N-terminal half of transportin SR2 interacts with HIV integrase

The karyopherin transportin SR2 (TRN-SR2, TNPO3) is responsible for shuttling specific cargoes such as serine/arginine-rich splicing factors from the cytoplasm to the nucleus. This protein plays a key role in HIV infection by facilitating the nuclear import of the pre-integration complex (PIC) that contains the viral DNA as well as several cellular and HIV proteins, including the integrase. The process of nuclear import is considered to be the bottleneck of the viral replication cycle and therefore represents a promising target for anti-HIV drug design. Previous studies have demonstrated that the direct interaction between TRN-SR2 and HIV integrase predominantly involves the catalytic core domain (CCD) and the C-terminal domain (CTD) of the integrase. We aimed at providing a detailed molecular view of this interaction through a biochemical characterization of the respective protein complex. Size-exclusion chromatography was used to characterize the interaction of TRN-SR2 with a truncated variant of the HIV-1 integrase, including both the CCD and CTD. These experiments indicate that one TRN-SR2 molecule can specifically bind one CCD-CTD dimer. Next, the regions of the solenoid-like TRN-SR2 molecule that are involved in the interaction with integrase were identified using AlphaScreen binding assays, revealing that the integrase interacts with the N-terminal half of TRN-SR2 principally through the HEAT repeats 4, 10, and 11. Combining these results with small-angle X-ray scattering data for the complex of TRN-SR2 with truncated integrase, we propose a molecular model of the complex. We speculate that nuclear import of the PIC may proceed concurrently with the normal nuclear transport.

During the last decades our understanding of HIV infection has considerably increased, and a wide range of drugs targeting different steps of the viral replication cycle has been developed. However, the current combination anti-retroviral therapy has significant limitations such as the inability to completely eradicate the virus, the emergence of drug-resistant strains, adverse effects, and high cost. For these reasons, the development of new anti-HIV drugs is still necessary (1).
HIV, a member of the superfamily Retroviridae, can infect both dividing and non-dividing cells (2). Unlike other retroviruses, such as Gammaretroviruses, which make use of mitosis to gain access to chromatin, HIV enters the nucleus in the form of the pre-integration complex (PIC) 2 by means of an active transport through the nuclear pore (3). The PIC contains several components, including the reverse-transcribed viral genome, a number of viral proteins such as matrix, nucleocapsid, capsid, viral protein R, and integrase (IN), as well as cellular co-factors. Activity of IN is essential for the proliferation of the virus, as this enzyme catalyzes the irreversible incorporation of the viral DNA into the host genome. IN belongs to the superfamily of DDE(D) nucleotidyltransferases (4). The enzyme contains three domains, the N-terminal domain, the catalytic core domain (CCD), and the C-terminal domain (CTD), all implicated in multimerization that is essential for the integration process. IN recognizes the blunt ends of the linearized doublestranded and reverse-transcribed viral DNA, resulting in a complex called the intasome (5). Apart from viral proteins, cellular proteins such as the barrier to autointegration factor 1 (BAF1), high-mobility group proteins, lamina-associated polypeptide 2␣ (LAP2␣), and lens epithelium-derived growth factor (LEDGF/p75) have been reported to be components of the PIC (6).
Both genome-wide siRNA screening as well as cell culture studies have identified TRN-SR2 as a co-factor of HIV replication, whereas the interaction between TRN-SR2 and HIV IN has been initially revealed through yeast two-hybrid studies (8,10,11). Subsequently, a range of experiments has indicated that TRN-SR2 is able to bind HIV integrase directly (8,15,(21)(22)(23)(24).
In addition, alternative hypotheses exist that postulate that TRN-SR2 engages in either direct (25,26) or indirect interaction with a capsid protein (20,27,28) upon the involvement of the polyadenylation specificity factor 6 (CPSF6) in the latter case.
TRN-SR2 is a karyopherin that belongs to the importin-␤ family and imports mostly serine/arginine (SR)-rich splicing factors to the nucleus, including the alternative splicing factor 1/pre-mRNA-splicing factor SF2 (ASF/SF2) and cleavage and polyadenylation-specific factor (CPSF6) (29,30). In addition, TRN-SR2 can also bind some non-SR proteins such as the RNA-binding motif protein 4 (RBM4) (31). TRN-SR2 consists of 20 consecutive Huntington, Elongation Factor 3, PR65/A, TOR (HEAT) repeats. Each repeat corresponds to two ␣-helices (denoted as helix a and helix b) connected by a loop and forming an ␣-hairpin (32,33). Structural studies of TRN-SR2 and other importins have revealed a characteristic solenoid-like overall structure of the molecule (28, 34 -37). Within each HEAT, the helix a is located on the outside and helix b on the inside of the solenoid.
It has been established that TRN-SR2 binds the substrate in the cytoplasmic compartment and subsequently docks to the nuclear pore complex, which enables its transport across the nuclear envelope. Nup358, a pivotal component of the nuclear pore complex located at the cytoplasmic side, has recently been attributed a role in transportin-mediated nuclear import (38). Once the TRN-SR2-based cargo complex reaches the nucleus, it interacts with Ras-related nuclear protein (Ran), the main regulatory co-factor of this machinery. A high nuclear concentration of RanGTP is maintained by the regulator of chromosome condensation 1 (RCC1) and the nuclear transport factor 2 (NTF2). The competitive binding of GTP-charged Ran induces cargo release (34). Previous mutagenesis and structural studies have shown that TRN-SR2 binds RanGTP via the N-terminal region, whereas the cargo attachment typically requires the arginine-rich helix 15b of TRN-SR2 (28,37). The TRN-SR2⅐RanGTP complex eventually returns to the cytoplasm where the RanGTPase-activating protein (RanGAP) together with the Ran-binding protein 1 (RanBP1) lead to the GTP hydrolysis of Ran, which in turn induces its release from TRN-SR2 (39). It was shown that, just like bona fide cargoes, HIV IN can be displaced from TRN-SR2 upon addition of RanGTP (40).
HIV depends on host factors at all stages of its replication cycle. Although viral proteins generally mutate quickly especially due to the error-prone activity of the reverse transcriptase, the interfaces between the viral and host proteins may not change considerably because the host proteins are practically invariable. Correspondingly, drugs targeting the virus/host interactions are much less likely to induce the selection of resistant strains. Fortunately, a number of interactions between HIV components and cellular factors have been discovered recently opening up new opportunities for the development of novel classes of antiretroviral agents (10,(41)(42)(43). Detailed understanding of the interaction between HIV IN and human TRN-SR2 at the molecular level is crucial toward a rational design of drugs that would target this interaction, which could lead to potent antivirals due to the bottleneck nature of HIV nuclear import.
Despite its importance, the complex formation between HIV IN and human TRN-SR2 has not yet been characterized in detail at the biochemical level. One reason is that the full-length IN progressively oligomerizes in solution to a variable extent (44,45), complicating quantitative in vitro studies. However, it was established previously that the interaction between TRN-SR2 and IN predominantly occurs via the CCD and CTD domains of the latter (21). Accordingly, next to the full-length protein, in this study we have employed an N-terminal truncation variant of integrase that includes the CCD and CTD domains. This variant forms dimers in solution (46). We have exploited size-exclusion chromatography (SEC) to investigate the interaction of TRN-SR2 with the CCD-CTD tandem and established the stoichiometry of the resulting complex. To pinpoint the specific interaction sites, we have used AlphaScreen binding assays involving the integrase and different truncated constructs of TRN-SR2 as well as short helical peptides thereof. Finally, we have employed small angle X-ray scattering (SAXS) to analyze the interaction between TRN-SR2 and IN in solution, which has yielded a 3D model of this molecular complex.

TRN-SR2 forms a stable complex with the CCD-CTD construct of HIV-1 integrase at a 1:2 molar ratio
We used an efficient expression and purification scheme based on producing the proteins of interest as His 6 -tagged small ubiquitin-like modifier (SUMO) fusions (47), followed by subtractive immobilized metal ion affinity chromatography and removal of the purification tag. Samples of highly pure recombinant human TRN-SR2 (103.9 kDa) as well as HIV-1 IN CCD-CTD tandem (24.7 kDa) could be obtained. We then used analytical SEC to systematically analyze the formation of the respective protein complex. The TRN-SR2 sample was run alone and pre-incubated with CCD-CTD in increasing molar ratios (1:1, 1:2, and 1:3). All runs reveal a single dominant peak that is increasing in intensity and somewhat shifting to the left as more CCD-CTD is being added (Fig. 1a). The elution position of the main peak obtained with a 2-fold excess of CCD-CTD corresponds to ϳ158 kDa according to a calibration by standard proteins, which is in good agreement with the theoretical value for the 1:2 complex (153 kDa). Moreover, although the mixtures containing a 1-or 2-fold molar excess of CCD-CTD do not show additional peaks in the chromatogram, the mixture with a 3-fold excess (as well as mixtures containing yet larger quantities of CCD-CTD, data not shown) reveals a distinct second peak at ϳ44 kDa, which corresponds to the uncomplexed CCD-CTD dimer. In line with this, for the 1:1 and 1:2 mixtures, electrophoretic analysis of the peak fraction reveals the same mass ratios of the two proteins as input, whereas for the 1:3 mix the peak fraction still appears close to the 1:2 ratio (Fig. 1b). Taking these observations together, we suggest that TRN-SR2 is strongly binding CCD-CTD in a 1:2 molar ratio. With a large excess of CCD-CTD, some further weak and presumably nonspecific binding is also possible. In addition, we have used the same SEC setup to analyze the mixtures of TRN-SR2 with an IN construct, including the CTD domain only. These experiments (data not shown) have confirmed also that the isolated CTD binds to TRN-SR2.

HIV-1 IN mainly binds to the N-terminal half of TRN-SR2
Previous studies have shown that the interaction between TRN-SR2 and IN predominantly occurs via the CTD of integrase, although the CCD is involved to a lesser extent (21). However, it remained obscure which parts of TRN-SR2 are responsible for the interaction with IN. Here, we have used the AlphaScreen methodology to systematically address this question.
For the initial mapping, we have prepared three consecutive fragments of TRN-SR2 each corresponding to roughly onethird of the solenoid-like molecule's length. These fragments are composed of residues 1-281, i.e. HEATs 1-6 and the first helix of HEAT 7 (helix 7a), residues 281-531, i.e. helix 7b and HEATs 8 -12, and residues 531-923, i.e. HEATs 13-20, respectively. The two former fragments include the regions involved in the binding of RanGTP (37), whereas the third fragment includes the SR-binding helix (residues 657-671), corresponding to helix 15b (28). These three TRN-SR2 fragments, each carrying an N-terminal His 6 tag, were tested for interaction with IN carrying an N-terminal GST tag (Fig. 2). The full-length protein and fragments 1-281 and 281-531 displayed affinity for IN (with apparent K d of 17 Ϯ 4, 53 Ϯ 14, and 17 Ϯ 1 nM, respectively), whereas the 531-923 construct did not bind to IN.
To obtain further insight into the regions involved, shorter fragments of TRN-SR2 corresponding to a single inside helix (helix b) from each of the 20 HEAT repeats have been prepared (except for helix 2b which could not be purified). These TRN-SR2 fragments were subjected to AlphaScreen assays with the full-length IN (Fig. 3, a-c). The same experiments were

Interaction of TRN-SR2 with HIV integrase
repeated with the CCD-CTD construct, yielding essentially the same result (data not shown). To check for aspecific binding, all TRN-SR2 peptides that had tested positive to binding IN were additionally used in control experiments with LEDGF/p75, which does not bind TRN-SR2. This control experiment ( Fig.  3d) was important because of the modest length of the peptides. As a result, a reliable, specific interaction with integrase could be established for three peptides corresponding to TRN-SR2 helices 4b, 10b, and 11b (Fig. 3e). In addition, helices 2b, 7b, and 8b showed some weak binding to IN. For the peptide corresponding to helix 13b, a clear signal was obtained with both IN and LEDGF/p75 suggesting aspecific binding. Finally, in line with the results obtained with fragment 531-923 of TRN-SR2, no binding to IN could be detected for the C-terminal helices 14b through 20b.
Thus AlphaScreen assays involving both the "one-third" constructs of TRN-SR2 and individual inside helices of each HEAT consistently indicate that the binding of IN occurs within the N-terminal half of TRN-SR2. Importantly, we have previously established that the same region of TRN-SR2 is involved in the complex formation with RanGTP (37). Because RanGTP effectively displaces full-length IN from TRN-SR2 (40), we conclude that RanGTP directly competes with HIV IN for binding to TRN-SR2. In addition, we have carried out further out-competition experiments with integrase fragments, including CCD-CTD and also isolated CTD. Also, these two fragments were readily displaced by RanGTP (Fig. 4), confirming our hypothesis.

Structural characterization of the molecular complex between TRN-SR2 and the CCD-CTD
Attempts to crystallize TRN-SR2 in complex with HIV integrase or fragments thereof have been undertaken previously, by us and others, without success. Instead, we have resorted to SAXS experiments that can provide information about the overall structure of the macromolecular complex in a native solution environment (48,49). Thus, we have compared the solution structure of the TRN-SR2⅐CCD-CTD complex and two other TRN-SR2-based protein complexes that have been previously resolved using X-ray crystallography, namely those with RanGTP (37) and with a fragment of ASF/SF2 (28), an important natural cargo of TRN-SR2 (29,50,51). The latter fragment contained residues 106 -230 of ASF/SF2, which corresponds to the second RNA recognition motif (RRM2) and a nuclear localization signal enriched in arginine and serine residues, the so-called RS domain.
Initially, the samples were subjected to SEC-SAXS analysis. Specifically, the samples were run through a SEC column, and the eluent was continuously analyzed using a synchrotronbased SAXS setup. In each case, the elution profile revealed a single main peak for the complex. Analysis of the SAXS curves for each complex yielded an essentially constant radius of gyration (R g ) value across the peak, suggesting a homogeneous sample (Fig. 5). The R g value at the peak of the TRN-SR2⅐CCD-CTD complex was 45.9 Å. In contrast, both the RanGTP and ASF/ SF2 complexes revealed clearly smaller R g values (33.7 and 35.3 Å, respectively). At the same time, the expected masses of the three complexes are similar (152.7 kDa for the 1:2 complex of TRN-SR2 with CCD-CTD, 128.4 kDa for the 1:1 complex with RanGTP, and 131.7 kDa for the 1:1 complex with the ASF/SF2 fragment). These data indicate that the TRN-SR2⅐CCD-CTD complex is considerably less compact than the other two complexes, because the volume characteristics are proportional to the third power of R g .
SAXS data for TRN-SR2 complexes with both CCD-CTD and RanGTP were also collected in batch mode using a concentration series. The obtained curves for the TRN-SR2⅐CCD-CTD complex show some increase of the low-angle signal with increasing protein concentration but perfectly overlap at intermediate and high angles (Fig. 6a). This indicates that this complex is prone to a mild self-association, in contrast to the ideally behaving TRN-SR2⅐RanGTP complex (Fig. 6b) (40), whereas both complexes are quite homogeneous. We have further used five alternative SAXS-based calculations to estimate the molecular weight of the TRN-SR2⅐CCD-CTD complex ( Table 1, see also "Experimental procedures"). All estimates, in particular the one obtained through the recently proposed method of Rambo and Tainer (52) (158.6 kDa), match well the expected mass for the 1:2 complex (152.7 kDa). This confirms the stoichiometry obtained using SEC, because the addition or removal of even a single CCD-CTD monomer would change the mass considerably (by 24.7 kDa).

Interaction of TRN-SR2 with HIV integrase
The batch mode SAXS curves for the TRN-SR2⅐CCD-CTD and TRN-SR2⅐RanGTP complexes (Fig. 6c) are noticeably different, suggesting distinct overall structures of the complexes. Also, the pair distance distribution function (Fig. 6d) again suggests that the TRN-SR2 complex with CCD-CTD is considerably more extended. The maximal dimension of the TRN-SR2⅐CCD-CTD complex is 132 Å compared with only 102 Å for the TRN-SR2⅐RanGTP complex (Fig. 6d).
Finally, we have used all available structural information to create an initial model of the TRN-SR2⅐CCD-CTD complex, and we then further refined it using the obtained SAXS data (Fig. 7a). To this end, crystal structures of free TRN-SR2 (PDB code 4C0P) and CCD-CTD dimer (PDB code 1EX4) were first manually docked together as rigid bodies using the molecular contact information from AlphaScreen and mutagenesis experiments. In integrase, two sets of residues have been previously identified as being important for the interaction with TRN-SR2: Phe-185/Lys-186/Arg-187/Lys-188 in the CCD and Arg-262/ Arg-263/Lys-264/Lys-266/Arg-269 in the CTD (21). In TRN-SR2, the contacts with CCD-CTD were revealed by our AlphaScreen assays. A further elastic deformation ("opening") of the TRN-SR2 solenoid was necessary to obtain a good fit between the theoretical scattering calculated from the final model and experimental SAXS data ( 2 ϭ 1.4, Fig. 7a), which is logical given the observed increase in the maximal particle dimension. The final structural model also superimposes well with the ab initio molecular envelope obtained directly from the SAXS data (Fig. 7a).

Discussion
By now, TRN-SR2 is widely accepted as a key player in HIV nuclear import. However, the exact mechanism of its involvement is a subject of ongoing debate. Here, we have demonstrated a direct interaction between recombinant, highly purified samples of both human TRN-SR2 and HIV-1 IN, and we performed a systematic in vitro characterization of the resulting macromolecular complex using a range of biochemical and structural methods.
Titration of TRN-SR2 with increasing amounts of the CCD-CTD fragment of HIV integrase followed by SEC runs revealed a 1:2 stoichiometry for the primary binding (Fig. 1). This correlates well with the fact that the CCD-CTD dimerizes via the CCD (46). It is logical to suggest that the dimeric state of CCD-CTD is preserved upon its engagement in a complex with TRN-SR2. At the same time, it is the CTD of integrase that is known to provide the main contribution toward binding to TRN-SR2 (21,53,54). Interestingly, isolated CTD dimerizes in solution (55), and such CTD dimers could be observed in the recently published structure of the HIV-1 intasome (56).
We have also investigated which parts of TRN-SR2 are responsible for the interaction with HIV IN. Using AlphaScreen methodology, we could establish that IN binds to two fragments of TRN-SR2 each corresponding to about one-third of the total solenoid-like molecule, namely the fragments comprising residues 1-281 (i.e. HEATS 1-7) and 281-531 (i.e. HEATS 7-12), respectively, whereas the remaining C-terminal fragment did not bind the IN. We have further delineated three individual peptides, corre-  168.6 128.9 Using volume of correlation and R g (52) 158.6 Expected true mass from sequence 152.7 (1:2 complex) 128.4 (1:1 complex)

Interaction of TRN-SR2 with HIV integrase
sponding to the inner ␣-helices of HEAT repeats 4, 10, and 11, as strong IN binders. These data indicate that HIV integrase mainly interacts with TRN-SR2 through its N-terminal half (up to and including HEAT 11). The final component of this work was the study of the TRN-SR2⅐CCD-CTD complex in solution using SAXS, a technique that had been previously employed to characterize both the uncomplexed TRN-SR2 (54) and the TRN-SR2⅐RanGTP complex (40). As a result, we could derive a SAXS-based 3D model of the TRN-SR2⅐CCD-CTD complex. It is interesting to compare this model with the available crystal structures of TRN-SR2 complexes with RanGTP and ASF/SF2 (Fig. 7) (28, 37).
Generally speaking, binding within the N-terminal half of TRN-SR2 that we have observed here for the HIV IN CCD-CTD appears to be a recurring feature for various interacting partners. In TRN-SR2 complex with RanGTP, the latter was shown to be interacting mainly with the HEAT repeats 1-9 but also reaching to the C terminus, in particular HEAT 18 (Fig. 7b) (37). The crystal structure with the ASF/SF2 fragment (residues 106 -230) (28) revealed binding of the RRM2 domain mainly to HEAT repeats 4 -7, with some involvement of the C-terminal HEATs 19 and 20. In addition, the phosphorylated RS domain was binding to HEATs 14 -17, which in particular includes the arginine-rich helix 15b (Fig. 7c).
In this respect, it not surprising that the full-length IN but also its CCD-CTD fragment and isolated CTD domain were readily expelled from the complex with TRN-SR2 upon RanGTP addition (Fig. 4). This process resembles the action of RanGTP on TRN-SR2-bound natural cargoes. At the same time, the described motif of IN interaction with TRN-SR2 appears to leave the helix 15b of the latter accessible to binding the RS motif of bona fide partners such as ASF/SF2. Previously, a TRN-SR2 mutant carrying multiple alanine mutations of residues involved in binding the RS motif was shown to bind IN equally efficiently as the wild-type TRN-SR2 (28). As a future possibility, point mutagenesis within the HEAT repeats 4, 10, and 11 shown here to interact with IN would help to delineate the specific interacting amino acids in TRN-SR2. However, a statistically rigorous scanning would require testing a large number of alanine mutations.
The solenoid-like TRN-SR2 molecule appears to be rather flexible. Previously, high-resolution X-ray crystallographic studies have revealed that tight intermolecular interactions make the solenoid-like TRN-SR2 wrap around either the ASF/ SF2 fragment or RanGTP, reducing the radius of gyration of the complex compared with free TRN-SR2 (28,37). At the same time, our new SAXS data reveal that the TRN-SR2⅐CCD-CTD complex is, by contrast, considerably less compact than the complexes of TRN-SR2 with either the ASF/SF2 fragment or RanGTP, as evidenced by an ϳ30% increase in the radius of gyration. This difference is also apparent from a side-by-side comparison of the respective molecular models (Fig. 7). Apparently, the weaker interaction of TRN-SR2 with CCD-CTD is leading to a more open solenoid structure of TRN-SR2, which is the dominant component of the complex.
Several studies of the intasome, the complex of IN and viral DNA, are available. A crystal structure of the prototype foamy virus intasome has revealed a tetramer of IN bound to DNA, which is the minimal functional state required for a concerted integration (57). A crystal structure of the Rous sarcoma virus intasome (58) and a cryo-EM structure of mouse mammary tumor virus intasome (59) showed the existence of an octameric state, whereas a cryo-EM structure of maedi-visna lentivirus intasome presented a tetramer-of-tetramers architecture (60). Most recently, by fusing the DNA-binding protein Sso7d to the N terminus of HIV-1 integrase, it became possible to obtain a sample suitable for cryo-EM, yielding the first 3D structure of the HIV-1 intasome, including four IN chains (56).
Thus, although the HIV-1 particles contain ϳ250 copies of IN, only four of them appear to be involved in the intasome. Other IN chains may be expected to have a structural role within the PIC and/or orchestrate the nuclear import. In particular, the CTDs, which do not directly interact with the viral or target DNA but are exposed on the surface of the PIC, could interact with TRN-SR2. Moreover, our data suggest that the interaction between IN and TRN-SR2 does not involve the SRbinding region, which implies that a simultaneous transport of natural cargoes and HIV IN would be feasible, suggesting a model wherein the viral PIC is a stowaway during the TRN-SR2-mediated nuclear import of bona fide cargoes, such as ASF/SF2 and CPSF6. Finally, the new molecular insights into this interaction obtained here validate the interaction interface toward a rational design of compounds with inhibitory effect. Such lead compounds have a potential of becoming a future class of antiretroviral agents.

Expression and purification of proteins for biochemical and SAXS studies
Full-length human TRN-SR2 was inserted into the pETH-SUK vector, a derivative of pETHSUL (47), using the In-Fusion system (Clontech) essentially as described previously (37). The expressed fusion product contained an N-terminal His 6 tag, the SUMO protein, a SUMO hydrolase cleavage site, and the TRN-SR2 sequence. The protein was overexpressed in Escherichia coli NiCO21 (DE3) strain (New England Biolabs) containing the pRARE2 plasmid (EMD Biosciences). Bacterial cultures were grown in ZYP-5052 autoinduction media at 24°C until A 600 nm reached ϳ4 and then for another 24 h at 18°C until A 600 nm reached ϳ16 (61). The cells were pelleted by centrifugation and resuspended in buffer A (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 15 mM imidazole, 5 mM ␤-mercaptoethanol). After addition of 5 l of a cold-active nuclease Cryonase (Clontech) per 100 ml, the cells were lysed using an EmulsiFlex-C5 homogenizer (Avestin) followed by sonication on ice for 2 min with 40% amplitude and 2/2-s on/off pulses using the Branson Digital Sonifier. The lysates were cleared at 18,000 ϫ g for 35 min at 4°C and then were applied on a Ni 2ϩ -NTA (GE Healthcare) column using buffer A, followed by elution with buffer B (buffer A supplemented with 500 mM imidazole). Subsequently, the tag was cleaved off by an overnight incubation with SUMO hydrolase in a 1:500 molar ratio at 4°C in buffer A. This was followed by a second purification through the Ni 2ϩ -NTA column, whereby the flow-through containing the cleaved protein was collected, and the uncleaved protein and the protease, both containing the His 6 -SUMO tag, were retained on the column. The buffer was subsequently exchanged for 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 5 mM DTT using an Ultra-15 ultrafiltration device (Amicon), which was followed by ion-exchange chromatography on a 5-ml HiTrap Q HP column (GE Healthcare) using a linear gradient of 50 mM to 1 M NaCl.
The IN construct corresponding to CCD-CTD (residues 52-273) was expressed and purified similarly to methods described above for TRN-SR2. The lysates were purified with a Ni 2ϩ -NTA (GE Healthcare) column using buffer A in the washing step, and the protein was eluted by using buffer B, both supplemented with 1 M NaCl. SEC on a HiLoad 16/60 Superdex 200 preparation grade column (GE Healthcare) was then performed in 20 mM Tris-HCl, pH 7.5, 1 M NaCl, 5 mM DTT.
Inactive Q69L mutant of Ran protein has been overexpressed, purified, and charged with GTP as described previously (37). Thereafter the protein was incubated with TRN-SR2, and the resulting complex was purified using SEC (37).
Truncated ASF/SF2, which includes residues 106 -230 corresponding to the RRM2 and RS domains, has been inserted into the pETHSUK vector and co-expressed with human CDClike kinase 1, resulting in in vitro phosphorylation of ASF/SF2. The expression and purification of ASF/SF2 as well as formation and isolation of its complex with TRN-SR2 were done essentially as described previously (28).

Complex formation between TRN-SR2 and CCD-CTD and analytical SEC
The proteins were mixed in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM DTT at different molar ratios and incubated on ice for 1 h. Thereafter, the sample (150 l) was applied on a Superdex 200 10/300 column (23-ml volume, GE Healthcare) in the same buffer, and the eluted fractions were analyzed on a 15% PAGE in the presence of SDS-PAGE and stained with Coomassie Blue.

AlphaScreen interaction assays
Cloning, expression, and purification of tagged protein constructs for AlphaScreen assay was done essentially as described previously (8,62). Briefly, the pGEX expression plasmid was used as a template for cloning individual helical peptides of TRN-SR2. E. coli BL21-CodonPlus (DE3) cells were grown to an A 600 nm of 0.6, and protein expression was induced with 0.5 mM isopropyl D-thiogalactopyranoside. After incubation at 37°C for 2 h, the bacteria were harvested and then resuspended in a lysis buffer (PBS, pH 7.4, 1 M NaCl, 1 mM DTT, 1 mg/ml lysozyme, 1 mM PMSF, 1 l of DNase/10 ml, and 1% Triton X-100). After complete cell lysis by sonication, the supernatant was cleared by centrifugation, and recombinant proteins were bound to the glutathione-Sepharose resin (GE Healthcare). After washing the resin with 20 volumes of washing buffer (PBS, pH 7.4, 1 M NaCl, 1 mM DTT), the GST-tagged peptides were eluted with 5 ml of elution buffer (washing buffer supplied with 20 mM reduced glutathione). The fractions were analyzed using SDS-PAGE for protein content, pooled, and supplemented with 10% glycerol.
AlphaScreen assays were performed at least in duplicate as described previously (21). In the first series of experiments, GST-tagged full-length integrase at a fixed concentration of about 80 nM was tested against increasing concentrations of His 6 -tagged TRN-SR2 constructs (full-length, 1-281, 281-531, and 531-928). Protein concentrations were determined by A 280 nm . The apparent dissociation constant (K d ) was calculated assuming a simple bimolecular reaction. In the second series, His 6 -tagged full-length IN and His 6 -SUMO-CCD-CTD (uncleaved fusion) were tested against GST-tagged full-length TRN-SR2 as well as its individual peptides. Finally, to check for aspecific binding, all TRN-SR2 peptides that had tested positive to binding integrase were additionally used in control experiments involving His 6 -tagged LEDGF/p75, obtained as described previously (63). In addition, out-competition experiments were performed using His 6 -tagged integrase (fulllength, CCD-CTD, or CTD) and GST-tagged TRN-SR2 and increasing concentrations of untagged RanGTP. To this end, optimal binding concentrations of IN constructs and TRN-SR2 were first determined using cross-titration experiments.

SAXS measurements
Synchrotron X-ray scattering data were collected at the beamline BM29 of European Synchrotron Radiation Facility (Grenoble, France) and the SWING beamline of Synchrotron Soleil (Paris, France). X-ray radiation wavelength () of about 1 Å and a PILATUS 1M detectors were used. The data were collected up to q ϭ 4 sin/ ϭ 0.5 1/Å, where 2 is the scattering angle.
First, the samples have been subjected to SEC separation at 10°C followed by an in-line SAXS measurement (SEC-SAXS). To this end, a Shodex KW404-4F column equilibrated with 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM DTT was used. 200 l of protein sample at ϳ10 mg/ml in the same buffer was injected. The eluent at 0.2 ml/min rate has been continuously analyzed by SAXS using 1-s frames. Initial processing, including radial averaging and buffer subtraction, was performed by an automatic pipeline (64). The ATSAS program package (65) and the DATASW program (66) were used for further processing.
Second, protein samples were also analyzed using a batch mode SAXS. To this end, a concentration series from ϳ0.6 to 8 mg/ml in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 10 mM DTT was used. No radiation damage was detected for 10 successive frames with a 2-s exposure. Comparison of the SAXS curves collected from samples at various concentrations indicated some increase of the low-angle signal with concentration for the TRN-SR2⅐CCD-CTD complex. Correspondingly, subsequent analysis of these complexes used a merged scattering curve that had been generated from the low-angle part of the lowest concentration curve (ϳ0.6 mg/ml) and the remaining part of the highest concentration curve (ϳ8 mg/ml), as an approximation of a dilute solution. The distance distribution function P(r) and the Porod volume were calculated using the program GNOM from the ATSAS package. The molecular weight of the TRN-SR2⅐CCD-CTD complex was estimated using five different methods as follows: (i) based on the I(0) value ratio compared with the TRN-SR2⅐RanGTP complex, using the known mass (128.4 kDa) of the latter as a reference; (ii) based on the excluded volume of the hydrated particle using the Porod invariant (65); (iii) using the excluded volume of the DAMMIN model (65); (iv) using the program SAXSMoW (67) that calculates the particle volume based on the experimental data regularized with GNOM; and (v) the method of Rambo and Tainer (52). Ab initio models without symmetry were created using low-resolution data in the range 0.011 Ͻ s Ͻ 0.24 Å Ϫ1 with DAMMIN (68). Ten independent runs were performed and averaged using DAMAVER (Table 1) (69).
The model of the rigid-body complex of TRN-SR2 and the CCD-CTD dimer was optimized through the program SASREF (70) using molecular contact restraints derived from mutagenesis and AlphaScreen data. When exact crystallographic conformations were used, the final fit between the scattering calculated from the model and the experimental SAXS data had a 2 of 3.2. Thereafter, the TRN-SR2 molecule was opened up using the normal mode elastic deformation server NOMAD-Ref (71) with amplitude corresponding to an ϳ3-Å root mean square deviation from the initial structure, and the same docking procedure with CCD-CTD dimer was repeated. This has allowed us to improve the fit to the experimental data ( 2 ϭ 2.0). Thereafter, the program SREFLEX (72) was used to apply additional normal mode deformations, yielding the final model with 2 ϭ 1.4.