Feedback Regulation of RTK Signaling in Development

Abstract

Precise regulation of the amplitude and duration of receptor tyrosine kinase (RTK) signaling is critical for the execution of cellular programs and behaviors. Understanding these control mechanisms has important implications for the field of developmental biology, and in recent years, the question of how augmentation or attenuation of RTK signaling via feedback loops modulates development has become of increasing interest. RTK feedback regulation is also important for human disease research; for example, germline mutations in genes that encode RTK signaling pathway components cause numerous human congenital syndromes, and somatic alterations contribute to the pathogenesis of diseases such as cancers. In this review, we survey regulators of RTK signaling that tune receptor activity and intracellular transduction cascades, with a focus on the roles of these genes in the developing embryo. We detail the diverse inhibitory mechanisms utilized by negative feedback regulators that, when lost or perturbed, lead to aberrant increases in RTK signaling. We also discuss recent biochemical and genetic insights into positive regulators of RTK signaling and how these proteins function in tandem with negative regulators to guide embryonic development.

INTRODUCTION

Receptor tyrosine kinases (RTKs) regulate virtually all aspects of embryonic development from early patterning to organogenesis (Lemmon and Schlessinger, 2010; Li and Hristova, 2006). The RTK superfamily encompasses 58 known receptors in humans which are classified into several multi-member subfamilies including, among others, fibroblast growth factor receptors (FGFRs), insulin and insulin-like growth factor receptors (IR and IGF-1R), platelet-derived growth factor receptors (PDGFRs), vascular endothelial growth factor receptors (VEGFRs), and epidermal growth factor receptors (EGFR/HER/ERBBs) (Lemmon and Schlessinger, 2010). Together, these receptors are involved in the entire spectrum of developmental processes. The intracellular signals initiated by RTK activation play pivotal roles in cell fate determination and morphogenesis, and many are highly conserved in evolution from the nematode Caenorhabditis elegans to humans (Pires-daSilva and Sommer, 2003). Furthermore, numerous diseases result from germline or somatic genetic changes that alter the activity, abundance, or cellular distribution of RTKs. Mutations in RTKs or proteins that facilitate their downstream signaling have been implicated in the onset and progression of a wide-range of diseases such as diabetes, inflammation, bone disorders, atherosclerosis, angiogenesis, and various cancers (Lemmon and Schlessinger, 2010).

RTK activation is triggered by binding of extracellular ligands, which leads to receptor oligomerization and auto-phosphorylation on tyrosine residues within the cytoplasmic domains. These phosphorylated residues create docking sites for phosphotyrosine-binding domain-containing proteins that couple RTK activation to downstream signaling pathways (Hubbard, 2004; Hubbard and Miller, 2007; Schlessinger, 2000). Interestingly, a large number of RTKs induce a similar set of downstream effectors, in particular those coupled to activation of the RAS/MAP kinase (MAPK) and phosphatidylinositide-3 kinase (PI3K)/AKT pathways (Blume-Jensen and Hunter, 2001; Ledda and Paratcha, 2007). What distinguishes the signaling outputs between distinct RTKs is often the duration and extent of pathway activation. Feedback regulators play a major role in fine-tuning these variables by attenuating or amplifying the signaling output. They can be already present and act prior to or immediately after receptor activation (early attenuators) (Haglund et al., 2003; Thien and Langdon, 2001) or can be transcriptionally induced by the pathways they eventually inhibit (late attenuators) (Table 1) (Casci and Freeman, 1999; Fiorini et al., 2002; Ghiglione et al., 1999; Golembo et al., 1996; Korsensky and Ron, 2016; Tsang and Dawid, 2004).

Table 1.

Classification of RTK signaling modulators according to their spatiotemporal feedback

Feedback	Modulator	Target/Pathway	Signaling Output	Mechanism of Action
Late, reversible	Anosmin 1	FGFR	Amplification	FGF-FGFR signaling complex assembly and stabilization
	DUSP6	FGF-MAPK	Attenuation	Dephosphorylation of ERK
	FLRT family	FGFR	Amplification	FGF-FGFR signaling complex activation
	MIG6	ErbB receptor family	Attenuation	Inhibition of ErbB receptor family dimerization
	SEF	FGF-MAPK/ERK & FGF-PI3K/AKT	Attenuation	FGF-FGFR signaling complex assembly and activation
	SPRED	FGF-RAS/RAF	Attenuation	Inhibition of RAF activation
	Sprouty family	RAS/MAPK & RAS/PI3K	Attenuation	Inhibition of downstream effector signaling
Early, reversible	PTEN	PI3K/AKT	Attenuation	Dephosphorylation of PIP3
	RKIP	RAF/MEK	Attenuation	Inhibition of RAF1 and MEK interaction
	SHP2	EGFR, FGF-MAPK/ERK, IR, & RET	Amplification	Dephosphorylation of inhibitory modulators of downstream effectors
Late, irreversible	CNPY1	FGFR	Amplification	FGFR maturation
	LRIG family	ErbB receptor family, MET, & RET	Attenuation	Receptor ubiquitination and degradation
	SOCS family	c-KIT, EGFR, FLT3, IGF-1R, & IR	Attenuation	Receptor ubiquitination and degradation
Early, irreversible	CBL family	EGFR, MET, PDGFR, & RET	Attenuation	Receptor ubiquitination and degradation
	NEDD4 & NEDD4L	ErbB receptor family, FGFR, IGF-1R, IR, NRTK1, & VEGFR	Attenuation	Receptor ubiquitination and degradation
	NRDP1	ERBB3	Attenuation	ERBB3 trafficking and ubiquitination
	SHISA2	FGFR	Attenuation	Receptor maturation and ubiquitination

Feedback regulators have also been shown to further control RTK-mediated signaling by modulating the localization of the ligands necessary for RTK activation. One of the characteristic features of RTK signaling in development is the gradation of signal output (Ashe and Briscoe, 2006). This gradation is traditionally thought to arise from a corresponding gradient in ligand concentration or receptor expression. However, it has recently been shown that the presence of regulators significantly contributes to this signal gradient. Indeed, the heparin sulfate proteoglycans (HSPGs), which serve as an amplifier of many RTK signaling pathways, most notably FGFRs, can control morphogen gradient formation by regulating the diffusion rates of ligands (Yan and Lin, 2009). Another example of such a mechanism is the action of CBL, a well-characterized negative feedback regulator of multiple RTK pathways such as EGFR, MET, and RET. This protein regulates the distribution of the Egfr ligand Gurken during Drosophila melanogaster embryogenesis by mediating endocytosis and subsequent degradation of the Egfr-Gurken complex (Chang et al., 2008). In this manner, feedback regulators are essential for not only controlling the level of signal output but also for tuning their spatiotemporal localization.

Here, we review several modulators of RTK signaling with an emphasis on those with known roles in development (Supplemental Table 1) and contributions to human congenital disorders (Figure 5). Our discussion highlights the insights gained from in vivo work in model organisms that can be used to further our biochemical understanding of RTK regulation through feedback pathways. Several other excellent reviews cover feedback regulators in more molecular detail (Avraham and Yarden, 2011; Lemmon et al., 2016; Mohapatra et al., 2013).

Figure 5. Characteristic features of craniofacial disorders associated with regulators of RTK signaling.

(A) Mild hypertelorism, epicanthic folds, broad nasal tip, full lips and café-au-lait spot on the left upper arm of a child who has Neurofibromatosis Legius syndrome with a SPRED1 mutation (Brems et al., 2007). (B) Dysmorphic features including hypertelorism, downslanting palpebral fissures, epicanthus, coarse facial features, and large, thick, low-set ears of an adolescent boy who has LEOPARD syndrome with a PTPN11 mutation (Santoro et al., 2014). (C) Characteristic craniofacial features including hypertelorism with downslanting palpebral fissures, full or ptotic upper eyelids, and low-set, posteriorly rotated ears with a thickened helix of a young boy who has Noonan syndrome with a PTPN11 mutation (Allanson et al., 2010). (D) Distinctive facial features including hypertelorism, ptosis, downslanting palpebral fissures, epicanthal folds, and low-set, posteriorly rotated ears of a woman who has Noonan syndrome-like phenotype with a C-CBL mutation (Martinelli et al., 2010). (E) Bilateral plantar creases and a flat appearing mid-face with a prominent forehead of a boy who has Macrocephaly/autism syndrome with a PTEN mutation (Butler et al., 2005). (F) Inversion of facial expression when smiling in a young girl who has Urofacial syndrome with a LRIG2 mutation (Stuart et al., 2013). Reprinted or adapted with permission.

Regulation of biosynthesis and maturation of RTKs

Recent studies have revealed that modulation of RTKs’ signaling outputs starts even prior to their arrival at the cell surface. A key aspect of this type of regulation occurs via quality control checks in the endoplasmic reticulum (ER), where newly synthesized receptors undergo folding and maturation by post-translational modification before being trafficked to the cell membrane. RTKs that are not properly synthesized, folded, or modified are degraded through a proteasome-dependent pathway. Several proteins, including the Canopy family, the Shisa family, and NRDP1, have recently been identified to regulate the strength of RTK-mediated signaling through interaction with and modification of receptors in the ER, consequently controlling the number of functional receptors at the cell surface.

The Canopy (CNPY) genes encode four putative ER-resident proteins hypothesized to be positive-feedback regulators of receptor maturation and trafficking (Do et al., 2012; Hart and Tapping, 2012; Hirate and Okamoto, 2006; Matsui et al., 2011), however, only a few studies to date have addressed their structure or function. In Danio rerio (zebrafish), cnpy1 expression is restricted to the midbrain-hindbrain boundary and can be induced by exogenous Fgf8. Knockdown of cnpy1 resulted in midbrain-hindbrain boundary defects with the appearance of an airplane “canopy” and impaired Fgf signaling in a cell-autonomous manner, indicating a positive-feedback relationship between cnpy1 and fgf8 (Hirate and Okamoto, 2006). Cnpy1 was also shown to positively regulate Fgf signaling for proper formation of Kupffer’s vesicle, which orchestrates left–right asymmetric body plan in zebrafish (Matsui et al., 2011). The closely related CNPY2 protein has been linked to FGF signaling in vitro: in mouse macrophages and human hepatocytes, FGF21 enhanced expression of Cnpy2, which resulted in stabilized expression of low-density lipoprotein receptors (Do et al., 2012). More recent work has identified CNPY2 as a HIF-1α-regulated angiogenic secreted factor that stimulates cell proliferation, migration, and angiogenesis in mouse models of cardiovascular pathologies and cancer (Guo et al., 2015b; Guo et al., 2015c; Ito et al., 2014; Taniguchi et al., 2017; Yan et al., 2016). Although these processes are known to be mediated by FGF signaling, no genetic interactions between CNPY2 and FGFs have been reported in vivo. While the involvement of Canopy proteins in FGFR signaling remains poorly understood, CNPY3 and CNPY4 were shown to regulate the toll-like receptors (TLRs), another class of single-pass transmembrane receptors. Co-expression of CNPY3 increased trafficking of exogenously expressed TLRs to the cell membrane via chaperone gp96, leading to elevated TLR-mediated signaling in vitro (Hart and Tapping, 2012). Interestingly, CNPY4 seems to exert an opposite effect and led to the downregulation of TLRs at the cell membrane, subsequently attenuating TLR-mediated signaling. Canopy proteins may play a similar regulatory role on RTKs to control trafficking, as CNPY1 directly interacts with FGFR1 and modulates the extent of mature N-linked glycosylation of the receptor (Matsui et al., 2011), however, further studies will be necessary to understand how Canopy proteins interact with and modulate RTK-mediated signaling.

In a manner potentially similar to the Canopy proteins, the nine Shisa proteins in vertebrates represent a novel class of ER-associated proteins that antagonize FGF-mediated signaling in a cell-autonomous manner by regulating receptor maturation. The founding member of the Shisa family, shisa1, was named based on its expression in the prospective head ectoderm and organizer in Xenopus laevis and in reference to a form of Japanese sculpture with a large head (Yamamoto et al., 2005). Misexpression of shisa1 in X. laevis resulted in enlarged cement glands and anterior head structures due to expansion of otx2 expression, which marks prospective forebrain and midbrain. Accordingly, morpholino knockdown of shisa1 reduced Fgf-mediated xbra expression at the mid-gastrula, and embryos exhibited small eyes and cement glands, suggesting that Shisa1 directs anterior-posterior axis formation through Fgf activity (Yamamoto et al., 2005). Subsequent studies of X. laevis, Gallus gallus (chick), and Mus musculus (mouse) showed that Shisa2 expression along the anteroposterior axis exerts negative regulatory effects on FGF signaling, suggesting that SHISA2 also plays a key role in the proper establishment of segmental patterning of the head (Supplemental Table 1) (Filipe et al., 2006; Furushima et al., 2007; Hedge and Mason, 2008; Nagano et al., 2006). Although the mechanism of action remains unclear, it has been suggested that the Shisa family members bind immature forms of receptors and utilize a conserved PY motif to interact with WW-domain-containing proteins such as the E3 ubiquitin ligase family of NEDD4 proteins, which are discussed below. In doing so, the Shisa proteins aid in bringing these proteins into proximity with immature forms of FGFR in the ER, resulting in ubiquitination of the receptor for retention and degradation (Pei and Grishin, 2012; Yamamoto et al., 2005).

Another example of a modulator of trafficking is the ubiquitin ligase NRDP1 which regulates ERBB3, the catalytically inactive (pseudokinase) member of the ErbB receptor family (Qiu and Goldberg, 2002). In zebrafish, nrdp1 is expressed in the neural crest, nervous system, and muscle during embryogenesis and significantly overlaps with expression of ERBB3, suggesting functional cooperation (Britsch et al., 1998; Lyons et al., 2005; Maddirevula et al., 2011). Knockdown of nrdp1 resulted in decreased expression of melanoblast markers and caused a significant reduction in pigmentation of embryos, a process driven by ERRB3 signaling (Maddirevula et al., 2011). As a RING finger-type ubiquitin ligase, NRDP1 regulates ERBB3 by controlling the abundance of receptor trafficked to the cell surface through constitutive ubiquitination of newly synthesized ERBB3 in the ER (Fry et al., 2011). The mechanism by which the cell regulates NRDP1 activity to fine-tune the precise level of receptor at the membrane was recently found to involve RTNA4. This member of the reticulon family of proteins, which control curvature of ER membranes, counteracts the NRDP1-dependent degradation of ERBB3 by sequestering NRDP1 into ER tubules. As a result, more ERBB3 can be trafficked to the cell surface, where it may engage growth factors and its co-receptors to initiate downstream signaling (Hatakeyama et al., 2016).

Regulation of ligand-receptor signaling complex formation

A single RTK can bind multiple different ligands, and a single ligand can bind to multiple receptors. The specificity of these interactions is primarily driven by relative ligand/receptor affinities and effective concentration of both the receptor and the ligand. While the abundance of the receptor is controlled primarily at the level of biosynthesis and internalization, the pool of available ligands can be significantly influenced by extracellular regulators. For example, HSPGs tightly bind growth factors to limit diffusion in the extracellular matrix and therefore increase their local concentration to drive paracrine signaling by FGF, EGF, MET, VEGF, and PDGF (Abramsson et al., 2007; Cecchi et al., 2012; Fager et al., 1992; Forsten and Schneider, 2005; Gengrinovitch et al., 1999; Rapraeger et al., 1991; Yayon et al., 1991). Other regulators operate intracellularly at the level of the receptor but also modulate the extent of productive ligand/receptor interactions. Recent studies have increased our understanding of such modulators and expanded our knowledge of similar types of feedback regulators beyond HSPGs. Here we discuss Anosmin 1, FLRT3, and SEF; and MIG6, which are FGFR and EGFR pathway-specific protein modulators, respectively, that interact with ligand-receptor signaling complexes to mediate assembly and activation.

The Anosmin 1 gene encodes an extracellular matrix-associated protein that is largely conserved from invertebrates to primates (de Castro et al., 2016), however, no ANOS1 ortholog has been identified in mouse and rat (de Castro et al., 2014). Therefore, the biological functions of Anosmin 1 have primarily been probed by overexpression of human Anosmin 1 in mouse and rat neurons, which led to effects on cell adhesion and migration and neurite outgrowth and branching (Bribian et al., 2008; Garcia-Gonzalez et al., 2016; Soussi-Yanicostas et al., 2002; Soussi-Yanicostas et al., 1998). In development, these processes contribute to cranial neural crest formation and several aspects of neurogenesis, as shown by in vitro and in vivo studies in C. elegans, D. melanogaster, chick, and zebrafish (Supplemental Table 1) (Di Schiavi and Andrenacci, 2013; Endo et al., 2012; Gianola et al., 2009; Murcia-Belmonte et al., 2016). Anosmin 1 enhances FGF2 signaling specifically through FGFR1 in a heparin sulfate (HS)-dependent manner (Bribian et al., 2006; Gonzalez-Martinez et al., 2004). Heparin-bound Anosmin 1 binds to a pre-formed FGF2/FGFR1 complex via extracellular FnIII domains to stabilize the complex, resulting in receptor activation (Figure 1A) (Cariboni et al., 2004; Hu et al., 2009). In both C. elegans and D. melanogaster, perturbation of the FnIII domains ablated biological activity of Anosmin 1 (Andrenacci et al., 2006; Bulow and Hobert, 2004).

Figure 1. Modulation of RTK signaling by regulation of the ligand-receptor signaling complex formation.

(A) Heparin-bound Anosmin 1 binds to a pre-formed FGF2/FGFR1 complex, promoting its assembly and resulting receptor activation. FGFR signaling induces expression of FLRT3 and SEF via transcriptional activation and translation. (B) FLRT3 complexes with FGFR to promote downstream signaling of the MAPK/ERK pathway via its intracellular domain. (C) SEF complexes with FGFR and blocks receptor phosphorylation and activation of the RAS/MAPK and PI3K/AKT signaling cascades. (D) EGFR signaling induces MIG6 expression via transcriptional activation and translation. MIG6 accumulates in the cytoplasm where it binds directly with the ligand-activated ErbB kinase domain to inhibit auto-phosphorylation. This interaction can direct trafficking of the MIG-bound EGFR from the plasma membrane to late endosomes, targeting the receptor for lysosomal degradation. Dashed lines connecting the human congenital disorder with the protein in the pathway encoded by the causative mutated gene. Syndromes noted in the text and/or Supplemental Table 1. HH, hypogonadotropic hypogonadism with or without anosmia; KS, Kallmann syndrome; FGFR, Fibroblast Growth Factor Receptor; GF, growth factor; HS, heparin-sulfate; P, phosphorylation; RTK, receptor tyrosine kinase.

The three Fibronectin-like domain-containing Leucine-rich Transmembrane (FLRT) genes encode a highly conserved family of glycosylated proteins that mediate cell recognition and FGF signaling in vertebrates in a manner that is distinct from HSPGs and Anosmin 1. Flrt3 was originally identified in X. laevis as a gene with a similar expression pattern to Fgf signaling molecules, particularly at the midbrain/hindbrain boundary (Bottcher et al., 2004). Gain- and loss-of-function of flrt3 or flrt2 phenocopy experiments that perturb Fgf signaling, including effects on gastrulation, microcephaly, anterior truncations, and induction of ectopic tail-like structures (Bottcher et al., 2004; Cho et al., 2013). In chick, flrt3 is necessary but not sufficient for proper formation of the limb organizer called the apical epidermal ridge (AER) and co-localized with fgf8 expression and Erk activity (Tomas et al., 2011). Flrt3 knockout mice are embryonic lethal due to fusion defects and impaired definitive endoderm migration, phenotypes attributed to FLRT3’s function as a cell-adhesion molecule (Egea et al., 2008; Karaulanov et al., 2006; Maretto et al., 2008; Tsuji et al., 2004). X. laevis biochemical analyses in vivo and in vitro revealed that FLRT proteins complex with FGFRs to promote downstream signaling of the MAPK/ERK pathway via their intracellular domain (Figure 1B) (Bottcher et al., 2004). Although rodent FLRT3 similarly physically interacts with FGFR1 (Haines et al., 2006), deletion of Flrt3 in mice had no effect on Fgf8 expression or the expression of known Fgf targets, despite expression of Flrt3 in well-known Fgf signaling centers such as the AER, the midbrain-hindbrain boundary, and the anterior visceral endoderm (Egea et al., 2008; Haines et al., 2006; Maretto et al., 2008). Taken together, these studies suggest that the degree of conservation of the FGF/FLRT3 positive feedback loop varies among species (Supplemental Table 1). Since Flrt3 null mice die at early stages of development, it will be worthwhile to investigate whether FLRT3 modulates FGF signaling at later stages using conditional knockout mice.

Similar to flrt3 in X. laevis, sef (similar expression to fgf gene) was originally identified in zebrafish as a gene whose expression domains overlapped with known signaling centers of Fgfs (Furthauer et al., 2002; Tsang et al., 2002). Loss- or gain-of-function of sef in zebrafish led to various developmental defects, including cephalic malformations, cyclopia, expansion of ventrally derived domains, and reduction of the dorsal-most mesoderm (Furthauer et al., 2002; Tsang et al., 2002). In X. laevis, misexpression of zebrafish sef in the ventral marginal region at the 4-cell stage resulted in posterior truncations and gastrulation defects and was accompanied by suppression of Fgf target genes (Tsang et al., 2002). SEF transcripts have since been detected in zebrafish, chick, and mouse in numerous structures, including somites, the developing brain, limbs, and fin buds (Supplemental Table 1) (Boros et al., 2006; Furthauer et al., 2002; Harduf et al., 2005; Lin et al., 2002; Tsang et al., 2002). Surprisingly, Sef null mice are viable and fertile and do not show any obvious morphological phenotype during embryonic development (Abraira et al., 2007; Lin et al., 2005; Mellett et al., 2015). Lack of severe defects in the Sef mutant mice may be due to compensatory effects by other feedback antagonists. Indeed, Sef and the similarly FGF-induced Sprouty genes, discussed below, are expressed in overlapping regions along the anterior–posterior axis of the mouse embryo (Furthauer et al., 2002; Lin et al., 2002; Minowada et al., 1999).

The prototypic sef in zebrafish encodes a transmembrane receptor-like glycoprotein that blocks phosphorylation of Fgfr and subsequent activation of the Ras/Mapk and PI3K/Akt signaling cascades (Figure 1C) (Furthauer et al., 2002; Harduf et al., 2005; Kovalenko et al., 2006; Kovalenko et al., 2003; Preger et al., 2004; Tsang et al., 2002; Xiong et al., 2003; Yang et al., 2003). In vitro studies with mammalian SEF not only replicated the FGFR-induced antagonism seen in other species but also revealed that SEF can inhibit signaling activated by other growth factors, including EGF, PDGF, and nerve growth factor (NGF) (Kovalenko et al., 2003; Preger et al., 2004; Ren et al., 2008; Torii et al., 2004; Ziv et al., 2006). Interestingly, alternative spliced isoforms of SEF have been identified in humans (Preger et al., 2004; Rong et al., 2007; Ziv et al., 2006). SEF-a is similar to the prototypic SEF reported in zebrafish and mice (Furthauer et al., 2002; Lin et al., 2002; Tsang et al., 2002; Yang et al., 2003), whereas SEF-b, which lacks a signal peptide for secretion, is localized to the cytoplasm and acts at the level of, or downstream from, MEK (Figure 2F) (Preger et al., 2004; Yang et al., 2003; Ziv et al., 2006). Although both isoforms interact with FGFR1, the outcome of this association is not identical, as the cell-surface SEF-a inhibits multiple FGF signaling pathways (Preger et al., 2004). Whether these isoforms function cooperatively or in the same developmental processes remains to be determined; to note, the SEF-b isoform exhibits a restricted pattern of expression in human tissues compared with SEF-a (Preger et al., 2004). Since RTKs deliver varied biological responses, it seems likely that SEF can interfere with RTK signaling at different levels to fine-tune signaling in a cell context- and isoform specific-manner.

Figure 2. Feedback modulators of intracellular signal transduction cascades.

FGFR signaling induces expression of the SPRED family, the Sprouty family, SEF, and nuclear DUSPs via transcriptional activation and translation to attenuate RAS/MAPK signaling. (A) Growth factor-activated RTKs induce GRB2-mediated recruitment of SHP2 to signaling complexes. GRB2 redirects activated SHP2 to other signaling proteins that normally inhibit RTK signaling, subsequently acting as a positive regulator. (B) Growth factor-activated RTKs recruit and activate PI3K. The PI3K lipid signaling intermediate is dephosphorylated by PTEN, thereby attenuating PI3K/AKT signaling. (C) SPRED proteins increase RAF recruitment to the plasma membrane and prolongs RAS/RAF complexation, withdrawing RAF from activation by phosphorylation. (D, D’) Sprouty proteins translocate to the plasma membrane where they are phosphorylated. This phosphorylation induces a confirmation change that allows Sprouty proteins to bind and disrupt the GRB2/SOS complex, RAS activation, and RAF activation, thereby attenuating RAS/PI3K and RAS/MAPK signaling. (E) RKIP binds to both RAF1 and MEK to prevent their physical interaction and MEK phosphorylation, thereby attenuating RAS/MAPK signaling. (F) SEF-b suppresses activation at the level of, or downstream from, MEK. (G) DUSP6 dephosphorylates ERK. Dashed lines connecting the human congenital disorder with the protein in the pathway encoded by the causative mutated gene. Syndromes noted in the text and/or Supplemental Table 1. BRRS, Bannayan-Ruvalcaba-Riley syndrome; CFC, cardio-facio-cutaneous syndrome; CS, Costello syndrome; CWS1, Cowden syndrome 1; HH, hypogonadotropic hypogonadism with or without anosmia; LDD, Lhermitte-Duclos disease; LPRD1, LEOPARD syndrome 1; MAS, Macrocephaly/autism syndrome; MC, Metachondromatosis; NFLS, Neurofibromatosis Legius syndrome; NS, Noonan syndrome; PS-like, Proteus-like syndrome; VACTERL, vertebral anomalies, anal atresia, congenital cardiac disease, tracheoesophageal fistula, renal anomalies, radial dysplasia, and other limb defects; GF, growth factor; P, phosphorylation; RTK, receptor tyrosine kinase.

Through an evolutionarily conserved modular domain named the ErbB binding region (EBR), the multi-adaptor protein Mitogen-Inducible Gene 6 (MIG6) mediates catalytic repression of ligand-bound ERBB receptors, namely EGFR, ERBB2, and ERBB4 (Anastasi et al., 2007; Hackel et al., 2001). Since ERBB3 signals as an obligate heterodimer with the other members of the ErbB family, MIG6 also inhibits its signaling and thus is a cellular inhibitor of the entire ErbB family. Knockout of Mig6 in mice resulted in aberrant lung development associated with high neonatal mortality (Ferby et al., 2006; Jin et al., 2009; Zhang et al., 2005), and surviving mice developed degenerative joint diseases and spontaneous tumors in organs including the skin, gastrointestinal tract, lung, and endometrium (Supplemental Table 1) (Ferby et al., 2006; Jeong et al., 2009; Jin et al., 2009; Zhang et al., 2005). Importantly, over proliferation and impaired differentiation of epidermal keratinocytes and the resulting skin tumors could be rescued by genetic or pharmacological suppression of EGFR, indicating that unrestrained EGFR activation and sustained signaling through MAPK was a result of loss of Mig6 (Ferby et al., 2006). Tissue-specific deletion of Mig6 in mouse hepatocytes caused hepatomegaly and fatty liver, a phenotype similar to that observed in mice homozygous for a gain-of-function Egfr allele (Ku et al., 2012; Natarajan et al., 2007; Reschke et al., 2010; Scheving et al., 2014). Receptor-induced phosphorylation of the MIG6 ERB domain stabilizes the MIG6/EGFR interaction and prevents activation of EGFR by blocking an allosteric site critical for activation within the receptor dimers (Figure 1D) (Park et al., 2015; Zhang et al., 2007; Zhang et al., 2005). Upon docking onto EGFR, MIG6 is also capable of recruiting components of the endocytic machinery, leading to receptor degradation independent of phosphorylation and ubiquitination (Frosi et al., 2010; Segatto et al., 2011; Walsh and Lazzara, 2014; Ying et al., 2010). This two-tiered mechanism of MIG6-mediated inhibition provides for immediate repression of EGFR signaling (kinase inhibition) coupled to longer term isolation from incoming ErbB receptor ligands (endocytosis) (Anastasi et al., 2016). Whether MIG6 exerts either of these inhibitory functions on other RTKs remains to be determined. MIG6 binding to ErbB receptors is dependent on a protein interface in the kinase domain unique to the ErbB family, so involvement of MIG6 with other RTKs would involve a different mechanism or could imply that these RTKs signal in cooperation with ErbB receptors. In vitro analyses suggest that MET could be a potential target of MIG6, as overexpression of MIG6 was able to inhibit the HGF/MET-induced cell migration and neurite outgrowth (Pante et al., 2005).

Receptor dephosphorylation

The phosphorylation status and signaling output of RTKs is determined by a balance between the intrinsic kinase activity of the receptor and the activities of protein tyrosine phosphatases (PTPs). PTPs have evolved in a number of families that are structurally and mechanistically distinct and control a broad spectrum of RTK signaling pathways (Ostman and Bohmer, 2001; Tonks, 2006). As such, they are arguably one of the most important regulators of the extent and intensity of RTK signaling. Animal studies thus far, however, have yielded only limited insights into specific functions of individual PTPs. While knockout mouse models have been made for all classical PTP genes except Ptpn18, Ptpn20, Ptpn21, and Ptpru (Hendriks et al., 2013), many of these knockout models displayed only mild developmental defects, suggesting significant functional redundancy between PTPs.

Ptpn11 is one of the few exceptions, as Ptpn11 knockout mice died at mid-gestation with multiple defects in mesodermal patterning (Qu et al., 1997; Saxton et al., 1997). Selective deletion of Ptpn11 in developing kidneys of mice caused reduced ureteric bud branching by downregulation of the transcription factors Etv4 and Etv5, which are targets of glial-derived neurotrophic factor (GDNF)/RET signaling and of other RTKs (Willecke et al., 2011). Ptpn11-deficiency in cardiomyocytes resulted in early postnatal lethality and dilated cardiomyopathy associated with increased IR signaling and decreased activation of ERK1/2 and JNK2 (Princen et al., 2009). Numerous other in vivo studies have linked Ptnp11 with FGF-dependent MAPK/ERK signaling and have revealed roles for Ptnp11 in patterning and specification of the optic vesicle, lens and lacrimal gland development, chondrogenesis, intestinal progenitor cell fate, lung branching morphogenesis, and formation of the midbrain-hindbrain boundary, among others (Supplemental Table 1) (Cai et al., 2013; Dee et al., 2016; Heuberger et al., 2014; Pan et al., 2010; Tefft et al., 2005; Yang et al., 2013). PTPN11 encodes the widely expressed non-receptor tyrosine phosphatase Src-homology 2 domain-containing phosphatase 2 (SHP2) (Dance et al., 2008), which in the absence of upstream stimulation, is kept in a low-activity state by an intramolecular interaction between the N-terminal SH2-domain and the catalytic phosphatase domain. Activation of RTKs and/or subsequent activation of scaffolding adaptor proteins leads to recruitment of SHP2 to signaling complexes, where engagement of the SH2-domains induces a conformational change that resolves auto-inhibitory interactions. SHP2-mediated dephosphorylation of FGFRs is controlled by the adaptor protein GRB2, which recruits SHP2 to the activated receptors (Figure 2A). GRB2 additionally redirects activated SHP2 to other signaling proteins, such as Sprouty or STAT proteins, that normally inhibit signaling through ERK1/2, AKT, or STAT5. In this manner, SHP2 can further promote RTK signaling (Ahmed et al., 2010; Ahmed et al., 2013; Hadari et al., 1998; Hanafusa et al., 2004; Tajan et al., 2015; You et al., 1999).

Dephosphorylation of signaling pathway components

The intracellular events downstream of activated RTKs are responsible for transduction and amplification of ligand-induced signaling. This most commonly involves protein phosphorylation, which ultimately results in changes in gene expression and other cellular effects. The PI3K/AKT and RAS/MAPK pathways are principal signaling mechanisms for controlling cell survival, proliferation, differentiation, and migration (Figure 2) (Mendoza et al., 2011), and as such, must be precisely spatially and temporally regulated. Phosphatase and Tensin homolog (PTEN) and Dual-Specificity Phosphatases (DUSPs) represent early and late attenuators, respectively, of RTK-induced intracellular signal transduction cascades. PTEN is the main negative regulator of the PI3K/AKT pathway, whereas DUSPs modulate activation of the RAS/MAPK pathway (Carracedo et al., 2008; Katz et al., 2011) (Figure 2). Multiple mechanisms and modes of crosstalk have been uncovered between these two pathways, further complicating our understanding of their complex roles in development (Mendoza et al., 2011).

The first, and probably still the clearest, indication that PTEN plays an essential role in regulation of cell growth came from early studies in D. melanogaster (Goberdhan and Wilson, 2003). Pten-deficient cells proliferated at a faster rate than their heterozygous counterparts, showed an autonomous increase in cell size, and formed enlarged organs (Gao et al., 2000; Goberdhan et al., 1999; Huang et al., 1999). In vitro and in vivo studies revealed that PTEN controls cell growth and proliferation by antagonizing growth factor-induced activation of the PI3K/AKT pathway. Specifically, PTEN preferentially dephosphorylates membrane-bound PIP₃ into PIP₂. This prevents PIP₃-mediated recruitment of AKT to the plasma membrane and its activation (Figure 2B) (Engelman et al., 2006; Maehama and Dixon, 1998). Numerous subsequent studies in mice and other model organisms have examined the functional role of Pten in various organs and tissues, yielding a diverse spectrum of phenotypes (Supplemental Table 1) (Knobbe et al., 2008). Knockout mouse models of Pten showed that deletion of a single allele resulted in lethal polyclonal autoimmune disorders and various forms of epithelial cancer (Di Cristofano et al., 1999; Di Cristofano et al., 1998). Because of the lethal nature of Pten loss, conditional deletion models have been used to address the roles of PTEN during development. Tissue-specific deletion of Pten in mouse neurons resulted in progressive macrocephaly, seizures, and ataxia, and neurons lacking Pten expressed high levels of phosphorylated AKT (Backman et al., 2001; Groszer et al., 2001; Kwon et al., 2001). In vitro and in vivo analyses revealed that PTEN also regulates cardiac hypertrophy and survival by blocking growth factor signaling through the PI3K/AKT pathway (Crackower et al., 2002; Schwartzbauer and Robbins, 2001). As PI3K pathway signaling is regulated in part by IR signaling and affects downstream proteins involved in metabolism such as mTOR, Pten-deficiency in hepatocytes led to massive hepatomegaly and steatohepatitis with triglyceride accumulation (Horie et al., 2004; Stiles et al., 2004).

DUSPs constitute a large heterogeneous subgroup of the PTP superfamily characterized by their ability to dephosphorylate tyrosine, serine, and threonine residues. Despite a fairly detailed understanding of the biochemical properties and catalytic mechanisms employed by DUSPs (Farooq and Zhou, 2004; Owens and Keyse, 2007), knowledge of their in vivo roles is still expanding (Supplemental Table 1). Of interest here, gene knockdown or overexpression studies in zebrafish, chick, and mouse first identified an in vivo role for DUSP6 in fin/limb bud patterning as a negative feedback regulator of the FGF-RAS/MAPK signaling pathway (Figure 2G) (Kawakami et al., 2003). FGF8 signaling induces expression of DUSP6, which encodes an ERK-specific DUSP, and establishes a negative feedback loop (Bermudez et al., 2010; Groom et al., 1996; Kawakami et al., 2003; Mourey et al., 1996; Muda et al., 1996). Targeted inactivation of Dusp6 in mice led to increased levels of phosphorylated ERK, the phosphorylated ERK target Erm, and transcripts initiated from the Dusp6 promoter itself (Li et al., 2007). Furthermore, Dusp6 knockout mice displayed cardiac hypertrophy and multiple skeletal abnormalities including dwarfism, defects in the middle ear bones and otic capsule, and premature fusion of the cranial sutures (craniosynostosis); histological analysis of the long bones revealed disorganization of chondrocytes in the growth plate (Li et al., 2007; Maillet et al., 2008; Urness et al., 2008). These same skeletal phenotypes are also found in mouse models of human disorders with constitutive activating mutations in FGFRs (Neben and Merrill, 2015; Ornitz and Marie, 2015), highlighting the relationship between FGF signaling and DUSP6. Although many agonists in addition to FGFs activate ERK1/2 during embryonic development, including EGF, NGF, HGF, VEGF, and PDGF, few studies have examined their regulation by DUSP6 in vivo (Bermudez et al., 2010).

Non-catalytic feedback modulators of signaling pathways

Sprouty (SPRY) and SPRED genes encode highly conserved protein families with no apparent enzymatic function that inhibit different steps of the RAS/MAPK signaling pathway and fine-tune RTK signaling in a cell context- and isoform specific-manner. Some evidence suggests that Sprouty proteins indirectly regulate the PI3K/AKT pathway, however, these effects are poorly understood (Castellano and Downward, 2011; Steelman et al., 2011). As late attenuators transcriptionally induced by growth factor activation, Sprouty and SPRED proteins adapt cells to longer term external stimulation, persisting on the timescale of hours (Volinsky and Kholodenko, 2013).

The first member of the Sprouty family identified was found in a screen for genes involved in development of trachea and eyes in Drosophila (Casci et al., 1999; Hacohen et al., 1998). Like Drosophila Sprouty, mammalian Sprouty proteins antagonize FGF signaling in tubular morphogenesis associated with tracheal/lung development (Figure 4A, A’) (Metzger et al., 2008; Shaw et al., 2007; Tefft et al., 1999) and angiogenesis (Taniguchi et al., 2007a; Taniguchi et al., 2009). Since these initial findings, the number of pathways and biological processes regulated by Sprouty proteins continues to expand, including submandibular parasympathetic gangliogenesis (Figure 4B-B’) (Knosp et al., 2015), ureteric branching (Figure 4C, C’) (Basson et al., 2005; Basson et al., 2006; Chi et al., 2004; Gross et al., 2003; Michos et al., 2010), external genitalia development (Figure 4D, D’) (Ching et al., 2014), endochondral bone formation (Figure 4E, E’) (Joo et al., 2016; Minowada et al., 1999), and branchial nerve development (Figure 4F-F”) (Simrick et al., 2011), among others (Supplemental Table 1).

Figure 4. Mouse models are invaluable in decoding the developmental roles of the regulator family of Sprouty proteins.

(A-A’) E12.5 Spry2^−/− mouse lung showing the normal ventral secondary branch (V1) and an ectopic branch (V*) that forms earlier and proximal to V1 (Metzger et al., 2008). (B-B’) Genetic deletion of Spry1;Spry2 in mice disrupts submandibular gland epithelial development resulting in a wide primary duct (white lines) and abnormal branching morphogenesis at E13 (Knosp et al., 2015). (C-C’) Kidneys and urogenital tract abnormalities in Spry1^−/− newborn pups. Normal ureters and abnormal hydroureters are indicated by red and yellow arrows, respectively. Ad, adrenal; Ki, kidney; Ut, uterus (Basson et al., 2005). (D-D’) Fusion of the preputial (Pre) and labioscrotal (LS) folds along the ventral surface of the genital tubercle (GT) is disrupted in E16.5 male Spry1^−/−;Spry2^−/− mice, resulting in the absence of an internalized urethra in the proximal GT (red arrow) (Ching et al., 2014). (E-E’) Von Kossa/Safranin-O staining of E18.5 femur sections showed more proliferating chondrocytes in the growth plate of Spry2^−/− mice than in that of wild type (Joo et al., 2016). (F-F”) E10.5 Spry1^−/−;Spry2^−/− mice have trigeminal defects, facial nerve defects, and glossopharyngeal and vagus cranial nerves display incomplete or irregular bridging between proximal and distal ganglia. Arrows highlight abnormal morphology, and asterisks indicate missing portions (Simrick et al., 2011). (G-G’) The region of the P0 Spry2^−/− mouse cochlea shown has four rows of outer hair cells (OHCs) instead of the three found in control and elsewhere in the Spry2 null organ of Corti. PC, pillar cells; IHC, inner hair cells (Shim et al., 2005). (H-H’) Increased FGF signaling in Spry1^−/−;Spry2^−/− mice resulted in disorganized and compacted rugae at P0 (Economou et al., 2012). (I-I’) Gross appearance of wild type and Spry2^−/−;Spry4^−/− at E12.5. The arrow and arrowhead indicate hemorrhage and edema, respectively (Taniguchi et al., 2009). (J-J’) Abnormal length and thickness of adult Spry2^−/−;Spry4^+/− incisor as well as the absence of a sharp tip (asterisk) (Klein et al., 2008). Reprinted or adapted with permission.

As the functions of Sprouty proteins in embryonic development have been reviewed previously by others (Cabrita and Christofori, 2008; Horowitz and Simons, 2008; Warburton et al., 2008), we highlight here the specific roles of these proteins in craniofacial and tooth development as an example of the types of effects these genes can have on RTK-mediated signaling. SPRY2 and FGF8/FGFR3 signaling is required for cell fate decisions in the mouse auditory sensory epithelium, as loss of Spry2 resulted in dramatic perturbations in organ of Corti cytoarchitecture (Figure 4G, G’) (Shim et al., 2005). Combined deletion of Spry1 and Spry2 in mice caused highly disorganized palatal rugae including broader and ectopic ruga formation (Figure 4H, H’), indicating that the FGF pathway is activatory in a Turing-type reaction-diffusion system for the striped pattern that establishes and maintains the palatal rugae (Economou et al., 2012). Spry2;Spry4 double knockout mice are embryonic lethal by E12.5 with craniofacial and limb morphogenesis abnormalities (Figure 4I, I’) (Taniguchi et al., 2007a; Taniguchi et al., 2009). The Spry4 loss of function phenotypes, including dwarfism and polysyndactyly, resemble mouse models of human disorders with activating mutations in FGFRs (Neben and Merrill, 2015; Ornitz and Marie, 2015; Taniguchi et al., 2007a), suggesting that loss of Spry4 results in hyperactivation of FGF signaling. Mice carrying single and various combinations of Sprouty mutant alleles also possess supernumerary teeth and display abnormalities in tooth size, shape, and micro-structure (Boran et al., 2009; Charles et al., 2011; Klein et al., 2008; Klein et al., 2006; Lagronova-Churava et al., 2013; Lochovska et al., 2015; Marangoni et al., 2015; Percival et al., 2017). For example, Spry2^+/−;Spry4^−/− mice develop a ‘tusk’-like incisor in their lower jaws due to the presence of enamel on the lingual surface (Figure 4J, J’) (Boran et al., 2009; Klein et al., 2008). Importantly, the lingual ameloblast phenotype can be rescued in the adult by reducing Fgf gene dosage (Klein et al., 2008), demonstrating the critical role of Sprouty genes in controlling the epithelial-mesenchymal FGF signaling loop.

The four mammalian orthologues of Sprouty proteins share sequence similarity to D. melanogaster Sprouty in the cysteine-rich C-terminus but significantly differ among each other and from the fly ortholog in the N-terminus (de Maximy et al., 1999; Leeksma et al., 2002; Mason et al., 2006). This sequence divergence at the N-terminus could dictate differential functions, potentially by mediating protein-protein interactions. Indeed, Sprouty proteins can interact directly with multiple downstream components of the RTK pathway, including FRS2, GRB2, RAF1, B-RAF, and SHP2. In most cases, however, it remains unclear how these associations modulate signaling. The best studied family members, SPRY1 and SPRY2, have been shown to antagonize RTK signaling at multiple levels, such as binding to the GRB2/SOS complex (Figure 2D) and inhibition of RAF1 activation by RAS (Figure 2D’) depending on the cellular context and/or the identity of the RTK (Mason et al., 2006). The phosphorylation of SPRY1 and SPRY2 at Tyrosine 53 and Tyrosine 55, respectively, induces a conformational change that has been shown to be essential for protein binding and modulation of RAS/MAPK signaling (Alsina et al., 2012; Guy et al., 2009; Hanafusa et al., 2002; Mason et al., 2004; Sasaki et al., 2003; Sasaki et al., 2001). Sprouty proteins may mediate their actions in part by increasing active forms of such phosphatases as PTEN (Edwin et al., 2006; Patel et al., 2013). In cultured cells, SPRY2 increased overall PTEN protein levels while decreasing PTEN phosphorylation, resulting in increased PTEN activity. This was reflected by diminished activation of AKT by EGF signaling and blocked cell proliferation (Edwin et al., 2006). In the context of EGFR signaling, SPRY2 levels are controlled through phosphorylation-dependent complex formation with C-CBL in vitro (Hall et al., 2003; Mason et al., 2004; Rubin et al., 2003). Binding of SPRY2 to C-CBL directs the proteolytic degradation of SPRY2 but also inhibits C-CBL-mediated degradation of EGFR, leading to sustained signaling activity (Egan et al., 2002; Ng et al., 2008; Rubin et al., 2003; Wong et al., 2002). This function may be limited to SPRY2, however, as SPRY4 suppression of MAPK/ERK activation by EGF stimulation did not result in interaction with C-CBL (Mason et al., 2004; Wong et al., 2001).

SPREDs (Sprouty-related PRoteins with an EVH1 Domain) are a family of membrane-associated, negative RAS/MAPK signaling modulators that possess structural and functional similarities to their relatives, the Sprouty proteins. There are four known mammalian SPRED proteins: SPRED1, SPRED2, SPRED3, and EVE-3, the last of which is a splice variant of SPRED3 (Kato et al., 2003; King et al., 2006; Wakioka et al., 2001). Spred1 knockout mice are viable and fertile but exhibit low body weight, a shortened face, and impaired hippocampus-dependent learning capabilities (Brems et al., 2007; Denayer et al., 2008; Inoue et al., 2005; Phoenix and Temple, 2010). Spred2 deficiency in mice suppressed aorta-gonad-mesonephros hematopoiesis and caused defects in bone morphogenesis, with the mice exhibiting a dwarfing phenotype and increase of early hematopoiesis (Bundschu et al., 2005; Nobuhisa et al., 2004). Overlapping expression patterns of different SPRED family members and their possible redundancy might preclude certain phenotypes from being observed in the single null alleles (Supplemental Table 1). Indeed, deletion of both Spred1 and Spred2 in mice, which have overlapping expression patterns in the heart, lung, liver, and bone, resulted in embryonic lethality with subcutaneous hemorrhage, edema, and dilated lymphatic vessels (Engelhardt et al., 2004; Kato et al., 2003; Stowe et al., 2012; Taniguchi et al., 2007b; Tuduce et al., 2010).

Like Sprouty proteins, SPRED proteins inhibit growth factor-mediated MAPK/ERK activation, albeit by different biochemical mechanisms. Overexpression of SPRED1 increases RAF recruitment to the plasma membrane and prolongs RAS/RAF interaction, thus withdrawing RAF from activation by phosphorylation (Figure 2C) (Bundschu et al., 2005; Wakioka et al., 2001). Subsequent studies confirmed that SPRED proteins also inhibit activation of RAS by the small GTPase RAP1 without affecting receptor phosphorylation (King et al., 2006; Nonami et al., 2005; Stowe et al., 2012). Recently, it was proposed that SPRED1-plasma membrane translocation is mediated in a B-RAF- and C-RAF-dependent manner to specifically disturb K-RAS but not H-RAS membrane anchorage (Siljamaki and Abankwa, 2016). This potential mechanism may explain why it has been difficult to pinpoint whether SPRED1 acts at the level of RAS or RAF.

Originally isolated from the bovine brain (Bernier and Jolles, 1984), RAF Kinase Inhibitor Protein (RKIP; also known as PhosphatidylEthanolamine-Binding Protein, PEBP1) was renamed based on its physiologically relevant inhibition of the RAS/MEK/ERK pathway (Yeung et al., 1999; Yeung et al., 2001). Expression of RKIP mRNA has since been detected in all mammalian tissues tested, with high levels in spermatids and brain oligodendrocytes, Purkinje cells, and specific cortical and hippocampal neuronal cell layers (Bernier and Jolles, 1984; Frayne et al., 1999; Theroux et al., 2007). RKIP deficient mice are viable but develop an olfaction deficit, a phenotype that correlates with the expression pattern of the gene in the brain (Theroux et al., 2007). Subsequent studies in model organisms have identified RKIP as critical for neurological functioning, photoreceptor degeneration, myogenesis, reproduction, and spermatogenesis (Supplemental Table 1) (Antoun et al., 2012; Gibbons et al., 2005; Murga-Zamalloa et al., 2011; Nixon et al., 2006; Subramanian et al., 2014; Yamamoto et al., 2012). RKIP inhibits RAF-1 mediated phosphorylation and activation of MEK by competitive physical association which disrupts the interaction between these kinases (Figure 2E). Overexpression of RKIP in vitro reduced cell proliferation and transformation and was accompanied by alterations in MEK-, ERK-, and AP-dependent transcription (Yeung et al., 1999). Interestingly, although RKIP can interact with B-RAF, depletion of RKIP did not affect B-RAF activation, indicating that RKIP may selectively limit the dynamic range of MAPK signaling in response to growth factors (Trakul et al., 2005).

Early attenuation of RTK signaling via receptor ubiquitination and degradation

Another common mechanism by which RTK signaling is downregulated is the removal of receptors from the plasma membrane via endocytosis. This can occur either reversibly, when internalized receptors are recycled back to the plasma membrane after a period of time, or irreversibly, when the downregulated receptors are sent for lysosomal degradation. RTK internalization and degradation are regulated upon growth factor induced RTK activation through ubiquitination of the intracellular receptor domains by E3 ubiquitin ligases (Figure 3) (Goh and Sorkin, 2013). The two main E3 ligases involved in RTK ubiquitination during development are the HECT-type ligase NEDD4 and the RING-type ligase CBL. Both NEDD4 and CBL regulate signal duration of multiple RTKs, and their disruption in development results in serious abnormalities.

Figure 3. Attenuation of RTK signaling by receptor ubiquitination and degradation.

(A) Growth factor activation of RTKs leads to recruitment of NEDD4 to the receptor complex. (B) Independent and (C) SOCS- or (D) LRIG-mediated mechanisms recruit CBL to the receptor complex. NEDD4 and CBL direct the ubiquitination of RTKs, resulting in receptor endocytosis and routing to early endosomes. RTKs can then either be recycled to the plasma membrane or targeted for lysosomal degradation, thereby attenuating receptor signaling. Dashed lines connecting the human congenital disorder with the protein in the pathway encoded by the causative mutated gene. Syndromes described in the text and/or Supplemental Table 1. PVNH7, Periventricular nodular heterotopia 7; NS-like, Noonan syndrome-like; UFS, Urofacial syndrome; GF, growth factor; P, phosphorylation; RTK, receptor tyrosine kinase; U, ubiquitination.

NEDD4 (Neuronal precursor cell Expressed and Developmentally Downregulated) proteins are found ubiquitously in eukaryotes and have expanded to nine known family members in mammals, with two of them, NEDD4 and NEDD4L (also known as NEDD4-2) being very closely related (Supplemental Table 1) (Scheffner and Kumar, 2014). Complete loss of Nedd4 in mice resulted in embryonic lethality at mid-gestation with pronounced heart defects, subcutaneous bleeding, and developmental delays (Fouladkou et al., 2010; Kawabe et al., 2010; Liu et al., 2009). Although NEDD4 has several additional substrates beyond RTKs, biochemical analysis suggested that the Nedd4 loss-of-function phenotype can be at least partially attributed to abnormal RTK signaling: the growth retardation in Nedd4 heterozygous mice is associated with reduced cell surface expression and signaling through IR and IGF-1R (Cao et al., 2008). That loss of Nedd4 resulted in loss of IGF-1R signaling contradicts earlier in vitro studies which reported that NEDD4 ubiquitinates and decreases stability of IGF-1R (Vecchione et al., 2003). This suggests that NEDD4 may fine-tune RTK signaling differently in distinct cell types. In agreement with a role for NEDD4 in promoting receptor degradation, Nedd4L-deficient mouse embryos showed increased expression of neurotrophic RTK 1 (NRTK1, also known as TRKA), a possible contributor to the pain sensitivity phenotype in heterozygous adults (Yanpallewar et al., 2016). The binding of NEDD4L to activated NRTK1 leads to receptor ubiquitination and down-regulation and to the modulation of neuronal survival in vitro (Figure 3A) (Arevalo et al., 2006; Yu et al., 2014). NEDD4 may also attenuate RTK signaling by regulating the levels of the tumor suppressor PTEN. In vitro studies demonstrated that NEDD4 was responsible for PTEN ubiquitination (Trotman et al., 2007; Wang et al., 2007), and subsequent studies in X. laevis confirmed that Nedd4-mediated ubiquitination of Pten promoted axonal and dendritic branching by allowing full activation of the PI3K/Akt pathway (Christie et al., 2010; Drinjakovic et al., 2010; Schmeisser et al., 2013). However, it does not appear that aberrant PTEN ubiquitination in mice played a role in impaired axon growth upon deletion of Nedd4 and Nedd4L (Hsia et al., 2014), suggesting that NEDD4 regulation of PTEN may only occur in specific biological contexts.

The first evidence that members of the CBL (Casitas B-lineage Lymphoma proto-oncogene) family of E3 ligases (cbl in D. melanogaster, SLI-1 in C. elegans, and CBL-3, CBL-B, and C-CBL in mammals) act as negative regulators of RTKs was provided by genetic screens in C. elegans and D. melanogaster (Supplemental Table 1). These early studies demonstrated that loss-of-function point mutations in the CBL homologs sli-1 and cbl resulted in aberrant signaling by the EGFR homologs LET-23 and Der, respectively (Jekely et al., 2005; Meisner et al., 1997; Pai et al., 2000; Wang et al., 2008; Yoon et al., 1995). Subsequent studies of the mammalian homologs have shown that c-Cbl- or Cbl-b-deficient mice are viable and fertile with only minor phenotypic differences, but combined deletion results in early embryonic lethality before mid-gestation (Mohapatra et al., 2013; Nakamura et al., 2001). This redundancy is consistent with in vitro work demonstrating that C-CBL and CBL-B work cooperatively to control the duration of EGFR signaling (Pennock and Wang, 2008). Upon ligand-induced receptor activation, phosphorylated CBL proteins complex with EGFR via a highly conserved TKB domain to facilitate receptor ubiquitination and degradation by a catalytic RING finger domain (Figure 3B) (de Melker et al., 2001; Haglund et al., 2003; Levkowitz et al., 1999; Levkowitz et al., 1998; Longva et al., 2002). The c-Cbl knockout phenotype is faithfully recapitulated by a mutation in the RING finger domain that eliminates its E3 ligase activity and resulted in more severe phenotypic changes than a loss-of-function mutation in the c-Cbl TKB domain (Thien and Langdon, 2005; Thien et al., 2003). Thus, TKB domain-mediated interactions with RTKs could not fully explain the spectrum of C-CBL functions. Further confirmation of this hypothesis came from a study analyzing homozygous knock-in mutation of Tyrosine 737, which eliminates the binding site for PI3K in the C-terminal tail of CBL, located outside of the TKB domain (Adapala et al., 2010a; Adapala et al., 2010b). Abrogation of the CBL/PI3K interaction resulted in perturbed RANKL-mediated signaling, leading to increased bone mass due to a cell-autonomous defect in osteoclast function, a phenotype not seen with other Cbl mutations (Adapala et al., 2010a; Adapala et al., 2010b).

Late attenuation of RTK signaling via receptor ubiquitination and degradation

By recruitment of E3 ubiquitin ligases to the receptor complex, members of the Leucine-Rich and Immunoglobulin-like domain (LRIG) and Suppressor of Cytokine Signaling (SOCS) families accelerate receptor ubiquitination and degradation (Figure 3C, D) (Gur et al., 2004; Laederich et al., 2004). Unlike NEDD4 and CBL ubiquitin ligases, the expression of LRIG1 and SOCS genes are induced by growth factor activation of RTKs via transcriptional activation and translation (Segatto et al., 2011). The LRIG1 transmembrane protein and SOCS cytosolic adaptor proteins have been shown to interact with RTKs and attenuate receptor signaling through both ligand-dependent and independent mechanisms.

Genetic approaches in vivo have confirmed the essential biological functions of LRIG proteins and have provided insight into the broad range of their effects on signaling pathways (Supplemental Table 1). Deletion of Lrig1 in mice leads to psoriasis-like epidermal hyperplasia and dramatically increased proliferation of the intestinal crypts and tracheal and bronchial epithelium (Karlsson et al., 2008; Lu et al., 2014; Luetteke et al., 1994; Suzuki et al., 2002). These phenotypes were correlated with a substantial increase in total and phosphorylated protein levels of EGFR, ERBB2, ERBB3, and MET in associated tissues (Suzuki et al., 2002), emphasizing the role of LRIG1 as a negative regulator of RTK signaling in vivo. Importantly, the skin and intestinal phenotypes in Lrig1-deficient mice could be rescued by genetic or chemical inhibition of EGFR phosphorylation, suggesting direct involvement of LRIG1 in controlling the strength of EGFR signaling (Luetteke et al., 1994). Indeed, extensive in vitro studies have demonstrated that LRIG1 attenuates the half-life of all four receptors of the ErbB family and of MET by amplifying C-CBL-mediated ubiquitination (Goldoni et al., 2007; Gur et al., 2004; Laederich et al., 2004; Rafidi et al., 2013; Rondahl et al., 2013; Shattuck et al., 2007; Stutz et al., 2008; Yi et al., 2011). Additional in vitro work suggests that limited proteolysis of the soluble ectodomain of LRIG1 may inhibit EGFR signaling by competing with ligand binding and stabilizing the receptor in the inactive monomeric state (Goldoni et al., 2007). LRIG1 was also shown to restrict RET recruitment to lipid rafts and to inhibit binding of its ligand GDNF preventing receptor activation (Ledda et al., 2008). While the functions of the other members of the LRIG family, LRIG2 and LRIG3, remain poorly understood, studies in X. laevis have demonstrated that lrig3 modulates Fgf-dependent Erk phosphorylation and Wnt signaling during neural crest cell specification and induction. When co-expressed in vitro, Lrig3 co-immunoprecipitated with Fgfr1 via its ectodomain, and this interaction was correlated with reduced levels of Fgfr1 protein (Zhao et al., 2008), suggesting that Lrig3 may attenuate Fgf signaling by the mechanisms similar to those described for EGFR and RET.

Although the biological roles of SOCS proteins have traditionally been considered in the context of cytokine receptor signaling through the JAK/STAT pathway in immunity and hematopoiesis, emerging evidence implicates SOCS proteins in the control of RTK signaling during development (Supplemental Table 1) (Trengrove and Gray, 2013). In vitro and in vivo studies demonstrated that SOCS2 exerts a dual role in the regulation of EGF signaling: Socs2 knockout mice displayed increased intestinal growth due to enhanced responsiveness to EGF (Michaylira et al., 2006), and cortical neurons derived from transgenic Socs2 overexpressing mice had increased neural outgrowth, apparently also due to enhanced EGF signaling (Goldshmit et al., 2004). The gigantism phenotype of Socs2-deficient mice suggests an important role for SOCS2 in the regulation of growth, possibly by modulating growth hormone and IGF-1R signaling (Greenhalgh et al., 2002; Metcalf et al., 2000). These mice exhibited prolonged STAT5B activation, and loss of Stat5b function partially relieved the growth enhancement. In contrast to Socs2 mutants, but similar to Nedd4 heterozygous mice, Socs6 knockout mice displayed a mild growth retardation thought to be due to perturbation of IGF-1R signaling (Krebs et al., 2002). Despite in vitro studies supporting a role for SOCS6 in neural stem differentiation and glucose metabolism (Choi et al., 2010; Gupta et al., 2011; Liu et al., 2008a; Liu et al., 2008b; Vlacich et al., 2010), mice deficient in Socs6 did not display phenotypic alterations consistent with such functions (Krebs et al., 2002). However, transgenic mice overexpressing Socs6 had altered glucose metabolism compared to wild type mice, with enhanced PI3K/AKT activation that was independent of increased IR or IGF-1R phosphorylation (Li et al., 2002). This suggests an additional mechanism by which SOCS6 regulates insulin signaling downstream of the receptor to control glucose metabolism. Similar to the engagement of LRIG1 with many RTKs, in vitro studies suggest that SOCS proteins regulate multiple RTKs including c-KIT, FLT3, IR, IGF-1R, and EGFR by enhancing their degradation via recruitment of E3 ubiquitin ligase complexes (Banks et al., 2005; Kario et al., 2005; Krebs et al., 2002; Nicholson et al., 2005; Trengrove and Gray, 2013). A subset of SOCS proteins – SOCS2, SOCS6, and SOCS7 – protect RTKs from SOCS-mediated degradation by interacting with the domains of other SOCS proteins responsible for the recruitment of E3 ubiquitin ligase complexes (Piessevaux et al., 2006). This suggests a role for these SOCS proteins in restoring cells to a responsive state for subsequent RTK stimulation.

Modulators of RTK signaling associated with human congenital disorders

Given the critical roles of RTK signaling in cell fate determination and morphogenesis, there has been great interest in understanding how RTK regulators are deregulated in human disorders. Indeed, both gain-of-function mutations, which lead to constitutive protein activation, and loss-of-function mutations, which lead to non-functional or dominant negative proteins, have been mapped to regulators of RTK signaling in human disease (Rauen, 2013; Tartaglia and Gelb, 2005). Importantly, mutations in the same gene can cause multiple conditions with wide phenotypic variability, and mutations in different genes can result in disorders with overlapping clinical features, linking these genes into overarching molecular networks. Studying the underlying pathophysiology of these disorders has uncovered novel regulators of RTKs, revealed new biological functions for those already identified, and advanced development of molecular-based therapies for treatment. We highlight here efforts that have provided information regarding human genetic disorders. Several other excellent reviews cover feedback regulators in cancer (Casaletto and McClatchey, 2012; Logue and Morrison, 2012; Regad, 2015)

The key role for Anosmin 1 in neuronal targeting and migration was determined by the identification of missense mutations that result in inactive protein in Kallmann syndrome (KS) (Bick et al., 1992). KS is a disorder characterized by hypogonadotropic hypogonadism (HH), defined as absent or incomplete sexual maturation by the age of 18 years, with or without anosmia. Less frequent symptoms include renal agenesis, cleft palate, mirror movements, and hearing loss (Tsai and Gill, 2006). FGFR1 loss-of-function mutations in an autosomal dominant form of KS first suggested that Anosmin 1 was involved in FGF signaling (Dode et al., 2003). Interestingly, missense mutations in members of the FGF8 set of co-regulated genes, or synexpression group, including DUSP6, SPRY4, FLRT3, and SEF, have also been identified in individuals with HH with or without anosmia (Miraoui et al., 2013). The functional characterization of these mutations may offer new insight into their molecular mechanisms of action and roles of these genes in regulation of FGF signaling in gonadotropin-releasing hormone biology.

The importance of genes that encode protein components or regulators of the RAS/MAPK pathway is elegantly demonstrated by germline mutations associated with a class of developmental disorders known as the RASopathies (Goodwin et al., 2015; Goyal et al., 2017; Jindal et al., 2015; Jindal et al., 2017; Rauen, 2013). In one of these conditions, Costello syndrome (CS), nearly all individuals have a heterozygous de novo germline mutation in HRAS that results in a constitutively active protein (Aoki et al., 2005; Estep et al., 2006), while in cardio-facio-cutaneous syndrome (CFC), patients have heterozygous activating germline mutations in KRAS, BRAF, MEK1, or MEK2, all components of the RAS/MAPK pathway (Niihori et al., 2006; Rodriguez-Viciana et al., 2006). Because of the common underlying pathway dysregulation, RASopathies exhibit numerous overlapping clinical phenotypes. Heterozygous inactivating mutations in SPRED1 cause Neurofibromatosis Legius syndrome (NFLS), a mild form of Neurofibromatosis 1 (NF1), which is characterized by multiple cafe-au-lait skin spots, variable dysmorphic features such as hypertelorism or macrocephaly, lipomas, and mild learning disabilities or attention problems (Figure 5A) (Brems et al., 2007; Brems et al., 2012). The similarities of NFLS and NF1 are explained by the shared underlying molecular mechanism: SPRED1 downregulates the RAS/MAPK pathway through neurofibromin, the NF1 gene product (Stowe et al., 2012). Interaction between these proteins facilities plasma membrane localization of neurofibromin, where it functions as a RAS GTPase-activating protein to negatively regulate RAS signaling (Adapala et al., 2010a; Dunzendorfer-Matt et al., 2016; Hirata et al., 2016; Martin et al., 1990; Stowe et al., 2012; Xu et al., 1990). Association of SPRED2 and SPRED3 with neurofibromin suggests that these isoforms may compensate for loss of Spred1 and thus helps explain the milder phenotype associated with NFLS in comparison with NF1 (Stowe et al., 2012).

The manifestations of another RASopathy, LEOPARD syndrome 1 (LPRD1), are numerous: multiple lentigines, electrocardiographic conduction abnormalities, ocular hypertelorism, pulmonic stenosis, abnormal genitalia, retardation of growth, and sensorineural deafness (Figure 5B) (Legius et al., 2002; Mendez et al., 1985). LPRD1-associated PTPN11 mutations lead to a catalytically defective SHP2 protein that acts in a dominant negative fashion and interferes with MAPK/ERK signaling (Digilio et al., 2002; Kontaridis et al., 2006; Tartaglia et al., 2006). In contrast, heterozygous missense mutations in PTNP11 that result in excessive SHP2 activity are a principal cause of Noonan syndrome (NS) (Fragale et al., 2004; Tartaglia and Gelb, 2005), a relatively common disorder characterized by short stature, facial dysmorphia, and a wide spectrum of congenital heart defects (Figure 5C) (Digilio et al., 2002; Tartaglia et al., 2001). How two mutations with opposite effects on catalytic activity result in syndromes with similar clinical symptoms is a fascinating open question. Genetic and biochemical studies in D. melanogaster and zebrafish successfully demonstrate that ptpn11 mutations associated with LPRD1 and NS result in distinct but similar phenotypes, and in the case of the zebrafish, recapitulate the craniofacial and cardiac defects of human patients (Supplemental Table 1) (Bonetti et al., 2014; Jopling et al., 2007; Oishi et al., 2006; Oishi et al., 2009; Stewart et al., 2010). Mouse models generated for the two most prevalent LPRD1 and NS PTPN11 mutations exhibit developmental defects, including reduced length, craniofacial abnormalities, congenital heart defects, with activation of the PI3K/AKT or RAS/ERK pathways, respectively (Araki et al., 2004; De Rocca Serra-Nedelec et al., 2012; Marin et al., 2011). Importantly, genetic deletion of ERK1/2 prevented cardiac abnormalities in a cardiomyocyte-specific SHP2 gain-of-function mouse model of NS (Nakamura et al., 2007), and injection of the MAPK/ERK kinase inhibitor U0126 in utero prevented craniofacial malformations in newborn pups (Nakamura et al., 2009). Similarly, pharmacological intervention with rapamycin, an inhibitor of mTOR, reversed the hypertrophic cardiomyopathy in a mouse model of LPRD1 (Marin et al., 2011). These studies suggest that some RASopathy-associated PTPN11 mutations can be rescued, opening a new therapeutic avenue for affected individuals. A NS-like phenotype has been associated with several additional genes including C-CBL (Figure 5D) (Martinelli et al., 2010; Niemeyer et al., 2010; Perez et al., 2010). In vitro studies showed that the C-CBL mutations found in patients impaired CBL-mediated degradation of cell-surface receptors in a dominant-negative fashion and caused dysregulation of intracellular signaling through RAS, explaining the overlapping phenotype in NS associated with RAS/MAPK pathway activating mutations (Martinelli et al., 2010; Schubert et al., 2006).

Germline mutations and deletions in PTEN that result in dysregulation of the PI3K/AKT pathway cause Bannayan-Ruvalcaba-Riley syndrome (BRRS) and Cowden syndrome 1 (CWS1) (Liaw et al., 1997; Marsh et al., 1999; Nelen et al., 1997; Zhou et al., 2003). BRRS and CWS1 are rare allelic disorders that share characteristics such as hamartomatous polyps of the gastrointestinal tract, mucocutaneous lesions, and increased risk of developing neoplasms (Blumenthal and Dennis, 2008). It has been suggested that both conditions and several other distinctive phenotypes associated with PTEN mutations be referred to as PTEN hamartoma tumor syndrome (Lachlan et al., 2007; Marsh et al., 1999; Nelen et al., 1997; Sarquis et al., 2006). Also included in this spectrum of disorders are PTEN-related Proteus syndrome (PS) and ‘Proteus-like’ syndrome, complex and highly variable disorders involving vascular malformations and hamartomatous overgrowth of multiple tissues associated with germline and tissue-specific somatic activating mutations in AKT1 or PTEN, respectively (Cohen, 2014; Lindhurst et al., 2011; Smith et al., 2002; Turner et al., 2004; Zhou et al., 2001; Zhou et al., 2000). These correlations demonstrate the critical involvement of PTEN in regulation of the pro-proliferative signals mediated by the PI3K/AKT pathway.

Expanding the array of clinically distinct phenotypes associated with PTEN mutations are VACTERL association and macrocephaly/autism syndrome (Butler et al., 2005; Reardon et al., 2001). VACTERL describes a constellation of congenital anomalies including vertebral anomalies, anal atresia, congenital cardiac disease, tracheoesophageal fistula, renal anomalies, radial dysplasia, and other limb defects (Khoury et al., 1983), whereas macrocephaly/autism syndrome is characterized by increased head circumference, abnormal facial features, and delayed psychomotor development resulting in autistic behavior or intellectual disability (Figure 5E) (Herman et al., 2007; Tsujita et al., 2016). Whether individuals affected with macrocephaly/autism syndrome and VACTERL association develop further clinical manifestations of other PTEN-associated syndromes is unknown. Mouse models with deficient Pten result in macrocephaly and autistic-like behavior with abnormal activation of PI3K/AKT pathway (Supplemental Table 1) (Chen et al., 2015; Clipperton-Allen and Page, 2014; Kwon et al., 2006; Page et al., 2009). Future analyses of specific disease-causing human PTEN mutations will prove useful in understanding the mechanisms underlying these heterogeneous phenotypes.

The recent discovery of autosomal recessive LRIG2 mutations in Urofacial syndrome (UFS) provides additional insight into LRIG2 function as a regulator of RTK signaling. UFS presents with urinary bladder dysfunction associated with abnormal facial expressions (Figure 5F) (Stuart et al., 2013). Interestingly, loss of function mutations in Heparanase-2, which regulates the availability and signaling of growth factors through processing of HSPGs, were also identified as causative for UFS (Daly et al., 2010; Pang et al., 2010). In fact, deletion of Hpse2 but not Lrig2 in X. laevis and mice caused UFS-like urological phenotypes. These observations suggest that HPSE2 might functionally overlap in its mode of RTK inhibition with LRIG2 (Supplemental Table 1) (Guo et al., 2015a; Roberts et al., 2014).

CONCLUDING REMARKS

From the single cell stage, RTKs guide the embryogenesis, development, and postnatal growth of nearly all organisms. Our understanding of the significant contribution that RTKs play has been enabled through extensive work in model organisms and by advances in elucidating the biochemistry, cell biology, and structure of these receptors. Equally important contributions have arisen from studies of human congenital disorders and clinical analyses of RTKs in diseases. These studies reveal that a complex network of proteins is required to guide RTKs during their lifetime in the cell, from their biosynthesis and maturation in the ER, subsequent trafficking to the cell surface, ligand-dependent activation triggering autophosphorylation and downstream signaling, and final desensitization by ubiquitination and endocytosis. Although we have come to appreciate the mechanisms by which many of these regulators contribute to the duration and extent of ligand-dependent RTK activation in development, there are still many exciting discoveries to be made about the individual steps that are necessary to properly regulate RTK signaling. Our knowledge of the specificity of individual regulators for different RTKs and insights into the mechanism of their function remains largely minimalistic. Even more lagging are structural studies on many of these targets, which are necessary to advance development of specific therapeutics for patients in which modulation of RTK regulators could be clinically beneficial. Lastly, it is likely that many feedback regulatory loops still remain to be discovered.

Highlights.

• RTK signaling plays pivotal roles in cell fate determination and morphogenesis.

• Feedback regulators fine-tuning the duration and extent of pathway activation.

• Positive regulators function in tandem with negative regulators in development.

• Germline mutations in RTK regulators cause human genetic disorders.