Nuclear transport receptor binding avidity triggers a self-healing collapse transition in FG-nucleoporin molecular brushes

Abstract

Conformational changes at supramolecular interfaces are fundamentally coupled to binding activity, yet it remains a challenge to probe this relationship directly. Within the nuclear pore complex, this underlies how transport receptors known as karyopherins proceed through a tethered layer of intrinsically disordered nucleoporin domains containing Phe-Gly (FG)-rich repeats (FG domains) that otherwise hinder passive transport. Here, we use nonspecific proteins (i.e., BSA) as innate molecular probes to explore FG domain conformational changes by surface plasmon resonance. This mathematically diminishes the surface plasmon resonance refractive index constraint, thereby providing the means to acquire and correlate height changes in a surface-tethered FG domain layer to Kap binding affinities in situ with respect to their relative spatial arrangements. Stepwise measurements show that FG domain collapse is caused by karyopherin β1 (Kapβ1) binding at low concentrations, but this gradually transitions into a reextension at higher Kapβ1 concentrations. This ability to self-heal is intimately coupled to Kapβ1-FG binding avidity that promotes the maximal incorporation of Kapβ1 into the FG domain layer. Further increasing Kapβ1 to physiological concentrations leads to a “pileup” of Kapβ1 molecules that bind weakly to unoccupied FG repeats at the top of the layer. Therefore, binding avidity does not hinder fast transport per se. Revealing the biophysical basis underlying the form–function relationship of Kapβ1-FG domain behavior results in a convergent picture in which transport and mechanistic aspects of nuclear pore complex functionality are reconciled.

Intrinsically disordered proteins (IDPs) that adorn the surfaces of biomolecular structures are thought to confer a host of unique functionalities not found in structured proteins (1). However, unlike their free-floating counterparts in solution (2), the properties of such surface-tethered IDPs can be particularly challenging to evaluate because of their inherent flexibility and conformational susceptibility to local interfacial constraints (3). Herein lies the crux of the nuclear pore complex (NPC) problem, in which in vitro efforts (4–11) to rationalize the collective form–function characteristics of the intrinsically disordered Phe-Gly (FG) domains (12) typically neglect the uncertainty regarding their numbers [approximately 200 divided amongst 11 different FG-bearing nucleoporins (Nups) (13)], their locations within the central NPC channel, and corresponding distances between neighboring anchoring sites (14). As the key components of the NPC barrier mechanism, the manner by which the FG domains impede nonspecific molecules (greater than 40 kDa) whilst granting karyopherins (Kaps) and their cargoes access between the nucleus and the cytoplasm (15–17) is likely to be influenced by such physical constraints. Accordingly, it remains unaccounted for how these contextual details can influence (i) FG domain barrier conformation, (ii) Kap-FG binding avidity (18) [i.e., Kapβ1, also known as importinβ or impβ, has an estimated 10 FG binding sites (19, 20)], and (iii) subsequent binding-induced conformational changes in the FG domains (21).

As Paine et al. wrote in 1975, “as solute size approaches the dimensions of the (nuclear) pore, solute–pore wall interactions become increasingly important. Specific site interactions…would also influence solute movements” (22). Yet, the sheer molecular complexity of the NPC (23) has for the most part motivated reductionist approaches to tease apart FG domain function in vitro (4–11). Not surprisingly, inherent differences in experimental approach and length scale that largely overlook such “solute–pore wall interactions” have led to differing views on the matter. Briefly, the selective phase model (5, 8) derives from the characteristics of macroscopic FG hydrogels whereby the FG domains form a “self-healing” sieve-like meshwork that only Kaps can dissolve through. The polymer brush model (4) is based on nanoscale validation of how surface-tethered FG domains entropically exclude nonspecific cargoes (24) whilst promoting Kap access by reversibly collapsing (6). On this basis, it has been postulated that permanently collapsed FG domains at physiological Kap concentrations might provide a hydrophobic “FG-rich layer” around the NPC walls for the surface diffusion of Kap-cargo complexes—i.e., “reduction of dimensionality” (25). Finally, the trees and brushes model (26) proposes that a bimodal distribution of collapsed and extended FG domain regions within the NPC provides distinct transport routes for Kap-cargo complexes and passive diffusion, respectively. To add to the confusion, mechanistic and kinetic views of NPC transport also appear to be at odds. One example pertains to the collapse of Nup153 FG domain brushes upon binding Kapβ1 at picomolar concentrations (6), which has been interpreted (10) to imply a substantially stronger binding affinity over reported K_D values (approximately 10 nM) (27). Indeed, the incompatibility of in vitro–obtained Kapβ1-FG domain binding affinities (27–29) to describe in vivo transport rates questions even the relevance of known K_D measurements (30).

In this work, we sought to correlate the conformational changes of surface-tethered FG domains directly to multivalent Kapβ1-FG binding interactions (i.e., binding avidity) using a surface plasmon resonance (SPR)-based assay that we developed for this purpose. Importantly, this allows for an in situ measure of FG domain surface density, conformational height change, and Kap-FG binding activity. As an analytical biosensing technique, SPR monitors the change in the angle of incidence light required to create surface plasmon resonance as molecular binding events occur in real time in proximity to the glass/metal sensor surface (31–33). In this way, rate and equilibrium constants are determined as analytes from solution bind to surface-tethered molecules (i.e., ligands). Moreover, because the mass of surface-bound molecules can be quantified, SPR is poised for our intended purpose, except for measuring conformational changes, which is formidable because it requires knowing the dielectric property (i.e., refractive index) of the surface-tethered layer (as governed by its changing thickness) (34, 35).

Results

Measuring FG Domain Conformational Changes by SPR.

To circumvent the SPR refractive index constraint, we begin with a general expression for the effective refractive index n_eff at the interface, which is the weighted sum of local indices (34):

[1]

where n(z) is the refractive index at height z perpendicular to the sensor surface and l_d is the characteristic evanescent field decay length. For a surface-grafted molecular layer of mean thickness d, n(z) = n_a for 0 ≤ z ≤ d and n(z) = n_s for d < z < ∞ where the subscripts a and s correspond to “adlayer” and “solvent,” respectively. Based on this definition, Eq. 1 becomes:

[2]

In the presence of noninteracting molecules, n_s is replaced by n_p, giving:

[3]

assuming (i) there is a homogeneous distribution of noninteracting molecules in the solvent, and (ii) the layer itself is unaffected by the particles (i.e., noninteracting by definition). Subtracting Eqs. 2 and 3 gives:

[4]

where Δn_eff is the change in the effective bulk refractive index (noting that n_a is now eliminated). Further, because the SPR response to changes in the bulk solution refractive index (n) (where n_PBS = 1.33411 for PBS and n_BSA = 1.33567 for PBS/BSA; SI Text) is approximately linear over a restricted range (i.e., Δn = 0.01) in the absence of adsorption from solution, Δn_eff can be replaced by the term R/m (34). Here, R is the SPR response resulting from the noninteracting molecules and m is the slope that relates the change in the SPR response to changes in n. By measuring R and m, Eq. 4 can be solved for the thickness d of a molecular layer:

[5]

Now, if a reference cell is implemented in addition to the sample cell (with respective parameters defined by subscripts 1 and 2; SI Text), calculating d₂ - d₁ gives:

[6]

noting that all the refractive indices are canceled out. Here, d₁ corresponds to a passivation layer of known thickness in the reference cell [i.e., HS - (CH₂)₁₁ - (OCH₂CH₂)₃ - OH (36); henceforth, C₁₇H₃₆O₄S]. We note that Eq. 6 provides the mathematical basis by which the height of a molecular layer can be assessed without the refractive index constraint using l_d = 350 nm (SI Text). Therefore, any subsequent height change Δd is computed by subtracting the initial layer height given as d₂(initial) from each measured d₂.

Fig. 1 describes the methodology of the assay. First, C₁₇H₃₆O₄S and the cysteine-modified N-terminal FG domain of human Nup62 (amino acids 1–240, 15 FG repeats; henceforth, cNup62) are covalently grafted via thiol binding to cell1 (reference) and cell2 (sample) in the SPR system, respectively. Nonspecific interactions are prevented by filling any exposed gold sites in cell2 with C₁₇H₃₆O₄S (37). R₁ and R₂ are the SPR response signals [in resonance units (RU)] that result from injecting BSA into cell1 and cell2, respectively. BSA molecules are almost entirely excluded from NPCs (38), which makes them ideal as innate molecular probes with minimal external influence on the FG domains. We further determined that m₂/m₁ is approximately 1, given the negligible difference in SPR sensitivity between cell1 and cell2 (SI Text). In accordance with Eq. 6, d₂(initial) can then be computed for cNup62 using the known value of d₁ = 2 nm for the C₁₇H₃₆O₄S layer (36). Moreover, it is useful to correlate d₂(initial) to the mean cNup62 grafting distance g_cNup62, which is estimated from the shift in the immobilization baseline ΔRU as shown in Fig. 1 (39, 40) (i.e., where 1,300 RU = 1 ng/mm²; SI Text). Stepwise height changes Δd in the cNup62 layer can thereafter be obtained by injecting BSA at each respective Kapβ1 concentration (c_Kapβ1). Likewise, any change in the surface density of bound Kapβ1 (ρ_Kapβ1) and the mean next-neighbor distance of Kapβ1 (g_Kapβ1) can be determined from subsequent changes in ΔRU.

Fig. 1.

Measuring binding activity and conformational height changes in situ by SPR. R₁ and R₂ that derive from the presence of innate noninteracting BSA probes (red) in cell 1 (C₁₇H₃₆O₄S) and cell 2 (cNup62), respectively, are used to calculate d₂ in Eq. 6. Stepwise changes Δd that follow Kapβ1 (green) binding are obtained by subtracting d₂(initial) from each d₂. For instance, Δd_final corresponds to the overall height difference between d₂(initial) and d₂(final)—i.e., the first and last BSA injections. g_cNup62 is obtained from the initial shift in the immobilization baseline ΔRU. Likewise, g_Kapβ1 and ρ_Kapβ1 are obtained from subsequent changes in ΔRU.

Kapβ1-FG Binding Causes a Nonmonotonic Collapse Transition in cNup62.

Fig. 2A shows a representative SPR measurement where d₂(initial) = 14.1 nm for g_cNup62 = 2.4 nm. Given that d₂(initial) > σ_cNup62 and g_cNup62 < σ_cNup62, where σ_cNup62 (= 8.5 nm) is the hydrodynamic size of cNup62, indicates that the FG domains form an extended molecular brush (4). We then calculated Δd, ρ_Kapβ1, and g_Kapβ1 for 16 titrations of c_Kapβ1 increasing from 0.1 nM to 13.4 μM. Striking nonmonotonic phase behavior emerges when Δd is plotted against ρ_Kapβ1 (Fig. 2B): (1) Up to c_Kapβ1 = 40 nM, Δd declines sharply (i.e., negative height change), reaching a minimum at ρ_Kapβ1 = 29.9 Da/nm² (g_Kapβ1 = 55.5 nm); (2) Δd undergoes a gradual increase that crosses over Δd = 0 at ρ_Kapβ1 = 1,010 Da/nm² (g_Kapβ1 = 9.8 nm) when c_Kapβ1 = 4 μM; and (3) Δd increases steadily (i.e., positive height change) until ρ_Kapβ1 = 1,442.6 Da/nm² (g_Kapβ1 = 8.2 nm) at the maximum c_Kapβ1 = 13.4 μM. Further evidence of these transitions can be drawn from correlations with Kapβ1-FG binding activity. In Fig. 2C, the quality of a single-component Langmuir isotherm fit (χ²) to R_eq deteriorates once c_Kapβ1 is increased past 4 μM, where K_D is approximately 400 nM. Conversely, χ² is minimized by a two-component fit (SI Text) over the entire c_Kapβ1 range, giving K_D1 = 347 nM and K_D2 = 95.9 μM, which suggests that a low-affinity binding phase emerges at higher c_Kapβ1 values. For comparison, a single-component fit giving K_D = 1.28 μM appropriately describes Kapβ1-FG binding to sparse non–brush-like cNup62 “mushrooms” [Fig. 2C, Inset; g_cNup62 = 11.0 nm and d₂(initial) = 2.5 nm].

Fig. 2.

Nonmonotonic behavior in a cNup62 brush caused by Kapβ1-FG binding. (A) Successive 3·30 s BSA injections follow 16 c_Kapβ1 titrations ranging from 0.1 nM to 13.4 μM on a cNup62 brush characterized by g_cNup62 = 2.4 nm and d₂(initial) = 14.1 nm. (B) The cNup62 brush undergoes collapse at low ρ_Kapβ1 (1), followed by recovery (2), which reaches pileup (3) upon crossing Δd equal 0. Included are the values of g_Kapβ1 and c_Kapβ1 (in parentheses) that correspond to each respective Δd measurement. (C) The steady-state (R_eq) SPR response across the entire c_Kapβ1 range (from A; 0.1 nM to 13.4 μM) is optimally fit using a two-component Langmuir isotherm (green) giving K_D1 = 347 nM and K_D2 = 95.9 μM. For single fits (K_D of approximately 400 nM), χ² is minimized at low terminal c_Kapβ1 values (grey, purple, and red) but deviates past c_Kapβ1 > 4 μM (blue and pink). Solid and dashed lines denote the actual fitted c_Kapβ1 range and the predicted K_D behavior, respectively. (Inset) A single K_D = 1.28 μM is found for Kapβ1 binding to sparse cNup62 mushrooms where g_cNup62 = 11.0 nm and d₂(initial) = 2.5 nm.

Although Δd accounts for an ensemble average of local height changes, the following qualitative outcomes can be rationalized from Gedankenexperiment (“thought experiment”; illustrated in SI Text). For phase 1, the following scenarios can be eliminated: (i) cNup62 accommodates Kapβ1 (Δd = 0); (ii) cNup62 engulfs Kapβ1 and swells (Δd > 0); and (iii) Kapβ1 stays “perched” on cNup62 (Δd > 0). As depicted in Fig. 3, the steep negative decline (Δd < 0) is caused by a local collapse of cNup62 around Kapβ1 due to multivalent Kapβ1-FG interactions. In phase 2, the “recovery” in Δd is a consequence of in-layer steric crowding as caused by a further addition of Kapβ1, which rearranges the FG domains into more entropy-favoring conformations. Subsequent cross-over occurs (Δd → 0; c_Kapβ1 = 4 μM) when ρ_Kapβ1 = 1,010 Da/nm², which closely approximates the expected surface density of a packed Kapβ1 monolayer (approximately 1,000 Da/nm²) [from small angle X-ray scattering data (41); SI Text]. Referring to Fig. 2C, K_D1 = 347 nM is relatively strong up until this point owing to maximal Kapβ1-FG binding within the cNup62 layer. However, correlating Δd > 0 and K_D2 = 95.9 μM in phase 3 indicates the formation of a weakly bound secondary “pileup” layer when excess Kapβ1 binds to unoccupied FG domain regions that protrude from the cNup62 layer.

Fig. 3.

Kapβ1-FG binding activity and cNup62 form–function are intimately coupled. (1) A local collapse of cNup62 occurs around Kapβ1 owing to strong multivalent Kapβ1-FG (dark green) binding at low ρ_Kapβ1. (2) Additional Kapβ1 molecules bind tightly in the cNup62 layer, driving unoccupied FG domains to extend or recover because of increasing in-layer steric repulsion whereupon the layer self-heals, reaching Δd = 0. (3) At high ρ_Kapβ1, a secondary layer of Kapβ1 (light green) binds weakly to unoccupied FG domain protrusions giving Δd > 0. Red dashed lines correspond to the cNup62 layer height as measured by BSA (red watermarked).

In Fig. 4A, the results of Δd vs. ρ_Kapβ1 obtained from cNup62 brushes with different g_cNup62 and d₂(initial) indicate that the collapse transition is a common feature during initial Kapβ1 binding. This is followed by a recovery phase with taller brushes requiring more Kapβ1 molecules (higher ρ_Kapβ1) to reach pileup. Recalling that ρ_Kapβ1 for a Kapβ1 monolayer is approximately 1,000 Da/nm² indicates that taller brushes [d₂(initial) > 14.1 nm] accommodate a secondary Kapβ1 layer to recover. For comparison, sparser mushroom-like cNup62 layers undergo a negligible collapse and reach pileup without recovering. A three-dimensional spatial description is shown in Fig. 4B, where the change in total mass-volume density Δυ (i.e., cNup62 and Kapβ1) is plotted against relative height change Δd/d₂(initial). During collapse, the linear increase in Δυ is dominated by a compaction of cNup62 because only small amounts of Kapβ1 are bound. Interestingly, the overlap indicates that Δυ scales with Δd/d₂(initial), that is, the total amount of space occupied is equally optimized within different cNup62 layers regardless of their initial brush conformation or amount of bound Kapβ1. During the initial stages of recovery, Δυ increases at constant Δd/d₂(initial) where the void volume of each layer is being filled with additional Kapβ1. Upon reaching pileup, Δυ approaches a saturated critical capacity that is maintained by increasing Δd/d₂(initial) (i.e., via FG domain rearrangements). While our interpretation is consistent with theoretical predictions (42), we note that pileup commences sooner for sparse cNup62 layers because of their isolation and lower capacity to bind Kapβ1.

Fig. 4.

Brushes collapse sparse layers do not. (A) Plot of Δd vs. ρ_Kapβ1, where the extent of collapse increases for taller cNup62 brushes (red > green > purple > grey) as compared to sparser layers (blue, pink). A greater amount of bound Kapβ1 is also required for taller brushes to recover before reaching pileup (red > green > purple > grey). Sparse cNup62 layers exhibit a negligible collapse followed by an immediate pileup without recovering (blue, pink). (B) Plot of the total (Kapβ1 and cNup62) mass-volume density change Δυ vs. relative height change Δd/d₂(initial). (1) For brushes (red, green, purple, grey), a linear increase in Δυ accompanies a 10% reduction in Δd/d₂(initial) because of cNup62 compaction upon collapse. Their overlap reveals that the total space occupied scales with the extent of collapse and is conserved. (2) The transition into recovery at Δυ of approximately 20 Da/nm³ proceeds with additional Kapβ1 binding without changing Δd/d₂(initial). Saturation at Δυ of approximately 70 Da/nm³ denotes FG domain reextension to maintain its capacity to accommodate more Kapβ1, marking the commencement of (3) pileup. Sparser conformations (blue, pink) have a low Kapβ1 capacity, and pile up at low Δυ without recovering.

Kapβ1-FG Binding Avidity Depends on cNup62 Conformation.

Fig. 5A summarizes the dependence of d₂(initial) on g_cNup62. Clearly, extended molecular brushes form at small g_cNup62, and transition into sparser layers or mushrooms at large g_cNup62. Because cNup62 (pI = 9.31) is net positively charged at pH 7.2, we deduce that this behavior is polyelectrolytic in nature (i.e., forming polyelectrolyte brushes), as suggested by Flory–Huggins theory (43) (SI Text). The corresponding plot of K_D vs. g_cNup62 in Fig. 5B reveals how nonmonotonic behavior is linked to Kapβ1-FG binding avidity. When g_cNup62 > σ_cNup62, single K_D values of approximately 10 μM reflect the limited propensity of individual cNup62 mushrooms to bind Kapβ1. This appears to split at g_cNup62 < σ_cNup62, where two binding constants (K_D1 and K_D2) emerge becoming more apparent at low g_cNup62 (Fig. 2C) because of the onset of brush formation. At low to moderate c_Kapβ1, strong binding (K_D1 of approximately 0.2 μM) accompanies collapse and recovery where Kapβ1 has access to FG repeats residing amongst neighboring FG domains, thereby reaching a maximum (K_D1 decreases) at small g_cNup62. This is consistent with prevailing sub-μM K_D values, noting that the highest Kap concentrations tested were below 1 μM (27–29). At large c_Kapβ1, however, in-layer steric crowding and a reduction of unoccupied FG repeats give rise to weaker binding (K_D2 ranging from 10 μM to 1 mM) that is associated with pileup. The large variation in K_D is therefore a hallmark of binding avidity that emerges from the myriad of Kapβ1-FG binding possibilities that derive from the inherent flexibility and conformational susceptibility of surface-tethered FG domains.

Fig. 5.

Brush height and Kapβ1 binding avidity are correlated via cNup62 grafting distance. (A) Dependence of d₂(initial) on g_cNup62, showing that cNup62 forms a molecular brush at low g_cNup62 (i.e., high surface grafting density) and transitions towards sparse mushrooms at high g_cNup62. A fit of the Flory–Huggins equation to d₂(initial) suggests that cNup62 is polyelectrolytic in nature (SI Text). (B) Kapβ1 binding affinity to cNup62 is modulated by g_cNup62. An intermediate single binding phase occurs at g_cNup62 larger than σ_cNup62 (= 8.5 nm; dotted line) because of the limited Kapβ1 binding capacity of sparse mushrooms. This splits at low g_cNup62 (i.e., in the brush regime), where strong binding to cNup62 (K_D1; dark green) occurs at low to moderate c_Kapβ1 (collapse and recovery), whereas weak binding (K_D2; light green) occurs at large c_Kapβ1 (pileup).

Discussion

We have shown that self-healing nonmonotonic FG domain behavior is intimately coupled to Kapβ1-FG binding activity as defined by their relative spatial arrangements (i.e., g_cNup62 and g_Kapβ1). This results from a competition between FG domain collapse (caused by multivalent Kapβ1-FG binding) and FG domain reextension, which maximizes the capacity of the layer to bind more Kapβ1. Supposing that only sparse FG domain mushrooms existed in the pore, one might expect an increase in passive transport owing to a reduction in barrier functionality; counterintuitively, however, selective transport could slow down because of a comparatively high Kapβ1-FG binding affinity (K_D of approximately 10 μM; Fig. 5B). Instead, our findings support a view where crowding is not only important for selectivity (44), but also essential for promoting fast Kapβ1 transport in the NPC (45). Indeed, the high FG domain surface density (small g_cNup62) data, which bears a close resemblance to the NPC (where up to 128 copies of Nup62 may be present; ref. 46), predicts that at least two Kapβ1 binding phases exist at physiological concentrations (c_Kapβ1 of approximately 10 μM; ref. 47): (i) strong binding (K_D1) amongst a population of semicollapsed FG domains at the pore walls; and (ii) weak Kapβ1 binding (K_D2) to unoccupied FG domain protrusions near the pore center. As follows, it is the weak binding phase in a Kapβ1-crowded pore that is key to promoting fast transport rates.

Altogether this is reminiscent of a “highway” effect, where Kap transport is slow at the pore walls but fast near the pore center (Fig. 6) as can be inferred from two-phase binding in NPC transport studies (48). More striking evidence can be found from the single molecule fluorescence studies of Ma et al. in terms of the preferred location of Kapβ1 along the NPC walls that leaves a narrow passage at the pore center for passive diffusion to proceed (49). The transition from more collapsed FG domain segments nearer the walls to unoccupied protrusions toward the center may certainly contribute to the inhomogeneous, viscous characteristics of the central channel (50). Nevertheless, understanding the collective FG domain response in the NPC will require an evaluation of the surface density for each different FG Nup and the effects of Kapβ1 binding avidity. The highway effect might also explain how increasing c_Kapβ1 sharply decreases NPC interaction time, thereby improving import efficiency (45). Without precluding the effect of weakly binding competitors (30), this might explain how in vivo NPC transport is fast despite strong binding avidity in vitro. It is noteworthy that the K_D2 measurements lie in close agreement with the range of weak μM to mM affinities anticipated to describe known NPC transport rates (i.e., approximately 10 ms) (30). Thus, binding avidity need not hinder fast transport per se.

Fig. 6.

The NPC transport highway where Kapβ1 traffic can proceed via at least two “lanes” at physiological concentrations. Slow transport is anticipated for strongly bound Kapβ1 molecules (dark green) that saturate semicollapsed FG domains around the pore walls. Fast transport occurs nearer the pore center, where Kapβ1 binds weakly to unoccupied FG domain protrusions (light green). Small passive molecules (red watermarked) may diffuse freely through the pore center.

Finally, our findings reconcile the key features postulated by different NPC models. While entropic exclusion rejects nonbinding molecules (4, 24), nonmonotonic behavior signifies that the FG domains are not permanently collapsed but undergo dynamic rearrangements during Kapβ1 transport (6). Importantly, this imparts a self-healing mechanism on surface-tethered FG domains in the NPC at the nanometer scale without requiring for hydrophobic FG cross-linking as argued from the basis of bulk FG hydrogels that take over several micrometers to reseal (8). Hence, one may consider the population of strongly bound Kapβ1 as integral constituents of the NPC (25). Nevertheless, the occurrence of the weak binding phase does bring into question the role of RanGTP in dissociating Kapβ1 from the FG domains (18). To clarify, we have also ascertained that Kapβ1 does not bind covalently to the underlying gold SPR surface (SI Text), thereby disputing allegations (10) that the FG domain collapse constitutes an in vitro artifact. Methodological differences aside, the mismatch in Kapβ1 concentrations may explain why FG domain collapse was observed for Nup153 brushes (c_Kapβ1 ≤ 33 nM) (6), but not for brushes of Nsp1 (c_Kapβ1≥200 nM) (10). In the future, experimentation ought to involve stepwise height measurements spanning from low-nM to approximately 10 μM Kapβ1 concentrations. On a related note, it should be instructive that Kap binding activity cannot be rationalized (10) from conformational FG domain behavior alone.

To conclude, we have uncovered Kapβ1-FG domain behavior that reconciles transport and mechanistic aspects of NPC functionality. Such insight can contribute to the functional design and optimization of biomimetic selective channels and nanopores (9, 11, 51). On a technical note, our SPR methodology affords the correlation of binding affinities, in-plane molecular arrangements, and conformational changes in situ. This can be powerful in resolving the form–function relationships of diverse surface-tethered IDPs (52–54) and other stimuli-responsive polymers (55, 56) on biological interfaces.

Methods

Cloning and Expression of Recombinant cNup62 and Kapβ1.

A comprehensive description of the following protocols can be found in ref. 11. Briefly, the N-terminal FG repeat domain of human Nup62 (amino acids 1–240) was subcloned by GenScript Inc. into pPEP-TEV vector at the BamHI and SalI restriction sites. One cysteine was added to its C terminus (Cys-Nup62) as a covalent tether to Au. The recombinant N-terminal His₆-tagged cNup62 and Kapβ1 were expressed in Escherichia coli BL21 (DE3) cells. The final protein purity was analyzed by SDS/PAGE (SI Text), and selected fractions were dialyzed against PBS (pH 7.2; Invitrogen) for further use.

Other Materials.

Ten mg/mL BSA (Sigma–Aldrich) was carefully dissolved in PBS; C₁₇H₃₆O₄S (Nanoscience) was dissolved until reaching 10 mM in ethanol and diluted with PBS to 1 mM before experimentation.

SPR Sensor Chip Preparation.

SPR bare gold sensor chips (SIA Kit Au) were from GE Healthcare. Upon removal from storage in an argon atmosphere, gold sensor surfaces were ultrasonicated in acetone and high-purity ethanol (Merck) for 15 min, respectively, and dried in a nitrogen gas stream followed by 60 min UVO cleaning (Model 42A-220; Jelight Company Inc.). The gold sensor surfaces were then ultrasonicated for another 15 min in ethanol, dried in a nitrogen gas stream, and mounted on the sample holder for immediate SPR usage. A comprehensive description of the SPR measurement protocol with error analysis can be found in SI Text.

Dynamic Light Scattering.

Hydrodynamic diameter measurements of Kapβ1 and cNup62 were made in PBS with the addition of 1 mM DTT using a Zetasizer Nano instrument (Malvern). This gave σ_h = 8.47 ± 0.45 nm (polydispersity index = 0.36 ± 0.08) for cNup62 and σ_h = 12.06 ± 2.09 nm (polydispersity index = 0.423 ± 0.19) for Kapβ1, using n = 1.45 and n = 1.330 as the refractive index for proteins and dispersant [i.e., water; t = 25.0 °C, viscosity = 0.8872 cP (1P = 0.1 Pa•s)], respectively.