Comparative study of the interactions between fungal transcription factor nuclear localization sequences with mammalian and fungal importin-alpha

Importin-α (Impα) is an adaptor protein that binds to cargo proteins (containing Nuclear Localization Sequences - NLSs), for their translocation to the nucleus. The specificities of the Impα/NLS interactions have been studied, since these features could be used as important tools to find potential NLSs in nuclear proteins or even for the development of targets to inhibit nuclear import or to design peptides for drug delivery. Few structural studies have compared different Impα variants from the same organism or Impα of different organisms. Previously, we investigated nuclear transport of transcription factors with Neurospora crassa Impα (NcImpα). Herein, NIT-2 and PAC-3 transcription factors NLSs were studied in complex with Mus musculus Impα (MmImpα). Calorimetric assays demonstrated that the PAC-3 NLS peptide interacts with both Impα proteins with approximately the same affinity. The NIT-2 NLS sequence binds with high affinity to the Impα major binding site from both organisms, but its binding to minor binding sites reveals interesting differences due to the presence of additional interactions of NIT-2-NLS with MmImpα. These findings, together with previous results with Impα from other organisms, indicate that the differential affinity of NLSs to minor binding sites may be also responsible for the selectivity of some cargo proteins recognition and transport.

Binding of NIT-2 NLS to the MmImpα. The crystal structure of MmImpα/NIT-2 NLS presented two fragments of NIT-2 NLS peptide bound to major and minor sites, which is similar to several monopartite NLS-MmImpα structures 7,53 . Electron density is present for seven peptide residues ( 917 SSKRQRR 923 ) at the major NLS-binding site, bound at positions P0-P6 of MmImpα. The peptide presents an average B-factor of 56.6 Å 2 (the average B-factor for the entire Impα is 43.4 Å 2 ) (Fig. 1). The residues bound to the core of the major NLS-binding site (residues 919-922; positions P2-P5) have average B-factors (53.6 Å 2 ) comparable to Impα. All these residues (919-922) present charged interactions between their side-chains and Impα side-chain residues (Fig. 3A).
Electron density is also present for six peptide residues ( 918 SKRQRR 923 ) at the minor NLS-binding site, bound at the positions P0′-P5′ of the MmImpα. The peptide presented an average B-factor of 43.0 Å 2 (the average B-factor for entire Impα is 43.4 Å 2 ) (Fig. 1). The residues bound to the core of the minor NLS-binding site (residues 919-922; positions P1′-P4′) have lower average B-factors (36.6 Å 2 ) compared to Impα. K919, R920 and R923 residues (positions P1′, P2′ and P5′) present charged interactions between their side-chains and Impα side-chain residues (Fig. 3A). The superposition of Cα atoms between NIT-2 and SV40 NLS peptides yields an RMSD of 1.02 Å for the major binding site (positions P1-P5) and 0.57 Å for the minor binding site (positions P1′-P4′). Interestingly, these values are higher than in previous comparisons with SV40 NLS 53 , reflecting the high structural variability at the N-and C-termini of both NIT-2 NLS peptides in comparison with SV40 NLS peptide (Fig. 4).
Binding of PAC-3 NLS to the MmImpα. The crystal structure of MmImpα/PAC-3 NLS presented two fragments of PAC-3 NLS peptide bound to major and minor sites. Electron density is present for seven peptide residues ( 297 SVKRRQI 303 ) at the major NLS-binding site, bound at positions P0-P6 of MmImpα. The peptide presents an average B-factor of 53.0 Å 2 (the average B-factor for entire Impα is 46.8 Å 2 ) (Fig. 2). The residues bound to the core of the major NLS-binding site (residues 299-302; positions P2-P5) have lower average B-factors (45.7 Å 2 ) compared to Impα. All these residues (299-302) present charged interactions between their side-chains and Impα side-chain residues (Fig. 3B).
Electron density is present for five peptide residues ( 298 AKRRA 302 ) at the minor NLS-binding site, bound at the positions P0′-P4′ of the MmImpα. The KRR residues were modeled at the positions P1′-P3′ and Ala residues were modeled at positions P0′ and P4′ due to the lack of electron densities for their side chains (Fig. 2). The residues bound to the core of the minor NLS-binding site (residues KRR; positions P1′-P3′) have higher average B-factors (61.0 Å 2 ) compared to Impα. These residues (positions P1′-P3′) present interactions between their side-chains and Impα side-chain residues (Fig. 3B). The superposition of Cα atoms between PAC-3 and SV40 NLS peptides yields an RMSD of 0.78 Å for the major binding site (positions P1-P5) and 0.31 Å for the minor binding site (positions P1′-P4′). Similar to NIT-2 NLS and SV40 NLS comparison (previous section), the comparison between PAC-3 and SV40 NLS peptides yielded higher than other equivalent comparisons 53 with SV40 NLS. As seen in Fig. 4, NIT-2, PAC-3 and SV40 NLS in both sites display high structural variability at the N-and C-termini.  53 and MmImpα/ P4(R) and P4(M) NLSs, PDB ID: 5KLR, 5KLT 9 ) resulted in an average RMSD of 0.3 Å. This low value reflects the high structural conservation of the MmImpα, which is independent of the NLS peptide bound to them. In contrast, a similar superposition between MmImpα/NIT-2 NLS and NcImpα/NIT-2 NLS resulted an RMSD of 5.07 Å. This RMSD difference is the result of a more concave structure of the NcImpα compared to MmImpα, as previously observed 41 , which belong to different Impα clades. Interestingly, despite the structural differences between NcImpα and MmImpα structures, the NIT-2 NLS peptide binds to NLS-binding sites with the exact same residues at each position of MmImpα and NcImpα (Fig. 3). The superposition of Cα atoms between NIT-2 NLSs from MmImpα and NcImpα structures yields an RMSD of 0.17 Å for the major binding site (positions P1-P5) and 0.51 Å for the minor binding site (positions P1′-P4′). The comparison of the NIT-2-NLS binding at the major site of MmImpα and NcImpα reveals that the contacts are very conserved, with the equivalent residues of both Impα proteins making contacts with NIT-2 NLS peptides. The same comparison for the minor binding site reveals interesting differences related to positions P3′ and P4′. While NIT-2 NLS interacts with N283 and G281 at position P3′ and with E354 N319 and R315 at position P4′ from MmImpα, no important interaction is observed between the NIT-2 NLS side-chain at positions P3′ and P4′ and NcImpα (Fig. 3). The structural data for the major and minor binding sites from both MmImpα and NcImpα are fully in agreement with the affinity assays (next section).

Calorimetric assays for the binding of NIT-2 and PAC-3 NLSs and MmImpα.
Representative thermograms of calorimetric titrations for both complexes are shown in Fig. 5. Binding isotherms for NIT-2 and PAC-3 NLS peptides and the Impα receptor were best fitted with a nonlinear regression model of two nonidentical and independent binding sites or one binding site. The data processing revealed that two NIT-2 NLS peptides bind to Impα, but only one PAC-3 NLS peptide binds to Impα. For NIT-2 NLS, the dissociation constant (K d ) was in the submicromolar range (~0.1 mM) and attributed to the major binding site, and the other constant corresponding to a 10-fold lower affinity was attributed to the minor binding site. In the case of the PAC-3 NLS, the K d for the only binding site was also in the submicromolar range. Enthalpic parameters (ΔH) for all assays showed favorable enthalpic values: −8.09 ± 0.29 (NIT-2 NLS, major binding), −3.74 ± 0.14 (PAC-3 NLS) and −0,44 ± 0,19 (NIT-2 NLS, minor binding).
In addition, aiming to further understand the binding of the PAC-3 NLS to MmImpα, two mutated peptides (N and C-termini mutated basic clusters) were tested by ITC using the same experimental conditions employed by PAC-3 NLS: i) 281 FDAAAAQFDDLNDFFGSVKRRQIN 304 and ii) 281 FDARKRQFDDLNDFFGSVAAAQIN 304 . ITC assays revealed that both mutated peptides present no measured interaction with Impα receptor (Suppl. Fig. 1), showing that the presence of both basic clusters are necessary for the PAC-3 NLS binding to Impα. www.nature.com/scientificreports www.nature.com/scientificreports/ The comparison between MmImpα/NIT-2 NLS and NcImpα/NIT-2 NLS 51 calorimetric assays reveals that K d values for the major binding site are exactly the same and thus are compatible with the conservation of residue interactions (Fig. 3). The same comparison for the minor binding site reveals that NIT-2 NLS has a higher affinity for MmImpα/NIT-2 NLS, which is also compatible with the higher number of interactions observed in the MmImpα/NIT-2 NLS structure compared to the NcImpα/NIT-2 NLS structure (Fig. 3). The comparison between MmImpα/PAC-3 NLS and NcImpα/PAC-3 NLS 52 calorimetric assays reveals that their K d values are the same considering the experimental error.

Discussion
comparison of monopartite nLSs binding to mammalian and fungal impα. More than 120 crystal structures of Impα have been solved since 1998 (S. cerevisiae Impα, PDB ID 1BK5 37 ) followed by the first mammalian Impα (M. musculus, PDB ID 1IAL 22 ) and cocrystallized Impα with NLS peptides 7,36 . Most of the Impα structures deposited in the Protein Data Bank are MmImpα complexed to NLS peptides from several organisms 10 but also synthetic NLS peptides 8 and small molecules 57 . In addition, H. sapiens Impα variants [12][13][14]16,48,57 , O. sativa 33,39 , A. thaliana 40 and N. crassa 41 structures were also solved. The analysis of these structures clearly demonstrates that the overall Impα structures are highly conserved among them 8,41 , and only their solenoid curvatures may vary, particularly between proteins from different phylogenetic families 58 . However, few structural studies have compared different Impα variants from the same organism or Impα of different organisms.
A study with different human Impα variants complexed to influenza A PB2 NLS 15 identified important differences among the variants: the Impα3 variant is more flexible than other variants; the Impα1 variant has the strongest autoinhibition and Impα 3 has the weakest inhibition. Two comparative studies between Impα from different organisms have also been performed 33,39,41 . OsImpα was solved complexed to the prototypical monopartite NLS from SV40 and with two synthetic "plant-specific" NLSs 41 . NcImpα was also solved complexed to SV40 NLS. Interestingly, the binding of the SV40 NLS to the major-binding sites from OsImpα, NcImpα and MmImpα were very similar 33,39,41 . The binding to the minor-binding site is shifted one position for OsImpα and NcImpα compared to MmImpα and presents some different interactions, particularly for the N-and C-termini of the peptide. Structural comparison and multiple alignment of Impα proteins show that some residues of the region near the minor site (Armadillo repeats 8 and 9) present in NcImpα (S402, E493 and K497), OsImpα (S394, E480 and K484) are not conserved in MmImpα (T402, S483, A487). These substitutions may prevent the binding of particular residues of NLS peptides to an Impα by steric hindrance or may cause different interactions of a particular peptide with different Impα proteins.  (Fig. 4). In addition, for the major binding site, the interactions for both peptides and receptors are conserved (Fig. 3). However, for the minor binding site, different interactions occur for the different receptors and NLS peptides (Fig. 3). These differences are related to specific sequential differences between both receptors, as previously reported 33,39,41 , but also with the shifted position of the SV40 NLS bound to MmImpα 7 that also present alternative binding modes to this site 7,24 . These alternative binding modes are likely related to the high content of sequential K/R residues of the SV40 NLS. The calorimetric assays described in this report (Table 2) are in agreement with the structural studies of these four complexes (NcImpα/NIT-2 NLS, MmImpα/NIT-2 NLS, NcImpα/SV40 NLS, MmImpα/SV40 NLS). The K d values are on the same order of magnitude for the major binding site (0.56, 0.56, 0.89, and 1.8 µM for the same complexes) and present a higher variation for the minor binding site (9.9, 5.7, 1.7, 23 µM).
NIT-2 NLS peptide binds to MmImpα and NcImpα with similar conformations at major and minor NLS binding sites according to the MmImpα/NIT-2 NLS and NcImpα/NIT-2 NLS crystal structures. Indeed, the ITC assays are completely in agreement with the structural data, which for the major binding site, the K d value is exactly the same for both proteins, and for the minor binding site, the K d value is on the same order of magnitude.  7 . Positions binding to the major (P 1 -P 5 ) and minor binding sites (P 1 ′-P 4 ′) are identified along the chains. This figure was generated using PyMOL v.1.8.6 64 program. www.nature.com/scientificreports www.nature.com/scientificreports/ However, a deeper analysis of NIT-2 binding to minor binding sites of these receptors reveals interesting differences. The presence of additional interactions of NIT-2-NLS with MmImpα compared to NcImpα, particularly at positions P3′ and P4′, may explain the higher affinity of this peptide to MmImpα (Fig. 2). Interestingly, in contrast with a previous comparison of NcImpα and OsImpα with MmImpα 33,39,41 , in which nonconserved residues from these Impα interact differently with the N-terminus of NLS peptide, the present study observed different interactions in the C-terminal region of peptide. The NcImpα and MmImpα residues are conserved (D280, N283, R315, N319, E354) ( Fig. 3)(Suppl. Fig. 2); thus, the different interactions for the same NLS peptide may be related to different structural concavities of both Impα, as observed for the high RMSD when both structures are superposed (subsection: Comparison between MmImpα structures and MmImpα/NIT-2 NLS and NcImpα/ NIT-2 NLS structures). Therefore, we suggest that both the N-and C-termini are able to confer specificity to particular NLS sequences that are able to bind at minor binding sites.
Thus, the structural and calorimetric study with MmImpα complexed to NIT-2 NLS revealed that this peptide binds as classical monopartite NLS (consensus sequence: KK/RX(K/R) 4 ) to a mammalian Impα, similar to a fungal Impα 51 . The NIT-2 NLS peptide ( 917 SSKRQR 923 ) interacts with high affinity to major binding sites of both receptors with K d of the same order of magnitude (~0.1 µM) compared to other classical monopartite NLSs with high affinity to MmImpα 56 . As other classical monopartite NLS 10,56 , the NIT-2 NLS peptide is also able to interact with the minor binding site one order of magnitude weaker than the major binding site (~1 µM). For the minor binding site, different interactions between a particular NLS with MmImpα and NcImpα are observed.

What is the role of PAC-3 NLS in PAC-3 protein transport? PAC-3 is a transcription factor that is
translocated to the nucleus at alkaline pH stress in N. crassa 52 . Calorimetric assays with the putative PAC-3 NLS and NcImpα demonstrated that this NLS peptide has a strong affinity (0.39 µM, Table 2) to NcImpα with a stoichiometry of 1:1 52 . Taking into account the calorimetric results and that its sequence resembles a bipartite consensus sequence (KRX 10-12 K(K/R)X(K/R)), with the exception of the P5 position (K/R), the authors of this study hypothesized that this NLS region is responsible for the recognition of the PAC-3 transcription factor by Impα. Thus, these components may form a complex that permits PAC-3 to be translocated to the nucleus under specific conditions. However, the authors of this study were not able to crystallize this complex to obtain structural information to confirm this hypothesis.
In the present study, we used the same PAC-3 NLS peptide and performed equivalent calorimetric and crystallographic studies using MmImpα. The calorimetric study demonstrated that the PAC-3 NLS peptide interacts with MmImpα with approximately the K d value considering the experimental error and with the same stoichiometry of 1:1. Furthermore, the calorimetric study with mutated N and C-termini basic clusters of PAC-3 NLSs and MmImpα revealed interesting results. As both mutated NLS peptides were not able to bind to the protein, it is possible to conclude that both clusters are necessary for the interaction between PAC-3 and MmImpα, thus PAC-3 is a bipartite NLS, as previously suggested in the study with PAC-3 and NcImpα 52 .
The crystal structure of the MmImpα/PAC-3 NLS complex revealed that two fragments of the PAC-3 NLS peptide bind to MmImpα. In the major NLS-binding site, seven peptide residues ( 297 SVKRRQI 303 ) were unambiguously observed bound at positions P0-P6 of MmImpα. However, no electron density was found in the linker region and, for the minor NLS-binding site, electron density was presented for five residue main chains, but only for three side chains (KRR). This sequence is not compatible with the expected sequence for the minor NLS-binding site 285 KRQ 287 (P1′-P3′), because the electron density in the position P3′ is compatible with Arg side chain. The presence of electron density for only three side chains in the minor NLS-binding site with higher B-factors compared to the entire protein indicates low affinity of this region to the protein or peptide staggering. Side chain electron densities for the positions P0′, P4′ and P5′ are typical for other bipartite or monopartite NLSs, such as NIT-2 NLS presented here (Fig. 1B) and in other previous studies 23,24,59 . Taking into account ITC assays with PAC-3 NLS and the previous structural studies with Impα, we suggest that peptide staggering is occurring with N-terminal sequence of the peptide ( 281 FDARKRQF 288 ). The presence of an Arg residue preceding the Lys residue and the absence of a basic residue after KR residues that would bind at the position P3′, may explain this phenomenon. NLS peptide staggering has been previous observed for other complexes, particularly for the SV40 TAg NLS 7 which presents a basic residue preceding the KR residues.
The lack of electron density in the linker region for PAC-3 NLS is also an intriguing result obtained here. However, previous structural results with bipartite NLSs also presented this common characteristic, such as for CBP80 8 , PRP20 35 , and PB2 12,15 . Thus, some features seem to be important for the stabilization of bipartite NLSs, in  Table 2. Thermodynamic constants of Impα/NLS complexes interactions. Data obtained by ITC assays.
Thus, the structural and calorimetric study with MmImpα complexed to N. crassa PAC-3 NLS revealed that this peptide binds to mammalian Impα. Considering that Impα structures are highly conserved 10 and, particularly, that their major and minor NLS-binding sites are also strictly conserved, we suggest that the binding of Bipartite NLS Monopartite NLS Table 3. Binding of nuclear localization sequences to specific binding clusters in Mus musculus importin-α.
PAC-3 NLS to MmImpα and NcImpα is similar. Indeed, the similarity of NIT-2 NLS binding to both MmImpα and NcImpα also supports this supposition. Thus, the present study confirms the hypothesis proposed by Virgilio and colleagues 52 and enables us to understand the structural determinants for the interaction between the PAC-3 transcription factor and NcImpα and its translocation to the nucleus of this fungus.

conclusions
In the present work, NLSs from different N. crassa transcription factors (NIT-2 and PAC-3) were studied by structural and calorimetric techniques in complex with M. musculus Impα. The comparison of these data with previous results 51, 52 revealed remarkable similarity of the interaction between these sequences and N. crassa or M. musculus protein receptors. The NIT-2 NLS peptide binds as a classical monopartite NLS with high affinity to the Impα major binding site for both organisms. Calorimetric assays demonstrated that the PAC-3 NLS peptide interacts with Impα from both organisms with approximately the same affinity and stoichiometry indicating that it is a bipartite NLS. Since the main docking event occurs between the NIT-2 and PAC-3 NLSs and Impα at the major binding site, we hypothesized that the full-length NIT-2 and PAC-3 interact similarly with Impα from these two organisms.
The analyses of NIT-2 NLS minor binding sites of both Impα proteins reveal some particular interactions that corroborate the different affinity values obtained in this study. The higher affinity of N. crassa NIT-2 by MmImpα instead of NcImpα is an unexpected result, but strongly indicates that the major binding site is the site used for the translocation of NIT-2 protein to the nucleus. In contrast, the comparison between MmImpα/SV40 NLS and NcImpα/SV40 NLS revealed a higher affinity of the SV40 NLS for the minor binding site of NcImpα than for MmImpα 41 . A similar result was also observed for rice Impα 33 . In light of these results, we hypothesized that the differential affinity for NLSs at the minor site may be a useful strategy for organisms that only have one Impα isoform to selectively recognize and transport different NLSs. experimental procedures protein expression and purification. The gene encoding the protein Impα from M. musculus was cloned into the pET30a expression vector. Recombinant MmImpα was cloned with a histidine tag and as a truncated protein (70-529) to avoid autoinhibition 22 . The clones were provided by Dr. Bostjan Kobe from the University of Queensland (Australia). The plasmid was expressed in Escherichia coli host strain Rosetta (TM) pLYS (Novagen), and the recombinant protein was purified by affinity chromatography according to Barros et al., 2012 59 . The protein was eluted with a 0.0-0.15 M imidazole linear gradient, concentrated using an Amicon dispositive, and the buffer was changed to 20 mM Tris-HCl, pH 8.0 and 100 mM NaCl for storage. The purified protein was stored at cryogenic temperature. NLS peptides NIT-2-NLS ( 915 TISSKRQRRHSKS 927 ) and PAC-3-NLS ( 281 FDARKRQFDDLNDFFGSVKRRQIN 304 ) were synthesized by GenOne with 98% purity. isothermal titration calorimetry. MmImpα and NIT-2-NLS were diluted at 40 μM and 800 μM, respectively, in buffer containing 20 mM Tris-HCl, pH 8.0 and 100 mM NaCl. The samples were submitted to ITC experiments, performed with a MicroCal iTC200 microcalorimeter (GE Healthcare), where the peptide sample was titrated into the protein sample. The affinity data were obtained at 20 °C from 20 titrations of 2 μL, with 240 s of interval between each titration and 800 rpm homogenization speed. Experiments with MmImpα and PAC-3-NLS (native and mutated) were performed under similar conditions but with a protein/peptide proportion of 1:10. Further experiments with mutated PAC-3 NLS were also performed with a protein/peptide proportion of 1:20. Control experiments were performed by titration of the peptide sample into the buffer, and the data obtained were subtracted from the peptide:protein titrations. Data were processed using Origin 7.0 software (Microcal Software, Northampton, MA) to obtain the thermodynamic constants of the interactions 60 . crystallization and structure solution. The complexes MmImpα/PAC-3-NLS and MmImpα/NIT-2-NLS were submitted to crystallization experiments using similar conditions as previous MmImpα/NLS peptide complexes 24,32,59 . Crystallization drops containing 1.0 μL of protein (18 mg/mL) 0.5 μL of peptide (5 mg/ mL) and 0.5 μL of reservoir solution were mounted in hanging-drop plates and stored at 18 °C. Single crystals were obtained with reservoir solutions containing 0.55 M sodium citrate (pH 6), 1.6 M sodium citrate and 10 mM DTT after 7-14 days. Crystals obtained were submitted to X-ray diffraction at the Brazilian Synchrotron Light Source (LNLS) in Campinas-SP, Brazil. X-ray data collected were processed using XDS software 61 , and the structures were obtained by Fourier synthesis using MmImpα/Ku80-NLS as a template 32 and refined using PHENIX 62 . Modeling of the peptides were performed using Coot 63 . All structural figures were generated using PyMOL v.1.8.6 64 program.