Morphogenesis Review

Cell, Vol. 96, 225–233, January 22, 1999, Copyright 1999 by Cell Press Morphogenesis Review

Brigid L. M. Hogan which organs as diverse as limbs, branchial arches, gen- Howard Hughes Medical Institute italia, and feathers arise from simple buds. By far the and Department of Cell Biology best understood of these systems is limb formation, Vanderbilt University Medical Center which has been the topic of recent extensive reviews Nashville, Tennessee 37232-2175 (Irvine and Vogt, 1997; Johnson and Tabin, 1997; Martin, 1998). Important new advances have also been made in our understanding of how feather buds develop in The exquisitely beautiful form of a developing embryo periodic patterns in the ectoderm of the chick embryo is the result of a coordination between the driving forces (Crowe et al., 1998; Jung et al., 1998; Noramly and Mor- of morphogenesis and the processes of cell growth, gan, 1998; Viallet et al., 1998). The insights that have proliferation, differentiation, and death. Morphogenesis come from studies on limbs and feathers, as well as is responsible for bringing cell populations together for simple organs in Drosophila such the tracheal system new inductive interactions and for building complex, (Krasnow, 1997) and the dorsal respiratory appendages three-dimensional structures such as hearts, limbs, of the egg (Wasserman and Freeman, 1998), are proving lungs, and eyes out of simple epithelial sheets and mes- very useful for understanding a more complex variation enchymal cell masses. Research over the past two de- of the budding theme, typically refered to as “branching cades has elucidated many of the genetic pathways morphogenesis.” This process, in which a tree-like or- underlying cell division, cell fate determination, and dif- gan is generated by a reiterated combination of bud ferentiation, and has shown them to be evolutionarily outgrowth, elongation of a stem, and subdivision of ter- conserved. A major challenge now is to explore the minal buds, underlies the development of organs such possibility that there is also a conserved “morphoge- as the mammalian pancreas, mammary gland, lung, netic code”—a set of rules common to processes that and kidney. In spite of the physiological importance of are used repeatedly in different combinations to make these organs, we are only just beginning to understand functional organs. These instructions fall into two cate- how their early morphogenesis is controlled. Following gories. First, there are basic subroutines that define es- some general observations, this review will focus on sentially mechanical operations such as the packaging one of these organs—the mammalian lung—as a model of cells into segments, the folding of epithelial sheets system. into tubes or cups, and the outgrowth of buds. Each of In vertebrates, budding invariably involves dynamic these modules utilizes sets of genes controlling proper- and reciprocal interactions between two cell popula- ties such as differential cell adhesion, cell motility, cell- tions—one mesenchymal and the other epithelial. For matrix interactions, and cytoskeletal organization. the purposes of this review, a primary bud is defined as The second category determines how these subrou- a knob-like cluster of progenitor cells located within tines are coordinated with cell proliferation and cell fate distinct boundaries that proliferate and move away from determination. This “project management” depends on signaling centers that arise in the organ primordia or the surface of a preexisting structure. A secondary bud progenitor fields at positions initially determined by the develops on a stalk derived from another bud, while primary embryonic axes. Each center is a group of cells branches arise by bifurcation of a terminal bud. Other that regulates the behavior of surrounding cells by pro- distinct mechanisms for subdividing a bud also exist. ducing positive and negative intercellular signaling mol- For example, a terminal, ampulla-like bud may be ecules. Evidence is beginning to accumulate that the cleaved into multiple lobules by the process of clefting, majority of these signaling factors are proteins encoded involving the ingrowth of mesenchyme and the deposi- by a relatively small number of conserved multigene tion of extracellular matrix. In some organs, such as the families, in particular the Fgfs, Bmps, Hedgehogs, Wnts, lung, budding, branching, and clefting each occur at and Egfs. The diverse biological activities of individual different stages of development (see below), while in ligands are regulated by antagonists, activators, or post- organs such as the salivary gland, for example, clefting translational modifiers that control, for example, the appears to predominate (Nakanishi and Ishii, 1989, for range over which the protein can function or its half- review). life in the environment. In addition, the signaling genes Simple bud formation involves several stages. The themselves are often transcriptionally regulated by posi- first is initiation, in which the boundaries of a bud primor- tive and/or negative feedback loops. An exciting pros- dium, and the signaling centers within it, are gradually pect for the future is the possibility of being able to established at a specific site on the anterior–posterior compute how each of these variables, in combination (A–P), dorsal–ventral (D–V), and medial–lateral (M–L) with the downstream subroutines they control, ulti- axes of the embryo. The second phase, proximal–distal mately affects the size, shape, and pattern of a particular (P–D) outgrowth, is usually, but not always, associated organ. with increased cell proliferation, mediated by at least one source of mitogen such as Sonic hedgehog (Shh) Budding Morphogenesis or Fgfs. During this phase, the extending bud retains One of the recurrent themes of vertebrate embryogene- its positional coordinates, which are important for later sis is the process of “budding morphogenesis,” by patterning events. Finally, there is cessation of outgrowth, due in part to the programmed death, displace- E-mail: [email protected]. ment, or inhibition of distal mitogen-secreting centers. Cell 226

Figure 2. Model for the Role of Fgfs and Shh in Chick Wing Bud Figure 1. Model of Feather Bud Formation and Patterning in Chick Outgrowth Skin Fgf10 (blue) is first expressed in the segmental plate (SP), intermedi- (A) Placode formation is initiated by a mesenchymal signal (gray) ate mesoderm (IM), and lateral plate mesoderm (LPM) over a broader that induces Shh and Fgf4 expression in the overlying ectoderm domain than where bud outgrowth will occur. By stage 14, Fgf10 (blue). Later, placodes also express Bmp2 and Bmp4 (red), leading expression is restricted to the forelimb progenitor field, possibly in to inhibition of placodal fate in the surrounding cells. part by signals from more medial mesoderm (thin arrows). At stage (B) Hypothetical distribution of signaling proteins based on a reac- 16, Fgf10 in the mesoderm directly or indirectly induces Fgf 8 expres- tion diffusion mechanism model of placode development. Both sion in the surface ectoderm (SE) of the presumptive apical ectoder- short-range activators (Fgf, Shh) and longer-range inhibitors (Bmps) mal ridge (AER). By stage 17, Fgf8 made in the AER acts on the are produced from the same source. Nearer the center, activators underlying mesoderm and maintains Fgf10 expression. Shh ex- override the inhibitors, and a feather primordium is established and pressed in the zone of polarizing activity (ZPA) functions directly or outgrowth initiated. Further from the center, inhibitors predominate indirectly (Yang et al., 1997) as a mitogen on the mesenchyme cells and an interbud domain is set up. Delta-1 in the mesenchyme is and also induces Fgf4 in the adjacent ectoderm and Bmp2 in the thought to both promote placode formation in the ectoderm via adjacent mesoderm (not shown). The posterior AER (by secreting Notch-1 and to inhibit placode fate laterally, via Notch-2 in the Fgf4) and the dorsal ectoderm (by secreting Wnt7a) maintain the dermis (Crowe et al., 1998; Viallet et al., 1998). activity of the ZPA. Figure adapted from Ohuchi et al. (1997). This reference should be consulted for more details. Expression of puta- tive inhibitory molecules, for example Bmps and vertebrate homo- How cessation of outgrowth is regulated and how final logs of Drosophila Sprouty, are not shown. Recent studies have organ size is always proportional to the overall size of established that murine sprouty 4 is expressed in the mesenchyme of the progress zone (de Maximy et al., 1999). the embryo are important topics that will not be discussed here. involves intercellular signaling via the Notch-delta pathway. Thus, delta-1 in the mesenchyme of the feather Spatiotemporal Specification of Bud Initiation primordium both promotes placode fate in the overlying Initiation of a primary bud appears to be a dynamic ectoderm and inhibits placode fate laterally (Jung et al., process in which initially broad domains of gene ex- 1998; Viallet et al., 1998). It remains to be seen how Fgfs, pression become gradually restricted. Such dynamic Bmps, and Wnts act in combination with Notch–Notch changes have been particularly well studied recently in ligand signaling to regulate the commitment of progeni- feather development (Crowe et al., 1998; Jung et al., tor cells to specific fates and lineages within the bud. 1998; Noramly and Morgan, 1998; Viallet et al., 1998) The limb bud also appears to develop gradually. (Figure 1). The first sign of the midline feather tract in Again, this has been well studied in the chick embryo, the chick embryo is a single ectodermal band of low- where bud initiation involves the establishment of a re- level Fgf4 and Shh expression running along the A–P stricted region of Fgf10 transcription in the lateral plate axis. This band gradually resolves into punctate groups mesoderm from an initially much broader expression of cells expressing first higher levels of these two genes, domain (see Figure 2) (Ohuchi et al., 1997; Martin, 1998). and then in addition, Bmp2 and Bmp4 in both the ecto- The importance of Fgf10 in limb outgrowth has recently derm and underlying mesoderm. Each localized expres- been confirmed by the fact that homozygous null mutant sion domain gives rise to a feather bud, separated from mouse embryos lack both fore- and hindlimbs, although its neighbor by an interbud zone. Based on the activity a scapula is present at the correct forelimb position (Min of ectopically applied proteins, a model has been pro- et al., 1998; Sekine et al., 1999). The precise mechanisms posed in which Shh and Fgf4 promote the proliferation limiting Fgf10 transcription just to the limb primordium and distal outgrowth of the buds by functioning as short- are currently unknown but may involve positive signals range activators, remaining closely associated with pro- from the intermediate mesoderm (Martin, 1998) and, ducer cells and the surrounding matrix. By contrast, the more speculatively, inhibitors in flanking regions. In any Bmps, produced from the same central source, act as case, bud localization on the A–P axis is downstream long-range inhibitors, tightening up the bud boundaries of Hox genes, since the position of the mouse forelimb and suppressing the initiation of new buds in the vicinity. is shifted anteriorly in Hoxb5 null mutants (Rancourt et Refinement of placode formation and patterning also al., 1995). Review 227

Bud Outgrowth and Shape Have One is the elaboration of physical constraints such as Multiple Determinants an inflexible ectodermal casing that forces cells to trans- Bud outgrowth in vertebrates is invariably associated locate distally like toothpaste through a tube, or con- with increased cell proliferation mediated by distal stricting collagen bundles that promote clefting of large sources of mitogens such as Fgfs and Shh. However, buds (Nakanishi and Ishii, 1989). Another process is several strategies may have to be employed to ensure cell movement, including convergence, extension, and that the enlarging bud acquires a specific shape and compaction. As will be discussed below, there is evi- does not just blow up like a balloon. The most important dence that lung epithelial cells translocate toward a dis- of these may be the one postulated above for feather tal source of chemoattractant, Fgf10 (Park et al., 1998; buds (Figure 1) and the tracheal system and respiratory see Figure 4). Some directed movement of proliferating appendages of the egg in Drosophila (Hacohen et al., mesenchyme may also occur in the limb bud in response 1998; Wasserman and Freeman, 1998), namely the es- to the AER (Vargesson et al., 1997) and in ureteric bud tablishment of a balance between self-enhancing mech- epithelium in response to Gdnf (Pepicelli et al., 1997; anisms that activate proliferation or extension distally Srinivas et al., 1999). and counteracting mechanisms that specifically inhibit Cell movement may also be important for determining these activities more proximally. Such a balance be- the position and physical dimensions of signaling cen- tween positive and negative pathways appears to be a ters. For example, the AER of the chick and mouse limb feature of many developmental processes in general, bud does not just differentiate in situ along the distal and of reaction-diffusion patterning mechanisms in par- rim. Instead, presumptive AER cells, probably induced ticular (Meinhardt, 1996; Wolpert et al., 1998). In bud by signals from the underlying mesoderm (Michaud et morphogenesis, the strategy will help to ensure that al., 1997), assemble there from a much wider domain, once outgrowth is underway, proliferation is higher dis- either as a result of cell sorting and convergent move- tally than proximally, the situation seen in vivo in chick ments (Altabef et al., 1997) or by the compaction or limb bud mesenchyme (Hornbruch and Wolpert, 1970; “zipping up” of an initially loose but coherent domain Cooke and Summerbell, 1980) and in lung endoderm in of presumptive AER cells (Bell et al., 1998; Loomis et culture (Mollard and Dziadek, 1998; Nogawa et al., 1998). al., 1998). Experiments in the chick suggest that the It is likely that the physical dimensions and location of process of assembling a distal AER with sharp bound- the mitogen-signaling centers have a profound influence aries, which cells do not cross, is regulated by signaling on the final shape of a bud-derived structure. For exam- via the Notch–Notch ligand-signaling pathway modu- ple, in the chick and mouse limb the apical ectodermal lated by Fringe (Irvine and Vogt, 1997; Johnston et al., ridge (AER), which promotes proliferation of the mesen- 1997; Zeller and Duboule, 1997). However, a full under- chyme by secreting Fgfs (Figure 2 and Martin, 1998), is standing of the role of the multiple Notch–Notch ligand a thin, tightly packed band of columnar epithelial cells family members at different stages of limb bud morpho- that is located distally and demarcated from the rest of genesis must await further studies. the ectoderm by sharp boundaries. The zone of polariz- In conclusion, differential cell proliferation and cell ing activity, or ZPA, that regulates A–P growth by secret- movement, coordinated by positive and negative recip- ing the mitogen Shh is located asymmetrically in the rocal feedback loops between signaling centers in the mesenchyme, near the posterior end of the AER. This mesenchyme and epithelium, appear to be common fea- arrangement may help to generate the flat and paddle- tures of bud outgrowth. To what extent are these fea- like shape of the limb. By contrast, in the second tures present in the developing lung? branchial arch of the chick embryo, Shh and Fgf8 are both coexpressed in an epithelial band along the entire The Dual Origin of the Vertebrate posterior margin, an arrangement conducive to generat- Respiratory System ing a flap-like structure (operculum) extending posteri- The two parts of the respiratory system of the mouse— orly over the branchial arches (Helms et al., 1997). the trachea and the lung—develop from the foregut at a The balance between positive and negative regulators time (E9.5) when it is a single tube of epithelial endoderm of proliferation within a bud may also have important surrounded by a simple layer of splanchnic mesoderm consequences for the overall patterning of the final and a thin outer covering of mesothelium (Figure 3). organ. In the limb bud, the so-called “progress-zone Each part develops by a completely different strategy model” predicts that the longer undifferentiated mesen- (for usefuldatabase see http://www.ana.ed.ac.uk/anatomy/ chyme cells remain under the influence of the AER (ex- database/lungbase/lunghome.html). The lung, which posed, for example, to high concentrations of Fgfs and consists of the two main bronchi and the respiratory Bmps), the greater the probability that they will differen- tree, is generated by branching morphogenesis of two tiate into distal cell types when they move out of the primary buds that arise on the ventrolateral wall of the zone (Wolpert et al., 1998). This model predicts that if foregut separated by the most ventral foregut. By con- mesenchyme cells in the progress zone proliferate trast, the unbranched trachea is formed by the division slowly, they will be exposed to the AER for longer, and of the foregut by a longitudinal septum into two tubes— the final limb, although shorter, will be distalized (Tabin, one dorsal (the esophagus) and one ventral (the trachea). 1998). Whether progress-zone models can be applied This D–V separation is absent or incomplete in mouse to other organs developing from buds is an intriguing embryos homozygous for null mutations in Shh, which possibility. is expressed in the ventral endoderm (Litingtung et al., Two other processes besides cell proliferation may 1998; Pepicelli et al., 1998). A similar or even more dra- play a role in determining bud outgrowth and shape. matic phenotype is seen in embryos with mutations in Cell 228

on liver (Zaret, 1998) and dorsal pancreas bud formation strongly implicate signals from the surrounding mesenchyme. For example, it has recently been shown that the appearance of dorsal pancreatic buds in the posterior foregut is regulated by factors produced by the overlying notochord that inhibit hedgehog activity in the adjacent endoderm (Kim and Melton, 1998). One candidate for a lung bud outgrowth-promoting factor is Fgf10, which is expressed in the splanchnic mesoderm overlying the lung buds at E9.5 (Bellusci et al., 1997b). In support of this hypothesis, Fgf10 null mutant embryos have no lungs buds at all, although the presumptive trachea apparently develops normally (Min et al., 1998; Sekine et al., 1999). In view of the finding mentioned earlier, that limb buds are shifted anteriorly in Hoxb5 mutants (Rancourt et al., 1995), it is interesting to note that lung buds arise close to the anterior boundary of Hoxb5 expression in the foregut mesenchyme (Bogue et al., 1996). Perhaps the morphogenesis of the lung and limbs in the ancestors of terrestrial vertebrates was coordinated by changes in the regulation of Fgf10 gene expression in the lateral mesoderm? Much more information is needed about the spatial expression of Hox genes along the proximal–distal axis of the lung.

Early Lung Morphogenesis Involves a Stereotypic Figure 3. Branching Morphogenesis of the Mouse Lung Pattern of Budding and Branching (A and B) Morphology of the left (L) and right (R) primary lung buds A striking feature of early lung development is the repro- of an E10.0 mouse embryo by scanning electron microscopy. (A) ducibility of the budding and branching pattern, which Ventral view with heart removed. (B) Cross section. In the L bud only the mesenchyme (m) is visible, but in the R bud both the mesen- is retained even in culture. The left primary bud grows chyme and the distal endoderm (en) surrounded by a basal lamina out as the future left bronchus and sprouts several sec- are revealed. ant, anterior; mes, mesothelium; eos, epithelium of ondary buds along its lateral side in a precise spatiotem- esophagus; tr, trachea. Photographs kindly supplied by Kathy Sulik. poral sequence (Figure 3C). The right bronchus also (C) Early pattern of secondary buds revealed by whole-mount in generates an asymmetric series of secondary buds, but situ hybriziation of E11.5 lung with a digoxygenin-labeled antisense the pattern is quite different. Most noticeably, the first riboprobe for Shh. Buds are numbered in the approximate temporal order in which they appear, although 2 and 3 arise almost simul- bud to arise projects dorsally, followed very closely in taneously. Scale bars ϭ 50 ␮m for (A), 100 ␮m for (B), and 250 ␮m time and space by a lateral and a ventral bud, and for (C). then by several lateral buds. The significance of this branching pattern is not trivial. The first three secondary buds on the right eventually give rise to separate lung Gli2 and Gli3 genes, which are expressed in the mesen- lobes, so that the right lung of the adult mouse is com- chyme and encode transcription factors downstream of posed of four distinct lobes, compared with only one Shh (Motoyama et al., 1998). Interestingly, Gli2Ϫ/Ϫ; on the left. This condition probably allows efficient pack- Gli3Ϫ/Ϫ double-mutant embryos show complete ab- ing of the lung into the thoracic cavity, along with the sence of any differentiation of the foregut endoderm heart and esophagus. into trachea, esophagus, or lungs, a more dramatic phe- The stereotypic budding pattern also reveals that lung notype than seen in ShhϪ/Ϫ embryos. These results sug- buds, like limb buds, have D–V and medial–lateral (M–L) gest that the D–V patterning of the foregut involves epi- axes, as well as L–R identity. As discussed in the review thelial-mesenchymal interactions regulated by Shh and by Beddington and Robertson (1998 [this issue of Cell]), Gli genes. Failure of the separation between esophagus L–R identity is likely to be downstream of Pitx2, which and trachea, as well as agenesis of the left lung and is expressed in the mesoderm of the left but not right hypoplasia of the right, is seen in embryos doubly homo- bud (Meno et al., 1998). What controls the M–L axis is zygous for mutations in RAR␣ and RAR␤2 (Mendelsohn not yet known. However, as discussed earlier, there is et al., 1994). Whether this phenotype is related to evidence that D–V patterning of the foregut and lungs changes in Hox and/or Shh gene expression in the fore- involves Gli gene expression in the mesenchyme (Mo- gut remains to be seen. toyama et al., 1998). In embryos homozygous for null mutations in Gli2, the right lung is not separated into Genetic Control of Bud Initiation distinct lobes. This can be traced back to the absence The mechanisms controlling the site of lung bud initia- of the ventral secondary bud and to the more lateral tion are not yet known. This is true also of the other position of the first dorsal bud, so that it lies close to organs that arise from buds along the A–P axis of the the second bud. Gli3Ϫ/Ϫ embryos show subtle morpho- foregut, namely the thyroid, liver, and pancreas. Studies metric abnormalities in individual lung lobes (Grindley Review 229

et al., 1997), but strikingly, Gli2Ϫ/Ϫ; Gli3ϩ/Ϫ embryos have a single lung fused across the midline, probably resulting from the presence of an ectopic lung bud arising in the ventral midline between the two primary buds (Moto- yama et al., 1998). By contrast, Gli2Ϫ/Ϫ;Gli3Ϫ/Ϫ embryos have no lung buds or trachea at all. Interpretation of these important findings requires further analysis, but they support the idea that the initiation of specific lung buds and their site of outgrowth is genetically controlled and dependent on epithelial and mesenchymal interactions.

Epithelial-Mesenchymal Interactions and Lung Morphogenesis Signaling between the mesenchyme and epithelium Figure 4. Fgf10 Acts as a Chemoattractant for Distal Lung Endo- continues to play a role in lung development after initial derm in Culture outgrowth of the primary buds. Evidence for this comes Mesenchyme-free distal endoderm from an E11.5 lung was incu- from several sources, including classical tissue trans- bated in Matrigel as described (Bellusci et al., 1997b). A 150 ␮m plantation experiments, identification of signaling genes diameter heparin bead–soaked in 50 ␮g/ml Fgf10 (kindly supplied expressed at high levels in the distal endoderm and by Dr N. Itoh) was placed 150 ␮m from the endoderm, which prolifer- mesoderm, and in vitro studies with purified proteins. ated and physically moved around the bead. Asterisks mark initial Transplantation studies first showed that mesen- endoderm positions (M. Weaver, personal communication). chyme from around distal lung buds will induce ectopic buds when grafted next to tracheal endoderm, at least division toward these external sources of Bnl protein up to E14.0. Moreover, the ectopic buds branch in a (Klambt et al., 1992; Lee et al., 1996). The number and lung-specific pattern and express genes characteristic position of the domains of Bnl production is invariant of distal endoderm (Shannon, 1994). By contrast, tra- and appears to be genetically “hardwired” under the cheal mesenchyme inhibits the outgrowth and branching control of A–P and D–V patterning genes (Krasnow, of distal endoderm, and recent experiments have shown 1997). that it induces the expression of genes normally tran- The identification of Fgf10 as a key regulator of early scribed in the proximal lung and trachea (Shannon et mouse lung development allows several questions to al., 1998). Taken together, these experiments support a be approached at a cellular and biochemical level. For key role for signaling factors made in the distal meso- example, is Fgf10 sufficient to reproduce the effect of derm in regulating bud initiation and outgrowth from the grafting distal mesenchyme around tracheal endoderm? endoderm. The answer is “no.” A heparin bead loaded with Fgf10 protein does not induce the formation of an ectopic bud The Role of Fgf10 in Lung Morphogenesis when placed adjacent to tracheal endoderm covered in A breakthrough in identifying the mesodermal factors tracheal mesoderm. However, isolated E11.5 tracheal controlling lung morphogenesis came with the tar- endoderm does respond to an Fgf10 bead in culture by geted expression of a dominant-negative Fgf receptor proliferating and forming multiple bud-like processes, (Fgfr2IIIb) in the endoderm of transgenic mouse em- albeit more slowly than distal endoderm (Park et al., bryos (Peters et al., 1994). In these embryos, the trachea 1998) (M. Weaver and B. L. M. H., unpublished results). and primary bronchi developed normally, but no sec- This suggests that the tracheal mesoderm inhibits, di- ondary or terminal buds were present. Of the several rectly or indirectly, the initial and/or sustained outgrowth Fgf genes transcribed in the developing lung, Fgf10 ap- of endoderm in response to an ectopic source of Fgf. peared to be the most likely candidate for regulating When placed adjacent to distal lung buds, Fgf10- budding, based on its highly localized and early expres- loaded beads will stimulate their outgrowth toward and sion, particularly since the Fgfr2 receptor is uniformly around the beads. Moreover, if pieces of distal mesen- expressed in the endoderm (Bellusci et al., 1997b; Park chyme-free endoderm are placed in Matrigel culture, et al., 1998). This hypothesis has recently been con- they will move toward a localized source of Fgf10 (Park firmed by the finding that Fgf10Ϫ/Ϫ embryos lack lung et al., 1998) (Figure 4). Precisely how the epithelial cells buds as well as limbs (Min et al., 1998; Sekine et al., sense the Fgf10 gradient and move in the Matrigel, and 1999). whether this involves changes in cell polarity and adhe- The dynamic temporal and spatial pattern of Fgf10 sion, cell rearrangement within the explanted bud, pseu- expression in the mouse lung is highly reminiscent of dopodial extensions, and/or expression of specific inte- the developing tracheal system in Drosophila where the grins, is not yet known. Study of this problem should gene encoding the Fgf ligand Branchless (Bnl) is tran- provide important insights into morphogenetic subrou- scribed segmentally in discrete clusters of cells outside tines. the initial tracheal sacs at positions where branches will form (Sutherland et al., 1996; Hogan and Yingling, 1998). Regulation of Fgf10 Expression in the Early Lung Tracheal cells, which uniformly express the Fgf recep- As discussed earlier, a common feature of budding sys- tor Breathless (Btl), migrate and elongate without cell tems seems to be a balance between self-enhancing Cell 230

feedback loops that positively regulate distal outgrowth and mechanisms that inhibit this process more proximally. Do such genetic interactions play any role in lung bud outgrowth and, if so, are the genetic pathways conserved with other budding systems such as feathers, limbs, and the Drosophila tracheal system? Candidates for factors that positively regulate Fgf10 in the distal mesenchyme are those encoded by genes expressed at high levels in the distal endoderm. These include Shh, Bmp4, and Wnt7b (Bitgood and McMahon, 1995; Hogan et al., 1997; Bellusci et al., 1997a; Litingtung et al., 1998; Pepicelli et al., 1998) (see Figure 3C for Shh). In homozygous null Shh embryos, two primary buds grow out, but mesenchymal proliferation is severely reduced and only a few, very abnormal cysts develop, supporting earlier evidence that Shh is a mitogen for the mesenchyme (Bellusci et al., 1997a). However, the mutant lungs show apparent upregulation of Fgf10 in the surrounding thin layer of mesoderm, including the mesoderm immedi- ately adjacent to the endoderm (Litingtung et al., 1998; Pepicelli et al., 1998). These findings establish that Fgf10 expression is not absolutely dependent on Shh but are consistent with a model in which high levels of Shh produced by the distal endoderm actually downregulate Fgf10 in the mesenchyme (Bellusci et al., 1997b). Further investigation is needed to see if Bmp4 also inhibits Fgf10 Figure 5. Two Models for Patterning of Bud Sites in the Early Mouse expression in the lung. Lung A potentially very important negative feedback mech- The figure shows schematically the generation of three secondary anism is predicted from studies on the genes affecting buds (1, 2, and 3) on the dorsal (D), ventral (V), and lateral sides, respectively, of a hypothetical right primary bud as it grows out the branching of the tracheal system in Drosophila. along the proximodistal axis (dotted line in A1). Model A predicts Among these genes is sprouty (spry), which encodes a that the future domains of Fgf10 expression (dotted blue circles) membrane-associated or secreted protein expressed at are “prepatterned” or “hardwired” at the time of primary bud out- high levels by the apical cells (Hacohen et al., 1998). Spry growth (A1). As the lung grows (A2), the domains start to express acts non–cell autonomously to inhibit the branching of Fgf10 in a proximal-to-distal sequence but retain their positions adjacent cells and is induced by the FGF ligand Bnl. Spry relative to the spatial coordinates of the primary bud. Secondary buds, as well as distal tips, express high levels of Bmp4 and Shh, therefore restricts the outgrowth of primary branches to although lower levels of the latter are produced throughout most of the apical tip cells nearest the source of Bnl. However, the endoderm. Model B predicts that lateral inhibitory mechanisms the inhibitory effect of Spry can be overcome in lateral from the distal tip endoderm of the primary and later secondary cells by high doses of exogenous FGF ligand, which buds (red lines) play a major role in determining budding pattern. accounts for the fact that new branches can be induced As the primary bud grows from stage B1 to B2, this lateral inhibition in vivo by ectopic expression of Bnl. At least four mam- is relieved first in the endoderm region where bud 1 will form, then in the region where bud 2 will arise, etc. As the buds grow out in malian sprouty genes have been identified, one of which response to low levels of Fgf10 in the mesoderm, they produce (sprouty 4) is expressed in the lung mesoderm (de Max- factors that upregulate expression of the gene in the overlying meso- imy et al., 1999). If, as seems likely, other members of derm and also inhibit other buds in the vicinity. Inhibitory mecha- the family are expressed in the endoderm, they may nisms also probably work in the mesenchyme to repress bud forma- serve to restrict the proliferative and motogenic re- tion dorsally and ventrally after buds 1 and 2 have formed. sponse to Fgf10 to a small population of endoderm cells at the distal tip of outgrowing buds and to inhibit lateral branching. The balance between Fgf10 levels in the necessarily mutually exclusive, models can be proposed overlying mesoderm and spry inhibition in the endoderm for how the budding pattern is regulated. Both assume might then play a role in promoting distal outgrowth and that a localized source of Fgf10 is required for the di- in determining the distance between one bud and the rected outgrowth of bud endoderm but is not necessarily next (Figure 5B). sufficient to determine the site of bud initiation. The first model is based on that proposed for early tracheal development in Drosophila, where the out- Two Models for the Early Morphogenesis growth of branches from the tracheal sac is determined of the Mouse Lung by small groups of cells that express the Fgf ligand If we are to understand early lung development at the Branchless. The location of these cells at specific points molecular level and formulate principles that can be outside the tracheal sac is genetically controlled and applied to other budding/branching systems, a great “hardwired” to the spatial coordinates of each segment deal more needs to be learned about the dynamics of (Krasnow, 1997). Figure 5 shows a variation of this model gene expression and cell lineage and about mutations applied to a hypothetical right primary bud early in affecting budding pattern. In the meantime, two, not mouse lung development. The sites of Fgf10 expression Review 231

that appear sequentially in the mesoderm as the primary bud elongates are genetically determined and related to the D–V, P–D, and M–L coordinates of the initial bud and ultimately to the foregut from which it is derived. Moreover, the sites of secondary bud outgrowth are independent of the activity of the distal tip of the primary bud. In the second model, the positioning of secondary buds is “generative” rather than “hardwired” and depends on a dynamic but reproducible interaction between multiple factors in both the endoderm and mesoderm that promote or inhibit bud initiation. According to this model (Figure 5; see also Bellusci et al., 1997b), the formation of secondary buds is inhibited within a specific distance from the distal tip of the primary bud by mechanisms acting laterally within the endoderm. Figure 6. A Leafy Seadragon, Phycodurus eques, from South Aus- tralia. For further details see Groves (1998) and http://www.nexus. This lateral inhibition may be mediated, for example, by edu.au/schools/kingscot/pelican/seadragon /sd_001.htm. homologs of spry, or by Bmp4, produced at the distal tip, by Notch–Notch ligand signaling, or by a combination of several such mechanisms. Once the proximal endoderm camouflages itself with frond-like appendages mimick- falls outside the lateral inhibition zone it can respond to ing seaweed (Figure 6 and Groves, 1998). hypothetical low concentrations of Fgf10 in the mesoderm, and a secondary bud is initiated. Factors released Acknowledgments by this new bud upregulate Fgf10 in the overlying mesenchyme and inhibit lateral buds in the flanking region, I am grateful for many stimulating discussions with my colleagues in and a new cycle of bud outgrowth begins. The interbud the Vanderbilt Developmental Biology Program, in particular Justin distance does not have to be fixed but can vary, de- Grindley and Molly Weaver, and with Mark Krasnow. Special thanks are also due to Susan Kidson. Many people have generously given pending, for example, on the time the tip has been grow- advice about limb development, especially Cindy Loomis, Cheryl ing. The specific sites at which buds appear can also Tickle, John Fallon, and Cliff Tabin. Work from my lab on lung devel- be influenced by inhibitory factors in the mesoderm and opment is supported by the National Institute for Child Health and by the deposition of extracellular matrix molecules. Human Development. I am an Investigator of the Howard Hughes It is possible that both of these models play a role in Medical Institute. vivo. For example, the first mechanism may predominate References in setting up the initial buds that give rise to the future main lobes, while the second mechanism may be more Altabef, M., Clarke, J.D.W., and Tickle, C. (1997). Dorso-ventral ecto- important later, particularly during the developmental dermal compartments and origin of apical ectodermal ridge in devel- phase when dichotomous branching predominates. Fi- oping chick limb. Development 124, 4547–4556. nally, in the last stages of lung development, the lobula- Beddington, R.S.P., and Robertson, E.J. (1999). Axis development tion of terminal buds into numerous alveoli by the very and early asymmetry in mammals. Cell 96, this issue, 195–209. different process of clefting comes into play. Bell, S.M., Schreiner, C.M., and Scott, W.J. (1998). The loss of ventral ectoderm identity correlates with the inability to form an AER in the In order to test these and other models, new tech- legless hindlimb bud. Mech. Dev. 74, 41–50. niques will have to be developed to follow lung morpho- Bellusci, S., Furuta, Y., Rush, M.G., Henderson, R., Winnier, G., and genesis and gene expression over time in organ culture. Hogan, B.L.M. (1997a). Involvement of Sonic hedgehog (Shh) in This challenge has recently been taken up for the devel- mouse embryonic lung growth and morphogenesis. Development oping mouse kidney, where it is now possible to follow 124, 53–63. the branching of the ureteric bud epithelium in organ Bellusci, S., Grindley, J., Emoto, H., Itoh, N., and Hogan, B.L.M. culture with a green fluorescent transgene marker (Srini- (1997b). Fibroblast growth factor 10 (FGF10) and branching morpho- vas et al., 1999) (see http://cpmcnet.columbia.edu/dept/ genesis in the embryonic mouse lung. Development 124, 4867–4878. genetics/kidney). In the pictures, separation of terminal, Bitgood, M.J., and McMahon, A.P. (1995). Hedgehog and Bmp genes are coexpressed at many diverse sites of cell-cell interaction ampulla-like buds into multiple branches can be fol- in the mouse embyro. Dev. Biol. 172, 126–138. lowed. It would also be very useful to be able to identify Bogue, C.W., Lou, L.J., Vasavada, H., Wilson, C.M., and Jacobs, particular cells in which signaling from ligands such as H.C. (1996). Expression of Hoxb genes in the developing mouse Bmps and Fgfs is active. New approaches such as these, foregut and lung. Am. J. Resp. Cell Mol. Biol. 15, 163–171. coupled with the analysis of branching patterns in mu- Cooke, J., and Summerbell, D. (1980). Cell cycle and experimental tant mice and the identification of mouse homologs of pattern duplication in the chick wing during embryonic development. genes important in budding and branching morphogen- Nature 287, 697–701. esis in other model organisms such as Drosophila are Crowe, R., Henrique, D., Ish-Horowicz, D., and Niswander, L. (1998). likely to yield important advances in the near future. A new role for Notch and Delta in cell fate decisions: patterning the feather array. Development 125, 767–775. Finally, one may ask if common strategies underlie de Maximy, A.A., Nakatake, Y., Moncada, S., Itoh, N., Thiery, J.P., branching morphogenesis in plants and animals. With and Bellusci, S. (1999). Cloning and expression pattern of a mouse this in mind, a fascinating creature to contemplate is homologue of Drosophila sprouty in the mouse embryo. Mech. Dev., the Leafy Seadragon, Phycodurus eques, a fish that in press. Cell 232

Grindley, J.C., Bellusci, S., Perkins, D., and Hogan, B.L.M. (1997). Min, H., Danilenko, D.M., Scully, S.A., Bolon, B., Ring, B.D., Tarpley, Evidence for the involvement of the Gli gene family in embryonic J.E., DeRose, M., and Simonet, W.S. (1998). Fgf-10 is required for mouse lung development. Dev. Biol. 188, 337–348. both limb and lung development and exhibits striking functional Groves, P. (1998). Leafy sea dragons. Sci. Am. 279, 84–89. similarity to Drosophila branchless. Genes Dev. 12, 3156–3161. Hacohen, N., Kramer, S., Sutherland, D., Hiromi, Y., and Krasnow, Mollard, R., and Dziadek, M. (1998). A correlation between epithelial M.A. (1998). sprouty encodes a novel antagonist of FGF signaling proliferation rates, basement membrane component localization that patterns apical branching of the Drosophila airways. Cell 92, patterns, and morphogenetic potential in the embryonic mouse lung. 253–263. Am. J. Respir. Cell Mol. Biol. 19, 71–82. Helms, J.A., Kim, C.H., Hu, D., Minkoff, R., Thaller, C. and Eichele, Motoyama, J., Liu, J., Mo, R., Ding, Q., Post, M., and Hui, C.C. (1998). G. (1997). Sonic hedgehog participates in craniofacial morphogene- Essential function of Gli2 and Gli3 in the formation of lung, trachea sis and is down-regulated by teratogenic doses of retinoic acid. and oesophagus. Nat. Genet. 20, 54–57. Dev. Biol. 187, 25–35. Nakanishi, Y., and Ishii, T. (1989). Epithelial shape change in mouse Hogan, B.L.M., and Yingling, J.M. (1998). Epithelial/mesenchymal embryonic submandibular gland: modulation by extracellualr matrix interactions and branching mophogenesis of the lung. Curr. Opin. components. Bioessays 11, 163–167. Genet. Dev. 8, 481–486. Nogawa, H., Morita, K., and Cardoso, W.V. (1998). Bud formation Hogan, B.L.M., Grindley, J., Bellusci, S., Dunn, N.R., Emoto, H., and precedes the appearance of differential cell proliferation during Itoh, N. (1997). Branching morphogenesis of the lung: new models branching morphogenesis of mouse lung epithelium in vitro. Dev. for a classical problem. Cold Spring Harbor Symp. Quant. Biol. 62, Dyn. 213, 228–235. 249–256. Noramly, S., and Morgan, B.A. (1998). BMPs mediate lateral inhibi- Hornbruch, A., and Wolpert, L. (1970). Cell division in the early tion at successive stages in feather tract development. Development growth and morphogenesis of the chick limb. Nature 226, 764–766. 125, 3775–3787. Irvine, K., and Vogt, T.F. (1997). Dorso-ventral signaling in limb devel- Ohuchi, H., Nakagawa, T., Yamamoto, A., Araga, A., Ohata, T., Ishi- opment. Curr. Opin. Cell Biol. 9, 867–876. maru, Y., Yoshioka, H., Kuwana, T., Nohno, T., Yamasaki, M., et al. Johnson, R.L., and Tabin, C.J. (1997). Molecular models for verte- (1997). The mesenchymal factor, FGF10, initiates and maintains the brate limb development. Cell 90, 979–990. outgrowth of the chick limb bud through interaction with FGF8, an Johnston, S., Rauskolb, C., Wilson, R., Prabhakaran, B., Irvine, K., apical ectodermal factor. Development 124, 2235–2244. and Vogt, T. (1997). A family of mammalian Fringe genes implicated Park, W.Y., Miranda, B., Lebeche, D., Hashimoto, G., and Cardoso, in boundary determination and the Notch pathway. Development W.V. (1998). FGF-10 is a chemotactic factor for distal epithelial buds 124, 2245–2254. during lung development. Dev. Biol. 201, 125–134. Jung, H.-S., Francis-West, P.H., Widelitz, R.B., Jiang, T.-X., Ting- Pepicelli, C., Kispert, A., Rowitch, D., and McMahon, A. (1997). GDNF Berreth, S., Tickle, C., Wolpert, L., and Chuong, C.-M. (1998). Local induces branching and increased cell proliferation in the ureter of inhibitory action of BMPs and their relationships with activators in the mouse. Dev. Biology 192, 193–198. feather formation: implications for periodic patterning. Dev. Biol. Pepicelli, C.V., Lewis, P.M., and McMahon, A.P. (1998). Sonic hedge- 196, 11–23. hog regulates branching morphogenesis in the mammalian lung. Kim, S.K., and Melton, D.A. (1998). Pancreas development is pro- Curr. Biol. 8, 1083–1086. moted by cyclopamine, a hedgehog signaling inhibitor. Proc. Natl. Peters, K., Werner, S., Liao, X., Wert, S., Whitsett, J., and Williams, Acad. Sci. USA 95, 14552–14557. L. (1994). Targeted expression of a dominant negative FGF receptor Klambt, C., Glazer, L., and Shilo, B.-Z. (1992). Breathless, a Drosoph- blocks branching morphogenesis and epithelial differeniation of the ila FGF receptor homolog, is essential for migration of trachael and mouse lung. EMBO J. 13, 3296–3301. specific midline glial cells. Genes Dev. 6, 1668–1678. Rancourt, D.E., Tsuzuki, T., and Capecchi, M.R. (1995). Genetic inter- Krasnow, M.A. (1997). Genes that control organ form: lesson from action between hoxb-5 and hoxb-6 is revealed by nonallelic non- bone and branching morphogenesis. Cold Spring Harbor Symp. complementation. Genes Dev. 9, 108–122. Quant. Biol. 62, 235–240. Sekine, K., Ohuchi, H., Fujiwara, M., Yamasaki, M., Yoshizawa, T., Lee, T., Hacohen, N., Krasnow, M., and Montell, D.J. (1996). Regu- Sato, T., Tagishita, N., Matsui, D., Koga, Y., Itoh, N., and Kato, S. lated breathless receptor tyrosine kinase activity required to pattern (1999). FGF10 is essential for the limb and lung development. Nat. cell migration and branching in the Drosophila tracheal system. Genet., in press. Genes Dev. 10, 2912–2921. Shannon, J.M. (1994). Induction of alveolar type II cell differentiation Litingtung, Y., Lei, L., Westphal, H., and Chiang, C. (1998). Sonic in fetal trachael epithelium by grafted distal lung mesenchyme. Dev. hedgehog is essential to foregut development. Nat. Genet. 20, Biol. 166, 600–614. 58–61. Shannon, J.M., Nielson, L.D., Gebb, S.A., and Randell, S.H. (1998). Loomis, C.A., Kimmel, R.A., Tong, C.X., Michaud, J., and Joyner, Mesenchyme specifies epithelial differentiation in reciprocal recom- A.L. (1998). Analysis of the genetic pathway leading to formation of binanant of embryonic lung and trachea. Dev. Dyn. 212, 482–494. ectopic apical ectodermal ridges in mouse Engrailed-1 mutant limbs. Development 125, 1137–1148. Srinivas, S., Goldberg, M.R., Watanabe, T., D’Agati, V., Al-Awqati, Q., and Costantini, F. (1999). Expression of green fluorescent protein Martin, G.R. (1998). The roles of FGFs in the early development of vertebrate limbs. Genes Dev. 12, 1571–1586. in the ureteric bud of transgenic mice: a new tool for the analysis of ureteric bud morphogenesis. Dev. Genet., in press. Meinhardt, H. (1996). Models of biological pattern formation: common mechanisms in plant and animal development. Int. J. Dev. Biol. Sutherland, D., Samakovlis, C., and Krasnow, M.A. (1996). branch- 40, 123–134. less encodes a Drosophila FGF homolog that controls tracheal cell migration and the pattern of branching. Cell 87, 1091–1101. Mendelsohn, C., Lohnes, D., Decimo, D., Lufkin, T., LeMeur, M., Chambon, P., and Mark, M. (1994). Function of the retinoic acid Tabin, C. (1998). A developmental model for thalidomide defects. receptors (RARs) during development. Development 120, 2749– Nature 396, 322–323. 2771. Vargesson, N., Clarke, J., Vincent, K., Coles, C., Wolpert, L., and Meno, C., Shimono, A., Saijoh, Y., Yashiro, K., Mochida, K., Ohishi, Tickle, C. (1997). Cell fate in the chick limb bud and relationship to S., Noji, S., Kondoh, H., and Hamada, H. (1998). lefty-1 is required gene expression. Development 124, 1909–1918. for left-right determination as a regulator of lefty-2 and nodal. Cell Viallet, J.P., Prin, F., Olivera-Martinez, I., Hirsinger, E., Pourquie, O., 94, 287–297. and Dhouailly, D. (1998). Chick Delta-1 gene expression and the Michaud, J., Lapointe, F., and Le Douarin, N. (1997). The dorsoven- formation of the feather primordia. Mech. Dev. 72, 159–168. tral polarity of the presumptive limb is determined by signals pro- Wasserman, J.D., and Freeman, M. (1998). An autoregulatory cas- duced by the somites and by the lateral somatopleure. Development cade of EGF receptor signaling patterns the Drosophila egg. Cell 124, 1453–1463. 95, 355–364. Review 233

Wolpert, L., Beddington, R., Brockes, J., Jessell, T., Lawrence, P., and Meyerowitz, E. (1998). Principles of Development. (Oxford: Ox- ford University Press). Yang, Y., Drossopoulou, G., Chuang, P.-T., Duprez, D., Marti, E., Bumcrot, D., Vargesson, N., Clarke, J., Niswander, L., McMahon, A., and Tickle, C. (1997). Relationship between dose, distance and time in Sonic Hedgehog-mediated regulation of anteroposterior polarity in the chick limb. Development 124, 4393–4404. Zaret, K. (1998). Early liver differentiation: genetic potentiation and multilevel growth control. Curr. Opin. Genet. Dev. 8, 526–531. Zeller, R., and Duboule, D. (1997). Dorso-ventral limb polarity and orgin of the ridge: on the fringe of independence. Bioessays 19, 541–546.