Regulation of organ size: Insights from the Drosophila Hippo signaling pathway

Abstract

Organ size control is a fundamental and core process of development of all multicellular organisms. One important facet of organ size control is the regulation of cell proliferation and cell death. Here we address the question, What are the developmental mechanisms that control intrinsic organ size? In several multicellular animals including humans and flies, organs develop according to an instructive model where proliferation is regulated by extracellular signals. However, the signals that regulate proliferation (and organ size) remain poorly understood. Recent data from flies have shed some light on the molecular mechanisms that regulate growth and size of organs. In this review, we will briefly discuss classic studies that revealed the mysteries of growth regulation. We will then focus on the recent findings from the Drosophila Hippo signaling pathway and its role in the regulation of organ size. Finally, we will discuss the mammalian Hippo pathway, and its implications in regulation of growth/proliferation during development and disease. Developmental Dynamics 238:1627–1637, 2009. © 2009 Wiley-Liss, Inc.

INTRODUCTION

A fundamental question in biology is, how is growth regulated during development to produce organs of particular sizes? Every organism begins life as a single-celled embryo, but spectacularly diverse sizes and shapes are formed as different species develop into their respective adult forms. Interestingly, each species develops to a certain stereotypical size and forms proportionately developed organs and organ systems. Furthermore, the growth rate of most species is predictable, despite the vast range of individual differences (in size), and in general, most multicellular organisms produce organs containing cells of reproducible size. Thus, a mechanism must exist for controlling size at the gross level for the regulation of organism or organ size, and at the finer level for controlling size at the level of a single cell. How is this dynamic regulation of tissue size achieved during development? On the other hand, once cellular proliferation begins in a developing embryo, how do cells know when to stop? Is the signal that instructs cells to exit the cell cycle and enter differentiation, the same as the “stop” signal that restricts cell proliferation? These questions have intrigued developmental biologists for a long time (reviewed in Conlon and Raff,1999).

Studies on growth control over the past several decades generated important insights into regulation of growth and size. Growth of organisms/organs is governed by both intrinsic (genetic) and extrinsic (nutrient availability, extracellular signals) cues during development (Stocker and Hafen,2000). Several cellular processes including nutrient uptake, cell division, and cell death contribute to growth during development. These cellular processes are integrated with other processes like signaling pathways, and constitute developmental programs that are sufficiently robust and plastic (flexible) to allow for changes in the size of a developing organ/organism. Together these processes give rise to an organism of proper size, pattern, and proportion (reviewed in Neufeld,2003). Recently, a new signaling pathway was identified in flies that controls tissue and organismal size by regulating cell numbers in growing organs. This pathway, called the Hippo signaling pathway, is a network of tumor suppressor genes and oncogenes, and is conserved from flies to humans (reviewed in Edgar,2006; Pan,2007; Reddy and Irvine,2008). In this review, we will discuss the role of the Hippo signaling pathway in regulation of growth and organ size control, and its implications in cancer—where regulation of cell proliferation and cell death is perturbed.

Classical studies lay the groundwork of our current understanding of the paradigms of growth and size regulation. Drosophila imaginal discs and other model organisms generated important insights about the growth characteristics of cells in growing tissues. The Drosophila imaginal discs are a very popular and well-characterized model system in which many aspects of pattern formation, growth regulation, and cell proliferation was experimentally approached (reviewed in French et al.,1976; Bryant and Simpson,1984; Cohen,1993). These studies revealed that (1) cells carry growth information that is intrinsic to the tissue or organ. Thus, the cells in a growing tissue recognize an endpoint in growth and stop dividing once the tissue has reached its target size. (2) Growth is plastic, as during the growth phase cells can alter their growth programs to regenerate both cell numbers and patterning information even when a large number of cells are lost/destroyed in developing tissues/organs. (3) During disc growth, proliferation is independent of patterning, and cell interactions that control growth are independent of cell lineage. (4) The regeneration experiments elegantly showed that intercellular interactions regulated growth during development, and that cell populations and not individual cells were units of growth. (5) Other studies showed that acceleration or deceleration of cell division rate (Morata and Ripoll,1975; Weigmann et al.,1997; Neufeld et al.,1998), or apoptosis is not the mechanism through which organ size is regulated. In summary, these studies revealed that the size of the developing organs depends largely on other as yet unknown cell–cell interactions.

Although the exact nature of the size-regulating pathway remained elusive, the classical studies in model organisms and Drosophila imaginal discs suggested that the determination of organ size is intimately linked to the relationship between cell growth, cell division, and the termination (restriction) of cell division. Developmental and cancer biologists have long pondered how cell proliferation is restricted, first during embryogenesis to allow morphogenesis and later in adults to maintain homeostasis. The premise for identifying genes that function to limit growth/proliferation is that if a signal exists that instructs the cells to stop proliferating and exit the cell cycle to differentiate, then such a signal can be genetically impaired and genes that are responsible for conveying the “stop growth” signal to cells/tissues can be identified. The advent of modern methods for generating genetic mosaics in the fruit fly, Drosophila melanogaster (Xu and Rubin,1993), prompted genetic screens in which genetically marked clones of cells with mutations in random genes were produced, and the adults were scored for abnormal (tumor-like) overgrowths. During the period of 1995–2005, several groups carried out genetic screens in fruit flies using tissue-specific genetic mosaics (reviewed in St Johnston,2002), to identify mutants that specifically affected growth but not patterning. One of the first tumor-suppressor genes identified by such screens was named large tumor suppressor (lats) by Gerald Rubin's group (Xu et al.,1995), and independently identified and named warts (wts) by Peter Bryant's group (Justice et al.,1995). Incidentally, lats/wts turned out to be the first molecularly defined member of the Hippo pathway that was discovered approximately 8 years later.

PATHWAYS REGULATING ORGAN SIZE: THE DISCOVERY OF HIPPO SIGNALING PATHWAY

The Identification of Hippo Pathway Mutants

A striking feature of many of the Hippo pathway components is their distinctive mutant phenotype, which results in dramatic overgrowth of mutant tissue in the imaginal discs and in adult organs (reviewed in Edgar,2006; Pan,2007; Saucedo and Edgar,2007). Thus, hippo (hpo), salvador (Sav), and warts (wts) were isolated based on their overgrown adult head phenotype in chemical mutagenesis screens. Mob-as-tumor-suppressor (mats) was a spontaneous mutation that became a part of the Hippo pathway in part due to the similarity of its mutant phenotype in adult heads and imaginal discs to other Hippo pathway mutants. Previously known genes like fat (ft), expanded (ex), and merlin (mer) were identified as components of the Hippo pathway based on a common/shared phenotype. Yorkie (Yki), on the other hand, was identified in a biochemical screen (yeast two-hybrid) for Warts interacting proteins.

The adult heads of flies of Hippo pathway mutants in which over 90% of cells were homozygous mutant were proportionately larger than the other structures but had a normal overall pattern. The mutant cells proliferated faster to out-compete the normal wild-type cells (Fig. 1). The underlying cause of overgrowth observed in the adult eyes, was studied in both the pupal retinae and the eye–antennal imaginal discs. In the adult eye sections, the pattern of ommatidial development was largely unaffected except for the spacing of the ommatidial clusters. To trace back this phenotype, the mid-pupal retinae were studied in which cellular outlines were marked using antibodies. Hippo pathway mutants exhibited a dramatic increase in the numbers of interommatidial cells when compared with wild-type (Fig. 1).

Figure 1. The overgrowth phenotype of Hippo pathway mutants. A–D: In comparison to normal flies (A), the adult heads of flies mutant for Warts (D), Salvador (C), and Hippo (B) show overgrowth and increase in overall size. The mutant cells are generated by a tissue-specific mosaic technique in which the normal cells (+/+) are killed due to the presence of a cell-lethal mutation, the mutant cells are marked by the presence of the white pigment cells, and the heterozygous cells appear orange/red. The eyes of flies mutant for Hippo, Salvador, and Warts are almost completely (95%) white, indicating that the entire eye is composed of mutant cells, whereas in normal wild-type flies using the same technique approximately 50% of the cells are heterozygous. E–H: Pupal retinae stained with an antibody against Discs large, a protein that marks cell outlines, clearly shows a shared phenotype of excess interommatidial cells in Hippo pathway mutants (f–h) compared with wild-type (e).

The extra interommatidial cells could be due to excessive cell proliferation, increased spacing of photoreceptor clusters during patterning, lack of apoptosis, or a combination thereof. Systematic analyses revealed that this phenotype results because Hippo pathway mutant cells proliferate faster than surrounding wild-type cells and because they do not terminate proliferation when imaginal tissues have reached their normal size. In fact, hpo mutant cells do not deregulate cell proliferation of terminally arrested cells, as cells differentiate normally. In addition, Hippo affects cell survival. In the wild-type pupal retina, extra interommatidial cells are removed by apoptosis, while extra hpo mutant cells are not. Thus, Hippo has a dual function in promoting cell proliferation arrest and apoptosis (Tapon et al.,2002; Harvey et al.,2003; Jia et al.,2003; Udan et al.,2003; Wu et al.,2003; reviewed in Harvey and Tapon,2007). These effects on proliferation arrest and apoptosis are not limited to the eye–antennal disc, but are observed in all disc-derived structures, including wings, halteres, legs, and thorax, suggesting that Hippo function is ubiquitously required. The ubiquitous expression pattern of several Hippo pathway components in imaginal discs supports this conclusion.

The molecular characterization of Hpo (Ste-20 kinase) and Sav (WW domain adaptor), together with the previously known Lats/Wts (Dbf-related kinase), shed important insights on the interactions between the proteins encoded by these genes, and a signaling pathway began to emerge (Jia et al.,2003; Udan et al.,2003; Wu et al.,2003; reviewed in Harvey and Tapon,2007). Since then, the addition of the Mob super-family protein Mats (Lai et al.,2005), two four point one, ezrin, radixin, moesin (FERM) domain-containing proteins Ex and Mer (Hamaratoglu et al.,2006), a transmembrane Proto-Cadherin Ft (Bennett and Harvey,2006; Cho et al.,2006; Silva et al.,2006; Willecke et al.,2006), the transcriptional coactivator Yki (Huang et al.,2005), and the transcription factor Scalloped (Sd; Goulev et al.,2008; Wu et al.,2008; Zhang et al.,2008b) laid down the frame-work of a signaling network from the membrane to the nucleus for the Hippo signaling pathway. Several targets of Hippo signaling were identified during the course of characterization of mutant phenotypes of pathway components, which include Cyclin E (CycE; Kango-Singh et al.,2002; Tapon et al.,2002), Cyclin B (CycB; Tyler and Baker,2007) and E2F (Nicolay and Frolov,2008; all regulating cell proliferation), Drosophila inhibitor of apoptosis protein 1 (DIAP1; apoptosis; Tapon et al.,2002), the cytoskeletal proteins ex (cell shape; Hamaratoglu et al.,2006), the microRNA bantam (Nolo et al.,2006), and target genes that function in other pathways and Hippo signaling like four-jointed (fj; planar cell polarity; Fanto et al.,2003; Cho and Irvine,2004), division abnormally delayed (dally) and dally-like (dlp; Drosophila HSPGs primarily function in the spread of morphogen signals; Baena-Lopez et al.,2008), and wingless (wg; Cho and Irvine,2004) involved in patterning and growth (reviewed in Edgar,2006; Yin and Pan,2007; Reddy and Irvine,2008).

The Kinase Cascade in the Hippo Pathway

The two protein kinases, Hpo and Wts form a core kinase cascade in the Hippo pathway. Hpo is a member of the Ste-20 superfamily of kinases, and acts upstream of Wts, a member of the nuclear Dbf2 related (NDR) family of protein kinases. Hpo activates Wts by direct phosphorylation without affecting the levels of Wts expression (Wu et al.,2003). Hpo-mediated Wts activation is facilitated by Sav, a WW domain-containing protein that probably functions as a scaffold, and also by Mats, a protein that binds directly to Wts and Hpo (Wu et al.,2003; Wei et al.,2007). When activated, Wts can phosphorylate Yki and reduce its activity, and as a consequence Yki is excluded from the nucleus (Dong et al.,2007). The phosphorylation sites important for activation of Hpo and Wts kinases and inactivation of Yki have been mapped in flies and in mammalian systems (Dong et al.,2007; Zhao et al.,2007; Hao et al.,2008; Oh and Irvine,2008; Zhang et al.,2008a). Yki promotes, either directly or indirectly, the transcription of genes that promote growth and cell-cycle progression as well as genes that inhibit apoptosis (Huang et al.,2005; Dong et al.,2007; Zhao et al.,2008a). Thus, increased activity of Hpo and Wts correlates with a reduction in tissue growth and a reduction of Hpo/Wts activity allows growth to occur.

Upstream Components of the Hippo Pathway

Following the discovery of the core kinase cascade, an important question that remained unanswered was: do extracellular signals regulate Hpo signaling? To understand the regulation of the Hippo pathway, a search for additional components of this pathway was initiated. Several candidate genes were tested for their ability to induce Hpo target genes in mutant clones induced in the imaginal discs, and for the signature “extra interommatidial cell” phenotype in the pupal retinae. These efforts led to the discovery of two proteins with FERM domains, Mer and Ex (Hamaratoglu et al.,2006), and the protocadherin Ft as activators of Hpo (Bennett and Harvey,2006; Cho et al.,2006; Silva et al.,2006; Willecke et al.,2006). Individually ex, mer, or ft had very weak effects on the extra interommatidial cell phenotype typical of the loss of Hippo signaling pathway (Willecke et al.,2006). However, loss-of-function clones of mer, ex double mutants and ft strongly induced CycE and Diap1, two well-known targets of the Hippo pathway.

Of interest, simultaneous loss of mer and ex resulted in (a) impaired endocytic trafficking, leading to up-regulation of multiple cell growth and proliferation pathways (Maitra et al.,2006); and (b) enormous overgrowth of the mutant tissue (more than either single mutant), that is similar to the phenotypes of loss of Hpo signaling (Hamaratoglu et al.,2006). Furthermore, characterization of the double mutant phenotype revealed that cells mutant for mer and ex showed all the defects of reduction/down-regulation of Hpo signaling, for example, increased growth, delayed cell cycle exit, ectopic cell survival, and up-regulation of downstream target genes, including diap1, cycE, and ex (Hamaratoglu et al.,2006). Further genetic and biochemical analysis of ex and mer double mutants showed that these two proteins act upstream of the Hpo signaling cascade to regulate proliferation arrest and apoptosis.

However, recently distinct functional requirements of Drosophila Mer and Ex in the Hippo pathway were reported (Pellock et al.,2007). Merlin and Expanded both contain a FERM domain, but structurally Drosophila Ex is distinct from Mer, and phylogenetically distinct from the ERM proteins (Boedigheimer et al.,1993; Bretscher et al.,2002). Earlier studies showed the additive phenotype of the double mutants, and had proposed a redundant function for Ex and Mer (McCartney et al.,2000). Unlike the mer, ex double mutants, loss of ex or mer has distinct effects on growth and apoptosis. mer and ex both induce transcriptional targets of Hippo signaling (CycE and Diap1), but these targets are not robustly induced in mer mutant cells in the imaginal discs (Pellock et al.,2007). Interestingly, loss of ex in the Drosophila eye discs leads to overgrowth as seen by ectopic proliferation (S-phases) in cells that normally cease proliferations and are postmitotic, whereas mer mutant cells induce very few ectopic S-phases, suggesting that ex plays a significant role in cell cycle exit. Loss of ex in the pupal retina has a weak effect such that few extra interommatidial cells survive, whereas mer clones show several extra interommatidial cells. Thus, compared with ex, the apoptotic mechanism is defective in mer mutant cells (Pellock et al.,2007). These data suggest that ex and mer may act together or separately, by means of signaling interactions that converge on the kinase cascade of the Hippo pathway. Alternatively, these observed differences in ex and mer function may reflect the spatial and temporal differences in the expression and/or function of ex and mer. However, the molecular mechanisms that connect ex and mer to the Hpo kinase cascade remain unclear.

Recently, Ft was shown to be the most upstream activator of the Hippo pathway (Bennett and Harvey,2006; Cho et al.,2006; Silva et al.,2006; Willecke et al.,2006, Taylor and Baker,2007). Ft encodes a transmembrane protein with multiple Cadherin repeats in its extracellular domain (Mahoney et al.,1991). This presented the possibility that Hpo signaling may indeed be regulated by an extracellular signal—probably a ligand expressed on the surface of an adjacent cell. The effects of loss of ft on tissue growth and planar polarity were well documented but the underlying mechanism by which ft regulated growth was not known. Loss of ft in mutant clones in imaginal discs caused the induction of transcriptional targets of the Hippo pathway. Interestingly, Ft was shown to regulate Hippo pathway activity by two distinct signaling interactions—one directly through ex and the Hpo kinase cascade leading to a Fat-Ex-Hippo pathway (Bennett and Harvey,2006; Silva et al.,2006; Willecke et al.,2006). The other mechanism showed that Ft directly binds to Dachs (D), an atypical myosin, which in turn regulates the abundance of Wts, thereby regulating the Hippo pathway independently of Hpo and other upstream components leading to the Fat-Dachs-Warts pathway (Cho et al.,2006; Feng and Irvine,2007; reviewed in Hariharan,2006; Yin and Pan,2007; Reddy and Irvine,2008).

In addition to Dachs, a second protocadherin and known interactor of Ft called Dachsous (Ds) was recently shown to be involved in Hpo signaling (Rogulja et al.,2008; Willecke et al.,2008). Ds is proposed to act as a ligand for the Ft receptor for their role in the regulation of planar polarity (Saburi and McNeill,2005). Although attractive, the molecular mechanism by which this binding can regulate growth is complex. Both loss-of-function and overexpression of Ds in imaginal discs caused transcriptional induction of Hpo target genes, albeit in a graded manner along the boundary of Ds expression (Willecke et al.,2008). Furthermore, this effect on the target genes was not restricted to the cells missing or overexpressing Ds, indicating that discontinuities in the levels of Ds induced Hpo target genes and presumably growth. Ds, therefore, may not simply act as a ligand, which would induce target genes proportional to its concentration. However, how are discontinuities in Ds levels translated into signal(s) that regulate growth? The answer to this question remains unclear, because the evidence so far suggests that Ds acts through Ft and Dachs to regulate the activity of Wts in both Ex-dependent (using the Fat-Ex-Hpo pathway; Fig. 2A) and Ex-independent (using the Fat-Dachs-Warts pathway; Fig. 2B) manner (Rogulja et al.,2008; Willecke et al.,2008). Furthermore, flies that express uniform Ds/Four-jointed expression have significantly reduced appendages like wings, legs, and so on, but growth is not abolished. The Ds boundary effect thus accounts for some but not all growth control (Rogulja et al.,2008; Willecke et al.,2008). These data indicate that other signaling mechanisms may act in addition to the Ds boundary effect to control imaginal disc size (Willecke et al.,2008). Thus, other currently unknown signals may act (a) in parallel to the Hippo pathway, or (b) act parallel to Ds in the Hippo pathway (for example, by means of Merlin) to regulate tissue growth.

Figure 2. Models of Hippo signaling. A: The Ft-Ex-Hpo model of Hippo signaling, where the transmembrane receptor Fat (Ft) interacts by means of Expanded (Ex) but not Merlin (Mer), which comprise the upstream regulators of Hippo signaling, interacts with the core kinase cascade of Hippo signaling (which include the Hippo kinase (Hpo), the adaptor protein Salvador (Sav) and Mats, and the downstream kinase Warts (Wts)). The core kinase cascade functions by inhibiting the phosphorylation of the transcriptional coactivator Yorkie (Yki). Unphosphorylated form of Yki binds with its cognate transcription factor Scalloped (Sd) and translocates to the nucleus to control expression of various Hippo target genes. Currently, all cell biological functions of Hippo signaling cannot be explained by the regulation of target genes controlled by the Yki/Sd complex. Thus, other transcription factors can partner with Yki to mediate Hippo functions. The Ft binding partner Dachsous (Ds) is also involved in Hippo signaling by means of both modes of Hippo signaling. B: The Ft-Dachs-Wts Model of Hippo signaling contends that the transmembrane protein Ft does not interact with Ex (or Hpo, Sav and Mats), but interacts with Wts through the atypical myosin, Dachs. Dachs negatively regulates the levels and activity of the Wts kinase, which in turn can signal to Yki. Ft-Dachs pathway is directly involved in the regulation of planar cell polarity. In addition to the Ft-Dachs-Wts interaction, another noncanonical interaction is that between Hippo and the Drosophila homolog of the RAS association factor 1 protein (RASSF1). Hpo and RASSF1 bind with each other by means of their SARAH domains to repress the expression of Sav, and activate Hippo signaling. C: The current model of the Mammalian Hippo signaling pathway shows the widespread structural and functional conservation between the mammalian and Drosophila proteins. The core kinase cascade between MST1/2 (Hpo homologs), hWW45 (Sav homolog), MOB (Mats homolog), and the Lats 1/2 (Warts homolog) is conserved at the molecular level. As in flies, the mammalian Lats1/2 in turn prevent the phosphorylation of YAP and TAZ, the two Yki homologs, and allow YAP/TAZ to translocate to the nucleus to bind the TEAD (Sd homolog) transcription factors to regulate Hippo signaling. The interactions upstream of the kinase cascade remain unclear (marked by dotted arrows and question marks); however, NF2 (Merlin homolog) has been shown to interact by means of YAP to regulate cell proliferation, survival, and contact inhibition. Database searches have identified Mammalian Ex but its role in vivo remains to be defined. The mammalian isoform of Fat (Ft4) has recently been implicated in breast cancer, but its functional interaction with mammalian Hippo pathway has not been worked out yet.

The Oncogene Yorkie: The Link Between the Kinase Cascade and Transcription

Yki, the Drosophila ortholog of the mammalian transcriptional coactivator Yes-Associated Protein (YAP), was identified as the missing link between Wts and transcriptional regulation in the Hippo pathway (Huang et al.,2005). Yki was identified in a yeast two-hybrid screen using the N-terminal noncatalytic domain of Wts as bait. Loss-of-function of Yki (named after Yorkshire Terriers, a small dog breed) in mutant clones in imaginal discs results in the development of smaller organs/structures. Yki is required for normal tissue growth and diap1 transcription. Overexpression of yki phenocopies loss-of-function mutations of hpo or wts, including elevated transcription of cycE and diap1, increased proliferation, defective apoptosis, and tissue overgrowth. Yki is a critical target of the Wts/Lats protein kinase and a potential oncogene (Huang et al.,2005). Yki is the first substrate identified for NDR family kinases and is phosphorylated and inactivated by Wts. Biochemical studies showed that Wts-mediated phosphorylation of Yki was stimulated by upstream components of the Hippo pathway (Huang et al.,2005).

At the time of its characterization, it was speculated that like other coactivators, Yki functions by interacting with DNA binding transcription factors. YAP, the mammalian homolog of Yki, is known to function as coactivator for several transcription factors, such as the p53 family member p73 (Strano et al.,2005), the Runt family member PEBP2α (Yagi et al.,1999; Vassilev et al.,2001), and the four TEAD/TEF transcription factors (Yagi et al.,1999; Vassilev et al.,2001; Strano et al.,2005). This interaction is generally mediated by WW domains of YAP, and the PPxY motifs of the cognate transcription factors or by other binding interactions independent of the WW domain. Indeed, the Drosophila TEAD domain transcription factor, scalloped (sd), was identified as the cognate factor that partnered with the N-terminal (TEAD binding) region of Yki to regulate Hpo target genes by three different approaches: yeast two-hybrid screen (Zhang et al.,2008b), GST pull-down experiments (Goulev et al.,2008), and identification of a Sd/TEAD-binding site in the Hippo Response Element (HRE) of the diap1 gene (Wu et al.,2008; Zhang et al.,2008b).

Scalloped and the Transcriptional Regulation of Hippo Target Genes

In Drosophila, Sd belongs to a family of evolutionarily conserved proteins characterized by the presence of a TEA/ATTS DNA-binding domain (Campbell et al.,1991,1992). Sd has been extensively characterized for its role in wing development and is known to physically interact with the product of the vestigial (vg) gene, where the dimer functions as a master gene controlling wing formation (Simmonds et al.,1998; Paumard-Rigal et al., 1998; Halder et al., 1998; Halder and Carroll,2001). The Vg-Sd dimer activates the transcription of several wing-specific genes, including sd and vg themselves. Sd binds with Yki to regulate the transcription of diap1, a target gene of the Hippo pathway (Wu et al.,2008). The interaction between Yki and Sd increases sd transcriptional activity both ex vivo in Drosophila S2 cells and in vivo in Drosophila wing discs and promotes Yki nuclear localization (Wu et al.,2008; Zhang et al.,2008b). The diap1 locus harbors a minimal Sd-binding Hippo Responsive Element (HRE) that mediates transcriptional regulation by the Hippo pathway. Genetic and biochemical studies showed that sd is required for yki-induced tissue overgrowth and target gene expression, and this sd function is conserved in its mammalian homolog (Goulev et al.,2008; Wu et al.,2008; Zhang et al.,2008b; Zhao et al.,2008b; reviewed in Bandura and Edgar,2008). Contrary to Yki, Sd is not expressed in all imaginal tissues. This indicates that Yki-Sd interaction acts in a tissue-specific manner and that other Yki partners must exist that remain to be identified.

Two Mechanisms of Hippo Regulation: The Fat-Ex-Hpo and the Fat-Dachs-Warts Pathways

Three reports showed that Fat functions through a linear pathway that involves Ft, Ex, Hpo, and Wts (Fig. 2A; Bennett and Harvey,2006; Silva et al.,2006; Willecke et al.,2006). This model is based on the following observations: (1) ft mutant cells affected the stability and localization of Ex to the apical surface of the cells. (2) Overexpression of an activated form of Ft could induce Hpo and Wts phosphorylation. (3) The ability of Ft to inhibit Yki activity was reduced by knockdown of hpo or wts by means of RNA interference using a transcriptional assay of Yki activity in cell culture. (4) The most convincing evidence came from genetic epistasis experiments that showed that, as opposed to the requirement of Hpo components for Ft mediated growth regulation, activation of the Hpo pathway at the level of ex or hpo mostly bypassed the requirement for ft. This placed Ft upstream of Ex and other Hpo components. Interestingly, this interaction does not involve Mer (which is proposed to act together with Ex), suggesting that Mer may receive inputs from other upstream signals. Together, these findings suggested that Ft (possibly acting as a receptor) signaled to Ex, which in turn signaled to the kinase cascade. Warts then negatively regulated the activity of the oncoprotein Yki, which then bound its cognate transcription factor Sd to regulate the expression of diap1.

Cho et al. (2006) demonstrated that Ft promoted Wts-mediated Hippo signaling by a different mechanism that involved an unconventional myosin Dachs (Fig. 2B). In the Fat-Dachs-Wts model, the path from Ft to Wts did not go through Hpo or other upstream components of the pathway like Ex or Mer. Three lines of evidence were presented to support this conclusion. (1) The genetic analysis of double-mutant combinations placed ft upstream of dachs, and dachs upstream of both hpo and wts. (2) The levels of Wts protein, but not the levels of Hpo, Sav, Mer, or Mats were reduced in ft mutant tissue. (3) When overexpressed in tissue culture cells, Dachs bound to Wts. Thus, Ft regulated the abundance of Wts by means of Dachs, while Hpo independently regulated the enzymatic activity of Wts. In this view, Wts was the point of convergence of two distinct upstream signals. This mechanism of Fat action was also reported by Feng and Irvine, where Ft acts in parallel to Ex to signal to Wts to regulate growth by means of Hippo signaling pathway (Feng and Irvine,2007).

Of interest, the experiments described in the two different models do not exclude or contradict each other. Indeed, many of the experimental results are consistent with Ft being able to regulate Wts activity in at least two different ways and also with Wts being regulated by signals that are independent of Ft (reviewed in Hariharan,2006). The observation that the phenotype of ftex double mutants is a little more severe than that of ex mutants suggests that Ft can regulate Wts independently of Ex. Also consistent with this conclusion are the results that show that the suppression of ft mutant phenotypes by the overexpression of Hippo pathway components is typically incomplete. In addition, the observation of reduced growth and decreased transcription of Yki targets in cells doubly mutant for ft and Dachs would seem to suggest that the elevated levels of Wts in these cells can still be activated by a pathway that is independent of Ft such as Mer. In addition, the fact that Sd is expressed in a spatially defined pattern and is not expressed in all imaginal discs, suggests that other transcription factors can bind with Yki to regulate the expression of other target genes in the Hippo pathway.

RASSF1 and the Hippo Pathway

Hippo signaling can also be regulated by other Ft-independent mechanisms, for example, through the interaction of the Ras Association Factor 1 (RASSF1) with Salvador and Hippo. RASSF1 belongs to a family of proteins frequently silenced in a variety of solid tumors, mainly by promoter methylation. Mammalian RASSF proteins were known to interact with Mst1 (the Hippo homolog), and after the discovery of the Hippo pathway in flies a homology domain was discovered between Sav, RASSF, and Hippo called the SARAH (Sav-RASSF-Hippo) domain. RASSF proteins typically contain a RA domain and a C-terminal SARAH domain (Polesello et al.,2006; Guo et al.,2007; Avruch et al.,2009). The dRASSF mutant flies are viable, and under controlled conditions uncover a clear growth defect in comparison to wild-type animals. RASSF expression was reduced in clones mutant for a hpo allele that lacked the SARAH domain. Further investigations reveal that the effect on dRASSF is posttranscriptional. Of interest, dRASSF levels were unaffected in clones mutant for other Hippo-pathway members, such as ex, sav, and wts (Polesello et al.,2006; Guo et al.,2007). These results suggest that direct binding to Hpo through its SARAH domain, rather than signaling through the Hippo pathway, is necessary for dRASSF stability. RASSF1 binds to Hpo but not with Sav, and does not form a ternary complex. Hpo was able to bind Sav and dRASSF, however, in two different Hpo complexes—a highly active Sav-associated pool and an inactive dRASSF-associated pool. Thus, Sav can promote Hpo activation. Salvador and RASSF repress each other's expression, and in cell culture, increasing amounts of Sav can displace RASSF bound to Hpo, suggesting that modulation of the inhibitory effects of dRASSF and the activator effects of Sav determine the level of Hippo activation, and organ size (Polesello et al.,2006; Guo et al.,2007).

THE MAMMALIAN HIPPO PATHWAY: REGULATION OF ORGAN SIZE

Most components of the Hippo pathway identified to date have one or more mammalian orthologs that probably function in an analogous manner to their Drosophila counterparts (Table 1). The components of the Hippo pathway are highly conserved in mammals, including YAP (Yes-associated protein), Lats1/2, Mob, Mst1/2, Sav, Mer, Ex1/2, and Ft4 (Yki, Wts, Mats, Hpo, Sav, Mer, Ex, and Ft homologs, respectively). The YAP-like protein Transcriptional coactivator with PDZ binding motifs (TAZ or WWTR1), is also involved in the mammalian Hippo pathway. Like Yki, YAP and TAZ are phosphorylated and inhibited by the Hippo pathway through cytoplasmic retention (Zhao et al.,2008a, b). Although there is high structural and functional similarity between YAP and TAZ, these proteins have distinct context-dependent functions (Hong and Yaffe,2006; Lei et al.,2008).

Table 1. Genes Involved in the Drosophila and Mammalian Hippo Pathway

Drosophila melanogaster	Human orthologs	Protein	Role in Hippo Signaling Pathway	Role in Human Cancer (Reference)
Fat	Fat4	Proto-Cadherin	Hippo Pathway Receptor (?)	Mutated in breast cancer (Qi et al.,2009)
Dachsous	DCHS1	Proto-Cadherin	Ligand for the Fat receptor (?)	Unknown
Expanded	FMRD6 (Willin)/ EX2	4.1 Superfamily FERM domain protein	Upstream regulators of the Hippo kinase cascade	Unknown
Merlin	NF2	4.1 Super-family FERM domain protein	Upstream regulators of the Hippo kinase cascade	Mutated in familial and sporadic schwanomas (Evans et al.,2000)
Hippo	MST1/2	Ste-20 family Protein Kinase	Bind Salvador and Mats, and phosphorylate Mats and Warts in response to upstream signals	Hypermethylated in soft tissue sarcoma (Seidel et al.,2007)
Salvador	hWW45	WW domain Adaptor protein	Binds Hpo, the Sav-Hpo complex more efficiently regulates the activity of the Warts kinase	Mutated in cancer cell lines (Tapon et al.,2002)
Mats	MOBK1B	Mob superfamily coactivator of protein kinases	Phosphorylated by Hpo and Wts, and binds Hpo, and Wts, to regulate Wts activity	Mutated in cancer cell lines (Lai et al.,2005)
Warts	Lat1/2	NDR family Protein Kinase	Phosphorylates Mats, Wts and Yki to regulate Hippo pathway activity	Silenced in breast tumors (Turenchalk et al., 1999; Zeng and Hong,2008)
Yorkie	YAP TAZ	Transcriptional Coactivator	Unphosphorylated form binds Sd, and translocates to the nucleus to regulate expression of transcriptional targets of Hippo signaling	Amplified (overexpressed) in breast tumors, colorectal cancer and several other solid tumors (Overholtzer et al.,2006; Zender et al.,2006)
Scalloped	TEAD	Transcription factor	Binds with Yki to regulate target gene expression.	Unknown
dRASSF1	RASSF1	RA domain containing RAS effector protein	Binds Hippo, complex may be disrupted by Sav in response to pathway activation	Hypermethylated in Lung and kidney cancers
Dachs	Unknown	Unconventional Myosin	Binds with Warts	Unknown

Human YAP, Lats1, Mst2, and Mob1 can functionally rescue the respective Drosophila mutants, suggesting the functional conservation of these proteins in mammals. Mst1/2, the Hpo homologs, can phosphorylate all three-core components and thus play a key role in the mammalian Hippo pathway. Mst1 and 2 phosphorylate (a) Lats1/2 on the activation loop and hydrophobic motif (Chan et al.,2005), (b) WW45 after physically interacting (binding) with WW45 (Callus et al.,2006), and (c) Mob1, which enhances its interaction with Lats1 (Praskova et al.,2008). Not only are the components and functional interactions conserved between the core Hippo pathway components from flies to mammals, but the mechanism by which YAP and TAZ are regulated by Hippo components is also conserved. Thus, Mst, WW45, Lats, and Mob induce YAP phosphorylation, cytoplasmic translocation, and inhibition (Zhao et al.,2007; Hao et al.,2008; Lei et al.,2008; Oka et al.,2008; Zhang et al.,2008a). As in flies, TEAD family transcription factors, homologs of the Drosophila Sd, are key mediators of YAP function in mammalian cells (Zhao et al.,2008a). However, in mammalian systems YAP/TAZ are also known to function with other transcription factors, such as the p53 family member p73 (Strano et al.,2005), and the Runt family member PEBP2α. Thus, Yki may partner with other transcription factors to regulate Hpo target genes in a tissue-specific or context-dependent manner. Based on current reports the conservation of the upstream components—Ft, Ex, and Ds—remains unclear in mammalian Hippo signaling. Therefore, the conservation of the Hippo pathway is not limited to structural homologies and signaling interactions among pathway components but also to the function of Hippo signaling in the regulation of organ size (Camargo et al.,2007; Dong et al.,2007). The evidence for the functional conservation came from experiments in which overexpression of YAP in the mouse liver was shown to cause an increase in liver size and eventually to tumor formation (Dong et al.,2007).

Although mammalian Hippo signaling is very similar to that of Drosophila, recent studies have also suggested that there are some differences between mammalian and Drosophila Hippo signaling pathways. For example, the role of RASSF1 in the regulation of MST1 (Hippo) appears to be opposite between flies and mammals. In mammals, RASSF1A can increase the MST1 activity and this activation is required for Fas-induced cell death (Oh et al.,2006). These differences may be due to the spatial and temporal differences in the expression and activity of the various RASSF isoforms and MST1/2 proteins. Similarly, the level of CycE (a transcriptional target of the Hippo pathway) is not significantly elevated in WW45 mutant mice (Lee et al.,2008). Despite these dissimilarities, the core wiring diagram of the Hippo signaling pathway is conserved between flies and mammals, and further studies in both flies and mammalian systems will add to our understanding of growth and size regulation.

Roles of Hippo Signaling in Cancer, Apoptosis, Regulation of Cell Shape, and Contact Inhibition

The Hippo pathway is emerging as a network of several tumor suppressor genes and (so far) one oncogene Yki. Interestingly, YAP has been shown recently to be a candidate oncogene in the human chromosome 11q22 amplicon (Overholtzer et al.,2006; Zender et al.,2006). In addition to genomic amplification, YAP expression and nuclear localization was also shown to be elevated in multiple types of human cancers (Zender et al.,2006; Dong et al.,2007; Zhao et al.,2007; Steinhardt et al.,2008). Furthermore, mutations of Lats1/2, Sav, and Mob have been implicated in tumorigenesis (St John et al.,1999; Tapon et al.,2002; Lai et al.,2005; reviewed in Harvey and Tapon,2007). RASSF1, the constitutive binding partner of MST1/2, and MST1/2 are frequently mutated by hypermethylation in several cancer types (Seidel et al.,2007; Avruch et al.,2009). NF2/ Merlin is a well-known tumor suppressor gene in mammals (Okada et al.,2007). Fat4 has recently been shown to be inactivated/deleted in breast cancers and primary tumor cell lines (Qi et al.,2009). Despite its conservation and intimate relationship with cancer, the Hippo pathway has not been systematically studied in mammalian cells.

Activation of Hippo signaling by overexpression of the Hpo kinase or Mst1/2 has been shown to activate apoptosis in response to DNA damage or other stress and apoptotic stimuli (Graves et al.,1998; Udan et al.,2003; Colombani et al.,2006; reviewed in Hay and Guo,2003). Increased cell death results in formation of reduced/smaller organs or appendages. Although loss of Hippo signaling clearly makes cells resistant to apoptosis and promotes cell survival, the molecular mechanism by which Hippo signaling regulates apoptosis remains unclear. This is because activation of Hippo signaling does not suppress the transcriptional expression of Diap1. Of interest, a Hpo-response element has been mapped in the Diap1 gene that is induced in response to loss of Hippo signaling. Thus, even though a molecular mechanism by which Hippo signaling likely regulates the expression of Diap1 gene has been reported, clearly the regulation of cell death by activation of Hpo signaling is complex (Wu et al.,2008). Mammalian YAP has been shown to have both pro- and antiapoptotic functions (Oka et al.,2008); however, it is still possible that under certain conditions such as DNA damage, YAP can be tyrosine phosphorylated by c-Abl, which selectively activates YAP transcriptional activity on p73 to induce apoptosis (Zeng and Hong,2008).

Hippo signaling is implicated in the regulation of cell shape because the plasma-membrane protein Mer/NF2 is involved in the regulation of cell shape, motility, and adhesion (Okada et al.,2007). Most studies conclude that membrane association is necessary for the growth-suppressing function of Mer. Mutant versions of Mer that cannot localize to the membrane cannot inhibit cell proliferation. In addition, Mer can associate with several membrane proteins, including both adhesion receptors and membrane receptors that regulate cell proliferation and differentiation (Curto and McClatchey,2008). Recent studies suggest that Merlin directly controls the surface availability and function of membrane receptors, and their trafficking by endocytosis in both Drosophila and mammalian cells (Manchanda et al.,2005; Maitra et al.,2006). Ex and Mer are localized under the plasma membrane and are scaffolding proteins that provide a regulated linkage between membrane proteins and the cortical cytoskeleton, and also participate in signal-transduction pathways.

How normal cells in culture stop proliferating upon achieving confluency has long been a topic of intense investigations. This phenomenon called “contact inhibition” is a hallmark of cells growing in culture. Contact inhibition is lost in cancer cells, however, the mechanism regulating this phenomenon remains poorly understood. Recently, several components of the Hippo pathway have been implicated in contact inhibition. Mer becomes dephosphorylated and activated in confluent cells (Shaw et al.,1998; Morrison et al.,2001), which has been reported to be both necessary and sufficient for contact inhibition. Lats2 and WW45 are also related to contact inhibition as their knockout mouse embryonic fibroblast (MEF) cells show loss of contact inhibition. Finally, YAP is phosphorylated and translocated to the cytoplasm by the Hippo pathway at high cell density in a Mer-dependent manner (Zhao et al.,2007). Moreover, a dominant-negative form of YAP restores contact inhibition in ACHN, a cancer cell line with activation of YAP due to WW45 mutation (Zhao et al.,2007). So far, the reports suggest that, when cultured cells reach confluence, cell–cell interactions lead to signaling events that activate the Hippo pathway. The activation of the Hippo pathway leads to the phosphorylation of YAP and its retention in the cytoplasm where it interacts with 14-3-3 proteins and reduced transcription of YAP/Hippo target genes. This likely sends the “stop proliferation” signal to cells in culture and thus play a crucial role in contact inhibition (Zhao et al.,2007,2008a). Identifying the upstream signal of this pathway might solve a long-standing mystery in cell biology.

FUTURE PERSPECTIVES

Several questions remain unanswered in the field of Hippo signaling and function. For example, is the activity of the Hippo pathway regulated by mechanisms beside Ft/Ds? Are there other regulators of the pathway? Experimental evidence so far suggests that several additional components of the Hippo pathway (for example, the protein that links Ex or Mer to Hpo, or the receptor that acts in conjunction with Mer) remain to be discovered. Given that Hippo signaling may play additional roles in addition to its roles in regulation of cell proliferation arrest, apoptosis, cell morphology, and contact inhibition, there may be other downstream targets. Do other transcription factors bind with Yki in tissues where Sd is not expressed? In addition, several biochemical mechanisms remain unresolved. For example, how Ft regulates Wts levels and Hpo activity, and the functional consequences of either the abnormal Ex localization in ft clones or the binding of Ds to Wts are still unknown. Can we measure normal levels of Hpo signaling (in physiological conditions instead of in vitro)? Finally, are there other signals that instruct cells to cease proliferation? In the coming years, we can expect several interesting discoveries that will add to our knowledge of Hippo signaling and further our understanding of growth regulation.

Acknowledgements

The authors thank Kevin Waits (Mercer University School of Medicine, Macon, Georgia) and Jaison J. Nainaparampil (University of Dayton, Dayton, Ohio) for their help and comments. We apologize to all authors whose work could not be cited due to space limitations. A.S. is supported by Start-up support from University of Dayton, grants from Ohio Cancer Research Associates and University of Dayton Research Council. M.K.S. is supported by Strart-up Grant from Mercer University School of Medicine, a Seed Grant from Mercer University, and a MEDCEN Foundation grant.