The Role of Protease Activity in ErbB Biology

Abstract

Proteases are now recognized as having an active role in a variety of processes aside from their recognized metabolic role in protein degradation. Within the ErbB system of ligands and receptors proteases are known to be necessary for the generation of soluble ligands from transmembrane precursers and for the processing of the ErbB4 receptor, such that its intracellular domain is translocated to the nucleus. There are two protease activities involved in the events: proteases that cleave within the ectodomain of ligand (or receptor) and proteases that cleave the substrate within the transmembrane domain. The former are the ADAM proteases and the latter are the γ-secretase complex and the rhomboid proteases. This review discusses the roles of each of these protease systems within the ErbB system.

ADAM PROTEASES

All ligands of ErbB1 are synthesized with a membrane tether, which must be cut in order to release the soluble growth factors [1]. This process, which is referred to as protein ectodomain shedding (Figure 1A), is therefore a critical regulator of the ability of ErbB-ligands to activate ErbB receptors in a paracrine manner (for previous reviews, see [1, 2]). Moreover, shedding also can be important for juxtacrine signaling through ErbB1, at least under certain conditions, as signaling of TGFα to adjacent ErbB1-expressing cells can be blocked with a metalloproteinase inhibitor [2, 3]. Since ErbB receptors have important functions in development and in diseases such as cancer [4, 5], there is considerable interest in elucidating the identity of the enzymes responsible for processing and releasing ErbB-ligands, and understanding their regulation during development and adult homeostasis as well as their dysregulation in disease.

Figure 1.

Protein ectodomain shedding refers to the proteolytic release of the extracellular domain or “ectodomain” of a membrane protein from its membrane anchor (A). ADAM (a disintegrin and metalloprotease) 10 and 17 as well as other enzymes, such as ADAMTS1 (a disintegrin and metalloprotease with thrombosponding motifs) have been implicated in the ectodomain shedding of ErbB-ligands and of ErbB3. A schematic representation of the domain organization of a typical ADAM (B) shows an amino-terminal metalloprotease domain, followed by a disintegrin domain, a cysteine-rich region, an EGF-like repeat, a transmembrane domain and a cytoplasmic domain. The cytoplasmic domain of ADAMs frequently contain signaling motifs such as phosphorylation sites or proline-rich SH3 ligand binding domains.

ADAMs 10 and 17 are the Principal Sheddases for ErbB1-ligands

In mammalian cells, two membrane-anchored enzymes belonging to the ADAM (a disintegrin and metalloprotease) family of membrane anchored metalloproteinases, ADAMs10 and 17, have emerged as key molecules in shedding of the seven ErbB1-ligands (Figure 1B shows a typical domain organization of an ADAM, for previous reviews on ADAMs, please see [2, 6–8]. Based on studies with cells lacking various ADAMs and with selective inhibitors, ADAM17 is considered the principal sheddase for TGFα, HB-EGF, amphiregulin, epiregulin, epigen [9–16], whereas ADAM10 was identified as the major sheddase for EGF and betacellulin [14, 17, 18]. The relevance of ADAM17 in shedding of TGFα, HB-EGF and amphiregulin in vivo was corroborated by studies in mice lacking ADAM17, which resemble mice lacking TGFα in that they have open eyes at birth and wavy whiskers [15], or mice lacking amphiregulin with respect to defects in mammary ductal development [12], or HB-EGF with respect to defects in heart valve development [10, 14]. Moreover, ADAM17 is responsible for the PMA-stimulated shedding of several isoforms of the ErbB ligand neuregulin [19, 20]. The role of ADAM10 in shedding EGF has been further corroborated using ADAM10 selective inhibitors [17]. However, an evaluation of the contribution of ADAM10 to activating EGF and BTC in vivo is currently not possible, because mice lacking ADAM10 die very early during embryogenesis (E 9.5) [21], so conditional knockout mice will be necessary to attempt to address the relevance of ADAM10 in shedding EGF and BTC in vivo.

Phorbol Esters are Activators of ADAM17

In light of the well-established role of proteolysis in activating ErbB1 ligands, it is important to understand how the responsible sheddases are regulated. Tumor promoting phorbol esters (PMA/ TPA) were among the first identified activators of protein ectodomain shedding [22–24]. Since then, ADAM17 has emerged as the principal ErbB1-ligand-sheddase that is stimulated by short-term treatment of cells with the PMA (25 ng/ml for <1hr) [14, 16, 18, 25]. Addition of PMA to cells leads to a burst of shedding of ADAM17-substrates that have accumulated in the late secretory pathway and on the cell surface, but once these substrates are consumed, shedding returns to constitutive levels even in the presence of PMA [16].

There are conflicting reports on the mechanism of activation of ADAM17 by PMA. Several studies have provided evidence that PMA stimulation increases relocalization and transport of ADAM17 to the cell surface and increases processing of ADAM17 by furin, which removes the inhibitory pro-domain[25, 26]. On the other hand, a study by Horiuchi et al. reported conditions under which PMA-stimulation of ADAM17-dependent shedding in cell based assays did not coincide with a detectable increase in pro-domain removal or relocalization to the cell surface, suggesting that these processes are not required for stimulation of ADAM17 [16]. Horiuchi et al. also showed that blocking transport through the secretory pathway with brefeldin A had no effect on the PMA-stimulation of ADAM17, so increased transport through the secretory pathway is unlikely to be a key component of the activation of ADAM17. Finally, several studies have reported that the cytoplasmic domain of ADAM17 becomes phosphorylated upon treatment of cells with PMA, suggesting that this could be required for its activation by PMA [26–28]. However, at least two studies have shown that the cytoplasmic tail of ADAM17 is dispensable for PMA stimulation[16, 29]. Clearly, more research will be necessary to understand how ADAM17 is stimulated by PMA and other, more physiological stimuli, such as G-protein coupled receptors (see below). In this context it is important to note that ADAM10 has been reported to respond to activation by PMA[30], albeit after longer treatment with higher doses of PMA. Under these conditions, PMA is known to stimulate a number of other processes that might indirectly affect shedding, such as increased biosynthesis and transport of membrane proteins through the secretory pathway. Nevertheless, in side-by-side experiments, short term treatment of cells for 1 hr or less with relatively low doses of PMA (25 ng/ml) only activates shedding by ADAM17, but not by ADAM10 [16, 31]. Stimulation with low doses of PMA for an hour or less can therefore be useful to identify candidate substrates for ADAM17, subject to further validation by additional criteria, such as evaluating shedding in the presence of selective ADAM17 inhibitors or in ADAM17-deficient cell types.

Impact of Other Signaling Pathways on ADAMs 10 and 17

A commonly used activator of ADAM10-dependent shedding is the calcium ionophore ionomycin [16, 18, 25, 31], which also activates ADAM17 [13]. Deletion of the cytoplasmic domain of ADAM10 results in reduced calcium influx-stimulated shedding of its substrates in rescue experiments with Adam10−/−cells [16], suggesting that the function of ADAM10 is regulated through interactions with cytoplasmic molecules, although these remain to be identified. The cytoplasmic domain of ADAM10 is also important for its sorting to the basolateral membrane in polarized epithelial cells [32]. The identification of a basolateral sorting motif in ADAM10 is consistent with the observation that shedding of the ADAM10 substrate EGF is restricted to the basolateral aspect of polarized epithelial cells [33].

Both calcium influx and PMA are pleiotropic activators of intracelluar signaling pathways, and therefore do not necessarily help understand the role of physiological signaling pathways in regulating the function of ADAMs 10 or 17. The discovery of a metalloprotease-dependent crosstalk between G-protein coupled receptors and ErbB1 [34], which was later found to depend on ADAMs such as ADAM17 [35, 36], provided one of the first insights into the regulation of ADAMs by a more physiological signaling pathway . Since then, this general concept has been corroborated by numerous additional reports of crosstalk between GPCRs and ErbB signaling, for example activated by angiotensin [37], or cannabinoids [38]. Moreover, tobacco smoke [39] and UVA [40] also activate metalloprotease-dependent ErbB/ERK signaling. In conceptually similar experiments, activation of the VEGFR2, a receptor with important roles in angiogenesis, also results in enhanced ADAM17-dependent release of ErbB ligands and activation of ERK [41]. Finally, studies of TNFα shedding in mice lacking ADAM17 in myeloid cells demonstrated that ADAM17 is activated when toll like receptors are stimulated with endotoxin (TLRs) [42], and more recently, TLR signaling has been implicated in ADAM17-dependent podosome disassembly [43], so it appears likely that TLR stimulation should also activate ADAM17-dependent crosstalk to ErbB receptors. Thus ADAM17, and in some cases additional ADAMs [36], are emerging as amplifiers of cell signaling and mediators of crosstalk between several prominent signaling pathways and ERK1/2, most likely through their ability to shed ErbB-ligands. However, since ADAM17 has many other substrates on the cell surface, ADAM17-dependent activation of ERK and other signaling pathways could conceivably also be due to processing of other membrane proteins.

Could other ADAMs contribute to ErbB-ligand Shedding?

In addition to ADAMs 10 and 17, several other ADAMs are capable of processing certain ErbB-ligands when they are overexpressed or dysregulated [16]. EGF, for example, can be released from cells by co-transfected ADAMs 8, 9, 12 or 19, and BTC by ADAMs 8, 12 or 19, and ADAM8 can cleave the ADAM17 substrates AMP and HB-EGF when overexpressed [16]. In addition, ADAMTS1 has been implicated in shedding of HB-EGF and AR and ErbB activation [44]. The ability of several different enzymes to shed ErbB1-ligands and activate the ErbB receptors or ERK suggests that the release of the soluble ectodomain of an ErbB-ligand is sufficient for activation of paracrine ErbB signaling, regardless of which enzyme is responsible. So any ADAM or other enzyme that can cleave one or more ErbB ligands could conceivably contribute to the paracrine activation of ErbB receptors under conditions where the enzyme is overexpressed together with its candidate substrates.

There are several examples of dysregulation of the expression of ADAMs other than ADAM10 or 17. In the case of ADAM9, dysregulation occurs in various carcinomas [45] and also in cells or tissues treated with reactive oxygen species [46, 47], and ADAM12 is also overexpressed in stromal cells adjacent to tumor cells [48]. Studies on ADAMs 8, 9, 12 and 15 have not yet uncovered evidence for short-term posttranslational upregulation of their shedddase activity, such as that seen for ADAM17 stimulated by PMA, or ADAM10 stimulated by Ca²⁺-influx (data not shown). Instead, these ADAMs appear to be constitutively active as their increased expression results in increased processing of substrates. Therefore, the relative contribution of these ADAMs to shedding of membrane proteins is likely to depend on transcriptional regulation or changes in protein stability or turnover, with higher protein levels correlating with higher activity.

What Determines the Substrate Selectivity of ADAMs in Cells?

An important question that arises with respect to proteolytic activation of EGFR-ligands is which enzyme is required for processing which substrate(s) under various conditions, and what underlying mechanism might account for any substrate selectivity [49]. An attractive hypothesis has been that ADAMs selectively target a subset of substrates, perhaps via an interaction between the extracellular domains of the enzyme and substrate, and that this accounts, at least in part, for the substrate selectivity. However, with respect to Ca²⁺-influx or PMA-stimulated shedding, domain swap experiments between the ADAM10 substrates EGF and BTC and the ADAM17 substrate TGFα suggest that the cleavage site is the most important determinant for the substrate selectivity of ADAMs 10 and 17, and not the cytoplasmic domain or ectodomain of these substrates [16]. Moreover, recent studies raise the possibility that the substrate selectivity of ADAMs10 and 17 might be determined, at least in part, by their partitioning into distinct membrane microdomains. Specifically, ADAM10 co-purifies with tetraspanins, whereas ADAM17 does not [50], and so this might be one mechanism that physically separates ADAM10 and 17 into different domains on the cell surface. The shedding activity of ADAM17, on the other hand, has been proposed to partition into lipid rafts, and this could represent one aspect of its substrate selectivity [51]. Finally, the function of both ADAMs is increased when cholesterol is removed from cell membranes by treatment with cyclodextrane [52], suggesting that membrane rigidity or perhaps the lateral mobility of these ADAMs and presumably also their substrates, could contribute to the regulation of their activity as sheddases. Clearly further studies will be required to understand the basis for the substrate selectivity and regulation of ADAMs under various physiological and pathological conditions.

γ-SECRETASE

γ-Secretase is an intramembrane aspartyl protease activity that mediates the processing of numerous transmembrane proteins, including the well described cleavage of Notch and the amyloid precursor protein [53]. Substrate cleavage occurs within the transmembrane domain and results in the liberation of the intracellular domain (ICD) from the membrane into the cytoplasm. γ-Secretase is the term applied to an activity that involves at least four transmembrane proteins that exist as complex: presenilin, nicastrin, PEN-2, and APH-1. Within this membrane complex, presenilin is the protease responsible for substrate cleavage, while nicastrin functions as a targeting subunit. The functions of PEN-2 and APH-1 concern maturation and stability of the complex. Presenilin is a multi-transmembrane domain molecule that exists as a zymogen until it is activated by endoproteolysis into N- and C-terminal fragments, which remain associated in a high molecular weight complex of the active enzyme.

ErbB-4 Cleavage and Nuclear Localization

Ectodomain proteolytic processing of ErbB-4 includes a basal level, which can be increased by TPA in all cells or by the addition of neuregulin (heregulin) to some but not all cells [54, 55]. This cleavage results in the formation of two receptor fragments: a 120 kDa ectodomain fragment that is released into the media and an 80 kDa membrane-bound fragment, termed m80 (Figure 2). Cleavage requires ADAM 17 (TACE) and it is likely that this is the enzyme that also executes cleavage of ErbB-4 between His651 and Ser652 within the extracellular stalk or ecto-juxtamembrane region [56, 57]. Hence, the m80 fragment includes eight ectodomain residues, the transmembrane domain and entire ICD of ErbB-4.

Figure 2. Ectodomain Cleavage of ErbB-4.

Adam17 interaction with ErbB-4, following the addition of TPA or neuregulin, results in an endoproteolytic cleavage between His651 and Ser652. This event results in release of a large ectodomain fragment into the media and the formation of a cell-associated truncated receptor termed m80.

Sensitivity to ectodomain shedding is likely determined, at least in part, by the length of the stalk region in various transmembrane proteins, as demonstrated for the selectins [58]. There are two ErbB-4 isoforms termed Jm-a, in which the ectodomain is sensitive to cleavage, and Jm-b, which is not cleavable [59]. The stalk region in Jm-b is much shorter (6 residues) than the corresponding region of Jm-a (16 residues). Since ADAM-mediated cleavage events do not involve a defined sequence or cleavage site in the substrate, it seems that longer stalk regions in substrates may simply permit accessibility of the protease. It is interesting to note that, like the Jm-b ErbB-4 isoform, ErbB-1, -2 and -3 also have relatively short stalk regions (6–9 resides) and are not subject to a significant level of metalloprotease mediated ectodomain cleavage [54].

It seems probable, though not formally demonstrated, that the shed ErbB-4 ectodomain may function to block receptor activation by binding neuregulin. The function of the m80 fragment, however, is known. The capacity of γ-secretase to cleave substrates requires that the substrate have a short ectodomain region [60]. There is no γ-secretase cleavage of transmembrane proteins in which the ectodomain exceeds 200 residues and cleavage is maximal when the ectodomain is reduced to 50 or fewer residues. Hence, the ADAM-mediated removal of a large portion of the ErbB-4 ectodomain is a prerequisite step for subsequent γ-secretase cleavage [61]. It has been shown that the nicastrin subunit of the γ-secretase complex recognizes transmembrane proteins with shortened or nub-like ectodomains and thereby acts as a targeting agent for intramembrane cleavage by presenilin [62]. This would predict that nicastrin recognition of the ErbB-4 m80 fragment initiates intramembranous cleavage (Figure 3). Subsequently, presenilin activity converts the ErbB-4 m80 fragment to a soluble s80 or ICD fragment that is found in the cytoplasm, nucleus and, in one report, mitochondria [61, 63, 64]. The C-terminus of ErbB-4 encodes a PDZ domain recognition motif and this is required for presenilin cleavage of the m80 fragment [65]. Deletion of this motif (TVV) does not influence ectodomain cleavage, but does abrogate presenilin association with the m80 fragment and production of the ICD fragment. Presenilin also contains a PDZ domain recognition motif, and, it is possible that a scaffold of PDZ domain containing proteins may be required for γ-secretase cleavage. Proteins that recognize the PDZ domain recognition motifs in ErbB-4 and presenilin have been reported.

Figure 3. γ-Secretase Intramembrane Proteolysis of the m80 ErbB-4 Fragment.

Available data suggests that the nicastrin subunit of the γ-secretase complex likely functions as a targeting subunit by recognition of the m80 fragment of ErbB4. This is followed by intramembrane cleavage of m80 and generation of an ICD (s80) fragment that is subsequently translocated to the nucleus and mitochondria. Also depicted is a hypothetical small peptide that would resemble the Aβ peptide produced by cleavage of APP.

Presenilin cleavage of substrates occurs within the transmembrane domain and, based on APP and Notch processing, this may occur at multiple sites, producing several species of ICD fragments that may have differing levels of metabolic stability based on the N-end rule [66]. Hence, mutation within the transmembrane domain can diminish cleavage and/or alter the metabolic stability of the ICD fragments. This is shown in the case of Notch where a transmembrane mutation appears to prevent cleavage, but actually results in a new ICD fragment that is very rapidly degraded due to the presence of a metabolically destabilizing N-terminal residue [67]. It has been reported that the Val675Ala [68] or Val673Ile [69]mutations within the ErbB-4 transmembrane domain abrogate γ-secretase cleavage, as judged by the inability to detect the ICD fragment. In view of the Notch mutagenesis data, it is not clear whether these mutations actually prevent cleavage or result in a less stable ICD fragment. Given the low level of ICD fragment normally detectable, a modest change in stability would render the fragment undetectable by the same methodology. In addition to the ICD fragment, γ-secretase cleavage of the m80 ErbB-4 fragment should also produce a small peptide (Figure 3) with an N-terminal Ser652 residue that would be analogous to the amyloid beta peptide derived from APP, which is considered to be causative in Alzheimer’s disease. This small ErbB-4 peptide, however, has never been detected.

There is now a lengthy and growing list of transmembrane substrates subject to γ-secretase processing [70]. The critical issue until recently has been whether this proteolytic pathway constitutes a signal transduction pathway necessary for biological responses. Alternatively, it has been argued that the ADAM/γ-secretase activities could represent a cellular mechanism to degrade the transmembrane domains of membrane proteins [71, 72], since this cannot be accomplished by the proteosome. This is suggested, in part, because, among the membrane molecules that are subject to this proteolytic pathway, in relatively few cases (Notch, ErbB-4, Syndecan-3) is the processing controlled by ligand binding [72, 73]. In many cases, processing is constitutive or accelerated by relatively non-specific reagents, such as TPA.

It is now clear that endogenous generation of the ErbB-4 ICD by γ-secretase is required for correct control of astrogenesis in the developing mouse [74]. In this system the ICD fragment interacts with TAB2, an adaptor protein, and thereby with N- CoR, a co-repressor, and chaperones this complex into the nucleus. A similar chaperone mechanism between the ErbB-4 ICD and STAT5 has been proposed to be operative in mammary differentiation [75]. It is clear that ErbB-4 is functionally involved in mammary development [76]. While ErbB-4 nuclear localization has been observed in normal and tumor mammary tissue and exogenous ICD expression provokes differentiation events, it has not yet been demonstrated that ErbB-4 cleavage is physiologically relevant in this tissue. Also, in line with a role of the ErbB-4 ICD fragment in various cell differentiation systems, is the report that γ-secretase inhibition prevents neuregulin generation of the nuclear ErbB-4 ICD in oligodendrocytes and maturation of this cell type [77].

In addition to STAT 5 and the TAB 2:N-CoR complex mentioned above, several other proteins [Eto-2 [78], YAP [79, 80], WWOX [81], ER [82], Mdm 2 [83], AIP4/Itch [84] have been reported to associate with the ErbB-4 ICD. Eto-2, YAP and ER are transcription factors/ co-activators and the ICD may regulate their nuclear localization similar to STAT 5 and N-CoR. WWOX is a cytoplasmic protein and its interaction with the ICD attenuates nuclear translocation of the ICD, while AIP4/Itch is a cytoplasmic ubiquitin ligase that modulates the levels of intact ErbB-4 and the ICD. The ICD is an active tyrosine kinase [85] that phosphorylates Mdm2, a regulator of p53 [83]. Interestingly, intact ErbB-4 does not detectable phosphorylate Mdm2. These results, many of which have been produced in experimental cell systems only, suggest that proteolytic processing of ErbB-4 by γ-secretase allows new signaling functions to occur.

Cleavage of Neuregulin-1

Juxtacrine growth factor signaling is defined by an interaction between ligand and receptor on the surfaces of adjacent cells. Perhaps the best understood examples are the interactions between ephrins and Eph receptors and the Notch receptor and its ligands. Available data indicate that juxtamembrane signaling also occurs between transmembrane precursor form of neuregulin-1 and ErbB-4 [86]. In these juxtacrine systems not only is the receptor subject to proteolytic processing, as described above, but the ligand is also cleaved with release of its ICD into the cytoplasm.

While the ephrins and Notch ligands and their respective receptors are cleaved by γ-secretase to release ICD fragments, the role of γ-secretase in the processing of neuregulin-1 is not clear. However, it is clear that a soluble ICD fragment is produced following interaction of the type III neuregulin-1 with ErbB-4 [86]. In one report [86], the capacity of the ICD fragment to induce transcriptional activity was assessed in 293 cells, using the Gal4-VP16 transactivation system in the context of a neuregulin-1 construct. Treatment with soluble ErbB-2: ErbB-4, as a ligand for neuregulin-1, provoked reporter transcriptional activity that was blocked by γ-secretase pharmacologic inhibitors. However, the study did not examine whether production of the ICD fragment was similarly sensitive to γ-secretase inhibitors.

In a separate report, the PSD-95 promoter was identified as an indirect target of the neuregulin-1 ICD [87]. Examination of ICD constructs indicated that, while γ-secretase activity is required for production of the ICD from neuregulin-1, it is not required for increased activity from the PSD-95 promoter. In part this discrepancy is attributed to a splice variant in the neuregulin-1 cytoplasmic domain referred to as the type “a” isoform. Formation of this ICD construct is γ-secretase independent and accounts for the majority of neuregulin-1 signaling to the PSD-95 promoter. Unfortunately, the protease that does provoke formation of the neuregulin-1 ICDa fragment was not identified.

Cleavage of HB-EGF

In the case of HB-EGF, a fragment containing the ICD is released from pro-HB-EGF and translocates to the nucleus following the addition of TPA to cells [88, 89]. However, this fragment also contains the transmembrane domain and is localized on the inner nuclear membrane [90]. The data indicate that the γ-secretase activity is not required for formation of this fragment. The nuclear HB-EGF fragment (termed HB-EGF-C), which is analogous to the m80 ErbB-4 fragment, does induce the nuclear export of the transcription factor promyelocytic leukemia zinc finger.

RHOMBOIDS

Drosophila genetic analysis led to the discovery of a different type of proteolytic control of ErbB signaling. Drosophila only has a single ErbB homolog, called the EGF receptor (EGFR), but which is actually equally similar to all four mammalian ErbB genes [91]. The Drosophila EGFR has multiple developmental roles, regulating in different contexts, for example, cell proliferation, survival, differentiation and migration[92]. Because of this central developmental role, the Drosophila EGFR pathway has been extensively genetically analyzed, allowing the identification of most of the main components of the pathway, and its key regulators. This kind of genetic analysis identified and highlighted the importance of a gene called rhomboid (now rhomboid-1) [93–99]. Of all the known components of the pathway, it was the only one to be necessary (its loss blocked EGFR activity), sufficient (its ectopic expression was enough to activate signaling almost anywhere), and whose dynamic expression pattern closely prefigured where EGFR signaling activity. Based on these three genetic criteria, Rhomboid-1 appeared to be the cardinal physiological activator of EGFR signaling in flies. Further analysis, using genetic mosaic techniques, was able to prove that rhomboid-1 was required in the cells that produced the TGFα-like ligand, Spitz, not in the signal receiving cells [99]. Like TGFα and most other mammalian EGFR ligands, Spitz is synthesized as a transmembrane protein and it had been previously shown that the extracellular EGF domain needed to be released from the membrane to be an active ligand of the EGFR [100, 101]. These observations led to the proposal that Rhomboid-1 participated in this proteolytic release and activation of soluble Spitz (Figure 4).

Figure 4. Genetic Model of Rhomboid-1 in Drosophila EGFR Pathway.

Note that this conceptual model, as inferred from genetic data, could not address the intracellular location of Rhomboid-1 in the signal emitting cell. Spitz is the Drosophila orthologue of TGFα. See Figure 5 for a more accurate cellular picture of Rhomboid-1 function.

This quite detailed model of rhomboid-1’s role in EGFR activation in Drosophila had been inferred solely from genetic evidence. Unfortunately, it was not obviously supported by the molecular evidence: rhomboid-1 was a rather featureless membrane protein with seven transmembrane domains but otherwise no motifs indicative of its function and, specifically, no obvious similarity to proteases [102]. The picture was clarified when it was finally shown that Rhomboid-1 was after all a protease, but a novel one, unrelated to any previously identified protease class: it was the first member of the intramembrane serine proteases [103] (Figure 5A). Once rhomboid’s activity as a protease was clear, the genetic model made sense. Rhomboid provides the proteolytic activity needed to release Spitz (and two other TGFα-like ligands) from the cell in which it is synthesized, allowing it to become a diffusible growth factor that can activate the EGFR. Rhomboid-1 therefore has an analogous role to that proposed for ADAM17 in mammalian TGFα release [9, 14, 15]. It is important to emphasize, however, that rhomboid is a completely unrelated to ADAM17, demonstrating that, although similar, this was not a mechanism conserved between Drosophila and mammals. This was somewhat surprising, given that major signaling pathways are highly conserved amongst metazoans.

Figure 5. Rhomboids are Intramembrane Serine Proteases.

A. Cartoon of Drosophila Rhomboid-1, with the catalytic serine and histidine residues indicated. B. Crystal structure of the E. coli rhomboid, GlpG (which has six TMDs); from data of Wang et al. [109]. In both panels, the catalytic serine is highlighted in red, and the TMD 4 is shown in green.

Intramembrane Proteases

As described above, rhomboid was the first intramembrane serine protease to be identified. Since then, it has become clear that it is the founding member of a large and widely conserved family of similar enzymes that exist in all kingdoms, from bacteria to humans [104, 105]. The defining features of rhomboids are that their active site residues are buried within the plane of the lipid bilayer of membranes, that they cleave proteins in or near to their transmembrane domains (TMDs), and that they use a proteolytic mechanism very similar to the classical serine proteases [Reviewed in [106, 107]]. Recently, rhomboids became the first intramembrane proteases to have high resolution crystal structures solved (Figure 5B) and this, combined with earlier biochemistry and genetics, has revealed much about how these unexpected and slightly mysterious enzymes work [108–113].

Having emphasized the novelty of rhomboids as intramembrane serine proteases, it is important to point out that they are actually one of four classes of intramembrane proteases that have been identified in the last 10 years. The first to be identified was the site-2 protease, an intramembrane metalloprotease that regulates cholesterol biosynthesis via the release of a fragment of the SREBP protein [114]. The most famous is presenilin [115, 116], an intramembrane aspartyl protease, discussed above. And the fourth class is the signal peptide peptidase/SPP-like enzymes, which are mechanistically related to presenilin but have an opposite orientation in the membrane [117, 118]. These four classes of intramembrane proteases share the common property of membrane buried active sites and the ability to cleave TMDs, but are otherwise distinct and are evolutionarily unrelated. As the newest entrants to the otherwise quite mature field of proteases, much is still unknown about their mechanisms, regulation, and biological functions, but the myriad important physiological and pathological processes that they control has made them the centre of much attention.

Control of Drosophila EGFR Signaling by Intracellular ompartmentalization

The type of genetic analysis that led to the identification of rhomboid as the primary activator of EGFR signaling in flies also indicated an unexpectedly central role for regulated intracellular trafficking of proteins in the control of growth factor signaling. Another membrane protein shown to be required for EGFR signaling was Star, a single-pass transmembrane protein that, like rhomboid, had no sequence motifs to indicate its function [97, 98, 119]. Several groups discovered that Star was necessary for trafficking Spitz, the TGFα-like ligand, from the endoplasmic reticulum, to the Golgi apparatus, the location of Rhomboid-1 [120–122] (Figure 6). Upon synthesis, Spitz is retained in the ER by a poorly characterized mechanism that is dependent on PLC-γ [123]. Star is required to overcome this ER retention, thereby allowing Spitz to be cleaved in the Golgi apparatus, from where it is secreted as a soluble and active growth factor. Thus the trafficking of Spitz regulates its encounter with rhomboid-1 and ultimately EGFR activation. Recall that genetic analysis proved that Star and rhomboid-1 are the cardinal activators of EGFR activity [93, 98, 99], implying that regulated trafficking and ligand activation in the signal emitting cell is the primary strategy for control of EGFR signaling in Drosophila.

Figure 6. Intracellular Trafficking Controls Contact between the TGFα-like Ligand, Spitz, and Rhomboid-1.

A. Spitz is retained in the ER by a PLCγ-dependent mechanism until interaction with Star allows it to translocate to the Golgi apparatus (B). C. In the Golgi apparatus (or other post-ER secretory organelle), Spitz is cleaved by Rhomboid-1; Star may also be cleaved. D. Upon cleavage, the soluble form of Spitz is traverses the distal part of the secretory pathway, and secreted from the cell (E).

Recently, this regulated trafficking has been shown to have even greater potential regulatory complexity by Shilo and colleagues. They showed that Star itself could be cleaved by a Rhomboid-1-dependent mechanism, limiting its ability to be recycled to the ER and thereby reducing the amount of Spitz that can be trafficked to the Golgi apparatus [124]. Yet another level of possible control is suggested by their further observations that two other rhomboids, Rhomboid-2 and Rhomboid-3, both of which can also cleave the Drosophila TGFα-like ligands, have proteolytic activity in the ER as well as the later secretory pathway, where rhomboid-1 is confined [125]. This is supported by earlier observations that unlike Rhomboid-1, Rhomboids -2 and -3 are not fully dependent on Star to process Spitz [126, 127]. These data have led the suggestion that Rhomboids -2 and -3 can attenuate EGFR signaling by cleaving Star and Spitz in the ER, as well as acting to promote signaling by acting in the later secretory pathway [125]. It remains to be clarified what the physiological importance of this potential elaboration of the Star-dependent trafficking mechanism is, but this work highlights the overall regulatory potential of intracellular trafficking as a growth factor control strategy. More broadly, exploiting the precision of the membrane trafficking machinery to regulate the contact between rhomboids and their substrates – both integral membrane proteins – turns out to be a primary control strategy in many other biological contexts [128].

Rhomboids in Mammalian EGFR Signaling?

An obvious question is whether the essential role of rhomboid in flies is recapitulated in mammals – after all, the pathways are in other respects highly conserved. The current weight of evidence tends to suggest that this is a case where Drosophila and mammals may have evolved different molecular mechanisms for achieving the same end – the release of soluble and active growth factor from a membrane tethered precursor. As described above, there is clear and compelling evidence that in mammals, the activation of TGFα and probably other EGFR ligands, is achieved by ADAM17 and related ADAM metalloproteases [9, 14, 15]. So there is currently no obvious place for rhomboids in the process. However, absence of evidence is not the same thing as evidence of absence and there are reasons to be cautious in interpreting current data. It is worth noting that the role of rhomboids in EGFR signaling has been conserved in C. elegans [129]. Although this does not directly address mammalian biology, it does indicate that the relationship is ancient and quite widespread. More directly, an interaction has been reported between a human rhomboid-related protein without catalytic activity and EGF-like ligands [130], although the meaning of these observations remains obscure. To date, no function has been reported for any mammalian rhomboid protease, except for a sub-family of mitochondrial rhomboids [131], so there is no genetic evidence to address a potential role for rhomboids in mammalian EGFR signaling. In summary, there is no compelling evidence either to support or to reject the idea that rhomboids are involved, at least in some contexts. If they are, it will be necessary to explain how such an activity is integrated with the already well-established role of ADAM metalloproteases. Presumably, knockout mice that remove one or more of the four active rhomboids in the secretory pathway provide the best hope of answering this longstanding question.

Another aspect of the regulatory logic of EGFR signalling in flies that is poorly understood in mammals is the role of intracellular protein trafficking and compartmentalisation in controlling the production of EGF family ligands. Regardless of whether rhomboid is involved (and notably there is no obvious ortholog of Star in mammals), there have been numerous suggestions that the trafficking of ligands in mammals is complex [132–135], but the physiological relevance of this and whether this is merely a constitutive part of the biosynthetic pathway or a regulated process is not yet clear. It is notable that the overall regulatory logic of producing EGF growth factors as membrane tethered precursors, followed by regulated shedding, is conserved between Drosophila and mammals. The lack of a mammalian Star does not rule out a similarly conserved control strategy using alternative molecular mechanisms, and there is considerable potential in further analysis of the role of intracellular trafficking in the control of mammalian EGFR signaling.