The jing Zn-finger transcription factor is a mediator of cellular differentiation in the Drosophila CNS midline and trachea

Transcriptional regulatory mechanisms play an important role in cell fate determination in both vertebrates and invertebrates. Transcription factors have multiple roles in controlling cellular competence or determination and also in regulating differentiation (Cau et al., 2000; Hallam et al., 2000; Moran-Rivard et al., 2001; Pierani et al., 2001; Portman and Emmons, 2000). The combinatorial nature of transcriptional regulation during cellular specification is emerging from studies of diverse embryonic tissues including the nervous system and respiratory system (trachea) (Boube et al., 2000; Jurata et al., 2000; Ma et al., 2000; Zelzer and Shilo, 2000). In addition, different categories of transcription factors can function at successive steps during cellular development. In this way, cellular responses may be dictated by the temporal and spatial characteristics of multiple and/or diverse types of transcription factors.

In Drosophila, the ventral midline and respiratory system (trachea) are ectodermal derivatives patterned by positional cues present in embryos. The ventral midline is patterned by the combinatorial actions of dorsal/ventral (D/V) and neurogenic genes that confine the expression of the bHLH-PAS transcription factor single-minded (sim) to the mesectoderm (Crews, 1998; Morel and Schweisguth, 2000). Development of the entire CNS midline requires the regulatory functions of sim and in the absence of sim function midline cells take on lateral neuroectodermal fates (Crews, 1998; Estes et al., 2001; Nambu et al., 1991; Xiao et al., 1996). The midline-inducing capabilities of sim were discovered by ectopic expression experiments (Nambu et al., 1991). Subsequent CNS midline gene regulation involves the combinatorial functions of three different transcription factors including bHLH-PAS, SOX and POU domain-containing proteins (Ma et al., 2000). CNS midline precursors give rise to midline glia and various interneuron and motoneuron lineages including two MP1 neurons, two UMI neurons, the MNB and VUMs (Klämbt et al., 1991; Bossing and Technau, 1994; Schmid et al., 1999).

The tracheal placodes are specified by TGFβ signaling along the dorsoventral axis and Wingless (WG) signaling along the anteroposterior axis (Affolter et al., 1994; de Celis et al., 1995; Wilk et al., 1996). These cues are responsible for independently activating primary genes such as the bHLH-PAS transcription factor trachealess (trh) and the POU domain transcription factor ventral veinless (vvl) (previously known as drifter), which are required in a combinatorial fashion for subsequent tracheal development (Boube et al., 2000; Isaac and Andrew, 1996; Llimargas and Casanova, 1997; Wilk et al., 1996; Zelzer and Shilo, 2000). In the absence of trh and vvl function tracheal cells fail to invaginate and tracheal tubules do not form (de Celis et al., 1995; Isaac and Andrew, 1996; Wilk et al., 1996). Ectopic trh expression can induce tracheal pits and therefore it has been considered an inducer of cell fates (Wilk et al., 1996). Within the trachea, the DPP pathway specifies the fates of branches that will give rise to the dorsal branch and lateral anterior and posterior branches (Affolter et al., 1994; Vincent et al., 1997; Wappner et al., 1997). Activation of the EGF receptor pathway is required for specifying the dorsal trunk and visceral branch (Wappner et al., 1997). In addition, the WNT pathway is required for specification of the dorsal trunk (Llimargas, 2000; Chihara and Hayashi, 2000).

The invertebrate ortholog of the aryl hydrocarbon nuclear translocator, known as tango (tgo), encodes a common partner for SIM and TRH (Oshiro and Saigo, 1997; Sonnenfeld et al., 1997; Zelzer et al., 1997). TGO is present in the cytoplasm and translocates to the nucleus upon expression of a dimerization partner such as sim, trh or Spineless-Aristapedia (ss) (Emmons et al., 1999; Ward et al., 1998). Therefore, the precise regulation of lineage-specific transcriptional regulators such as sim, trh and ss is critical. In both the CNS midline and trachea, TGO::SIM and TGO::TRH heterodimers activate common target genes containing asymmetrical E-box sites with an ACGTG core (Crews, 1998; Zelzer and Shilo, 2000). These E-box sequences are sufficient to drive both midline and tracheal expression and are required for the expression of known target genes including the breathless fibroblast growth factor receptor and the repulsive guidance molecule slit (Battye et al., 1999; Glazer and Shilo, 1991; Kidd et al., 1999; Sonnenfeld et al., 1997; Wharton et al., 1994).

In this study, we have used genetic and cellular analysis to establish novel roles for the jing zinc-finger transcription factor in the differentiation of CNS midline and tracheal cells. A genetic approach to identify molecules required for the commitment of CNS midline and tracheal cells led to the identification and characterization of the jing locus during embryogenesis. The jing locus has been previously identified in genetic screens and recently characterized for its role in border cell migration in Drosophila ovaries (Karpen and Spradling, 1992; Liu and Montell, 2001). During embryogenesis, jing transcripts and protein are detected in the CNS midline, trachea and segmental ectodermal stripes. Gene dosage and overexpression experiments reveal that appropriate levels of jing in the CNS midline and trachea are crucial for formation of CNS commissural and longitudinal axons as well as tracheal tubules, respectively. Loss-of-function mutations in jing are associated with reductions in cell-type gene expression and inappropriate apoptosis of CNS midline and tracheal precursors. These results therefore establish that jing is required in a positive manner to promote cellular differentiation and survival in embryonic ectodermal lineages.

The wild-type strain was w¹¹⁸. The tango¹ (tgo¹) allele and the tgo¹trh⁸ double mutant strain have been previously described (Sonnenfeld et al., 1997). Double mutant stocks were generated by standard genetic procedures using P element jing alleles and null alleles of single-minded (sim^H9) and trachealess (trh¹) (Isaac and Andrew, 1996; Nambu et al., 1990). The second and third balancer chromosomes in these stocks carried the lacZ-markers CyOP[wg-lacZ] and TM3P[ubx-lacZ]. The slit (sli) reporter P[ sli 1.0 HV-lacZ] was used to assess transcription of the sli gene (Ma et al., 2000; Wharton et al., 1994). The breathless^H82Δ3 P-element excision allele (Klämbt et al., 1992) and the null sli¹ allele (Nüsslein-Volhard et al., 1984; Rothberg et al., 1990) were used to assess genetic interactions with multiple jing alleles.

P-element lethal stocks were obtained from the Indiana University Drosophila Stock Center (Bloomington, Indiana). The jing-lacZ enhancer trap strain (jing⁰¹⁰⁹⁴) contains an embryonic recessive lethal insertion of the P-element P[PZ] originally designated l(2)01094 (BDGP) (Karpen and Spradling, 1992; Spradling et al., 1999). l(2)01094^K03404 is a second embryonic lethal insertion of the P[lacW] P-element in the jing gene (from I. Kiss, T. Laverty and G. Rubin) (Liu and Montell, 2001) and we refer to this allele as jing^K03404. Both alleles do not complement the lethality of a jing deficiency Df(2R)ST1 (Liu and Montell, 2001).

The P[sim-UAS] and P[prd-Gal4] strains were used to ectopically express sim under control of a prd enhancer in a background heterozygous for the jing-lacZ enhancer trap (Brand and Perrimon, 1993; Ward et al., 1998; Xiao et al., 1996). P[sim-GAL4] and P[btl-GAL4] were used as drivers to overexpress jing in the CNS midline and trachea, respectively (Shiga et al., 1996; Ward et al., 1998).

Genomic DNA surrounding the P element insertions in jing⁰¹⁰⁹⁴ and jing^K03404flies, and including jing-coding sequences, was sequenced and deposited into GenBank (AF285778). Expressed sequence tags (ESTs) LD10015, LD36562 and LD10101 were identified by database searching, obtained from Research Genetics (Birmingham, AL) and subjected to DNA sequence analysis. Gel-purified fragments of PCR-generated genomic and EST DNA were sequenced on both strands by dye terminator cycle DNA sequencing (Perkin Elmer) using an ABI PRISM Genetic Analyzer. LD36562 and LD10101 sequences were identical to FlyBase Genome Annotation Database (GadFly) identifier CG9403 (http://flybase.bio.indiana.edu/annotl) (Adams et al., 2000).

The jing full-length cDNA (LD36562) was cloned into pUAST (Brand and Perrimon, 1993) and together with pUChspΔ2,3 P element helper plasmid was injected into y w embryos and w⁺ transformants were selected (Spradling, 1986). Three transgenic UAS-jing lines produced similar results in overexpression experiments using P[sim-GAL4] and P[btl-GAL4] as drivers (Shiga et al., 1996; Ward et al., 1998).

For antibody production, the JING peptide VPAASANKNKRTAAG (amino acids 81-95) was synthesized (Eastern Quebec Proteomics Core Facility), coupled to KLH (Sigma) and used to generate anti-JING rat antisera (PRF&L). JING antibody specificity was confirmed by examining embryos homozygous for a deficiency in jing (Df(2R)ST1) and after ectopic expression of jing (prd-GAL4/jing-UAS).

The following antibodies were used: rat anti-JING (1:100; this work); mAb anti-β-galactosidase (anti-β-gal) (Promega); rabbit polyclonal anti-β-gal (Promega); rat polyclonals anti-Single-minded (anti-SIM) and anti-Trachealess (anti-TRH) (Sonnenfeld et al., 1997; Ward et al., 1998); anti-Wrapper and mAb 1D4 anti-FASCICLIN II (Noordemeer et al., 1998; Lin et al., 1994) (gifts from C. S. Goodman); mAb anti-Slit (a gift from Spyros Artavanis-Tsakonas); mAb 2A12 (a gift from N. Patel); and rabbit anti-Odd-skipped (ODD) (Skeath and Doe, 1998; Spana et al., 1995) (a gift from Jim Skeath). The following antibodies were obtained from the Developmental Studies Hybridoma Bank: mAb BP102; 22C10/FUTSCH (Fujita et al., 1982; Hummel et al., 2000); and mAb 4D9 anti-Engrailed/Invected (Patel et al., 1989).

Embryo staging was carried out according to Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein, 1985) and processing for light microscopy was undertaken according to standard protocols (Patel, 1994). JING protein distribution was determined by staining whole-mount embryos with rat anti-JING at 1:100 dilution. Antibody staining was visualized using HRP- or rhodamine-conjugated secondary antibodies (Jackson). Balancer chromosomes carrying lacZ for the second (CyOP[wg-lacZ]) and third chromosomes (TM3P[ubx-lacZ]) were used to identify homozygous mutant embryos after anti-β-gal staining. HRP-labeled embryos were analyzed by light microscopy using a Zeiss Axioskop.

In situ hybridization was performed on w¹¹⁸ whole-mount embryos as described (Tautz and Pfeifle, 1989). DNA probes were generated by random priming using wild-type Drosophila genomic DNA (GenBank Accession AF285778) and jing cDNA (LD36562) as templates. All probes showed identical expression patterns. jing DNA probes were labeled with dig-11-dUTP (Boehringer Mannheim) and their specificity determined by in situ hybridization to embryos carrying the Df(2R)ST1 deficiency. Embryos were analyzed by light microscopy.

jing enhancer trap expression was analyzed after staining heterozygous jing⁰¹⁰⁹⁴ embryos with anti-β-gal and a secondary antibody conjugated with FITC or rhodamine at 1/250. TUNEL (TMR red; Roche) staining was carried out according to previous procedures and was double stained with anti-SLI, -TRH or -Odd-skipped (Booth et al., 2000). Fluorescently labeled embryos were mounted in 4% n-propyl gallate to inhibit photobleaching and analyzed on a Zeiss Axiovert 100 TV confocal microscope. Optical sections of 1 μm were recorded in line average mode. All figures were processed using Adobe Photoshop software.

An F₂ lethal complementation screen using ethylmethane sulfonate (EMS) was performed as described (Grigliatti, 1986; Sonnenfeld et al., 1997). Three lethal EMS-induced jing mutations were tested for genetic complementation inter se and with deficiency Df(2R)ST1 (Lui and Montell, 2001). All mutant chromosomes were balanced over CyOP[wg-lacZ] marked balancers. The jing³ EMS-induced allele was sequenced according to previous procedures (Sonnenfeld et al., 1997). Embryonic lethal and viable excision (reversion) jing⁰¹⁰⁹⁴ and jing^K03404 alleles were obtained by standard procedures (Bellen et al., 1989; Robertson et al., 1988). Lethal and viable excisions were isolated by loss of eye color and by genetic complementation and were mapped by PCR using jing-specific primers and DNA sequence analysis.

Dosage-sensitive genetic interactions between two loci are a good indicator that two gene products are functionally related. We identified the l(2)01094 gene (BDGP) in a search for P-element-induced mutations displaying severe CNS axon phenotypes in double heterozygous combination with null mutations in sim. The l(2)01094 gene has been previously isolated in genetic screens and recently characterized for its role in border cell migration in Drosophila ovaries (Karpen and Spradling, 1992; Liu and Montell, 2001). l(2)01094 encodes a zinc-finger transcription factor called jing and we refer to this P-element-induced allele as jing⁰¹⁰⁹⁴ (Liu and Montell, 2001).

To address whether jing dosage is important for CNS midline development, jing P-element insertion mutant alleles were placed in heterozygous combination with null mutations in genes whose primary effects arise from the CNS midline, including sim and sli mutations. We also tested hypomorphic tgo mutations. CNS axon and midline cell development were assessed in double heterozygous embryos by BP102, anti-SIM or anti-SLI staining (Klämbt et al., 1991; Rothberg et al., 1990; Ward et al., 1998). jing⁰¹⁰⁹⁴ alleles perturb CNS axon formation and midline cell development in double heterozygous combination with sim^H9 (Fig. 1C,I; Fig. 2C), tgo¹ (Fig. 1E,I), and sli¹ (Fig. 1I; Fig. 2I,K). For example, 54% of jing and sim double heterozygotes show improper commissural and longitudinal axon formation (‘stalled axons’). A smaller percentage of jing⁰¹⁰⁹⁴ and sim^H9 double heterozygotes (7.7%) show ‘collapsed axon’ phenotypes similar to those of sim or sli homozygotes (Fig. 1B,I) (Nambu et al., 1990; Rothberg et al., 1990). The phenotypes of jing and sim double heterozygotes are insertion dependent as they revert to wild- type after precise excision of the P element in jing⁰¹⁰⁹⁴ flies (Fig. 1D,I; not shown).

CNS axon and midline cell development are perturbed in embryos triple heterozygous for jing, sim and tgo (Fig. 1F,I; Fig. 2B,E,G). However, unlike in homozygous sim mutants, the midline cells in jing, sim and tgo triple heterozygotes are specified but then fail to differentiate properly, as determined by their displacement from the ventral nerve cord (a feature characteristic of apoptotic cells) and loss of Sim immunoreactivity by stage 15 (Fig. 1B; Fig. 2B,E,G) (Sonnenfeld and Jacobs, 1995). The ventral displacement of the CNS midline cells occurs after reduction of one copy of jing and sim, suggesting that these effects are specific for the midline (Fig. 2C). Triple heterozygotes also show alterations in repulsive signaling mechanisms. Fasciclin 2-positive longitudinal axons collapse into a single tract along the midline in jing⁰¹⁰⁹⁴; tgo¹ sim^H9 triple heterozygotes stained with 1D4 monoclonal antibody (Fig. 1G,H) (Van Vactor et al., 1993). These phenotypes are, therefore, similar to those of homozygous mutations in sli¹, which affect midline repulsion mechanisms and cause the ventral displacement of midline cells (Sonnenfeld and Jacobs, 1994; Battye et al., 1999; Kidd et al., 1999).

To characterize the relationship between jing and CNS midline further, we removed one copy of both jing and sli and analyzed the development of the CNS axons and midline cells. Reducing one copy of both jing and sli is associated with collapsed axons (55%), the ventral displacement of SIM⁺ midline cells (38%) and reductions in SLI immunoreactivity (40%) in stage 14 embryonic nerve cords compared with wild type (Fig. 1I; Fig. 2I-K). By comparison, 57% of sim^H9 and sli¹ double heterozygotes have collapsed axons (Fig. 1I) and ventrally displaced midline cells (45%; Fig. 2I), which is consistent with the established regulatory role of sim (Fig. 2H) (Ma et al., 2000; Wharton et al., 1994). Comparison of SIM and SLI immunoreactivity in jing and sli double heterozygotes therefore reveals that although midline cells are present in these embryos, they do not adequately express sli. In summary, these results imply that jing dosage may be important for the regulation of sli.

We next assessed whether jing dosage is important for tracheal development by analyzing jing in trans-heterozygous combination with mutations in genes whose function is specific for the embryonic trachea. Tracheal tubule development was analyzed in double heterozygous embryos by staining with mAb 2A12, which in wild-type embryos stains the lumen of all tracheal tubules (Fig. 3A). Tracheal tubules do not form in homozygous trh mutants (Fig. 3B) (Isaac and Andrew, 1996; Wilk et al., 1996; Sonnenfeld et al., 1997). Tracheal tubule formation is defective after both trh and jing are reduced by only one copy each. For example, 51% of embryos double heterozygous for jing⁰¹⁰⁹⁴ and trh¹ show a significant loss of most tracheal branches by stage 15 (Fig. 3C,E). In addition, jing⁰¹⁰⁹⁴ and trh¹ double heterozygotes are sensitive to the dose of tgo, as 69% of embryos triple heterozygous for these mutations (jing⁰¹⁰⁹⁴; trh¹ tgo¹) show tracheal phenotypes (Fig. 3E). jing mutations also show dominant interactions with a direct target of TGO and TRH heterodimers, the fibroblast growth factor receptor known as breathless (btl) (Klämbt et al., 1992; Oshiro and Saigo, 1997). Ninety eight percent of jing⁰¹⁰⁹⁴and btl^H82Δ3 double heterozygotes show tracheal phenotypes that affect the formation of transverse connectives and visceral branches (Fig. 3D,E).

In conclusion, the genetic analysis presented here provides strong evidence that proper dose of jing in combination with that of trh or btl is important for tracheal tubule patterning. If jing functions in a parallel pathway to that of trh and btl these results would then indicate that the pathways must converge on a common component that is necessary for tracheal tubule formation.

The jing locus encodes a transcription factor with homology to the mouse transcription factor AEBP2 (Lui and Montell, 2001). jing-coding sequence corresponding to the expressed sequence tag (EST) LD36562 rescues jing mutant effects in the ovary, confirming the identity of this gene (Lui and Montell, 2001). Fig. 4A shows the proximity of embryonic lethal jing P element insertions to a transcription unit (LD10015) adjacent to the jing 5′ regulatory region. Given the proximity of the jing P elements to the LD10015 transcription unit, it was important to determine whether the latter was affected by these insertions. We therefore performed in situ hybridization on embryos homozygous for jing P element insertional mutations (jing⁰¹⁰⁹⁴ and jing^K03404) using digoxigenin-labeled LD10015 EST as a probe. LD10015 mRNA was detected in embryos homozygous for either jing⁰¹⁰⁹⁴or jing^K03404 and therefore we conclude that LD10015 transcription is unaffected by lethal jing P element insertions (data not shown).

Analysis of genomic DNA sequence (GenBank Accession Number, AF285778) surrounding two lethal P-element insertions in jing reveals that there are three putative DNA binding sites for Tgo::Sim and Tgo::Trh (CMEs), and one for the HMG SOX protein called Fish-hook (also known as Dichaete, D – FlyBase) (TACAAT) in the 5′ regulatory region of jing (Fig. 4A) (Ma et al., 2000; Oshiro and Saigo, 1997; Sonnenfeld et al., 1997; Wharton et al., 1994). This raises the possibility that jing may be a direct transcriptional target of bHLH-PAS heterodimers and SOX proteins including TGO:SIM, TGO:TRH or Fish-hook (Crews, 1998; Ma et al., 2000).

Point mutations in jing were isolated by a chemical mutagenesis. From a screen of 6344 EMS-mutagenized second chromosomes, three novel jing mutations were isolated for failure to complement the embryonic lethality of jing^K03404 genetically, therefore defining a single complementation group. jing EMS-induced mutations are homozygous embryonic lethal and are lethal in trans to jing P element-induced mutations and a deficiency Df(2R)ST1 covering the jing locus (Liu and Montell, 2001). Based on phenotypic analysis of the CNS and trachea, the jing EMS-induced alleles were placed in the following allelic series of phenotypic severity: jing³>jing²>jing¹. Molecular analysis of jing³ reveals a single nucleotide change in the coding region of this gene, confirming the identity of this complementation group (Fig. 4B). The jing³ mutation results in the conversion of tryptophan¹²⁰⁰ (w¹²⁰⁰) to a premature stop codon located in the middle of the second zinc-finger motif (Fig. 4B). Given the importance of the zinc-finger motifs and a nuclear localization signal to DNA binding, the molecular nature of the jing³ mutation is consistent with its strong loss-of-function and hemizygous phenotypes. The phenotype of jing³ mutant embryos is therefore shown in phenotypic analyses.

The expression pattern of jing was studied throughout embryogenesis with a jing-lacZ enhancer trap line (jing⁰¹⁰⁹⁴), digoxigenin-labeled jing DNA probes and a rat JING antibody. jing mRNA and protein product are first detected during precellular blastoderm stages, suggesting that Drosophila embryos contain a maternal supply of jing (data not shown). A discernable jing expression pattern is apparent from stage 9, as jing transcripts and protein accumulate in the CNS midline, neuroectoderm and trachea (Fig. 4C-E).

In the wild-type stage 9 CNS, jing mRNA is distributed in a dorsoventral pattern that is not continuous between segments (Fig. 4C). To determine the identity of the jing-expressing CNS cells, co-localization studies were performed using a jing-lacZ enhancer trap and confocal microscopy. Embryos carrying the jing-lacZ enhancer trap and stained with anti-β-gal and anti-SIM show co-localization in subsets of CNS midline cells during stage 9 (Fig. 4D, arrow). As SIM localizes only to midline cells in the CNS, this result confirms the midline expression of jing (Crews, 1998). During stage 9, jing transcription also occurs in the neuroectoderm and in the supraoesophageal ganglion (Fig. 4C,D).

During stage 10, JING protein is present in the tracheal placodes (Fig. 4F). A pair of JING-positive cells flank the tracheal placodes dorsally (Fig. 4F). The jing-lacZ enhancer trap is also expressed in TRH-positive tracheal cells in the anterior of each placode (Fig. 4G). The jing-lacZ enhancer trap is co-expressed with trh and tgo from stage 10 until stage 16 of embryogenesis (data not shown). JING protein is detected in all tracheal branches throughout embryonic tracheal development, consistent with a role for jing throughout tracheal tubulogenesis (Fig. 4H).

During stage 12/3, jing transcripts and protein product are present in CNS midline cells and segmental ectodermal stripes (Fig. 4I-K). By stage 14, jing is strongly expressed in midline glia that occupy a characteristic dorsal position in the ventral nerve cord (Fig. 4L,M). Weaker jing expression is detected in ventrally positioned midline neurons (Fig. 4L, black arrowhead). To determine the subcellular localization of JING in the CNS, wild-type embryos were stained with anti-JING and analyzed by confocal microscopy. By this method, JING protein can be detected within the nuclei of the midline glia (Fig. 4N, arrow) and to a lesser degree in midline neurons (Fig. 4N, arrowhead). JING protein is not detectable by confocal microscopy in cells of the lateral neuroectoderm, as opposed to jing-lacZ expression (see Fig. 5A) (data not shown).

To further analyze jing-lacZ, we assessed its expression in homozygous jing and sim mutants using monoclonal anti-β-gal. During stage 14, the jing-lacZ enhancer shows strong expression in CNS midline cells (Fig. 5A, arrow) and weaker expression in lateral CNS cells (Fig. 5A, arrowhead). In homozygous jing⁰¹⁰⁹⁴ mutants carrying the jing-lacZ P element insertion, lacZ expression is reduced in the entire CNS suggesting that this insertion affects jing gene expression and that jing expression may be controlled by autoregulation (Fig. 5B). By contrast, in stage 15 sim^H9 homozygotes, jing-lacZ enhancer expression is absent in the CNS midline (Fig. 5C, arrow) but still present in the lateral CNS (Fig. 5C, arrowhead) and other areas of embryonic jing expression (data not shown). This result confirms the midline identity of the jing-lacZ-expressing cells.

To assess the midline identity of jing-lacZ enhancer expression further, we determined whether sim activates the jing-lacZ enhancer by in vivo ectopic expression experiments. The ability of sim to induce midline gene expression ectopically has been established (Nambu et al., 1991; Wilk et al., 1996; Xiao et al., 1996; Zelzer et al., 1997). sim expression was targeted to the pair-rule ectodermal stripes of the paired (prd) gene using GAL4/UAS (Brand and Perrimon, 1993) and by crossing flies containing the P[prd-GAL4] driver, and heterozygous for the jing-lacZ enhancer, with flies containing P[UAS-sim] (Ward et al., 1998). The progeny were stained with anti-SIM to confirm ectopic expression (Fig. 5D) and with anti-β-gal to identify ectopic jing-lacZ expression (Fig. 5E). Ectopic expression of sim is sufficient to activate jing-lacZ in ventrally positioned cells in pair-rule ectodermal stripes (Fig. 5E,F, arrows). The ventral activation of jing-lacZ by sim is consistent with previous results showing the activation of midline-specific genes by ectopic sim expression (Xiao et al., 1996). In summary, the results shown here provide strong evidence that jing expression occurs in CNS midline cells.

The jing expression pattern and gene dose effects in the CNS midline and trachea suggest that jing function may be important for the development of both systems. Therefore, CNS axon and tracheal tubule development was assessed in jing homozygous mutant embryos stained with monoclonal antibodies BP102 and 2A12, respectively. In jing³ homozygous mutant embryos, commissural growth cones are often absent in the midline at stage 12 when compared with wild type (Fig. 6A,B). By stage 14, homozygous jing³ mutants show losses of longitudinal connections and reduced commissures compared with wild type (Fig. 6C,D). Embryos double mutant for jing and sim display phenotypes similar to those of sim homozygotes (Fig. 6E). Therefore, the sim embryonic CNS axon phenotype is epistatic to that of jing, implying that jing functions downstream of sim (Avery and Wasserman, 1992).

The GAL4/UAS system was used to determine the effects of overexpressing jing in the CNS midline (Brand and Perrimon, 1993). Flies containing P[sim-GAL4] were crossed to flies containing P[jing-UAS] and their progeny stained with BP102 to assess CNS axon formation. Expression of one copy of P[jing-UAS] specifically in the CNS midline is sufficient to inhibit commissural and longitudinal axon formation (Fig. 6F). Therefore, the jing midline overexpression phenotype is similar to that resulting from jing loss of function (Fig. 6D), and phenotypes of jing and sim double heterozygotes (Fig. 1C). These results demonstrate that appropriate jing dose is a requirement for proper CNS axon development in the CNS midline. Interestingly, we observe a similar CNS axon phenotype after overexpression of sim in the CNS midline (data not shown).

The homozygous jing CNS phenotype suggests an alteration in the mechanisms that guide CNS axons. Fasciclin 2 staining using 1D4 mAb, shows that longitudinal fascicles stall within segment boundaries causing breaks in the longitudinal tracts in 95% of jing³ mutant segments (Fig. 6H, arrowhead; n=210 segments) (Van Vactor et al., 1993). A subset of normally ipsilateral axons of the most medial fascicle project instead contralaterally in jing³ mutants (Fig. 6H, arrows; n=210 segments). As ipsilateral fascicles are prevented from crossing the midline in wild-type embryos (Fig. 6G), these results suggest that midline repulsive mechanisms are perturbed in jing mutant embryos (Hummel et al., 1999b; Kidd et al., 1999).

We next wanted to determine whether jing is involved in tracheal patterning due to its expression pattern and its dose-sensitive effects with mutations in genes controlling tracheal development. Embryos homozygous for a jing deficiency (Df(2R)ST1) and jing³ mutations are associated with losses of the dorsal trunk, severely disrupted transverse connectives and absences of the visceral branch (Fig. 6I,J) compared with wild-type (see Fig. 3A). Embryos doubly mutant for jing and trh lack all tracheal tubules and display phenotypes identical to trh homozygous mutants (Fig. 6K). Therefore, trh loss-of-function is epistatic to jing loss-of-function, implying that jing functions downstream of trh.

To determine the effects of overexpressing jing in the trachea, flies containing the P[breathless (btl)-GAL4] driver were crossed to those containing P[jing-UAS]. Progeny from this cross were stained with 2A12 antibody and tracheal tubule development was analyzed by light microscopy. Overexpression of jing in the trachea is associated with defects in dorsal trunk fusion, as well as improper formation of the transverse connective, dorsal branch and visceral branch (Fig. 6L). Therefore, jing overexpression tracheal phenotypes are similar to jing loss-of-function tracheal phenotypes.

Cell type-specific markers were used to follow CNS midline development in homozygous jing mutant embryos. Midline cells were identified using anti-SIM and the glial-specific marker anti-Slit (Nambu et al., 1990; Rothberg et al., 1990). Expression of sli was assessed in homozygous jing mutant embryos using the lacZ reporter P[1.0 HV, sli-lacZ] (Ma et al., 2000; Wharton and Crews, 1993).

There are reductions in the number of SIM-positive and sli-lacZ expressing midline cells in homozygous jing³ mutants compared with wild-type embryos during stage 9 and 11, respectively (Fig. 7A,B,E,F). This clearly demonstrates that the early differentiation of midline lineages requires jing function. By later stages of embryogenesis (stage 15), SIM and SLI immunoreactivity is drastically reduced in jing mutant nerve cords (Fig. 7D,H) compared with wild-type (Fig. 7C,G). The presence of SLI-positive cellular profiles in macrophages outside the VNC suggests that midline lineages are lost by cell death. Similar results were obtained using anti-Wrapper as a marker of glial identity (data not shown).

To address whether midline glia enter apoptotic pathways, jing mutant embryos were double-labeled with anti-SLI and TUNEL, and the occurrence of apoptotic glia monitored from stages 12 to 15 (Gavrieli et al., 1992). On average, there are one or two apoptotic midline glia within an entire nerve cord of a stage 12 wild-type embryo (Fig. 7I,I′, arrowhead; n=11 embryos). By contrast, every nerve cord segment in jing³ mutant embryos contains apoptotic glia in addition to the presence of more TUNEL-positive profiles in the CNS (Fig. 7J,J′; n=13 embryos). The increased occurrence of apoptotic glia correlates with reductions in SLI immunoreactivity in the midline of jing³ mutant embryos (Fig. 7J, arrowheads) and establishes that jing function is required for midline glial survival.

Enhancer traps and antibodies were used to follow the development of individual motoneurons (VUMs, 22C10) and interneurons, such as the midline precursors (MP1, dMP2, vMP2; P223, anti-ODD and 22C10) and the median neuroblast (MNB; anti-Engrailed) in wild-type and homozygous jing³ mutant embryos (Fujita et al., 1982; Hummel et al., 2000; Schmid et al., 1999; Skeath and Doe, 1998; Sonnenfeld and Jacobs, 1994; Spana et al., 1995). jing loss-of-function mutations are associated with reductions in the expression of all neuronal markers tested. There are absences of immunoreactivity in the VUMs, MNB and MP1 neuronal lineages in some VNC segments in jing³ mutant embryos (Fig. 8B,D,F,H) compared with wild type (Fig. 8A,C,E,G). There is a loss of ODD immunoreactivity as early as stage 10 in MP neurons in homozygous jing³ mutants compared to wild-type (Fig. 8I,J). Similar reductions in the number of immunoreactive vMP2 and dMP2 are observed by 22C10 staining of stage10 homozygous jing³ mutant embryos (Hummel et al., 2000) (data not shown).

Within a particular VNC segment in jing³ mutants, there is a loss of Engrailed (EN)-positive neurons while the number of EN-expressing neuroectodermal cells remains equal to that in wild-type embryos (Fig. 8D, arrow). In addition, jing mutant embryos displaying reduced 22C10 staining of the VUMs in the CNS midline do not show any visible defects in peripheral nervous system development (data not shown). These results strongly suggest that the primary site of jing CNS function is at the midline.

In summary, these results demonstrate that midline neuronal and glial populations do not differentiate without proper jing function and suggest a positive role for jing in promoting CNS midline cell development.

To determine the role of jing during tracheal development, a phenotypic analysis of homozygous jing mutant embryos was performed using antibodies to TRH as a marker of cell identity and to EN for identifying the anterior border of the trachea (Glazer and Shilo, 2001; Isaac and Andrew, 1996; Wilk et al., 1996). Initial defects in tracheal morphogenesis occur during tracheal placode stages in embryos homozygous mutant for all jing alleles (Fig. 9B). This correlates with the nuclear localization of JING in tracheal placode cells (Fig. 4). The number of TRH-positive precursors in stage 10 homozygous jing³mutant embryos is approximately 22% of the expected number of wild-type cells (Fig. 9A,B). The relatively normal pattern of ectodermal segmentation in jing³ mutant embryos, as revealed by EN staining, suggests that the improper differentiation of tracheal cells in these mutants is not likely to result from indirect effects of ectodermal patterning (Fig. 9B,F). These results also reveal that the positioning of tracheal placodes in jing³ mutants is not altered from that of wild-type embryos.

To determine the fate of tracheal lineages we analyzed the pattern of cell death by double labeling wild-type and jing³ mutant stage 11 embryos with TUNEL and anti-TRH (Fig. 9C,D) (Gavrieli et al., 1992). Cell death is not common in the tracheal pits of wild-type stage 11 embryos (Fig. 9C). On average, there is a maximum of three TUNEL- and TRH-positive cells within an entire stage 11 wild-type embryo (n=24). By contrast, there is an average of 20 TUNEL- and TRH-positive precursors in stage 11 jing³ mutant embryos (Fig. 9D; n=34). There is also an increase in the number of apoptotic profiles surrounding the tracheal pits in jing³ compared with wild-type embryos (Fig. 9C,D, arrowhead). Cell death is observed by TUNEL labeling throughout embryogenesis in all tracheal branches in homozygous jing³ mutant embryos, suggesting that the requirement for jing function is not branch specific (data not shown).

In jing³ homozygous mutant embryos, tracheal cells invaginate but the tracheal branches do not migrate properly anteriorly across EN-positive stripes as they do in wild-type embryos (Fig. 9E,F). In addition, fewer TRH-positive cells express EN in homozygous jing³ mutant embryos compared with wild-type at stage 12 (Fig. 9E,F). By stage 15 in jing³ mutant embryos, parts of the dorsal trunk, the dorsal branch and transverse connectives are missing and correlate with a loss of cells by apoptosis (Fig. 9I; data not shown). In addition, the visceral branch does not form in jing³ mutant embryos (Fig. 9I). Therefore, the EGFR-dependent visceral and dorsal trunk branches appear more severely affected than the Dpp-dependent dorsal and ganglionic branches, as well as the transverse connectives in jing³mutant embryos (Fig. 9I). Despite the death of tracheal cells in jing mutant embryos, the overall embryonic pattern of cell death is not significantly altered by the end of embryogenesis from that of wild-type embryos (Fig. 9H,J). Therefore, the tracheal defects in jing mutants are not likely to result from widespread defects in embryonic differentiation.

In this study, we provide evidence that proper functioning of the jing zinc-finger transcription factor is required in the CNS midline and trachea for patterning of CNS axons and tracheal tubules, respectively. The jing expression pattern, embryonic lethality, and loss- and gain-of-function phenotypes are consistent with an embryonic role for this locus. Given the recent characterization of jing during the migration of border cells in Drosophila ovaries, these studies highlight the important role of this gene during cellular differentiation (Lui and Montell, 2001).

Loss-of-function jing alleles result in aberrant expression of all CNS midline and tracheal markers tested. Loss of midline en expression in segments that have no detectable changes in lateral en expression shows that jing mutants have defects specific to the midline. The loss of CNS midline and tracheal cells in homozygous jing mutant embryos is at least partially mediated by cell death. Therefore, during embryogenesis jing is required for terminal differentiation and viability of CNS midline and tracheal cells.

The results presented here show that CNS midline neurons and glia do not differentiate properly in homozygous jing mutant embryos. Several lines of evidence support this. The expression of cell-type-specific markers of midline neuronal and glial identity is altered in jing mutants compared with that in wild-type embryos. For example, expression of the sli-lacZ 1.0 HV reporter initiates in six midline glia in each wild-type nerve cord segment during stage 11 (Wharton and Crews, 1993). By contrast, sli-lacZ 1.0 HV reporter expression in jing mutants initiates in only an average of three midline glia per nerve cord segment by stage 11. In addition, there are reductions in the number of SIM-positive midline cells and ODD-positive/22C10-positive MP neurons by stage 9 in jing³ homozygous mutant embryos, respectively. Therefore, early midline glial and neuronal differentiation is aberrant in homozygous jing mutant embryos. By the end of embryogenesis, many neuronal and glial cell type markers are barely detectable in homozygous jing mutant ventral nerve cords.

The loss of sim, sli, odd and 22C10/futsch expression in jing mutants may reflect improper activation/regulation of gene expression or may be secondary to cell loss. To address this issue, we analyzed the pattern of cell death in the CNS midline of jing mutant embryos. Apoptosis occurs in the midline glial lineage in wild-type embryos and begins during stage 12 to refine the number of cells from six to an average of three per nerve cord segment by the end of embryogenesis (Sonnenfeld and Jacobs, 1995; Zhou et al., 1995). In homozygous jing mutants, however, there are more apoptotic glia during stage 12 than in wild-type embryos and this correlates with the loss of SLI-positive glia. It is, therefore, likely that the loss in CNS midline gene expression in jing mutants results from a loss of cells. In summary, the loss in expression of cell identity markers and inappropriate cell death lead us to conclude that midline neurons and glia do not differentiate properly in jing mutant embryos.

The arthropod ventral nerve cord is characterized by the ladder-like pattern of the major CNS axon tracts. The nerve cord is segmental and each neuromere is connected by longitudinal axons, which are separated by anterior and posterior commissures. Disruption of this pattern by jing gain-of-function specifically in the CNS midline reveals the requirement for proper jing function within these cells for axon patterning. In addition, homozygous mutant jing embryos display reductions in CNS midline cells while neuroectodermal and peripheral nervous system development is unperturbed. Together, these results show that jing mutations have strong effects on the CNS midline and that jing dosage is crucial for their development.

Genetic analysis of axon patterning in the Drosophila CNS has revealed the important role of neuron-glial function in this process (Klämbt et al., 1991; Hummel et al., 1999b). Mutations leading to reductions in midline neuron numbers correlate with a reduction in the number of commissural tracts, whereas mutations leading to reductions in midline glia numbers show fused commissure phenotypes (Hummel et al., 1999a). These observations are consistent with the hypothesis that midline neurons (such as the VUMs) are required to attract commissural growth cones initially to the CNS midline, whereas midline glia are required subsequently for the organization of commissural axons (Hummel et al., 1999b). Based on these observations, we propose that defects in the differentiation of midline neuronal precursors, such as the VUMs, in jing loss-of-function mutants inhibit the attraction of commissural growth cones to the CNS midline during stage 12. As the attraction of commissural axons to the CNS midline precedes the separation of anterior from posterior commissures, the defects in midline neuronal differentiation and the associated lack of growth cones in the midline of jing mutants probably mask subsequent defects in glial-associated functions (Klämbt et al., 1991). During axon patterning, the MP1 interneurons participate in the formation of specific longitudinal pathways (Lin et al., 1995; Hidalgo and Brand, 1997). Therefore, the defects in MP1 neuronal differentiation in jing mutants may account for the inhibition in the formation of the longitudinal connectives.

Signals generated by CNS midline cells control the commissural axon pattern by either guiding growth cones toward the midline or preventing them from crossing the midline (Battye et al., 1999; Harris et al., 1996; Kidd et al., 1999; Tessier-Lavigne and Goodman, 1996). Defects in glial-associated functions occur in the CNS of homozygous jing mutant embryos. Reduced glial numbers and SLI production in jing mutants are consistent with the reduction in midline repulsion of longitudinal pathways as visualized by Fasciclin 2 staining (Fig. 6H). The remaining SLI protein product in stage 12 jing mutant nerve cords, however, is apparently sufficient to prevent a total collapse of the longitudinal connectives, as observed in homozygous sim and sli mutations (Nambu et al., 1990).

This work has identified multiple roles for jing in tracheal morphogenesis. The earliest function of jing is to allocate the correct number of cells to the tracheal placodes. Several lines of evidence support this. The number of tracheal placode cells is significantly reduced from wild-type in homozygous jing mutant embryos. In addition, tracheal precursors die in jing mutant embryos, suggesting that jing is essential for their differentiation. As JING localizes to the nuclei of tracheal placode cells and contains potential DNA-binding and transactivation domains, it is possible that it regulates genes essential for the differentiation and survival of tracheal precursors (Mitchell and Tijian, 1989).

Although loss of jing function affects cellular differentiation in all tracheal lineages, it appears to have more severe effects on dorsal trunk and visceral branch development. The dorsal trunk and visceral branches derive from the same position in the tracheal placode and are induced by Epidermal Growth Factor Receptor (EGFR) (Wappner et al., 1997). EGFR is activated in the central portion of the tracheal placodes by the restricted expression of rhomboid (rho) (Bier et al., 1990; Llimargas and Casanova, 1997; Sturtevant et al., 1996; Wappner et al., 1997). The defects in dorsal trunk and visceral branch formation in homozygous jing mutant embryos are similar to those in embryos homozygous mutant for Egfr signaling (Llimargas and Casanova, 1997; Wappner et al., 1997). Given that mutations in Egfr pathway genes do not affect tracheal placode cell numbers, we propose that jing may function prior to EGFR signaling.

Several lines of evidence suggest that jing functions specifically in tracheal cells. First, we detect JING protein within nuclei of tracheal precursors and differentiated lineages. Second, defective placodes in jing mutants are observed in hemisegments with normal en expression patterns indicating that defects in the metamerization process do not cause the jing tracheal phenotype. However, we cannot rule out the possibility that Hedgehog signaling in segmental ectodermal stripes is affected by jing mutations. A requirement for hh in determining proper tracheal placode numbers in some hemisegments has recently been shown (Glazer and Shilo, 2001). Third, the most severe defects in tracheal patterning in jing mutant embryos occur in the dorsal trunk and visceral branch, suggesting that there is some specificity to jing tracheal function. Last, overexpression of jing specifically in the trachea results in defects in tracheal patterning that resemble jing loss-of-function phenotypes.

Based on genetic and phenotypic analyses, we propose a role for jing downstream of sim and trh during CNS midline and tracheal development, respectively. First, jing expression is not observed prior to that of either sim or trh in the CNS midline and trachea, respectively. jing expression is detected in the CNS midline during stage 9, which is after the initiation of sim expression and establishment of midline fates (Crews, 1998). JING protein is present in tracheal precursor nuclei, coincident with TRH during stage 10. Second, the CNS axon and tracheal phenotypes of homozygous jing mutations are less severe than those of homozygous sim and trh mutations, respectively. However, we cannot rule out that maternal JING may rescue the effects of zygotic jing mutations or that jing functions in a combinatorial fashion and therefore may not display severe phenotypes (Ma et al., 2000). Third, jing can be activated by ectopic expression of sim, suggesting that sim may regulate jing. The presence of three E-box ACGTG core sites in the 5′ regulatory region of jing suggest that this regulation may be direct. Fourth, the sim and trh embryonic phenotypes are epistatic to that of jing, as shown by double mutant analysis. Finally, jing mutations genetically interact with mutations in bHLH-PAS target genes such as sli and btl (Ma et al., 2000; Oshiro and Saigo, 1997). The ventral displacement of midline cells in jing and sli double heterozygotes strongly suggests that jing is required for proper sli regulation.

In Drosophila ovaries, jing is an essential downstream target of the vertebrate homolog of the basic region/leucine zipper transcription factor CCAAT enhancer-binding protein (C/EBP), which is required for border cell migration (Liu and Montell, 2001). The JING protein is most similar to the mouse protein AEBP2, which was identified by its binding to a regulatory sequence in the adipocyte Ap2 gene. Given that C/EBP also binds this sequence, it is proposed that JING and C/EBP coordinate cell differentiation in a co-operative manner (He et al., 1999; Liu and Montell, 2001).

An interesting parallel between the ovarian and embryonic pathways involving jing is the activation of btl. btl is expressed in embryonic tracheal and midline glial cells, as well as border cells of the ovary, and is essential for their migration (Glazer and Shilo, 1991; Klämbt et al., 1992; Murphy et al., 1995). btl is a direct target of C/EBP and TRH::TGO heterodimers in vitro (Murphy et al., 1995; Oshiro and Saigo, 1997). Therefore, the strong dominant interactions between jing and btl in the embryonic trachea coupled with the role of jing in border cell migration implies an important link between the maternal and embryonic pathways involving jing. C/EBP is expressed in the embryonic trachea after btl expression begins (Rorth et al., 1992). Therefore, C/EBP probably does not regulate jing or btl in the trachea, as it does in border cells. Furthermore, C/EBP expression is not sufficient to cause ectopic btl expression and therefore, it is proposed that this transcription factor carries out the gene expression program initiated by other factors (Murphy et al., 1995). In this way, C/EBP can function in very different pathways, including fat metabolism in adipocytes, long-term memory in Aplysia neurons and cell migration in the ovarian border cells (Murphy et al., 1995). The transcriptional capabilities of jing have not yet been tested and therefore it is not known whether jing co-operates with TGO::TRH or TGO::SIM heterodimers in activation of btl or other targets.

We propose that bHLH-PAS heterodimers may activate jing transcription by binding any or all of the three CNS midline elements (CMEs) present in the 5′ regulatory region of jing (Fig. 10). The initiation or maintenance of jing transcription may also require the function of additional transcription factors such as VVL in the CNS midline and trachea or Fish-hook in the CNS midline (Ma et al., 2000; Llimargas and Casanova, 1997; Boube et al., 2000; Zelzer and Shilo, 2000). The presence of a Fish-hook DNA-binding site (TACAAT) adjacent to the CMEs in the jing 5′ regulatory region suggests this possibility (Fig. 4A, Fig. 10).

Genetic and phenotypic evidence suggests that jing functions in a different manner from either vvl or fish. vvl and fish have combinatorial regulatory roles in CNS midline or tracheal pathways, and therefore vvl and fish-null embryos do not show strong phenotypes (Boube et al., 2000; Ma et al., 2000; Zelzer and Shilo, 2000). By contrast, the defects in cell numbers and increased cell death during stage 10 and 11 in jing mutants precede and are different from the defects in homozygous vvl and dfr-fish double mutants in the CNS midline or from dfr homozygotes in the trachea (Boube et al., 2000; Ma et al., 2000; Zelzer and Shilo, 2000). Therefore, activation of jing transcription in the CNS midline or trachea cannot be controlled exclusively by Fish-hook or VVL.

jing encodes a putative DNA-binding protein with putative transcriptional regulatory domains, and its product can be seen in the nuclei of CNS midline and tracheal cells. Based on jing expression patterns and phenotypes, we propose that jing participates in the activation of genes downstream of both SIM::TGO and TRH::TGO in the CNS midline and trachea, respectively. Whether jing targets are also regulated by bHLH-PAS, POU or SOX combinatorial transcriptional activities remains to be determined. Nevertheless, JING function is required to promote the differentiation of CNS midline and tracheal lineages, which, in the absence of jing function, do not differentiate and instead undergo apoptosis.

How does jing promote cellular differentiation? We propose that jing regulates the transcription of important survival factors including those of the EGFR pathway. For example, the rho gene product regulates the processing of the EGF receptor ligand Spitz and is expressed at stage 9/10 in the CNS midline and in the center region of the tracheal placodes (Bier et al., 1990; Schweitzer et al., 1995; Golembo et al., 1996; Wappner et al., 1997). Proper function of EGFR pathway genes is required for the survival of cells most highly affected by jing mutations, including the CNS midline glia, the tracheal dorsal trunk and visceral branches (Scholz et al., 1997; Sonnenfeld and Jacobs, 1994; Wappner et al., 1997). Furthermore, the rho regulatory region is controlled by TGO::TRH:DFR interactions and also contains multiple CAAT sites similar to those bound by AEBP2, the protein most related to JING (He et al., 1999; Lui and Montell, 2001; Zelzer and Shilo, 2000). This raises the possibility that jing may be involved in a combinatorial fashion in the regulation of bHLH-PAS target genes. However, this hypothesis does not account for the role of jing in EGFR-independent process such as survival of CNS midline neurons and dpp-dependent tracheal branches. One can then argue that jing function in the CNS midline and trachea is generic, and that JING carries out the gene expression programs initiated by bHLH-PAS, POU domain and SOX transcription factors. If the transcriptional programs are not maintained by jing, cells enter default apoptotic pathways. Alternatively, JING may be responsible for directly activating unknown survival factors in midline neurons and dpp-dependent tracheal branches.

We thank H. Lipshitz and Angelo Karaiskakis for help with P element-mediated transformation. We are grateful to the Ottawa General Hospital Research Institute for use of their confocal microscope. Thanks to N. Sant, C. Stefanski and C.-A. Chenard for help isolating jing mutations. We thank S. Crews, M. Krasnow and S. Hayashi for reagents and flies. We thank C. S. Goodman, S. Artavanis-Tsakonis and J. Skeath for antibodies. BP102, 4D9 anti-ENGRAILED/INVECTED and mAb 22C10 antibodies were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. This work was supported by a Canadian Institutes for Health Research (CIHR) operating grant to M. J. S.