Physiological processes are strictly organized in space and time. However, in cell physiology research, more attention is given to the question of space rather than to time. To function as a signal, environmental changes must be restricted in time; they need not only be initiated but also terminated. In this review, we concentrate on the role of one specific protein family involved in biological signal termination. GTPase activating proteins (GAPs) accelerate the endogenously low GTP hydrolysis rate of monomeric guanine nucleotide-binding proteins (GNBPs), limiting thereby their prevalence in the active, GTP-bound form. We discuss cases where defective or excessive GAP activity of specific proteins causes significant alteration in the function of the nervous, endocrine, and hemopoietic systems, or contributes to development of infections and tumors. Biochemical and genetic data as well as observations from human pathology support the notion that GAPs represent vital elements in the spatiotemporal fine tuning of physiological processes.
Intercellular communication occurs mostly via interaction of soluble or surface-bound ligands with specific receptor proteins localized in the plasma membrane or at intracellular sites. The receptor-ligand interaction may induce various biochemical reactions such as changes of enzyme activity, protein-protein interactions, or transmembrane ion movements, resulting in cellular responses. To transmit information, signals have to be dynamic, i.e., be terminated and reelicited with definite time courses. Studies on cellular signaling concentrate mostly on activation processes, whereas counteracting mechanisms resulting in signal termination or downregulation are less frequently investigated. However, examples of human pathology alongside with phenotypes of genetically modified animals clearly indicate the fundamental physiological importance of signal terminating processes. This review focuses on a family of proteins that can be regarded as professional cellular signal downregulators or terminators.
Guanine nucleotide-binding proteins (GNBPs or G proteins) function as regulated time switches, the general functional scheme of which is commonly summarized in the so-called GTPase cycle (FIGURE 1). In the resting state, they are usually bound to GDP. They are activated with the help of guanine nucleotide exchange factors (GEFs) that promote dissociation of GDP and rebinding of intracellularly abundant GTP. In the GTP-bound, active form they interact with specific target proteins, thereby switching on enzyme activity or induce the formation of protein complexes. Intrinsic hydrolytic activity of G proteins transforms the bound GTP into GDP, and in this inactive form, the processes regulated by the given G protein become switched off. The generally low rate of GTP hydrolysis (GTPase activity) is encoded in the primary structure of the individual G proteins and varies in a broad range. However, it is not a constant property but is dynamically modulated by regulatory proteins. GTPase activating proteins (GAPs) enhance the intrinsic GTP hydrolyzing activity of monomeric G proteins by up to five orders of magnitude (44, 365) and thereby accelerate the switch-off of the signal. Regulators of G protein signaling (RGSs) are GAPs for Gα subunits of heterotrimeric G proteins carrying out similar functions in G protein signaling.
In the present review, we direct the attention on the important physiological role played by GAPs. Our paper is by far not meant to give a comprehensive list of GAPs. In this respect, we refer the readers to excellent recent reviews (32, 39, 163, 169, 308, 351) and apologize to authors whose work we could not cite due to space limitations. We concentrate on examples where long-term and intensive research accumulated clear evidence to support the vital necessity for turning off G proteins by GAPs in the right place and at the right time. We chose only examples where a coherent picture has emerged up to now, and typical phenotypes are observable either in human pathology or in genetically modified animals. Accordingly, we grouped our examples with respect to physiological functions rather than to groups or families of molecules.
II. GUANINE NUCLEOTIDE-BINDING PROTEINS
Different types of GNBPs function at different locations in the cell. Heterotrimeric GNBPs (also known as heterotrimeric G proteins) consist of three different subunits (α, β, and γ) with the α-subunit containing the GTP/GDP binding site. They are localized to the plasma membrane where they interact with G protein-coupled receptors (GPCR). The human genome contains 17 Gα, 5 Gβ, and 14 Gγ genes (384). Ligand binding to a GPCR initiates the activation of the interacting G protein that results in exchange of GDP for GTP on the α subunit and dissociation of the active GαGTP from the complex of βγ. Both GαGTP and βγ subunits have important physiological functions such as regulation of enzyme activity (adenylyl cyclase, phospholipase C, phosphodiesterase, GRK) or modulating channel function (K+, Ca2+). Gα subunits possess high GTPase activity that allows relatively fast decay of the biological signal and reassociation of GαGDP with βγ. Physiological role and regulation of heterotrimeric G proteins have been summarized in several excellent reviews (123, 146, 239, 250, 253, 384, 385).
Monomeric (or small) GNBPs (also known as small G proteins or small GTPases) consist of one single polypeptide chain of ∼20 kDa (348). They are localized in various compartments of the cell, both in the plasma membrane and in the membrane of different vesicles but also in the cytosol and the nucleus. Most small GNBPs contain hydrophobic (COOH-terminal farnesyl or geranyl-geranyl or NH2-terminal myristoyl) chains that are important both for localization and in interaction with certain regulatory proteins (13, 152, 212, 222, 240, 279, 310). COOH-terminal modifications require the so-called CAAX-box motif (cysteine-aliphatic-aliphatic-any amino acid) and differ in the detailed nature of the lipid and attachment site (FIGURE 2) (273, 348, 390). Small GNBPs are activated by specialized regulatory proteins, the GEFs (FIGURE 1, and below). Their inherent GTPase activity is two or three orders of magnitude lower than that of Gα subunis of heterotrimeric G proteins; thus the spontaneous decay of the active conformation is very slow. GTP hydrolysis by small G proteins is dramatically accelerated by GAPs. The human genome codes for over 150 monomeric G proteins (“Ras superfamily”) (383), which participate in the regulation of almost all cellular functions.
A. Structure of the G Domain
At the heart of the biochemical activity of GNBPs lies the catalytic machinery implemented in the globular core module termed G-domain (FIGURE 2). It is made up by a central β-sheet surrounded by α-helices (365). The nucleotide (GTP or GDP) is bound in a shallow pocket on the surface of the protein that is lined by the characteristic sequence motifs G1-G5 (FIGURE 2) defining the determinants for nucleotide binding specificity and hydrolysis (46). These include the typical P-loop motif (G1) mediating most of the commonly high nucleotide affinity, the guanine base binding motif (G4, G5), and the switch regions (G2, G3) that sense the presence of the γ-phosphate and are key to the conformational differences defining GTP-bound ON and GDP-bound OFF states, respectively (FIGURE 1) (45, 46, 365). This structural arrangement has historically been described with Ras (267, 355) and is found in the vast majority of GNBPs (387). In keeping with specific physiological requirements, various structural variations are found in the various subfamilies. These include NH2- or COOH-terminal peptide or helical extensions as well as the insertion of structural elements or whole domains in loops connecting the core structural elements (365).
The GTPase reaction is generally slow, ranging from 1–2 min−1 with Gα subunits (123) to 0.039 min−1 for RhoA (316) and to virtually undetectable GTP hydrolysis for Arf proteins (170). The mechanism of GTP hydrolysis has been studied in detail in many laboratories, with Ras being the probably best-studied example. The spatial constraints imposed by the P-loop as well as the interactions of the switch regions with the β,γ-phosphate moiety are key components of the catalytic machinery and are major biochemical determinants of the pathogenicity of oncogenic Ras mutants that are unable to hydrolyze GTP at a rate sufficient to terminate the downstream signal (4, 356). In the currently accepted mechanism, Ras employs substrate-assisted catalysis with GTP itself acting as a general base of G protein-mediated nucleotide hydrolysis (312), with the nucleophilic water molecule stabilized by the conserved glutamine of switch II. After the revelation of the first GAP mechanisms (290, 309), GTPase mechanistic research with other GNBPs concentrated on the GTP hydrolysis as catalyzed by the GAP components, thus considering the catalytic machinery as a heterodimeric complex made of the GNBP and its cognate GAP (308).
B. Families and Function of Monomeric GNBPs
Traditionally, monomeric (small) GNBPs are grouped into five families defined by Ras, Rho/Cdc42/Rac, Rab, Arf, and Ran as representatives (TABLE 1), although some proteins cannot be assigned to any of these families (348). The individual families have distinct and typical functions that are summarized below. However, recent reports describe more and more cross-interaction between the traditional functions of the individual families, and the boundaries become less defined (100, 244, 281, 333).
1. Ras family
Ras genes have originally been discovered as the cellular counterpart of a rat sarcoma virus-derived retroviral gene that is associated with oncogenic transformation (94) encoding H-Ras and K-Ras. Together with the later discovered N-Ras, these proteins constitute the founding members of the canonical Ras proteins (22). Additional members of the Ras family include R-Ras, M-Ras, Ral, Rap1/2, and Rheb and a few less explored examples (348).
Ras is activated by growth factors, cytokines, and hormones via different plasma membrane receptors. First discoveries reported activation of Ras by growth factors such as EGF or PDGF via receptor tyrosine kinase (RTK) pathways involving the adaptor protein Grb2 and the exchange factor SOS (56) (FIGURE 3). Subsequently, other pathways of SOS activation have been characterized as well as other RasGEFs have been discovered, which relay Ras activation to production of the second messenger cAMP or to Ca2+ signaling (57, 231).
GTP-bound Ras couples to multiple distinct downstream effectors (FIGURE 3). Interaction is usually mediated by the so-called Ras binding domain (RBD), a ubiquitin structure like protein module that senses the nucleotide bound state of Ras mediated by the switch regions (148, 386). The first discovered and best characterized Ras effectors are the Raf kinases that control, via activation of the MAP kinase cascade, gene expression, cell proliferation, differentiation, and survival (231, 383). Another well-characterized direct effector of Ras is represented by the phosphatidylinositol 3-kinases (PI3K) that, via enrichment of phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3] in the membrane, regulate several kinases, the most versatile being perhaps Akt (also called protein kinase B). In addition to protein and lipid kinases, active Ras also interacts with regulatory proteins of other small GTPases mediating cross-talk to other superfamily members, specifically Rac and Ral (36). The most important signaling pathways initiated by active Ras proteins are summarized in FIGURE 3. On a pathophysiological level, prominent features of Ras-related signaling processes have been described in the context of tumorigenesis (22, 42, 172, 214) and in brain function (175, 220, 377).
Rap proteins (Rap1A/B, Rap2A/B, Rheb) are close Ras homologs (311) and as such are counted in the same subfamily. They have originally been found antagonizing Ras-induced transformation but can oppose other actions of Ras including regulation of cell growth and differentiation, integrin-dependent responses, and synaptic plasticity (343, 344, 407). Although closely related to Ras, they are functionally different (410). Rap proteins control primarily cell adhesion, cell junction formation, cell secretion, and cell survival (38, 41, 183). Rheb turned out to be an important element relaying growth factor signaling to translation and protein synthesis on ribosomes (215).
2. Rho family
The most prominent members of the Ras homologous (Rho) family of small GTPases are RhoA, Rac1, and Cdc42 (TABLE 1), which are ubiquitously expressed, whereas expression of Rac2 is restricted only to hematopoietic cells (328).
The major function of Rho family small GNBPs is regulation of cell morphology, cell polarity, cell motility, cell adhesion, as well as endocytosis and exocytosis. They are involved in vital biological processes such as embryonic development, tissue renewal from stem cells (e.g., skin, intestinal epithel, hematopoiesis), wound healing, immune surveillance, or influence of tumor metastasis (107, 160, 241, 359). Many of these functions are achieved via regulation of the actin cytoskeleton and the actin-binding motor protein myosin (135, 284, 285).
The first fundamental observations have been made on fibroblasts microinjected with a constitutively active form of Rho family GNBPs. Expression of constitutively active Rho induced the formation of stress fibers, i.e., parallel actin bundles within the cell body (286). Constitutively active Rac initiated formation of lamellipodia which are broad, flat cellular extensions (287). Finally, constitutively active Cdc42 induced the formation of long, thin protrusions, called filopodia (261). These phenotypical changes could be reproduced by stimulating the cells with growth factors such as PDGF, EGF, or insulin (lamellipodia) or lysophosphatidic acid (LPA; stress fibers) (135, 261).
Cell migration is directed by environmental cues such as extracellular matrix proteins, attractive or repellent chemical stimuli via integrins or plasma membrane receptors. To start directed movement, the cell has to polarize, i.e., an asymmetric alteration occurs in the distribution of membrane components. Cdc42 is the key regulator of cell polarization, and it becomes activated toward the front of the migrating cell. Cdc42 initiates the formation of filopodia that can be regarded as fine sensors of the environment. Filopodia contain long parallel bundles of actin filaments. Extension of nonbranching actin filaments is organized by the Cdc42-target proteins Diaphanous-related formins Dia 1–3 (285) (FIGURE 4). Polarization also involves typical distribution of phosphoinositides and their synthesizing and hydrolyzing enzymes: PI3K that is required for synthesis of phosphoinositides phosphorylated in position 3′ is enriched at the front, whereas the phosphatase PTEN is accumulated in the tail (223, 288, 303). Accordingly, at the leading edge an enrichment of PI(3,4,5)P3 and phosphatidylinositol 3,4-bisphosphate [PI(3,4)P2] can be typically observed.
Activation of Rac at the leading edge, via RacGEF(s) localized by receptor-initiated protein and/or lipid interactions, results in development of lamellipodia that contain a branched network of actin filaments. The branching proteins Arp2/3 and the adaptor proteins WASP/WAVE, which are direct effectors of active Cdc42 and Rac, respectively, play the central role (FIGURE 4). Lamellipodia form focal adhesions to extracellular matrix proteins or neighboring cells, which provide traction sites for forward movement. Inhibition of Rac activation prevents migration, whereas inhibition of Cdc42 activation transforms directed migration into random movements (11).
Finally, retraction of the tail involves contraction by the actomyosin complex, where a kinase regulated by active Rho (ROCK) participates in activation of myosin. Elaborate fluorescent microscopic techniques involving energy transfer between interacting proteins have indicated a fine spatial and temporal sequence in activation of the different members of Rho family GTPases in migrating cells (121, 188). In the spatio-temporal orchestration of the cellular processes, several positive-feedback loops and cross-talk between the members of Rho family GTPases have also been revealed (288).
Beyond their central role in cytoskeletal organization, Rho family members also function as direct regulators of important enzymes. Rac1 and Rac2 are key components in the enzyme complex responsible for superoxide formation in phagocytes (NADPH oxidase, NOX2) (2) and in many other tissues by its homologs NOX1 and NOX3 (30). Rho is a critical regulator of smooth muscle contraction via its effector Rho kinase (ROCK), mainly via inhibition of myosin light-chain phosphatase (289) and phosphorylation of myosin light chain (234). In addition to protein kinases, several kinases regulating phospholipid metabolism (e.g., PI4P-5-kinase, PI3K, diacylglycerol kinase, phospholipase D, or phospholipase C isoforms) have been shown to be regulated by one or several members of the Rac/Rho family (160).
Last but not least, members of the Rho subfamily of small GTPases also play a fundamental role in the control of cell cycle progression. Both Rho, Rac, and Cdc42 are able to activate the JNK and p38 MAP kinase pathways (120, 266, 352, 353), and Rac and Cdc42 were shown to stimulate cyclin D1 transcription (224, 381). This way Rho family proteins play an essential role in G1/S transition (73).
3. Rab and Arf family
Small GNBPs of these two families participate in the organization of intracellular vesicular traffic including endo-, exo-, and transcytosis as well as anterograde and retrograde traffic between ER and Golgi (25, 85, 340, 403) (FIGURE 5).
In mammals, the Rab family consists of ∼60 different proteins that are localized in the membrane of the various subcellular organelles and function in specifying membrane identity (64). Most intracellular vesicles contain more than one type of Rab protein; however, these may be segregated in different vesicular domains (336). The large number of Rab proteins and their regulatory proteins in mammalian cells serves probably the fine organization of trafficking of many different types of cargo molecules (118).
In their active, GTP-bound form, Rab proteins bind to specific effector proteins and initiate the formation of large protein complexes stabilized both by protein-protein and protein-lipid interactions. Perhaps the best studied function of Rab proteins is in tethering, i.e., specific identification of membrane compartments, that precedes vesicle docking and fusion (FIGURE 6). Tethering factors [such as the early endosome antigen 1 (EEA1) localized to early endosomes] contain coiled-coil regions and multiple binding sites for the same or different Rab proteins. In this way, they provide a first connection between two identical or different membrane compartments, a prerequisite for activation of the fusion machinery, including the SNARE complexes (340).
Whereas Rabs are present on almost every type of intracellular organelle, members of the Arf family have a more restricted function. Arf1 is localized to the Golgi membrane where it is involved in regulation of the assembly of coat complexes around budding vesicles. The other well-investigated member of the Arf family, Arf6, is localized in the plasma membrane, where it regulates the endocytic pathway via organization of both clathrin-dependent and clathrin-independent processes (85). Cooperation between Rab and Arf proteins is represented by an increasing number of reports on the existence of golgin proteins with multiple binding sites for different members of the Rab and Arf family of GTPases (58, 111).
Both Rab and Arf proteins were reported to regulate several enzymes of the phospholipid metabolism (85, 340). The resulting local changes in the phospholipid concentrations (mainly of phosphoinositides) seem to provide significant contribution to the assembly of the membrane-associated complexes.
4. Ran family
The Ras-like nuclear protein (Ran) is a key regulator of nucleocytoplasmic transport (190, 380). The active, GTP-bound form of Ran is significantly enriched in the nucleus, due to the exclusive localization of the two regulatory proteins. The Ran-GEF RCC1 (for regulator of chromosome condensation) is associated with the chromatin, whereas RanGAP prevails in the cytoplasm. RanGTP binds to the nuclear transport proteins named importins and exportins; however, this binding alters the affinity to the transported cargo in opposite directions. RanGTP enhances the dissociation of the cargo molecule from importin, but it promotes the association of cargo molecules with exportins.
The function of Ran is not restricted to the interphase, but plays important roles also in other phases of the cell cycle (71, 201). In the metaphase, RanGTP enriched around the chromatin stimulates microtubule polymerization and the assembly of the mitotic spindle, whereas a decrease in the local concentration of RanGTP impairs the alignment of chromosomes (190). In the telophase RanGTP is required for reconstruction of the nuclear membrane (71).
C. Regulatory Proteins Acting on Monomeric GNBPs
Two types of regulatory proteins act on each family of small GNBPs (39, 44) (FIGURE 1). GEFs catalyze the exchange of the bound nucleotide, and in the cellular environment, where the GTP concentration largely exceeds the GDP concentration, they bring about the activation of the small GNBPs. In contrast, GAPs accelerate the slow intrinsic hydrolysis of GTP and thereby promote the deactivation of the small G protein. The third type of regulatory proteins, guanine nucleotide dissociation inhibitors (GDIs), act only on small GTPases of the Rho and Rab family.
In the resting state, GNBPs are bound to GDP. To become activated, the usually tightly bound GDP has to be exchanged for GTP, requiring the activity of GEFs. GEFs play a similar role as the ligand-receptor complex plays in activation of the α-subunit of heterotrimeric G proteins. These proteins essentially destabilize nucleotide binding, thereby allowing release of GDP and binding of the intracellularly abundant GTP (44). GEFs are generally family specific, and the GEF domain is frequently embedded as a module of 20–30 kDa in multi-domain proteins. The best explored examples include the Ras-specific Sos, the Rho-family Dbl-homology and DOCK, and the Arf-specific Sec7 domains. GEF mechanisms have been explored intensively since the cloning of the first mammalian GEF (329) more than 15 years ago, and structural information is available for a number of examples (44, 66).
Although the GEFs for the various families are unrelated in either sequence or structure, the derived mechanisms in general appear to follow similar strategies. Accordingly, GEFs induce rearrangements of the nucleotide binding residues, especially the switch regions and the P-loop to break up the nucleotide binding pocket. Expelling the catalytic cofactor Mg2+ also seems to represent a common feature of GEF mechanisms (44).
Regulation of GEFs occurs via a variety of strategies as reflected in their multidomain modular architecture. They include protein-protein interactions, lipid binding, interaction with second messenger molecules [cAMP, diacylglycerol (DAG), Ca2+], and posttranslational modification such as phosphorylation. Localization and local activation along with phosphorylation appear to be predominant strategies how GEFs act in the right place at the right time (44). Signaling complexes consisting of a GEF, its GTPase substrate, and the effector protein of the GTPase have been identified for the RhoGEF p115RhoGEF (161) and the RacGEFs Tiam1 (76) and Cool-2/α-Pix (20), indicating the ability of multidomain GEFs to carry out also scaffolding function. In the case of p63RhoGEF, the crystal structure of the complex of the GEF, its regulator Gα and its target RhoA has been solved (211), providing the molecular mechanism for activation of a monomeric G protein by a plasma membrane G protein-coupled receptor (320, 392).
In the human genome, ∼150 genes code for proteins with potential GAP activity. The sizes of these proteins range from 50 to 250 kDa. Several GAPs are not monospecific; rather, they recognize and interact with a few small GNBPs of the same family. GAPs acting on Rac/Rho family GTPases are especially abundant: the human genome contains ∼70 potential GAPs (32, 275), which are in almost threefold excess over Rac/Rho small GTPases. Inverse relation holds for the Rab family: 38 GAPs act on 60 Rab proteins (TABLE 1).
GAP activities are normally family specific, with the catalytic machinery implemented in modules with molecular weights ranging from ∼20 to ∼50 kDa. Family specific GAP domains differ in their structural topology (fold) and catalytic mechanisms to complement the G protein active site (FIGURE 7). While Ras- and RhoGAPs employ a catalytic arginine provided in trans (249, 290, 309), genuine RapGAPs use a catalytic asparagine that is believed to play the role of the catalytic glutamine in Ras to position the water molecule or derived nucleophile for nucleophilic attack (89). In addition, mixtures of strategies have been reported for RabGAPs (269) and RanGAP (314). As a general strategy, GAPs seem to stabilize the G protein switch regions in the majority of the examples and use a catalytic residue to contribute to in trans transition stabilization (308). RGSs represent a large group of GAPs acting on heterotrimeric G protein α-subunits that appear to primarily adjust the existing catalytic machinery but apparently do not provide a catalytic residue (297).
By acceleration of GTP hydrolysis, GAPs terminate the active state of the G protein and promote the formation and prevalence of the GDP-bound, inactive form of the GTPase. Whether this results in inhibition of the regulated biological process or rather in termination of the signal depends on the rate of activation of the G proteins by GEFs or GPCRs (296). When the rate of activation largely exceeds the rate of GTP hydrolysis, then signal termination will be dominant, whereas in the case of comparable activation and hydrolysis rates, substantial inhibition of signaling will occur (45). The same GAP can inhibit signaling or enhance termination (or influence both) in different processes of the same cell (296).
The substrate specificity of different GAPs has been determined first under in vitro conditions, measuring the GTP hydrolytic activity of bacterially expressed small GTPases in the absence and presence of suspected GAPs (316, 356). Later investigations carried out in cellular models indicated that the substrate specificity determined under in vitro or in vivo conditions may diverge (68, 69, 127, 235, 373). This may be due to structures outside of the catalytic domain of GAPs or to posttranslational modification of the GAPs or the substrate GTPases.
GAPs are complex proteins containing a wide variety of different domains (FIGURE 8) that allow interactions with proteins or lipids and represent targets of many regulatory processes. In fact, a broad spectrum of regulatory processes has been described such as phosphorylation/dephosphorylation, protein-protein interaction, lipid binding, and degradation/resynthesis (33) that are able to modify the physiological effect of various GAPs. These regulatory processes may alter either the catalytic activity of GAPs, or their localization and consequently their access to the GNBP substrates. Interestingly, there are examples where phosphorylation alone or in combination with lipid binding alters the substrate preference of a GAP (200, 206, 228). In a complex system like the cell, several modifying interactions may take place simultaneously, and in many cases, clarification of the precise mechanism in comparative in vitro and in vivo studies is still missing.
In the case of several GAPs (p50RhoGAP, oligophrenin, Abr, chimerins), autoinhibition due to protein folding has been described (75, 102, 108, 162, 240) that limits the activity of the relevant GAP. Various regulatory mechanisms were shown to be required to relieve this autoinhibition (75, 102, 108, 162, 240). Alternatively, the posttranslational addition of lipid tail to the substrate GNBP was necessary to allow the interaction with the relevant GAP (240).
Different cells express their own “GAP repertoire.” Early studies have identified six different Rac/RhoGAPs in neural tissue (216) and three Rac/RhoGAPs in neutrophilic granulocytes (122). Expression of multiple GAPs with similar substrate specificity in the same cell raises the question of specific or overlapping functions. The multidomain structure and the large differences in the domain composition of individual GAPs suggest that these proteins may participate in large protein complexes and function also as scaffolds. In fact, IQGAPs, which contain a RasGAP-homology domain and bind the GTP-bound active form of Rac or Cdc42, are devoid of GAP activity and function rather as Rac- or Cdc42 effector scaffolds (48). Likewise, the breakpoint cluster region of the p85 α-subunit of PI3K is structurally homologuous to the canonical RhoGAP module but has no detectable GAP activity towards Rho family members (243).
Examples of cross-talk between small GNBPs and GAPs have been demonstrated, e.g., Rac was shown to interact with p190B thereby influencing its RhoGAP activity (59) or Arf1 was shown to recruit ARHGAP10 to the Golgi membrane (100) or the activity of RA-RhoGAP was shown to be increased by Rap1 binding (395). A few GAPs have been reported to have dual or even triple G protein specificity. These include the Ras/Rap specific SynGAP (187) and GAPIP4BP/Rasal/CAPRI (82) proteins. Equipped with a canonical RasGAP module it turned out that RapGAP activity required portions outside the catalytic GAP module (194, 276). Biochemical studies suggested that the dual specificity is mediated by components of the Rap protein itself (337). Another interesting example is provided by ARAP1 that contains both an ArfGAP and a RhoGAP domain, but the two activities are regulated separately (232) and the protein couples vesicular traffic with cytoskeletal movements (244).
GDIs bind to the GDP-bound form of the relevant small G protein and protect them from the effects of GEFs. Binding of Rho family GNBPs to RhoGDIs involves both protein-protein interactions and binding of the prenyl tail of the small GTPase into the lipid binding pocket of RhoGDI (152, 166, 310). This way, RhoGDI is able to extract its target from the plasma or intracellular membranes and sequester it in an inactive form into the cytosol. Similarly, RabGDI is able to bind Rab proteins with two geranyl-geranyl tails (13, 279) and shuttle them through the cytosol to another vesicle where GDI displacement factor (GDF) contributes to the liberation of GDI and association of Rab with the appropriate vesicular membrane.
III. ROLE OF GAPs IN THE NERVOUS SYSTEM
In the developing nervous system migration of cells (or entire cellular layers), formation and guidance of axons as well as dendritic branching and spine formation depend largely on organization of the actin cytoskeleton. The central role of various members of the Rho/Rac family of small GTPases (see sect. IIB2) in all these processes has been shown and reviewed previously (106, 128, 209, 254, 255, 280, 375).
Antagonism between Rho and Rac activation has been observed both in neurite outgrowth and dendrite formation. Typically, activation of Rac induces lamellipodia and growth cone formation, promoting neurite outgrowth, whereas activation of Rho results in growth cone collapse and neurite retraction (184). A similar antagonism has been reported in axon pathfinding directed by the guidance molecule netrin-1 (202). However, the situation may not be generalized, as growth cone collapse induced by other guidance molecules (semaphorin 3A and 3D) has been reported to be mediated via active Rac (165, 363).
In addition to effects on neurite outgrowth, Rac and Rho activity had also different effects on dendritic development and synaptogenesis: activated Rac initiated whereas activated Rho reduced the formation of dendritic spines (248, 350). Genetic manipulation of the expression of Rac/Rho GNBPs suggested that axon growth, guidance, and branching are organized by different signaling pathways (255).
It is not exclusively the Rho/Rac family of small GNBPs that plays the central role in development and functioning of the nervous system. Members of the Ras/Rap family controlling the activity of the MAPK pathway, gene expression, and protein synthesis have also been shown to be key elements (397).
Investigations deciphering the molecular pathways leading to activation of small GTPases have mainly focused on exchange factors and have successfully identified several GEFs that function in different molecular complexes (193, 260, 376). However, the activity level of any small GTPase depends on the fine balance between activating and inactivating factors, i.e., on the ratio of GEF to GAP activities. In the following sections we describe examples where alteration of the GAP activity results in disturbance of the spatiotemporal fine regulation of different small GTPases with consequent disturbance of the development and function of neural activities.
A. Axon Guidance
Development of the neural network requires precise guidance of outgrowing axons. Pathfinding is directed by attractive and repulsive cues provided by soluble or transmembrane guidance molecules such as netrins, Slits, semaphorins, or ephrins (96). Signaling through the appropriate receptor molecules in the plasma membrane of developing neurons regulates cytoskeletal dynamics resulting in elongation or retraction of the growing axon. Interestingly, the same family of receptors may mediate both repulsive and attractive signals. Small GTPases have been implicated in the signaling pathway of several guidance receptors.
1. Eph receptor signaling
Axons of the neurons in layer 5 of the motor cortex descend in the corticospinal tract and form synapses with spinal motoneurons on the contralateral side. They are prevented from crossing the midline at spinal level by repelling cues provided by ephrinB3 gradient along the midline of the spinal cord. EphrinB3 is a ligand for the tyrosine kinase receptor EphA4, and signaling via the EphrinB3/EphA4 pathway was discovered to play a specific role in formation of the motor system.
In 2007, four different teams published concordant results on involvement of α2-chimerin in EphA4 signaling. Chimerins are the products of two related genes (α- and β- chimerin), both occurring in two splice variants. They function as GAPs for the small GNBP Rac. In addition to the typical GAP domain, α1- and β1-chimerin possess a DAG binding C1 domain. The longer variants (α2- and β2-chimerin) have also an NH2-terminal SH2 domain that allows interaction with phosphotyrosines (FIGURE 8).
The four studies provide a comprehensive picture on the essential and specific role of α2-chimerin in EphA4 signaling (31, 159, 327, 378). In biochemical experiments, binding of α2-chimerin to EphA4 has been demonstrated in a yeast two-hybrid system (378), by affinity chromatography (31), and by coimmunoprecipitation (159). The interaction has been localized to the NH2-terminal SH2 domain of α2-chimerin and the juxtamembrane tyrosines of EphA4 (31). Most of the data suggest that the kinase activity of EphA4 is required for the binding and results in tyrosine phosphorylation of α2-chimerin (31). EphA4 signaling reduced the level of active Rac, and this effect depended both on tyrosine phosphorylation and on the GAP activity of α2-chimerin (327).
In α2-chimerin knock-out animals, a remarkable change in motoric coordination has been observed: the alternating movement of the hindlimbs has been changed to synchronized movement, and the mice showed a rabbit-like hopping movement (FIGURE 9). In histological analysis, recrossing of the descending corticospinal axons has been shown in the spinal grey matter. Electrophysiological data are consistent with the histological finding: in α2-chimerin−/− animals unilateral stimulation of the motor cortex resulted in bilateral movement and EMG activity in the hindlimbs (31). Recrossing of the midline was not restricted to descending corticospinal axons but could be observed also in the case of axons of spinal interneurons resulting in dysfunction of left-right coordination of spinal central pattern generator neurons (378). The described alteration in motoric control is a phenocopy of that observed earlier in EphA4- or ephrinB3-deficient animals (77, 99, 191, 192). Recrossing of the midline was restricted to motoric axons, whereas proprioceptive sensory projections (which depend on plexinA1 guidance) have not been disturbed in α2-chimerin−/− animals (31).
Neurons isolated from the motor cortex of wild-type animals show typical growth cone collapse upon treatment with ephrinB3. Ephrin-induced growth cone collapse was significantly impaired in cultured neurons of α2-chimerin−/− animals (31, 378). This phenomenon could be reproduced in cultured neurons by silencing α2-chimerin expression or by expressing GAP-deficient α2-chimerin (327). The need of local reduction in Rac activity for growth cone collapse is consistent with the general observation that activated Rac induces axon outgrowth (see above).
The remarkable phenotype of α2-chimerin−/− animals developed in spite of the undisturbed expression of α1-chimerin (31). Taken together, the presented data suggest that the specific interaction between EphA4 receptor and the RacGAP α2-chimerin results in spatially restricted downregulation of Rac that is inevitable for correct transmission of repellent cues presented by ephrinB3 gradient in the midline of the spinal cord.
The role of α2-chimerin in axon guidance has been supported by observations made in cases of a human disorder, the Duane's retraction syndrome (233). This congenital eye movement disturbance is caused by aberrant innervation of extraocular muscles by axons of the abducens and oculomotor nuclei resulting typically in restricted outward gaze (FIGURE 10). In several patients, hypoplasia of the given nerves could be demonstrated by magnetic resonance imaging. Genetic analysis of the affected families located the mutation to CHN1, the gene of α-chimerins. Three of the seven identified mutations localized to the segment that is only present in α2-chimerin; thus the probable cause of the disease is a misfunction of α2-chimerin. In vitro analysis of the proteins with the identified mutations revealed that all of them were gain-of-function mutations, mainly due to increased binding to membrane phospholipids via the C1 DAG-binding domain of α2-chimerin. Overexpression of α2-chimerin in the oculomotor neurons of developing chick embryos resulted in premature termination of the oculomotor axons. In which receptor pathway is α2-chimerin involved in cranial nerves has to be determined later. However, the observations are consistent with previous findings on genetically modified mice: loss-of-function of α2-chimerin caused aberrant axon growth and loss of sensitivity to repellant cues, whereas gain-of-function mutations of α2-chimerin cause premature termination of axon growth.
Most recent data have revealed the role of the RapGAP TSC2 (or tuberin) in downstream signaling of neuronal EphA receptors (257). Stimulation of retinal ganglion cells (RGCs) with ephrinA1 decreased the ERK1/2-dependent phosphorylation of TSC2 with consequent increase of the GAP activity and inactivation of the small GTPase Rheb. Activity of Rheb is a crucial factor in regulation of the mTOR pathway and protein synthesis. Increased mTOR signaling was demonstrated in RGC upon expression of a constitutively active mutant of Rheb (257). The role of local protein synthesis in growth cone dynamics and axon guidance has been demonstrated earlier (50, 208). Growth cone collapse induced by ephrinA1 was reduced in RGCs both by expression of a constitutively active mutant of Rheb and by silencing of TSC2. These data suggest that a decrease of local protein synthesis as a consequence of decreased activity of the Rheb-mTOR pathway contributes to the growth cone collapse induced by EphA receptor ligands. In line with this hypothesis, Tsc-deficient mice (Tsc+/−) had increased mTOR activity in the axons of the RGCs and aberration of the axonal projection to the lateral geniculate nucleus has been detected (257). Mutations of TSC2 have been observed in patients suffering from tuberous sclerosis. In addition to tumor growth (see sect. VII), these patients also develop epilepsy, autism, or intellectual disabilities, which may be consequences of aberrant axon guidance.
2. Semaphorin signaling
Another family of axon guidance molecules is constituted by semaphorins that interact with plexin receptors. The biological function of semaphorins is not restricted to the nervous tissue, as they also play a role in shape changes and migration of epithelial and endothelial cells.
Semaphorin-induced responses were shown to depend on the expression of p190A (GRLF1 in FIGURE 8), a GAP acting on Rac/Rho family small GTPases (23). By coimmunoprecipitation, a direct interaction between p190A and both plexin A1 and B1 could be demonstrated. Furthermore, stimulation of the cells by semaphorin ligand transiently increased the association of p190A with the plexin receptor consistent with a transient decrease in the level of active RhoA (23). Depletion of cellular p190A levels impaired or prevented several semaphorin-stimulated responses such as the neurite outgrowth in PC12 neuroblasts, collapse and integrin-mediated adhesion of fibroblasts, migration of SKBR3 mammary carcinoma cells, or repulsion of primary endothelial cells. Collapse and adhesion of fibroblasts could be rescued by expression of wild-type but not GAP-deficient p190A in p190A−/− cells, indicating that the GAP activity of p190A was required for the semaphorin-dependent response.
Impairment of neurite outgrowth in p190A-deficient cells is consistent with findings in p190A knockout animals (52). This mutation is perinatally lethal due to multiple and serious disturbance of neural development. Histological analysis of the embryos indicated inhibition of neurite outgrowth and guidance at several locations of the brain such as subcortical and cortical axons or the anterior commissure where projections do not reach and do not cross the midline (51, 52).
3. Axon branching
While Rho-family members control the assembly of cytoskeletal structures during neural development, Ras-like GNBPs seem to be responsible for the promotion of signal transduction pathways involved in axonogenesis (136). The consequences of the loss of RasGAPs in neuronal development have been investigated specifically with neurofibromin and SynGAP in mice. Loss of neurofibromin in neurons leads to increased axon collateral branching after dorsal root injury, correlating with elevated activation of the MAPK pathway (294). Neurofibromin is also required for barrel formation in the mouse somatosensory cortex (210), and its ablation in neurons induces abnormal development of the cerebral cortex and reactive gliosis in the brain (408). While in those studies the importance of the RasGAP was primarily derived from upregulation of the MAPK pathway, a study in PC12 cells demonstrated the role of the GAP domain of neurofibromin for neurite outgrowth by overexpressing an inactive mutant, which resulted in impaired neurite outgrowth (400).
B. Formation of Dendritic Spines
Mental retardation has been shown to be related to disturbance of the size and density of dendritic spines, the major sites for excitatory synapses in the central nervous system (280). The first gene associated with X-linked mental retardation was OPHN1, which codes the protein oligophrenin-1, a GAP for Rho family small GNBPs (34). All the mutations identified in OPHN1 are loss-of-function alterations with decreased or absent expression of oligophrenin-1 (401).
Oligophrenin-1 shows a widespread distribution in the nervous system, present both in dendritic shafts and spines and axons and axon terminals (127). Downregulation of oligophrenin-1 in CA1 pyramidal neurons in hippocampal slices resulted in shortening of dendritic spines that was detectable already 48 h following the treatment. After 8 days also the spine density was significantly decreased (245). Both the morphological and the behavioral and memory changes could be reproduced in oligophrenin-1 knockout mice (176).
Oligophrenin-1 is able to enhance both in vitro and in vivo the GTPase activity of all three major members of the Rho/Rac family, i.e., RhoA, Rac1, and Cdc42 (127). However, the fact that the observed morphological changes of dendritic spines could be mimicked with constitutively activated Rho and prevented by inhibitors of Rho kinase, suggests that oligophrenin-1 affects spine morphology by reduction of local Rho activity.
In a recent study, oligophrenin-1 was shown to be enriched in dendritic spines upon activation of NMDA receptors and to be coimmunoprecipitated with the AMPA receptor subunits GluR1 and GluR2 but not with the NMDA receptor subunit NR1. Overexpression of oligophrenin-1 enhanced whereas its downregulation decreased synaptic transmission via AMPA but not NMDA receptors. At the same time, an increase or a decrease of spine density could also be observed. All these effects depended on the intact GAP activity of oligophrenin-1. Both expression of oligophrenin-1 and inhibition of Rho kinase decrease the NMDA-induced internalization of the AMPA receptor subunit GluR2. Thus oligophrenin-1, by local downregulation of Rho activity, seems to stabilize AMPA receptors and thereby it promotes synaptic transmission and spine formation (245).
In addition to its effects on the postsynaptic side, important function has been ascribed to oligophrenin-1 also in presynaptic axon terminals (247). It has been demonstrated that reduction or lack of oligophrenin-1 impairs recycling of synaptic vesicles by delaying endocytosis. At high-frequency stimulation, an impairment of synaptic transmission could be revealed when oligophrenin-1 was downregulated in the presynaptic neurons.
Taken together, presently available information indicates that downregulation of Rho activity by oligophrenin-1 allows replenishing of synaptic vesicles on the presynaptic side and stabilizes AMPA receptors on the postsynaptic side. Both effects are critical for efficient synaptic transmission. Reduced expression of oligophrenin-1 in case of mutation of the OPHN1 gene impairs transmission and synaptogenesis in developing brain, explaining (at least partially) the appearance of mental retardation.
Morphology of dendritic spines is also affected by α1-chimerin (60). This shorter splice variant appears during embryogenesis at a later timepoint than its longer homolog, α2-chimerin, but its expression is continued in adulthood. Expression of α1-chimerin is strongly dependent on neuronal activity as pharmacological blockade can decrease its expression by 60% within 48 h (60). Activation of mGluRs or mAChRs induced detectable translocation of α1-chimerin to the plasma membrane. This effect depended on phospholipase C (PLC)-β activity and the ability of α1-chimerin to bind to DAG. Overexpression of α1-chimerin in Purkinje cells resulted in ∼50% reduction in dendritic length and branching points (60), and this pruning effect depended on both the GAP activity and ability of DAG binding. In contrast, downregulation of α1-chimerin increased the length of dendritic protrusions. These findings are consistent with the general observation that activated Rac increases spine generation (see above). Although the functional consequences of alteration of dendritic morphology are not known, α1-chimerin seems to be involved in regulation of the ratio of active Rac and Rho at defined locations in the nervous system.
C. Development of the Vestibular System
Mice lacking both Bcr and Abr, two similar RacGAPs (see details in sect. VB1), exhibit a distinct phenotype characterized by clumsy movements, hyperactivity, and persistent circling (167, 168). Disturbances have been detected both in cerebellar and in vestibular development.
In mice, cerebellar development occurs within the first 3 postnatal weeks when granule cells typically migrate from the surface of the cerebellar cortex below the Purkinje cell layer. This migration is dependent on processes of the Bergmann glial cells.
At birth, the cellular structure of the cerebellar cortex did not differ in wild-type and Bcr plus Abr double null mutant animals. However, serious differences occurred within the first 2 weeks: in double null mutant animals, granule cells were found in an aberrant location on the cerebellar surface (167). Bcr and Abr are normally expressed both in Purkinje and granule cells as well as in astrocytes. Detailed morphological investigation could not reveal any alteration in Purkinje cells or granule cells of double null animals, whereas serious defects were observed in organization of cerebellar glial cells. These cells showed also an increased level of activated, GTP-bound Rac. Consistent with Rac hyperactivation, double null astrocytes showed increased spreading, had elevated basal phosphorylation of p38 MAP kinase, and reacted more vigourously to activation by lipopolysaccharide (LPS) or EGF. How far these morphological and biochemical alterations affect electrophysiological activity and coordination of cerebellar functions has not yet been determined.
In addition to cerebellar disturbances, also a dysgenesis of the vestibular system has been observed in Abr plus Bcr double null mutant animals (168). The neuroepithel was found to be detached from the underlying connective tissue in the utriculus and sacculus, and otoconia were absent in the double null animals. No similar alteration was observed either in the semicircular canals or in the cochlea, and the animals were not deaf.
Further research is needed to clarify whether there is any common cellular mechanism underlying the two detected aberrations affecting movement coordination in Bcr−/− × Abr−/− animals. However, it is striking that in spite of general expression of both RacGAPs in various parts of the nervous system, specifically two locations related to motor coordination are affected by defective expression of both Bcr and Abr.
D. Cognitive Functions
Since the MAPK pathway has been demonstrated to be involved in the regulation of memory formation and synaptic plasticity (220, 377), it is probably not surprising that deficiency of the RasGAP neurofibromin (encoded by the tumor suppressor gene NF1) is associated with impairment of cognitive functions such as learning and memory (3, 283, 298). Mice in which one neurofibromin copy has been genetically inactivated show clear deficits in spatial learning in a water maze (332). Furthermore, mice in which the alternatively spliced exon 23a has been deleted show learning deficits but normal development and tumor predisposition (79). Exon 23a encodes a 23-residue insertion within the GAP domain (258) that has been reported to affect the GAP activity in vitro (14). These results suggested that modulation of GAP activity by the presence of exon 23a encoded sequence is responsible for the observed learning deficits (79). Decreasing Ras activity by genetic or pharmacological manipulation in vivo rescues the learning deficits caused by the deletion of exon 23a, supporting the idea that indeed hyperactive Ras is responsible for these defects (78). Lack of neurofibromin in Drosophila is also associated with a learning phenotype (131). The importance of the GAP-related domain and its activity for long-term memory in flies has been demonstrated by testing known GAP-impairing NF1 mutations in long-term memory tests (151). Specifically, learning deficits in mice are due to increases in ERK activation leading to increased levels of synapsin phosphorylation, enhanced GABA release, and as a consequence inhibition of LTP and learning deficits (81). Apart from neurofibromin, the brain specific SynGAP has also been demonstrated to regulate ERK/MAPK signaling, synaptic plasticity, and learning (182), consistent with both neurofibromin and SynGAP being components of the NMDA-receptor proteome (155).
IV. ROLE OF GAPs IN THE ACTION OF INSULIN ON GLUCOSE UPTAKE
One of the most characteristic effects of insulin is the stimulation of glucose uptake into fat and muscle cells. This action is achieved by enrichment of a specific isoform of glucose transporters, the GLUT4 in the plasma membrane (FIGURE 11). Prevalence of GLUT4 on the cell surface depends on the dynamic equilibrium of exocytosis and endocytosis of specialized membrane vesicles. Insulin accelerates the exocytosis and impairs the endocytosis of GLUT4-containing vesicles. In skeletal muscle cells also physical exercise is able to mobilize GLUT4 transporters to the cell surface.
Although the above facts were known for decades, the signaling pathway from insulin receptors to mobility of GLUT4-containing vesicles remained an enigma until the 21st century. Several elements such as insulin receptor substrate (IRS), PI3K, and Akt, a PI3K-dependent serine/threonine kinase, have been identified in the signaling pathway from insulin receptor to GLUT4 translocation. The breakthrough came in 2002–2005 with the discovery of AS160, an Akt substrate of 160 kDa present in adipocytes and muscle cells (171). AS160 was shown to become phosphorylated by Akt on several amino acids, and this phosphorylation (especially that of T642) was essential for insulin-dependent translocation of GLUT4 (306, 402). The most intriguing discovery was that AS160 proved to be a GAP for several Rab proteins, and the GAP activity was required for insulin-dependent regulation of glucose transport (103, 306). Knocking down of AS160 (or TBC1D4 in the RabGAP nomenclature; referred to in this review as AS160/TBC1D4) resulted in an increase of GLUT4 on the surface of resting cells and a decrease in insulin-stimulated exocytosis of GLUT4 and glucose uptake (103). This phenotype could be reversed by expression of wild-type AS160/TBC1D4 but not by a protein where the critical arginine of the GAP domain has been mutated (103). Expression of a protein where the Akt phosphorylation sites were mutated to alanine prevented the insulin action, but this could be compensated by a GAP-domain mutant (306).
AS160/TBC1D4 was shown to be an efficient GAP for Rabs 2A, 8A, 10, and 14, and all these Rabs have been detected in GLUT4-containing vesicles (198, 225). In adipocytes, Rab10 seems to be the major target of AS160/TBC1D4 on GLUT4 vesicles as overexpression of hydrolysis-deficient Rab10 increased, whereas knocking down endogenous Rab10 decreased the insulin-stimulated translocation of GLUT4 (305, 307). Interestingly, in skeletal muscle cells, Rab8A seems to be the substrate of AS160/TBC1D4 (156, 157).
With the effect of insulin and other agonists of GLUT4 translocation (such as adiponectin, ALCAR [5-aminoimidazole-4-carboxamide-1-β-d-ribofuranoside], berberine or physical exercise) taken together, the picture arises (FIGURE 11) that constant RabGAP activity of AS160/TBC1D4 is required for retaining GLUT4 vesicles in the cell. Insulin stimulation results in phosphorylation of AS160/TBC1D4 on the critical T642 and several other amino acids by Akt. Apparently, phosphorylation by itself is sufficient for a decrease of the RabGAP activity of AS160/TBC1D4 (341). Promoting the GTP-bound, active form of critical Rab proteins (Rab10 in adipocytes, Rab8A in muscle cells) allows the correct docking and promotes fusion of GLUT4-containing vesicles with the plasma membrane.
Two recent clinical observations support the above mechanism. On one hand, it is well known that glucocorticoids induce insulin resistance and impair glucose uptake. In human and murine adipocytes, it has been found that dexamethasone treatment reduced the insulin-stimulated phosphorylation of AS160/TBC1D4 at T642 and brought about a proportional impairment of GLUT4 translocation and glucose uptake. No change could be detected in insulin-induced phosphorylation or activity of Akt (256).
Finally, in a genetic screen of severely insulin-resistant patients, a truncation mutation of AS160/TBC1D4 has been found in a young patient characterized by acanthosis nigricans and extreme postprandial hyperinsulinemia (88). The RabGAP domain of AS160/TBC1D4 has been deleted due to the truncation. Expressing this mutant protein in adipocytes resulted in increased basal and reduced insulin-stimulated GLUT4 levels on the cell surface. The clinical findings support the physiological importance of the RabGAP activity of AS160/TBC1D4 in regulation of GLUT4 trafficking and cellular glucose uptake.
Following the discovery of TBC1D4, a homolog has also been revealed. The new protein, TBC1D1, is ∼50% identical to AS160/TBC1D4 and has the same specificity for Rabs (291); however, it is preferentially expressed in skeletal muscle (63). Phosphorylation by Akt or by AMP-dependent protein kinase (AMPK) reduces the RabGAP activity of TBC1D1 and increases GLUT4 translocation to the plasma membrane (274, 291). The importance of TBC1D1 seems to be minor in adipocytes, but it may have a more significant role in skeletal muscles in transmitting signals on physical activity (63). The fact that R125W variant of TBC1D1 has been genetically associated with obesity (342) indicates its physiological relevance.
Thus both the results of physiological experiments and the available clinical observations support the critical role of RabGAPs in controling glucose uptake into adipocytes and muscle cells even if the two major metabolic compartments carry out their basically similar job with partially different molecular sets.
V. ROLE OF GAPs IN BLOOD CELLS
Formation and physiological function of blood cells require correct adhesion and migration, processes where the involvement of small GTPases of the Rho/Rac family has been amply documented (160, 241, 359) (and sect. IIB2). In human blood peripheral granulocytes, three different GAPs acting on Rho family GTPases (p50GAP, Bcr, and p190GAP) have been identified at the protein level (122). In addition, the presence of β2-chimerin has been documented in T lymphocytes (61, 330, 331). Information on the potential function of these proteins derives from experiments on genetically modified animals or cell lines.
The protein of 50 kDa molecular mass named alternatively as Cdc42GAP or p50RhoGAP (or RhoGAP1 in FIGURE 8) was the first Rho/Rac family GAP that had been cloned (24, 196). When tested with prenylated small GTPases, it reacts under in vitro conditions equally well with Rac, Rho, or Cdc42 (235). It will be referred to in this review as p50GAP.
Genetic deletion of p50GAP resulted in mice of remarkably small body and organ size at birth and high (∼90%) perinatal lethality, although the ratio of homo- and heterozygous animals followed the Mendelian distribution (373). In the investigated tissues, the active form of Cdc42 was elevated, whereas the level of active Rac or Rho has not been altered, suggesting that the major substrate of p50GAP was Cdc42 (346, 373, 374).
An analysis of the hematopoietic system of the p50GAP-deficient animals revealed serious disturbance of hematopoiesis with remarkable decrease in total bone marrow cell number as well as in the number of stem and progenitor cells. Erythropoietic precursors were more seriously affected than granulocytic precursors, although in peripheral blood, both red cells, neutrophils, and platelets were decreased (374).
Investigating the potential reasons of disturbed hematopoiesis, it was revealed that the cell cycle progression of wild-type and p50GAP-deficient cells did not differ, but the deficient cells showed significantly enhanced apoptosis. Increased apoptosis depended on the elevated activity of JNK in p50GAP-deficient cells and could be prevented by downregulation of JNK.
A mixture of hematopoietic stem and progenitor cells isolated from p50GAP−/− animals showed impairment in adhesion to fibronectin through integrin α4β1 and in migration directed by chemoattractive gradient. In line with defective adhesion and migration, hematopoietic stem and progenitor cells of p50GAP-deficient animals showed significantly reduced engraftment when transplanted into irradiated wild-type animals (374).
Peripheric granulocytes deficient in p50GAP also showed decreased chemotactic migration, whereas their random motility has been increased (346). In a gradient of chemotactic agent, wild-type cells formed membrane protrusions at the leading edge and moved towards the source of the stimulant, whereas p50GAP-deficient granulocytes formed multiple extensions in various directions and moved in random directions, often away from the chemoattractant source (346). These findings are consistent with previous observations on the key role for Cdc42 in orientation of the cells. In the absence of p50GAP, spatiotemporal control of Cdc42 activation is apparently disturbed, resulting in randomized orientation attempts of the cell. In contrast to disturbance of orientation, p50GAP-deficient PMN showed no alteration in adhesion to fibrinogen, and their directed migration on fibrinogen surface or across the epithelial layer has even been increased (347). These findings indicate that organization of cell orientation and motility depends on different signaling processes.
B. Phagocytic Functions
1. Bcr and Abr
These two GAPs have significant similarity: both possess a COOH-terminal RhoGAP domain and a tandem DH-PH domain characteristic for Rac/Rho family GEFs (FIGURE 8). In fact, the isolated DH domain of both Bcr and Abr was shown to have moderate GEF activity toward Cdc42, Rac, and Rho (69). The gene of Bcr is subject to crossing over with the gene of Abl tyrosine kinase forming the Philadelphia chromosome, and the fusion protein is involved in development of leukemia (91, 144, 370). However, the GAP activity of the proteins is not related to the development of leukemia.
Bcr was the first Rac/Rho family GAP that has been studied in a genetically deficient animal model (371). Under neutral conditions, the mice had no significant phenotype. However, they reacted vigorously upon exposition to LPS: they developed serious intestinal necrosis, lost weight, and died within a few days. An increase of reactive oxygen species (ROS) production and tissue damage due to agressive phagocytes underlie the clinical symptoms. In contrast to Bcr, Abr-deficient mice had no overt phenotype (68).
Phagocytes represent a double-edged sword: their activity is inevitable for elimination of bacteria, but in the case of pathological activation, they are able to cause serious damage to the host. Hyperactivity of phagocytic cells underlies the development of sepsis. The remarkable phenotype of Bcr−/− animals initiated the detailed investigation of the role of the two related GAPs in phagocytic cells. In phagocytes, both migration, engulfment of particles, and granule exocytosis depend on extensive and precisely organized changes of the actin cytoskeleton mediated by Rac/Rho family small GTPases. Furthermore, production of superoxide, the precursor to all other reactive oxygen species, is carried out by the NADPH oxidase, and one of the essential subunits of this enzyme is Rac (depending on the cell type and species Rac1 or Rac2) (368).
Under in vitro conditions, the isolated GAP domain of Bcr and Abr reacted with Rac and Cdc42 (69, 145) and in case of prenylated GTPases, also with Rho (207). However, when overexpressed in cells, both Bcr and Abr reduced the amount of active, GTP-bound Rac but did not affect either RhoGTP or Cdc42GTP (68) indicating that in phagocytes, both GAPs recognize mainly Rac as a substrate.
Macrophages isolated from double knockout (Abr × Bcr)−/− animals showed a remarkable, elongated morphology and significant increase in directed motility both in vitro and in vivo. In addition, their phagocytic capacity has been increased, and prolongation of the prevalence of RacGTP was observed (68). Also, PMA-stimulated superoxide production and LPS-induced release of the proteolytic enzyme matrix myeloperoxidase 9 has been elevated (83). Neutrophilic granulocytes of double knockout animals showed enhanced release of primary granules, but no change was observable in release of secondary granules or secretory vesicles (83).
These alterations of phagocytic functions contribute to the development of the serious clinical condition that was observed in these animals in two different models of sepsis. Administration of Escherichia coli LPS induced acute inflammation of the lungs, characterized by cellular infiltration and fluid leakage. The same symptoms occurred in both single-knockout animals, but they were more serious in double knockouts. Whereas in wild-type and single-knockout animals a tendency of recovery was evident after 48 h, the double knockouts showed no sign of recovery and died within 2 days. Similar enhancement of intestinal damage has been observed after cecum ligation and puncture of (Bcr × Abr)−/− animals (83).
The detailed investigation of phagocytic functions in single-knockout animals indicated that Bcr and Abr participate in similar reactions; however, slight differences have been observed both in superoxide production and in enzyme release. Thus the function of the two similar GAPs seems to be only partially overlapping, and both molecules are required to keep the reactivity of phagocytic cells under control and prevent extensive tissue damage in case of bacterial infection.
p190A [also known as glucocorticoid receptor DNA binding factor 1 (Grlf1, FIGURE 8) or p190RhoGAP] is one of two (p190A and p190B) homologous GAPs acting on Rac/Rho family GTPases (318), yet regulating different functions.
P190A is phosphorylated by Src-family tyrosine kinases upon stimulation by growth factor receptors or adhesion proteins (52, 104, 246), and this phosphorylation promotes the association of p190A with p120RasGAP directing its relocalization from the cytosol to the cell periphery (47, 153, 237). P190A is an ubiquitous protein (319), and it was shown to be involved in decreasing Rho activity upon integrin signaling in fibroblasts (17, 27, 147, 246). It also plays a central role in directing axon outgrowth of neurons (51, 52), and its association with semaphorins has been detailed above (see sect. IIIA2).
Expression of p190A has been shown in neutrophilic granulocytes both at the mRNA and at the protein level (122, 251). Translocation to the plasma membrane has been detected in stimulated neutrophils (101), and superoxide production was decreased by p190 under in vitro conditions (149). In human neutrophils, Src-dependent translocation and activation of p190A was observed upon β2-integrin stimulation (95). However, in contrast to these findings, a careful investigation of the phenotype of mice transplanted with p190A−/− bone marrow has not revealed any significant alteration in the in vitro or in vivo properties or activity of neutrophils or in the development of autoimmune arthritis (251). These data suggest a cell-type or species specific role of p190A in integrin signaling.
C. T-Cell Activation
Expression and RacGAP activity of β2-chimerin has been shown in Jurkat cells, a T lymphocyte cell line. Stimulation of the cells with high concentration of PMA induced a redistribution of β2-chimerin from cytosolic localization to the cell membrane, and this redistribution depended on binding of the chimerin C1 domain to membrane DAG (330). A detailed investigation revealed an important role of β2-chimerin in signal transduction following stimulation of the T-cell receptor (TCR) (61, 331). Overexpression of β2-chimerin inhibited both adhesion and interfered with activation of the key regulator NF-AT upon stimulation of TCR. These effects depended on the ability of β2-chimerin to downregulate RacGTP (61, 330). Upon activation of the TCR, β2-chimerin was shown to be tyrosine-phosphorylated by Lck, a receptor-dependent Src family tyrosine kinase. This phosphorylation interferes with the binding of the C1 domain of β2-chimerin to the membrane phospholipid DAG and reduces the RacGAP activity of β2-chimerin. Lack of phosphorylation of β2-chimerin results in enhanced RacGAP activity with consequent decrease of the level of active RacGTP and functional alterations such as reduced activation of NF-AT and decreased IL-2 production upon TCR stimulation (331). Hyperactivated forms of β2-chimerin (either due to blocked tyrosine phosphorylation or release of autoinhibition) interfere with formation of the immunologic synapse between T-cell and antigen-presenting cells contributing to impaired TCR stimulation.
Interestingly, TCR stimulation, via Lck, also contributes to activation of Vav1, a GEF for Rac (6, 137, 391). Thus phosphorylation via Lck influences the level of active RacGTP on two parallel pathways: both by activation of the RacGEF Vav1 and by inhibition of the RacGAP β2-chimerin. The rapid elevation of local [RacGTP] is apparently essential for full activation of T lymphocytes, and interference with one of the two parallel Rac-activating pathways results in detectable deficit of the biological function (331).
VI. ROLE OF GAPs IN HOST-PATHOGEN INTERACTIONS
The infectious capacity of a potentially pathogenic microorganism depends on its ability to enter and invade epithelial layers and to succumb or evade various protective mechanisms of the innate and adaptive immune system. Small GTPases of the Rho/Rac family play key roles in maintaining the integrity of the epithelial layer, in phagocytosis, and in production of ROS or cytokines, all being functions that determine the fate of an invading microorganism. It is thus no surprise that microorganisms have evolved a broad variety of effective methods that target small GTPases, and several of these molecules have GEF or GAP activity for members of the Rho/Rac family. Remarkably, the GAP or GEF modules encoded by these proteins employ similar regulatory mechanisms like their eukaryotic counterparts yet implemented in structural entities unrelated to those of the eukaryotic Rho family regulators (116, 238, 338, 393).
A. Pathogenic Yersinia Species
Three species of the Yersinia genus are pathogenic for humans and animals: Y. enterocolitica and Y. pseudotuberculosis cause enterocolitis and mesenteric lymphadenitis, whereas Y. pestis is the pathogenic agent of plague. All three pathogenic Yersinia species contain a virulence plasmid (pYv) that encodes the constituents of a type III secretion system (TTSS). TTSS are characterized by a needlelike projection of the bacteria that contacts the host cell. The proteins secreted via the TTSS are elements of a translocation channel or effector proteins injected into the cytoplasm of the host cell. Pathogenic Yersinia species secrete six different effector proteins, named Yersinia outer protein (Yop) E, T, O, H, J, and M. All of the Yops have remarkable functions within the host cell, affecting mostly the cytoskeleton.
YopE has in vitro GAP activity toward RhoA, Rac1, and Cdc42 but not other small GTPases (35, 369). In vivo GAP activity towards RhoA, Rac1, Cdc42, and RhoG has been confirmed in HeLa and human umbilical vein endothelial cells (HUVEC) (9, 295). GAP activity of YopE depends on the critical arginine in position 144 (35, 369).
Injection of YopE into monolayers of cultured HeLa or HUVEC results in disruption of the actin filaments, rounding up and eventually detachment of the cells (15, 35, 369). At the same time, uptake of the bacteria is seriously impaired (35). All these effects were shown to depend on the GAP activity of YopE as they are absent if the catalytically inactive R144 mutant toxin is injected and could be prevented by expression of constitutively activated Rac1 and RhoA (35, 369).
In addition to YopE, several other Yersinia toxins target the actin cytoskeleton via small GTPases. YopT is a cysteine protease that is able to remove the prenyl tail mainly from RhoA and to a lesser degree also from Rac and Cdc42 (322). As a result, YopT induces the release of RhoA from the plasma membrane and disruption of stress fibers in infected cells (5, 409). YopO was shown to have serine/threonine kinase activity, and on this basis, it has been named also Yersinia protein kinase A (YpkA). Recently, the crystal structure of the COOH-terminal part of YopO revealed a domain homologous to RhoGDIs, and this region of YopO was shown to bind to and inhibit exchange on Rac1 and RhoA (278). Thus the concerted action of YopE, YopT, and YopO is able to inactivate the Rho-family GTPases, remove them from the plasma membrane, and sequester them in a protein complex in the cytosol.
Concerted action of the Yops results in inhibition of the respiratory burst; impairment of phagocytosis; killing of pathogenic Yersinia species by macrophages and neutrophils (140, 302); inhibition of secretion of pro-inflammatory mediators such as IL-8, tumor necrosis factor (TNF)-α, or interferon (IFN)-γ (53, 367); and induction of apoptosis of macrophages and dendritic cells (301, 405). As a result, enteropathogenic Yersinia species colonize typically the Peyer's patches and mesenteric lymph nodes where they replicate in an extracellular form. Highly virulent strains are also able to disseminate systemically and reach the spleen or liver. Isolated deficiency in YopE decreased virulence and systemic spreading of the bacteria more vigorously than deficiency in YopT or YopO (366). Thus the GAP activity of YopE towards Rac/Rho family GTPases significantly contributes to the overall pathogenicity of the Yersinia species.
B. Pseudomonas aeruginosa
Pseudomonas aeruginosa is an opportunistic pathogen bacterium that causes severe infections such as pneumonia typically in patients with cystic fibrosis, burn wounds, or immunodeficiency (37). When investigated in cell culture, P. aeruginosa was shown to exert a cytotoxic effect characterized by rounding and detachment of epithelial cells, disruption of the actin cytoskeleton, decrease of viability, and impairment of phagocytosis (114).
Toxicity of P. aeruginosa depends on four effector proteins (ExoU, ExoS, ExoT, and ExoY) secreted via a TTSS similar to that of Yersinia species. The biochemical activity of all four exotoxins has been determined: ExoU has phospholipase A2 activity, ExoY functions as an adenylate cyclase, whereas ExoS and ExoT are bifunctional toxins. Both ExoS and ExoT have ADP-ribose transferase (ADPRT) activity localized to the COOH-terminal part of the molecules, whereas the NH2-terminal part of both ExoS and ExoT was shown to have in vitro GAP activity for Rho, Rac, and Cdc42 GTPases (124, 185). In vivo GAP activity of ExoS has been confirmed both in epithelial type cells (186, 404) and macrophages (292) and for ExoT in HeLa cells (174). Plasma membrane localization is essential for the in vivo GAP activity of ExoS (404), and cooperation was shown between ExoS GAP activity and intracellular RhoGDI in effects upon the actin cytoskeleton (345).
Both enzyme activities contribute to the toxic effects of ExoS and ExoT of P. aeruginosa. ADP-ribosylation of small GTPases and other proteins playing a role in organization of the actin cytoskeleton seems to be responsible for impairment of the barrier function of epithelial surfaces. Antiphagocytic effects allowing the bacteria to avoid intracellular killing and degradation seem to be more attributable to the GAP activity of the exotoxins on Rho family GTPases. In an animal model of acute pneumonia, the persistence of P. aeruginosa in the lungs was impaired both by deficiency of the GAP and ADPRT activity of ExoS (323), supporting the bifunctional biological effect of the exotoxin.
C. Salmonella Species
The effect of Salmonella species on intestinal epithelial cells provides an interesting example of sequential activation and deactivation of Rho-family GTPases.
Infection with different Salmonella species occurs by oral ingestion and depends on uptake of the microorganisms into epithelial cells of the small intestine. Entry of Salmonella into nonphagocytic cells requires profound rearrangement of the cytoskeleton. Membrane ruffles and lamellipodia are formed at the site of pathogen-host interaction, and macropinocytosis is increased leading to gradual engulfment of the microorganism and sequestration into membrane-bound vesicles [Salmonella containing vesicles (SCV)] (272). Parallelly, MAP kinases (p38 and JNK) are activated, and local inflammation is initiated resulting in increased secretion and diarrhea. Internalization of Salmonella into epithelial cells absolutely requires injection of specific effector proteins by the TTSS encoded by the Salmonella pathogenicity island 1 (SPI-1). Three of these bacterially synthetized proteins turned out to affect small GTPases in the epithelial cells.
In epithelial cells in the first phase following infection, activation of the small GTPases Rac and Cdc42 occurs, initiating actin polymerization and branching via the Arp2/3 complex, leading to formation of membrane ruffles and lamellipodia and activation of JNK. These cytoskeletal changes are dependent on two proteins secreted via the TTSS of the bacteria: SopE and SopB. SopE turned out to be a GEF for Rac1, Rac2, Cdc42, and a few other members of the Rho subfamily, but not for Ras or Ran proteins (139). A homolog of SopE (SopE2) has also been identified, and it was shown that one of the two SopE proteins was expressed by all pathogenic Salmonella strains (339). SopB is an inositol phosphate polyphosphatase that contributes to activation of Cdc42 (272). Mutant strains lacking both SopE proteins and SopB are completely unable to initiate cytoskeletal rearrangement and to invade the host cell (119).
However, these cytoskeletal changes are reversible, and they are typically reversed in 2–3 h (130, 272). The reversal phase depends on another bacterial protein injected via the TTSS into the host cell. The critical protein SptP has a double function: the NH2-terminal part is an active GAP for Rac and Cdc42, whereas the COOH-terminal part has tyrosine phosphatase activity (116, 117). The GAP activity of SptP allows reversibility of the cytoskeletal rearrangement of the host cell, as recovery of the cellular morphology was equally low when cells were infected with Salmonella strains lacking SptP or expressing a GAP-deficient mutant SptP (116). In addition to the cytoskeletal changes, SptP also reversed the activation of JNK. In this effect, both the GAP and the tyrosine kinase activity of the protein has a role (116, 242).
SptP possesses a critical arginine and accelerates the GTP hydrolysis on the target small GTPases by a mechanism similar to mammalian GAPs (116). Despite the lack of significant sequence or structural homology between SopE and eukaryotic RhoGEFs, SopE appears to act on Cdc42 also in a similar way to its eukaryotic counterparts (55).
Although SopE and SptP, the GEF and GAP for Rac and Cdc42, respectively, are delivered at the same time in equivalent amount to the host cell, there is a time shift in the activation and inactivation of the small GTPases. The puzzle has been solved by demonstrating a significantly different half-life for the two regulatory proteins in the host cell (189). Both SopE and SptP are degraded by the proteasomal pathway, but with largely different speed: SopE cannot be detected 30 min after injection, whereas SptP remained detectable for more than 3 h (189). Intracellular stability of the proteins is controlled by the NH2-terminal secretion domain and the half-life of chimeric proteins (NH2-terminal SopE + COOH-terminal SptP or NH2-terminal SptP + COOH-terminal SopE) correlated with the identity of the NH2-terminal part. Inhibition of proteasomal degradation of SopE prevented cellular recovery after Salmonella infection (189).
Transient activation of the host cell cytoskeleton is required for the entry of Salmonella bacteria in the epithelial cells. The following rapid recovery of the cell morphology ensures survival of the infected cell and, in this way, long-term intracellular residence for the pathogen. The sequential activation and deactivation of the small GTPases responsible for the cytoskeletal changes are achieved in case of Salmonella strains by simultaneous injection of two proteins with opposing effects but largely different half-life.
VII. ROLE OF GAPs IN TUMORIGENESIS
As detailed in section II, small GNBPs regulate a diverse array of cellular functions such as proliferation, survival or apoptosis (Ras), adherence, migration, gene expression (Rac/Rho), or vesicular traffic that determine the exposure or internalization of plasma membrane constituents (Rab, Arf). Evidently, hyperactivation of various small GTPases may increase proliferation and metastatic development of different cells.
Hyperactivation of small GNBPs can occur by mutation of amino acids essential for GTP hydrolysis, rendering the protein constitutively active. As one of the most important oncogenes, RAS genes are found mutated in a variety of malignancies (214), with the highest incidence in pancreatic (12), lung (293), and colorectal (43) tumors. Oncogenic activation in most cases results in impaired and GAP-insensitive GTPase activity (4, 22, 357). Interestingly, somatic gain-of-function mutations were only very rarely identified in members of the Rac/Rho family (364). Alternatively, overexpression of GEFs can result in increased activity of small GTPases. Typically overexpression of different GEFs for the Rac/Rho family GTPases has been associated with development of cancer (54, 90, 105, 110, 227, 282, 406).
In the following we describe a few examples where lacking GAP activity has been revealed as the cause of tumorigenesis.
A. Neurofibromatosis Type 1 (Neurofibromin)
Deregulation of cellular (normal) Ras by the inactivation of relevant RasGAP proteins may lead to phenotypes characterized by upregulation of Ras, i.e., elevated levels of GTP-bound Ras in cells with a number of implications. The most prominent and probably best explored example is reflected in a long known disorder, termed neurofibromatosis type 1 (NF1), also called von Recklinghausen neurofibromatosis. Patients have an increased risk to develop typical neurofibromas (283), essentially benign tumors of the peripheral nerve sheath that, however, can turn into malignancy in a significant number of cases (18). In addition, the disease is associated with numerous symptoms including pigment anomalies of the skin (cafe au lait spots) or iris (Lish nodules), bone deformations, vascular and cardiac abnormalities, and learning disabilities (18, 283). Genetic alterations in the tumor suppressor gene NF1 are responsible for the pathogenesis of the disease (70).
NF1 encodes the cytoplasmic RasGAP neurofibromin (320 kDa) (92, 133, 134). Its RasGAP activity is located in a central portion reported to comprise between 300 and 400 residues (21, 217, 394) that has been narrowed down to 230 residues sufficient for its enzymatic function (8). It is followed by a bipartite phospholipid binding module composed of a glycerophospholipid binding Sec14-homology and a pleckstrin homology (PH)-like domain (16, 84, 382) (FIGURE 8), the funcion of which is yet unclear. Alternative splice variants have been reported, with one of them leading to the so-called type II transcript expressed in Schwann cells (258) carrying a 21-residue insertion within the GAP domain biochemically associated with reduced GAP activity (14). Of the reported NF1 associated alterations, the majority results in premature stop codons most likely leading to truncated transcripts removed by the surveillance machinery. Ten percent of the alterations are missense mutations, single/double residue deletions, or peptide insertions (324, 361). Provided that these mutations do not affect protein stability or folding, they have a high potential to act as functional missense mutants in the cell and are thus excellent tools to probe the molecular functions at high sequence resolution.
The role of neurofibromin's RasGAP activity for tumorigenesis has been impressively demonstrated by a number of studies that report increased levels of activated (i.e., GTP bound) Ras and consequently a hyperactivated MAPK pathway (FIGURE 3) in NF1-deficient tumor tissues derived from Schwannomas and malignant peripheral nerve sheath tumors (29, 40, 87, 93, 109, 129, 178, 199, 325, 358). In addition, loss of NF1 has been demonstrated to result in abnormal growth of hematopoietic cells along with activation of the Ras signaling pathway (40), as has been found in types of myeloid leukemia (321). It is believed that controlling Ras activity defines neurofibromin's primary role as a tumor suppressor, although loss of heterozygosity has not been established in all examples. Indeed, the RasGAP domain harbors a cluster of missense mutations that have been studied in biochemical detail (7, 181). Several of these affect the catalytic machinery and thus GAP activity (132, 181, 203, 317, 360). Most impressively, the mutation of the catalytic arginine to proline, as found in an NF1-patient family and has been associated with heavy tumor formation, virtually abolishes GAP activity (181). The hyperproliferative phenotype of NF1-depleted/deficient Schwann cells can be rescued by the retroviral expression of the GAP domain of neurofibromin in a RasGAP-dependent mechanism but not of that of the similar one of p120GAP (150, 354), suggesting that at least in those cells the catalytic domain is equipped with the requirements to effectively interact with activated Ras.
The earlier notion that the NF1 gene carries sporadic mutations in several types of tumors (315) has been reinforced by recent cancer genome studies indicating that NF1 is mutated in a significant number of rather aggressive malignancies including glioblastoma (252, 270), lung adenocarcinoma (97, 379), ovarian cancers (304), and soft tissue sarcoma (26), suggesting that NF1 may act as a deregulated tumor suppressor gene in these types of cancers.
B. Colorectal Cancer (RASAL)
Apart from gain-of-function mutation in Ras or loss-of-function mutations in neurofibromin, epigenetic mechanisms have been demonstrated to be involved in cancer development of certain tissues. Colorectal cancer (CRC) progression is frequently associated with aberrant Ras activation, although not all CRC cells carry activating Ras mutations. Based on these observations, researchers investigated CRC cells for expression of 12 RasGAPs and found that the Ca2+-regulated RasGAP like protein (RASAL) that decodes the frequency of Ca2+ oscillations is silenced in CRC cells from multiple tumors, most likely by CpG methylation (164, 262). Similarly, another RasGAP, the human DOC-2/DAB2 interactive protein (hDAB2IP), has been demonstrated to be a tumor suppressor gene inactivated also by aberrant methylation of CpG islands in a number of malignancies including prostate (65, 226) and gastrointestinal (98) cancers. The respective phenotype can be rescued by ectopic expression of RASAL in these cells. In RASAL-depleted cells, Ras was indeed upregulated with respect to normal cells correlating RASAL expression to the regulation of the MAPK pathway. The results suggest that RASAL and DAB2IP act as tumor suppressor-like proteins in the respective tissue cells and that epigenetic silencing of a GAP represents a previously unknown mechanism of Ras activation in certain types of cancers.
C. Tuberous Sclerosis (Tuberin)
Tuberous sclerosis is an autosomal dominant tumor syndrome characterized by deregulation of cell proliferation that results in the formation of hamartoma-like neoplasias/tumors in many organs (125). TSC2 is one of the two genetic determinants of the disease (277), encoding the Rheb-specific RapGAP tuberin (204, 215). Tuberin is a 200-kDa protein with a COOH-terminal 160-residue domain homologous to RapGAP1 (FIGURE 8). Several missense mutations have been identified in the RapGAP domain of tuberin that appears to be a target for such mutations in tuberous sclerosis (213, 219). Mapping those mutations onto the structure of the canonical RapGAP reveals the highly conserved catalytic asparagine (Asn290, Asn1643) (89) mutated to lysine (213) or isoleucine (19). The corresponding RapGAP mutant was entirely inactive (or insoluble as in the case of Asn1643Ile), implying that loss of RapGap activity may be constitutional for TSC pathogenesis, similarly like mutation of the catalytic arginine in neurofibromin has been found mutated in an NF1-family that was affected with heavy tumor development, resulting in an 8,000-fold reduced GAP activity (181).
D. Liver Cancer (DLC1)
“Deleted in liver cancer” (DLC1/2; FIGURE 8) (also known as ARHGAP7 or STARD12) has been identified as a RhoGAP (67, 398) that inhibits the proliferation in hepatocellular carcinoma cells in a cellular context that requires RhoGAP activity (389) but employs also RhoGAP-independent mechanisms (142). In the case of DLC1, tumor suppression did not occur if a catalytically inactive point mutation was introduced (389). An epigenetic mechanism that involves silencing by hypermethylation of CpG islands has been proposed contributing to gastric cancerogenesis (179, 388). Cellular studies in cell lines derived from a variety of tumors suggest that DLC1/2 may play a tumor-suppresive role beyond liver cancer (205).
E. Breast Cancer (β2-Chimerin)
Stimulation of epidermal growth factor receptors (EGFRs) results in activation of both Ras and Rac (195), the latter being mediated by the exchange factor Vav (236, 349) (Figs. 3 and 4). Parallelly, PLC-γ is recruited and activated by the phosphorylated intracellular tail of EGFR (259, 299), resulting in production of cytosolic IP3 and membrane-attached DAG. It has been shown recently that EGF induces the redistribution of β2-chimerin, a RacGAP with a DAG-binding C1 domain (FIGURE 8), from the cytosol to the plasma membrane where it interacts with activated Rac (372). Downregulation of β2-chimerin resulted in significant prolongation of the prevalence of RacGTP, indicating an essential role of β2-chimerin in termination of the EGFR-induced Rac signal (372).
The biological significance of the downregulation of Rac signaling is enlighted by the finding that the mRNA level of β2-chimerin is significantly decreased in several breast cancer cell lines as well as in 70% of the investigated breast cancer tissues (396). Hyperactivation of Rac resulting in enhanced proliferation has been reported earlier in human breast cancer models (229). Expression of the full-length protein in β2-chimerin-deficient MCF-7 breast cancer cells resulted in decrease of proliferation and cell cycle arrest at the G1/S phase. Parallel dose-dependent decrease was observed in the level of RacGTP and cyclin D1. All these effects depended on the GAP activity of β2-chimerin (396). In addition to breast cancer cells, downregulation of β2-chimerin has also been observed in high-grade gliomas (399), and a progressive loss of β2-chimerin expression was detected in microarrays of benign duodenal adenomas and adenocarcinomas (396).
Taken together, β2-chimerin seems to play an essential and specific role in termination of EGFR-induced Rac signaling, and partial loss of this GAP is sufficient to enhance cell proliferation.
VIII. CONCLUDING REMARKS
GNBP function as programmed binary switches: in the GTP-bound, active conformation, the regulated biological processes are switched on, whereas in the GDP-bound, inactive state, these processes are switched off. The dynamism of the system is determined by the rate of GTP hydrolysis by the GNBP, a property that depends on the protein structure. Ultimately, the timing program is coded in the protein sequence. GAPs accelerate the rate of GTP hydrolysis by small GNBPs, and RGSs are GAPs for the Gα subunits of heterotrimeric GNBPs. In this way, GAPs and RGSs can be regarded as program modifiers that exert constant control over the regulated biological function. Depending on the rate of the opposite, activating process, this control may manifest itself as inhibition or termination of the physiological process. The detailed examples substantiate that loss of GAP activity may result in an increase in the intensity and/or the duration of the regulated physiological process, in most of the cases leading to pathological consequences. However, GAPs are themselves the subject of diverse and complex regulation, allowing in this way the very specific and fine temporal tuning of diverse cellular processes.
The large number of GAPs with identical or similar substrate preference has raised the question about specific or overlapping functions. Although it is too early to give a final answer, two lines of observations allow some speculations. On one hand, the diverse multidomain structure of GAPs suggests that these proteins participate in well-defined molecular complexes confined to specific sites within the cell. Replacement of one multidomain protein by another complex protein with a similar GAP domain can probably be regarded rather as an exception than the general pattern. On the other hand, the very specific symptoms and unique phenotype alterations arising in case of deficient GAP activity of one particular protein in spite of undisturbed expression of several other GAPs of the same or similar substrate specificity, also suggest a nonredundant role for these proteins.
The examples of alteration of physiological functions detailed in this review support the view that GAPs represent key elements in the precise spatiotemporal regulation of a wide variety of cellular functions.
This work was supported by grants from the Hungarian National Research Fund (OTKA K81277 and K75084) and TÁMOP (grants 4. 2. 1/B-09/1/KMR-2010-0001 and 4. 2. 2/B10/1-2010-0013). S. Welti has been supported by the Peter and Traudl Engelhorn Stiftung (Germany) and by grants (to K. Scheffzek) from the Baden-Württemberg Stiftung and the Federal Ministry of Education and Research (Germany).
No conflicts of interest, financial or otherwise, are declared by the authors.
E. Ligeti is indebted to Professors András Spät, Anna Faragó, and Péter Enyedi for stimulating discussions and critical reading of the manuscript and to Roland Csépányi-Kömi for devoted editorial help.
Address for reprint requests and other correspondence: E. Ligeti, Dept. of Physiology, Semmelweis University, Tűzoltó u. 37–47, H-1094 Budapest, Hungary (e-mail:).