Rho-Family Small GTPases: From Highly Polarized Sensory Neurons to Cancer Cells

The small GTPases of the Rho-family (Rho-family GTPases) have various physiological functions, including cytoskeletal regulation, cell polarity establishment, cell proliferation and motility, transcription, reactive oxygen species (ROS) production, and tumorigenesis. A relatively large number of downstream targets of Rho-family GTPases have been reported for in vitro studies. However, only a small number of signal pathways have been established at the in vivo level. Cumulative evidence for the functions of Rho-family GTPases has been reported for in vivo studies using genetically engineered mouse models. It was based on different cell- and tissue-specific conditional genes targeting mice. In this review, we introduce recent advances in in vivo studies, including human patient trials on Rho-family GTPases, focusing on highly polarized sensory organs, such as the cochlea, which is the primary hearing organ, host defenses involving reactive oxygen species (ROS) production, and tumorigenesis (especially associated with RAC, novel RAC1-GSPT1 signaling, RHOA, and RHOBTB2).


Introduction
The small GTPases (also known as small G proteins (~21 kDa)) of the Rho-family (Rho-family GTPases) form a subfamily of the Ras superfamily of small GTPases. Human and murine Rho-family GTPases include 21 members subdivided into eight families classified as typical (classical) or atypical depending on their mode of regulation [1] (Table 1). Rho, Rac, Cdc42, and RhoF/RhoD are members of the classical subfamilies (Table 1) and their functions are regulated by cycling between an inactive GDP-bound form and an active GTP-bound form. Rnd, RhoH, RhoU/RhoV, and RhoBTB are atypical subfamilies and predominantly exist in the GTP-bound-form. The GTP-bound status is regulated by guanine nucleotide exchange factors (GEFs), GTPase-activating proteins (GAPs), and Rho-specific guanine nucleotide dissociation inhibitors (RhoGDIs) [2,3]. In humans, there are~80 RhoGEFs classified into the Dbl family (69 members) and the DOCK family (11 members) [2,3]. The RhoGDI family comprises the RhoGDIα, RhoGDIβ, and RhoGDIγ subtypes [4]. RhoGDIα is ubiquitously expressed, whereas RhoGDIβ is expressed mainly in hematopoietic cells and also in cancer cells. RhoGDIγ is expressed primarily in the brain. RhoGDIα and RhoGDIβ are expressed in phagocytes. The atypical subfamily of Rho-family GTPases is regulated by expression, post-transcriptional modifications, and/or interaction with other proteins but not by GEFs or GAPs [1,5]. Not all Rho-family GTPases are regulated by RhoGDIs [1,6]. Table 1. Knockout (KO) mice availability, congenital diseases, and tumorigenesis-associated Rho-family GTPases. Blue indicates typical (classical) Rho-family GTPases including the Rho, Rac, Cdc42, and RhoD/RhoF subfamilies. Red indicates atypical Rho-family GTPases including the Rnd, RhoH, RhoU/RhoV, and RhoBTB subfamilies. CNS: central nervous system, HD: host defenses, TKS: Takenouchi-Kosaki syndrome. +: conditional floxed mice. + C : conventional KO mice.

RhoBTB1
RhoBTB2 + (CNS anomalies) [39] suppression [5] RhoBTB3 The illustration on the left shows phagocytes with a phagosome containing internalized pathogens (indicated by pink square). The middle illustration shows the magnified phagosome with a Nox2 complex (Rac is one of six components) on the membrane. Superoxide (O2 − ) and reactive oxygen species (ROS) are generated in the phagosome.
The illustration on the right shows the otoliths on the otolithic membrane in the vestibule. Superoxide from Nox3 is essential for the synthesis of otolith; however, the origin of superoxide (that is Nox3- The illustration on the left shows phagocytes with a phagosome containing internalized pathogens (indicated by pink square). The middle illustration shows the magnified phagosome with a Nox2 complex (Rac is one of six components) on the membrane. Superoxide (O 2 − ) and reactive oxygen species (ROS) are generated in the phagosome. The illustration on the right shows the otoliths on the otolithic membrane in the vestibule. Superoxide from Nox3 is essential for the synthesis of otolith; however, the origin of superoxide (that is Nox3-expressing cells in the inner ear) is still controversial. (C) The scheme represents how Rho-family GTPases are involved in tumorigenesis.

Role and Function of Rac in Hearing
The function of Rac1 in hearing was elucidated using Foxg1-Cre or Pax2-Cre mice with Rac1 knockout (KO). The Rac1-KO (Foxg1-Cre;Rac1 flox/flox and Pax2-Cre;Rac1 flox/flox ) mice showed embryonic lethality, incomplete development of HC planar cell polarity (PCP), and abnormal and fragmented hair bundles [50]. Mice with Rac1 and Rac3 double knockout (DKO) presented with exacerbated Rac1-KO phenotypes [51]. These pioneering studies unveiled the functions of Rac in the cochlea. However, Foxg1 [52,53] and Pax2 [54] are activated before Atoh1 in the cochlea, the master regulator of HC differentiation [55]. These genes are also active in the telencephalon (Foxg1), the brain stem, and throughout the OC epithelia including the HCs, SCs, and their precursors [56]. Further studies are required to reveal the specific functions of Rac in the HCs.

Role and Function of Cdc42 in Hearing
We recently reported that Cdc42 plays essential roles in the maintenance of cochlear HCs. We generated Cdc42-KO mice under the control of the Atoh1 promoter (Atoh1-Cre;Cdc42 flox/flox ) [57]. After normal morphological maturation, the Cdc42-KO mice showed progressive sensorineural hearing loss (SNHL), particularly at high frequencies, and HC loss accompanied by various stereociliary abnormalities starting at postnatal day 14 (P14) (scattered, short, long, and fused) predominantly at the IHCs of the basal turn. Cdc42 acts on the membranes covering the stereocilia (especially the upper half) and the apical junctional complexes (AJCs) in the cochlear HCs [57]. Cdc42 functions at the AJCs as a complex with atypical subfamily of protein kinase Cs (aPKCs) [57]. However, its maintenance mechanism at the stereociliary membranes remains unknown. Active Cdc42 at the upper half of the stereociliary membranes may be involved in an "tip turnover" model of stereolilia, in which actin turnover in stereocilia occurs only at the tips, but not shafts [47,[58][59][60]. The Cdc42 expression levels in the HCs resembled those in SCs [61]. Another group examined the Cdc42-KO phenotypes in Fgfr3-iCre-ER T2 ;Cdc42 flox/flox mice (Cdc42-KO in SCs (Deiters' and pillar cells)). Cdc42 was knocked out using Fgfr3-iCre-ER T2 mice, in which Cre functioned with the assistance of tamoxifen administration (between P2 and P4 or P16 and P18). The mice presented with impaired apical polarization of SCs, no cochlear HC loss, and impaired wound healing in the SCs after ototoxin administration in the adults [62]. To examine the function of Cdc42 in HCs (especially OHCs), different tamoxifen administration intervals were applied at E13.5 and E14.5 or at E15.5 and E16.5 against Fgfr3-iCre-ER T2 ;Cdc42 flox/flox mice. The mice showed impaired planar cell polarity (PCP) of the OHCs, fragmented stereociliary bundles, short stereocilia, and scattered OHC loss at P6 [63]. Their phenotypes were stronger than those observed in our Cdc42-KO mice [57]. While those studies used promoter-Cre mice, ours tested Atoh1-Cre transgenic mice. However, another group used Fgfr3-iCre-ER T2 mice. Following are possible reasons why the phenotypes differed between these groups: (1) although the Atoh1 and Fgfr3 promoters function in both HCs and SCs, the Atoh1 promoter operates predominantly in the former rather than the latter [56,64,65]; (2) unequal proportions of Cdc42 KO in the HCs and SCs may differentially affect the HC phenotypes since HC-SC interaction is a critical step in terminal HC differentiation [55].

Deafness in Patients with Active DIA1 Mutations Downstream of RhoA
Fifteen formin proteins, which nucleate and elongate unbranched/straight actin filaments, are found in mammals and are classified into eight subfamilies [94,95]. DIA1 is one of three members (DIA1-DIA3) of the diaphanous-related formin (DRF) subfamily and is a downstream target of the RHOA signaling pathway [95]. DIA1 has two alternative splicing variants: DIA1-1 consists of 1263 amino acids (aa) and DIA1-2 has 1254 aa. In the latter case, in-frame exon 2 (27 nucleotide pairs; 9 aa) is skipped.
From the study of Cdc42 in cochlear HCs, we found that Cdc42 knockdown (KD) enhances RhoA signaling activation [57]. We hypothesized that the activation of DIA1 by RhoA at least partially accounts for the phenotypes observed in Cdc42-KO cochlear HCs. DIA1 is the causative gene of the first type of non-syndromic SNHL with autosomal dominant inheritance, namely, DFNA1, which was reported in 1997 [96]. However, only one Costa Rican family was reported with this defect, and the molecular mechanism of DFNA1 was not disclosed. In 2016, we reported a novel heterozygous nonsense mutation in DIA1-2 causing a constitutively active mutant of DIA1, namely, p.R1204X (p.R1213X in DIA1-1) [97]. The p.R1204X/R1213X mutation is located in the C-terminal diaphanous autoregulatory domain (DAD) and disrupts an inhibitory intramolecular interaction between the N-terminal diaphanous inhibitory domain (DID) and the DAD [97]. Patients with the p.R1204X mutation present with progressive SNHL, generally at high frequencies [97]. Transgenic mice with the p.R1204X mutation have a similar phenotype including progressive loss of cochlear HCs, particularly in the OHCs of the basal turn of the cochlea, progressive SNHL typically at high frequencies, various abnormal stereocilia (sparse, short, long, fused, and dislocated) of the HCs, and deformed cell-cell junctions with the SCs [97]. Other groups also reported DFNA1 caused by the constitutively active DIA1-1 mutants p.R1213X [98,99], p.R1210Serfs31X [100], p.R1210Glyfs31X [101], and E1192_Q1220del [101]. The first heterozygous mutation reported as a cause of DFNA1, p.Ala1221Valfs22X, is located outside the DAD (D1188-G1217 in DIA1-1) [96]. Nevertheless, all other mutants in DIA1 manifesting as progressive SNHL are located in the DAD and produce constitutively active DIA1 mutants. However, the molecular mechanism of the original DFNA1 mutant is still unclear.

Deafness Associated with Rho-Family GTPases Other than Rac and Cdc42
Other Rho-family GTPases may also be involved in hearing. Mice with RhoA deletion in the OHCs and SCs (Deiters' and pillar cells; Fgfr3-iCre-ER T2 ;Cdc42 flox/flox ) under tamoxifen administration at E13 and E14 showed only mild abnormalities of the developing OHCs (enlargement of apical cell surface) and extrusion of the OHCs in the endolymph. However, the PCP of their OHCs and SCs were normal [112]. FAB65B is the causative gene of DFNB104 and binds to RhoC at the tapering stereociliary base. Conventional Fam65b-KO mice presented with hearing loss as a consequence of morphological abnormalities at the base of the stereocilia [113]. HOMER2 belongs to the homer family of post-synaptic density scaffolding proteins. It binds to CDC42 through its CDC42-binding domain (CBD) in the CC domain. Patients with a p.R185P mutation in HOMER2 showed hearing loss. Conventional Homer2 KO-mice had progressive SNHL [114]. ARHGFE6 is a specific Rho-GEF of RAC and CDC42 and one of the causative genes of X-linked mental retardation (XLMR). A reciprocal X/21 translocation presented with SNHL and severe intellectual disability [115]. Mice whose Arhgef6 isoform 1 was lost by genome editing had stereociliary disorganization and progressive HC loss and SNHL [116].

Roles of Rho-Family GTPases in Other Sensory Organs
The Rho-family GTPases may also function in other highly polarized sensory cells/organs because they participate in polarization events [45]. For example, Cdc42 is involved in the development of the outer segment and the connecting cilia of retinal photoreceptor cells, which are modified and specialized primary cilia [79,80,117] (Figure 2).
The first biological function ascribed to Rac was Nox2 activation [130]. The addition of either Rac1-GTP [131] or Rac2-GTP [132] was essential for high-level superoxide production in cell-free Nox2 assays. The role of Rac2 in vivo was later confirmed in hematopoietic cell-specific Rac2-KO mice [133] and the aforementioned RAC2 mutant patients [23,24,26]. Superoxide production from Rac1-KO neutrophils was normal [134]; however, that from DKO neutrophils of Rac1 and Rac2 had higher reducing power than that from Rac2-KO neutrophils [135]. Rac2 also participates in B lymphocyte activation and development [26,136] as well as those of T lymphocytes [25,26]. DKO of Rac1 and Rac2 almost completely suppress B lymphocyte development [78].
Translocation (targeting) of Rac from the cytosol to the phagosomal membrane (phagosome) is independent of that for the ternary phox protein complex (p47 phox -p67 phox -p40 phox ) [143,144]. The latter is one of two cytosolic activator complexes in Nox2-based oxidase [123]. The separation of these mechanisms may account for the very tight regulation of Nox2-based oxidase. We reported the isoform-specific targeting mechanism of three Rac isoforms from the cytosol to the phagosome via their C-terminal polybasic (PB) motifs (KKRKRK in Rac1, RQQKRP in Rac2, and KKPGKK in Rac3) [120].
Rac1 directly targets the phagosome with the highly positively charged PB motif. In contrast, Rac2 initially targets the endomembranes which then fuse to the phagosome. Another study corroborated the delivery of Rac2 to phagosomes via the endomembranes (including granule/vesicle membranes) [145].

Regulation of Superoxide Production by RhoGDIs
The regulatory mechanism of the cytosol-membrane cycle of the RhoGDI-Rac complex is not fully understood. This complex is one of two cytosolic Nox2-based oxidase activators. It was proposed without evidence that the Rac dissociates from RhoGDI in the cytosol. We reported that the RhoGDIα-Rac1 and RhoGDIβ-Rac1 complexes translocate/target the phagosome then Rac1 dissociates from them on the membrane [125]. Another study also reported the dissociation of Rac from RhoGDIβ on the phagosome [156]. The dissociation mechanism has been most extensively studied in phagocytosis. However, our proposal may also apply not only to phagocytosis [139,157] but to other cell types and signaling pathways as well [4]. In general, the cytosol-membrane cycle of the RhoGDIα/β-RhoGTPase complex in which the RhoGTPases are activated (released from RhoGDIs and converted to GTP-bound form) on membranes may be regulated as follows: (1) interaction between the C-terminal PB motif in RhoGTPases and the negatively charged residues in the N-terminal of RhoGDIα/β to form a dimer complex in the cytosol [120,125], (2) anionic phospholipids such as PI(3,4,5)P 3 , PA, and phosphatidylserine produced on membranes [120,158,159] compete for binding to the PB motif in RhoGTPases with the N-terminal negatively charged residues in RhoGDIα/β, (3) RhoGTPases are released with the help of RhoGDI dissociation factors such as GEF and phosphorylation in RhoGDIα/β [160][161][162], thereby promoting their membrane insertion and activation, and (4) electric repulsion between the anionic membranes and the negatively charged residues in the N-terminal of RhoGDIα/β promotes the dissociation of RhoGDIα/β from the membranes [125].

Recent Advances in Rac Involvement in Tumorigenesis
Rac1 is well known as a tumor progression factor (Table 1) because it participates in cell migration/invasion and proliferation [163]. Nevertheless, the exact mechanisms of Rac1 in tumorigenesis have not been fully unveiled. Recently, several activating RAC mutants including RAC1 (p.P29S, p.P29T, p.N92I, pC157Y, and p.A159V) and RAC2 (p.P29L and p.P29Q) indicated that RACs are oncogenic driver genes in human melanoma and head and neck squamous cell carcinoma patients and in cancer cell lines [14,15,21,22,27]. Although these studies reveal the roles of Rac in tumorigenesis and tumor progression, the mechanism of the involvement of Rac in tumorigenesis remains obscure. Using Forster resonance energy transfer (FRET) biosensors, Hirata et al. discovered that C6 glioma cells penetrating the brain parenchyma have high Rac1 and Cdc42 activity and low RhoA activity [164]. They also found that Dock9, a GEF for Cdc42, is an upstream mediator of this invasion process [164]. It was then demonstrated that Rac1 and Cdc42 activity is variable in C6 glioma [165]. Cells with high Rac1 or Cdc42 activity invade more efficiently and have stronger invasion-inducing signaling networks than cells with low Rac1 or Cdc42 activity [165]. Rac1b is an alternatively spliced Rac1 isoform with a short additional exon (3b; 57 nucleotides; 29 aa). Rac1b overexpression has been reported for breast, colon, and lung cancers [166]. In mouse lung and colon tumor models, Rac1b sufficed to initiate tumorigenesis but was assisted by inflammation or K-ras mutation [167,168].

A Novel Downstream Target of Rac Signaling, GSPT1, Is Associated with Tumorigenesis
We recently reported that the G1 phase of the primary astrocyte cell cycle is prolonged by both Rac1-KO and Rac1-KD. G1 to S phase transition 1 (GSPT1) is a novel transcriptionally regulated downstream target of Rac1 [169]. GSPT1 was initially identified as an essential gene for G1 to S phase transition in the cell cycle. However, it was later renamed as the eukaryotic releasing factor 3a (eRF3a) [170]. GSPT1/eRF3a is a GTP-binding protein participating in translation termination as a ternary complex with eRF1 and GTP. The eRF1 recognizes the terminal codon and releases the completed protein product after the hydrolysis of GTP by the intrinsic GTPase of eRF3a [170,171]. We found that Rac1-GSPT1 signaling promotes astrocyte cell cycle progression and proliferation (astrogliosis) after CNS damage such as spinal cord injury (SCI) and irradiation injury. The promotion of the G1 to S phase transition may be regulated by Rac1 via increases in cyclin D1 levels [172,173]. Genome-wide association studies (GWAS) showed that testicular germ cell tumors are susceptible to GSPT1 upregulation [174]. Xiao et al. also reported that HCT116 colorectal cancer cells express GSPT1 at high levels [175]. GSPT1 may be an antitumor drug target through ubiquitination and degradation [176]. GSPT1 may also be involved in lung and colorectal carcinoma cell migration [175,177]. In our studies, however, GSPT1 KD demonstrated no such effect in primary astrocytes, immortalized LN229 astrocytes, or A431 epidermoid carcinoma cells [169]. The reason for the discrepancy between our study and previous reports is unknown. Nevertheless, differences in the cell types and lines used may account for it. Another explanation is the difference in the cell migration assay used. A scratch assay is influenced by both cell migration and proliferation. Human GSPT1 is polymorphic and has a glycine repeat (7, 8, 9, 10, 11, or 12) in the N-terminus. The most common type is the 10 repeat [178]. Several studies reported that the presence of GSPT1 with a 12 glycine repeat is correlated with increased risks of gastric, breast, and colorectal cancers [179][180][181]. However, the exact mechanism linking the number of glycine repeats and tumorigenesis remains unknown. The RAC1-GSPT1 signaling axis may participate in an oncogenic mechanism and is a candidate target for a novel tumor therapy.

Conclusions
The functions of the Rho-family GTPases are gradually being elucidated, and innovative scientific and industrial applications are being developed for them. Studies using genetically engineered mice/animals (such as KO and knock-in) and information gathered from patients with genetic mutations are powerful tools to disclose the functions of Rho-family GTPases in vivo. Research on the Rho-family GTPases associated with sensory organs/cells involved in hearing, balance, and vision as well as those involved in various cancers, must be continued so that we gain a better understanding of the physiological role of these proteins in vivo. This will then facilitate the development of novel therapeutic modalities for sensory diseases and cancers.
Funding: The APC was funded by the Foundation for Promotion of Cancer Research in Japan.