Note: The chemical shift assignment of full-length HK022 Xis_C28S is available in the BioMagResBank under the accession number BMRB-5539; the atomic coordinates and the structure factors (PDB_ID 1PM6) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ, USA.
H. Rüterjans, Institute of Biophysical Chemistry, J.W. Goethe-University of Frankfurt, Marie-Curie Str. 9, 60439 Frankfurt, Germany. Fax: + 49 69 798 29632, Tel.: + 49 69 798 29631, E-mail: email@example.com
Heteronuclear high-resolution NMR spectroscopy was employed to determine the solution structure of the excisionase protein (Xis) from the λ-like bacteriophage HK022 and to study its sequence-specific DNA interaction. As wild-type Xis was previously characterized as a generally unstable protein, a biologically active HK022 Xis mutant with a single amino acid substitution Cys28→Ser was used in this work. This substitution has been shown to diminish the irreversibility of Xis denaturation and subsequent degradation, but does not affect the structural or thermodynamic properties of the protein, as evidenced by NMR and differential scanning calorimetry. The solution structure of HK022 Xis forms a compact, highly ordered protein core with two well-defined α-helices (residues 5–11 and 18–27) and five β-strands (residues 2–4, 30–31, 35–36, 41–44 and 48–49). These data correlate well with 1H2O-2H2O exchange experiments and imply a different organization of the HK022 Xis secondary structure elements in comparison with the previously determined structure of the bacteriophage λ excisionase. Superposition of both Xis structures indicates a better correspondence of the full-length HK022 Xis to the typical ‘winged-helix’ DNA-binding motif, as found, for example, in the DNA-binding domain of the Mu-phage repressor. Residues 51–72, which were not resolved in the λ Xis, do not show any regular structure in HK022 Xis and thus appear to be completely disordered in solution. The resonance assignments have shown, however, that an unusual connectivity exists between residues Asn66 and Gly67 owing to asparagine-isoaspartyl isomerization. Such an isomerization has been previously observed and characterized only in eukaryotic proteins.
Knowledge about the molecular mechanisms of viral site-specific integration/excision in prokaryotes and eukaryotes can be widely employed in biotechnological and medical applications, such as site-specific genomic targeting, drug design, vector construction, etc. The structural determinants of integrative recombination were studied extensively in various viruses during the 1990s[1–4], and almost all proteins participating in the bacteriophage λ recombination system have been structurally characterized [5–8]. However, very little is known about the structural basis of excisionase function, although its importance is generally recognized.
Two closely related bacteriophages –λ and HK022 – use common mechanisms for integration/excision of their genomes during a life cycle. The phage-encoded integrase, Int, recognizes attP (on the phage chromosome) and attB (on the bacterial chromosome) core sites and performs site-specific recombination with the help of the cell-encoded integration host factor, IHF, resulting in integration of the circular phage DNA into the cellular chromosome. The attP and attB sites generate the prophage sites attR and attL, which flank the inserted phage DNA. The reverse reaction (called excisive recombination) leads to excision of the prophage by recombination between attR and attL and regeneration of the attP and attB sites. The excision requires an additional enzyme – the excisionase (Xis). The cellular protein factor for inversion stimulation, FIS, enhances the excision, but cannot replace Xis [9,10]. Xis plays a key role in reorienting the recombination directionality. It has been shown that Xis from bacteriophage λ binds cooperatively to the two tandemly arranged specific DNA sites X1 and X2, which are located in the long P-arms of attR. The X2 site overlaps with the FIS binding site F, and in vivo the λ Xis bound to X1 cooperates either with FIS or with a second λ Xis molecule . When λ Xis occupies the X1 and X2 sites (or Xis and FIS occupy the X1 and F sites, respectively) the DNA becomes significantly bent – up to 140°. λ Xis also enhances the binding of Int to its second specific site, P2, whose affinity for Int is quite weak . These two factors that are promoted by Xis, DNA bending and Int binding to P2, have been proposed to change the directionality of the recombination.
The Xis proteins from both λ and HK022 phages are identical 72-residue proteins, with the exception of Gly59 which is substituted by Ser in HK022 Xis . Their binding sites are very similar and the proteins can be interchanged [14–16]. Structural studies of Xis were hampered by protein instability and a high intracellular toxicity . However, site-directed mutagenesis, in combination with functional studies, provided valuable information about the participation of different λ Xis regions in DNA binding, and Xis–Int, Xis–Xis and Xis–FIS interactions. Site-directed mutagenesis revealed that the N-terminal part of λ Xis (residues 1–53) is involved in both the DNA specific recognition and the interaction with FIS or a second Xis molecule, whereas the C-terminal part (residues 54–72) is important for the interaction with Int . Alterations of the C-terminal part, such as single amino acid substitutions or even its entire deletion, did not completely inactivate the in vivo excisive recombination activity of λ Xis, although its efficiency was substantially reduced . For example, when λ Xis is mutated at positions 57, 60, 62, 63, 64 or 65, the λ Xis–Int interaction is prevented, probably as a result of an inability to form the required structural motif. For this particular region (residues 59–65) an α-helical structure has been suggested by various prediction methods .
Recently, the 3D structure of the N-terminal fragment (residues 1–55) of the excisionase from λ phage was reported . This λ Xis N-terminal fragment (with a Cys28→Ser substitution) displays a tertiary fold characteristic for the ‘winged’ helix family of DNA-binding proteins [20,21]. It was found that at pH 5.0 the λ Xis solution structure consists of two antiparallel β-strands and two α-helices, whereas residues 51–72 appeared not to show any regular structure. Additionally, a model of sequence-specific DNA interactions was proposed, based on the protein structure. However, the reasons for Xis instability in vitro and in vivo, as well as the role of the C-terminal tail, remained unclear. In order to better understand the functional organization of excisionases, a structural and thermodynamic study of the excisionase from the λ-like coliphage HK022 was carried out in this work.
As wild-type λ Xis was previously characterized as a structurally unstable protein , in the present study the 3D structure of full-length HK022 Xis containing a single amino acid substitution, Cys28→Ser (Xis_C28S) (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results), was obtained by NMR spectroscopy. The validity of this structure was demonstrated for a wide range of external conditions by means of differential scanning calorimetry (DSC) and NMR spectroscopy. Comparison of the selected NOE patterns of HK022 Xis_C28S and wild-type HK022 Xis (Xis_wt) revealed complete structural identity of the two proteins. The main fold of the full-length HK022 Xis_C28S is very similar to that reported for the λ Xis N-terminal fragment, with a few minor (but important) differences. Contrary to the structure reported by Sam and co-authors , residues 2–4 of HK022 Xis adopt a β-strand conformation, forming a three-stranded antiparallel β-sheet together with residues 35–36 (β-strand 3) and residues 41–44 (β-strand 4). This makes the full-length HK022 Xis structure more similar to those of the typical ‘winged’-helix proteins, e.g. the bacteriophage Mu repressor DNA-binding domain . The family of excisionases comprises 63 different enzymes that are able to change or modulate the directionality of recombination . The structural and thermodynamic study of HK022 Xis presented here provides a useful insight into their structural organization and functionality.
Materials and methods
Preparation of protein and DNA samples
The previously constructed pPG14 plasmid , containing the HK022 Xis gene linked with a His-tag and a thrombin cleavage site at the protein N terminus, was used, in this work, for the production of Xis_wt protein. The Cys28→ Ser single amino acid substitution was performed with reference to the pPG14 construct, and the resulting plasmid (pPG14_C28S; I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results) was employed for the production of Xis_C28S.
A previously designed protein isolation and purification procedure (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results) was slightly modified in order to achieve optimal stable isotope labelling. The Escherichia coli BL21*(DE3)/pLysS strain was freshly transformed with the plasmids pPG14 or pPG14_C28S prior to protein overexpression in modified ECPM1-x media (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results), containing 1 g·L−1 unlabelled NH4Cl, 40 g·L−1 unlabelled glycerol and Trace Elements I solution . The cells were incubated in a fermenter, with intensive aeration, to slightly beyond the mid-log phase (A600 = 1.3–1.6), after which the cells were collected by centrifugation and resuspended in 4.0 L of ECPM1-x media without any sources of nitrogen (for the preparation of 15N-labelled protein) or containing neither nitrogen nor carbon sources (for the preparation of 13C,15N-labelled protein). After a short starvation period (20 min), the cells were supplemented with either 1.0 g·L−1 of 15NH4Cl or 1.0 g·L−1 of 15NH4Cl, 0.5 g·L−1 of 13C-labelled glucose and 1.5 g·L−1 of 13C-labelled glycerol. Cell growth was continued for another 20 min before the addition of isopropyl thio-β-d-galactoside (1.0 mm final concentration) to induce expression of the Xis gene. After 2 h 45 min of induction, the cells were harvested by centrifugation and lysed using a French press. All samples of Xis_wt were supplemented with 10 mm 2-mercaptoethanol to protect the thiol group of the protein against oxidation. The cleared cell lysate was subjected to Ni2+ chelating chromatography. Protein elution was performed using a linear gradient of 50–400 mm imidazole.
The collected protein was cleaved with thrombin, and a subsequent preparative gel filtration, through a 2.6 × 60 cm Superdex 75 column, was employed as the final step of protein purification. Homogeneity of the protein was verified by SDS/PAGE and mass spectrometry (MALDI-TOF). The fractions containing > 99% pure Xis protein were used for the NMR sample preparation and other applications.
For all NMR sample preparations, the selected Xis fractions were exhaustively dialyzed against a 100-fold excess of NMR buffer [50 mm sodium phosphate (pH 6.5), 100 mm NaCl, 0.2 mm EDTA (disodium salt), 0.03% NaN3] and subsequently reduced in volume to a final Xis concentration of 0.1–1.2 mm, depending on the experiment. The protein concentration was derived from the optical density of the samples using a calculated extinction coefficient of 13 940 mm−1·cm−1 (A0.1%1cm = 1.614) at 280 nm; 5%2H2O as lock substance and 0.1 mm 4,4-dimethyl-4-silapentane-1-sulfonate as internal proton chemical shift standard were added to the samples. All NMR samples of Xis_wt also contained 1.0 mm dithiothreitol. Glycerol (6%) was added to the NMR buffer for monitoring the Xis–DNA interaction in order to reduce aggregation of the complex. Usually, 300 µL samples were placed into 5 mm NMR tubes (Shigemi, Allison Park, PA, USA) under argon protection.
The Xis samples for DSC were dialysed against the buffers, and then filtered and degassed prior to filling of the calorimetric cell. The buffers used in this work consisted of 25 mm buffer species (sodium acetate for the pH range 4.5–5.5; sodium phosphate for the pH range 6.0–7.0) and 100–400 mm NaCl. For all Xis_wt samples, 1 mm dithiothreitol was added to the buffer prior to dialysis. The pH values were measured at 25 °C without corrections for the temperature dependence.
A synthetic 20 bp DNA duplex was used in this work for the DNA binding experiment. It contained 15 bp of the natural HK022 Xis binding site, X1 [14,24], with stabilizing GCG and GC sequences at the 5′- and 3′ termini, respectively. Two single-stranded DNA oligonucleotides – 5′-GCGATATGTTGCGTTTTGGC-3′ and the complementary sequence (purchased from Carl Roth, Karlsruhe, Germany) – were annealed, and the double-stranded X1 was purified by gel filtration in buffer containing 40 mm K2HPO4 and 100 mm NaCl, pH 6.0. The double-stranded X1 DNA was concentrated and equilibrated with the corresponding NMR buffer in an Amicon ultrafiltration device (membrane MWCO = 0.5 kDa).
The DSC data were recorded using the SCAL-1 scanning microcalorimeter (SCAL, Pushchino, Russia) at a pressure of 2.0 atm. The optimal heating rate (60 K·h−1) was established experimentally. The data were sampled and processed using the service program wscal, based on the principles described by Filimonov et al.  and Privalov & Potekhin .
For calculating the experimental partial molar heat capacity function, Cp,pr(T), the partial specific volumes of Xis_wt and Xis_C28S were assumed to be 0.73 mL·g−1. The protein concentration, derived from the absorbance of the samples, varied in the DSC samples from 1.0 to 2.9 mg·mL−1. The molecular weights of Xis_wt and Xis_C28S were calculated, from the amino acid sequence, to be 8.635 kDa and 8.619 kDa, respectively.
The analyses of the Cp,pr(T) functions were performed as described previously [25,27–30].
All NMR spectra for resonance assignments and structure determination were collected at 303 K on Bruker DMX 500 and DMX 600 spectrometers, equipped with 5 mm triple-resonance (1H/13C/15N) probes with XYZ-gradient capability. Proton chemical shifts were referenced relative to internal 4,4-dimethyl-4-silapentane-1-sulfonate; 15N and 13C chemical shifts were referenced indirectly using the corresponding chemical shift ratios .
3D Triple-resonance [15N,1H]-TROSY-HNCO , (HCA)CO(CA)NH  and [15N,1H]-TROSY-HNCACB  spectra were collected for the sequential backbone resonance assignments. Side-chain resonance assignments were achieved using the following experiments: 3D HBHA(CBCA)(CO)NH , 3D H(CC)(CO)NH-TOCSY , 3D (H)C(C)(CO)NH-TOCSY  and a 15N-edited 3D TOCSY-HSQC . The resonances of aromatic ring protons were assigned using 2D clean [1H-1H]-TOCSY spectra [39,40] recorded with spin-lock times of 59.3 ms and 5.6 ms, and 2D [1H-1H]-NOESY spectra in 1H2O and 2H2O.
3D Heteronuclear NMR spectra were collected for determining the interproton distances in HK022 Xis_C28S. A 15N-edited 3D NOESY-HSQC experiment , employing water flip-back  and gradient sensitivity enhancement , was acquired with a mixing time of 100 ms. 13C-Edited 3D NOESY-HSQC spectra (mixing time 70 ms) were recorded in two different versions; optimized for Hα/Hβ NOE-correlations (3D NOESY-[13C,1H]-HSQC) and for methyl group NOE-correlations (3D NOESY-(CT)-[13C,1H]-HSQC).
The proton exchange experiments were carried out at 600 MHz and temperatures of 288 K and 298 K. A reference [1H,15N]-HSQC spectrum was recorded with fully protonated Xis_C28S. The sample was then lyophilized and dissolved in the same volume of ice-cold 2H2O. A series of identical [1H,15N]-HSQC spectra were acquired every 15 min during the first 2 h, and thereafter every 30 min, until all amide protons were completely exchanged with 2H (after 12 h).
[15N,1H]-TROSY and homonuclear 1D spectra were collected to establish the Xis–DNA interaction at 288 K. A reference [15N,1H]-TROSY spectrum was recorded using 0.5 mL of a 0.4 mm Xis_C28S sample (20 mm sodium phosphate, 100 mm NaCl, 6% glycerol, pH 6.5); equivalent spectra were acquired after each titration step with 25 µL of 1.2 mm X1 (20 bp DNA duplex) until a protein/DNA ratio of 1 : 3 was reached. The reverse order of titration, when 25 µL of a 1 mm Xis sample was added stepwise to 0.5 mL of a 0.4 mm X1 solution, was monitored by 1D (SW = 32.2 p.p.m.) and [15N,1H]-TROSY spectra up to a protein/DNA ratio of 3 : 1.
The NMR spectra were processed and analyzed on Silicon Graphics workstations using the xwinnmr 2.6, aurelia 2.7.5 (Bruker BioSpin, Rheinstetten, Germany) and felix 97 (Accelrys, San Diego, CA, USA) programs.
Restraint generation and structure calculation
The NOE-based distance restraints were extracted from 3D 13C- and 15N-edited NOESY-HSQC spectra and homonuclear 2D NOESY spectra in 1H2O and 2H2O. Automated assignments of the NOEs, based only on chemical shifts, were obtained using the program nmr2st.
The structures were calculated using a simulated annealing protocol with torsion angle dynamics (dyana 1.5 ), combined with an iterative structure refinement procedure . Using the program glomsa, 37 stereospecific assignments were obtained for the prochiral methylene and isopropyl groups of 29 residues, including γ1/γ2 of six Val residues and δ1/δ2 of three Leu residues. A cis-proline residue entry was added to the standard dyana residue library for correct calculation of Pro32, which has been identified as a cis isomer. No additions were made in the standard dyana residue library with respect to the IsoAsp66 residue, as the C-terminal tail of HK022 Xis was shown to be disordered in solution and no NOE violations were found with the usage of the standard Asn66 residue.
For calculating the final structure ensemble, 868 NOE-derived distance restraints and 34 hydrogen bond restraints (dHO ≤ 2.1 Å and dNO ≤ 3.1 Å) were employed. Subsequent restrained energy minimization, carried out using the discover module of the insight 97 software package (Accelrys, San Diego, CA, USA), was performed with the 20 best dyana conformers. The minimized structures were analyzed using procheck-nmr 3.4 . The structure images were prepared using the molmol program .
Expression and purification of the Xis_wt and Xis_C28S proteins
The isolation and purification of Xis was hampered by the fact that the protein can adopt a non-native conformation, characterized by a significant decrease in structural integrity and functional activity. This conformation tends to form high-order aggregates and seems to consist of disulfide-bridged dimers, as evidenced by SDS/PAGE. Therefore, an amino acid substitution was designed to replace the SH group of Cys28 with the OH of Ser in order to decrease the irreversibility of the Xis transition to the non-native state (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results).
The replacement of Cys28 with Ser in Xis resulted in a reasonable stabilization of the protein against aggregation in various buffer systems and allowed optimization of the isolation/purification scheme in order to reach a sufficient protein yield. Interestingly, the optimized conditions could also be successfully applied to the isolation, purification and storage of Xis_wt. Comparison of the biophysical characteristics and biological activity of Xis_wt and Xis_C28S revealed a high similarity in almost all protein parameters (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results). The same amino acid substitution was used recently to stabilize the N-terminal fragment of λ Xis; an equal ability of both the wild-type and the mutant protein to bind specific DNA and to initiate excisive recombination in vivo was also demonstrated .
Thermodynamic characterization of HK022 Xis
The instability of the Xis protein has been a significant obstacle for structural investigations. In this work, DSC was used to study the denaturation of Xis in order to define the stabilization energy (ΔG) of Xis_wt and Xis_C28S under various experimental conditions and thus determine the optimal conditions for NMR experiments. The thermodynamic parameters of Xis denaturation under these conditions are summarized in Table 1.
Table 1. Thermodynamic parameters of the HK022 Xis_wt denaturation. The buffers used for data collection contained a 25 mm buffer species (sodium acetate for pH 5.5; sodium phosphate for pH 6.5 and 7.0) and 100 mm NaCl. ΔCp calculated from the temperature dependence of ΔH is equal to 3.1 kJ·mol−1·K−1. ° Indicates the value of the corresponding parameter at 25 °C. FN defines the protein native state population.
ΔH (Tm) (kJ·mol−1)
a Data for the Xis_wt thermal denaturation in the presence of 400 mm NaCl. b Data for the Xis_C28S thermal denaturation.
It was found that the thermal denaturation of Xis is highly reversible and cannot be responsible for the previously observed irreversible inactivation of the protein. In Fig. 1A, two repetitive scans of the same Xis_wt sample did not show any significant difference, demonstrating the ability of the protein to reconstitute the initial tertiary structure after denaturation. This observation was also supported by NMR data (not shown).
The influence of pH on the Xis_wt thermal denaturaion is illustrated in Fig. 1B. A visible increase of the protein stability was observed when the pH was raised from 4.7 to 7.0. At pH 4.7, the denaturation midpoint temperature (Tm) was so low, and consequently the denaturation enthalpy (ΔH) so small, that the molar partial heat capacity function [Cp,pr(T)] did not contain a substantial peak. Although the Cp,pr(T) of the protein showed an intensive heat absorption peak (Tm = 41.6 °C) when recorded at pH 5.5, the thermodynamic analysis of this function indicated a significant denatured state population (FD) of Xis_wt already at room temperature (Table 1). In contrast, at pH 6.5 and 7.0 the protein stability was higher; the FD did not exceed 3%, even at 30 °C.
Furthermore, it was found that the stability of Xis could be significantly increased by raising the salt concentration. Figure 1C shows two Cp,pr(T) functions obtained for Xis_wt at pH 5.5 and NaCl concentrations of 100 and 400 mm. In the latter, the Tm was shifted by +5.7° (from 41.6 to 47.3 °C). Thermodynamic analysis indicated that this salt-induced stabilization is mostly entropic in nature as the ΔH values were almost equal for low- and high-salt conditions at the same temperature.
A detailed analysis of the Xis_wt and Xis_C28S thermal denaturation was performed for Cp,pr(T) functions obtained under the same conditions as the NMR experiments (Fig. 2). Direct comparison of the thermodynamic parameters of Xis_wt and Xis_C28S indicated that the influence of the Cys28→Ser amino acid substitution on the protein stability was very small (Table 1). The difference in Tm values (51.4 °C for Xis_wt vs. 51.9 °C for Xis_C28S) was within the limits of experimental error.
The high reversibility, the independence of Tm on protein concentration, and the presence of only one heat absorption peak suggest a simple monomolecular two-state scheme for Xis denaturation. Indeed, a Cp,pr(T) function that was simulated based on this assumption (Fig. 2A, dots) fits the experimentally obtained curve reasonably well.
The temperature dependence of ΔH is shown in Fig. 2B. The specific (J·g−1) presentation of ΔH revealed that the Xis denaturation enthalpy at 130 °C (ΔHspecific = 35 J·g−1) was significantly lower than those values reported for small globular proteins (50 J·g−1 ± 15% at 130 °C ). This difference suggests that the Xis molecule is not completely structured. The structured region of the protein is apparently limited to the segment encompassing residues 2–50. After correction for the size of this cooperative unit (apparent molecular mass of 6.070 kDa instead of 8.635 for full-length Xis), the ΔHspecific of 50 J·g−1 at 130 °C corresponds well with the expected value.
The calculated temperature dependences of FN and FD at the NMR conditions used are presented for Xis_wt in Fig. 2C. According to the analysis, FD starts to increase rapidly at temperatures only above 35 °C. Thus, 30 °C (FD = 0.03) was chosen as an optimal temperature for the NMR experiments.
Although this thermodynamic study revealed an equal stability of Xis_wt and Xis_C28S, the latter showed a significantly lower tendency to aggregate at high concentrations and remained native for a longer time. Thus, Xis_C28S was chosen for the structural study by means of heteronuclear NMR spectroscopy.
Assignment of backbone and side-chain resonances in HK022 Xis_C28S spectra
Nearly complete backbone and side-chain resonance assignments were achieved for Xis_C28S (Fig. 3A), except for the 13C resonances of the aromatic rings, the acidic carboxyl groups and the amide resonances in the side-chains of Arg and Lys. In contrast to the previously published λ Xis N-terminal fragment assignment , the backbone amide proton resonances of Tyr2 and Thr4 have been identified for HK022 Xis_C28S. Despite different experimental conditions, the only other significant difference in the backbone amide resonances of the two assignments is the position of Glu45 HN.
The [1H,15N]-HSQC spectrum of Xis_wt corresponds almost entirely to Xis_C28S (Fig. 3B). The most strongly shifted HN resonances of HK022 Xis_wt (shown in red) were assigned with the help of 3D 15N-edited NOESY spectra. Differences in chemical shift values occurred only in close proximity to Cys28 (residues Arg26–Phe31) and did not affect the amide resonances of residues sequentially more than four amino acids distant from Cys28. Moreover, comparison of the 15N-edited 3D NOESY-HSQC spectra of Xis_wt and Xis_C28S, revealed the almost complete structural identity of these two proteins (data not shown).
In the course of the sequence-specific resonance assignment, an unusual connectivity was observed between Asn66 and Gly67. This connectivity was identified as an isoaspartyl linkage (Fig. 4A), which has been previously reported for Asn–X pairs in several other proteins [51–53]. Recently, an isoaspartyl linkage was described for the Asn306–Gly307 pair in malate synthase G, based on the juxtaposition of expected and observed signs of the N306 Cα and Cβ signals in 3D and 4D heteronuclear NMR spectra . In the current study, the isoaspartyl linkage was determined based on the relative sign of the signals in the 3D [15N,1H]-TROSY-HNCACB spectrum.
Figure 4B presents the sequential connectivities of Xis_C28S residues 65–69, as observed in the 3D [15N,1H]-TROSY-HNCACB (left panel) and 3D (HCA)CO(CA)NH (right panel) spectra. In the HNCACB experiment, the resonance signals of 13C directly coupled to 15N are always positive, whereas the relayed 13C resonance signals are negative. The sequential connectivity involving the Cβ resonance of IsoAsp66 is therefore positive in sign, whereas the Cα signal is negative (Fig. 4B, left panel, G67 strip). In contrast, the intraresidual connectivities involving the Cα and Cβ resonances of IsoAsp66 show the usual signs (Fig. 4B, left panel, N66 strip).
In the (HCA)CO(CA)NH experiment, the intraresidually observed IsoAsp66 13C(O) resonance [i.e. the 13C(OO–) at 179.2 p.p.m.; Fig. 4B, right panel, N66 strip] is shifted downfield owing to the negative charge; it does not correspond to the sequentially observed carbonyl (i.e. the Cγ of the IsoAsp66 residue at 176.2 p.p.m.; Fig. 4B, right panel, G67 strip). Both spectra unambiguously demonstrate the IsoAsp66–Gly67 linkage in Xis_C28S; resonances corresponding to the usual Asn66–Gly67 residue pair were not observed.
The solution-state structure of HK022 Xis
A compact, well-resolved structure has been calculated for full-length HK022 Xis_C28S. The 20 final conformers, superposed at the well-structured Tyr2–Val50 region of Xis_C28S, are shown as a stereo-view representation in Fig. 5A. The root mean square deviation (RMSD) value of the backbone atoms in this region is 0.83 Å; excluding residues 12–17 (the flexible loop between the two α-helices), these structures can be superposed with a backbone RMSD of 0.71 Å. The statistics of the final structure calculation are summarized in Table 2.
Table 2. Structural statistics of the 20 energy-minimized conformers of HK022 Xis_C28S. RMSD, root mean square deviation.
a The number of included H-bond restraints is indicated in parentheses. b N, Cα, C′. c Values for the structured part only (residues 2–50).
The global structure of HK022 Xis consists of two small antiparallel β-sheets and an L-shaped α-helical motif (Fig. 5B). The α-helices 1 (residues Leu5–Arg11) and 2 (Leu18–Glu27) are separated by a loop L (residues Glu12–Ser17) and are oriented nearly orthogonal to each other. The first antiparallel β-sheet consists of three β-strands: β-strand 1 (residues Tyr2–Thr4), β-strand 3 (residues Val35–Lys36) and β-strand 4 (residues Tyr41–His44). Residues Asp37–Glu40 form a reverse β-turn T between β-strands 3 and 4. The β-strands 2 (residues Ile30–Phe31) and 5 (residues Val48–Lys49) form the second β-sheet. The size of this structure element is rather small and was not recognized by any secondary structure prediction program. On the other hand, the proton-exchange data and a detailed structural analysis suggest that there is a strong hydrogen bond interaction between the amide proton of Phe31 and the carbonyl oxygen of Val48. The experimentally determined NOE patterns within the β-sheets of the HK022 Xis_C28S structure are presented in Fig. 6.
The triproline segment (residues Pro32–Pro34) acts as a linker between β-strands 2 and 3. The first residue in this segment, Pro32, adopts a cis conformation (as supported by NOE data) whereas the other two residues, Pro33 and Pro34, are trans prolines. The bulge structure (residues 45–47), which connects β-strands 4 and 5, is positioned just opposite to the triproline segment (Fig. 5B).
The residues from Asp51 to Ser72 are not included in any regular structure element in HK022 Xis, as indicated by the lack of any medium- or long-range NOEs. Hence, the C terminus is largely disordered and displays very high local RMSD values.
Interaction of HK022 Xis with specific DNA (X1)
A series of [15N,1H]-TROSY and 1D 1H spectra was recorded in order to identify the amino acids in the Xis molecule that are directly involved in the protein–DNA interaction. Unfortunately, strong association/aggregation was observed under all conditions tested, both when the DNA was titrated to the protein and vice versa. Even with the usage of relatively dilute protein samples (0.1 mm), association/aggregation was found to be prevalent. It should be pointed out that aggregation had already started at the very first titration steps (at a protein–DNA molar ratio of 10 : 1) and affected nearly all Xis in solution. This behavior suggests that Xis may change its structure when bound to DNA, thus facilitating or inducing Xis–Xis and/or Xis–FIS interactions.
Although the large size of the associates/aggregates did not permit the identification of direct contacts between protein and DNA, it was possible to define a part of Xis that was not as strongly influenced by the protein–DNA complex formation. The [15N,1H]-TROSY spectrum of Xis_C28S in the presence of its site-specific DNA (a 20 bp DNA duplex containing the X1 site) at a 1 : 3 molar ratio is shown in Fig. 7. At this ratio, all Xis molecules are bound to X1 DNA. Hence, only the HN resonances of very mobile amino acids should be detected in this spectrum. Besides a few signals that could not be directly assigned, the C-terminal amide resonances (starting with Leu52 HN) were all observable at the same positions as in uncomplexed Xis_C28S.
Bacteriophages λ and HK022 are closely related lambda-like Enterobacteria viruses. They use common mechanisms for integration/excision of their genomes during their life cycle. Both phage excisionases show almost complete sequence identity; only one out of 72 residues differs in these two proteins.
Although the 3D structures of the two excisionases were therefore expected to be very similar, some differences were nevertheless clearly observed. First, HK022 Xis consists of five β-strands, whereas λ Xis features only two regular β-strands and two extended sequence segments . Second, the ‘wing’β-sheet of HK022 Xis_C28S consists of three β-strands and is therefore structurally more similar to other proteins of this class, such as the DNA-binding domain of the Mu repressor . Third, the backbone atoms of the region comprising the ‘wing’β-sheet (residues 2–4, 35–36 and 41–44 of HK022 Xis and residues 16–18, 47–48 and 58–61 of the Mu repressor) could be superposed with an RMSD value of 0.49 Å, whereas superposition with the same region of λ Xis led to an RMSD value of 1.30 Å (Fig. 5C, left panel). In Fig. 5C (right panel), the reverse β-turns of these three proteins are compared. Again, the flexibility and direction of this structural element in full-length HK022 Xis were more similar to those in the DNA-binding domain of the Mu repressor .
This structural difference between λ and HK022 Xis may be the result of differences in the conditions of the NMR investigations, i.e. pH 6.8 in the current study vs. pH 5.0 in the work of Sam et al. . At pH 5.0, the spectra of truncated λ Xis did not reveal the Tyr2 and Thr4 backbone amide proton resonances that are crucial for the identification of the NOE contacts in the first β-sheet. However, the spectra of full-length HK022 Xis_C28S and Xis_wt, acquired under the same experimental conditions reported by Sam et al., still displayed these resonances. This suggests that there may be some basic differences in the structural organization and thermodynamic stability between the full-length and truncated Xis proteins.
Of particular interest is the segment of three consecutive proline residues (Pro32–Pro33–Pro34) connecting β-strands 2 and 3. The NOE patterns and structure calculations of Xis_C28S revealed a cis–trans–trans arrangement of these prolines. Currently, only six protein structures are available in the Brookhaven Data-Bank that have a triproline motif with the first proline in a cis configuration. These include cytochrome c oxidase (PDB entries 1AR1, 2OCC, and 1OCO), endo-β-N-acetylglucosaminidase F1 (2EBN), myelin basic protein (1QCL), protocatechuate 4,5-dioxygenase (1BOU and 1B4U), thiaminase I (2THI), and cytotoxic T lymphocyte-associated antigen 4 (1DQT, 1I85, and 1I8L). All of these proteins feature a cis–trans–trans triproline motif, except for the latter which shows a cis–trans–cis configuration. Interestingly, these triproline sequences are always located on the protein surface – directly at the interaction site, in those cases where protein–protein contacts have been observed. This suggests a similar role for the Pro32–Pro33–Pro34 segment on the Xis surface, upon interaction with ligands such as FIS or another Xis molecule.
The amino acid substitution of Cys28→Ser did not significantly change either the structure or stability of HK022 Xis. The differences observed in the NMR spectra of HK022 Xis_wt and Xis_C28S occurred almost exclusively in the immediate vicinity of residue 28 (Fig. 3B), indicating that the overall structure of Xis_wt is not affected by the substitution. The DSC study revealed that the stabilities of the two proteins are approximately equal (Table 1). Thus, one can assume that the disulfide bridge between two Xis molecules can be formed only after perturbation of the native Xis structure, as observed during Xis aggregation (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results) or the association/aggregation upon interaction of Xis with specific DNA.
Spontaneous aspartate isomerization and deamidation of asparaginyl residues can serve as an initial site of spontaneous nonenzymatic degradation of proteins or peptides in vivo and in vitro. Such a degradation plays an important role in biological systems and serves as a molecular clock . In this work, an IsoAsp–Gly connectivity was unambiguously identified for the prokaryotic regulatory protein Xis, although it was previously observed and described as an attribute of regulation only in eukaryotic organisms. It is still not clear whether the isomerization in Xis (and subsequent protein degradation) also occurs in vivo; however, such a type of negative regulation of both λ and HK022 Xis may be physiological relevant, as even a small amount of Xis in infected cells can turn the phage life cycle to the lytic pathway. It has been shown previously that the λ Xis concentration in E. coli is strongly regulated by cellular proteases ; this isomerization could serve as a signal for protein self-degradation, thus playing an additional regulatory role. In fact, spontaneous degradation of Xis has been observed and found to cause significant problems in the protein preparation (I. Kleinhaus, K. Werner, H. Rüterjans and V. V. Rogov, unpublished results).
On the other hand, the unusual IsoAsp66–Gly67 connectivity in Xis does not affect the general fold or functional properties of the free protein. It has been shown  that this isomerization generally occurs in highly mobile protein regions and is accelerated at neutral or basic pH. As a high mobility of the HK022 Xis C terminus was reliably demonstrated in the current work, a possible functional meaning of this isomerization/degradation could be to prevent the formation of the predicted C-terminal α-helix (amino acids 59–64)  and to subsequently reduce the ability of Xis to recruit Int to the P2 site.
The previous investigations of excisionase interaction with specific DNA were usually performed with large (> 100 bp) DNA duplexes containing X1, X2 or X1/X2 sequences [20,24,57]. Unfortunately, the tendency of Xis to associate/aggregate when bound to short DNA duplexes (a 20 bp duplex containing the X1 site was used in the current work) does not allow a direct structural investigation of the protein–DNA complex. In our NMR experiments, no changes in the positions of the protein amide and DNA imide resonances were observed when one of the macromolecules was added in excess. [15N,1H]-TROSY spectra of the protein bound to DNA only showed HN resonances of C-terminal residues (Leu52–Ser71). These results support the previously suggested hypothesis that only the N-terminal, structured part of Xis is involved in the recognition of DNA [18,57]. On the other hand, the results also indicate that the C-terminal α-helix, predicted in the work of Wu and co-authors for Xis residues 59–64 , is not induced by the Xis–DNA interaction.
We have determined the solution-state structure of the full-length excisionase from λ-like coliphage HK022, containing the single amino acid substitution Cys28→Ser. This structure was shown to be stable under a wide range of experimental conditions and to be identical to the wild-type HK022 Xis structure; the denaturation of HK022 Xis is reversible and cannot be the origin of the previously observed irreversible inactivation of Xis in vivo and in vitro. The organization of the secondary structure elements in HK022 Xis is slightly different compared with the closely related bacteriophage λ excisionase; the presence of an additional antiparallel β-strand at the N-terminus (residues 2–4) makes the HK022 Xis structure more similar to the Mu-repressor DNA-binding domain.
We have found that the triproline segment (residues Pro32–Pro33–Pro34) adopts a cis–trans–trans conformation. This region could therefore play a key role in the specific protein–protein interaction that occurs during excisive recombination, similar to other proteins that display a cis–trans–trans triproline motif at the surface.
According to the NMR data, the C-terminal part of Xis is definitely not involved in the protein–DNA interaction, but it might serve as a specific site for the Xis–Int interaction that initiates excision. Moreover, the Asn66–IsoAsp isomerization at the C terminus could be involved in the protein self-regulation in vivo.
We gratefully acknowledge support from the Frankfurt University Center for Biomolecular Magnetic Resonance. We thank Prof. E. Yagil for making the pPG14 plasmid available to us and Dr G. Küllertz (Max Planck Research Unit, Halle, Germany) for a database search of the cisPro-Pro-Pro motif. P.P. thanks the Ministry of Education, Science and Sport of Slovenia for financial support. V.R. thanks Dr S. Potekhin for helpful discussion.