Interrogation of 3D-swapped structure and functional attributes of quintessential Sortase A from Streptococcus pneumoniae

The anchoring of the surface proteins to the cell wall in gram-positive bacteria involves a peptide ligation reaction catalyzed by transpeptidase sortase. Most bacterial genomes encode multiple sortases with dedicated functions. Streptococcus pneumoniae (Sp) carries four sortases; a housekeeping sortase (SrtA), and three pilin specific sortases (SrtC1, C2, C3) dedicated to the biosynthesis of covalent pilus. Interestingly, SrtA, meant for performing housekeeping roles, is also implicated in pilus assembly of Sp. The allegiance of SpSrtA to the pathogenic pilus assembly makes it an ideal target for clinical inhibitor development. In this paper, we describe biochemical characterization, crystal structure and peptide substrate preference of SpSrtA. Transpeptidation reaction with a variety of substrates revealed that the enzyme preferred elongated LPXTG sequences and transferred them equally well to both Ala- and Gly-terminated peptides. Curiously, crystal structure of both wild type and an active site (Cys to Ala) mutant of SpSrtA displayed inter-twined 3D-swapped dimers in which each protomer generated a classic eight stranded beta-barrel “sortase fold”. Size-exclusion chromatography and sedimentation equilibrium measurements revealed predominant presence of a dimer in equilibrium with its monomer. The crystal structure-based Cys-Cys distance mapping with defined chemical cross-linkers established the existence of 3D-swapped structure in solution. The swapping in SpSrtA, unprecedented for sortase family, may be physiologically relevant and meant to perform regulatory functions. Sp: Streptococcus , SrtA: Sortase A, Streptococcus pneumoniae Δ59SpSrtA: construct with Δ81SpSrtA: SpSrtA construct with truncation of from RP-HPLC: reverse-phase high performance liquid chromatography, Fmoc: Fluoromethoxycarbonyl, NMP: N-methyl-2-pyrrolidone. RrgA (Rlr-regulated gene A), RrgB (Rlr-regulated gene B)


Introduction
Streptococcus pneumoniae (S. pneumoniae) is a gram-positive bacterium and a commensal of human nasopharyngeal cavity. Under favourable conditions, S. pneumoniae can infect other host tissues and cause pneumonia, meningitis, endocarditis, and cellulites [1]. The surface proteins of S. pneumoniae play critical roles in bacterial pathogenesis by facilitating attachment of the bacterium to the host and promoting degradation of host-tissue components [2][3][4][5][6] Several surface proteins of S. pneumoniae, and other gram-positive bacteria, contain a LPXTG type of pentapeptide sequence in their C-terminal region that serves as a recognition motif for a class of cysteine transpeptidases called sortase [7].The bacterial genomes generally encode multiple sortases including a ubiquitous housekeeping sortase A (SrtA) which covalently anchors the surface proteins to the cell wall peptidoglycan [8].
Other sortases are meant for dedicated functions, such as, assembly of large protein appendages built by covalent linking of one or more pilin subunits [9]. The sortase-mediated covalent anchoring of proteins proceeds through a thioacyl-enzyme intermediate generated by nucleophilic attack of the thiol group of the catalytic Cys residue on the scissile T-G peptide bond of the LPXTG motif [10]. Subsequent reaction of the thioacyl-enzyme intermediate with the terminal amine group of the pentaglycine/dialanine branch of the peptidoglycan results in the formation of a peptide bond leading to the covalent attachment of the surface protein to the peptidoglycan. Sortases involved in pilin assembly use epsilon amine of a specific Lys residue present in the pilin subunit to resolve the thioacyl-enzyme intermediate and link the pilin chains by an isopeptide bond [9].
(Δ59SpSrtA) from the genomic DNA of S. pneumoniae Strain R6 using specific primers (Table S1). The PCR product was digested with enzymes NdeI and HindIII and subsequently ligated into NdeI/HindIII digested pET28c vector. The Δ81SpSrtA (residue 82-247) was subcloned from pET28c-Δ59SpSrtA construct, using primers listed in Table S1, following the same protocol as above. DNA sequencing was done to verify the identity of individual clones.
The respective constructs were transformed into E. coli BL-21 cells for expression.
The transformed cells were grown in LB media, containing kanamycin (50 μg/ml) till mid-log phase (OD600~0.5) at 37 °C. Protein expression was induced by addition of 0.5 mM IPTG and grown at 30 °C for 5 hrs. The cells were harvested, re-suspended in a buffer composed of 10 mM Tris-HCl (pH 7.5), 40 mM NaCl, 2 mM β-mercaptoethanol and lysed by sonication.
The supernatant was collected after sonication and purified using Ni-NTA based affinity chromatography as described previously [17].

Site-directed mutagenesis
Site-directed mutagenesis was carried out according to the QuickChange (Stratagene) protocol. Briefly, mutation was introduced by PCR using pET28c-Δ81SpSrtA as the template DNA and mutagenic primers to introduce desired mutation in the template DNA. The PCR product was digested with Dpn1 for 1 hr at 37 °C and the reaction mixture was transformed into E. coli DH5α cells. Plasmid DNA was isolated from E. coli DH5α cells and mutation was confirmed by DNA sequencing. The plasmid DNA was transformed into E. coli BL-21 cells for expression. The mutagenic primers are listed in Table S2.

Synthesis and purification of peptide substrates
Peptides were synthesized using standard solid phase method using Fmoc chemistry on a peptide synthesizer (Applied Biosystems, ABI 433A or Advanced Chemtech, Model ACT90).

Crystallization, diffraction data collection and processing
The Ni-NTA purified protein was desalted on a PD-10 column, concentrated using a Millipore Amicon filters (10 kDa cut-off) and used for crystallization. Conditions supporting the growth of wild type Δ81SpSrtA crystals were initially ascertained using the hanging-drop vapordiffusion and micro-batch under-oil methods at 20 °C with commercially available crystallization screens as reported previously [17]. The best diffracting crystal was grown using the under-oil method in a solution containing 25 mg/mL of protein (10 mM Tris buffer, pH 7.5), 0.2 M tri-ammonium citrate and 20% (w/v) PEG 3350, pH 7.0 along with 40% v/v (±)-1,3-butanediol as an additive agent. Diffraction quality crystals of the Δ81SpSrtA(C207A) mutant protein were grown in a condition containing 0.2 M tri-ammonium citrate and 20% (w/v) PEG 3350, pH 7.0, with 1.0 M guanidine hydrochloride as an additive agent. Crystals grew to their maximum dimensions in 25-30 days for the wild-type and 5-7 days for the mutant.
X-ray diffraction data from the wild-type protein crystal were collected at home source (λ = 1.5418 Å) without any cryoprotectant, whereas diffraction data for the mutant protein (C207A) crystal were collected at the synchrotron (λ = 0.9772 Å, BM-14, ESRF) using 20% glycerol as cryoprotectant. Crystals of the wild type and mutant protein diffracted to 2.70 Å and 2.48 Å, respectively. The diffraction data for both the crystals were processed using MOSFLM [18] and scaled using SCALA from the CCP4 program suite [19]. Data-collection and processing statistics for the crystals are presented in (Tables1 and 2). The quality of data sets was assessed using SFCHECK [20] which suggested no twinning in the data.

Structure solution, refinement and analysis
Structure of Δ81SpSrtA was solved by molecular replacement using Phaser [21] with the homologous S. pyogenes SrtA (78% sequence identity, PDB id: 3fn5, [22]) as the search model. The molecular-replacement (MR) solution yielded four molecules in the crystal asymmetric unit, consistent with expectation from Matthews coefficient and solvent content.
The MR solution was subjected to 10 cycles of rigid body refinement followed by several cycles of restrained refinement using the program REFMAC5 [23] from the CCP4 package [19], with alternate rounds of inspection and manual model building in COOT [24]. A few rounds of simulated annealing were also performed using phenix.refine from the PHENIX suite [25] followed by TLS refinement in REFMAC5. 5% of the measured reflections were kept aside for calculating Rfree [26]. The electron density map showed breaks in the backbone between P187 and R189 in all four chains. The structure was finally modeled as a swapped dimer, which led to a convergence of Rwork/Rfree (18.10% / 23.37%) ( Figure S1).
The Δ81SpSrtA(C207A) model was built using molecular replacement with S. pyogenes SrtA as model, and not the wild type enzyme, since 3D-swapping was not assured. Subsequently, the model was refined using the same protocol employed for Δ81SpSrtA. 3D-swapping was observed in the 2Fo-Fc and Fo-Fc maps of the mutant protein as well. The final model was refined with Rwork = 18.23% and Rfree = 22.48%.
The stereochemical acceptability of the structures was analysed using MOLPROBITY [27] and validated using PROCHECK [28] Structural superposition was performed using SUPERPOSE [29]. Domain motions and dimer interface statistics were evaluated using the DYNDOM server [30] and PISA server [31] respectively. Coordinates and structure factors for Δ81SpSrtA and Δ81SpSrtA(C207A) have been deposited in the Protein Data Bank with accession codes 4o8l and 4o8t, respectively.
The hinge region was estimated using the method proposed by Shingate and Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200631/897978/bcj-2020-0631.pdf by guest on 16 December 2020 Sowdhamini [32]. First, the swapped monomer structure of Δ81SpSrtA was superposed with the unswapped monomer from the closest homolog S. pyogenes; the swapped domain remained unaligned. The unaligned C-terminal stretch was then superposed separately with the equivalent region from S. pyogenes SrtA. Stretches that were unaligned in both instances were likely to form the hinge region (Table S4). Further, in the swapped dimer, the hinge loop is 'extended', whereas in the unswapped homolog it forms a turn structure; hence the difference in backbone torsion angles was also evaluated using the method proposed by Bennett and co-workers [33]. We calculated delta-torsion as √(∆φ 2 + ∆ψ 2 ), where φ, ψ are the backbone torsion angles. The stretch of non-superposable residues with deltatorsion > 30° (Table S5) was considered as the hinge region.

Transpeptidation assay
Transpeptidation reaction between LPXTG (donor peptide substrate) and AAKY/GGGKY  (Table S6). For, assessing the effect of Ca 2+ on Δ81SpSrtA activity, transpeptidase reaction was carried out as above but in the presence of 5 mM CaCl2.

Size exclusion chromatography
Size-exclusion chromatography was carried out by FPLC on a Superdex200 (10/300 GL) column. The column was pre-equilibrated with degassed buffer composed of 50 mM Tris Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200631/897978/bcj-2020-0631.pdf by guest on 16 December 2020 (pH 7.5) containing 150 mM NaCl at a flow rate of 1 ml/min. Protein sample was loaded on the column using a 100 μl injection loop and elution was monitored at 280 nm. The fractions corresponding to individual peaks were pooled and concentrated using protein concentrator (Amicon, Millipore, 10 kDa MW cut-off).

Analytical ultracentrifugation
Sedimentation equilibrium experiments were performed using an Optima XL-A analytical ultracentrifuge equipped with an An60Ti rotor (Beckman Inc). Sedimentation equilibrium studies were carried out at 13000, 17000, 23000, and 27000 rpm, at 20 C using six-channel charcoal-filled centerpieces. Data were collected by scanning samples at 280 nm with the resolution of 0.003 cm and average of 5-7 scans per step. The partial specific volume and solvent density were calculated using SEDNTERP. Equilibrium data were edited with WinREEDIT and analyzed by nonlinear least squares using WINNONLIN and subsequent simulations were performed using SCIENTIST (Micromath, Salt Lake City, UT, USA).
Absorbance profiles were analyzed using a sum of two exponential model which describes the monomer-dimer equilibrium as given below (eq 1-4).

2M
M2 ( where , is the extinction coefficient of monomer, and l is the path length of the sedimentation cell (1.2 cm).

N-terminal truncated versions (Δ59SpSrtA and Δ81SpSrtA) of SpSrtA are active
The amino acid sequences of SrtA from the pathogenic S. pneumoniae (TIGR4 strain) and non-pathogenic R6 strain are identical except for a single amino acid (N233D) substitution. SrtA sequence corresponding to residues, 60-247 and 82-247, respectively, was expressed in E coli as described previously [17]. Both the constructs contained a hexa-His tag and were purified using Ni-NTA affinity chromatography ( Figure S2). The ES-MS mass of purified Δ59SpSrtA was found to be 23493 Da which was in accord with the calculated mass of the protein (23495 Da) without a Met residue ( Figure S3). Likewise, Δ81SpSrtA yielded a mass of 20949 Da that fit the computed mass of the des-Met protein.
Thus, N-terminus Met residue was processed by E. coli aminopeptidases in both constructs of SpSrtA. pneumoniae. Accordingly, sortase-mediated transpeptidation reaction was examined with native and appropriately modified LPXTG pentapeptides using AAKY and GGGKY as nucleophile acceptors ( Figure 1A, Figure S4).
RPHPLC analyses of the transpeptidation reaction mixture carried out in the presence of ∆59SpSrtA with native LPNTG (free termini) and AAKY or GGGKY as substrates did not yield any product. The results were similar and no product was seen when either end of LPNTG pentapeptide (acetylation of the amino terminus or amidation of the carboxyl end) was modified. However, blocking of both the ends of the pentapeptide motif (Ac-LPNTG-NH2) facilitated the transpeptidation reaction with both AAKY and GGGKY in a similar fashion leading to a product yield of about 8%.
Next, the effect of Ala residue at the LPNTG termini was evaluated using ALPNTGA.
Under identical conditions as above, heptapeptide ALPNTGA produced transpeptidation yields of 10-12% which was marginally higher than the yield for Ac-LPNTG-NH2. Further, we placed Gln at N-terminus of LPNTG because Gln residue precedes the LPNTG sorting Subsequently an Ala residue was added at the N-terminus of QLPNTGA to see if extension of the motif has any bearing on the transpeptidation reaction. Interestingly, AQLPNTGA peptide produced about 25% yield that was almost two-fold or more than that obtained when termini modified LPNTG or ALPNTGA or QLPNGA peptides were used as donor substrates. Finally, we extended the AQLPNTGA sequence by placing a Tyr residue at the N-terminus to further evaluate the effect of peptide length on transpeptidation reaction as also to use Tyr residue for spectroscopic estimation of peptide concertation. The transpeptidation yield of about 28-30% associated with YAQLPNTGA was marginally higher than that obtained with AQLPNTGA. Taken together, the above results indicate that Δ59SpSrtA preferred relatively longer peptide substrates. This feature of S. pneumoniae SrtA was in sharp contrast with S. aureus SrtA which was shown to process Ac-LPNTG-NH2 as effectively as the octapeptide YALPNTGK [34].
Next, we assessed the transpeptidase activity of Δ81SpSrtA to see if further truncation (22 residues) was detrimental to enzyme activity. The transpeptidation assay was carried out using YAQLPNTGA as the substrate peptide and AAKY or GGGKY as the acceptor peptide ( Figure 1B). Analysis of transpeptidation reaction with both the acceptor peptides proceeded in a similar fashion, and the transpeptidation reaction was found to be independent of calcium ion ( Figure 1C). Thus, both versions of the enzyme namely, Δ81SpSrtA and Δ59SpSrtA, were active.

SpSrtA crystallizes as a 3D-swapped dimer
Both truncated forms of sortase, namely, Δ59SpSrtA and Δ81SpSrtA, could be crystallized. However, Δ59SpSrtA crystals were recalcitrant to diffraction. The preliminary account of crystallization and data collection at 2.9 Å resolution for the wild type Δ81SpSrtA was reported earlier [17]. In the present work, diffraction data of Δ81SpSrtA crystals was collected at 2.7 Å (data collection and processing statistics in Table 1, 2). The crystal asymmetric unit contains two independent homodimers which are not related by Unlike the structures of other SrtA homologs (from Streptococcus pyogenes, PDB id: 3fn5; Streptococcus agalactiae, PDB id: 3rcc [35] and Staphylococcus aureus, PDB id: 1t2p [36] (X-ray structure), 2kid [37] (NMR structure)), the Δ81SpSrtA monomer alone does not form the conserved eight-stranded beta-barrel sortase-fold, which is unique to the sortase superfamily. One monomer exchanges the C-terminal end, consisting of the β7 and β8 strands and a helix, with another monomer, thereby forming a 3D-swapped dimer. There are two such dimers in the asymmetric unit, comprising of chains A and B, and chains C and D respectively. Strands 1-6 from one monomer and the swapped strands 7-8 from the other monomer constitute the 8-stranded beta-barrel fold. Residues Pro187, Asp188, Arg189 and Val190 in the β6/β7 loop are estimated to form the hinge region based on the differences in structural superposition with its closest unswapped homolog, S. pyogenes SrtA (Table S4), and also from the differences in torsion angles of the equivalent residues between the two structures (Table S5). It may be noted that Pro residues have been found to be present in the hinge loop in many 3D-swapped proteins observed so far.
Structural superposition of the four Δ81SpSrtA monomers show that the stretch from A strong dimer interface is formed between chains A and B, and between chains C and D, whereas weaker interfaces (crystallographic interfaces) are formed between A and C, A and D, and B and C. This is evident from the large interface area (> 10000 Å 2 ) and large number of inter-chain hydrogen bonds and van der Waals contacts across the AB and CD interfaces (interface statistics in Table S7). As a result of the dimerization via 3D-swapping, a secondary interface is formed by residues 113-120 (in β2/β3 loop), 125 (in β3/β4 loop), 143-146 (part of β4/H4 loop after active site His141), 187-190 (hinge region in β6/β7 loop), residues 205-212 (β7/β8 loop containing active site Cys207) and active site Arg215 ( Figure   2C). Interestingly, the equivalent regions (β2/β3, β4/H1 and β7/β8 loops) in Bacillus anthracis SrtC (PDB id: 2ln7) have been implicated in the formation of a transient dimer [38].
It is pertinent to mention here that the structure of S. pneumoniae D39 Sortase A (PDB id: 5dv0) which shares 98% sequence identity with our SpSrtA (4o8l), and used 4o8l The data for 5dv0 has been reported at a poor resolution of 3.3 Å and the description of the structure has not been published thus far. However, PISA analysis shows that the assembly of 5dv0 represents a thermodynamically stable homodimer with a buried surface area of 10710 Å 2 (∆G of -61.0 kcal/mol) which is similar to the present (4o8l) structure (Table S7).
The structure of 5dv0 when viewed as biological assembly with the cyclic symmetry C2 in PyMol reveals a 3D-swapped structure arising from exchange of the β7-β8 strands of the monomer ( Figure S6).

Structural features of the active site
The active site of sortases is comprised of a catalytic triad of highly conserved His, Cys and Arg residues. In prototypic S. aureus SrtA, the active site is composed of Cys184, His120 and Arg197.The equivalent residues in SpSrtA are His141 (at C-terminal of β4), Cys207 (at C-terminal of β7) and Arg215 (at N-terminal of β8). We carried out mutation of each putative active site residue and generated three single mutants (H141A, C207A and R215A respectively) of Δ81SpSrtA. The mutants were found to be inactive in standard transpeptidation assays, establishing the critical role of individual residues in catalysis.
The Cys207 residue is found to be oxidized to its sulphenic acid form in chain B ( Figure S7); a similar feature has been observed for one of the structures of S. pyogenes SrtA (PDB id: 3fn6). 3D-swapping leads to two sets of active site residues at the open interface ( Figure 3A). His141 from one monomer, and Cys207 and Arg215 from the other monomer of a swapped dimer constitute a pair of active site residues. In the present structure, the Arg215 residue is partially exposed, whereas the Cys207 and His141 residues are buried, unlike the SrtA from S. pyogenes and S. agalactiae. The superposition of the active site residues from the four chains in the crystallographic asymmetric unit depicts some flexibility in the side chains of Cys207 and Arg215 ( Figure 3B) but that of His141 is rigid. Of the two catalytic Arg residues (Arg215) at the dimeric interface, one is found to interact with

Comparison with other known SrtA structures
The individual domains in the swapped dimer are similar to the sortase fold observed in homologous structures. SpSrtA has the highest sequence similarity with SrtA from S. pyogenes (78% sequence similarity, 67% identity; PDB id: 3fn5), followed by that from S. agalactiae (77% sequence similarity, 58% identity; PDB id: 3rcc). Figure 4 shows the structure-based sequence alignment of SpSrtA with the other sortases of known structure.
The RMSD for superposition of chain B of 3fn5 on the larger domain of chain A (up to the

Absence of a Ca 2+ binding site
The prototype sortase, S. aureus SrtA has a flexible β6/β7 loop. A 310 helix is formed in β6/β7 loop upon substrate-binding, which immobilizes the loop and leads to the formation of hydrophobic contacts around the Leu in LPXTG sorting motif. This is further stabilized by Ca 2+ -binding to the C-terminus region of the loop [39]. This has led to the idea of an 'induced fit' mechanism for its activity. In contrast, S. pyogenes SrtA and S. agalactiae SrtA have a preformed substrate-binding site that doesn't require large scale conformational changes for substrate-binding [40]. hydrogen bonds with Gln129 and Asp195, and with Asn135 instead of Glu133, which has moved away from the equivalent position occupied by Asp112 in S. aureus SrtA. The above description is consistent with the calcium-independent transpeptidation activity of Δ81SpSrtA shown in Figure 1C. The S. pyogenes SrtA behaves in much the same way as

Comparison of S. aureus
Δ81SpSrtA and does not require Ca 2+ for its activity. The equivalent residues in S. pyogenes SrtA are Lys126, Gln129, Gly133 and Asp196.

Δ81SpSrtA exists predominantly as a dimer in solution
We performed size-exclusion chromatography to assess the oligomerization status

DISCUSSION
Sortase enzymes are implicated in bacterial pathogenesis, and also serve as effective protein labelling tool [8]. Therefore, structure and specificity elucidation as a prelude to developing clinical inhibitors or improved sortases for protein engineering applications remains a topical subject in sortase research. Here we report the peptide substrate specificity and crystal structure of housekeeping SrtA from S. pneumoniae. The housekeeping SrtA encoded, along with pilus subunits (RrgA, B, and C) and pilin sortases (SrtC1, SrtC2, SrtC3), in the rlrA pathogenicity islet is apparently the most crucial transpeptidase among S. pneumoniae sortases because of its additional role in the assembly of pathogenic pilus [11].
We generated two N-terminal truncated versions of SrtA ( 59SpSrtA and 81SpSrtA) to explore the limit of truncation for expression of a soluble protein endowed with catalytic activity. We examined the donor substrate specificity of the above constructs with the classic LPNTG motif using AAKY or GGGKY as an acceptor peptide. Curiously, end-capped LPNTG pentapeptide (acetylated/amidated) or Ala extended heptapeptide (ALPNTGA), which was earlier shown to be a robust substrate for archetypal housekeeping SrtA of S. aureus [34], turned out to be poor substrates for 59SpSrtA. In contrast, transpeptidation yield with an octapeptide (AQLPNTGA) or a nonapeptide (YAQLPNTGA) was found to be almost two-fold higher (13% vs 28%) than that of the heptapeptide. Interestingly, 81SpSrtA also displayed a donor LPXTG substrate preference similar to 59SpSrtA and produced comparable transpeptidation yields against both, AAKY and GGGKY, acceptors. Notably, only Gly-based amine-acceptors, such as GGGKY, are effective substrates for quintessential housekeeping SaSrtA. Besides, unlike SaSrtA, transpeptidation reaction of SpSrtA was independent of calcium ion. The cumulative results indicate that substrate recognition propensity of SpSrtA is quite different from quintessential housekeeping SrtA of S. aureus [34].
Although both constructs of SpSrtA were endowed with catalytic activity, only 81SpSrtA produced good quality crystals for structure analysis. Interestingly, 81SpSrtA displayed a 3D-swapped structure comprising of two independent homodimers in the asymmetric unit wherein each protomer generated a characteristic 'sortase fold' by The occurrence of a 3D-swapped dimer as observed in SpSrtA is not known among the sortase family of transpeptidases but dimer-monomer equilibrium has been reported in some cases [42][43][44]. Sortase A of S. aureus (SaSrtA) has been shown to exist in a monomerdimer equilibrium in solution [42]. However, functional evaluation of individual species of SaSrtA produced curious results [45];

Competing interest
The authors declare no conflict of interest.

Funding
This work was supported by core grants to the National Institute of Immunology from the . Each dimer contains two characteristic 8-stranded beta barrel "sortase fold", but each fold in the dimer is constituted by the combination of beta-strands from both the chains, that is, a complete fold is made through 3D-swapping. Folded catalytic domains are labeled as protomer-A, protomer-B, protomer-C and protomer-D depending upon the chain identifier contributing first six beta strands. B) Flexibility of the monomers in the 3D-swapped dimer. Superposition of the larger domain (β1-β6 strands) in the swapped monomers chain A (Green) and chain C (Blue) through DynDom server (Hayward et al., 1997) showed a rotation of ~29º around the rotation axis C) Primary and Secondary interfaces in the swapped dimer. The primary interface is formed by sections of two monomers in the 3Dswapped structure, whereas it would be part of the same monomer in an unswapped structure. The secondary interface is the new interface formed due to the proximity of the two protomers in the dimer.  are shown on top of the sequence alignment, where coils and arrows represent helices and strands, respectively. α, β, and TT correspond to α-helix, β-strand, 310-helix and β-turn respectively. The residues shown with blue background correspond to the active site residues. Sequence alignment was done with ClustalW (http://www.ebi.ac.uk/clustalw) and coloured using ESPRIPT [46].   Rmerge † (%) 9.4 (52.9) 9.2 (54.7) Overall B factor from Wilson plot (Å 2 ) 66.5 53.9 No. of molecules in the asymmetric unit 4 (Two dimers) 6 (Three dimers)