Plasticity of an Ultrafast Interaction between Nucleoporins and Nuclear Transport Receptors

1The plasticity of intrinsically disordered proteins (IDPs) is thought to be key to their highly diverse roles in the eukaryotic interactome and a variety of vital processes such as transcription, epigenetic regulation mechanisms, and transport through nuclear pore complexes (NPCs). The central channel of the NPC is filled with phenylalanine-glycine-rich proteins, called FG-nucleoporins (FG-Nups) that are intrinsically disordered. FG-Nups build up an approximately 30-nm-thick permeability barrier through which large molecules (>40 kDa) can only be shuttled when bound to a nuclear transport receptor (NTR) with passage times as fast as 5 ms. Due to the intrinsic dynamics of the FG-Nups, even state-of-the-art electron tomographic studies are not able to visualize them within the central NPC channel, despite their millimolar concentrations. Consequently, the molecular structure of the permeability barrier and its general mode of action are widely debated (for a review see ).

2The key to understanding the observed nucleocytoplasmic transport phenomena resides in a description of the binding mode between FG-Nups and NTRs, for which a molecular analysis of the FG-Nup⋅NTR interaction is a prerequisite. Our current understanding of the molecular basis of FG-Nup⋅NTR interactions is in large part derived from X-ray crystallographic structures or molecular dynamics (MD) simulations of NTRs in the presence of short FG-peptides (up to ∼13 amino acids in length), as well as binding measurements with different NTRs or mutated NTR binding pockets. Even for FG-Nups alone, only overall chain dimensions or long-range interactions within the Nups have so far been analyzed in solution. Notably, even such fundamental binding characteristics as the equilibrium dissociation constant (Kd) between Nups and NTRs are still matter of discussion - estimates range from a few nM to several mM. However, high Kd (low affinity, ∼mM) values are not easily compatible with high specificity of the transport process, while low Kd values (∼nM range) cannot easily explain high transport rates, since these might be expected to correlate with long residence times whereas NTRs must encounter many FG-Nups while crossing the thick barrier.

3Fast protein binding also typically requires proper orientation of the protein binding partners as well as conformational adaption of the IDP to bind to a folded protein. Those can occur prior to or during binding, as described by either of the two prevalent models for protein binding namely conformational selection and induced fit. While such a conformational shift or fit can present the rate-limiting step of binding, fast binding is warranted in many biological processes. Several binding rate enhancing effects have been suggested or observed experimentally, such as maintenance of a degree of disorder (termed “fuzziness”; ) by conformational funneling, a large capture radius of the flexible IDPs, and the involvement of long-range electrostatic interactions to steer (attract) proteins together.

4In this work, we characterize the conformational plasticity of Nups from human and yeast in the presence of structurally and functionally diverse NTRs. A focus was a PxFG-rich domain of the Nup153 (Nup153FGPxFG) as its size permitted a combination of nuclear magnetic resonance (NMR), single molecule Förster resonance energy transfer (smFRET), and molecular dynamics (MD) simulations to characterize local, residue specific, as well as long-range implications of Importinβ binding to Nup153FGPxFG conformation and dynamics. Additional Brownian dynamics (BD), fluorescence stopped-flow and single molecule transport experiments with functional NPCs in permeabilized cells, revealed the detailed kinetics of the complex formation between Nup and NTR. Using this molecular, integrative structural biology approach we propose a mechanism whereby Nups contribute low-affinity minimalistic binding motifs that act in concert to create a polyvalent complex. The global Nup structure and dynamics are largely unaffected by the interaction, thereby ensuring ultrafast binding and unbinding of individual motifs—a result that explains how nuclear transport can be fast yet specific, and that may have general implications for the mechanism of action of other IDPs that exhibit a multiplicity of binding motifs.

5We initially characterized the structure and dynamics of Nup153FGPxFG using high resolution NMR (Figure 1A, sequences given in Supplemental Experimental Procedure). Complete assignment of the backbone resonances (Figure S1) allowed us to develop a multi-conformational model of the protein in solution using a combination of Flexible-Meccano and the genetic algorithm ASTEROIDS. Representative ensembles comprising 200 conformers were selected on the basis of the experimental chemical shifts and were in excellent agreement with 1DN-NH and 1DCα-Hα residual dipolar couplings and small angle X-ray scattering (SAXS) curves that were not used in the selection process (Figures 1B–1D). The amino acid specific backbone dihedral angle distributions determined from the ensemble selections (Figure S1) show that negligible secondary structure is present.

6We labeled Nup153FGPxFG with a donor (Alexa488) and acceptor dye (Alexa594) for FRET at its C- and N terminus, respectively. This allowed us to measure average distance between the dyes as well as the dynamic properties of the protein using histograms relating FRET efficiency (EFRET) and donor lifetimes (τ) of single molecules (sm), a method widely used to detect even minute changes in structure and dynamics, for example when IDPs bind, fold or expand.

7We added unlabeled Importinβ to the FRET labeled Nup153FGPxFG and followed the smFRET response. While the diffusion of Nup153FGPxFG in the absence and presence of Importinβ confirmed the binding of Importinβ under single molecule conditions (Figures 2 and S2), we detected neither substantial changes in EFRET nor in the width of the histograms indicating absence of significant changes in the distance distribution (Figure S2 shows an all F to all A negative control). Indeed, the EFRET populations of the unbound and bound Nup153FGPxFG also overlay very closely with respect to τ, which indicates similarly fast dynamics of both forms (Figures 2 and S2 for detailed analysis of structure and dynamics).

8As smFRET is compatible with large proteins, we were able to repeat the same experiments for the same PxFG region within the full-length Nup153FG (601 amino acids), finding similar characteristics, and suggesting that our truncated Nup153FGPxFG largely retains the conformational sampling from within the whole Nup153FG (Figure S2).

9In order to determine the general nature of this binding mode, we repeated the experiments with two different FxFG-rich regions of Nup153FG, as well as the GLFG-rich yeast Nup49 and several different NTRs: i) transportin 1 (TRN1), a transport receptor involved in the import of proteins containing an M9 recognition sequence, ii) nuclear transport factor 2 (NTF2), the import receptor of RanGDP and iii) chromosomal region maintenance 1 (CRM1), a major exportin. While TRN1 and CRM1 have a similar molecular weight and superhelical structure as Importinβ, NTF2 is a much smaller, β sheet-rich dimer. As detailed in Figure S3, despite the very distinct functionalities of the different NTRs, the smFRET and FCS measurements of the different Nups and NTRs indicate similar binding characteristics as for the Nup153FG⋅Importinβ complex.

10To characterize the effects of Importinβ binding on Nup153FGPxFG at atomic resolution, we titrated Importinβ into a solution of 15N labeled Nup153FGPxFG and measured 1H-15N HSQC spectra at different molar ratios. Peak intensities, as well as 1HN and 15N chemical shifts of Nup153FGPxFG, were analyzed for each titration step (Figures 3 and S4). Resonance line broadening, associated with small changes in both 1HN and 15N chemical shifts, was observed around all F’s in the Nup sequence (Figure 3A). Binding was clearly highly localized, and limited to F’s, with only F and the immediately adjacent amino acids being affected by the interaction. Interestingly, one single F, which is not associated with a G, is also involved in binding to Importinβ, showing the largest chemical shift changes in the 1H-15N HSQC spectrum during titration with Importinβ (Figure 3A and S4). 15N relaxation rates measured as a function of molar ratio of Importinβ suggest that, overall, the molecule remains flexible in the complex with the transverse relaxation (R2) increasing significantly upon Importinβ titration only around the interaction sites (Figures 3C and S4), in agreement with the above smFRET-based observations that global disorder and flexibility are not affected by Importinβ binding. Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion experiments (Figure S4) suggested that fast exchange (< 10 μs) between the bound and unbound form of Nup153FGPxFG gives rise to the increased R2 rates around the interaction sites, which makes it possible to estimate a residue-specific Kd,individual for each position in Nup153FGPxFG with Importinβ (Figures 3E, 3F and S4) from the population weighted R2 measurements. Interestingly, the FG-specific affinities to Importinβ are not identical across the Nup153FGPxFG sequence, implying a contribution of inter-FG residues to binding, although all FG-specific Kd,individual values lie in the millimolar range.

11Strikingly, when studying the binding to different NTRs like TRN1 and NTF2 (Figure S4), despite exhibiting different binding preferences for FG-Nups, their binding modes are remarkably similar to that of the Importinβ complex. The same regions in Nup153FGPxFG are affected by the interaction, again with very low residue specific affinities, with the Nup remaining overall flexible when bound while interacting only locally as seen from both chemical shift changes, in the case of NTF2, and remarkably similar locally elevated transverse relaxation rates in TRN1 (Figure S4). Comparison of 13C backbone chemical shifts measured in the free and NTF2-bound forms of Nup153FGPxFG demonstrates that the protein backbone remains flexible upon interaction, sampling effectively the same conformational equilibrium in the free and bound state (Figure S4).

12We note that during the publication process of this work, localized interaction was also reported for the yeast Nsp1 with Kap95 (the yeast homolog of Importinβ) using NMR, suggesting that a similar interaction mechanism may also be conserved across species.

13To further quantify the action of multiple FG-repeats, we designed a Nup construct, in which all F of Nup153FGPxFG except F1374, the strongest interaction site for Importinβ, were replaced by A (Figure S1). Titration of Importinβ into this Nup153AGPxAG,F1374 mutant resulted in strongly reduced peak broadening and negligible chemical shift changes compared to Nup153FGPxFG (Figure S4). As in the case of Nup153FGPxFG, 15N R2 relaxation rates of Nup153AGPxAG,F1374 at the interaction site exhibited a linear dependence on Importinβ concentration (Figure 3E). However the effective Kd,individual from F1374 within Nup153AGPxAG,F1374 reveals significantly weaker binding for this interaction site than for F1374 when situated within the wild-type (WT) protein (Kd,individual = 7.3 mM compared to 0.8 mM, Figure 3). This result clearly shows that presenting multiple equivalent binding sites to the binding partner has a measurably positive effect on the effective affinity of the individual interaction site.

14We employed MD simulations to investigate the experimental observations of Nup153FGPxFG⋅Importinβ association from NMR and smFRET. From a broad ensemble of Nup153FGPxFG obtained from unbiased MD simulations in explicit solvent (Movie S1), we incubated different conformers with the N-terminal portion of Importinβ (from here named ImportinβN) and monitored their binding for a total simulation time of 2 μs (Figures S5 and S6, and Table S1). The association of Nup153FGPxFG to ImportinβN was repeatedly observed within the simulated timescale and occurred in a specific manner (Figures 4 and S5, and Movie S2). FG-repeats docked into previously identified binding pockets on the surface of ImportinβN and even formed contacts similar to those previously observed crystallographically upon interaction between Importinβ and Nsp1-derived peptides (Figures 4C and S6). Binding was reduced and less specific for Nup153FGPxAG (Figure S5), in agreement with NMR and smFRET (Figures S1, S2, and S4).

15We suggest that the high solvent exposure of Fs in the unbound state (typically contained within the hydrophobic interior of folded proteins) (Figure S5) renders them readily available for Nup153FGPxFG⋅ImportinβN association, without requiring any global structural transitions in either partner (Figures 4D, S6, Movie S2).

16The ability to monitor spontaneous Nup153FGPxFG⋅ImportinβN association on the sub-microsecond timescale suggests an ultrafast association (Figure S5). Underlining the generality of our observation, we were also able to monitor such a spontaneous binding event when repeating simulations for an FxFG-rich region of Nup153 binding to ImportinβN (Figure S5, Movie S3, sequences given in Supplemental Experimental Procedure). However, force field inaccuracies and limited sampling prohibit the reliable extraction of an association rate, and we therefore studied the interaction further through fluorescence stopped-flow experiments (FSF) and Brownian dynamics (BD) simulations.

17Stopped-flow kinetics monitoring fluorescence anisotropy (r) can be used to study binding mechanisms and measure the association rate (kon) between proteins. The binding of Importinβ to Nup153FG site-specifically labeled with Cy3B elicits detectable changes in r, due to slowed rotational motion. Since Nup153FGPxFG has only a very small overall binding affinity toward Importinβ, we could not detect a sufficiently strong signal change in the anisotropy measurements in the tested and experimentally feasible concentration range (Figure S7). Thus, for FSF, we used fluorescently labeled full-length Nup153FG. We performed rapid mixing experiments under pseudo-first order conditions in “physiological” transport buffer. A monoexponential function does not describe well the observed anisotropy changes in Figure 5 (Figure S7 and Table S2). This is likely a result of having multiple different binding motifs and/or the ability of multiple Importinβ to engage into binding a single Nup, which adds another level of complexity (multivalency). A biexponential equation is able to describe the kinetics, resulting in two kobs per Importinβ concentration (Figures 5A, 5B, and S7). The fluorescence anisotropy at the end of the reaction was used to calculate the apparent Kd,app (Figure 5C). Remarkably, by performing experiments at multiple NTR concentrations we extracted an ultrafast kon,ultrafast = 1.5·109 M−1s−1 (Figure 5B) for the major component (average amplitude of 70%), while the second component was still very fast, with a kon,fast = 6.1·107 M−1s−1 at room temperature. These FSF measurements report on overall formation of the Nup153FG⋅ Importinβ complex i.e., one or more F binding. While we provide all results and further analysis details in Figure S7 and Table S2, for later discussion we focus on the fastest measured kon,ultrafast.

18We next estimated association rates from BD simulations, which compared to MD permit larger statistical sampling, at the cost of freezing the internal dynamics of the binding partners. Upon successful complex formation, starting from the conformations obtained from MD, the association rate was estimated (Figure S7) to be around 109 M−1 s−1 (Figure 5E), in agreement with stopped-flow measurements.

19BD simulations carried out without the contribution of apolar desolvation generated a drastic decrease of the estimated kon,BD by around two orders of magnitude, while the absence of electrostatic interactions had a negligible effect (Figures 5E and S7, and Table S2D and S2E). These observations complement our evidence for an association mainly favored by the energetic gain of sequestrating F residues from the solvent and burying them into the ImportinβN binding pockets.

20While desolvation effects cannot easily be tested experimentally, high ionic strength buffers can be used to shield long-range electrostatic interactions. We thus performed a salt titration ranging from 0.05 to 1 M ionic strength (using NaCl), permitting an estimate of kon under infinite electrostatic shielding by extrapolation using a Debye-Hückel-like approximation (Figures 5D and S7 and Table S2B). In line with the BD simulations, we obtained a kon,elect off of 2.9·108 M−1s−1, i.e., binding remains very fast even under electrostatic shielding.

21Additional stopped-flow measurements probing different Nup153FG regions (FxFG-, PxFG-rich) with diverse NTRs (NTF2, TRN1, Importinβ) are shown in Figure S7 and Table S2C. In all cases, we observed similar remarkably fast kinetics yielding consistent results for kon > 5·108 M−1s−1.

22Previously, solid phase binding assays indicated that the Importinβ double mutant (I178D/Y255A, termed ImportinβDA) has a more than 60-fold lower Kd for binding to full-length Nup153FG as compared to Importinβ WT. kon,BD dropped by only 40% compared to Importinβ WT, which we confirmed by experimental FSF studies (drop of kon,FSF by 30%, Figure 5). However, fluorescence anisotropy measurements revealed an ImportinβDA titration curve (Figure 5C) that confirms altered binding as compared to Importinβ WT, as e.g., due to an increase in koff.

23The efficiency of an NTR to bring cargo across the NPC barrier can be assayed using standard NPC transport assays. In these assays, a fluorescent cargo (NLS-MBP-eGFP) recognized by the Importinβ transport machinery is incubated with permeabilized cells in the presence of a functional transport system and the resulting nuclear fluorescence is measured. In line with the previously reported lower Kd of ImportinβDA, cargo accumulated slower compared to Importinβ WT measurements (Figures 6A and 6B) which can e.g., be due to a lower barrier crossing time, a reduced docking efficiency to the NPC or cargo release from the NPC for example.

24A prediction from our kinetic analysis is that the actual speed of barrier crossing, which involves several binding and unbinding steps between NTR and FG repeats should be rather similar for WT and mutant Importinβ, as changes in kon were small, and if at all, a higher koff for the mutant would make crossing even faster (see discussion).

25In contrast to the “bulk” transport assay, the speed of barrier crossing (characteristic crossing time) can be measured directly using single molecule (sm) tracking assays (Figure 6C), in which individual Importinβ molecules are fluorescently labeled and tracked while they cross from one side of the NPC to the other. This yielded a typical value of 6.9 ± 0.2 ms for Importinβ and 6.1 ± 0.5 ms for ImportinβDA for barrier crossing (Figures 6D and 6E). We note that this crossing time is near the sampling limit of our technology, and thus faster crossing times cannot easily be captured.

26The realization that many proteins are disordered has attracted considerable attention to the study of the molecular mechanisms controlling their interactions, including the role of disorder in promoting or facilitating binding. In particular, very little is known about the binding mechanisms involved in complex processes such as nucleocytoplasmic transport, where NTRs have to engage in multiple, specific binding and unbinding events while traversing a tens of nanometer thick permeability barrier.

27In this study, we have used a multidisciplinary approach to investigate the molecular mechanism underlying the interaction process between NTRs and Nups. In general, from our three core findings a coherent view emerges on how multiple rapid, yet specific protein interactions can be achieved.

28Based on our smFRET measurements, we found that Nup153FGPxFG resembles full-length Nup153FG with respect to its dynamics (Figures 2 and S2). Upon interaction with Importinβ, Nup153FGPxFG remains flexible, engaging with Importinβ only locally, as is evident from peak broadening in the respective 1H-15N HSQC spectra as well as R2 relaxation rates (Figures 3, S1, and S4). Local backbone sampling even of the interacting F was not measurably modified upon interaction. The conformers of Nup153FGPxFG that were subjected to ImportinβN binding in the MD simulations were also devoid of large-scale conformational changes, and interactions were only observed between individual surface exposed residues of Nup153FGPxFG and ImportinβN.

29It appears therefore that globally, the FG-Nup maintains its conformational ensemble as shown by smFRET. This observation is sound, as IDPs frequently use motif binding to engage with their binding partners. Our observation suggests an extraordinarily small motif (the side chain of F), which would be difficult to identify from large-scale bioinformatics approaches.

30The observed binding mode appears distinct from other single motif binding interactions, as well as from mechanisms that involve global conformational transitions, such as folding upon binding (Figure 7). The intrinsic flexibility of the Nup, the repeated occurrence and short length of the binding motif seem to create a highly reactive binding surface, which renders the individual FG-motifs prone to bind at any time without compromising the Nup’s inherent plasticity.

31The maximal association rate in the absence of electrostatic forces for a binary interaction system (in which all collisions are productive) can be approximated by the Einstein-Smoluchowski diffusion limit, which yields a theoretical kon of ∼109 M−1s−1 for the interaction of proteins of the size of Nup153FG and Importinβ.

32Very high association rates have been observed previously in the presence of long-range electrostatic attractions (108-1010 M−1s−1) for example for the barnase/barstar interaction, as well as for small IDP complexes studied by NMR. In the absence of electrostatic steering, this upper limit is typically never reached, as successful collisions require proper orientation of the binding partners. Consequently, most experimentally observed association rates at high salt concentrations fall into the regime of 104–106 M−1s−1.

33Our ensemble FSF kinetics (for Nup153FG) and BD simulations (for Nup153FGPxFG) show a kon of ∼109 M−1s−1 (Figure 5) supporting the aforementioned idea of a strongly reactive binding surface. We specifically observe an influence of apolar desolvation energies in the BD simulation and electrostatics are not found to play a major role in association. This applies apparently to both, Nup153FGPxFG, which is uncharged and was tested in BD, as well as Nup153FG, which has several charges in the N-terminal regions (Figures 5D and S2). Even in the limiting case of electrostatic shielding we found complex formation to still have a remarkably fast kon,FSF (Figures 5D, 5E and Table S2B).

34While experimentally bridging the gap between our molecular-level description of the small binary Nup⋅NTR complex (160 kDa) in solution to the actual in vivo transport mechanisms (involving ∼120 MDa NPCs) is still a challenging quest, the sm transport experiments (Figure 6) underline that the initially unexpected kinetic findings for the ImportinβDA mutant are in line with the finding in functional NPCs.

35According to ensemble titration fluorescence curves, we have observed an apparent local equilibrium constant (Kd,app) between Nup153FG and Importinβ in the nanomolar regime (Figure 5C and Table S2). However, we report millimolar affinities per FG-motif from our NMR measurements within Nup153FGPxFG (Figures 3 and S4), in line with a recent computational model. Our NMR studies further suggest that individual FG-motifs bind independently of each other, as the 15N R2 rates are similar to the values of the unbound Nup between the FG-repeats. Nevertheless, the sum of FG-motifs influences the effective binding strength of individual FGs to Importinβ, as can be seen by comparing the effective Kd for F1374 in the WT and the Nup153AGPxAG, F1374 mutant (Figure 3, S1, and S4).

36While these estimates of Kd values (from NMR and ensemble fluorescence) were measured on different Nup constructs, they also report on two different properties: the binding of Importinβ to a larger region of Nup153FG (fluorescence anisotropy) and to a single FG-motif (NMR), and illustrate an important characteristic of the system, namely the importance of polyvalent interactions, which is exploited also by other transport receptors (Figure S4). While an individual FG-motif might be unlikely to be bound, the chances that at least one FG-motif within the Nup molecule is bound may remain high. This stabilizing effect of multivalency/polyvalency is well known, and is even used as a design principle in enhancing the affinity of ligand interactions with multi-site targets where ligands are connected in tandem via short linkers. Stability enhancements achieved in such experiments can approach four-to-five orders of magnitude and are primarily due to substantial decreases in the global dissociation rate, i.e., in a multivalent system the molecules only separate as a result of a dissociation event if all other motifs are unbound.

37To demonstrate generality of these three core findings, we performed additional smFRET, FCS (Figures S2 and S3), NMR (Figure S4), MD (Figure S5 and Movie S3), and FSF experiments (Figure S7 and Table S2C) on a variety of different Nups from human and yeast, including the most common motif in vertebrates (FxFG) and the crucial GLFG sequence in yeast, for a diverse set of NTRs (NTF2, TRN1, CRM1, Importinβ). All results are in close agreement, highlighting the universal nature of the observed mechanism.

38Currently, several models are discussed on how a permeability barrier in the NPC can be built; among those are the selective phase, the brush, the reduction of dimensionality and the karyopherin centric model, etc., as well as mixtures of those. These models vary mainly over how FG-Nups are arranged and potentially interlinked inside the NPC to create a tight barrier. However, common to all these models is that the concentration of FG-repeats of about 50 mM creates a very crowded environment, which is roughly in line with stoichiometric measurements of Nups and the overall size of the central channel. Independently of the transport model assumed, mobility of an NTR inside the barrier is thus largely limited by the koff and kon of the interaction between FG-Nups and NTRs. This is also the case if FGs interact with FGs inside the barrier as proposed in the selective phase model, as long as these interactions are highly dynamic and do not pose a substantial energetic barrier or rate-limiting step to be melted. That we do not observe obvious FG-FG interactions in our studies is thus not necessarily inconsistent with such a model.

39If we were to naively consider the characteristic time for a single Nup and Importinβ to separate based on commonly measured fast kon and affinities (e.g., Kd (Nup⋅NTR) ∼100 nM and kon ∼106 M−1s−1 → unbinding time (UT) ∼100 ms), it appears impossible that Importinβ could cross a 50 mM FG-filled pore within 5 ms. This is the previously described “transport paradox,” in which high specificity is somehow coupled with rapid transport.

40Our work (down to picosecond and atomic resolution) is largely compatible with the existing barrier models, as it addresses on a molecular mechanistic level how an NTR could rapidly pass through a dense barrier. Using a simple model of a bivalent system, we already expect an order of magnitude difference between the dissociation rate for an individual motif and that for the whole protein. We have also observed extremely rapid association rates (∼109 M−1s−1) and in Supplemental Experimental Procedures (two toy models) we outline that if we consider a very rough estimate for the characteristic time for an individual motif unbinding event (UT∼1 μs) for full-length Nup153 (>24 valencies), it becomes clear that the Importinβ could “creep” through the dense FG-motif plug of the pore within the short transport time. Such movement is consistent with our (Figure 6) and other NTR diffusion studies through NPCs in intact cells and various model systems.

41In this case, nature has achieved a combination of high specificity with fast interaction rates. This is based on many individual low-affinity motifs paired with a binding mode that requires relatively little energy or time investment for the Nup to transit between free and bound conformations, and provides a rationale for the fast, yet specific, nuclear transport. While rapid binding can in principle be realized between proteins of single binding elements (e.g., driven by strong electrostatics), the proofreading emanating from the multiplicity and rapid repetition of many such events is what contributes to specific transport.

42We note that the transport paradox goes far beyond the relevance for the transport mechanism, since transient, but targeted interactions are central to the emerging view of highly dynamic protein (and other biomolecular) interaction networks. Furthermore, FG-repeats are also present in stress and P granules. It seems likely that such ultrafast binding mechanisms are also important for other biological recognition processes, where individual interaction motifs only make weak contributions, as e.g., in the recognition of glycans, or other very short linear motifs, like WG motifs in small RNA pathways, or binding of proteins to epigenetic marks, like many histone modifications.

43In addition, ultrafast association is achieved by using the unique plasticity of multivalent disordered proteins, which is distinct from mechanisms where orientation specific binding is required for complex formation. This represents an additional biological advantage for IDPs in comparison to folded proteins, and might have further facilitated their enrichment in organisms of higher complexity.

44The proteins were purified essentially as described in following routine column chromatography and then transferred into the respective measurement buffers. Labelling of Nup153FG (amino acids 875 to 1475 of the full length Nup153; numbering with respect to the full length protein as in ‘UniProt: P49790’) was performed using routine procedures to introduce Alexa488 as a donor and Alexa594 as an acceptor dye for smFRET experiments (and analog for other dyes), as described in.

45Spectral assignments of 13C, 15N Nup153FGPxFG were obtained from a set of BEST-TROSY-type triple resonance spectra: HNCO, intra-residue HN(CA)CO, HN(CO)CA, intra-residue HNCA, HN(COCA)CB, and intra-residue HN(CA)CB. For the measurements of RDCs, 13C, 15N Nup153FGPxFG was aligned in 12 mg/ml Pf1 phages yielding a D2O splitting of 2.16 Hz. RDCs were measured using BEST-type HNCO and HN(CO)CA experiments. 15N relaxation dispersion was carried out at Nup153FGPxFG/Importinβ concentrations of 250 μM and 180 μM, respectively, applying CPMG frequencies between 25 and 1,000 Hz. All experiments were performed in Na-phosphate buffer (pH 6), 150 mM NaCl, 2 mM DTT, 5 mM MgCl2, at 25°C and at a 1H frequency of 600 MHz if not noted otherwise.

46The conformational space available to disordered Nup153FGPxFG was sampled using the Flexible-meccano statistical coil description and representative ensembles in agreement with experimental chemical shifts were selected using ASTEROIDS and the ensemble was subsequently cross-validated against experimental RDCs and SAXS.

47SmFRET measurements of dual labeled freely diffusing proteins were performed on a confocal geometry detecting donor and acceptor intensities (from which the FRET efficiency EFRET is calculated) as well as fluorescence lifetimes (τ) on a custom built multiparameter setup as previously described.

48The association kinetics were monitored by following the fluorescence anisotropy change of Nup153FG labeled at the indicated position with Cy3B (see sequences in Supplemental Experimental Procedures) upon binding to different concentrations of NTRs, under pseudo-first order conditions. Anisotropy (r) was calculated from fluorescence intensities measured with polarizing filters in the parallel (‖) and perpendicular (⊥) position.

49Each trace was obtained by averaging ≥30 traces and background fluorescence was then subtracted. The anisotropy traces where fit with a biexponential function to determine kobs. The different kobs were plotted against the respective NTR concentrations and were linearly fit to obtain the association constant (kon) from the slope.

50The used BioLogic (Grenoble, France) stopped-flow equipment permits automatic titration and repeated technical replicates, which typically yield a small standard deviation. We derived an experimental error of ∼20% in kon measurements between different biological replicates. To be conservative, we thus do not show (the typically lower) standard deviations from technical replicates.

51Routine reconstitution of the nucleocytoplasmic transport machinery in permeabilized cells was used and fluorescence cargo (NLS-MBP-eGFP) was imaged on a confocal microscope (Leica, Mannheim) at the indicated time points.

52For single molecule tracking of NTRs, the same assay was used, but Importinβ-Alexa488 at single molecule concentration was tracked with an acquisition time of 2ms on a previously described home built imaging microscope.

53All data analyses for FSF, FCS, smFRET and tracking were performed with custom written routines in IgorPro (Wavemetrics, OR).

54The Nup153FGPxFG fragment was modeled on the basis of its sequence that also included the exogenously inserted residues used for labeling of the fragment with fluorophores. For the binding simulations, Nup153FGPxFG or Nup153FGFxFG were randomly placed in a box of dimensions 15 × 15 × 15 nm3 together with the N-terminal segment of ImportinβN (PDB: 1F59). Brownian Dynamics (BD) simulations were performed starting from the MD complex that showed a specific association between the partners, and resembled the crystallographic binding pose as reported by ref..

55S.M., D.M., I.V.A., designed and performed experiments, analyzed data and co-wrote the manuscript. M.R.J., N.B., C.K., S.T., J.C. provided additional reagents and analysis tools. S.L.S. designed experiments and analysis methods and co-wrote the manuscript. M.B., F.G. and E.A.L. conceived the project and co-wrote the manuscript.