Recognition of linear targeting signals

Protein cargoes commonly use NLSs and NESs to bind their cognate Kaps for import and export, respectively. Only four classes of NLSs and one class of NESs have been defined, each for a specific Kap, and targeting signals for many Kaps remain undefined^14,15. Each NLS/NES class has distinct sequence patterns and binds to different sites on their cognate Kap. NLSs/NESs do share some common features: they are often found in intrinsically disordered regions (IDRs) or at termini of cargoes; they become ordered, adopting either extended or helical conformations, upon binding Kaps; many NLSs are enriched in basic residues as Kaps are acidic; and the peptides in an NLS/NES class can be highly diverse in sequence. These common features and the exceptions are discussed below.

NLS and NES classes.

The oldest class of NLS is the highly basic classical NLS (cNLS) that binds IMPα proteins (seven subtypes in humans; see Supplementary Table 1)¹⁶, which are adaptors for IMPβ. The two subclasses of cNLS are the monopartite cNLS that is 5–7 residues long (consensus: K-K/R-X-K/R) and the bipartite cNLS with 2 linker-connected basic clusters (consensus: K/R-K/R-X_10–12-K/R_3/5, where X_10–12 is a linker of 10–12 residues and K/R_3/5 refers to three basic residues within five consecutive residues)^17,18. Recent findings suggest that the linkers can be much longer than the classical ones of 10–12 residues and may contain α-helical elements or even entire folded domains^19,20. cNLSs bind two sites, major and minor, on the concave surface of the IMPα armadillo (ARM) repeat domain; most monopartite cNLSs bind to the major site and a few bind to the minor site, whereas bipartite cNLSs occupy both sites²¹. The residues that form the concave surface of IMPα, including the major and minor cNLS binding sites, are virtually identical between IMPα subtypes¹⁶. The amino-terminal IMPβ-binding (IBB) region of IMPα contains an NLS-like₄₉KRRNV₅₃ segment that binds the major cNLS-binding site of unliganded IMPα and is thus autoinhibitory^22,23. IMPβ binding to the IBB frees up the IMPα major site to bind cNLSs on cargoes, increasing cargo-binding affinity by >100-fold^24–27 (BOX 1; TABLE 1). However, some cNLSs bind to IMPα with such high affinity as to displace the IBB in the absence of IMPβ; examples include cNLSs of several inner nuclear membrane proteins (Supplementary Box 2). The extent of IBB autoinhibition varies for different IMPα subtypes, conferring subtype specificity for some cargoes that can bind IMPα in the absence of IMPβ^28,29. The invariant cNLS binding surface shared by all IMPα subtypes suggests that subtype specificity is determined by interactions outside the cNLS-binding sites^16,29. Additional examples of IMPα subtype specificity are discussed below (also reviewed elsewhere¹⁶).

Armadillo (ARM) repeat domain

A protein domain typically 40 residues long that shares homology with repeating units in the ARM protein family. it contains two or three helices per repeat, which stack into a solenoid arrangement.

The second type of NLS defined is the proline-tyrosine NLS (PY-NLS) that binds KAPβ2 or KAPβ2b. PY-NLSs are 15 to >100 residues long, have overall positive charge and bind in mostly extended conformations to a negatively charged site of KAPβ2 (REFS^1,30–32) (FIG. 2). PY-NLSs are very diverse in sequence and best defined by a set of physical, chemical and sequence rules. Sequence motifs include an N-terminal hydrophobic or basic motif and a carboxy-terminal R/K/H-X_2–5-P-Y or R/K/H-X_2–5-P-ϕ motif (ϕ, hydrophobic; H, histidine; P, proline; Y, tyrosine). The N-terminal, R/K/H and PY motifs quasi-independently contribute to overall binding energy and can be used in combination such that two of the three motifs can sometimes confer import activity. Hence, some PY-NLSs do not have a clearly recognizable PY motif³⁰.

Open in a separate window

Fig. 2 |

Comparison of unliganded, cargo-bound and RAN-bound importins.

Surface representations of nuclear and cytoplasmic complexes of importins (grey) with RAN–GTP (teal) or cargo (purple). All importins are structurally aligned at residues 1–200 and displayed in the same orientation. Dashed lines mark top of first HEAT repeat and bottom of last HEAT repeat of each unliganded importin to help visualize importin conformational changes and differences in superhelical compactness. Cargo release by RAN–GTP results from either steric clash between RAN–GTP and cargo or conformational changes to the importin superhelix. PDB IDs (left to right): IMPβ [PDB:2BKU, PDB:1UKL, PDB:3W5K, PDB:1QGK, PDB:3ND2]; KAPβ2 [PDB:1QBK, PDB:2H4M, PDB:2QMR]; KAP121 [PDB:3W3Z, PDB:3W3X, PDB:3S3T]; and TNPO3 [PDB:4C0Q, PDB:4C0O, PDB:4C0P]. ASF, splicing factor 1; Cl, Canis lupus; hnRNP, heterogeneous nuclear RNP; Hs, Homo sapien; IBB, amino-terminal IMPβ-binding; IK-NLS, isoleucine-lysine nuclear localization signal; Kaps, Karyopherins; Mm, Mus musculus; PY-NLS, proline-tyrosine nuclear localization signal; RRM, RNA-recognition motif; RS-NLS, arginine-serine repeat nuclear localization signal; Sc, Saccharomyces cerevisiae; SREBP2, sterol regulatory element-binding protein 2.

The third NLS class is the isoleucine-lysine NLS (IK-NLS) that binds yeast KAP121 (FIG. 2). IK-NLSs have the consensus sequence K-V/I-X-K-X_1–2-K/H/R, where the valine (V)/isoleucine (I) and the lysine (K) in the fourth NLS position are Kap121-binding hotspots and Kap–cargo specificity determinants^33,34. Little is known about sequences that bind the homologous human IPO5, but KAP121 residues that contact IK-NLSs are conserved in IPO5, suggesting that it likely also binds IK-NLS-containing cargoes³³. Both the IK-NLS and monopartite cNLS contain multiple lysine residues and are compact, but with different sequence patterns (K-V/I-X-K-X_1–2-K/H/R versus K-K/R-X-K/R). Little is known about whether they bind KAP121/IPO5 and IMPα interchangeably. One study showed that overexpression of IK-NLS of Ubiquitin-like-specific protease 1 (ULP1) inhibited the KAP121 but not the IMPα pathway³⁵, but potential cross-interactions need further investigation.

The fourth class of NLS is the TNPO3-binding arginine-serine repeat NLS (RS-NLS) found in RS domains (>50 residues long, >40% RS dipeptide) of many serine-arginine (SR) splicing factors. The RS-NLS is often immediately C-terminal to an RNA-recognition motif (RRM) domain. Many RS-NLSs must be phosphorylated on serine residues to bind TNPO3 (REFS^36–38); the alternating arginine and phospho-serine residues bind complementary acidic and basic patches on TNPO3 (FIG. 2). Some cargoes, such as pre-mRNA cleavage factor CPSF6, have arginine-glutamate/aspartate instead of phospho-RS dipeptides³⁶.

At this time, the only known export signal class is the one that binds CRM1, hence it is simply called the NES. NESs are usually 8–15 residues long and contain 4–5 hydrophobic anchor residues that interact with CRM1, arranged with various spacings¹⁴. The first NES sequences discovered had many leucines, but it is now obvious that NESs are very diverse in sequence and structure^14,39,40. NESs can bind in both polypeptide directions and occupy different extents of a structurally invariant hydrophobic groove on the convex/outside surface of the ring-shaped CRM1 (REFS^14,41–44) (FIG. 3a). NES peptides adopt various helix, β-strand, loop or turn structures that position hydrophobic side chains into hydrophobic pockets in the groove of CRM1. The only conserved structural element of the NES is a single helical turn that forms hydrogen bonds with a conserved CRM1 lysine that acts as a selectivity filter¹⁴. An active NES has to meet only this one structural requirement of three or four residues and the rest of it can be structurally diverse, resulting in the large sequence variability of NESs. At this time, there are 11 NES consensus patterns (TABLE 1), but there may be more to be uncovered. The very relaxed sequence requirements of an NES allow CRM1 to recognize thousands of diverse NESs in broadly functioning protein cargoes.

Open in a separate window

Fig. 3 |

Comparison of unliganded, cargo-bound and RAN-bound exportins and biportins.

Surface representations of Karyopherins (Kaps; grey), Kap–RAN–GTP (teal)–cargo (purple) or Kap–RAN–GTP complexes. All Kaps are aligned, displayed in same orientation and marked with dashed lines as in FIG. 2. a | Exportin structures. Autoinhibitory carboxy-terminal helix/loop (red) of CRM1 and XPO5 are shown in the unliganded exportins. Locations of 3′ overhangs of pre-microRNA (pre-miRNA) and tRNA, recognized by different regions of XPO5 and LOS1, respectively, are also indicated. RAN–GTP binding to XPO1 and XPOT cause ring compaction for cargo binding, whereas RAN–GTP binding to CSE1 and XPO5 opens them up to accommodate cargoes. b | Biportin IPO13 structures. Cargo release from IPO13 is induced by RAN–GTP using steric clash mechanism and conformational changes to the Kap superhelix, to bind specific import and export cargoes. PDB IDs (left to right): CRM1 [PDB:3GJX, PDB:4FGV]; CSE1 [PDB:1WA5, PDB:1Z3H]; XPO5 [PDB:3A6P, PDB:5YU7]; LOS1 [PDB:3ICQ, PDB:3IBV]; and IPO13 [PDB:2X19, PDB:3ZJY, PDB:2XWU, PDB:2X1G, PDB:3ZKV]. Cl, Canis lupus; Ct, Chaetomium thermophilum; Dm, Drosophila melanogaster; eIF1A, eukaryotic translation initiation factor 1A; Mm, Mus musculus; NES, nuclear export signal; Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe.

Recognition of folded domains

Some protein cargoes use folded domains rather than linear NLSs/NESs to bind their Kaps. Although IMPα predominantly binds to linear cNLSs, a few cargoes exclusively use folded domains to bind a distinct non-cNLS-binding site on IMPα, thus conferring IMPα subtype specificity (for example, Ebola virus secondary matrix protein VP24 binding to IMPα6 (REF.⁵⁶)). Several IMPβ−cargo interactions also involve globular cargo domains. Examples include the cholesterol regulator sterol regulatory element-binding protein 2 (SREBP2) and the transcription factor zinc finger protein SNAI1, which bind IMPβ through their dimeric helix–loop–helix leucine zipper domain and zinc finger domains, respectively. Both IMPβ-binding domains are enriched in basic residues, and the flexible IMPβ superhelix adopts different helical pitches and uses different sites to bind these very differently sized and shaped domains^57,58 (FIG. 2). Due to sterically clashing binding sites and/or allosteric conformation changes of IMPβ, the binding of these domains and the IMPα IBB to IMPβ is mutually exclusive, with the IBB binding with higher affinity. It is likely that more cargoes will be found to bind IMPβ through folded globular domains.

Other importins also bind folded domains of their import cargoes. IPO9 was recently shown to bind the highly basic histone H2A–H2B dimer by wrapping around its globular histone-fold domain⁵⁹. IPO9 was also reported to bind very differently folded cargo regions such as the C-terminal HEAT repeats of serine-threonine-protein phosphatase PP2A and the homeobox domain of developmental regulator Aristaless^60,61. Similar to IMPβ, IPO9 probably uses different sites and binding modes to bind structurally diverse cargo domains. IPO4, IPO5, IPO7, IPO8 and IPO11 all bind diverse folded cargo domains but little information is available about these interactions^1,11. IPO5 binds to a helical region of the high-density lipoprotein component apolipoprotein A-I, IPO8 binds the globular single-domain eIF4E and IPO11 binds the compact five-helix matrix domain of the Rous sarcoma virus Gag polyprotein^1,62.

Biportins IPO13, XPO4 and XPO7 also bind folded domains of their import cargoes. Recent structures showed IPO13 using different sites to bind the diverse domains of two different import cargoes: the exon junction complex MAGO-Y14 (REF.⁶³) and SUMO-conjugating enzyme UBC9 (REF.⁶⁴) (FIG. 3b). The yeast homologue of IPO13, PDR6, also imports yeast UBC9 but uses a binding mode different from that of IPO13 (REFS^12,64,65). IPO13 and PDR6 bind different regions of UBC9 using their N-terminal and C-terminal HEAT repeats, respectively. XPO4 binds the three-helix HMG box domains of import cargoes Sex-determining Region Y protein (SRY) and SRY-box transcription factor 2 (SOX2)⁶⁶. A recent proteomics study identified structurally diverse candidate import cargoes for XPO7 but it is not yet known whether they bind using linear elements or folded domains¹¹.

Many export protein cargoes use folded domains to bind exportins (other than CRM1 and MSN5) and biportins. The exportin CSE1 (yeast homologue of CAS) wraps around both the ARM domain of autoinhibited KAP60 (yeast IMPα homologue) and RAN–GTP (FIG. 3a). XPO6 exports the profilin–actin complex, which has only folded globular domains and no IDRs; however, XPO6 binding to profilin–actin has not been biochemically demonstrated and it is unclear whether additional factors, potentially with linear motifs, are involved. Notably, the nuclear tumour suppressor protein RASSF1A appears necessary for RAN–GTP interaction with XPO6 and for nuclear export of profilin–actin⁶⁷. Biportins IPO13, XPO4 and XPO7 also recognize diverse folded domains, such as histone fold heterodimers and helical homeodomains, in their export cargoes (reviewed elsewhere¹). IPO13 binds the folded domain of export cargo eukaryotic translation initiation factor 1A (eIF1A) that binds to a site distinct from those for import cargoes⁶⁸ (FIG. 3b). XPO4 binds the MH2 domain of SMAD3 (REF.⁶⁹), as well as the SH3-like domain and OB domain of eIF5A (REFS^70,71).

In summary, Kaps use many different binding sites to bind diverse folded cargo domains. The flexible superhelical architecture also enables Kaps to adopt various conformations to accommodate diverse cargo domains.

The combined use of motifs

A Kap may also bind to both a linear element and folded domain in a cargo. For example, folded domains immediately adjacent to linear cNLSs (such as in RCC1) require certain IMPα subtypes with more flexible ARM domains such as IMPα3 to accommodate the bulky fold^29,72,73. The unusual IMPβ–IPO7 heterodimer binds both the globular histone-fold domain of histone H1 and its C-terminal disordered tail⁷⁴. TNPO3 binds both the RS-NLS and the RRM domain of splicing factor 1 ASF1 (also known as SF2) (FIG. 2); the RS-NLS is the primary TNPO3-binding determinant and the RRM a secondary element^1,36,75,76. Many TNPO3 cargoes have RRM domains but many candidate TNPO3 cargoes identified by proteomics do not⁷⁵; it is unclear how prevalent TNPO3–RRM interactions are. Lastly, SNUPN binds CRM1 in a multipartite manner using an N-terminal NES, a folded globular m³g cap-binding domain and a disordered C-terminal tail, all binding to the outer convex surface of CRM1 (REFS^41,42) (FIG. 3a). It is not known whether other cargoes also use folded domains, with or without an NES, to bind CRM1.

RAN–GTP-regulated cargo import

For nuclear import, the binding of RAN–GTP versus cargo is mutually exclusive. This selection is driven by the existence of a steric clash between import cargo and RAN–GTP when binding importin (steric mechanism), by the diverse conformational changes of the importin upon binding RAN–GTP that abolish cargo binding sites (allosteric mechanism) or a combination of the two mechanisms (FIGS 1b–d and ​and2).2). TNPO3 uses a steric clash mechanism, with overlapping RAN–GTP and RS-NLS binding sites⁷⁵, whereas IMPβ primarily rearranges HEAT repeat packing into a conformation that is incompatible with cargo binding^96,97. Other importins use combinations of steric and allosteric mechanisms. A long acidic loop between the helices of HEAT repeat 8 in KAPβ2, which is disordered in the KAPβ2–cargo complex, undergoes a drastic conformational change when it interacts with RAN–GTP, allowing it to engage the PY-NLS binding site and displace the cargo⁹⁴. For KAP121, RAN–GTP binds at a site that partially overlaps with the IK-NLS binding site, and also causes the KAP121 superhelix to elongate, closing the IK-NLS binding pockets³³.

Although, in most cases, binding of RAN–GTP to the importin is sufficient to release the cargo, there are a few reports of importin–cargo complexes where RAN–GTP alone is not able to mediate this release. The most clearly demonstrated example is the RAN–GTP inability to release histones H2A–H2B from IPO9 (REF.⁵⁹) or from its yeast homologue KAP114 (REFS^98,99). RAN–GTP is also unable to release another KAP114 cargo, the TATA-binding protein¹⁰⁰, or the mRNA-binding protein NPL3 from its importin, MTR10 (REF.¹⁰¹). Other examples with conflicting data also exist, such as the release of RNA-binding proteins NAB2 and NAB4/HRP1 from yeast KAP104 (KAPβ2 homologue)^31,102; further studies are needed to confirm the RAN–GTP insensitivity of this release. In all of these cases, nucleic acid targets (chromatin, RNA) of the cargoes are required along with RAN–GTP for efficient cargo release, suggesting that locations of cargo release in the nucleus may be specified by downstream functions of the cargoes. The mechanisms of this cargo release mediated by the combined effect of the nuclear target and RAN–GTP are unknown at this time.

RAN–GTP-regulated cargo export

To enable nuclear export, exportins show little or no binding to cargo or RAN–GTP alone, and instead bind both cooperatively. XPO5/MSN5 is an exception as it binds stably to RAN–GTP alone and to some primary miRNA precursors^95,103,104, perhaps suggesting additional miRNA processing roles for this exportin¹⁰⁴. Coupled cargo and RAN–GTP binding involves various large conformational changes in most exportins and direct contacts between RAN–GTP and cargo that stabilize the ternary complex (FIGS 1b–d and ​and3a;3a; CRM1 is an exception, see below). Unliganded CSE1 and XPO5 form closed-ring structures that open up when binding both cargo and RAN–GTP to accommodate both molecules inside the exportin solenoids^{91,93,103,105}. Conversely, unliganded CRM1 and LOS1 adopt open conformations, but undergo compaction when binding cargo and RAN–GTP to interact with both molecules favourably; compaction of CRM1 is one of the many conformational changes that leads to the opening of the NES binding groove (reviewed elsewhere¹⁰⁶ and more below) whereas the LOS1 superhelix clamps onto both RAN–GTP and cargo^41,42,92,107. Additionally, binding of RAN–GTP displaces C-terminal autoinhibitory tails/helices of both CRM1 and XPO5, to facilitate the opening or closing of the exportin superhelix, respectively^103,108,109.

As mentioned earlier, CRM1 is the only exportin where the RAN–GTP and cargo binding sites are distant, with no direct RAN–GTP–cargo contacts. This is because CRM1 uses its outer/convex surface to bind cargoes, whereas Ran–GTP interacts with the inner/concave surface. Coupling of RAN–GTP and cargo binding is thus driven by conformational changes to open or close the NES-binding groove in CRM1. The precise mechanisms for CRM1 export cargo dissociation vary in yeast and human even though both involve RanBP1. Yeast RanBP1 binds the inner CRM1 surface in the RAN–GTP–CRM1–cargo complex, which induces a conformational change that closes the NES-binding groove to release cargo¹¹⁰. By contrast, human RanBP1 strips RAN–GTP from the RAN–GTP–CRM1–cargo complex to cause subsequent cargo release¹¹¹.

Changes to cargoes

Studies of the IMPα–IMPβ and CRM1 systems show that higher cNLS/NES binding affinities correlate with greater nuclear or cytoplasmic localization within an optimal range, respectively^27,112 (TABLE 1). Post-translational modifications (PTMs) in or adjacent to signals in the cargo that bind to Kaps can change Kap–cargo affinities and thus cause a change in their cellular localization (FIG. 4b). The effects of PTMs are usually highly dependent on their position within the full-length cargo proteins. This topic was previously reviewed^113,114 and we highlight recent examples here. PTMs of NLSs/NESs, such as phosphorylation, acetylation or methylation, are known to affect Kap affinities. In some instances, the effects are predictable; for example, phosphorylation of the MSN5-binding cargo segments and the TNPO3-binding RS-NLS is critical for Kap binding, whereas arginine methylation of the FUS RGG regions impairs FUS interaction with KAPβ2 (REF.¹¹⁵). In other cases, the effects are difficult to predict; for example, phosphorylation of cNLSs inhibits import of some cargoes but enhances import of others. Recent examples of cNLS phosphorylation include serine phosphorylation of the bipartite cNLS linkers of tumour suppressor protein 53BP1 (REF.¹¹⁶) and peptide hormone GLP1 (REF.¹¹⁷) that decreased importin binding and nuclear import. Acetylation of lysines in the cNLSs of the glucose metabolic enzyme PFKFB3 (REF.¹¹⁸) and the antiviral protein IFIX¹¹⁹ also inhibited import, but lysine acetylation of the tyrosyl-tRNA synthetase (TyrRS) cNLS promoted import; in the latter case, the unacetylated cNLS is helical and acetylation is suspected to destabilize the helix to generate an extended and active cNLS¹²⁰. Effects of NES modifications are similarly difficult to predict. Phosphorylation near the NES of transcription factor TFEB¹²¹ increased CRM1 binding but phosphorylation of the hypoxia-inducible factor HIF2α NES is inhibitory¹²². Methylation of a lysine adjacent to the NES of the transcriptional regulator YAP1 also inhibited CRM1 binding¹²³.

Modifications that change the accessibility of NLSs/NESs to Kaps can also affect Kap–cargo binding (FIG. 4b). Conformational changes of the cargo or changes in cargo oligomerization may expose or sequester targeting signals. Both types of changes may be induced by PTMs or by binding to additional macromolecule partners. Classic examples include tetramerization of the tumour suppressor protein p53 that masks its NES, and the masking of the NLS in the transcription factor NF-κB/RELA by its binding to a regulatory partner protein IκB (reviewed in REF.¹²⁴). More recently, dephosphorylation of serine residues distant from the putative NLS of transcription factor YB1 was suggested to increase NLS accessibility and nuclear entry during the late G2/M phase of the cell cycle, mediated by a conformational change¹²⁵. By contrast, phosphorylation of the apoptotic regulator caspase 2 causes binding of scaffolding protein 14-3-3 that masks the caspase 2 cNLS, inhibiting its nuclear import and activation^126,127. Similar control of NES accessibility has also been observed. Acetylation of the N-terminal IDR region in the fly homeobox protein Ubx, which is far from its NES, seems to mask the NES via an unknown mechanism, to inhibit export¹²⁸.

Mutations of NLSs and NESs, such as those that occur in neurodegenerative diseases and cancers, change cargo localization by disrupting functional targeting signals or by generating new ones. Many mutations in and/or adjacent to the PY-NLSs of mRNA binding proteins FUS and heterogeneous nuclear RNPs (hnRNPs) A1 and A2B1 were found in patients with familial amyotrophic lateral sclerosis and multisystem proteinopathy^129–132. Mutations in the PY-NLSs of hnRNPs H1 and H2 are also thought to drive the mental retardation, X-linked, syndromic, Bain type (MRXSB) disease^133,134. All these mutations likely decrease KAPβ2 binding and disrupt nuclear import. Conversely, frameshift mutations in nucleophosmin create potent NESs that lead to aberrant CRM1-mediated nuclear export in a subtype of acute myeloid leukaemia¹³⁵. The W1038X truncation mutation in tumour suppressor PALB2 also exposes an otherwise occluded NES, mislocalizing the protein in breast/ovarian cancers¹³⁶. Alternatively, amyotrophic lateral sclerosis-linked mutations in superoxide dismutase 1 (SOD1) cause SOD1 misfolding that exposes an NES for CRM1 export¹³⁷.

In summary, modifications to cargoes, both within and outside Kap-binding regions, can change Kap–cargo binding affinities in many different ways. These changes to cargoes ultimately result in their altered subcellular localization.

Import of the replisome core

Assembly of the multiprotein DNA replication machinery, or replisome, on DNA has been reviewed extensively, but its trafficking to the nucleus is rarely discussed. Components of the eukaryotic replisome core include the CMG helicase (named for its components — cell division control protein 45 (CDC45), minichromosome maintenance complex (MCM2–MCM7) and the GINS complex (SLD5–PSF1–PSF2–PSF3)), three DNA polymerases (Polε, Polδ and Polα-primase), processivity clamp proliferating cell nuclear antigen (PCNA) and clamp loader replication factor C (RFC), clamp unloader RFC-like complexes (RLC) and the single-stranded DNA-binding protein RPA (Fig. 5a). The components of the CMG helicase seem to be imported separately prior to their assembly in the nucleus. Yeast IMPα–IMPβ is responsible for importing the bipartite cNLS-containing CDC45 (REFS^158,159) and the MCM2–MCM7 complex (where MCM2 and MCM3 have cNLSs)^160–162. The latter is assembled in the cytoplasm by loading factor CDT1, which is imported with the MCM complex to load it onto DNA^163,164. It is not known how the GINS complex is trafficked. Mechanisms of trafficking for the remaining replisome components are largely unclear, with some conflicting reports. The p66 and p125 subunits of human Polδ have putative cNLSs^165,166. However, the C-terminal α-helical putative cNLS of p125 is atypical and it is unclear whether it is functional for import¹⁶⁷. No information is found on nuclear import of Polε. The mouse Polα-primase is imported as a heterotetrameric complex; putative cNLSs in its p180 and p54 subunits direct import of smaller assemblies but are dispensable for import of the larger complex^168,169. A bipartite cNLS in p180 is buried in a folded domain, unlikely to drive import^170,171. The homotrimeric PCNA binds IMPβ but not IMPα, and the IMPβ-binding region (conserved from human to yeast) is partially located at the trimer interface^172,173, suggesting that PCNA is trafficked as a monomer either by binding directly to an importin or by binding the cNLS-containing cyclin-dependent kinase inhibitor p21 (REF.¹⁷⁴). The large subunit of the penta meric human RFC clamp loader, replication factor C subunit 1 (RFC1), and that of an RLC clamp unloader, ATPase family AAA domain-containing protein 5 (ATAD5), have putative N-terminal NLSs that have not been verified experimentally^175,176. Nuclear import of RPA requires its C-terminal tetramerization domain, but there are conflicting reports of different importers including KAP95 and MSN5 in yeast, and the involvement of RIP as an IBB-containing IMPβ adaptor for RPA in the frog^46,177–179.

Open in a separate window

Fig. 5 |

Nuclear trafficking of three different gene expression machineries.

a | Most replisome components are believed to be imported by importin IMPα–IMPβ, but only two proteins — cell division control protein 45 (CDC45) and minichrosome maintenance (MCM) complex — have validated classical nuclear localization signals (cNLSs; yellow triangles), whereas various others have only unverified NLSs or their NLSs have not be identified (?-NLS; dashed, yellow triangles). Xenopus laevis replication protein A (RPA) may be imported by IMPβ via adaptor RPA-interacting protein (RIP), which contains an IMPβ-binding (IBB) region. A few components of the replisome may be imported by other Karyopherins (Kaps), and trafficking of others, such as GINS complex and polymerase ε (Polε), are unstudied (?).Similarly, only a few components seem to have nuclear export signals (NESs; red triangles) and are exported by CRM1 — it is unclear whether others are recycled back to the cytosol. b | Import of the yeast RNA polymerase II (Pol II) complex. Interacting with RNA polymerase II protein 1 (IWR1) acts as an adaptor and binds IMPα–IMPβ to import fully assembled RNA Pol II. IWR1 is recycled back to the cytosol by CRM1 exportin. Alternative import pathways of partially assembled RNA Pol II were proposed, such as via passive diffusion of small components or additional factors such as Regulator of transcription 1 (RTR1), as well as Glycine-proline-asparagine (GPN)-loop GTPase proteins. These factors are exported by CRM1 possibly via the NES within GPN protein NPA3. c | Several ribosomal proteins (RPs), such as uL14, uL23, eS7 and eS26, are imported by multiple importins. Others, such as uL11, are imported by specific importins, and yet others are imported bound to their chaperones (red ovals). Nuclear RAN–GTP dissociates most Kap–RP complexes, releasing the RPs into the nucleoplasm for complex assembly; exceptions include those that require dedicated chaperones, called escortins, for release or other, unexplored mechanisms (?). Assembled and mature pre-40S and pre-60S subunits are exported either by CRM1 or alternate pathways that involve mRNA export receptor MEX67-MTR2 (not pictured). ACL4, assembly chaperone of RPL4; ATAD5, ATPase family AAA domain-containing protein 5; BCP1, Bacterioferritin co-migratory protein 1; GINS, ‘go-ichi-ni-san’ in Japanese meaning ‘5-3-2-1’ (SLD5-PSF1-PSF2-PSF3); LTV1, low-temperature viability protein 1; NMD3, nonsense-mediated mRNA decay protein 3; PCNA, proliferating cell nuclear antigen; PNO1, partner of NOB1 (nuclear integrity protein 1 (NIN1) binding protein 1); PY-NLS, proline-tyrosine nuclear localization signal; RFC1, replication factor C subunit 1; RIO2, RIO (right open reading frame)-type kinase 2; SYO, symportin 1; TSR2, 20S ribosomal RNA accumulation protein 2; YAR1, yeast ankyrin repeat-containing protein 1.

Some replisome components are found to also exit the nucleus. The MCM complex is exported after replication in subsequent S, G2 and M phases — probably by CRM1 (the NES was identified in yeast MCM3) — which is regulated by cyclin-dependent kinases^162,180. Human p125 is also exported by CRM1 but no NES has been reported¹⁶⁶. CRM1 exports human PCNA by binding an NES accessible only in the monomer¹⁸¹, consistent with export of monomeric PCNA and its assembly and disassembly on DNA.

In summary, a few replisome components have been found to carry cNLSs and some others may ‘piggy-back’ on cNLS-containing proteins, such as the PCNA binding to p21, for import. The use of cNLSs for the import of the replisome components is accompanied by the involvement of KAP60 in regulating G1/S and G2/M transitions in yeast, suggesting that key DNA replication and cell cycle checkpoint proteins are imported by IMPα–IMPβ to coordinate cell cycle progression with genome replication^182,183. However, import mechanisms for the majority of replisome components remain unknown. Similarly, CRM1-mediated export of three replisome components was reported, but it is unclear whether more components are actively exported out of the nucleus and how this export impacts the activity of the replisome.

Import of RNA polymerase II

We focus on RNA polymerase II (RNA Pol II) to describe nuclear import of a large transcription machinery (FIG. 5b). The 12-subunit RNA Pol II is fully assembled in the cytoplasm before import (reviewed in REF.¹⁸⁴). Yeast RNA Pol II nuclear localization protein IWR1 contains a cNLS conserved from yeast to human, which binds IMPα–IMPβ, and is believed to be the major import adaptor for assembled RNA Pol II¹⁸⁵; the effect of the human homologue of IWR1 has not been verified. Subunit sub-assemblies and small individual subunits can also enter the nucleus in yeast, prior to assembly, by active transport and passive diffusion, respectively¹⁸⁶. Regulator of transcription 1 (RTR1; RPAP2 in human) may be the import factor that mediates the entry of individual large subunits of the RNA Pol II and subunit assemblies^186,187; however, no NLS has been identified for RTR1. RPAP2 is also implicated in the import of human RNA Pol II¹⁸⁷. Glycine-proline-asparagine (GPN)-loop GTPases were also reported as putative import factors in both human and yeast (reviewed in REF.¹⁸⁴). Pull-down mass spectrometry studies revealed possible interactions of human GPN1/2/3 with IMPα2 (although no cNLSs were identified)¹⁸⁸, but no such interaction was shown for yeast NPA3 (GPN1 homologue)¹⁸⁹. However, some studies suggested that GPNs act upstream of IWR1 as cytosolic chaperones for RNA Pol II assembly, rather than as import adaptors^190,191.

Several RNA Pol II import adaptors are exported from the nucleus via CRM1, but export of the polymerase has not been reported. The RNA Pol II machinery may be passively retained in the nucleus due to its large size and association with DNA, consistent with studies showing that the nuclear proteome is primarily maintained by complex assembly and passive retention rather than active export⁶. IWR1 is displaced from RNA Pol II by transcription initiation factors and then exported via a putative NES¹⁹². RTR1/RPAP2 is exported in a CRM1-dependent manner (but no NES has been identified)^187,193, possibly through binding NES-containing GPNs (the NES was so far identified in GPN1/NPA3 (REF.¹⁹⁴); reviewed elsewhere¹⁸⁴).

In summary, generally, RNA Pol II complex assembles in the cytoplasm and uses import adaptors for nuclear import which are recycled back to the cytoplasm by CRM1. However, other mechanisms for RNA Pol II nuclear import, involving individual subunits and their smaller assemblies, were also reported. The mechanism of function of these import adaptors and their evolutionary conservation remain mostly unclear. The contribution of the multiple import mechanisms to RNA Pol II biogenesis and function also remains to be addressed.

This work was funded by the National Institute of General Medical Sciences (NIGMS) of the National Institutes of Health (NIH) under Awards R01GM069909 and R35GM144137 (Y.M.C.), the Welch Foundation Grant I-1532 (Y.M.C.), Cancer Prevention Research Institute of Texas (CPRIT) Grant RP180410 (Y.M.C.), support from the Alfred and Mabel Gilman Chair in Molecular Pharmacology, Eugene McDermott Scholar in Biomedical Research (Y.M.C.), the Gilman Special Opportunities Award (H.Y.J.F.) and NIGMS Molecular Biophysics Training Program T32GM131963 (C.E.W.).