Recognition Elements in the Histone H3 and H4 Tails for Seven Different Importins*

N-terminal tails of histones H3 and H4 are known to bind several different Importins to import the histones into the cell nucleus. However, it is not known what binding elements in the histone tails are recognized by the individual Importins. Biochemical studies of H3 and H4 tails binding to seven Importins, Impβ, Kapβ2, Imp4, Imp5, Imp7, Imp9, and Impα, show the H3 tail binding more tightly than the H4 tail. The H3 tail binds Kapβ2 and Imp5 with KD values of 77 and 57 nm, respectively, and binds the other five Importins more weakly. Mutagenic analysis shows H3 tail residues 11–27 to be the sole binding segment for Impβ, Kapβ2, and Imp4. However, Imp5, Imp7, Imp9, and Impα bind two separate elements in the H3 tail: the segment at residues 11–27 and an isoleucine-lysine nuclear localization signal (IK-NLS) motif at residues 35–40. The H4 tail also uses either one or two basic segments to bind the same set of Importins with a similar trend of relative affinities as the H3 tail, albeit at least 10-fold weaker. Of the many lysine residues in the H3 and H4 tails, only acetylation of the H3 Lys14 substantially decreased binding to several Importins. Lastly, we show that, in addition to the N-terminal tails, the histone fold domains of H3 and H4 and/or the histone chaperone Asf1b are important for Importin-histone recognition.

Importins bind and transport H3 and H4 into the nucleus (20 -25). There are a total of at least 10 different Importins in human cells (26,27). Co-immunoprecipitation, in vitro binding with recombinant proteins, and nuclear localization studies in yeast and permeabilized HeLa cells showed that several Importins can bind and import H3 and H4. These Importins are yeast Kap95, Kap104, Kap123, and Kap121; their human homologs Imp␤, Kap␤2, Imp4, and Imp5; and three additional human Importins, Imp7, Imp9, and the Importin adaptor Imp␣ (12-17, 28, 29). Although multiple Importins can bind and import H3 and H4, Imp4 is consistently the most abundant Importin that co-purifies with the histones, suggesting that it is the major/primary nuclear importer of H3/H4 in human cells (17,28,30). Similarly, the homolog of Imp4 in yeast, Kap123, is also the most abundant Importin that co-purifies with H3 and H4 from yeast cytosol (16).
Histones H3 and H4 each consist of an N-terminal disordered tail region followed by a globular histone fold domain (1, 2) (Fig. 1A). Previous studies showed that N-terminal tails of histones H3 and H4 are necessary and sufficient for nuclear import, consistent with the presence of NLS-like sequences in the tails. Removal of either the H3 or H4 tail does not prevent nuclear import, but simultaneous removal of both tails produced non-viable S. cerevisiae and caused loss of nuclear H3-H4 in Physarum polycephalum (12,31). Of the Importins that bind and import H3 and H4, only classes of NLSs for Imp␣/Imp␤, Kap␤2, and Imp5 are known. Imp␣ binds directly to the classical NLS (c-NLS), Kap␤2 binds the entirely distinct PY-NLS, and Imp5 recognizes a short lysine-rich NLS named the IK-NLS (also distinct from c-NLS) (24,26,27). Classical NLSs contain either one or two clusters of basic residues (consensus sequences K(K/R)X(K/R) or (K/R)(K/R)X 10 -12 (K/R)3/5 where X is any amino acid and (K/R)3/5 is three lysines or arginines in five consecutive residues) (32)(33)(34). The PY-NLS is defined by loose sequence motifs (N-terminal hydrophobic or basic motifs and a C-terminal (R/K/H)X 2-5 PY motif), structural disorder, and an overall basic charge (35). The IK-NLS is defined by the consensus motif K(V/I)XKX 1-2 (K/H/R) (36). Examination of sequences in the H3 and H4 tails revealed no recognizable c-NLS or PY-NLS (35,37). A previous report suggested that residues 35-40 of the H3 tail resemble an Imp5-specific IK-NLS (36). Classes of NLS that bind Imp4, Imp7, and Imp9 have not yet been defined. 20 -30% of cytoplasmic H3 histones are acetylated at Lys 14 and/or Lys 18 , and all cytoplasmic H4 are acetylated at both Lys 5 and Lys 12 (30). However, the effect of H3 and H4 tail acetylation on histone import is controversial. Mutations of H3 and H4 tail lysines to glutamines (acetylation mimics) in yeast abolished nuclear accumulation, slowed growth, or caused loss of viability, suggesting that acetylation impairs nuclear import (12). In contrast, H3 and H4 acetylation in P. polycephalum led to increased nuclear accumulation, and acetylated H4 tail peptides were reported to bind Imp4 better than unacetylated peptides, suggesting that acetylation may promote nuclear import (17,31).
Here, we biochemically map binding determinants in the H3 and H4 tails for Imp␤, Kap␤2, Imp4, Imp5, Imp7, Imp9, and Imp␣. Structural analysis revealed that Kap␤2 binds residues 11-27 of the H3 tail, which resemble a basic PY-NLS that is missing its PY epitope (19). This basic segment of H3 is also important for binding Imp␤, Imp4, Imp5, Imp7, Imp9, and Imp␣. In addition, an IK-NLS-like motif at H3 residues 35-40 is used to bind Imp5, Imp7, Imp9, and Imp␣. The first 20 residues of the H4 tail, enriched in basic and glycine residues, interact with Imp␤, Kap␤2, Imp4, Imp7, and Imp9. The H4 tail also uses an IK-NLS-like motif to bind Imp5, Imp7, Imp9, and Imp␣. As we uncovered the Importin binding determinants, we also examined effects of histone tail acetylation on Importin interactions. Finally, we assembled a complex of the histone chaperone Asf1b bound to the full-length H3/H4 dimer and compared Importin interactions of this three-protein complex with those of the histone tails alone.
Dissociation constants (K D values) of Kap␤2-H3 tail interactions were previously measured by isothermal titration calorimetry (ITC) (19). H3 tail binds Kap␤2 tightly with a K D of 77.1 nM, which is comparable with affinities of known Kap␤2-PY-NLS interactions (35, 38 -42). Here, we measured the K D values of MBP-H3 tail and MBP-H4 tail binding to Imp5 and Kap␤2 by ITC. MBP-H3 tail binds Imp5 with a K D of 57.2 (43.9, 89.2) ( Table 1; numbers in parentheses represent the 68.3% confidence intervals on K D as calculated using F-statistics and errorsurface projection method). The H4 tail binds Kap␤2 and Imp5 ϳ10-fold more weakly (H4 tail-Kap␤2, K D ϭ 871 nM (737.3, 924.5); H4 tail-Imp5, K D ϭ 619 nM (506.4, 692.3)). To verify that the N-terminal MBP tag has no effect on Importin binding, we first showed by ITC that MBP alone does not bind Imp5 (supplemental Fig. 6). ITC analysis of Imp5 and the H3 tail peptide (residues 1-47 with no MBP tag) produces a K D of 60.4 nM, which is very similar to the K D of 57 nM for Imp5 binding to MBP-H3 tail.
We were not able to measure K D values of histone tail binding to Imp␤, Imp4, Imp7, Imp9, or Imp␣ by either ITC, microscale thermophoresis, or fluorescence anisotropy. The Importins aggregate in conditions for these biophysical experiments. For example, the Importins aggregated as a result of stirring in the ITC cell due to heating during microscale thermophoresis experiments and at higher titration concentrations necessary for the fluorescence anisotropy experiments. Therefore, band densities for bound Importins and histone tails in pulldown binding assays were measured, and their ratios were compared in a histogram to estimate relative strengths of H3 tail and H4 tail binding to Imp␤, Kap␤2, Imp4, Imp5, Imp7, Imp9, and Imp␣ (Fig. 1B). Neither immobilization to glutathione-Sepharose or dimerization of the GST tag in GST-H3 tail has any effect on Importin binding as the K D values obtained by titrating Imp5 onto immobilized GST-H3 tail via pulldown binding assays (apparent K D ϭ 66 nM; supplemental Fig. 3) and by ITC of Imp5 with GST-H3 tail (K D ϭ 47.1 nM; Table 1 and supplemental Fig.  6) are similar to the K D obtained by ITC for the monomeric MBP-H3 tail ( Table 1).
Comparison of bound Importin bands suggests that Importin-H3 tail interactions can be roughly divided into two groups: 1) Kap␤2 and Imp5, known from ITC to bind tightly (K D values Ͻ100 nM), and 2) Imp␤, Imp4, Imp7, Imp9, and Imp␣, which all seem to bind more weakly. Apparent K D values estimated by fitting pulldown titration data for Imp␣, Imp␤, Imp4, Imp7, and Imp9 (ϳ150 -500 nM) are consistent with the Importin-H3 tail affinity trend shown in Fig. 1B (supplemental Fig. 3). Imp␣ binds the histone tails more weakly than the SV40 c-NLS (Fig.  1C). The Importin-H4 tail affinity trend is similar, but H4 tail binds at least 10-fold more weakly than H3 tail as we compare K D values for Kap␤2 and Imp5. H4 tail binds most strongly to Imp5 (K D ϭ 619 nM (506.4, 692.3)) and more weakly to the other six Importins. We validated the Importin-histone tail affinity trends by performing pulldown assays similar to those in Fig. 1B but using 1 ⁄ 10 the amount of proteins (gel stained with SYPRO Ruby protein stain; supplemental Fig. 4).
Interactions of H3 Tail Residues 11-27 with Importins-A crystal structure of Kap␤2 bound to the entire H3 tail showed only H3 residues 11-27 bound to the PY-NLS binding site of the Importin (19). H3 residues 11 TGGKAPRKN 19 bind in extended conformation and contribute a significant portion of the total binding energy (Lys 14 is a hot spot for binding Kap␤2), whereas residues 20 LATKAARK 27 form an ␣-helix (19). We used a series of H3 tail mutants, ITC, and qualitative pulldown binding assays to examine the contributions of H3 residues 11-27 to the binding of different Importins. The large majority of H3 tail side chains that contact Kap␤2 are from basic residues (19). Interestingly, interactions of the H3 tail with the other six Importins also appear to be dominated by electrostatic interactions as observed by the salt dependence of binding strengths (supplemental Fig. 5). Therefore, we mutated individual basic residues and all basic residues within H3 residues 11-27 and examined effects of the mutations on Importin binding by ITC (Imp5 (Table 1) and Kap␤2 (19)) and by pulldown binding assays ( Fig. 2A). Significance of changes in Importin binding due to H3 tail mutations were assessed by performing one-way ANOVA tests and t tests on the raw Importin/histone ratios ( Fig. 2A). Single site mutant H3 tail(K14A) shows large decreases in binding to Imp␤, Kap␤2, Imp4, Imp7, and Imp9 but does not affect Imp␣ and Imp5 binding. Single mutations of Arg 17 and Lys 23 moderately decrease binding to Imp␤, Kap␤2 (R17A only), Imp4 (R17A only), Imp7, and Imp9 but do not affect Imp␣ and Imp5 binding. In contrast, single mutations of Lys 18 , Arg 26 , and of Lys 27 show little to no effect in binding any of the Importins.
In summary, the basic segment that spans H3 residues 11-27 is important for binding Imp␤, Kap␤2, Imp4, Imp5, Imp7, Imp9, and Imp␣. Here, residue Lys 14 is a hot spot for binding all the Importins except for Imp5 and Imp␣. This basic segment is likely not the sole binding element for Imp5.
Mutations of the H3 Tail and Nuclear Localization in Cells-Nuclear localization of EYFP 2 -H3 tail proteins (wild type and mutants) was examined in live HT1080 cells using a spinning disk confocal microscope (Fig. 3D). Like the SV40 NLS, wild-type H3 tail localizes exclusively to the nucleus. However, the H3 tail(K14A/K17A/K18A/K23A/R26A/K27A) and H3 tail(K14/K17/K18/K23/R26/K27/ 35 VKKPHR 40 to Ala) mutants localize to the cytoplasm. The single site H3(K14A) mutant shows decreased nuclear accumulation, probably due to decreased binding to most but not all Importins as the mutant protein is still able to bind Imp␣ and Imp5.
Similar pulldown assays were performed with H3(28 -135) and H4 tail(21-102), which lack tail basic epitopes that bind Importins (Fig. 6, B and C). Removal of the basic epitopes decreased binding to all seven Importins, but the overall binding affinity trend remains similar to that for the Asf1b(1-169)bound full-length H3/H4 dimer. In summary, along with the basic epitopes in the N-terminal tails, the histone fold domains of H3 and H4 and/or the histone chaperone Asf1b is important for Importin-histone recognition and Importin specificity.

Discussion
We have mapped sequence elements in H3 and H4 tails that bind Imp␤, Kap␤2, Imp4, Imp5, Imp7, Imp9, and Imp␣. The results inform on both the NLS organization in the histone tails and general sequence elements that bind different Importins.
Importin-Binding Epitopes in H3 and H4 Tails-The basic segment in H3 residues 11-27 contributes most of the binding energy for interactions with Imp␤, Kap␤2, and Imp4, but interactions with Imp5, Imp7, Imp9, and Imp␣ involve an additional downstream IK-NLS-like motif. The H4 tail also uses either one or two basic regions to bind multiple Importins. Imp␤, Kap␤2, Imp4, Imp7, Imp9, and Imp␣ bind the basic segment between H4 residues 5 and 20, whereas Imp5 binds an additional IK-NLS-life motif.
Nuclear import cargoes usually have specific NLSs that bind one of the 10 Importins (26). However, two classes of highly abundant proteins, histones and ribosomal proteins, are exceptions as they can bind multiple Importins (12-16, 28, 29, 43). For example, the ribosomal protein L23A (rpL23A) binds Imp␣/␤, Kap␤2, Imp5, and Imp7 through a ␤-like import receptor binding (BIB) sequence (43). Ribosomal proteins like rpS7 and rpL5 also have BIB-like sequences and bind multiple Importins (43). BIBs were suggested to have originated from ancestral nuclear import signals prior to the divergence of Importins to gain specialized/distinct NLS binding sites (43). Although Importins have evolved to bind distinct signals, many of them may retain binding to these ancestral non-specialized BIB sequences. 59 KYPRKSAPRRNK 70 in the rpL23A BIB aligns with 14 KAPRKQLATKAAR 26 in the H3 tail and 8 KGLGKG-GAKRHR 20 in the H4 tail. Sequence similarity and the ability to bind multiple Importins suggest that the N-terminal basic segments of H3 and H4 tails may be ancestral NLSs like BIBs.
NLSs in the Histone Tails That bind Kap␤2, Imp␤, and Imp4 -Kap␤2, Imp␤, and Imp4 bind solely to the N-terminal basic segments of the H3 and H4 tails. Imp␤ is known to bind NLSs of very different lengths, sequences, and structural elements, but electrostatic contacts are common features in all these Imp␤-NLS interactions (24, 26, 27, 44 -47). Electrostatics are also important for H3 and H4 tails binding to Imp␤, but in the absence of Imp␤-histone tail structures, we cannot predict the NLS conformations or the locations of their binding sites on Imp␤. A Kap␤2-H3 tail structure shows that H3 residues 11-27 form a PY-NLS variant that is missing the canonical PY motif (19). No structural information is available at this time for how Imp4 binds cargoes/NLSs, but binding to only the N-terminal basic segments of the H3 and H4 tails suggests that NLSs for Imp4 may be generally compact and monopartite.
The second epitope of the H3 bipartite c-NLS is in the IK-NLS motif ( 35 VKKPHR 40 ) that is also used to bind Imp5. Structures of Kap121 (yeast Imp5) suggested that the IK-NLS is the most important binding element in Kap121 and Imp5 cargoes (36). Interestingly, our results suggest that IK-NLS motifs in H3 and H4 tails are only a portion of their NLSs for Imp5 with relatively minor contribution to binding energy. Furthermore, unlike IK-NLSs in Kap121 cargoes Pho4p, Spo12p, and binding partner Nup53p, which are specific for Kap121 (36), IK-NLS motifs in H3 and H4 are also part of bipartite c-NLSs that bind Imp␣. Finally, structures of Imp7 and Imp9 are not yet available, but our results suggest that both Importins can bind long sprawling NLSs with multiple basic epitopes.
Importins in cells (17,28). Asf1 binds the histone folds of the H3/H4 dimer, thus leaving H3 and H4 tails free to bind Importins (9, 10). Furthermore, although many different Importins can bind and import H3 and H4 tails, Imp4 and its yeast homolog Kap123 are consistently the most abundant Importins that co-purify with H3 and H4 in lysates (16,17,28,30). Our H3 and H4 tail binding data show no specificity for Imp4 over the other Importins. Imp4 is in fact one of the weakest binding Importins. The specificity of H3/H4 for Imp4 in cells must therefore lie outside of the histone tails, perhaps in the histone folds, and/or involves Asf1. Binding studies of Asf1b-H3/H4 showed that Imp4 is one of the strongest binders for the histone-chaperone complex, suggesting that the H3/H4 dimer histone fold and/or Asf1b must be important for Imp4 specificity. Kap123 is the most abundant Importin in budding yeast. Its abundance may be key in accomplishing a nuclear import rate that is 5-12-fold more rapid than other Importins (57). If Imp4 is similarly abundant in human cells, its high cellular concentration may be an advantage for Imp4-mediated H3/H4 import.
When Lys 14 (binding hot spot for Imp␤, Kap␤2, Imp4, Imp7, and Imp9) in the H3 tail is acetylated, binding to Imp␤, Kap␤2, Imp4, Imp5, Imp7, Imp9, and Imp␣ is decreased. Therefore, nuclear import of the small pool of cytoplasmic H3/H4 with acetylated H3 tail may be affected. In contrast, all new H4 is persistently acetylated prior to nuclear import. We showed that diacetylation of H4 tail Lys 5 and Lys 12 has little effect on binding to most Importins. It is important to note that unacetylated H4 tail binds Importins weakly (K D values Ͼ600 nM) and may contribute little toward nuclear import of the H3/H4 dimer. It is currently difficult to predict how acetylation of the full-length H3/H4 dimer or the Asf1b-H3/H4 complex affects Importin binding. Studies of Importin-histone complexes beyond histone tails will be important to resolve current controversies regarding histone acetylation and nuclear import.
For pulldown binding assays, bacteria expressing the GSThistone tail proteins were lysed by sonication and centrifuged. The supernatants were incubated with GSH-Sepharose followed by extensive washes with TB buffer containing 10% glycerol. Immobilized GST-histone tail proteins were stored in TB buffer containing 40% glycerol at Ϫ20°C.
To purify MBP-histone tail proteins, bacterial lysates were incubated with amylose beads (New England BioLabs), and the fusion proteins were eluted with buffer containing 20 mM Tris, pH 7.5, 50 mM NaCl, 2 mM EDTA, 2 mM DTT, 10% glycerol, and 10 mM maltose. Eluted proteins were further purified by ion exchange chromatography.
Pulldown Binding Assays-Purity of Importins for pulldown assays was verified by SDS-PAGE and Coomassie Blue staining (Fig. 1E). Normalization of Coomassie Blue staining and confirmation of linearity of staining were performed for GST-H3 tail (supplemental Fig. 2). Qualitative comparisons of Importinhistone tail binding in Fig. 1B were performed by incubating immobilized GST-H3 tail or GST-H4 tail on glutathione-Sepharose beads (20 nmol of GST-histone tail in each binding assay) with a 500 nM concentration of each purified Importin in TB buffer in a total volume of 800 l for 30 min at 4°C followed by extensive washing with the same buffer. Bound proteins were visualized using SDS-PAGE/Coomassie Blue. Gels were subject to densitometry analysis using ImageJ. The density of the Importin band was divided by the density of the GST-histone tail band in the same gel lane. Importin inputs were visualized by SDS-PAGE/Coomassie Blue to ensure that similar concentrations of Importins were used for all binding assays, and excess unbound Importins in the flow-through were also monitored.
All other pulldown binding assays were performed by incubating of 0.8 nmol immobilized GST-H3 tail or GST-H4 tail proteins (ϳ2 M) with an ϳ4 M concentration of each purified Importin in TB buffer in a total volume of 100 l for 30 min at 4°C followed by extensive washing with the same buffer. For RanGTP dissociations assays, ϳ10 M purified RanGTP was added to immobilized GST-H3 tail proteins that are bound to Importins followed by extensive washing. Activities of Imp␤, Kap␤2, Imp4, Imp5, Imp7, and Imp9 were verified by their binding to RanGTP, and the activity of Imp␣ was verified by binding to the classical NLS of the SV40 T antigen. Bound proteins were visualized using SDS-PAGE/Coomassie Blue. Gels were subjected to densitometry analysis using ImageJ. Density of the Importin band was divided by the density of the GSThistone tail band in the same gel lane. The density ratios were then normalized to the ratio of the Importin band over the wild-type GST-H3 tail band or wild-type GST-H4 tail band. Relative band densities of experiments performed in triplicate are plotted with standard errors in histograms generated with GraphPad Prism. One-way ANOVA and t tests were performed on all pulldown binding assay data.
K D values were also estimated from pulldown assays. 20 nM to 1 M concentrations of each Importin in total volumes of 0.4 -15 ml (to ensure a molar excess of Importins to the H3 tail) were titrated onto 0.2 nmol of GST-H3 tail immobilized on glutathione-Sepharose beads in a series of pulldown binding assays. Relative densities of the gel bands from three separate experiments were measured using ImageJ. The data were fitted to a simple bimolecular equilibrium relationship in GraphPad Prism to obtain the K D values.
Measuring Dissociation Constants with Isothermal Titration Calorimetry-Binding affinities of MBP-H3 tail proteins to Kap␤2 and Imp5 were measured using ITC. ITC experiments were performed with a Malvern ITC200 calorimeter (Malvern Instruments, Worcestershire, UK). Proteins were dialyzed against buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, 10% glycerol, and 2 mM ␤-mercaptoethanol. 200 -400 M MBP-H3 tail proteins were titrated into a sample cell containing 20 -40 M recombinant Kap␤2 or Imp5. ITC experiments were performed at 20°C with 19 rounds of 4-l injections. Data were plotted and analyzed using NITPIC and Sedphat, and the data were visualized using GUSSI. For error reporting, we used F-statistics and error-surface projection method to calculate the 68.3% confidence intervals of the fitted data (61). Histograms to compare K D values were generated by GraphPad Prism.
Nuclear-Cytoplasmic Localization of H3 Tail Proteins in Cells-Cellular localization of EYFP 2 -H3 tail fusion proteins overexpressed in HT1080 cells were observed as described previously (62). EYFP 2 -H3 tail expression constructs were cloned into the pEYFP 2 vector. Live cell images were collected using a spinning disk confocal microscope system (Nikon-Andor, Nikon, NY) and MetaMorph software. Image analysis was performed similarly with ImageJ. Experiments were performed in duplicates or triplicates with a total of Ͼ150 transfected cells.