF019981, and AI608217) derived from three independent cDNA clones (IMAGE: 1226169, 3470654, and 789910). By sequencing the entire insert of these overlapping clones, we obtained a 752-bp consensus sequence (GenBank accession no. AB080658), which consists of two exons perfectly corresponding to nucleotides 1698–1985 and 2504–2946 of the mouse Pax1 genome sequence AF285175 (Fig 1). The presence of a poly(A) stretch and the consensus splicing signals GT/AG indicates that the 752-bp fragment is part of a real gene. Notably, this gene overlaps with Pax1 exon 0 by 183 nucleotides in an antisense orientation (Fig 1). This gene does not appear to have a real open reading frame, because the sequence contains a number of stop codons in all three reading frames. These data suggest that this gene is a noncoding antisense gene for Pax1 (thus referred to hereafter as Aspax1). The complete primary structure and function of Aspax1 remains to be elucidated.

Extending our gene search, we examined the mouse genome sequence maps (NCBI: http://www.ncbi.nlm.nih.gov and EBI: http://www.ensembl.org). We found Nkx2-2 and Foxa2 (formerly Hnf3ß) located 180 kb upstream and 672 kb downstream of Pax1, respectively (Fig 2A). According to the annotation for the mouse chromosome 2 genome sequence (NW_000178), there are four (LOC241703, LOC241704, LOC228736, and LOC228737) and six (LOC241705, LOC228740, LOC228742, LOC241706, LOC241707, and LOC241708) potential genes in the intervals between Nkx2-2 and Pax1 and between Pax1 and Foxa2, respectively. Aspax1 is not annotated in NW_000178. These putative genes are proposed as gene models on the basis of computer predictions. However, with the exception of LOC228737, no experimental data or EST evidence support the predictions. By further BLAST analysis, the LOC228737 sequence (XM_149219: 650 bases) turned out to correspond to the 3' part of a full-length RIKEN cDNA sequence (AK045921: 4346 bases). By comparing AK045921 with the genome sequence (NW_000178), we found that the AK045921 gene consisted of six exons, spanning over an 150-kb genome interval between Nkx2-2 and Pax1 (Fig 2B). Interestingly, the 5' part of AK045921 (nucleotides 2–281) overlaps to part of Nkx2-2 (nucleotides 798–1077 of NM_010919) in an antisense orientation (Fig 2B). AK045921 does not appear to contain an open reading frame, as there are a number of stop codons in all three reading frames. Thus, AK045921 may represent a noncoding antisense gene for Nkx2-2 (thus referred to hereafter as Asnkx2-2).

In summary, we found Nkx2-2 and Foxa2 as the nearest known genes on either side of Pax1. In the interval between Nkx2-2 and Pax1, we confirmed the presence of two genes, Asnkx2-2 and Aspax1, and they may be noncoding antisense genes for Nkx2-2 and Pax1, respectively. The Pax1^Un-s deletion ablates the last exon of Asnkx2-2, at least two known exons of Aspax1, and all exons of Pax1 (Fig 2B).

BAC transgenic rescue:
We performed the BAC transgenic rescue experiment with three Pax1-positive clones, BAC42 (130 kb), BAC213 (75 kb), and BAC132 (100 kb), which together covered the Pax1^Un-s deletion interval (Fig 2A). We established several BAC transgenic lines (BAC42, four lines; BAC213, one line; BAC132, two lines) and crossed transgenic mice with Pax1^Un-s mice. We could not observe any rescue of the Pax1^Un-s phenotype in the presence of a BAC transgene in heterozygous Pax1^Un-s mice.

Ectopic expression of Nkx2-2 in Pax1^Un-s mutants:
We examined expression of Nkx2-2 in Pax1^Un-s mutant embryos at E10.5 by whole-mount in situ hybridization (Fig 4). Nkx2-2 is normally expressed in the ventral region of the neural tube adjacent to the floor plate (PRICE et al. 1992 ; Fig 4A and Fig F). Notably, Nkx2-2 is ectopically upregulated in the sclerotome compartment of somites and in limb buds of Pax1^Un-s mutant embryos (Fig 4C, Fig D, Fig H, and Fig I). Normally, Pax1 is expressed in both of these embryonic tissues (Fig 4E and Fig J). Expression of Nkx2-2 in the neural tube is not affected in Pax1^Un-s mutants. Interestingly, ectopic expression of Nkx2-2 is not detected in the pharyngeal pouch endoderm, another expression domain of Pax1. Since Pax1^null homozygous embryos do not exhibit such ectopic expression of Nkx2-2 (Fig 4B and Fig G), Pax1 deficiency itself cannot account for this phenomenon. Comparison of Pax1^Un-s heterozygotes (Fig 4C and Fig H) with homozygotes (Fig 4D and Fig I) demonstrated that the level of Nkx2-2 expression in these ectopic sites increases depending on the copy number of the mutant allele, suggesting that a cis-acting mechanism drives this ectopic activation of Nkx2-2. We also examined expression of Foxa2 in Pax1^Un-s mutant embryos at E10.5, but we could not detect any change in its expression (data not shown).

View larger version (66K): In this window In a new window Download PPT slide   Figure 4. Ectopic expression of Nkx2-2 in somites and limb buds of Pax1Un-s mutants. Whole-mount in situ hybridization shows expression of Nkx2-2 (A–D and F–I) and Pax1 (E and J) in wild-type (A, E, F, and J), Pax1null/null (B and G), Pax1Un-s/+ (C and H), and Pax1Un-s/Un-s (D and I) embryos at E10.5. (F–J) Transverse views at the level of the upper thoracic region. The red arrowheads in C and D indicate ectopic expression of Nkx2-2 in forelimb buds. The red and black arrowheads in E point to normal Pax1 expression in the forelimb buds and pharyngeal pouch endoderm, respectively.

Downregulation of Bapx1 in the sclerotome of Pax1^Un-s mutants:
Another homeobox transcription factor of the NK-2 type, Bapx1 (or Nkx3-2), has been shown to be a major mediator of Shh signaling to induce sclerotome chondrogenesis (MURTAUGH et al. 2001 ). Furthermore, it has been recently identified as a transcriptional target of Pax1 and its paralog Pax9 (RODRIGO et al. 2003 ). The results from these studies together suggest that the main role of Pax1 (and Pax9) in sclerotome development is to activate its downstream effecter Bapx1. Thus we examined expression of Bapx1 in Pax1^Un-s mutant embryos (Fig 5). In targeted Pax1^null mutants, Bapx1 expression is either unchanged in heterozygotes or only slightly reduced in homozygotes as a result of compensation by Pax9 (RODRIGO et al. 2003 ). In contrast, in Pax1^Un-s mutants, expression of Bapx1 is significantly reduced, which is already apparent in Pax1^Un-s heterozygous mutants (Fig 5B) and is even more prominent in Pax1^Un-s homozygous mutants (Fig 5C).

View larger version (82K): In this window In a new window Download PPT slide   Figure 5. Downregulation of Bapx1 in the sclerotome of Pax1Un-s mutants. Whole-mount in situ hybridization shows expression of Bapx1 in (A) the control (+/+), (B) heterozygotes (Un-s/+), and (C) homozygotes (Un-s/Un-s) at E10.5. Note that the positive staining in homozygotes is significantly reduced (C). A slight reduction in Bapx1 expression is recognized in heterozygotes, especially in the anterior part of the trunk (the cervical-to-thoracic regions; B).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

To explain the characteristic features of the Pax1^Un-s deletion mutation, we previously proposed two working hypotheses: the co-deletion model and the ectopic activation model (WILM et al. 1998 ). In the present study, we determined the deletion interval and clarified the molecular lesion of the Pax1^Un-s mutation. Our results indicate that the Pax1^Un-s mutation affects four genes within or nearby the deletion: Pax1, Nkx2-2, Asnkx2-2, and Aspax1. We have shown that the Pax1^Un-s deletion ablates all exons of Pax1, the last exon of Asnkx2-2, and at least the last two exons of Aspax1. Furthermore, we have demonstrated that in Pax1^Un-s mutants Nkx2-2 is ectopically activated in normally Pax1-positive domains. Thus, the Pax1^Un-s mutation leads to both the co-deletion and the ectopic activation of additional genes. Nevertheless, the co-deletion of Asnkx2-2 and Aspax1 may not contribute significantly to the Pax1^Un-s phenotype. Asnkx2-2 and Aspax1 are most probably noncoding antisense genes for Nkx2-2 and Pax1, respectively. Noncoding antisense RNA genes are usually involved in post-transcriptional regulation (EDDY 2001 ; ERDMANN et al. 2001 ). If this were the case, Asnkx2-2 and Aspax1 would function only in the context of the functions of their respective sense genes. Indeed, Asnkx2-2 is unlikely to have other functions that are not related to Nkx2-2. The knockout allele of Nkx2-2 reported by SUSSEL et al. 1998 was created by deleting both exon 1 and exon 2 of Nkx2-2, the latter of which contained exon 1 of Asnkx2-2 (see Fig 2B). Thus, this strategy inactivates both Nkx2-2 and Asnkx2-2 at the same time. Nevertheless, the observed defects in the knockout mutants are all related to Nkx2-2-positive tissues, strongly suggesting that Asnkx2-2 functions in the context of Nkx2-2 or that Asnkx2-2 is dispensable. The disturbance of Asnkx2-2 alone could lead to an enhanced translation of Nkx2-2, but this effect is possible only when Nkx2-2 sense mRNA is present. Therefore, the ectopic expression of Nkx2-2 should play a primary role, and the contribution by the disturbance of Asnkx2-2, if at all, should be secondary. Likewise, the disruption of Aspax1 and Pax1 at the same time may not result in a mutant allele with a more extreme phenotypic effect than that of a Pax1 deficiency. In any event, the functions of Asnkx2-2 and Aspax1 have to be determined in future studies to assess the contributions of these two genes to the Pax1^Un-s phenotype. Currently, we are not able to determine their expression profiles by Northern and in situ hybridization analyses, suggesting that they are expressed at low levels and/or that their transcripts are unstable, which is often the case with antisense RNA genes (STORZ 2002 ). In addition, our result from the BAC transgenic rescue experiment does not support the presence of more genes within the deletion interval that contribute to the Pax1^Un-s phenotype.

On the other hand, ectopic expression of Nkx2-2 in the somites and limb buds of Pax1^Un-s mutant embryos is remarkable and strikingly similar to normal expression of Pax1. Thus, this result suggests that cis-regulatory elements or enhancers, which normally direct Pax1 expression, still remain in the Pax1^Un-s genome and drive expression of Nkx2-2 in Pax1^Un-s mutants. Pax1 and Nkx2-2 show different and discrete expression profiles in the wild-type situation. Therefore, we can assume an insulation mechanism that limits enhancer activities of Pax1 or Nkx2-2 enhancers to their respective genes. We propose a model to explain how the Pax1-like expression of Nkx2-2 in Pax1^Un-s mutants occurs, as depicted in Fig 6 (insulator model). In this model, the enhancers that direct expression of Pax1 in the somites and limb buds are located outside the deletion interval, i.e., >65 kb downstream of Pax1 in the wild-type genome. Consistent with this assumption, we could not see any rescue for the skeletal defects in Pax1^null mutants by the BAC transgenes (BAC42, BAC213, and BAC132; see Fig 2A) in multiple lines, suggesting that the BAC intervals defined by the Pax1-positive BAC clones do not contain cis-regulatory elements sufficient for the expression of Pax1 in the sclerotome and the limb buds. Furthermore, in this model we assume that some cis-element that defines a boundary between Pax1 and Nkx2-2 is located up to 50 kb upstream of Pax1. Such a cis-element can be regarded as an insulator, as defined in a recent review (WEST et al. 2002 ). This insulator blocks the action of the Pax1 enhancers on the Nkx2-2 promoter in the normal situation (Fig 6, top). In Pax1^Un-s mutants, the insulator is ablated by the deletion, thereby allowing the Pax1 enhancers to illegitimately communicate with the Nkx2-2 promoter (Fig 6, bottom). On the other hand, the absence of ectopic Nkx2-2 in the pharyngeal pouch endoderm of Pax1^Un-s embryos suggests that the Pax1 pharyngeal pouch endoderm enhancer is located within the Pax1^Un-s deletion interval. Alternative to this insulator model, the insulation mechanism might be established simply by the distance between the two loci. If it is the case, the shortened distance between the Nkx2-2 promoter and the Pax1 enhancers would account for the regulatory interference in the Pax1^Un-s allele, although this explanation is inconsistent with a distance-independent manner of enhancer actions (BLACKWOOD and KADONAGA 1998 ). The molecular basis of the gene insulation here remains to be elucidated in future studies.

View larger version (14K): In this window In a new window Download PPT slide   Figure 6. A model for the mechanism of ectopic activation of Nkx2-2 in Pax1Un-s mutants. Enhancers (e) of Pax1 expression in the pharyngeal pouch endoderm (pe), somites (s), and limb buds (lb) are indicated by ovals. Note that in this model pe is located inside the Pax1Un-s deletion, while the s and lb enhancers reside distal to the 3'-deletion breakpoint. In the wild-type situation, a putative cis-element or insulator (ins) directs the interaction of the Pax1 enhancers only with the promoter (p) of Pax1, not with the Nkx2-2 promoter, thereby establishing a boundary between Pax1 and Nkx2-2. The Pax1Un-s deletion eliminates all exons and the promoter of Pax1, together with the pe enhancer and the insulator, thus allowing the s and lb enhancers to interact with the Nkx2-2 promoter.

What could ectopically expressed Nkx2-2 lead to? Since Nkx2-2 is implicated in the cell-type specification during the development of the central nervous system and pancreatic islets (SUSSEL et al. 1998 ; BRISCOE et al. 1999 ; MCMAHON 2000 ; QI et al. 2001 ; SOULA et al. 2001 ), ectopic expression of Nkx2-2 might have a potential to change the fate of sclerotomal cells. We tested this possibility by examining expression of Nkx6-1 as a marker that is often coexpressed by Nkx2-2-positive cells in the ventral neural tube (MCMAHON 2000 ) and in the pancreas (SANDER et al. 2000 ), but we could not detect any change in Nkx6-1 expression (data not shown). Thus it seems unlikely that cells with ectopic expression of Nkx2-2 in Pax1^Un-s mutants change their fates.

We have demonstrated that Bapx1 (or Nkx3-2) is downregulated more significantly in Pax1^Un-s mutants than in Pax1-null mutants. Since Bapx1 is the main effecter of Shh signaling to induce sclerotome chondrogenesis (MURTAUGH et al. 2001 ), this observation accounts well for the difference in the phenotype severity between the Pax1^Un-s and Pax1-null mutations. Bapx1 is a downstream target of Pax1 and Pax9, but Bapx1 expression in Pax1-null mutants is only slightly reduced due to the compensation by Pax9 (RODRIGO et al. 2003 ). Therefore, in addition to the loss of Pax1, there must be an additional mechanism that results in the enhanced downregulation of Bapx1 in Pax1^Un-s. We assume that ectopically activated Nkx2-2 may fulfill this role. Bapx1 and Nkx2-2 are NK homeobox transcription factors and share a high degree of structural similarities, including the presence of the TN domain, NK2 homeodomain (characterized by the presence of tyrosine at homeodomain position 54), and NK2-specific domain (NK2-SD; TRIBIOLI et al. 1997 ). It has been shown that Bapx1 acts as a transcriptional repressor to exert its chondrogenic function in the sclerotome (MURTAUGH et al. 2001 ). Bapx1 can be converted into a transcriptional activator by appending a strong activator domain from VP16 (designated as Nkx3.2-VP16 in MURTAUGH et al. 2001 ). When Nkx3.2-VP16 is overexpressed in chick somites, sclerotome chondrogenesis is inhibited (MURTAUGH et al. 2001 ) and expression of endogenous Bapx1 is downregulated, due to the disturbance of the positive autoregulatory loop that maintains expression of Bapx1 (ZENG et al. 2002 ). Thus these defects caused by overexpressed Nkx3.2-VP16 in somites mirror what we have demonstrated in Pax1^Un-s mutants, suggesting the possibility that ectopically expressed Nkx2-2 in the sclerotome may act as a transcriptional activator, thereby interfering with Bapx1 function. Nkx2-2 has a potent activator domain at the C-terminal part, and the NK2-SD domain, presumably together with a cofactor, masks the transcriptional activity of this domain (WATADA et al. 2000 ). Therefore, under an abnormal condition (i.e., ectopic expression) the activity of the C-terminal domain may not be properly masked by NK2-SD, leading to transcriptional activation by Nkx2-2.

In conclusion, we propose that a combination of the Pax1 deficiency and the ectopic activation of Nkx2-2 by the Pax1 enhancers is the molecular basis of the Pax1^Un-s mutation. This combined molecular defect accounts for the Pax1^Un-s phenotype that is significantly enhanced but is restricted to Pax1-positive tissues.

	FOOTNOTES

Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AF285175, AB080656, AB080657, and AB080658.
¹ These authors contributed equally to this work.
² Present address: Graduate School of Frontier Biosciences, Osaka University, Suita, Osaka 565-0871, Japan.
³ Present address: Division of Cardiovascular Medicine, Vanderbilt School of Medicine, Nashville, TN 37232.

	ACKNOWLEDGMENTS

We thank M. Karasawa for her technical advice for the deletion interval analysis and H. Hamada and the members of his laboratory for help with whole-mount in situ analysis. We acknowledge E. Knapik and J. Favor for critical reading of the manuscript. We also acknowledge R. Hill for the Bapx1 probe, the Resource Center of the German Human Genome Project for distributing EST clones, and the University of Oklahoma for BAC sequencing through the NIH-funded Genome Sequencing Network. This work was supported in part by grants from the Deutsche Forschungsgemeinschaft and the GSF-National Research Center for Environment and Health, and in part by the Ministry of Education, Culture, Sports, Science and Technology of Japan, with Special Coordinating Funds for Promoting Science and Technology (K.A.).

Manuscript received February 13, 2003; Accepted for publication May 27, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

BALLING, R., U. DEUTSCH, and P. GRUSS, 1988  undulated, a mutation affecting the development of the mouse skeleton, has a point mutation in the paired box of Pax1.. Cell 55:531-535.[Medline]

BLACKWOOD, E. M. and J. T. KADONAGA, 1998  Going the distance: a current view of enhancer action. Science 281:60-63.[Abstract/Free Full Text]

BLANDOVA, Z. K. and I. EGOROV, 1975  Sut allelic with un. Mouse News Lett. 52:43.

BRISCOE, J., L. SUSSEL, P. SERUP, D. HARTIGAN-O'CONNOR, and T. M. JESSELL et al., 1999  Homeobox gene Nkx2.2 and specification of neuronal identity by graded Sonic hedgehog signaling. Nature 398:622-627.[Medline]

CARNINCI, P., Y. SHIBATA, N. HAYATSU, Y. SUGAHARA, and K. SHIBATA et al., 2000  Normalization and subtraction of cap-trapper-selected cDNAs to prepare full-length cDNA libraries for rapid discovery of new genes. Genome Res. 10:1617-1630.[Abstract/Free Full Text]

DAHL, E., H. KOSEKI, and R. BALLING, 1997  Pax genes and organogenesis. Bioessays 19:755-765.[Medline]

DEUTSCH, U., G. R. DRESSLER, and P. GRUSS, 1988  Pax 1, a member of a paired box homologous murine gene family, is expressed in segmented structures during development. Cell 53:617-625.[Medline]

DIETRICH, S. and P. GRUSS, 1995  undulated phenotypes suggest a role of Pax-1 for the development of vertebral and extravertebral structures. Dev. Biol. 167:529-548.[Medline]

EDDY, S. R., 2001  Non-coding RNA genes and the modern RNA world. Nat. Rev. Genet. 2:919-929.[Medline]

ERDMANN, V. A., M. Z. BARCISZEWSKA, M. SZYMANSKI, A. HOCHBERG, and N. DE GROOT et al., 2001  The non-coding RNAs as riboregulators. Nucleic Acids Res. 29:189-193.[Abstract/Free Full Text]

GRÜNEBERG, H., 1950  Genetical studies on the skeleton of the mouse. II. Undulated and its "modifiers. J. Genet. 50:142-173.

GRÜNEBERG, H., 1954  Genetical studies on the skeleton of the mouse. XII. The development of undulated. J. Genet. 52:441-455.

ISEGAWA, Y., J. SHENG, Y. SOKAWA, K. YAMANISHI, and O. NAKAGOMI et al., 1992  Selective amplification of cDNA sequence from total RNA by cassette-ligation mediated polymerase chain reaction (PCR): application to sequencing 6.5 kb genome segment of hantavirus strain B-1. Mol. Cell. Probes 6:467-475.[Medline]

LETTICE, L., J. HECKSHER-SORENSEN, and R. HILL, 2001  The role of Bapx1 (Nkx3.2) in the development and evolution of the axial skeleton. J. Anat. 199:181-187.[Medline]

MANSOURI, A., G. GOUDREAU, and P. GRUSS, 1999  Pax genes and their role in organogenesis. Cancer Res. 59:1707s-1710s.

MCMAHON, A. P., 2000  Neural patterning: the role of Nkx genes in the ventral spinal cord. Genes Dev. 14:2261-2264.[Free Full Text]

MOUNT, S. M., 1982  A catalogue of splice junction sequences. Nucleic Acids Res. 10:459-472.[Abstract/Free Full Text]

MURTAUGH, L. C., L. ZENG, J. H. CHYUNG, and A. B. LASSAR, 2001  The chick transcriptional repressor Nkx3.2 acts downstream of Shh to promote BMP-dependent axial chondrogenesis. Dev. Cell 1:411-422.[Medline]

NEUBÜSER, A., H. KOSEKI, and R. BALLING, 1995  Characterization and developmental expression of Pax9, a paired-box-containing gene related to Pax1. Dev. Biol. 170:701-716.[Medline]

PABST, O., H. HERBRAND, and H. H. ARNOLD, 1998  Nkx2-9 is a novel homeobox transcription factor which demarcates ventral domains in the developing mouse CNS. Mech. Dev. 73:85-93.[Medline]

PRICE, M., D. LAZZARO, T. POHL, M. G. MATTEI, and U. RUTHER et al., 1992  Regional expression of the homeobox gene Nkx-2.2 in the developing mammalian forebrain. Neuron 8:241-255.[Medline]

QI, Y., J. CAI, Y. WU, R. WU, and J. LEE et al., 2001  Control of oligodendrocyte differentiation by the Nkx2.2 homeodomain transcription factor. Development 128:2723-2733.

RODRIGO, I., R. E. HILL, R. BALLING, A. MUNSTERBERG, and K. IMAI, 2003  Pax1 and Pax9 activate Bapx1 to induce chondrogenic differentiation in the sclerotome. Development 130:473-482.[Abstract/Free Full Text]

SANDER, M., L. SUSSEL, J. CONNERS, D. SCHEEL, and J. KALAMARAS et al., 2000  Homeobox gene Nkx6.1 lies downstream of Nkx2.2 in the major pathway of beta-cell formation in the pancreas. Development 127:5533-5540.[Abstract]

SOULA, C., C. DANESIN, P. KAN, M. GROB, and C. PONCET et al., 2001  Distinct sites of origin of oligodendrocytes and somatic motoneurons in the chick spinal cord: oligodendrocytes arise from Nkx2.2-expressing progenitors by a Shh-dependent mechanism. Development 128:1369-1379.[Abstract]

SPÖRLE, R. and K. SCHUGHART, 1998  Paradox segmentation along inter- and intrasomitic borderlines is followed by dysmorphology of the axial skeleton in the open brain (opb) mouse mutant. Dev. Genet. 22:359-373.[Medline]

STORZ, G., 2002  An expanding universe of noncoding RNAs. Science 296:1260-1263.[Abstract/Free Full Text]

SUSSEL, L., J. KALAMARAS, D. J. HARTIGAN-O'CONNOR, J. J. MENESES, and R. A. PEDERSEN et al., 1998  Mice lacking the homeodomain transcription factor Nkx2.2 have diabetes due to arrested differentiation of pancreatic beta cells. Development 125:2213-2221.[Abstract]

TIMMONS, P. M., J. WALLIN, P. W. RIGBY, and R. BALLING, 1994  Expression and function of Pax 1 during development of the pectoral girdle. Development 120:2773-2785.[Abstract]

TRIBIOLI, C., M. FRASCH, and T. LUFKIN, 1997  Bapx1: an evolutionary conserved homologue of the Drosophila bagpipe homeobox gene is expressed in splanchnic mesoderm and the embryonic skeleton. Mech. Dev. 65:145-162.[Medline]

WALLACE, M. E., 1980  Undulated alleles. Mouse News Lett. 63:10.

WALLIN, J., J. WILTING, H. KOSEKI, R. FRITSCH, and B. CHRIST et al., 1994  The role of Pax-1 in axial skeleton development. Development 120:1109-1121.[Abstract]

WALLIN, J., H. EIBEL, A. NEUBÜSER, J. WILTING, and H. KOSEKI et al., 1996  Pax1 is expressed during development of the thymus epithelium and is required for normal T-cell maturation. Development 122:23-30.[Abstract]

WATADA, H., R. G. MIRMIRA, J. KALAMARAS, and M. S. GERMAN, 2000  Intramolecular control of transcriptional activity by the NK2-specific domain in NK-2 homeodomain proteins. Proc. Natl. Acad. Sci. USA 97:9443-9448.[Abstract/Free Full Text]

WEST, A. G., M. GASZNER, and G. FELSENFELD, 2002  Insulators: many functions, many mechanisms. Genes Dev. 16:271-288.[Free Full Text]

WILM, B., E. DAHL, H. PETERS, R. BALLING, and K. IMAI, 1998  Targeted disruption of Pax1 defines its null phenotype and proves haploinsufficiency. Proc. Natl. Acad. Sci. USA 95:8692-8697.[Abstract/Free Full Text]

WRIGHT, M. E., 1947  Undulated: a new genetic factor in Mus musculus affecting the spine and tail. Heredity 1:137-141.

ZENG, L., H. KEMPF, L. C. MURTAUGH, M. E. SATO, and A. B. LASSAR, 2002  Shh establishes an Nkx3.2/Sox9 autoregulatory loop that is maintained by BMP signals to induce somitic chondrogenesis. Genes Dev. 16:1990-2005.[Abstract/Free Full Text]

This article has been cited by other articles:

				C. Kokubu, U. Heinzmann, T. Kokubu, N. Sakai, T. Kubota, M. Kawai, M. B. Wahl, J. Galceran, R. Grosschedl, K. Ozono, et al. Skeletal defects in ringelschwanz mutant mice reveal that Lrp6 is required for proper somitogenesis and osteogenesis Development, November 1, 2004; 131(21): 5469 - 5480. [Abstract] [Full Text] [PDF]