Functional Plasticity of the BNIP-2 and Cdc42gap Homology

FEBS Letters 586 (2012) 2674–2691

journal homepage: www.FEBSLetters.org

Review Functional plasticity of the BNIP-2 and Cdc42GAP Homology (BCH) domain in cell signaling and cell dynamics ⇑ Catherine Qiurong Pan b, Boon Chuan Low a,b, a Cell Signaling and Developmental Biology Laboratory, Department of Biological Sciences, 14 Science Drive 4, National University of Singapore, 117543 Singapore, Singapore b Mechanobiology Institute Singapore, 5A Engineering Drive 1, National University of Singapore, 117411 Singapore, Singapore article info abstract

Article history: The BNIP-2 and Cdc42GAP Homology (BCH) domains constitute a new and expanding family of Received 13 March 2012 highly conserved scaffold protein domains that regulate Rho, Ras and MAPK signaling, leading to Revised 16 April 2012 cell growth, apoptosis, morphogenesis, migration and differentiation. Such versatility is achieved Accepted 16 April 2012 via their ability to target small GTPases and their immediate regulators such as GTPase-activating Available online 21 April 2012 proteins (GAPs) and guanine nucleotide exchange factors (GEFs), their ability to form intra- Edited by Marius Sudol, Gianni Cesareni, molecular or inter-molecular interaction with itself or with other BCH domains, and also by their Giulio Superti-Furga and Wilhelm Just ability to bind diverse cellular proteins such as membrane receptors, isomerase, caspases and metabolic enzymes such as glutaminase. The presence of BCH and BCH-like domains in various Keywords: proteins and their divergence from the ancestral lipid-binding CRAL–TRIO domain warrant the need BCH domain to examine closely their structural, functional and regulatory plasticity in isolation or in concert CRAL–TRIO domain with other protein modules present in the same proteins. GTPase Ó 2012 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. Kinase Modular Scaffold Signaling Protein–protein interaction Dynamics

1. Introduction combination, and how their potential functions are regulated across molecular, cellular and tissue levels. For example, the 1.1. Modular protein domains defines specificity, integration and adaptor protein, Grb2 contains both Src-homology-2 (SH2) and crosstalk of signaling networks Src-homology-3 (SH3) domains which bridge interaction between the phosphorylated tyrosine residue on the receptor tyrosine Protein domains are the autonomous, minimal modular entities kinase and the core PXXP motif of the RasGEF, SOS, respectively, that constitute the basic structural or/and functional units of a thereby transducing signals upon binding of ligands at the extra- polypeptide. They can act as protein docking sites, enzymatic units cellular space to intracellular milieu [2]. Precise molecular recogni- or regulatory devices and are classified into different families tion is also important for transcriptional regulation. For example, according to their conserved primary sequences. Each family of do- under the HIPPO signaling which is crucial for the control of organ main recognizes unique interaction motif at their target sites. As size and tumor suppression, the LATS kinase via its PPxY motif, such, proteins exploit different architecture designs of modular do- targets the WW domain of the transcriptional regulator, Yes mains in order to exert common, overlapping or distinct functions, associated protein (YAP) to phosphorylate YAP, leading to seques- particularly in mediating protein–protein and protein–lipid inter- tration of YAP in cytoplasm by protein 14-3-3. This mechanism actions that form the basis for intracellular and intercellular signal- helps reduce the YAP-induced cell proliferation and tumor ing [1]. suppression [3]. Similarly, in protein trafficking and targeting, As cells sense both biochemical and physical environments, it is interaction between the pleckstrin homology (PH) domain of the important to define how these domains work in isolation and in molecular motor, dynamin and phosphoinositide is essential for endocytosis [4,5]. To further enhance the efficiency of multi-component signaling ⇑ Corresponding author at: Cell Signaling and Developmental Biology Laboratory, Department of Biological Sciences, 14 Science Drive 4, National University of networks, cells employ specific scaffold proteins that utilize their Singapore, 117543 Singapore, Singapore. Fax: +65 6872 6123. multiple protein domains to recruit and organize different targets E-mail address: [email protected] (B.C. Low). into higher order macromolecular complexes at specific location

0014-5793/$36.00 Ó 2012 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.febslet.2012.04.023 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2675 and to ﬁne-tune the activity and crosstalk among them. For exam- tion often lead to developmental defects and disease, less is known ples, the activity of Raf-Mek-Erk or Mek-Erk nodes downstream of about how this ‘‘3G’’ (GTPase, GEF, GAP)-signalome itself is regu- Ras is supported by different scaffold proteins at different locales, lated by cellular factors and how their functions and regulation namely the KSR (cell membrane), MP1 (late endosome), Sef (Golgi), are linked to other signaling networks downstream of membrane paxillin (focal adhesions), IQGAP1 (cytoskeleton) and b-arrestin receptor and MAP kinase modules. We therefore hypothesized that (early endosome) [6]. Such distribution of signaling nodes by BCH domain in BNIP-2 and Cdc42GAP could serve as a binding distinct hubs ensure localized and more speciﬁc control in the acti- platform for GTPases and possibly their immediate regulators, or vation (and deactivation) of Erk during different cellular processes. they could form multimeric form of BNIP-2, as sometimes seen Last but not least, coordination of cell adhesion dynamics require for other homologous domains used in protein dimerization/ the FERM (four-point-one, ezrin, radizin, moesin) domain of oligomerization. FAK (focal adhesion kinase) interacting with phospholipids Based on structural modeling, site-directed mutagenesis,

PtdIns(4,5)P2 at the cell membrane to release the autoinhibition binding studies and GTPase assays, we revealed that BCH do- in FAK. This facilitates the recruitment of Src tyrosine kinase and main of BNIP-2 indeed binds to itself and the homologous BCH formation of focal adhesion complex via integration of many more domain of Cdc42GAP via a unique motif, 217-RRKMP-221 [11]. scaffold proteins, kinases, phosphatases, GTPases and their Interestingly, BNIP-2 BCH domain also binds Cdc42, a partner immediate regulators [7,8]. of the Cdc42GAP, via a sequence motif similar to the CRIB Therefore, precise locales and timing for the molecular recogni- (Cdc42/Rac1-Interactive Binding) motif commonly present in tion by different combinations of protein domains in cis and in the effector proteins of Cdc42/Rac1 [16,17] (see Fig. 3A and C trans will help define the specificity, connectivity, crosstalk and for details later). Unlike binding of CRIB motif which is abso- feedback among various signaling nodes. Acting in concert, they lutely GTP-dependent (in order to allow activation of effectors regulate important functions such as cytoskeleton rearrangement, pathway), the BCH domain of BNIP-2 does not distinguish be- cell–cell adhesion, cell–matrix interaction, protein trafficking, tween the unloaded, GDP and GTP-bound form of Cdc42 organelle and membrane dynamics, cell metabolism and immune in vitro [9]. However, it binds preferentially to the dominant response, gene expression and protein synthesis and stability, lead- negative form of Cdc42, T17N but not with the constitutively ac- ing ultimately to cell growth, cell death, cell differentiation, cell tive form, G12V [12]. Subsequently, we showed that these pro- migration and invasion. files are also observed in binding of Rho by the other homologous BCH domains of BNIP-S [33], BNIP-XL [18] and 2. BCH domain as an emerging regulatory scaffold domain Cdc42GAP [19] and are distinct from that conferred by the GTPase-binding domains of GEFs and GAPs [14,15,20]. Further- 2.1. Discovery of the prototypical BCH domain from BNIP-2: from more, BNIP-2 appeared to promote the inactivation of Cdc42 receptor tyrosine kinase to small GTPase, MAP kinase and beyond by stimulating the apparent GTPase activity of Cdc42 in vitro and such effect is abrogated upon binding by FGFR [9] or nega- To understand the fundamental mechanism in cell signaling it is tively regulated by the homophilic interaction of BNIP-2 or its important to identify what and how one particular signaling pro- heterophilic interaction with Cdc42GAP [11]. Since then, by tein forms its protein interaction networks and how this process using yeast 2-hybrid, proteomics-based mass spectrometry and is regulated in vitro and in vivo. In our effort to decipher how acti- candidate approaches, we and others have expanded the family vation of the fibroblast growth factor receptor (FGFR) tyrosine ki- of BNIP-2 interactomes. We have further delineated how some nase could lead to novel signaling pathway, yeast 2-hybrid was of these BCH domains could in fact either activate or inactivate employed and BNIP-2 was identified as a candidate partner for specific Rho GTPases signaling depending on various cellular the cytoplasmic tail of the receptor [9]. Interestingly, only the ki- contexts the proteins are present, and they could directly or nase-dead mutants of the FGFR could ‘‘trap’’ and co-immunopre- indirectly control multiple signaling and metabolic pathways. cipitate with BNIP-2, indicating that their interaction was very These processes involve specific members and conformation of transient and BNIP-2 was likely to be a substrate of the receptor. Rho and Ras family of small GTPases and selective GEFs and Further in vitro kinase assays and detection with phosphotyrosine GAPs, growth promoting FGF receptor tyrosine kinase, myogenic antibody confirm BNIP-2 as a bone fide substrate of the FGFR both Cdo receptor and the associated JLP/p38-MAP kinase, Mek2, in vitro and inside the cells [9]. microtubule-based kinesin motor, anti-apoptotic Bcl2, peptidyl BNIP-2 was first identified as one of the three interacting part- prolyl-isomerase Pin1 and glutaminase, a metabolic enzyme ners for the anti-apoptotic Bcl2 and viral E1B 19 kDa protein, hence essential for cell growth and neuronal differentiation (Fig. 1). named Bcl-2 and Nineteen kilodalton-Interacting Protein-2, BNIP-2 Moreover, BCH domains can undergo auto-inhibitory intra- [10]. However, its precise biological functions remained unknown molecular interaction and form oligomer with identical BCH do- then. Our further sequence analyses revealed 61% similarity be- mains (homophilic interaction) or with other homologous BCH tween the C-terminus of BNIP-2 with the N-terminus, non-cata- domains (heterophilic interaction). They also collaborate with lytic region of Cdc42GAP (p50RhoGAP or ARHGAP1) and named other protein modules and their interactomes on the same pro- the region BNIP-2 and Cdc42GAP Homology (BCH) domain teins (in cis) or with different proteins (in trans) in order to en- (Fig. 1; [9,11,12]). As this domain is associated with the Rho sure faithful signaling integration and crosstalk. The basis for GTPase-activating domain (GAP) domain of Cdc42GAP, we specu- such plasticity for BNIP-2 and several BCH-containing proteins lated that it could be linked to function and regulation of GTPases will be further highlighted and discussed below in the context signaling. Among key molecular switches, Ras and Rho small GTP- of their versatile binding capabilities through their multiple ases control cell signaling and cytoskeleton networks by cycling binding motifs, their dynamic disposition within the cells, their between their inactive GDP-bound form and active GTP-bound synergism (or antagonism) with other protein domains in cis or state [13]. This process is tightly regulated by their guanine nucle- in trans, their potential regulation by transcription, alternative otide exchange factors (GEFs) that catalyze the exchange of GDP transcription start sites, alternative RNA splicing and post-trans- for GTP [14] or by their GTPase-activating proteins (GAPs) that help lational modifications, and their molecular evolution which could hydrolyze the GTP to GDP [15]. Although the biochemical functions shed further lights on the common and divergent properties of of GTPases, GEFs and GAPs are well established and their deregula- BCH domains with other protein modules. 2676 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

1 167 314 BNIP-2

1 72 205 CDC42GAP 439 (p50RhoGAP / ARHGAP1)

GTPases, GEFs, GAPs BCH glutaminase FGFR Cdo receptor Pin1 Bcl2

morphogenesis, motility, differentiation, trafficking, apoptosis

Fig. 1. The BCH domain and its expanding protein interaction networks. Regions in blue depict the conserved domains between the C-terminus of BNIP-2 and the N-terminus of Cdc42GAP, hence named BNIP-2 and Cdc42GAP Homology (BCH) domain. This novel protein module which is present in diverse proteins can interact with Rho and Ras small GTPases, their GEFs and GAPs, membrane receptors, glutaminase, Pin1, E1B-19 kDa protein, Bcl2 and other cellular proteins where they regulate diverse signaling pathways, leading to changes in cell morphology, motility, differentiation, protein trafﬁcking and apoptosis. BCH domain can also interact with itself (homophilic) or with other homologous BCH domains (heterophilic). For clarity, proteins that bind outside the BCH domains (heterophilic) are not shown here although they could co-regulate speciﬁc processes with the interactome of BCH domains.

2.2. Multiplicity and versatility of BCH domain-containing proteins adding to its evolutionary importance and diverse functionality [26]. BCH domain is completely absent in prokaryotes and the 2.2.1. BCH domain defines a distinct functional subclass of the CRAL/ most primitive BCH domain can be identified from slime mold TRIO superfamily (Dictyostelium), coanoflagellate (Monosiga), alveolates (Plasmo- Although initially identified as a region of high protein se- dium, Cryptosporidium), green alga (Chlamydomonas) and yeast. quence homology between BNIP-2 and Cdc42GAP, BCH domain Similarly, CRAL_TRIO domains are identified in many lower species also shares 14% sequence identity with the CRAL_TRIO domain, of alveolates. This indicates that BCH domains evolved from their the lipophilic domain of the Saccharomyces cerevisiae Sec14p pro- ancestors more than 1500 Mya ago with the appearance of Protists tein, suggesting that both domains could have evolved from a com- [26,27]. mon ancestral sequence. CRAL_TRIO domain was first identified in These BCH-containing proteins can be divided into three un- cellular retinaldehyde binding protein (CRALBP) and Trio (a RhoG- ique subgroups, namely those associated with BNIP-2 type se- EF) [21]. Other proteins such as tyrosine phosphatase [22], a- quence (Group I), with macro-domain (Group II) and with tocopherol transfer protein [23], RasGAP neurofibromatosis type- RhoGAP protein domains (Group III) (Fig. 2). And most signifi- 1 protein (NF1) and several other RhoGEFs such as Dbl, Duo, Dbs, cantly, all these BCH domains contain a hallmark signature motif Kalirin [24], all are thought to possess similar protein domains. R(R/K)h(R/K)(R/K)NL(R/K)xhhhhHPs, where ‘h’ refers to any large However, each member binds diverse small hydrophobic ligands. and hydrophobic residue and ‘s’ is small and weakly polar resi- For examples, the CRAL_TRIO domain of Sec14L group binds phos- due; Ala, Thr, Gly, Ser. This motif is absent from any of the photidylinositol, aTTP has affinity for tocopherol and CRALBP binds CRAL–TRIO domains, including those from Trio and NF1. Trio retinaldehyde [25]. The ligand specificities of the RhoGEFs and Ras- and NF1 are two members of the CRAL–TRIO group that are most GAP groups (represented by NF1 proteins) is currently unknown. closely related to the BCH domains (Figs. 2 and 3A and B). Further However, unlike the BCH domains, none of the CRAL–TRIO do- analyses of the gene-structure and protein domain context indi- mains is known to be associated with any specific protein–protein cate that BCH domain-containing genes have evolved through interaction. Likewise, BCH domains are not known to interact with gene duplication, intron insertions and domain swapping events any lipid molecules inside the cells. This observation raises several and this divergence could have occurred as early as the appear- important questions. How do BCH domains evolve and attain such ance of protists during evolution. Interestingly, sequences coding versatile functions? Do BCH and CRAL–TRIO domains still share for BCH domains can undergo alternative RNA splicing, leading to some of their ancient features and exhibit lipids binding and pro- splicing variants of BNIP-2, BNIP-2-Similar [28] and BNIP-2 Extra tein–protein interaction capabilities? Long [18] from the Group I. At least 4 splicing variants are also We have recently conducted extensive genome-wide, cross- identified in BCH domain-containing, proline-rich RhoGAP species analyses with 100 CRAL_TRIO and similar domains, protein, BPGAP [29] from the Group III. Interestingly, conservation including the putative BCH sequences, from eight representative of certain alternatively spliced sequence such as species and compared the pattern of their sequence conservation, ‘‘YEEEKFKKRQKR’’ in BNIP-2b appears to be highly similar to a se- gene structure, including the exon/intron evolution, and examined quence present in BNIP-2 Homology (BNIP-H, or Caytaxin; [30,31] the possible structural folds of BCH domain based on computa- and this sequence is somewhat less similar to that of the corre- tional modeling [26]. The results reveal a clear distinction of BCH sponding region in BNIP-XLa (Fig. 2). Interestingly, in addition domains as a novel subclass within the CRAL_TRIO superfamily to alternative RNA splicing, different transcriptional initiation and highlight several unique differences between the two. To date, sites are also used by the gene locus BMCC1/PRUNE2 to encode BCH domains have been identified in at least 175 distinctive pro- three more extended isoforms of BNIP-XL (see details later). All teins from the slime molds, plants, yeasts, insects, fish to human, these dynamic features are likely to confer different properties C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2677

Fig. 2. Domain architecture and classifications of BCH domain-containing proteins. Domain architecture for the Group I, Group II and Group III type BCH-containing proteins as defined by the presence of single-BCH, or BCH domain that is associated with the macro and RhoGAP domains, respectively. The percentages indicate the extents of amino acid sequence identities compared to the prototypical BNIP-2 BCH domain. Together, they define a distinct subclass of proteins that could have evolved from a common ancestral module shared with primarily lipid-binding CRAL–TRIO protein domain. Shown also are amino acid sequence inserts within the BCH domains that arise from their alternative RNA splicing. Only some of the splicing variants are represented here. between the isoforms and also between the homologs, further Cdc42/Rac1, respectively (Fig. 3A, C and D). Among various BCH enriching for the functional plasticity among the BCH domains. domains, the BCH domain of BNIP-2 has been validated to contain Further sequence analyses also revealed significant conserva- a novel Cdc42-binding motif (285-VPMEYVGI-292) within its CRIB- tion in two GTPase-binding motifs, that are similar to the Rho- like region ([32]; Fig. 3C). Likewise, the RBD-like motifs in BNIP-S binding domain (RBD) and the Cdc42/Rac interactive binding [33], BNIP-XL [18] and Cdc42GAP [19] have also been experimen- (CRIB) domain commonly found in effector proteins of Rho and tally validated (Fig. 3D). Indeed, these GTPase-binding motifs play 2678 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

Fig. 3. Multiple sequence alignment and phylogenetic relationship among BCH domains. (A) Multiple sequence alignment of various BCH domains from Homo sapiens BNIP- 2a (NP_004321), BNIP-XLa (AY439213), BNIP-H (AAO63019), BNIP-Sa (AY078983), BPGAP1 (AF544240), Cdc42GAP/p50RhoGAP (Q07960), GDAP2 (CAI22738), TRIO (AAC34245), NF1 (P21359) and Saccharomyces cerevisiae Sec14p lipid-binding domain (P24280) were generated using ClustalW (http://www.ebi.ac.uk/Tools/services/web/ toolform.ebi?tool=clustalw2) and displayed by BOXSHADE. (B) Phylogenetic tree of the BCH domains from Homo sapiens BNIP-2a (NP_004321), BNIP-XLa (AY439213), BNIP- H (AAO63019), BNIP-Sa(AY078983), BPGAP1 (AF544240), Cdc42GAP/p50RhoGAP (Q07960), GDAP2 (CAI22738), TRIO (AAC34245), NF1 (P21359) and Saccharomyces cerevisiae Sec14p lipid-binding domain (P24280), were generated using ClustalW as above. (C, D) Multiple sequence alignment of BCH domains from Homo sapiens BNIP-2a (NP_004321), BNIP-XLa (AY439213), BNIP-H (AAO63019), BNIP-Sa (AY078983), BPGAP1 (AF544240), Cdc42GAP/p50RhoGAP (Q07960), GDAP2 (CAI22738), TRIO (AAC34245) with the CRIB motif (C) of Homo sapiens PAK1 (AAH50377), Saccharomyces cerevisiae STE20 (Q03497), Candida albican (CLA4 AAAB68613) or with the RBD (Rho-binding domain) motif (D) of Rhophilin-1 (NP_443156), PRK1/PKN (NP_002732) and Rhotekin (AAH17727). The requirement for the CRIB-like motif of BNIP-2 (highlighted in green box in Figure A), RBD-like motifs of BNIP-XL, BNIP-Sa and Cdc42GAP (highlighted in red box in Figure A) and the BCH-BCH binding motifs of BNIP-2 and BNIP-S (highlighted in brown box in Figure A) had been validated by experiments [18,19,32,33]. A cryptic motif ‘‘285-VPMEYVGI-292’’ that is critical for binding to Cdc42, is also shown in red box in Figures A and C. All sequence alignments described above were generated using the default setting with protein weight matrix Gonnet, gap open 10, gap extension 0.2, gap distance 5, iteration none and with NJ clustering. C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2679

Fig. 3 (continued) important roles in mediating cell morphogenesis, cell migration sequesters RhoA from inactivation by the adjacent GAP domain, and cell differentiation, as will be discussed in details next. Inter- thus preventing cell rounding and retraction [19] (Fig. 4E). In con- estingly, although all BCH domains confer BCH/BCH homophilic trast, the BCH domain of BNIP-Sa promotes cell rounding and and heterophilic interactions, it appears that different sets of mo- apoptosis by activating RhoA while simultaneously preventing tifs are employed for different groups. For example, the motif RhoA from inactivation by Cdc42GAP by forming a heterophilic ‘‘RRKMP’’ used on BNIP-2 [11] is only conserved in BNIP-H and BCH/BCH complex [33] (Fig. 4D). In comparison, the BCH domain BNIP-XL, but it is absent from BNIP-S, Cdc42GAP and BPGAP1 of BNIP-XL inhibits cell transformation by sequestering oncogenic (Fig. 3A). Indeed, mutating the corresponding region in Cdc42GAP Lbc RhoGEF from activating RhoA [18] (Fig. 4C). Extending this, failed to disrupt its ability to form homophilic and heterophilic the BCH domain of the brain-specific BNIP-H/Caytaxin promotes complex [11]. BNIP-S, on the other hand, utilizes a separate motif neurites outgrowth by acting in concert with the other non-BCH 215-ATWYVKA-221 for its homophilic interaction (Fig. 3A) entity as a scaffold to traffic glutaminase on a kinesin motor and whereas it binds to Cdc42GAP at a unique GAP-Binding Motif (aa towards the tip of differentiating neurons [30,35] (unpublished 133–147) that overlaps with the Rho-binding region within an ex- data) (Fig. 4B). Last, but not least, the BCH domain of BPGAP1, a tended RBD-like motif (aa 133–177) [33]. Taken together, BCH do- homolog of Cdc42GAP, promotes ERK activation and cell migration mains have evolved to acquire multiple functional motifs that by collaborating with the RhoGAP domain and its proline-rich mo- contribute to its functional versatility. tifs that interact with cortactin [36], endophilin-2 [37], Mek2 and Pin1 isomerase [38] (Fig. 4F). The mechanistic action and the pos- 2.2.2. The BNIP-2 and BPGAP family proteins and their interactomes – sible regulation for these versatile processes are further discussed mechanisms of action and regulation below. To further explore the extents of distribution, function and regulation of BCH domain-containing proteins, several novel se- 2.2.2.1. BNIP-2 with morphogenesis, cell growth, apoptosis and quences encoding multiple homologs of BNIP-2 were identified differentiation. Since the earlier work that described BNIP-2 as a from human and other species and their biochemical and cellular partner for the pro-survival Bcl2 and E1B 19 kDa [10], we and oth- functions delineated. In the following sections, we will highlight ers have delineated the biochemical and cellular functions of BNIP- how BNIP-2, BNIP-S, BNIP-XL, BNIP-H, BPGAP1 and Cdc42GAP em- 2 as a core regulatory protein in multiple signaling gateways, ploy their unique functional motifs in their BCH domains to engage including the FGF receptor tyrosine kinase [9], Cdo receptor [34], multiple targets and how they act in concert with other protein do- Cdc42 [9,12,32], Cdc42GAP [9,11,32] which all engage its BCH domains to give rise to emerging functions and properties (Table 1 mains (Fig. 5A). In particular, expression of BNIP-2 in epithelial and and Fig. 4). For examples, in the context of 3G-signalome, the fibroblast cells induce extensive changes in cell morphology and BCH domain of BNIP-2 induces cell protrusions by activating membrane protrusions by targeting and activating Cdc42 with its Cdc42 via a yet unidentified regulator [32] or acting as a scaffold unique Cdc42-binding motif 285-VVMEYVGI-292 present within for the activation of Cdc42 that promotes muscle differentiation the core CRIB-like motif of the BCH domain ([32]; Fig. 5A). An adja- upon coupling to the active myogenic Cdo receptor and p38-MAPK cent region to this motif is also required for the concerted activa- activation [34] (Fig. 4A); whereas the BCH domain of Cdc42GAP tion, but the target protein, believed to be a GEF that help 2680 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

Table 1 BCH domain as a multifunctional regulatory scaffold domain. BCH domain-containing proteins regulate diverse signaling pathways and functions by utilizing their conserved BCH domains and their interplay with other associated domains.

Proteins Target proteins Functions Regulation BNIP-2 family BNIP-2 (Bcl2 and 19 kDa Targets of BCH domain: E1B 19 kDa (1) Binding to viral E1B 19 kDa protein (1) Homophilic or heterophilic interac- interacting protein-2) and Bcl2 [10]; Cdc42 [11,12,32], and Bcl2 promotes cell survival [10] tions via BCH domains affects Cdc42 BNIP-2 [11] and Cdc42GAP (2) Substrate of FGFR [9]. Its C-terminus activity [11] (p50RhoGAP) [11], Cdo receptor [34]; shares high sequence homology with (2) Phosphorylation by FGFR disrupts its FGF receptor-1 [9]; Rho and RhoGEF the N-terminus of Cdc42GAP, hence binding to Cdc42 and Cdc42GAP [9] (unpublished data) termed BCH domain [11] (3) Estrogen treatment reduces BNIP-2 Targets of other domain: caspases (3) BCH domain binds Cdc42 via a CRIB- expression and its pro-apoptotic [41], granzyme B [42] like motif and leads to drastic cell activity [43] elongation and membrane protrusions (4) Cleavage by caspases and granzyme B where it is localized to [32]. It also might lead to the release of pro-apop- forms BCH/BCH complex with itself totic BCH-containing fragment [41,42] or Cdc42GAP via a different motif [11] (5) Splicing within BCH domains gener- (4) Promotes myogenic and neuronal dif- ates BNIP-2a and BNIP-2b with poten- ferentiation by acting as a scaffold that tially different properties and links Cdo receptor to activation of functions Cdc42, leading to JLP-scaffolded acti- (6) Contextual signaling: versatility in vation of p38a/b [34,40] engaging different Rho GTPases and (5) Human BNIP-2 is cleaved by caspases their GAPs or GEFs, leading to con- 3, 6, 8, 10 in vitro, or by caspases 3, 8 text-dependent regulation of GTPase and 9 in cells [41] signaling (6) Also subjected to cleavage by granzyme B [42] (7) Contains a putative kinesin-binding motif for intracellular trafficking; possibly acting as scaffold for cargo trafficking [35] (8) Retards MDCK epithelia cell spreading and collective cell migration by activation of Rho (unpublished data) (9) Binding to BPGAP1 enhances its GAP activity towards Rho and reducing cell proliferation (unpublished data) BNIP-H (BNIP-2 Homology; Targets of BCH domain: glutaminase (1) Specific expression in brain, especially (1) Intramolecular inhibition that is CAYTAXIN) KGA [30], peptidyl–prolyl isomerase in the cerebellum and hippocampus released by NGF stimulation or active Pin1 [55] [30] Mek2 [55] Targets of other domain: kinesin-1 (2) Mutation in ATCAY gene leads to Cay- (2) Binding to CHIP (E3 ligase) and poly- light chain [35], caspases 3, 7 [57], man ataxia characterized by hypoto- ubiquitination is likely for degrada- kinesin-1 heavy chain (unpublished nia, mental dysfunction and tion [56] data) cerebellar ataxia [31] ⁄⁄Binding sites of Mek2 [55] and (3) Promotes neuronal differentiation and Implicated in disease: Mutation in ATCAY Ubiquitin E3 ligase CHIP [56] have relocalises glutaminase to neurite ter- gene leads to Cayman ataxia [31] and is not been mapped minals and regulates glutamate pro- also linked to the ‘‘jittery’’, ‘‘hesitant’’ duction [30] and ‘‘sidewinder’’ ataxia models in mice (4) Transports mitochnodria on kinesin-1 [46,47] [35] (5) As an adaptor/scaffold for transporting KGA on kinesin-1 along microtubules of neurites (unpublished data) (6) Binds E3 ligase CHIP and undergoes polyubiquitination [56] (7) As a presynapse substrate of caspases- 3 and -7. Contains 102-DETD-105 cleavage motif, releasing BCH fragment that might regulate Mek2 signaling [57] (8) Interacts with Pin1 upon NGF stimulation or stimulated by active Mek2; Pin1 disrupts the binding between BNIP-H and glutaminase [55] BNIP-XL (BNIP-2 Extra Targets of BCH domain: RhoA (1) One of the 4 splicing variants encoded (1) Alternative splicing encoded from the Long; BMCC1-isoform 4) (strongest binding for T19N, F30L but by the BMCC1 gene (also known as BMCC1 locus; further splicing within no binding to G14V and Q63L), RhoC PRUNE2). BNIP-XL (BMCC1 isoform- BCH domain generates BNIP-XLa and but not RhoB [18]; BCH domain 4) itself also undergoes alternative BNIP-XLb interacts with DH–PH domain of Lbc splicing within BCH domain, generat- (2) BMCC1 expression is downregulated RhoGEF and possibly p115RhoGEF ing BNIP-XLa and BNIP-XLb [18] after NGF-induced differentiation but also [18] (2) Associated with neuronal apoptosis upregulated during the NGF-deple- Targets of other domain: Multiple and is expressed in brain, spinal cord tion-induced apoptosis [66] caspases [41] and dorsal root ganglia; predomi- (3) Subjected to cleavage by caspases [41] Lbc binds to N-terminus and BCH nantly in the neurons of the cranial Implicated in disease: domain of BNIP-XL via its proline nerve motor nuclei and motor neurons (1) BMCC1-3 is a good prognosis marker region and DH–PH domains, of the spinal cord [66] for neuroblastomas in children [66] respectively [18] C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2681

Table 1 (continued)

Proteins Target proteins Functions Regulation BMCC1/PRUNE2 binds GTP, 8-oxo- (3) BNIP-XL restricts binding of RhoA with (2) BMCC1-1 is identiﬁed as a susceptibil- GTP and putative protein targets its RhoGEF Lbc, thus reducing RhoA acti- ity gene for Alzheimer’s disease [71] identiﬁed but not yet validated [69] vation and cellular transformation [18] and BMCC1-2 as a biomarker for leio- (4) Also binds to p115RhoGEF but the myosarcomas [72] function is unknown [18] (3) BMCC1-3 and BNIP-XL (BMCC1-4) is (5) Cleaved by caspases 3,6, 7, 8, 9 and 10 proapoptotic protein in neuronal cells in vitro, or by caspases 8 and 9 in cells [66] and an antagonist of Rho-medi- [41] ated cellular transformation [18], (6) Binds GTP and 8-oxo-GTP and several respectively putative protein targets [69], and also microtubule–associated protein, MAP6 [93] BNIP-S (BNIP-2 Similar; Targets of BCH domain: RhoA (1) Induces apoptosis by targeting (1) Splicing variants, BNIP-Sa and BNIP- BNIPL) (stronger preferences for T19N but no Cdc42GAP and RhoA with its BCH Sb. BNIP-Sa induce cell rounding and binding to G14V) [33], Bcl2 [74], domain. This disrupts Cdc42GAP and apoptosis whereas BNIP-Sb is local- BNIP-S [28], BNIP-2 [28], Cdc42GAP RhoA association thereby increasing ized to the nucleus and has no effect [28,33,74], proto-Lbc RhoGEF [18] RhoA activation for cell rounding and on cell morphology [28] ⁄⁄Binding sites of the cell- apoptosis [33] proliferation-related proteins MIF (2) BCH domain of BNIPL-2 interacts with and GFER have not been mapped [73] Bcl2 and Cdc42GAP during apoptosis [74] (3) BNIPL-2 promotes invasion and metastasis via Cdc42 activation and upregulation of CD44 in human hepatocellular carcinoma [75] RhoGAP Family Cdc42GAP Targets of BCH domain: RhoA [19], (1) GAP for Cdc42 and Rho [77,81,83,85] (1) Auto-inhibition [88] (p50RhoGAP;ARHGAP1) Rab5 [89], Rab11 [89], BNIP-Sa (2) BCH domain acts as a local modulator (2) Sequestered by BNIP-Sa from inacti- [28,33] to sequester RhoA from inactivation vating RhoA, leading to cell rounding ⁄⁄Binding sites of Nudel has not been by its adjacent RhoGAP domain [19] and caspase-independent apoptosis mapped [92] (3) Target to endosome via BCH domain [33] and form complex with Rab11 and Implicated in disease: Rab5 [89] (1) Up-regulated in Waldenstrom macro- (4) Nudel sequesters Cdc42GAP at the globulinemia [79] leading edge of the cell to increase (2) Homozygous knockout embryo and active Cdc42 in presence of phosphor- newborn mice have reduced organ ylated ERK [92] and body size, due to elevated JNK- (5) Promotes cell migration [81,82] mediated apoptosis [81,83,84] (6) Suppress muscle cell differentiation by inactivating Cdc42 [34] BPGAP1 (BCH containing, Targets of BCH domain: RhoA, BNIP-2 (1) Induces membrane protrusions and (1) Splicing variants, BPGAP1–4 [29] proline-rich RhoGAP- BCH domain; K-Ras and SmgGDS cell migration via interplay of BCH (2) Intramolecular inhibition at proline- 1;ARHGAP8) (unpublished data) domain, GAP domain and proline-rich rich region, released by active Mek2 Targets of other domain: Cortactin region [29] acting as scaffold to promote interac- [36], Endophilin 2 [37], Pin1 [38], (2) Induces motility by translocating tion between Pin1 and BPGAP1 and Mek2 [38] Cortactin to cell periphery [36] suppresses BPGAP1-induced cell ⁄⁄Binding sites of Cdc42 and Mek2 (3) Induces EGFR endocytosis and ERK motility and ERK activation [38] have not been mapped activation by coupling to endophilin- Implicated in disease: Upregulated in pri- 2 [37] mary colorectal tumors [78] and cervical (4) BCH induces chronic K-Ras/Erk activa- cancer [80] tion and PC12 neuronal differentiation that is suppressed by another Ras modulator, SmgGDS (unpublished data) (5) BNIP-2 promotes BPGAP1’s GAP activity towards RhoA and downregulates cell proliferation (unpublished data)

activate Cdc42, remains elusive. Although BCH domains are highly interaction with Cdc42 are abrogated upon binding and phosphor- conserved across different members, this Cdc42-binding motif for ylated by FGF receptor [9]. Although the BCH domain of BNIP-2 ap- BNIP-2 is only conserved in BNIP-XL and BNIP-H (Fig. 3C). How- pears insensitive towards either the GDP-bound or GTP-bound ever, no binding of Cdc42 have been observed for these two pro- form of Cdc42, its strong preference for the dominant negative teins, indicating that unique structures surrounding this motif form of Cdc42–T17N and its near-complete loss of binding with could play an important role in the substrate recognition. the constitutive active form, Cdc42–G12V [12] support the abso- BNIP-2 utilizes another distinct motif 217-RRKMP-221 to medi- lute requirement of Gly-12 to promote their interaction, rather ate its homophilic and heterophilic BCH/BCH interactions. The than depending on the activity per se. The ability of BCH domains homophilic interaction of BNIP-2 or its heterophilic binding with to still recognize either active or inactive form of Cdc42 (and Rho in Cdc42GAP could affect their ability to promote the inactivation of other examples) appears to be consistent with its scaffold function Cdc42 in vitro, although such interactions do not affect their bind- to bring any forms of GTPase and their immediate regulators in ing to Cdc42 [11]. Interestingly, such interaction as well as its context-dependent manner. The signiﬁcance of this recurring 2682 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

A B C

D F

Fig. 4. Functional plasticity and mechanisms of action by the BCH domains. BCH domains are involved in targeting small GTPases, their GAPs, GEFs, signaling receptors, cargo enzymes or other cellular proteins to elicit diverse cellular responses. These are shown by the following models, depicted clockwise. (A) BNIP-2 induces extensive cell elongation and membrane protrusions in single epithelial cells via the BCH domain binding to Cdc42 [32]. It also acts as a scaffold protein that couples the promyogenic receptor, Cdo and activation of Cdc42 with the activation of p38a/b MAPK on the scaffold, JLP. The alliance of these two scaffold proteins promotes myogenic differentiation [34]. (B) BNIP-H (or Caytaxin) promotes neurite outgrowth by engaging its BCH domain as a scaffold/adapter to traffic and localize KGA (kidney-type glutaminase) towards the neurites termini [30]. BNIP-H itself can be trafficked by kinesin motor in hippocampus [35] supporting our observations that BNIP-H can act as an adaptor for transportation of cargo proteins or signaling complexes along neurites (unpublished data). (C) BNIP-XL suppresses stress fiber formation and cellular transformation by engaging its BCH domain to sequester RhoA from activation by RhoGEF, proto-Lbc; The N-terminal region of BNIP-XL interacts with the proline-rich region of Lbc. These concerted interactions reduce the level of active RhoA, leading to disruption of stress fibers and Lbc-induced cell transformation [18]. (D) BNIP-Sa induces cell retraction, cell rounding and apoptosis, all via its BCH domain sequestering RhoA and also by displacing Cdc42GAP/p50RHoGAP from inactivating RhoA via its BCH/BCH interaction with Cdc42GAP. Together, such mechanisms sequester RhoA from immediate inactivation by Cdc42GAP thereby facilitating the activation of RhoA, leading to cell rounding and apoptosis [33]. (E) The BCH domain of Cdc42GAP sequesters RhoA from being inactivated by the adjacent RhoGAP domain, thus reducing GAP-mediated Rho inactivation and preventing cell rounding [19]. (F) BPGAP1 enhances cell motility and ERK1/2 activation. The BCH domain of BPGAP1 induces short pseudopodia in MCF7 cells [29]. The domain interplay of the BCH, PRR (proline-rich region; labelled P) and RhoGAP domains lead to enhanced cell motility and ERK1/2 activation. For example, through the PRR, BPGAP1 mediates translocation of cortactin from cytosol to membrane periphery for cell migration [36] and it also engages EEN/Endophillin II to increase EGF receptor endocytosis and ERK1/2 activation [37]. In addition, BPGAP1 can be negatively regulated by the peptidyl-prolyl cis/trans isomerase Pin1 via targeting both the PRR and the RhoGAP of BPGAP1; thereby suppressing the BPGAP1-induced cell motility and ERK1/2 activation. Intriguingly, such interaction between BPGAP1 and Pin1 is promoted by the active regulatory scaffold, Mek2 by aiding the relief of the auto-inhibited site in BPGAP1 [38]. theme is also seen in the binding of Rho to the BCH domains of [40]. While providing one novel mechanism for specificity of BNIP-S, BNIP-XL and Cdc42GAP. p38a/b activation during myogenic and neuronal differentiation, Extending its involvement in anchoring membrane receptor sig- forming scaffold alliances is becoming a distinctive mode of cell naling, BNIP-2 has also been shown to play a crucial scaffold func- surface receptor signaling. tion to support Cdo receptor activation of Cdc42, in concert with Adding to its versatility, human BNIP-2 is subjected to cleav- p38a/b-MAPK activation by another scaffold protein, JLP – all trig- age by caspases 3, 6, 8, and 10 in vitro or at least by caspase 3, 8 gered under the same Cdo receptor upon activation through N-cad- and 9 inside the cells [41]. It can also be cleaved by granzyme B herin ligation. This mechanism brings both scaffold proteins [42]. All these processing occur outside the BCH domain of BNIP- together to activate myogenic differentiation [34,39]. Interestingly, 2. Granzyme B triggers apoptosis via the cleavage of a repertoire Cdo receptor binding sites have been mapped to encompass aa of cellular proteins, leading to caspase activation and mitochon- 261–292, which include the Cdc42-binding motif. The intricate drial depolarization. As Granzyme B cleaved recombinant BNIP-2 mechanism underlying the activation of Cdc42 by this Cdo-Bnip- in vitro and endogenous BNIP-2 was cleaved during the natural 2-Cdc42 complex remains to be determined. Given the increasing killer cell-mediated killing of tumor cells [42] these observations evidence that BCH domains can also interact with GEFs and GAPs, raise the possibilities that cleavage of BNIP-2 by caspases and it is tempted to speculate that either one such GEF is to be re- granzyme could lead to well controlled and timely release of cruited to activate the pathway, or/and a Cdc42GAP has been inac- the potent pro-apoptotic BCH domains. However, the mechanism tivated. In the later scenario, overexpression of Cdc42GAP in the by which BNIP-2 causes apoptosis remains unclear, although we same model system of C2C12 myoblasts suppressed the myogenic have shown that the corresponding BCH domain in BNIP-S lead differentiation. Through the same module of Cdo-Bnip-2-Cdc42, to apoptosis by activating through a caspase-independent but BNIP-2 is also involved in promoting neuronal differentiation Rho-dependent pathway [33]. As BNIP-2 harbors a Rho-binding C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2683

A BNIP-2

BCH domain signature motif R(R/K)h(R/K)(R/K)NL(R/K)xhhhhHPs 235 250 granzyme putative kinesin- cleavage site binding motif IEAD 94 EFEWED 99 35 38 183 211 259259 294294

1 RBD BCH CRIB 314 167 -like -like EF-hand motif EIDLDGLDTPSE 217 221 285 292 79 90 RRKMP VPMEYVGI BCH-BCH Cdc42-binding 83 caspase 86 binding motif motif cleavage site

• The BCH domain binding sites for FGFR, Bcl2, E1B 19- KDa and Cdc42GAP have not been mapped

• Cdo receptor binding site; aa 261-292

B BNIP-H (CAYTAXIN) BCH domain signature motif R(R/K)h(R/K)(R/K)NL(R/K)xhhhhHPs 259 274 kinesin-binding WED motif ELEWED 115 120 207 235 283 318 1 191 RBD BCH CRIB 371 -like -like 102 105 DETD caspase-3 cleavage site site I site II

242 RRRMP 245 putative BCH-BCH binding motif

KGA binding site: Pin1 binding site: a) Site I – residues 191-235 a) Site I – residues 191-206 b) Site II – residues 288-331 b) Site II – residues 286-332

•BCH domain binding sites for Mek2 and CHIP have not been mapped

Fig. 5. Linguistics of BCH domains. Schematic diagrams of (A) BNIP-2, (B) BNIP-H, (C) BNIP-XL with BMCC1-1/PRUNE2 and (D) BPGAP1, as the representative members of BCH domain-containing proteins, highlighted with their known or putative functional motifs, including the Rho-binding domain, CRIB-like motif, BCH signature motif, BCH/BCH interaction motif, caspase or granzyme cleavage sites and kinesin-targeting motifs. Note: Ã indicates the putative Rho binding domain based on close homology to known RBD in BNIP-XL, BNIP-S and Cdc42GAP (see Fig. 3D) while ÃÃ indicates putative Cdc42/Rac interactive binding domain based on close homology to known CRIB-like motif in BNIP-2 (see Fig. 3C). For BMCC1/PRUNE2, the sequence used is from accession number BAJ08045.1 [69]. domain (RBD)-like sequence similar to those in BNIP-S and to inactivate Rho, leading to greater loss of stress fibers and reduced Cdc42GAP (Fig. 3A and D), the released fragment of BCH with cell proliferation (unpublished data). Taken together, these findings RBD might activate Rho towards apoptosis. Consistent with its support the notion that both BNIP-2 and BNIP-S are likely to act as role in regulating apoptosis, expression of BNIP-2 mRNA has tumor suppressor instead of being proto-oncogenic. been shown to be downregulated upon estrogen treatment, a re- To explore the multifunctional nature of BNIP-2, we recently gime thought to confer neuroprotection against neurodegenera- employed fluorescent reporter construct coupled to Total Internal tive diseases such as Alzheimer’s disease [43]. Its expression is Reflection Fluorescent (TIRF) microscopy that measured the activ- also downregulated in the heart upon coxsackievirus B3 infec- ity of BNIP-2 especially its dynamics near the membrane. We ob- tion in mice [44], but up-regulated in lung cancers [45], further served that BNIP-2 undergoes dynamics distribution between the indicating its complex nature of regulation in cell growth and endosomes and cell protrusions along the microtubules, but with cell death. We have further observed that BNIP-2 retards MDCK major concentrations being detected at the protrusive tips. And, epithelial cells spreading and collective cell migration by activat- this dynamics is recapitulated by the expression of the BCH do- ing Rho/ROCK/myosin signaling cascade (unpublished data) main alone (unpublished data). In this regard, the N-terminus of whereas in fibroblasts, BNIP-2 enhances the ability of BPGAP1 BNIP-2 carries a putative kinesin-binding WED motif 94- 2684 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

C BNIP-XL BCH domain signature motif R(R/K)h(R/K)(R/K)NL(R/K)xhhhhHPs 668 683 caspase cleavage site P-loop ETDNSDL DSGPGWS 170 176 332 339 616 644 692 727 1 600 RBD BCH CRIB 769 -like -like

650 RRRMP 654 putative BCH-BCH binding motif putative kinesin- binding motif coiled-coil 1139-DLDWDD-1144 sequence P-loop 1283 / 1311 2654 / 2660 1 2922 BMCC1 BCH 3062 (PRUNE2)

617 proline-rich sequence 2745 PRUNE2 Homology 1916 / 1921 BNIP-2 Homology

D BPGAP1

BCH domain signature motif R(R/K)h(R/K)(R/K)NL(R/K)xhhhhHPs 99 114

51 75 123 158 R233 1 34 167 RBD BCH CRIB P GAP 433 -like -like

176 189 PPPTKTPPPRPPLP 256 Pin1 PPI- 260 binding motif DDYGD

•Residues P184 and P186 (underlined) are critical for binding to the SH3 domain of Cortactin and Endophilin-2.

•186-PPLP-189 motif is the auto-inhibited region which is relieved upon binding to active Mek2. This motif is also targeted by WW domain of Pin1. The peptidyl-prolyl isomerase (PPI) domain of Pin1 targets the GAP domain.

•Residue R232 is the arginine finger motif for the RhoGAP domain where Rho inactivation is required for endophilin 2-mediated endocytosis of EGFR

•The BCH domain binding sites for Rho, Ras and SmgGDS have not been mapped

Fig. 5 (continued)

EFEWED-99 similar to that identified on BNIP-H (Caytaxin) (Fig. 5A especially in the cerebellum and hippocampus [30]. Several forms and B) that is required for its trafficking inside the cells [35]. This of mutations at the human and mouse BNIP-H loci have been observation suggests that BNIP-2 could itself translocate towards linked to a recessive congenital ataxia in human Cayman cerebellar protrusive membranes to exert its scaffold functions there or it ac- ataxia (thus named ‘‘Caytaxin’’) [31], and also linked to the ‘‘jit- tively recruits and targets cargos (possibly signaling complexes) to tery’’, ‘‘hesitant’’ and ‘‘sidewinder’’ ataxia models in mice [46,47]. the right cellular compartments. The physiological significance and This disease is associated with hypotonia, variable psychomotor the mechanism of this function remain to be explored. Further- retardation, truncal ataxia and intention tremor, scoliosis, and ocu- more, it remains to be determined how alternatively spliced vari- lar abnormalities. Interestingly, the two types of BNIP-H gene ants of BNIP-2 could be generated and regulated to impact on mutation associated with Cayman ataxia were predicted to cause any functions that have been described so far. Taken together, point mutation (e.g. S310R) or truncation within the BCH domain BNIP-2 regulates diverse cellular functions by engaging multiple [31]. BNIP-H deficiency in rats also leads to generalized dystonia partners and is likely to be activated under different cellular con- [48], another neurologic movement disorder characterized by sus- text, pending on the biochemical signals and their impending tar- tained muscle contractions that produce abnormal movements or get proteins. postures. Furthermore, expression of BNIP-H is regulated in a developmental and spatial-specific manner and present in presyn- 2.2.2.2. BNIP-H (Caytaxin) with neuronal differentiation, neurotrans- aptic cytosol [49] while its deficiency could disrupt signaling in mission and ataxia. Unlike BNIP-2 which is ubiquitously expressed, cerebellar cortex [50]. While all these genetics studies firmly expression of BNIP-H (or Caytaxin) is highly specific to the brain, establish that BNIP-H is required for proper brain function, nothing C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2685 is known about how BNIP-H executes its function(s) and regulated region 1) located at the human chromosome 9 (9q21.2) [66–69]. at the molecular, cellular and physiological levels. To achieve this This gene, spanning 295 kbp, is also known as PRUNE2 as it encodes objective, we employed proteomics/MALDI-TOF and identified a the largest isoform (BMCC1 isoform-1/BMCC1-1/PRUNE2-1/ kidney-type phosphate-activated glutaminase (KGA) as a binding NM_015225, 3088aa; or BAJ08045.1, 3062aa) that harbors a region partner for BNIP-H. KGA is a metabolic enzyme responsible for of homology with the PRUNE, a His-Asp-Asp super family proteins the production of the glutamate which can be used as a form of at its N-terminus. It also harbors a coiled-coil region, a proline-rich neurotransmitter. We showed that BNIP-H expression is upregu- region and a putative P-loop sequence (Fig. 5C). A slightly shorter lated in retinoic acid-induced neuronal differentiation in embry- isoform consisting of 3062aa may result from an alternative splic- onic carcinoma cell P19 and BNIP-H could relocalise KGA to the ing [69]. Interestingly, the Prostate Cancer Antigen 3 gene (PCA3), a tips of neurons and control the steady-state level of glutamate highly specific biomarker upregulated in prostate cancer, is located [30]. Our model suggests that deregulated level of glutamate in within the intron 6 of BMCC1 but in the opposite orientation [67]. BNIP-H-deficient individuals could render glutamate excitotoxicity In contrast, BMCC1 isoform-2 (BMCC1-2/PRUNE2-2/C9orf65/ or/and deregulated glutamatergic activation. This could then lead BC019095/NM_138818) comprises the first 6 exons of BMCC1 to ataxia and dystonia and possibly other excitotoxicity-based neu- but it terminates immediately upstream of the PCA3 gene. The iso- ronal damage [51–54]. Subsequently, we have also identified sev- form-3 (BMCC1-3/BMCC1/ABO50197; 2724aa) does not overlap eral more novel protein targets for BNIP-H, among them are the with BMCC1-2 but it comprises 13 distinct exons (exons 7–19) heavy chain of kinesin-1 (previously named KIF5B) motor, Rab positioned immediately downstream of PCA3 [66]. BNIP-XL is the small GTPases, Mek and Pin1 isomerase. We showed that nerve isoform-4 (BMCC1-4/KIAA0367/AY43213; 769aa) which has the growth factor stimulates the interaction of BNIP-H with pepti- start site located still further downstream within the second exon dyl–prolyl isomerase Pin1, a process reproduced by the presence of BMCC1-3. All three isoforms-1, -3 and -4 encode the highly con- of a constitutive active form of Mek2 such that it involves the served BCH protein domain. Interestingly, BNIP-XL mRNA itself can release of an auto-inhibitory intramolecular interaction on undergo alternative RNA splicing, generating a 769 amino acid- BNIP-H. Consequently, two binding sites for Pin1 and KGA have long protein (BNIP-XLa) and a 732 amino acid-long protein been mapped and shown to be overlapping (Fig. 5B) such that (BNIP-XLb)(Fig. 2). The shorter variant results from alternative Pin1 disrupts the interaction between BNIP-H and KGA [55]. splicing of exons 11 and 12 (with respect to exon numbering in Interestingly, the mutation S310R found in a mutant form of BNIP-XL), which introduces an in-frame stop codon. Multiple se- BNIP-H lies within one of the binding sites overlapping with those quence analyses of BNIP-XL with BNIP-2, BNIP-Sa, BNIP-H/Cay- of Pin1 and KGA. Its precise impacts on BNIP-H binding to all taxin and BPGAP1 indicates the closest homology to BNIP-2 known cellular partners still remain unclear. across the entire protein (58% identity, 74% similarity) whereas BNIP-H has been shown to act as an adaptor to transport mito- its BCH domain is most similar to that of BNIP-H/Caytaxin (76% chondria on the kinesin-1 light chain along the neurites [35].A identity, 90% similarity). kinesin-binding motif (115-ELLEWED-120) has been identified. All isoforms of BMCC1 have been linked to, or identified as pos- How BNIP-H functions to transport KGA and regulated by differen- sible biomarker for, various human pathologies. For example, tiation signals and whether there exist cargo proteins other than BMCC1-1 was initially thought to offer an improved diagnosis for KGA, all remain to be further investigated. Interestingly, unlike prostate cancer because of its upregulation with PCA3 gene in pros- BNIP-2 which targets Cdc42, BNIP-H interacts predominantly with tate cancers and in response to androgen [67]. However, both Rab GTPases and co-localized with them in endosomes and along genes may not appear to be tightly co-regulated [70]. BMCC1-1 neurites, reflecting their tight regulation by vesicular-based trans- has also been identified as a susceptibility gene for Alzheimer’s dis- port system (unpublished data). Furthermore, BNIP-H undergoes ease [71] while BMCC1-2 as a biomarker for leiomyosarcomas [72]. polyubiquitylation by CHIP, a E3 ligase protein, probably for its BMCC1-3 and BNIP-XL (BMCC1-4) have been shown to be a degradation [56]. Similar to BNIP-2 and BNIP-XL (see later), proapoptotic protein in neuronal cells [66] and an antagonist of BNIP-H is also subjected to cleavage by caspases. Interestingly, Rho-mediated cellular transformation [18], respectively. Expres- BNIP-H is shown to be a presynapse substrate for caspase-3 and sion of BMCC1-3 enhances neuronal apoptosis upon NGF depletion caspase-7 that target the highly conserved 102-DETD-105 motif, and is expressed in brain, spinal cord and dorsal root ganglia releasing the BCH domain that appears to regulate Mek2 signaling (DRG). It also offers a good prognostic marker for childhood through a yet unknown mechanism [57]. Since caspases activity is neuroblastomas [66]. Overall, BMCC1-1 mRNA and proteins are essential for many other non-apoptotic functions in cell signaling, predominantly expressed in the neurons of the cranial nerve motor learning and memory [58–60], differentiation of neural stem cells nuclei and motor neurons of the spinal cord and to lesser extent, in [61], dendritic [62]/axon [63] pruning and cell migration [64,65], other nerve tissues [68,69]. DRG neurons express higher levels of it remains to be seen how these processes are linked to the control BMCC1-1 in their soma compared with adjacent cells. Moreover, of neuronal differentiation mediated by BNIP-H. As dysregulation their expression is most profound in adult nerve tissues than those of BNIP-H function lead to cerebellar ataxia, dystonia, and possibly in fetal or neonatal nerve tissues. other related neurological disorders, our current attempt to under- These expression profiles suggest that various BMCC1 isoforms, stand the mechanistic function of BNIP-H and its checkpoints at including BNIP-XL, may contribute to the maintenance of mature cellular and molecular levels could pave way to our future under- nervous systems and inhibition of unwarranted cell growth and standing on the neuronal remodeling, plasticity and neurotrans- proliferation. However, the underlying mechanisms for their action mission at synapses, and with a long term objective of remain unknown. In contrast to the positive regulation of BNIP-2 identifying key strategic target(s) for intervention in such diseases. and BNIP-S on their cognate GTPases inside the cells, i.e. activating Cdc42 and RhoA, respectively, BNIP-XL affects actin cytoskeletal 2.2.2.3. BNIP-XL with tumor suppression and neuronal differentia- reorganization and suppresses cellular transformation by inhibit- tion. BNIP-XL initially represents the longest form of the BNIP-2 ing Rho signaling instead [18]. At the molecular levels, the BCH do- family proteins, with its sequence homology extended beyond main of BNIP-XL interacts with RhoA and RhoC but not with RhoB. the highly conserved BCH domain, hence we named it BNIP-2 Extra Furthermore, it binds to specific conformers of RhoA and also Long [18]. To date, it is shown to be one of the four major isoforms mediates association with the catalytic DH–PH domains of Lbc, a encoded via differential initiation sites from the complex BMCC1 RhoA-specific guanine nucleotide exchange factor (RhoGEF). Con- gene (for BCH motif-containing molecule at the carboxyl terminal sistent with the observations made by other members, BNIP-XL 2686 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 does not recognize the constitutive active G14V and Q63L mutants of the BCH domain. Full-length BNIP-S is present in the cytosol of RhoA but it targets the fast-cycling F30L and the dominant-neg- and its overexpression cause extensive cell rounding, leading to ative T19N mutants. Whereas overexpression of BNIP-XL reduces caspase-independent apoptosis. However, BNIP-Sb is localized to active RhoA levels without changing its level of expression, knock- the nucleus and fail to exert any changes in cell morphologies or down of BNIP-XL expression has the reverse effect. The constitutive cell fates [28]. The transcripts for both variants can be detected active G14V and Q63L are locked in an active conformation mim- in different cancer lines. It appears that the expression for the icking GTP-RhoA in vitro, because residues for GAP recognition in full-length version is more abundant in the male reproductive or- the P-loop (residues 12–19) and switch II (residues 62–68) regions gan of mice [28], or in the human placenta and lung [73]. Most sig- have been mutated. By contrast, the ‘fast-cycling’ F30L is spontane- nificantly, this apoptotic effect is induced by the BCH domain ously active and demonstrates enhanced GDP to GTP exchange, which triggers an initial phase of extension followed by extensive while retaining regular GTP hydrolysis rates catalyzed by GAPs. retraction. And this effect can only be blocked by co-expressing The dominant negative T19N mutant on the other hand, sequesters dominant negative mutant form of RhoA, indicating that this and inhibits multiple endogenous RhoGEFs, preventing RhoA acti- induction is likely to involve activation of Rho [33]. BNIP-S does vation leading to elevated GDP-RhoA levels in vivo. Collectively, not bind Rac1 or Cdc42. However, BNIP-S BCH domain binds to these results suggest that BNIP-XL inhibits distinct RhoA pathways, the unloaded or GDP-bound form of RhoA but not to the GTP- including stress fiber assembly and transformation. Such result is bound RhoA. And when co-expressed inside the cells, BNIP-S inter- consistent with its function as an antagonist for cell growth in acts strongly with the dominant negative mutant of RhoA, T19N, fibroblasts or as a pro-apoptotic factor in neuronal cells. Further but not with the constitutively active RhoA-G14V [33]. Sequence mutagenesis studies revealed that BNIP-XL utilizes both its BCH comparison between its BCH domain and REM Class 1 Rho-binding domain and the N-terminus to target the DH–PH moiety and the domains reveals a putative RBD-like motif which when deleted proline-rich region of Lbc, respectively [18]. Consequently, BNIP- lead to the loss of RhoA binding and failure to execute apoptosis. XL inhibits Lbc-induced oncogenic transformation. Given the All these observations indicate that BNIP-S binds to inactive importance of RhoA and RhoGEF signaling in tumorigenesis, RhoA and that could lead to its activation, leading to cell rounding BNIP-XL could suppress cellular transformation by preventing sus- and apoptosis. Interestingly, unlike BNIP-2 which uses the same tained Rho activation in concert with restricting RhoA and Lbc BCH/BCH binding motif ‘‘RRKMP’’ for its both homophilic and het- binding via its BCH domain. This could provide a general mecha- erophilic interaction with BCH domains, BNIP-S does not possess nism for regulating RhoGEFs and their target GTPases. The identi- such motif. Instead, it uses a separate motif 215-ATWYVKA-221 fication of a specific BNIP-XL mutant that uncouples GEF for its homophilic BCH/BCH interaction [33] (Fig. 3A), and this interactions while retaining RhoA binding should address the spe- interaction is required for its apoptotic activity [28]. The binding cific contributions of BNIP-XL in linking actin cytoskeleton rear- site for Cdc42GAP is instead located at the region proximal to the rangements and oncogenic transformation. Interestingly, some BCH domain (aa 133–147), overlapping part of the extended Rho- BMCC1 isoforms have been shown to bind GTP, its oxidized form binding region of the BCH domain. Consequently, overexpression 8-oxo-GTP (an agent known to cause mutagenesis and cell death), of BNIP-S can capture RhoA for further activation while separate the T-cell activation RhoGAP, TAGAP [69] and microtubule–associ- pools of BNIP-S sequester Cdc42GAP via their heterophilic interac- ated protein MAP6 [93]. While their mechanism of binding and tion. Acting in concert, this mechanism leads to RhoA activation, function remain to be established, these observations imply a com- cell rounding and apoptosis (Fig. 4D). However, it remains to be plex regulatory scaffold function by BNIP-XL and possibly other seen how BNIP-S binding to RhoA could lead to activation of RhoA BMCC1 isoforms in regulating GTPase signaling and other cell in the absence of negative regulation by Cdc42GAP. Perhaps, BNIP- growth/death signaling pathways. Adding to the dynamic roles of S could recruit a RhoGEF in their close proximity since dominant these proteins in neuronal cell growth and development, BNIP- negative mutant of RhoA could inhibit the pro-apoptotic effect of XL, similar to BNIP-2 and BNIP-H, can also undergo cleavage by BNIP-S. In this regard, BNIP-S has been shown to interact with pro- multiple caspases-3, 6, 7, 8, 9 and 10 in vitro and at least by casp- to-Lbc RhoGEF [18] but it remains to be confirmed by means of ge- ases 8 and 9 inside the cells, albeit at site distinct from that of netic knockdown and mutant studies to see if this binding is BNIP-2 [41] (Fig. 5C). Such actions could result in the release of specific and linked to this process. the BCH domain or its smaller fragments that are crucial in regulat- In addition to involvement of Rho signaling in its pro-apoptotic ing apoptosis or engaging small GTPases signaling in cell dynamics effect, BNIP-S has also been shown to interact with two cell-prolif- control. Surprisingly, unlike BNIP-2, BNIP-H and BNIP-S which all eration related proteins, MIF (macrophage migration inhibitory carry a highly conserved ‘‘E(L/F)EWED’’ kinesin-1 light chain bind- factor) and GFER (growth factor erv1 (Saccharomyces cerevisiae)- ing motif, BNIP-XL do not posses such a motif. Instead, a similar like) and help suppress colony formation and cell proliferation motif ‘‘DLDWDD’’ is found in its longer isoform BMCC1-1/PRUNE2 [73]. Another variant of BNIP-S (BNIPL-2), also interacts with Bcl- (aa 1139–1144) and BMCC1-3 (aa 777–782; [35]), which are lo- 2 and Cdc42GAP via its BCH domain [74]. Similar to the proposed cated outside the region of homology to the BNIP-2 Family proteins function of BNIP-2, BNIPL-2 is thought to suppress the anti-apopto- (Fig. 5C). It remains to be seen whether BNIP-XL traffics in the cell tic effects of Bcl2 and the activity of Cdc42GAP towards Cdc42. Fur- via other motor system or it could ‘‘piggyback’’ on these longer iso- thermore, BNIPL-2 can also promote invasion and metastasis via forms via their BCH/BCH interaction motif. Cdc42 activation and upregulation of CD44 in human hepatocellular caricinoma [75]. However, none of these studies had completely 2.2.2.4. BNIP-S with apoptosis. Comparing the functions of BNIP-2, delineated their curical binding sites on the BCH domains in order BNIP-H and BNIP-XL, BNIP-S (or BNIPL for BNIP-2-Like) represents to establish their molecular interaction and functional relationship. the most potent form of pro-apoptotic BCH containing protein despite sharing high sequence homology (86% similarity) with the 2.2.2.5. Cdc42GAP and BPGAP1 with cell morphogenesis, migration, BCH domain of BNIP-2 or 72% similar across the whole protein invasion and neuronal differentiation. RhoGAPs function as negative [28]. Phylogenetic analyses suggest that BNIP-S could have ap- regulators by activating the intrinsic Rho GTPase activity, convert- peared earlier than the others during the evolution [26]. BNIP-S ing their active GTP-bound state to the inactive GDP-bound state. can undergo alternative RNA splicing with the retention of an in- The human genome encodes more than 80 RhoGAPs with distinc- tron that introduces a non-sense mutation. This results in a trun- tive arrays of protein domain/motifs and exerts diverse physiolog- cated variant, BNIP-Sb, that is devoid of the last 71 amino acids ical outcomes [76]. These domain/motifs could potentially act as C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2687 regulator for either their GAP activity, subcellular localization or mechanism for regulating the local activity of Cdc42GAP towards connecting to various signaling networks via protein–protein Rho. This mechanism could have important bearings on previously interactions. The Cdc42GAP (p50RhoGAP or ARHGAP1; [77]) and unappreciated function of BCH domains or the ‘‘BCH-like’’ CRAL– its homolog BPGAP1 (or ARHGAP8; [29,78]) are two BCH do- TRIO domains in other RhoGAPs or RhoGEFs. In a mechanism sim- main-containing RhoGAPs (Fig. 2) that functions biochemically as ilar to the activation of RhoA by BNIP-S which blocks Cdc42GAP a GAP for Cdc42 and Rho. Expression of Cdc42GAP is up-regulated from acting on Rho through their heterophilic BCH/BCH binding in Waldenstrom Macroglobulinemia [79] whereas expression of [33], the function of Cdc42GAP can also be suppressed at the lead- BPGAP1 is elevated in primary colorectal tumors [78] and cervical ing edge of cells through its sequestration by Nudel, a cytoplasmic cancer [80]. However, the molecular bases and cellular outcomes regulator for dynein that is activated by active Erk [92]. Conse- for their elevated expression remain unclear. The Cdc42GAP regu- quently, more Cdc42 is activated. lates diverse functions, including cell migration [81,82] and muscle In comparison, we have identified BPGAP1 (or ARHGAP8), a cell differentiation by inactivating Cdc42 [34]. Its homozygous close homolog of Cdc42GAP, and showed that it activates cell pro- knockout embryos/new born mice displayed reduced organ and trusions and cell migration via the interplay of its BCH domain, a body size, owing to increased spontaneous JNK-mediated apopto- proline-rich region (PRR) and a GAP domain [29]. At least 4 alter- sis [81,83,84]. Although the biochemical and structural properties native RNA splicing variants have been described, including those of the GAP domain of Cdc42GAP are already well defined [77,85– present in the BCH domain (Fig. 2), thus adding to the versatility 87], little is known on how their GAP functions inside the cells and functional diversity in the BPGAP family proteins [29]. Specif- are regulated. In particular, the impact of the BCH domain that is ically, the BCH domain of BPGAP1 induces Cdc42/Rac-dependent located N-terminus proximal to the GAP domain remains largely membrane protrusions that are necessary for its physical and func- undetermined. tional coupling to cortactin and endophilin-2 which bind to its pro- We recently showed that the BCH domain on Cdc42GAP could line-rich region (PRR). Cortactin helps form branching actin serves as a local modulator to sequester RhoA from being inacti- network [36] whereas endophilin-2 promotes EGF receptor endo- vated by its adjacent GAP domain ([19]; Fig. 4E). This BCH domain cytosis for enhanced ERK signaling [37]. This process is also aug- cannot distinguish between the GDP-bound and GTP-bound form mented by the activity of the RhoGAP domain that inactivates of RhoA. Further mutagenesis revealed a novel RhoA-binding motif Rho. Furthermore, we showed that active Mek2 stimulates the re- (residues 85–120) within the BCH domain (Fig. 3A and D). Deletion lease of an auto inhibition on the PRR, thus promoting binding of of this motif significantly reduced BCH inhibition on GAP-mediated the peptidyl–prolyl isomerase Pin1 to the PRR and the RhoGAP do- cell rounding in human cervical epithelial cells, whereas its full main to suppress BPGAP1-induced acute Erk activation and cell suppression also required an intramolecular interaction motif (res- motility (Figs. 4F and 5D; [38]). Interestingly, such mechanism is idues 169–197). also employed by the BCH domain of BNIP-H where Mek2 pro- Activation of RhoA can lead to distinct cellular and physiological motes Pin1 binding and disruption of the binding of its cargo, glu- outcomes, depending on the cell types and the environment they taminase KGA [55]. It remains to be investigated whether this are in, the types of processes (e.g. cell spreading, cell adhesion ver- mechanism would affect the versatility of the BCH domain in sus cell motility) and the involvement of different constituent pro- BPGAP1. teins such as upstream regulators or downstream effectors that We had earlier reported on the role of BCH domain in stimulat- propagate the signals at different locales and timing. Cdc42GAP ing ERK signaling, through a yet unknown mechanism [37]. Using expression (hence inactivation of Rho) would lead to cell protru- proteomics pulldown and candidate approaches, we have since sion in epithelial breast cancer MCF7 cells. When BNIP-S is overex- identified several novel partners, including Ras and one of its reg- pressed, this protein induces cell rounding and apoptosis by ulators, SmgGDS. We show that BCH domain binds to K-Ras and displacing Cdc42GAP from inactivating RhoA. Separate pools of that leads to enhanced Ras/ERK signaling necessary for the neurite BNIP-S would then bind RhoA and that lead to RhoA activation pos- outgrowth and differentiation. Knockdown of SmgGDS potentiates sibly by engaging a RhoGEF. It is likely that, while preventing RhoA K-Ras/BCH-mediated Erk activation further, revealing a novel tri- from being inactivated by the GAP domain of Cdc42GAP, formation partite regulation of Ras/Erk by the BCH domain (unpublished of BNIP-S/RhoA complex could trigger additional signaling path- data). And through its heterophilic BCH/BCH interaction of BNIP- way(s) that couples the activation of Rho to apoptosis. In contrast, 2, we are examining how BNIP-2 and BPGAP1 would co-regulate when Cdc42GAP is expressed in cervical epithelial HeLa cells, this Rho, Ras and Cdc42 signaling. protein posseses minimal GAP activity towards RhoA owing to sequestration of RhoA by the adjacent BCH domain in cis [19], hence RhoA is not inactivated and HeLa cells remain cuboidal. 3. Conclusions and future perspective However, when this sequestration is lost (evidenced by deleting the RhoA-binding motif in the BCH domain), the GAP will inacti- Protein domains evolve to change their biochemical and bio- vate Rho. Under this condition, inactivation of Rho leads to collapse physical properties to confer different roles in recognizing either of stress fibers and cell adhesion, and cells become rounded up in- specific or diverse targets and to function at unique or multiple lo- stead of rendering protrusion. These observations therefore high- cales. They usually do not work in isolation but they interact with light the plasticity and contextual signaling of Rho on cell other molecule(s) that carry the same or with distinct protein morphogenesis under the influence of BCH domains. modules or other non-peptide ligands – all defined by their unique Recent studies also showed that the BCH domain undergoes sequence motifs. Many domains also co-exist by various copy autoinhibition in vitro [88] and is required for its endosomal local- numbers and permutation in one particular protein. Such designs ization and binding to Rab5 and Rab11 [89]. Interestingly, the other help thrive for their optimal signaling capabilities, forming more plant RhoGAPs commonly referred to as RopGAPs (Rop, Rho of sophisticated yet well coordinated network of signaling hubs. plants) are associated with a Cdc42/Rac interactive binding (CRIB) The properties and functions of these domains can also be modified motif at their N-terminus [90]. This CRIB motif can help regulate by biochemical and mechanical signals which cells receive from the local GAP activity by forming high affinity complexes with spe- the external environments. Therefore, optimal design and engi- cific Rho proteins and GAP domains and acts as lid for binding and neering of domain architecture equipped with single or multiple releasing Rho of plants [91]. Such a sequestration of substrate in cis motifs allow them to execute as well as receiving feedback for bet- by the tandem Rho-binding domains therefore provides a novel ter integration of cell signaling. 2688 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

In this review, we have highlighted how BCH domains have ac- domains of GEFs, GAPs and GDIs? Do other BCH-like quired many of these key features that enable them function as domains also possess some of the binding properties such versatile regulatory scaffold modules. Through their multiple func- as GTPase binding? This is of special interest, as the CRAL– tional motifs, some of which obvious but some are cryptic, they TRIO domains present in the RasGAP (e.g. NF1) and RhoGEF interact with a broad spectrum of targets, ranging from membrane (e.g. Trio, DBS) could indeed engage some of these GTPases receptors, different GTPases and their immediate regulators such to regulate cell morphogenesis, migration, apoptosis and dif- as GEFs and GAPs, kinase such as Mek2, isomerase Pin1, anti-apop- ferentiation. Identification and systematic studies of other tosis protein Bcl2 and key metabolic enzymes such as glutaminase, BCH-domain-containing proteins will also help to provide impacting ultimately on cell growth, apoptosis, morphogenesis, further insights into the mechanistic functions of this migration, metabolism and differentiation. While all these obser- emerging class of protein module. vations point to the functional plasticity of BCH domains and fur- (2) How are signaling pathways connected by their binding to ther underscore the importance of BNIP-2 family and BPGAP BCH domains? Are these pathways merely operating in par- family proteins in cell dynamics and cell fates control, the precise allel by different pools of BCH domain proteins or are they mechanisms underlying these dynamic processes and their physi- physiologically or physically linked? To date, we have ological impacts remain to be further elucidated. Here, we reflect observed the significant impact of BCH domains either on several interconnected issues related to the structural, func- directly or indirectly via other domains they are associated tional, mechanical, evolutionary, physiological aspects of BCH do- with, on various signaling nodes, especially in the FGF and mains and the cellular environments these proteins are exposed Cdo receptor signaling, various Rho, Ras, Rab GTPases, some to, and aim to understand them at the molecular, cellular and tis- kinase, isomerase, caspases, Bcl2 and even the metabolic sue levels (Fig. 6). Some of these are summarized below: pathway. Under what conditions are they functionally linked? And given that some BCH domains are already tar- (1) What are the structural bases underlying the plasticity ver- gets of kinases, ligase and isomerase, what are the signals sus specificity of BCH domains in recognizing multiple and that modify their properties through phosphorylation, ubiq- diverse targets? In particular, within the same class of part- uitination and isomerization and what are the impacts on ner, such as Rho, how do they distinguish between the active their functions? G14V which they do not bind versus inactive mutant T19N (3) What are the dynamic features of BCH domains that help which they bind best, despite that the same BCH domain them achieve their multifunctional roles? And how are these binds both GDP-bound and GTP-bound form of Rho equally functions regulated in space and time at the cellular to tis- well? What are the atomic and molecular bases that distin- sue/organismal levels. For examples, BNIP-H and BNIP-2 guish BCH domain from other conventional GTPase-binding are actively transported along the microtubules during neu-

Unique GTPases binding Signaling crosstalk Rho, Cdc42, Ras, Rab, active vs Rho, Ras, RTK, Mek, MAPK inactive conformers; binding by (p38/ERK), caspases Structures other BCH or BCH-like domains and metabolic pathways complex with targets Protein modification homophilic or heterophilic phosphorylation by RTK, (BCH-BCH) complex Mek, other kinases ubiquitination isomerization

Spatiotemporal Cytoskeletal dynamic trafficking network with cargo and actin and microtubule signaling proteins dynamics linked to morphogenesis and migration

Lipidomics Tissue / Organogenesis lipids and phosphoinositides zebrafish, mice , rats targeting or regulatory as disease models

Gene regulation Splicing variants, redundancy, transcription tissue specificity microRNA, translation physiological and hormonal regulation

Fig. 6. Future perspective: elucidating the functional plasticity of BCH scaffold protein domains across molecular, cellular and physiological levels. To further understand the molecular and cellular bases for the physiological functions and regulation by the BCH domains, future works will aim at (i) solving the structures of BCH domains with their target proteins including those that confer unique GTPase-binding profile (e.g. Rho, Cdc42, Ras, Rab) and their dependence on specific conformation (e.g. dominant negative mutant T17N is most preferred but constitutive G12V is the least or no binding) and the BCH/BCH interactions; (ii) exploring their diverse and dynamic signaling crosstalk between cell growth, proliferation, differentiation regimes (e.g. receptor tyrosine kinases, GTPases, p38 and Erk) and key metabolic pathways (e.g. glutaminase) under the influence of biochemical or/and mechanical cues such as substrate rigidity; (iii) examining their roles on actin and microtubule dynamics, leading to cell morphogenesis and cell migration; (iv) deciphering their post-translational protein modifications (via phosphorylation, ubiquitination and isomerization) and intracellular trafficking as well as identification of cargo proteins such as metabolic enzymes and signaling complexes; (v) examining their possible binding and regulation by lipids and phosphoinositides; (vi) exploring the significance of alternative RNA splicing variants in physiological and hormonal regulation; (vii) identifying their modes of gene regulation and expression by transcription, micro-RNA and translation, and (viii) elucidating their distinct roles in tissue/organogenesis using various vertebrate models. All these issues will be systematically examined and their precise functions/steps elucidated. For clarity, the model presented here depicts the complex between BNIP-2 and Cdc42GAP via their BCH domains (orientation is hypothetical) while each could still act as distinctive scaffold module to regulate binding and activity of 3G-Signalome, i.e. GTPases, GEFs (+ sign) and GAPs (À sign). C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2689

ronal maturation and differentiation. Are they trafficking controlling normal physiology but they also offer us exciting pros- and delivering any cargos or are they being recruited to pects of exploiting them as potential targets of therapeutic inter- the neurites tips before exerting their scaffold functions vention. Furthermore, it is plausible now to think about creating there? What are those cargos? Are they metabolic enzymes novel cellular processes and functions by re-wiring specific net- or signaling complexes? Moreover, with their ability to rec- works or modifying their responses using synthetic scaffold pro- ognize multiple GTPases on the same domains, e.g. via their teins. It is with these motivations that we should continue to tandem RBD-like and CRIB-like motifs present in BNIP-2 BCH define the form (structure), the meaning (interactomes) and the domain, it is tempted to speculate that this BCH domain can context (when, where and regulation) of such linguistic designs also regulate Rho in addition to Cdc42 signaling. If so, what of BCH domains that their functional plasticity and biological sig- are the consequences on the spatial and temporal activities nificance can be fully appreciated. of these GTPases? (4) As cells sense both biochemical and mechanical cues under Acknowledgments various conditions, it will be important to address whether function(s) of BCH domains, that is usually linked to receptor We thank all the members and collaborators of BCL Lab, past and Rho signaling, is itself subjected to both the biochemical and present, and also colleagues who are working on this subject and mechanical triggers. And, how those signals influence for their dedication. This work is supported in part by the Biomed- the dynamic distribution of BCH domain proteins and their ical Research Council of Singapore (07/1/21/19/506), by the ability to regulate the actin and microtubule network associ- Ministry of Education (Tier-2 Grant T208A3121) and by the ated with extensive morphogenesis and motility. Mechanobiology Institute, co-funded by National Research Foun- (5) Are BCH domains certified ‘‘lipid-free?’’ Despite relatively dation and the Ministry of Education of Singapore. low sequence identity between the BCH domain and CRAL–TRIO and other lipids binding domains, several studies References including ours have not excluded entirely the ability of BCH domain to recognize certain lipids for their function or reg- [1] Pawson, T. and Nash, P. (2003) Assembly of cell regulatory systems through protein interaction domains. Science 300, 445–452. ulation. It is still plausible that certain species of lipids [2] Li, N. et al. (1993) Guanine-nucleotide-releasing factor hSos1 binds to Grb2 including the phosphoinositides, could act as ligands to help and links receptor tyrosine kinases to Ras signalling. Nature 363, 85–88. regulate functions of the BCH domains instead of serving as [3] Sudol, M. (2010) Newcomers to the WW Domain-Mediated Network of the Hippo Tumor Suppressor Pathway. Genes Cancer 1, 1115–1118. the substrates for transportation. However, the authenticity [4] Achiriloaie, M., Barylko, B. and Albanesi, J.P. (1999) Essential role of the of their binding would require further confirmation in vivo dynamin pleckstrin homology domain in receptor-mediated endocytosis. Mol. or cell-based system. Cell. Biol. 19, 1410–1415. (6) As BCH domains confer functional diversity, the proteins and [5] Kenniston, J.A. and Lemmon, M.A. (2010) Dynamin GTPase regulation is altered by PH domain mutations found in centronuclear myopathy patients. the genes from which they are encoded are expected to be EMBO J. 29, 3054–3067. under tight physiological regulation. This could take in the [6] Brown, M.D. and Sacks, D.B. (2009) Protein scaffolds in MAP kinase signalling. forms of gene transcription, protein expression and stability, Cell. Signal. 21, 462–469. [7] Frame, M.C., Patel, H., Serrels, B., Lietha, D. and Eck, M.J. (2010) The FERM and of course their ability to undergo alternative RNA splic- domain: organizing the structure and function of FAK. Nat. Rev. Mol. Cell Biol. ing. The unique pattern of RNA splicing, as we frequently 11, 802–814. observe in BNIP-2, BNIP-S, BNIP-XL and BPGAP1 calls for [8] Lietha, D., Cai, X., Ceccarelli, D.F., Li, Y., Schaller, M.D. and Eck, M.J. (2007) Structural basis for the autoinhibition of focal adhesion kinase. Cell 129, 1177– attention on their changes in the splicing pattern in response 1187. to hormonal and stress signals. Similarly, their regulation by [9] Low, B.C., Lim, Y.P., Lim, J., Wong, E.S. and Guy, G.R. (1999) Tyrosine microRNA and tissue specificity factors should also be exam- phosphorylation of the Bcl-2-associated protein BNIP-2 by fibroblast growth factor receptor-1 prevents its binding to Cdc42GAP and Cdc42. J. Biol. Chem. ined in the right context of cell signaling; 274, 33123–33130. (7) Last but not least, as functionalities and plasticity of BCH [10] Boyd (1994) Adenovirus E1B 19 kDa and Bcl-2 proteins interact with a domain containing proteins are being examined, we should common set of cellular proteins. Cell 79, 1121. [11] Low, B.C., Seow, K.T. and Guy, G.R. (2000) The BNIP-2 and Cdc42GAP homology consider developing various vertebrate models to mimic domain of BNIP-2 mediates its homophilic association and heterophilic and interrogate the actual physiological response associated interaction with Cdc42GAP. J. Biol. Chem. 275, 37742–37751. with the signaling pathways. These model systems of mice [12] Low, B.C., Seow, K.T. and Guy, G.R. (2000) Evidence for a novel Cdc42GAP to zebrafish allow us establish necessary platforms to detect domain at the carboxyl terminus of BNIP-2. J. Biol. Chem. 275, 14415–14422. [13] Wennerberg, K., Rossman, K.L. and Der, C.J. (2005) The Ras superfamily at a their impacts on developmental processes as well as for drug glance. J. Cell Sci. 118, 843–846. screening. Coupled with genetic manipulations of knock- [14] Schmidt, A. and Hall, A. (2002) Guanine nucleotide exchange factors for Rho down and functional rescues with pathway/partner-specific GTPases: turning on the switch. Genes Dev. 16, 1587–1609. [15] Bernards, A. and Settleman, J. (2004) GAP control: regulating the regulators of mutants in the cell culture and knockout/knockdown or small GTPases. Trends Cell Biol. 14, 377–385. knock-in for the in vivo models system, can we clearly deci- [16] Hoffman, G.R. and Cerione, R.A. (2000) Flipping the switch: the structural basis pher some of the true meaning behind the multiplexing for signaling through the CRIB motif. Cell 102, 403–406. [17] Bishop, A.L. and Hall, A. (2000) Rho GTPases and their effector proteins. roles of BCH domain containing proteins. Biochem. J. 348 (Pt. 2), 241–255. [18] Soh, U.J. and Low, B.C. (2008) BNIP2 extra long inhibits RhoA and cellular With the advent of high- and super-resolution live microscopy transformation by Lbc RhoGEF via its BCH domain. J. Cell Sci. 121, 1739–1749. [19] Zhou, Y.T., Chew, L.L., Lin, S.C. and Low, B.C. (2010) The BNIP-2 and Cdc42GAP and the availability of relevant biosensors for signaling proteins, homology (BCH) domain of p50RhoGAP/Cdc42GAP sequesters RhoA from one could trace the spatiotemporal dynamics of BCH domain con- inactivation by the adjacent GTPase-activating protein domain. Mol. Biol. Cell taining proteins or the BCH domains alone and their physical local- 21, 3232–3246. [20] Bourne, H.R., Sanders, D.A. and McCormick, F. (1991) The GTPase superfamily: ization with their partners under different cellular contexts and conserved structure and molecular mechanism. Nature 349, 117–127. genetic backgrounds. And with various biophysical and structural [21] Bankaitis, V.A., Mousley, C.J. and Schaaf, G. (2010) The Sec14 superfamily and determination tools such as NMR to map conformational shifts mechanisms for crosstalk between lipid metabolism and lipid signaling. and X-ray crystallography for detailed binding sites, one can start Trends Biochem. Sci. 35, 150–160. [22] Gu, M.X., York, J.D., Warshawsky, I. and Majerus, P.W. (1991) Identification, generating more precise mutants to interrogate their functional cloning, and expression of a cytosolic megakaryocyte protein-tyrosine- plasticity and dynamics distribution in the cells and in tissues. phosphatase with sequence homology to cytoskeletal protein 4.1. Proc. Natl. Such versatile properties not only underlie their central roles in Acad. Sci. 88, 5867–5871. 2690 C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691

[23] Min, K.C., Kovall, R.A. and Hendrickson, W.A. (2003) Crystal structure of [49] Hayakawa, Y., Itoh, M., Yamada, A., Mitsuda, T. and Nakagawa, T. (2007) human alpha-tocopherol transfer protein bound to its ligand: implications for Expression and localization of Cayman ataxia-related protein, Caytaxin, is ataxia with vitamin E deficiency. Proc. Natl. Acad. Sci. U.S.A. 100, 14713– regulated in a developmental- and spatial-dependent manner. Brain Res. 14718. 1129, 100–109. [24] Aravind, L., Neuwald, A.F. and Ponting, C.P. (1999) Sec14p-like domains in NF1 [50] Xiao, J., Gong, S. and Ledoux, M.S. (2007) Caytaxin deficiency disrupts and Dbl-like proteins indicate lipid regulation of Ras and Rho signaling. Curr. signaling pathways in cerebellar cortex. Neuroscience 144, 439–461. Biol. 9, R195–R197. [51] Pizoli, C.E., Jinnah, H.A., Billingsley, M.L. and Hess, E.J. (2002) Abnormal [25] Panagabko, C., Morley, S., Hernandez, M., Cassolato, P., Gordon, H., Parsons, R., cerebellar signaling induces dystonia in mice. J. Neurosci. 22, 7825–7833. Manor, D. and Atkinson, J. (2003) Ligand specificity in the CRAL–TRIO protein [52] Choi, D.W., Maulucci-Gedde, M. and Kriegstein, A.R. (1987) Glutamate family. Biochemistry 42, 6467–6474. neurotoxicity in cortical cell culture. J. Neurosci. 7, 357–368. [26] Gupta, A.B., Wee, L.E., Zhou, Y.T., Hortsch, M. and Low, B.C. (2012) Cross- [53] Rothstein, J.D. et al. (1996) Knockout of glutamate transporters reveals a major species analyses identify the BNIP-2 and Cdc42GAP homology (BCH) domain role for astroglial transport in excitotoxicity and clearance of glutamate. as a distinct functional subclass of the CRAL_TRIO/Sec14 superfamily. PLoS Neuron 16, 675–686. One 7, e33863. [54] Choudary, P.V. et al. (2005) Altered cortical glutamatergic and GABAergic [27] Feng, D.F., Cho, G. and Doolittle, R.F. (1997) Determining divergence times signal transmission with glial involvement in depression. Proc. Natl. Acad. Sci. with a protein clock: update and reevaluation. Proc. Natl. Acad. Sci. U.S.A. 94, U.S.A. 102, 15653–15658. 13028–13033. [55] Buschdorf, J.P., Chew, L.L., Soh, U.J., Liou, Y.C. and Low, B.C. (2008) Nerve [28] Zhou, Y.T., Soh, U.J., Shang, X., Guy, G.R. and Low, B.C. (2002) The BNIP-2 and growth factor stimulates interaction of Cayman ataxia protein BNIP-H/ Cdc42GAP homology/Sec14p-like domain of BNIP-Salpha is a novel apoptosis- Caytaxin with peptidyl-prolyl isomerase Pin1 in differentiating neurons. inducing sequence. J. Biol. Chem. 277, 7483–7492. PLoS One 3, e2686. [29] Shang, X., Zhou, Y.T. and Low, B.C. (2003) Concerted regulation of cell [56] Grelle, G. et al. (2006) Identification of VCP/p97, carboxyl terminus of Hsp70- dynamics by BNIP-2 and Cdc42GAP homology/Sec14p-like, proline-rich, and interacting protein (CHIP), and amphiphysin II interaction partners using GTPase-activating protein domains of a novel Rho GTPase-activating protein, membrane-based human proteome arrays. Mol. Cell. Proteomics 5, 234–244. BPGAP1. J. Biol. Chem. 278, 45903–45914. [57] Itoh, M. et al. (2011) Cayman ataxia-related protein is a presynapse-specific [30] Buschdorf, J.P., Li Chew, L., Zhang, B., Cao, Q., Liang, F.Y., Liou, Y.C., Zhou, Y.T. caspase-3 substrate. Neurochem. Res. 36, 1304–1313. and Low, B.C. (2006) Brain-specific BNIP-2-homology protein Caytaxin [58] Kudryashova, I.V., Onufriev, M.V., Kudryashov, I.E. and Gulyaeva, N.V. (2009) relocalises glutaminase to neurite terminals and reduces glutamate levels. J. Caspase-3 activity in hippocampal slices reflects changes in synaptic Cell Sci. 119, 3337–3350. plasticity. Neurosci. Behav. Physiol. 39, 13–20. [31] Bomar, J.M. et al. (2003) Mutations in a novel gene encoding a CRAL–TRIO [59] Bravarenko, N.I., Onufriev, M.V., Stepanichev, M.Y., Ierusalimsky, V.N., Balaban, domain cause human Cayman ataxia and ataxia/dystonia in the jittery mouse. P.M. and Gulyaeva, N.V. (2006) Caspase-like activity is essential for long-term Nat. Genet. 35, 264–269. synaptic plasticity in the terrestrial snail Helix. Eur. J. Neurosci. 23, 129–140. [32] Zhou, Y.T., Guy, G.R. and Low, B.C. (2005) BNIP-2 induces cell elongation and [60] Huesmann, G.R. and Clayton, D.F. (2006) Dynamic role of postsynaptic membrane protrusions by interacting with Cdc42 via a unique Cdc42-binding caspase-3 and BIRC4 in zebra finch song-response habituation. Neuron 52, motif within its BNIP-2 and Cdc42GAP homology domain. Exp. Cell Res. 303, 1061–1072. 263–274. [61] Fernando, P., Brunette, S. and Megeney, L.A. (2005) Neural stem cell [33] Zhou, Y.T., Guy, G.R. and Low, B.C. (2006) BNIP-Salpha induces cell rounding differentiation is dependent upon endogenous caspase 3 activity. FASEB J. and apoptosis by displacing p50RhoGAP and facilitating RhoA activation via 19, 1671–1673. its unique motifs in the BNIP-2 and Cdc42GAP homology domain. Oncogene [62] Williams, D.W., Kondo, S., Krzyzanowska, A., Hiromi, Y. and Truman, J.W. 25, 2393–2408. (2006) Local caspase activity directs engulfment of dendrites during pruning. [34] Kang, J.S. et al. (2008) A Cdo-Bnip-2-Cdc42 signaling pathway regulates Nat. Neurosci. 9, 1234–1236. p38alpha/beta MAPK activity and myogenic differentiation. J. Cell Biol. 182, [63] Nikolaev, A., McLaughlin, T., O’Leary, D.D. and Tessier-Lavigne, M. (2009) APP 497–507. binds DR6 to trigger axon pruning and neuron death via distinct caspases. [35] Aoyama, T., Hata, S., Nakao, T., Tanigawa, Y., Oka, C. and Kawaichi, M. (2009) Nature 457, 981–989. Cayman ataxia protein caytaxin is transported by kinesin along neurites [64] Geisbrecht, E.R. and Montell, D.J. (2004) A role for Drosophila IAP1-mediated through binding to kinesin light chains. J. Cell Sci. 122, 4177–4185. caspase inhibition in Rac-dependent cell migration. Cell 118, 111–125. [36] Lua, B.L. and Low, B.C. (2004) BPGAP1 interacts with cortactin and facilitates [65] Li, J. et al. (2007) Caspase-11 regulates cell migration by promoting Aip1- its translocation to cell periphery for enhanced cell migration. Mol. Biol. Cell Cofilin-mediated actin depolymerization. Nat. Cell Biol. 9, 276–286. 15, 2873–2883. [66] Machida, T. et al. (2006) Increased expression of proapoptotic BMCC1, a novel [37] Lua, B.L. and Low, B.C. (2005) Activation of EGF receptor endocytosis and gene with the BNIP2 and Cdc42GAP homology (BCH) domain, is associated ERK1/2 signaling by BPGAP1 requires direct interaction with EEN/endophilin with favorable prognosis in human neuroblastomas. Oncogene 25, 1931– II and a functional RhoGAP domain. J. Cell Sci. 118, 2707–2721. 1942. [38] Pan, C.Q., Liou, Y.C. and Low, B.C. (2010) Active Mek2 as a regulatory scaffold [67] Clarke, R.A. et al. (2009) New genomic structure for prostate cancer specific that promotes Pin1 binding to BPGAP1 to suppress BPGAP1-induced acute Erk gene PCA3 within BMCC1: implications for prostate cancer detection and activation and cell migration. J. Cell Sci. 123, 903–916. progression. PLoS One 4, e4995. [39] Lu, M. and Krauss, R.S. (2010) N-cadherin ligation, but not Sonic hedgehog [68] Li, S. et al. (2011) The expression and localization of Prune2 mRNA in the binding, initiates Cdo-dependent p38alpha/beta MAPK signaling in skeletal central nervous system. Neurosci. Lett. 503, 208–214. myoblasts. Proc. Natl. Acad. Sci. U.S.A. 107, 4212–4217. [69] Iwama, E., Tsuchimoto, D., Iyama, T., Sakumi, K., Nakagawara, A., Takayama, K., [40] Oh, J.E., Bae, G.U., Yang, Y.J., Yi, M.J., Lee, H.J., Kim, B.G., Krauss, R.S. and Kang, Nakanishi, Y. and Nakabeppu, Y. (2011) Cancer-related PRUNE2 protein is J.S. (2009) Cdo promotes neuronal differentiation via activation of the p38 associated with nucleotides and is highly expressed in mature nerve tissues. J. mitogen-activated protein kinase pathway. FASEB J. 23, 2088–2099. Mol. Neurosci. 44, 103–114. [41] Valencia, C.A., Cotten, S.W. and Liu, R. (2007) Cleavage of BNIP-2 and BNIP-XL [70] Salagierski, M., Verhaegh, G.W., Jannink, S.A., Smit, F.P., Hessels, D. and by caspases. Biochem. Biophys. Res. Commun. 364, 495–501. Schalken, J.A. (2010) Differential expression of PCA3 and its overlapping [42] Scott, G.B., Bowles, P.A., Wilson, E.B., Meade, J.L., Low, B.C., Davison, A., Blair, PRUNE2 transcript in prostate cancer. Prostate 70, 70–78. G.E. and Cook, G.P. (2010) Identification of the BCL2/adenovirus E1B– [71] Potkin, S.G. et al. (2009) Hippocampal atrophy as a quantitative trait in a 19K protein-interacting protein 2 (BNIP-2) as a granzyme B target genome-wide association study identifying novel susceptibility genes for during human natural killer cell-mediated killing. Biochem. J. 431, 423– Alzheimer’s disease. PLoS One 4, e6501. 431. [72] Price, N.D. et al. (2007) Highly accurate two-gene classifier for differentiating [43] Belcredito, S., Vegeto, E., Brusadelli, A., Ghisletti, S., Mussi, P., Ciana, P. and gastrointestinal stromal tumors and leiomyosarcomas. Proc. Natl. Acad. Sci. Maggi, A. (2001) Estrogen neuroprotection: the involvement of the Bcl-2 U.S.A. 104, 3414–3419. binding protein BNIP2. Brain Res. Brain Res. Rev. 37, 335–342. [73] Shen, L., Hu, J., Lu, H., Wu, M., Qin, W., Wan, D., Li, Y.Y. and Gu, J. (2003) The [44] Sall, A. et al. (2010) Pro-apoptotic activity of mBNIP-21 depends on its BNIP-2 apoptosis-associated protein BNIPL interacts with two cell proliferation- and Cdc42GAP homology (BCH) domain and is enhanced by coxsackievirus B3 related proteins, MIF and GFER. FEBS Lett. 540, 86–90. infection. Cell. Microbiol. 12, 599–614. [74] Qin, W. et al. (2003) BNIPL-2, a novel homologue of BNIP-2, interacts with Bcl- [45] Bonner, A.E., Lemon, W.J., Devereux, T.R., Lubet, R.A. and You, M. (2004) 2 and Cdc42GAP in apoptosis. Biochem. Biophys. Res. Commun. 308, 379–385. Molecular profiling of mouse lung tumors: association with tumor [75] Xie, L. et al. (2007) BNIPL-2 promotes the invasion and metastasis of human progression, lung development, and human lung adenocarcinomas. hepatocellular carcinoma cells. Oncol. Rep. 17, 605–610. Oncogene 23, 1166–1176. [76] Tcherkezian, J. and Lamarche-Vane, N. (2007) Current knowledge of the large [46] Gilbert, N., Bomar, J.M., Burmeister, M. and Moran, J.V. (2004) Characterization RhoGAP family of proteins. Biol. Cell 99, 67–86. of a mutagenic B1 retrotransposon insertion in the jittery mouse. Hum. Mutat. [77] Lancaster, C.A., Taylor-Harris, P.M., Self, A.J., Brill, S., van Erp, H.E. and Hall, A. 24, 9–13. (1994) Characterization of rhoGAP. A GTPase-activating protein for rho- [47] Kapfhamer, D., Sweet, H.O., Sufalko, D., Warren, S., Johnson, K.R. and related small GTPases. J. Biol. Chem. 269, 1137–1142. Burmeister, M. (1996) The neurological mouse mutations jittery and [78] Johnstone, C.N. et al. (2004) ARHGAP8 is a novel member of the RHOGAP hesitant are allelic and map to the region of mouse chromosome 10 family related to ARHGAP1/CDC42GAP/p50RHOGAP: mutation and expression homologous to 19p13.3. Genomics 35, 533–538. analyses in colorectal and breast cancers. Gene 336, 59–71. [48] Xiao, J. and Ledoux, M.S. (2005) Caytaxin deficiency causes generalized [79] Hatjiharissi, E. et al. (2007) Proteomic analysis of waldenstrom dystonia in rats. Brain Res. Mol. Brain Res. 141, 181–192. macroglobulinemia. Cancer Res. 67, 3777–3784. C.Q. Pan, B.C. Low / FEBS Letters 586 (2012) 2674–2691 2691

[80] Song, J.Y., Lee, J.K., Lee, N.W., Jung, H.H., Kim, S.H. and Lee, K.W. (2008) [86] Rittinger, K., Walker, P.A., Eccleston, J.F., Nurmahomed, K., Owen, D., Laue, E., Microarray analysis of normal cervix, carcinoma in situ, and invasive cervical Gamblin, S.J. and Smerdon, S.J. (1997) Crystal structure of a small G protein in cancer: identification of candidate genes in pathogenesis of invasion in complex with the GTPase-activating protein rhoGAP. Nature 388, 693–697. cervical cancer. Int. J. Gynecol. Cancer 18, 1051–1059. [87] Barrett, T. et al. (1997) The structure of the GTPase-activating domain from [81] Wang, L., Yang, L., Filippi, M.D., Williams, D.A. and Zheng, Y. (2006) Genetic p50rhoGAP. Nature 385, 458–461. deletion of Cdc42GAP reveals a role of Cdc42 in erythropoiesis and [88] Moskwa, P., Paclet, M.H., Dagher, M.C. and Ligeti, E. (2005) Autoinhibition of hematopoietic stem/progenitor cell survival, adhesion, and engraftment. p50 Rho GTPase-activating protein (GAP) is released by prenylated small Blood 107, 98–105. GTPases. J. Biol. Chem. 280, 6716–6720. [82] Szczur, K., Xu, H., Atkinson, S., Zheng, Y. and Filippi, M.D. (2006) Rho GTPase [89] Sirokmany, G., Szidonya, L., Kaldi, K., Gaborik, Z., Ligeti, E. and Geiszt, M. CDC42 regulates directionality and random movement via distinct MAPK (2006) Sec14 homology domain targets p50RhoGAP to endosomes and pathways in neutrophils. Blood 108, 4205–4213. provides a link between Rab and Rho GTPases. J. Biol. Chem. 281, 6096–6105. [83] Wang, L., Yang, L., Burns, K., Kuan, C.Y. and Zheng, Y. (2005) Cdc42GAP [90] Wu, G., Gu, Y., Li, S. and Yang, Z. (2001) A genome-wide analysis of Arabidopsis regulates c-Jun N-terminal kinase (JNK)-mediated apoptosis and cell number Rop-interactive CRIB motif–containing proteins that act as Rop GTPase targets. during mammalian perinatal growth. Proc. Natl. Acad. Sci. U.S.A. 102, 13484– Plant Cell Online 13, 2841–2856. 13489. [91] Schaefer, A., Höhner, K., Berken, A. and Wittinghofer, A. (2011) The unique [84] Wang, L., Yang, L., Debidda, M., Witte, D. and Zheng, Y. (2007) Cdc42 plant RhoGAPs are dimeric and contain a CRIB motif required for affinity and GTPase-activating protein deficiency promotes genomic instability and specificity towards cognate small G proteins. Biopolymers 95, 420–433. premature aging-like phenotypes. Proc. Natl. Acad. Sci. U.S.A. 104, 1248– [92] Shen, Y. et al. (2008) Nudel binds Cdc42GAP to modulate Cdc42 activity at the 1253. leading edge of migrating cells. Dev. Cell 14, 342–353. [85] Barfod, E.T., Zheng, Y., Kuang, W.J., Hart, M.J., Evans, T., Cerione, R.A. and [93] Arama, J., Boulay, A.C., Bosc, C., Delphin, C., Loew, D., Rostaing, P., Amigou, E., Ashkenazi, A. (1993) Cloning and expression of a human CDC42 GTPase- Ezan, P., Wingertsmann, L., Guillaud, L., Andrieux, A., Giaume, C. and Cohen- activating protein reveals a functional SH3-binding domain. J. Biol. Chem. 268, Salmon, M. (2012) Bmcc1s, a Novel Brain-Isoform of Bmcc1, Affects Cell 26059–26062. Morphology by Regulating MAP6/STOP Functions. PLoS One 7, e35488.