Localized JAK/STAT signaling is required for oriented cell rearrangement in a tubular epithelium

The generation of complex shapes, both of the whole embryo and its constituent tissues and organs, depends on the coordination of multiple cellular processes; these include cell shape change, proliferation, cell migration and epithelial-mesenchymal transitions. An essential contributor to the form-generating processes of gastrulation and epithelial morphogenesis is oriented cell rearrangement. During gastrulation in frogs and fish, cells intercalate in the mediolateral axis, and converge towards the midline; this process has been termed convergent extension and results in dramatic elongation of the embryonic anteroposterior axis (reviewed byKeller et al., 2000;Wallingford et al., 2002). Although these cell rearrangements during gastrulation occur primarily among mesenchymal cells, cell rearrangement can also occur among cells that are bound together in an epithelium. Elongation of a number of epithelial sheets,such as the Drosophila germband and the C. elegans dorsal epidermis, has been shown to occur by mediolateral intercalation relative to the midline (Irvine and Wieschaus,1994; Heid et al.,2001). Elongation of epithelial tubes such as the sea urchin archenteron, C. elegans intestine, insect renal tubule andDrosophila posterior spiracles and hindgut is also driven by cell rearrangement; in these cases, cell intercalation is oriented circumferentially (perpendicular to the tube proximodistal axis) rather than towards the midline (Fig. 1B)(Ettensohn, 1985;Leung et al., 1999;Skaer, 1993;Brown and Castelli-Gair Hombria,2000; Iwaki et al.,2001) (reviewed by Wallingford et al., 2002).

An unsolved problem of great interest is the source of the polarizing information that causes rearranging cells to intercalate in one axis, and not in another. Although convergent extension in both frogs and fish has been shown to require the planar cell polarity pathway initiated by non-canonical Wnt signaling, the nature of the positional cues that orient the rearranging cells has not been defined (reviewed byWallingford et al., 2002). Similarly, known genes required for oriented cell rearrangement in the C. elegans dorsal epidermis and Drosophila posterior spiracles encode putative transcription factors that are thought to confer morphogenetic capacity, rather than regulate expression of cue-providing molecules(Heid et al., 2001;Brown and Castelli-Gair Hombria,2000). As localized, secreted signals have been demonstrated to guide various types of cell migration and epithelial outgrowth in bothDrosophila and mammals (Duchek et al., 2001) (reviewed byMetzger and Krasnow, 1999;Hogan and Kolodziej, 2002;Moore, 2001;Rollins, 1997), an appealing hypothesis is that rearranging cells orient with respect to spatially localized cell signaling molecules.

The conserved JAK/STAT (Janus Kinase/Signal Transducer and Activator of Transcription) signaling pathway is widely used; it has been shown to play a required role in a variety of processes including hematopoiesis, sex determination, lymphocyte migration, and border cell migration (reviewed byWard et al., 2000;Luo and Dearolf, 2001;Sefton et al., 2000;Moore, 2001;Vila-Coro et al., 1999;Silver and Montell, 2001;Beccari et al., 2002). A suggestion that JAK/STAT signaling might be involved in epithelial cell rearrangement comes from the observation that the ligand for theDrosophila JAK/STAT pathway, Unpaired (Upd; Os — FlyBase) is expressed in a highly localized position at the anterior of the embryonic hindgut epithelium (Iwaki et al.,2001). Mutants that block the elongation of the hindgut, which occurs largely by cell rearrangement, also alter the localized pattern ofupd expression (Iwaki et al.,2001).

Exploiting the genetic simplicity of Drosophila, we present evidence that the JAK/STAT pathway orients cell rearrangement in the hindgut,a simple epithelial tubule. In contrast to the situation in mammals where there are four different JAKs and seven different STATs(Imada and Leonard, 2000), theDrosophila genome encodes only one known ligand (upd), one receptor (the cytokine-like domeless, dome, also known as master of marelle), one JAK (hopscotch, hop) and one STAT(Stat92E, also known as marelle) (reviewed byCastelli-Gair Hombria and Brown,2002). For examining cell movement during tubulogenesis, the hindgut is a particularly useful model as it elongates by cell rearrangement without either cell proliferation or apoptosis(Iwaki et al., 2001) (reviewed by Lengyel and Iwaki, 2002). Analysis of loss-of-function mutants shows that the key components of theDrosophila JAK/STAT pathway are required to achieve a fully elongated hindgut; gain-of-function (overexpression) studies show that uniform high level activation of the pathway is not sufficient, while localized production of ligand is necessary to promote oriented cell rearrangement. Our results support a model in which an anteroposterior gradient of ligand activates STAT activity in a similar gradient, leading to orientation of cell rearrangement. This is the first example of a required role for JAK/STAT signaling in orienting cell rearrangement that drives elongation of an epithelium.

Mutant alleles used (the strongest available) were:upd^os1A (a deletion)(Ferrus et al., 1990);hop^C111(Perrimon and Mahowald, 1986);Stat92E⁰⁶³⁴⁶(Hou et al., 1996);drm³, drm⁶ anddrm^P1 (a deletion)(Green et al., 2002);byn^apro(Murakami et al., 1995);bowl¹ (Wang and Coulter, 1996); lin²(Hatini et al., 2000);wg^7L74(Tearle and Nusslein-Volhard,1987); and twi¹(Nusslein-Volhard et al.,1984). For temperature shift experiments, the temperature-sensitive allele wg^IL114(Nusslein-Volhard et al.,1984) was used. Three GAL4 lines that drive expression uniformly throughout the hindgut were used: bynGAL4 drives strong expression starting at stage 7, while fkhGAL4 and 455.2GAL4 drive progressively weaker expression starting at stages 8 and 9, respectively(San Martin and Bate, 2001;Fuss and Hoch, 2002;Iwaki and Lengyel, 2002) (D. D. I., unpublished). drmGAL4 drives expression only in the small intestine domain of the hindgut (Green et al., 2002). UAS lines used were UAS-upd(Zeidler et al., 1999),UAS-hop^TML(Harrison et al., 1995),UAS-Stat92E (from Steven Hou, unpublished) andUAS-dome^ΔCYT3.2 (deletion of the Dome cytoplasmic domain) (Brown et al.,2001).

Germline clones of hop and Stat92E were made using the FLP-DFS technique (Chou et al.,1993). For hop, larvae of the genotypehop^C111FRT101/ovo^D1FRT101; hsFLP38 were heat shocked at 37°C for 2 hours on each of days 4 and 5 after egg deposition. The eclosed females of this genotype were then mated to FM7, act-lacZ males to obtain progeny, half of which lacked both maternal and zygotic contributions of hop (referred to ashop^m-z-). For Stat92E, larvae of the genotype hsFLP₁₂₂;FRT82B ovo^D1/FRT82BStat92E^P1681 were similarly heat shocked and mated to Stat92E⁰⁶³⁴⁶/TM3 ftz-lacZ males to obtain Stat92E^m-z-progeny.

Whole-mount in situ hybridization was carried out as described(Pignoni and Zipursky, 1997;Tautz and Pfeifle, 1989). Digoxigenin-labeled RNA probes (Roche Molecular Biochemicals) were made from cDNA templates of upd (Harrison et al., 1998), dome(Brown et al., 2001),hop (Binari and Perrimon,1994), Stat92E (Hou et al., 1996; Yan et al.,1996), drm (Green et al., 2002), Ser(Thomas et al., 1991),hh (Lee et al., 1992)and dri (Gregory et al.,1996). Antibody staining of embryos was performed using standard techniques (Ashburner, 1989). Antibodies (and dilutions) used were α-Crb (1:100), α-Wg (1:100)and α-En (1:5), all available from the Developmental Studies Hybridoma Bank at the University of Iowa Department of Biological Sciences;α-Stat92E (Chen et al.,2002) (1:1000); α-phosphorylated histone H3 (Upstate Biotechnology, 1:500); α-Con(Meadows et al., 1994) (1:30);and α-β-Galactosidase (Promega, 1:1000, Cappel, 1:500). For transverse sections, embryos were fixed in either 2% glutaraldehyde plus 1%osmium tetroxide, or in 4% formaldehyde, embedded in Epon and sectioned at 2μm. Sections were stained with 0.025% Methylene Blue and 0.01% Toluidine Blue in 0.025% sodium tetraborate buffer. Light microscopy was carried out using a Zeiss Axiophot microscope, and images were captured with a Sony DKC-5000 digital camera. Confocal microscopy was carried out using a Carl Zeiss LSM 310 with a 40× objective and 1.5× digital zoom; images were acquired and processed using Zeiss LSM software. Hindgut lengths were measured as previously described (Iwaki et al., 2001) using images acquired with a Hamamatsu camera and Axiovision software. Total hindgut cell number in embryos of different mutant genotypes was determined by counting nuclei (identified by anti-β-Galactosidase staining) that were expressing lacZ due to the presence of either byn^apro orbynGAL4,UAS-lacZ.nls, following previously published procedures(Iwaki et al., 2001). The number of cells in the hindgut circumference was determined by counting cells in eight to ten serial transverse sections through the large intestine of three to four different embryos. Embryos were staged according to Campos-Ortega and Hartenstein(Campos-Ortega and Hartenstein,1997).

The Drosophila hindgut epithelium elongates and narrows during a 10-hour period of embryogenesis (Fig. 1A), driven largely by cell rearrangement(Fig. 1B) (reviewed byLengyel and Iwaki, 2002). This cell rearrangement must be oriented by some type of polarity, which probably arises from locally expressed signaling molecules. The developing hindgut is patterned into three morphologically distinct regions along its anteroposterior axis: the small intestine at the most anterior (to which the renal tubules connect), the large intestine and the rectum at the most posterior. Genes encoding at least seven cell signaling molecules are expressed differentially in these three regions(Fig. 1C). The putative transcriptional regulator encoding genes drumstick (drm),bowel (bowl) and lines (lin) control patterned expression of these signaling molecules (in particularupd), and are concomitantly required for cell rearrangement and elongation of the hindgut (Fig. 1C-K). The signal-encoding genes Ser, Dl, hh, dpp andspi, however, are not required for hindgut cell rearrangement(Iwaki and Lengyel, 2002;Fuss and Hoch, 2002;Hoch and Pankratz, 1996;Takashima and Murakami, 2001)(data not shown).

wg is expressed throughout the hindgut primordium starting at stage 5; by late stage 10 it is expressed at the most anterior and most posterior of the hindgut (Fig. 1C). To assess the required role of wg expression at these different stages, we carried out temperature shift experiments with a temperature-sensitive wg mutant (see supplementary figure S1). While early lack of wg function results in a dramatically smaller hindgut, the elimination of wg function during stage 10 (prior to the period of major elongation of the hindgut) or later allows essentially normal hindgut elongation. Thus, localized expression of Ser, hh, spi, dpp,Dl and wg does not appear to play a required role in hindgut cell rearrangement.

upd, encoding the ligand for the Drosophila JAK/STAT pathway, is only expressed in the small intestine(Fig. 1C,H) and is regulated by genes controlling hindgut cell rearrangement. In drm andbowl mutants, expression of upd is missing from the small intestine (Fig. 1I,J), while inlin mutants, upd expression is expanded throughout much of the hindgut (Fig. 1K). These results raise the possibility that localized Upd might provide an orienting cue for rearranging hindgut cells.

If it plays a role in hindgut cell rearrangement, upd must be expressed before and during the period of major hindgut elongation, i.e. between stages 11 and 16 (Fig. 1A); genes encoding the other known components of theDrosophila JAK/STAT signaling pathway, summarized inFig. 2A, should also be expressed at the same stages, both within and adjacent toupd-expressing cells. We used in situ hybridization to characterize the expression of upd, dome, hop and Stat92E during stages just prior to and during hindgut elongation; characterization of Stat92E protein expression is presented in a subsequent section.

Expression of upd in the hindgut is first detected at stage 9 in a narrow ring of cells that will become the small intestine(Fig. 2B). Expression in the prospective small intestine is maintained during stages 10 and 11(Fig. 2C), where it can be seen just posterior to the everting renal tubules (note that in the hindgut at these germband-extended stages, `posterior' is toward the head). During stages 12-14, when the hindgut undergoes a major part of its elongation, updexpression is seen throughout the now distinct small intestine(Fig. 2D). Expression ofupd is maintained throughout the small intestine during the remainder of embryogenesis.

In addition to upd, we examined expression in the hindgut of genes encoding the JAK/STAT receptor (dome), JAK (hop) and STAT(Stat92E). hop is expressed uniformly throughout the embryo,including the hindgut as it elongates (data not shown)(Binari and Perrimon, 1994). Expression of both dome and Stat92E is detected weakly at the anterior of the hindgut beginning at stage 9, becomes significantly stronger by stage 11, and is maintained through stage 14(Fig. 2E-J). For both the receptor- and STAT-encoding genes, expression domains in the hindgut epithelium overlap with and extend beyond the narrow domain of updexpression (Fig. 2, brackets). Most significantly, expression of dome and Stat92E extends to a more posterior position in the hindgut epithelium than does expression ofupd (Fig. 2, compare C,F,I, brackets). Thus, the mRNA expression of the ligand, receptor and STAT components in the hindgut prior to and during its elongation is consistent with a role for JAK/STAT signaling in hindgut cell rearrangement.

To assess the requirement for upd and JAK/STAT signaling in hindgut elongation, we examined hindgut morphology, length and circumference in wholemounts and transverse sections of embryos deficient for upd, dome,hop or Stat92E. In embryos lacking zygotic updfunction, the hindgut reaches only about half its normal length and is somewhat wider (Fig. 3B). Hindguts are also incompletely elongated and wider in embryos lacking both maternal and zygotic activity of either hop or Stat92E(Fig. 3C,D). Embryos lacking zygotic dome function have hindguts only slightly shorter than wild type (data not shown), presumably because of the maternal contribution ofdome (Brown et al.,2001). However, when a dominant negative form of Dome(UAS-dome^ΔCYT3.2) is expressed uniformly usingbynGAL4, hindguts are significantly shorter and wider(Fig. 3I). Length measurements reveal that upd-, hop- and Stat92E-deficient, as well as Dome^DN- expressing hindguts, while not as short and wide as those of drm, bowl and lin embryos (compareFig. 1B-D), are nevertheless 40-50% shorter than those of wild-type (summarized inFig. 3Q). Consistent with their wider appearance in whole-mount embryos, upd, dome, hop andStat92E-deficient hindguts have a greater number of cells in their circumference (19-27) than do wild-type hindguts (12)(Fig. 3F-H,M,Q). Overall, the shorter and wider appearance of the hindgut is roughly similar among embryos lacking the different components of the JAK/STAT pathway.

The shorter hindgut length is not due to a deficiency of cells, nor is the excess number of cells in the circumference due to overproliferation, as the total number of hindgut epithelial cells is 96, 89 and 85% of wild-type forupd, hop and Stat92E mutant embryos, respectively(Fig. 3Q). Further, staining ofupd embryos with anti-phosphorylated histone H3, a marker of mitosis,did not detect more cell division than observed in wild type (data not shown). The number of cells in the wild-type hindgut at stage 11 is ∼50; by stage 16, this number has been reduced dramatically (to ∼12) by cell rearrangement (Iwaki et al.,2001). Because by stage 16 the number of cells in the circumference of upd, hop and Stat92E embryos is reduced only partially (to roughly 20-30), while total cell number is essentially the same as seen in wild-type (Fig. 3Q), it must be concluded that, in the JAK/STAT loss-of-function mutants tested, there is a defect in hindgut cell rearrangement.

If localized JAK/STAT signaling provides an orienting cue for cell rearrangement, then ectopic JAK/STAT signaling throughout the hindgut would be expected to disrupt this process. To test this, we used bynGAL4 to drive various UAS constructs uniformly in the hindgut epithelium. Uniform expression of upd results in hindguts that elongate to only about 65%of the wild-type length and have about 50% more than the normal number of cells in their circumference (Fig. 3J,N,Q). In the testis, eye and hemocytes, ectopic expression ofupd or activated JAK causes increased cell proliferation(Kiger et al., 2001;Tulina and Matunis, 2001;Chen et al., 2002;Luo et al., 1997). However,total cell number in hindguts ectopically expressing upd is only 76%of normal (Fig. 3Q); consistent with this, staining with anti-phosphorylated histone H3 did not detect excess cell proliferation (data not shown). As uniform expression of upd in the hindgut does not result in an increase (but rather a reduction) in hindgut cell number, the excess number of circumferential cells seen in updoverexpressing hindguts must arise from a defect in cell rearrangement.

To ask whether localized activation of other components of the JAK/STAT pathway is required for hindgut elongation, we expressed uniformly an activated form of the Drosophila JAK(UAS-hop^TML). Similar to what was seen for uniform expression of upd, total cell number did not differ significantly from wild type, but the resulting hindguts were shorter and had more circumferential cells (Fig. 3K,O,Q). When both UAS-hop^TML andUAS-Stat92E are driven by bynGAL4(Fig. 3L,P), the hindgut elongation defect is more severe than UAS-hop^TMLalone.

The ectopic expression studies presented here demonstrate that, while components of JAK/STAT signaling are required, activation of the pathway at uniformly high levels throughout the hindgut is not compatible with normal cell rearrangement. Experiments to be presented in a subsequent section further support the idea that spatially restricted JAK/STAT signaling is necessary for hindgut cell rearrangement.

As proper cell rearrangement is correlated with correct hindgut patterning(Iwaki et al., 2001), it could be argued that, rather than affecting the orientation of rearranging cells directly, upd and JAK/STAT signaling control cell rearrangement by affecting patterning. To test this idea, we assessed gene expression characteristic of the different hindgut regions(Fig. 4A) in updmutants and in embryos ectopically expressing upd throughout the hindgut. Examination of all markers tested (except wg, see below)supports the conclusion that all three domains — small intestine, large intestine, and rectum — as well as boundary cell rows and rings are present and correctly patterned in upd mutant hindguts(Fig. 4B;upd^os1A row). Similarly, all three hindgut domains and boundary cells are present when upd is uniformly expressed in the hindgut (Fig. 4B; bynGAL4; UAS-upd row). The only domain missing from hindguts lacking upd is a small, wg-expressing region at the extreme anterior of the small intestine that is established during stages 10/11 (Fig. 4A,B). As our temperature-shift experiments showed that activity of wg is not required for hindgut elongation after stage 10, this defect in wgexpression cannot be the basis for the effect of upd loss-of-function on hindgut elongation.

In summary, all domains of the hindgut, with the exception of a small number of cells at its anteriormost tip, are correctly patterned in hindguts either lacking or uniformly expressing upd. We therefore conclude that, rather than affecting patterning primarily and morphogenesis secondarily, upd and JAK/STAT signaling directly affect the cell rearrangement that elongates the hindgut.

If localized expression of upd at the anterior of the hindgut epithelium is required for hindgut cell rearrangement, then expression ofupd in this domain should rescue the elongation defect in embryos lacking upd. We therefore used drmGAL4 to drive expression of UAS-upd only at the anterior of the hindgut in embryos lackingupd function. The resulting hindguts appeared morphologically normal,with an average length ∼93% that of wild type(Fig. 5C,D,I). Thus, anteriorly localized (in the small intestine) expression of upd in anupd mutant background is sufficient for elongation.

Is restriction of upd expression only to the hindgut anterior necessary for the cell rearrangement that drives elongation? To answer this,we used the uniformly expressed bynGAL4 to drive upd in anupd mutant background. Rather than rescuing (like thedrmGAL4 driver), this resulted in severely defective, short and wide hindguts (Fig. 5A,B). It could be argued that the early and high level of expression driven bybynGAL4 causes a level of activation of JAK/STAT pathway that inhibits cell rearrangement. We therefore performed the same experiment withfkhGAL4 and 455.2GAL4, which (respectively) drive expression uniformly in the hindgut at progressively lower levels and later times(Iwaki and Lengyel, 2002;Fuss et al., 2000;San Martin and Bate, 2001) (D. D. I., unpublished). We observed neither rescue of the upd hindgut phenotype nor a phenotype more defective than that of upd alone (in contrast to the result with bynGAL4; data not shown). Thus,anteriorly localized upd expression appears to be both necessary and sufficient for cell rearrangement in the large intestine.

An important issue is whether upd plays a permissive role in the hindgut (giving cells the ability to rearrange), or an instructive role(orienting cells as they rearrange). As none of three different levels of uniform upd expression rescued the upd loss-of-function phenotype, it seems unlikely that the function of JAK/STAT signaling in the hindgut is simply to promote the capacity of cells to rearrange. Rather, the data support the notion that the required role of localized expression ofupd is to provide an instructive cue that orients cell rearrangement.

Although upd cannot entirely mediate the effect of drm on hindgut cell rearrangement (as drm hindguts are shorter and wider than those of upd; Fig. 5E,F), we asked whether, and to what extent, expression ofUAS-upd under control of drmGAL4 could rescue thedrm loss-of-function phenotype. Strikingly, expression ofupd at the anterior of the drm mutant hindgut is sufficient to bring about significant rescue of the drm hindgut phenotype, as assessed in both wholemounts and transverse sections(Fig. 5G,H). Compared withdrm hindguts, the partially rescued hindguts are 45% longer, and have 35% fewer cells in their circumference(Fig. 5I). The rescue of thedrm hindgut phenotype by anteriorly expressed upd thus demonstrates that upd is a key mediator of drm function in the hindgut.

Given the small number of cells in its circumference by stage 16(Fig. 3E)(Iwaki et al., 2001), the large intestine must undergo significant cell rearrangement as it elongates. Yet upd, which we have shown controls cell rearrangement throughout much of the hindgut, is expressed only in the small intestine, and not in the large intestine. Because action of Upd over a distance has been described in the eye disc (Zeidler et al.,1999), it seemed possible that Upd produced in the small intestine might control cell rearrangement by activating JAK/STAT signaling in the large intestine (Fig. 1C,H). During stages 11 and 12, as cell rearrangement is elongating the hindgut, the prospective large intestine is 50 to 60 μm in length(Fig. 6A,B); this is the distance over which Upd would have to diffuse to affect the entire large intestine.

In addition to activating phosphorylation of Stat92E(Luo and Dearolf, 2001),expression of upd has also been shown to upregulate Stat92E protein levels during embryogenesis (Chen et al.,2002). We therefore investigated expression of Stat92EmRNA and protein as reporters for receipt of the Upd signal. Antibody staining shows that, in the stage 11 hindgut, Stat92E protein is greatly reduced in the absence of upd function, and dramatically upregulated whenupd is uniformly expressed (Fig. 6C,D). This regulation appears to occur at the transcriptional level, as Stat92E mRNA is similarly reduced in the absence ofupd function, and upregulated when upd is uniformly expressed in the hindgut (Fig. 6E,F). We conclude that staining with anti-Stat92E identifies cells receiving Upd signal.

During stages 11 to 12, Stat92E protein is detected in a domain that extends more posteriorly than the upd expression domain at the hindgut anterior (compare Fig. 2C and Fig. 6G,I);this can be seen most clearly when both upd mRNA and Stat92E protein are labeled (Fig. 6K). Sagittal sections reveal that Stat92E is present in the nuclei of the hindgut epithelium at stages 11 and 12, (Fig. 6H,J); therefore, during these stages, at least some of the large intestine epithelial cells are receiving Upd signal. Significantly, Stat92E appears to be present in the hindgut as a gradient, with the most anterior cells (at the point of renal tubule evagination) most strongly labeled; during stages 11 to 12, this gradient extends to at least 20% to 45% of the length of the large intestine primordium (Fig. 6G-J). Because of limits in dynamic range of the anti-Stat92E assay, we cannot presently determine whether, at some point, this gradient extends over the entire large intestine.

Stat92E is present not only in the nuclei of the hindgut epithelium, but also in the nuclei of the hindgut visceral mesoderm(Fig. 6J; see supplementary figure S2). However, as determined by staining with anti-Connectin, the hindgut visceral mesoderm is normal in upd mutant embryos (see supplementary figure S2); furthermore, the hindguts of twi mutants, which lack all mesoderm, are able to undergo a significant amount of elongation (see supplementary figure S2) (San Martin and Bate, 2001). Thus, even though cells of the visceral mesoderm receive the Upd signal, they do not appear to play a significant role in the rearrangement of cells within the hindgut epithelium.

The expression of Stat92E in the hindgut epithelium, while suggestive, does not reveal whether receipt of Upd signal by prospective large intestine cells is required for their rearrangement. Because in an upd mutant embryo expression of upd at the anterior of the hindgut rescues elongation(Fig. 5C), we asked whether expression of dominant active Hop (hop^TML) might also rescue. As a control, we expressed hop^TML uniformly in the hindgut; similar to what we observed for upd, this results in a dramatic upregulation of Stat92E (Fig. 6, compare D with L). A key finding is that, in contrast to the essentially complete rescue of the upd hindgut phenotype whenupd is expressed anteriorly (Fig. 6M), anterior expression of hop^TML in anupd background does not rescue hindgut elongation(Fig. 6N). These results demonstrate that activation of the JAK/STAT pathway only in cells of the prospective small intestine (i.e. the domain that expresses upd) is not sufficient for normal hindgut cell rearrangement. Thus, upd is required in a non-cell autonomous fashion for hindgut cell rearrangement.

The observed upregulation of Stat92E in cells of the hindgut posterior to the small intestine is consistent with the cell non-autonomous function ofupd in the hindgut. The fact that cell rearrangement is severely abnormal when uniform levels of Stat92E and Hop^TML are driven together (Fig. 3L,P) further supports the idea that Stat92E must be distributed non-uniformly, i.e. as a gradient, in order for cells to rearrange. The observed gradient of Stat92E,reaching to at least 20% of the length of the large intestine during stage 11,and to at least 45% during stage 12 (Fig. 6H,J), suggests that there is a corresponding gradient of the activating ligand Upd in the hindgut. Both the distance (roughly 20 to 40μm, see Fig. 6A,B) and time(stages 9 through 12, roughly 4-6 hours) over which this postulated Upd gradient is established are consistent with parameters described for formation of gradients of Upd, Dpp and Wg in imaginal disc tissues(Zeidler et al., 1999;Entchev et al., 2000;Teleman and Cohen, 2000;Strigini and Cohen, 2000), as described below.

We have shown that elongation of the Drosophila hindgut by cell rearrangement requires the Upd ligand and the JAK/STAT pathway components Dome(receptor), Hop (JAK) and Stat92E. As elongation does not occur when expression of ligand or activation of the pathway is uniform, but only when the source of ligand is localized to the hindgut anterior, the requirement for localized JAK/STAT signaling in hindgut elongation can be characterized as instructive, rather than permissive. As patterning is normal in hindguts both lacking and uniformly expressing upd, the required role of JAK/STAT signaling in hindgut morphogenesis is likely via direct effects on cell movement.

The rescue of the upd phenotype by anteriorly localized expression in the hindgut of upd, but not of activated JAK (Hopscotch),demonstrates that there is a requirement for upd function that is not cell autonomous. In other words, upd is required in cells (those of the large intestine that undergo the greatest rearrangement) different from cells that produce it (those of the small intestine). A number of examples have been described in which localized expression of a signaling molecule(including Upd) is required non-autonomously for cell rearrangement,morphogenesis or motility. In the Drosophila eye imaginal disc,expression of Upd at the midline is required to establish a dorsoventral polarity that orients ommatidial rotation(Zeidler et al., 1999). In both Drosophila tracheae and the vertebrate lung, branching morphogenesis of the epithelium depends on localized expression of FGF in adjacent mesenchyme (reviewed by Metzger and Krasnow, 1999; Hogan and Kolodziej, 2002).

Localized activation of JAK/STAT signaling has been shown to play a role in cell motility in a number of contexts. In Drosophila, localized expression of Upd in the anterior polar cells of the egg chamber acts to coordinate the migration of the adjacent border cells(Silver and Montell, 2001;Beccari et al., 2002). In mammals, cytokines expressed in target tissues act to attract both migrating lymphocytes and tumor (reviewed by Moore,2001; Muller et al.,2001; Murphy,2001; Vila-Coro et al.,1999). Our finding that localized (only in the small intestine)expression of upd is both necessary and sufficient for rearrangement of cells in the large intestine indicates that Upd must have an organizational, action-at-a-distance function in controlling cell rearrangement during tubular morphogenesis.

The rescue experiments discussed above establish that there is a cell non-autonomous requirement for upd in hindgut elongation. Consistent with this, there is evidence that Upd is present and required in an anteroposterior gradient in the hindgut. Prior to and during hindgut elongation, both Stat92E mRNA and Stat92E protein are detected not only in the small intestine epithelium (and the visceral mesoderm surrounding the small intestine), but also in the epithelium posterior to the small intestine; this expression of Stat92E appears to be in a gradient. In theDrosophila eye imaginal disc, a gradient of Upd is required to orient the rotation of ommatidial cell clusters(Zeidler et al., 1999); in addition, there is evidence for a gradient of Upd and Stat92E in patterning of the follicular epithelium of the Drosophila egg chamber (R. Xi, J. R. McGregor and D. A. Harrison, unpublished). As expression of Stat92E depends onupd (this work) (Chen et al.,2002), it is likely that Upd protein is present in the hindgut epithelium as an anteroposterior gradient, with its highest level in theupd-expressing cells of the small intestine, and lowest level in posterior, upd non-expressing cells of the large intestine. Expression of SOCS36E (suppressor of cytokine signaling at 36E), which is regulated by upd, overlaps with and extends significantly beyond the domain of upd expression(Karsten et al., 2002),further supporting the idea that there is a gradient of Upd in the hindgut. A model that summarizes the observed localized expression of upd mRNA,a gradient of Stat92E protein, and cell rearrangement leading to elongation is shown in Fig. 7.

In the Drosophila eye imaginal disc, anti-Upd staining and the behavior of clones of mutant cells that have lost components of the JAK/STAT pathway indicate that Upd is present in a gradient that extends at least 50μm beyond its midline mRNA expression domain(Zeidler et al., 1999). In theDrosophila hindgut, we have shown that Stat92E is a reliable reporter for the presence of Upd. Two to four hours after upd is first expressed at the anterior of the hindgut (stage 9), Stat92E can be detected at least 30-40 μm from the site of upd expression (stages 11 and 12). These time and distance parameters are similar to those observed during generation of the Upd gradient in the eye, and the Dpp and Wg gradients in wing imaginal discs, which form over distances of roughly 40-80 μm in 1-8 hours (Zeidler et al., 1999;Entchev et al., 2000;Teleman and Cohen, 2000;Strigini and Cohen, 2000). Thus, it is reasonable to imagine that a gradient of Upd is established in the developing hindgut in a short enough time frame to affect cell rearrangement.

The essential consequence of JAK/STAT signaling is activation of the STAT protein, which leads to altered transcriptional programs(Darnell, 1997;Horvath, 2000). STAT has been shown in a number of contexts to be required for cell motility(Sano et al., 1999;Yamashita et al., 2002;Silver and Montell, 2001;Beccari et al., 2002), and therefore probably regulates expression of genes controlling cytoskeletal assembly and cell adhesion. In these contexts, however, activation of STAT does not appear to be required to orient cell movement, but rather to facilitate or promote it. As Stat92E is required for hindgut elongation, and its protein product appears to be present in a gradient along the anteroposterior axis, this raises the intriguing question of how a gradient of a transcription factor might orient cell rearrangement.

Our results demonstrate a new role for Upd and the JAK/STAT pathway, namely the control of cell rearrangement that elongates a tubular epithelium. Given the widespread occurrence of JAK/STAT signaling, further analysis of the mechanism by which JAK/STAT signaling controls hindgut cell rearrangement inDrosophila is likely to provide insights into the control of cell motility during many processes, including organogenesis, wound healing and cancer metastasis.

We thank D. Harrison for sharing information prior to publication, J. Axelrod, U. Banerjee, D. Harrison, K. Lyons and members of the Lengyel laboratory for helpful comments on the manuscript, and S. Yu for excellent technical assistance. K. A. J. and D. D. I. were supported by USPHS National Research Service Awards GM07185 and GM07104, respectively. This work was supported by NIH grant HD09948 to J. A. L.