The SUMO E3 ligase activity of RANBP2 and the RANBP2/RANGAP1*SUMO1/Ubc9 (RRSU) complex
In addition to binding RAN, RANBP2 stably associates with the RAN GTPase-activating protein 1 (RANGAP1) (Mahajan et al., 1997; Saitoh et al., 1997; Matunis et al., 1998; Swaminathan et al., 2004) throughout the cell cycle. The interaction requires SUMO-1 conjugation to RANGAP1 (Matunis et al., 1996; Mahajan et al., 1997) and the presence of the SUMO ubiquitin-like-conjugating Ubc9 enzyme (Zhang et al., 2002; Zhu et al., 2006), an E2 enzyme that transfers SUMO peptides to SUMO chains, analogous to enzymes acting in the ubiquitination cascade.
RANBP2 is possibly the most abundant SUMO E3 ligase in the cell, and as such has a prominent role in SUMO modification of proteins. Most RANBP2-dependent functions are likely mediated by its activity in SUMO conjugation of target proteins.
The RANBP2 SUMO E3 ligase domain lacks homology to other known SUMO or ubiquitin E3 ligases (Pichler et al., 2004). It is characterized by two 50 aminoacid-long internal repeats, IR1 and IR2 (43% identical), separated by a 20 aminoacid-long linker. Both IR1 and IR2 can bind to Ubc9 and catalyze sumoylation of substrates in vitro, though IR2 has lower affinity for Ubc9 than IR1 (Pichler et al., 2004; Tatham et al., 2005).
RANBP2 is quantitatively engaged in complexes with sumoylated RANGAP1 and Ubc9. This binding requires RANBP2's IR1 and the SUMO-interaction motif SIM1, as well as Ubc9. Once IR1 and SIM1 are occupied by RanGAP1*SUMO1 and Ubc9, the E3 ligase activity depends on IR2 (Werner et al., 2012). In that sense, the entire RANBP2/RANGAP1-SUMO complex (called RRSU complex) is viewed as a multisubunit SUMO ligase. Indeed, RRSU effectively sumoylates the physiological substrate Borealin in vitro (Klein et al., 2009), whereas free RANBP2 does not (Werner et al., 2012).
After NPC disassembly at NEB, RANBP2 remains associated with RANGAP1-SUMO1 and Ubc9 (Swaminathan et al., 2004); RRSU associates with the mitotic spindle and a fraction is recruited to KTs after MTs attach to them (Joseph et al., 2002; Joseph et al., 2004).

Structural functions at the nuclear rim and NPCs
RANBP2 has structural roles at the NE besides nuclear transport proper. The development of in situ SUMOylation assays has revealed that both the nuclear rim and PML nuclear bodies are major sites of SUMOylation; RANBP2 inactivation abolished SUMOylation processes along the nuclear rim and reduced the number of PML bodies, while not affecting the nuclear lamina (Saitoh et al., 2006). The loss of nuclear PML bodies has been observed in tumorigenesis, particularly in colon cancer. Satow et al. (2012) reported that β-catenin overexpression disrupts PML bodies in colon cancer cell lines and inhibits RANBP2-dependent SUMOylation of specific PML-associated proteins. The data suggest that RANBP2 is required for SUMOylation of proteins associated with the formation of particular subnuclear structures, the loss of which impinges on nuclear functions in cancer cells.
RANBP2 also recruits motor proteins at the NE to regulate NE breakdown at the onset of mitosis. Through its zinc finger domain, RANBP2 binds the COPI coatomer complex, which coats the Golgi vesicles and contributes to membrane remodelling at the Golgi; the RANBP2-derived zinc finger domain alone dominantly interferes with COPI recruitment to the nuclear rim and inhibits NE breakdown (Prunuske et al., 2006). Interestingly, RANBP2 acts cooperatively with Nup153, the most nuclear of the NUPs, which contains a distinct zinc finger domain, in coordinating NE breakdown.
RANBP2 also binds to BICD2 (homologous to Drosophila Bicaudal D), an adaptor between motor proteins and their cargo, and recruits BICD2 to NPCs in the G2 phase of the cell cycle (Splinter et al., 2010). BICD2 in turn regulates dynein-dynactin motor complexes at NPCs, and thus centrosome tethering to the NE prior to mitotic entry. BIC2 is also required for the antagonistic activity of kinesin-1, which pushes centrosomes apart. The balance between dynein and kinesin-1 opposite activities governs centrosomal positioning, and hence sites where centrosomes will nucleate the spindle MTs to form asters and later spindle poles; RANBP2 recruitment of BICD2 to the NPCs just before NE breakdown represents a most upstream step in this cascade of events.
A specialized version of this process takes place in radial glial progenitors (RGPs), from which neurons, glia, and brain adult stem cells originate. RGP nuclei migrate basally during G1, then apically during G2 via dynein, and eventually divide at the ventricular surface. Hu et al. (2013) discovered that apical nuclear migration requires dynein recruitment at NPCs by two cooperating G2-specific mechanisms: the "RanBP2-BicD2" pathway acts first, and "Nup133-CENP-F" operates sequentially. This work identifies spatially regulated mechanisms, implying that only restricted regions of neurogenic tissues are permissive for mitosis: in this context, RANBP2 is essential for dynein control of apical nuclear migration, nuclear membrane remodelling and centrosome dynamics prior to mitosis.

Cell differentiation-associated functions
The NPC is not a static transport gate and undergoes dynamic remodelling during differentiation. In myogenic differentiation, myoblasts fuse to form syncithial myotubes. By atomic force microscopy, NPCs have been found to undergo structural differences during C2C12 myogenic cell differentiation from myoblasts to myotubes, parallel to an increased amount of RanBP2 at NPCs (Asally et al., 2011). siRNA-mediated depletion of RanBP2 in myoblasts suppresses differentiation to myotubes. Thus, RanBP2 is required for NPC remodelling in myogenesis, suggesting that a re-adaptation of transport mechanisms, and of the gateways through which these take place, is required as myotubes fuse and many nuclei become immersed in a common cytoplasm.
RANBP2 can carry out more specialized functions in a tissue-specific manner. RANBP2 is highly abundantly expressed in the vertebrate retina. Its cyclophilin domain (which led Ferreira et al., (1996) to classify RANBP2 as Type-II cyclophilin), and the RB4 domain, interacts with opsin, a retinal transmembrane protein; the RBD4 and cyclophilin domains are therefore proposed to act as a functional "supradomain" with a chaperone function for opsin in the retina. Along with this chaperone function, RANBP2 associates through its cyclophilin-like domain with subunits of the 19S regulatory complex of the 26S proteasome in the neuroretina (Ferreira et al., 1998), and thus contributes to control the stability of proteins that it "chaperones" in the retina.
Related to RANBP2 transport functions but independent from them, Cai et al., (2001) identified a novel domain between RBD2 and RBD3 capable of direct association with two MT-based kinesin motors, KIF5B and KIF5C, in neuronal cells. Preventing the interaction of the RANBP2 kinesin-binding domain (KBD) with KIF5B / KIF5C in neuronal cells caused perinuclear clustering of mitochondria, deficits in mitochondrial membrane potential and cell shrinkage (Cho et al., 2007): thus RANBP2 modulates kinesin-dependent mitochondria transport and function. The RBD2, KBD and RBD3 domains of RANBP2 are proposed to constitute a tripartite domain (R2KR3), modulating mitochondrial transport via kinesin subtypes in subsets of neuroretinal cells (Patil et al., 2013).
Aslanukov et al. (2006) discovered yet another association of RANBP2, via its leucine-rich domain, with Cox11, a mitochondrial metallochaperone, and HKI (hexokinase type I), defined as the "pacemaker" of glycolysis. Cox11 inhibits HKI activity, but RANBP2 suppresses this inhibition. Consequently, RANBP2 haploinsufficient mice show markedly decreased HKI and ATP levels in the central nervous system, with deficits in growth rates and glucose catabolism (Aslanukov et al., 2006). These mice also show absent or severely reduced cell death response to light-induced oxidative stress in the retina (Cho et al., 2010). RANBP2 cell type-conditional mice models, selectively lacking RANBP2 either in rod or in cone photoreceptors (Cho et al., 2013), showed that RANBP2 ablation in cone photoreceptors promoted their non-apoptotic death, while rod photoreceptors underwent cone-dependent non-autonomous apoptosis. Thus, RANBP2 modulates cell type-specific and distinct pathways of cell death - a key feature of neurodegenerative diseases.

Mitosis
- Mitotic spindle organization
Chromosome segregation at mitosis is crucial to the maintenance of genomic stability, a process often disrupted in cancer. A role of RANBP2 in chromosome segregation was first suggested by the finding that RANBP2 accumulates at the mitotic spindle in prometaphase, and in part at KTs upon MT attachment (Joseph et al., 2002; see figure 2); at these sites RANBP2 remains associated in complex with RANGAP1-SUMO1, suggesting that some of its functions entail RANGTP hydrolysis at specific mitotic sites.
In RNAi-based studies in human cells, RANBP2 down-regulation caused multipolar spindles, with supernumerary poles lacking centrioles (hence suggestive of MT dysfunction) as well as defects in chromosome congression and segregation (Salina et al., 2003; Joseph et al., 2004; Klein et al., 2009; Hashizume et al., 2013). Consistent findings in Caenorhabditis elegans embryos (Askjaer et al., 2002) suggest that mitotic functions of RANBP2 are conserved across species in which this protein is present.
- Mitotic microtubule-kinetochore interactions
Further studies showed that RANBP2 depletion resulted in aberrant KT morphology, associated with mis-localization of RANGAP1 and other KT proteins, e.g. Mad1, Mad2, Zw10, Mis12, CENP-A, CENP-E, CENP-F and dynein; RANBP2 depletion also caused lengthened prometaphase duration and chromosome misalignment at metaphase, but the simultaneous depletion of RANBP2 and either Mad1 (Salina et al., 2003) or Mad2 (Joseph et al., 2004), two major spindle assembly checkpoint (SAC) regulators, restored normal prometaphase duration; these findings suggest that RANBP2 depletion-dependent abnormalities activate the SAC. Interestingly, RANBP2 depletion yields unstable KT-MT interactions, suggesting that the concentration of RSSU complex at MT-attached KTs contributes to the functional connections between the spindle and chromosomes prior to chromosome segregation (Joseph et al., 2004). In conditions under which RSSU targeting to KTs was prevented, discrete attachments between MTs and KTs were not maintained, yielding high rates of chromosome mis-segregation (Salina et al., 2003; Joseph et al., 2004; Arnaoutov et al., 2005). Indeed, RANBP2 hypomorphic mice develop severe aneuploidy (Dawlaty et al., 2008).
RSSU targeting to KTs is highly regulated in human cells and requires i) MT attachment to KTs, and ii) proteins that stabilize MT interaction with KTs, e.g. Hec1/Ndc80 and Nuf2 (Joseph et al., 2004). There is therefore a functional cross-talk between proteins that regulate MT/KT interactions, and RSSU recruitment to KTs, which in turn reinforces these interactions. Interestingly, CRM1 is required for the RSSU complex recruitment to KTs (Arnaoutov et al., 2005) while importin beta overexpression inhibits it (Roscioli et al., 2012).
However, neither endogenous RANBP2, nor GFP-tagged RANBP2 constructs, localize to KTs in MEFs (Hamada et al., 2011). Some cell cycle checkpoints are not yet fully proficient in embryonic cell cycles and some of their regulatory mechanisms may diverge from those operating in somatic cells.
It is worth noting that cells subjected to extended RANBP2 RNAi for longer times eventually escape the mitotic arrest with defective MT/KT connections, originating multinucleated cells and/or citokynesis defects (intracellular bridges) followed by cell death (Salina et al., 2003; Joseph et al., 2004). In terms of cancer prognosis, these data suggest that a very narrow threshold, probably modulated by the genetic background of the cells, defines whether defective RANBP2 expression is pro-tumorigenic (by inducing genetic instability in cells that remain viable) or anti-tumorigenic (by preventing normal cell division altogether and inducing the death of the severely aberrant cell products).

Regulation of the SUMO conjugation pathway in mitosis
MEFs with reduced RANBP2 levels are viable and display no overt nuclear transport abnormalities compared to wild-type, yet develop severe aneuploidy associated with chromosome segregation defects, including anaphase bridges (Dawlaty et al., 2008). Chromatin bridges in anaphase are typical of cells in which DNA decatenation is impaired by mutation or inhibition of topoisomerase II alpha (Topo IIa) (Bhat et al., 1996).
Studies in S. cerevisiae (Takahashi et al., 2006), Xenopus egg extracts (Azuma et al., 2003) and human cells have shown that Topo IIa is subjected to sumoylation (Azuma et al., 2003; Azuma et al., 2005; Mao et al., 2000). Indeed, Dawlaty and coworkers observed that i) RANBP2 hypomorphic MEFs fail to accumulate Topo IIa at inner centromeres in mitosis, and ii) RANBP2 SUMO E3 ligase activity is required for Topo IIa SUMO conjugation and inner centromere targeting, to enable decatenation of sister centromeres prior to anaphase onset (Dawlaty et al., 2008).
RANBP2 also associates with the chromosomal passenger complex (CPC) during mitosis and stimulates sumoylation of Borealin (Klein et al., 2009); this, however, affects neither CPC assembly nor its localization; RANBP2-dependent Borealin SUMOylation might be required for CPC interaction with an as yet unidentified protein(s) at centromeres (Klein et al., 2009).

RANBP2 in cell viability
As remarked, RANBP2 inactivation causes early embryonic lethality (Aslanukov et al., 2006; Dawlaty et al., 2008). Hamada and coworkers (2011) studied mitotic cell viability using a Cre-mediated RANBP2 conditional knockout approach. The incidence of chromosome missegregation was 100% for RANBP2-null MEFs, yet these cells did not die during faulty mitosis and rarely died during the next 12 hours after mitotic exit, suggesting that the mitotic errors caused RANBP2 knock-out are not the primary cause of cell death (Hamada et al., 2011). Rescue experiments, expressing various RANBP2 portions in a RANBP2-null background, revealed that a short N-terminal fragment corrected transport defects and restored cell viability, suggesting prominent NPC dysfunction, rather than mitotic failure, as the cause of cell death (Hamada et al., 2011). By contrast, RANBP2 siRNA-silenced HeLa cells underwent prolonged metaphase followed by mitotic catastrophe in live cell imaging (Hashizume et al., 2013); the use of a fluorescently-tagged import reporter demonstrated that, under these conditions, RANBP2-depletion-induced mitotic death is not a side effect of failed nuclear import. The discrepancy between these models remains to be explained.