Crystal structures of the amino-terminal domain of LpoA from Escherichia coli and Haemophilus influenzae

New crystal structures of the N-terminal domain of LpoA from Escherichia coli and Haemophilus influenzae revealed the ability of this domain to flex between relatively flat and curved structures and to vary the width of its central groove.

Abstract

The bacterial periplasmic protein LpoA is an outer membrane lipoprotein and an activator for the cross-linking activity of PBP1A, a bifunctional peptidoglycan synthase. Previous structures of the amino-terminal (N) domain of LpoA showed it to consist entirely of helices and loops, with at least four tetratricopeptide-like repeats. Although the previously determined ortho­rhombic crystal structure of the N domain of Haemophilus influenzae LpoA showed a typical curved structure with a concave groove, an NMR structure of the same domain from Escherichia coli was relatively flat. Here, a crystal structure of the N domain of E. coli LpoA was determined to a resolution of 2.1 Å and was found to be more similar to the H. influenzae crystal structure than to the E. coli NMR structure. To provide a quantitative description for these comparisons, the various structures were superimposed pairwise by fitting the first half of each structure to its pairwise partner and then calculating the rotation axis that would optimally superimpose the second half. Differences in both the magnitude of the rotation and the direction of the rotation axis were observed between different pairs of structures. A 1.35 Å resolution structure of a monoclinic crystal form of the N domain of H. influenzae LpoA was also determined. In this structure, the subdomains rotate 10° relative to those in the original orthorhombic H. influenzae crystal structure to further narrow the groove between the subdomains. To accommodate this, a bound chloride ion (in place of sulfate) allowed the closer approach of a helix that forms one side of the groove.

1. Introduction  

Bacterial cell walls are primarily comprised of peptidoglycan (PG), a polymer of disaccharides cross-linked through peptide substituents. PG provides protection against internal osmotic pressure (turgor), functions in cell division and defines cell shape. Besides the recently characterized PG polymerase RodA encoded by most bacteria (Meeske et al., 2016 ▸), Gram-negative bacteria also require a bifunctional enzyme, specifically either PBP1A or PBP1B, which can polymerize and cross-link PG and may play a role in PG repair. LpoA is an outer membrane lipoprotein that is essential for activating the cross-linking (transpeptidase) activity of PBP1A (Typas et al., 2010 ▸; Paradis-Bleau et al., 2010 ▸). LpoA is composed of two domains: a primarily helical amino-terminal (N) domain and a carboxyl-terminal domain involved in interactions with PBP1A. Analyses of multiple crystal structures of full-length LpoA from Haemophilus influenzae (HiLpoA), together with small-angle X-ray scattering results, suggested that LpoA is a flexible molecule that is capable of extending through gaps or holes in PG in order to interact with the inner-membrane-bound PBP1A (Sathiyamoorthy et al., 2017 ▸). Although the interdomain linker is responsible for much of the flexibility of the molecule, twists and bends of up to 5° have been observed within the N domain and are likely to affect the overall length of the molecule.

The structures of the N domain from both E. coli LpoA (EcLpoA) and HiLpoA contain repeats of the helical tetratricopeptide-like (TPR-like) motif, which is a helix–turn–helix motif that is often involved in protein–protein interactions. Consecutive TPR-like motifs typically induce a superhelical twist to form a concave groove that is ideal for binding peptide segments of other proteins (Grove et al., 2008 ▸). For example, a conformational change of the TPR-containing protein MamA is associated with binding of its putative ligand, suggesting possible mechanisms for TPR-regulated functionality (Zeytuni et al., 2011 ▸).

The structure of the N domain of E. coli LpoA was first determined using NMR (EcNMR) and, although composed of TPR-like helix–turn–helix motifs, was observed to be relatively flat without a significant superhelical twist (Jean et al., 2014 ▸). When the orthorhombic crystal structure of the N domain of H. influenzae LpoA (HiOrt) was reported (Sathiyamoorthy et al., 2017 ▸), its subdomains 1 and 2 were noted to each be very similar to the corresponding sub­domains in EcNMR. However, when the two intact N-domain structures were superimposed, based solely on least-squares fitting of their subdomains 1 (specifically residues 33–148), their subdomains 2 were related by a 45° rotation (see Fig. 2 in Sathiyamoorthy et al., 2017 ▸).

To better understand the range of conformations accessible to the LpoA N domain, we set out to determine its structure in additional crystal environments. We determined the tetragonal crystal structure of the N domain of EcLpoA (EcTet) and found it to be closer in structure to HiOrt than to EcNMR. We also determined a 1.35 Å resolution structure of the N domain of HiLpoA in a monoclinic space group; inspection of this structure showed it to have a more concave cleft than the previously reported orthorhombic form (PDB entry 4p29; Sathiyamoorthy et al., 2017 ▸).

2. Materials and methods  

2.1. Macromolecule production  

To express the EcLpoA N domain, the gene corresponding to residues 31–252 of EcLpoA was amplified using PCR from genomic DNA (E. coli strain MG1655) and cloned into pMCSG7 (Stols et al., 2002 ▸). The primers used to amplify the domain from EcLpoA and to linearize the pMCSG7 plasmid are shown in Table 1 ▸. The resulting construct contained an N-terminal His₆ tag and TEV protease cleavage site (detailed in Table 1 ▸). The construct was confirmed by carrying out Sanger sequencing in both the forward and reverse directions.

Table 1. Expression plasmid construction.

	EcTet (31–252)	HiMon (33–253)
Source organism	E. coli MG1655	H. influenzae Rd
DNA source	pCB40 plasmid DNA (Paradis-Bleau et al., 2010 ▸)	H. influenzae Rd genomic DNA
Cloning method	Isothermal (NEBuilder HiFi, NEB)	Restriction enzymes
Forward primer	agaacctgtacttccAATCCACTCCCGATCAGTCCACTG†	CATGCCATGGCGAATTTCACGCAAACCTTACAA‡
Reverse primer	cgttatccacttccaatattttaTTTAAACGCTTTTACGTTAACCAAC†	GCCGACGTCGACTTGTTGGAAATTAAGCAATGTAAG§
Forward primer to amplify plasmid	taaAATATTGGAAGTGGATAACGGATC
Reverse primer to amplify plasmid	GGATTGGAAGTACAGGTTCTCG
Cloning vector	pMCSG7 (Stols et al., 2002 ▸)	pETBlue-2 (Novagen)
Expression vector	pMCSG7	pETBlue-2
Expression host	E. coli Origami 2(DE3) (Novagen)	E. coli Tuner(DE3)/pLacI (Novagen)
Complete amino-acid sequence of the construct produced	mhhhhhhssgvdlgtenlyfqSTPDQSTAYMQGTAQADSAFYLQQMQQSSDDTRINWQLLAIRALVKEGKTGQAVELFNQLPQELNDAQRREKTLLAVEIKLAQKDFAGAQNLLAKITPADLEQNQQARYWQAKIDASQGRPSIDLLRALIAQEPLLGAKEKQQNIDATWQALSSMTQEQANTLVINADENILQGWLDLQRVWFDNRNDPDMMKAGIADWQKRYPNNPGAKMLPTQLVNVKAFK¶	MANFTQTLQKDANASSEFYINKLGQTQELEDQQTYKLLAARVLIRENKVEQSAALLRELGELNDAQKLDRALIEARISAAKNANEVAQNQLRALDLNKLSPSQKSRYYETLAIVAENRKDMIEAVKARIEMDKNLTDVQRHQDNIDKTWALLRSANTGVINNASDEGNAALGGWLTLIKAYNDYIRQPVQLSQALQSWKNAYPNHAAATLFPKELLTLLNFQQVEHHHHHH ††

^†The primer 5′ extensions shown in lower case base-pair with sequences on the linearized plasmid for the isothermal cloning procedure.

^‡The NcoI site for cloning is underlined.

^§The SalI site for cloning is underlined.

^¶Lower case indicates the sequence in EcTet (31–252) that was removed after TEV proteolysis.

^††The underlined N-­terminal MA and C-terminal His₆-tag residues were added to the native sequence during cloning.

The pMCSG7-EcLpoA-N plasmid was transformed into E. coli Origami 2(DE3) cells (Novagen). The cells were grown in two 2.8 l flasks, each containing 450 ml Terrific Broth. After inoculation and initial growth, the flasks were transferred to an incubator at 22°C and IPTG was added to 0.45 mM for induction. The cells were harvested 16 h after induction by centrifugation at 5900g, after which they were resuspended with DNase and protease-inhibitor tablets and frozen at −80°C. The cells were lysed via three passes through an Emulsiflex-C3 homogenizer at 4°C. The clarified lysate was applied onto a 5 ml column packed with HisPur Cobalt Resin (Thermo Fisher), washed and then eluted with 180 mM imidazole. The protein was treated with TEV protease overnight at 4°C and then passed through a HisPur Ni–NTA Resin column (Thermo Fisher) to remove cleaved His₆ tag and TEV protease. Concentrated protein (25 mg ml⁻¹) was frozen in aliquots containing 10% glycerol. Before crystallization trials, the aliquots were passed through a gel-filtration column (Superdex HR 75 10/300) to remove glycerol. The predicted sequence of the mature EcLpoA N-domain protein was serine followed by residues 31–252 of EcLpoA (formula weight 25 154).

The cloning, expression and purification of residues 33–253 of the HiLpoA N domain were identical to the procedures described previously (Sathiyamoorthy et al., 2017 ▸). The gene fragment with NcoI and SalI restriction sites corresponding to residues 33–253 of HiLpoA was amplified using PCR from genomic DNA isolated from the H. influenzae Rd strain (ATCC catalog No. 9008). After digestion, the purified fragment was ligated into the pETBlue-2 (Novagen) plasmid. The plasmid encoding the N-terminal residues MA and the C-terminal His₆ tag was transformed into E. coli Tuner(DE3)/pLacI (Novagen) cells for expression. The formula weight of the purified protein was predicted to be 26 224. Details of the construct are shown in Table 1 ▸.

2.2. Crystallization  

Initial crystals of the N domain of EcLpoA grew from condition A1 of the Morpheus screen (Molecular Dimensions). The N domain of HiLpoA was previously reported to form orthorhombic crystals (Sathiyamoorthy et al., 2017 ▸). In the current work, we grew monoclinic crystals of the N domain of HiLpoA from different conditions. Crystals were harvested into precipitant containing 10% glycerol and were immediately cooled in liquid nitrogen. Crystallization information is given in Table 2 ▸.

Table 2. Crystallization.

	EcTet (31–252)	HiMon (33–253)
Method	Sitting-drop vapor diffusion	Hanging-drop vapor diffusion
Plate type	24-well sitting drop	Linbro (Hampton Research)
Temperature (K)	295	295
Protein concentration (mg ml−1)	19	10
Buffer composition of protein solution	150 mM NaCl, 25 mM Tris–HCl pH 7.5	0.1% β-mercaptoethanol, 1 mM EDTA, 0.1 mM benzamidine, 20 mM Tris–HCl pH 8.0
Composition of reservoir solution	0.03 M magnesium chloride hexahydrate, 0.03 M calcium chloride dihydrate, 5% glycerol, 10% PEG 20 000, 17% PEG monomethyl ether 550, 0.1 M Buffer System 1 pH 6.5†	20% polyethylene glycol (PEG) 10 000, 0.1 M HEPES pH 7.5
Volume and ratio of drop	2 µl, 1:1
Volume of reservoir	1 ml

^†Buffer System 1, 1.0 M stock solution: 30 ml 1 M 2-(N-morpholino)ethanesulfonic acid (MES) pH 3.1 was titrated with 24.1 ml 1 M imidazole pH 10.2 to a final pH of 6.5.

2.3. Data collection and processing  

Diffraction data were collected from the tetragonal crystals of the N domain of EcLpoA (EcTet) on the LS-CAT beamline 21-ID-D at the Advanced Photon Source (APS), Argonne National Laboratory and were processed with HKL-2000 (Otwinowski & Minor, 1997 ▸). Diffraction data for the monoclinic crystals of the N domain of HiLpoA (HiMon) were collected on the DND-CAT beamline 5-ID-B at APS and were processed with d*TREK (Pflugrath, 1999 ▸). Details of data collection and processing are presented in Table 3 ▸.

Table 3. Data collection and processing.

Values in parentheses are for the outer shell.

	EcTet (31–252)	HiMon (33–253)
Diffraction source	Beamline 21-ID-D, LS-CAT, APS	Beamline 5-ID-B, DND-CAT, APS
Wavelength (Å)	0.9772	0.9793
Temperature (K)	140	140
Detector	Dectris EIGER X 9M	MAR MX225 CCD
Crystal-to-detector distance (mm)	250	150.8
Rotation range per image (°)	0.5	0.8
Total rotation range (°)	200	180
Exposure time per image (s)	0.25	6
Space group	P43212	P21
a, b, c (Å)	70.001, 70.001, 97.802	68.988, 36.927, 95.359
α, β, γ (°)	90, 90, 90	90, 108.08, 90
Mosaicity (°)	0.30–0.49	0.55–1.03
Resolution range (Å)	22.7–2.10 (2.18–2.10)	34.50–1.35 (1.40–1.35)
Total No. of reflections	160658 (8220)	301389 (12818)
No. of unique reflections	14746 (1368)	91428 (6067)
Completeness (%)	99.23 (94.08)	90.3 (59.9)
Redundancy	10.9 (5.7)	3.28 (2.13)
〈I/σ(I)〉	17.14 (2.13)	13.1 (2.9)
R meas	0.116 (0.829)	0.055 (0.314)†
CC1/2	0.998 (0.715)	‡
Overall B factor from Wilson plot (Å2)	27	16

^†For the N domain of HiLpoA , the redundancy-independent merging R factor R _meas was calculated by multiplying the R _merge value of 0.046 by [N/(N − 1)]^1/2, where N is the data redundancy.

^‡Not available.

2.4. Structure solution and refinement  

Initial attempts to solve the structure of EcTet by molecular replacement in Phaser (McCoy et al., 2007 ▸; Adams et al., 2010 ▸) with search models containing each subdomain of EcNMR (PDB entry 2mhk; Jean et al., 2014 ▸) were unsuccessful. The two sub­domains of HiOrt (29% sequence identity to EcTet), after pruning side chains that were not common with EcTet in Sculptor (Bunkóczi & Read, 2011 ▸), were used as search models. Molecular-replacement calculations in Phaser resulted in a log-likelihood gain (LLG) of 312 and a translation-function Z-score (TFZ) of 17.1, and confirmed that the space group for these data was P4₃2₁2. The structure was manually built with Coot (Emsley et al., 2010 ▸) and refined in PHENIX (v.1.12; Adams et al., 2010 ▸) with data to 2.1 Å resolution (99% complete) and three TLS domains to a final R _work and R _free of 0.177 and 0.213, respectively. Waters were added and validated in Coot. Simulated-annealing OMIT maps were consulted to minimize the bias from the starting model (Supplementary Fig. S1). Details of the refinement and model validation are summarized in Table 4 ▸. The coordinates and structure factors were deposited in the PDB as entry 6dr3. Residue numbering corresponds to that of GenBank sequence NP_417616.1 and not to the sequence numbering for EcNMR (PDB entry 2mhk).

Table 4. Structure solution and refinement.

Values in parentheses are for the outer shell.

	EcTet (31–252)	HiMon (33–253)
PDB entry	6dr3	6dcj
Resolution range (Å)	22.7–2.10 (2.18–2.10)	34.44–1.35 (1.40–1.35)
Completeness (%)	99.23 (94.08)	90.52 (60.70)
σ Cutoff	None	None
No. of reflections, working set	13194 (1230)	91406 (6065)
No. of reflections, test set	1465 (137)	7328 (490)
Final R cryst	0.177 (0.249)	0.155 (0.228)
Final R free	0.213 (0.280)	0.190 (0.279)
No. of non-H atoms
Protein	1776	3652
Ion	0	2
Water	151	714
Total	1927	4368
R.m.s. deviations
Bonds (Å)	0.002	0.007
Angles (°)	0.44	1.17
Average B factors (Å2)
Overall	38	25†
Protein	38	22
Ion	n.a.	21
Water	40	35
Ramachandran plot
Most favored (%)	97.29	99.54
Allowed (%)	2.71	0.46
Rotamer outliers (%)	0.53	0.51
MolProbity score/clashscore	1.01/1.41	1.03/2.46

^†The HiMon structure was refined with anisotropic B factors. Isotropic equivalent values are shown in this table.

In solving the HiMon structure, molecular-replacement calculations with Phaser using the intact structure of HiOrt (PDB entry 4p29) as a search model (Sathiyamoorthy et al., 2017 ▸) were unsuccessful. Breaking the search model into two subdomains as above successfully placed two copies of each of the two subdomains in the asymmetric unit, consistent with two monomers. The LLG and TFZ values were 2681 and 33.3, respectively, for the molecular-replacement solution. As above, the HiMon structure was modified in Coot and refined with PHENIX with data to 1.35 Å resolution (90% complete) and anisotropic B factors to a final R _work and R _free of 0.155 and 0.190, respectively. TLS refinement was not performed as it was redundant with the anisotropic B factors. Hydrogens were included in their riding positions. Details of the refinement and model validation are summarized in Table 4 ▸ and an OMIT map is shown in Supplementary Fig. S2. Coordinates and structure factors were deposited in the PDB as entry 6dcj.

2.5. Structure comparisons  

All structural superimpositions were performed by least-squares fitting using the super routine of PyMOL (v.2.2; Schrödinger). This routine in PyMOL iteratively rejects paired residues if the interatomic distance exceeds two times the overall r.m.s. deviation. Object transformation matrices from super were converted into an equivalent angular rotation about a screw axis (GEM; E. Fauman, unpublished work), which was displayed as a cylindrical arrow using a compiled graphics object in PyMOL. To ascertain the difference in curvature (the relative difference in subdomain 2 rotation) between two LpoA N-domain structures, sub­domains 1 of the structures were first superimposed. Sub­domains 2 were then superimposed and the matrix from this latter transformation was converted into an equivalent rotation as above. All figures were prepared in PyMOL.

3. Results  

3.1. Comparison of EcTet and EcNMR (PDB entry 2mhk)  

As described in Section 2, the protein that was crystallized to form EcTet was designed to include Ser30 followed by residues 31–252 of EcLpoA. Two subdomains consisting of 13 helices and the loops between them were observed in EcTet, with subdomain 1 consisting of helices H1–H7 and spanning residues 30–147 and subdomain 2 consisting of helices H8–H13 and spanning residues 148–252 (Fig. 1 ▸). Residues 242–247 form one turn of an α-helix, while two turns were observed in both EcNMR and HiOrt. Only weak density was observed corresponding to residues 247–252, indicating that these residues were likely to be partially disordered.

Figure 1.

Cartoon representation of the EcTet domain, highlighting the TPR-like motifs in different colors. Left, top view looking into the groove located above helix H7. Right, side view showing the concave shape of the domain and the locations of the subdomains.

Subdomains 1 and 2 of EcTet were individually superimposed on the respective subdomains of EcNMR, and analysis of this superposition confirmed similar local folds (subdomain 1, r.m.s.d. on C^α atoms of 2.0 Å; subdomain 2, r.m.s.d. on C^α atoms of 1.5 Å; see Table 5 ▸). However, superimposing the full-length N domains showed quite a large difference in backbone structure (r.m.s.d. of 3.5 Å), with this difference explained by different rotations between the two subdomains in the two structures. The difference in domain curvature (calculated as described in Section 2) showed that EcNMR subdomain 2 required a 33° rotation around the axis displayed in Fig. 2 ▸ in order to optimally superimpose it onto EcTet. That is, in contrast to EcNMR, EcTet was observed to exhibit more of a superhelical twist and a more pronounced concave groove between the two subdomains.

Table 5. Summary of structural comparisons of LpoA N domains.

	R.m.s.d.† (Å)
Aligned structures	Overall	Subdomain 1	Subdomain 2	Rotation† (°)	Translation‡ (Å)
EcNMR/EcTet	3.7 (209/222)	2.0 (112/118)	1.5 (94/105)	33	0.4
HiOrt/EcTet	2.4 (199/221)	0.9 (82/118)	0.8 (77/104)	17	0.0
HiMon/EcTet	3.2 (206/221)	1.0 (99/116)	1.2 (77/104)	27	−1.0
HiOrt/HiMon	1.0 (207/221)	0.3 (96/116)	0.4 (80/105)	10	−0.5

^†Values in parentheses are (number of C^α atoms superimposed/total number of C^α atoms).

^‡The rotation of subdomain 2 along the inter-subdomain screw axis relative to subdomain 1.

^§The translation along the inter-subdomain screw axis.

Figure 2.

Superimposition of the EcNMR structure (gray) and the EcTet crystal structure (magenta) accomplished by fitting their subdomains 1. The green cylindrical arrow indicates the 33° rotation axis relating the subdomains 2 of the two superimposed structures. Right, side view showing the rotation axis end-on.

3.2. Comparison of the overall conformations of EcTet and HiOrt  

After aligning subdomain 1 of the EcTet crystal structure onto that of HiOrt, subdomain 2 of EcTet must be rotated by 17.4° around the axis shown in Fig. 3 ▸ to superimpose it on subdomain 2 of HiOrt. That is, the EcTet concave curvature, while more pronounced than that for EcNMR, was observed to be less prominent than that of HiOrt. A similar comparison with the high-resolution HiMon structure (see Section 3.3 below) required a 27° rotation around a similar axis in order for EcTet subdomain 2 to best superimpose on that of HiMon (Table 5 ▸).

Figure 3.

Superimposition of HiOrt (cyan) and EcTet (magenta) accomplished by fitting their subdomains 1. Right, side view showing the 17° rotation between the respective subdomains 2.

The differences between the overall conformations of EcTet and HiOrt coincided with, and may have originated from, differences between the junctions of their two subdomains, specifically the loop between helices H7 and H8. Inspection of HiOrt (cyan) showed its H7 and H8 helical axes to be parallel to each other and the loop between the helices (residues 148-RKD-150) to be relatively short (Fig. 4 ▸ a). Inspection of the EcTet structure (magenta) showed its H8 helix to make an angle of approximately 25° with H7 and the loop between them (residues 146-SQGRPSI-152) to be longer (Fig. 4 ▸ b). The inspection also showed its H7 helix to be about one turn shorter than H7 in HiOrt. The change in conformation is not explained by the differences in the loop sequence. The difference between the interhelix angles of EcTet and HiOrt appeared to be consistent with the different rotational relationships between their two subdomains.

Figure 4.

Helices H7 and H8 from both HiOrt (cyan) and EcTet (magenta) after fitting their subdomains 1. A front view (a) and side view (b) show H7 and H8 to be mostly parallel in HiOrt, but H8 is rotated by about 25° in EcTet, requiring a longer loop (residues 146–152) between the two helices.

3.3. Comparison of HiMon with HiOrt  

As for HiOrt (PDB entry 4p29; Sathiyamoorthy et al., 2017 ▸), the HiMon crystal structure was also found to contain two molecules per asymmetric unit. In HiMon, molecules A and B superimposed with an r.m.s.d. of 0.3 Å for 194 C^α atoms of a total of 211 residues in each molecule. HiMon molecules A and B were calculated to be related by a local 176° rotation axis and a 12.2 Å translation along an axis parallel to, but not coincident with, the crystallographic b axis. Moreover, β-turn 189–192 appeared to be better ordered in molecule A than in molecule B, which is likely to be because of the crystal-packing environment. No interpretable electron density corresponding to residue 216 of either molecule was observed, and hence this residue was omitted from the deposited PDB coordinates.

Since the A and B molecules were essentially identical in both HiOrt and HiMon, the respective B chains were arbitrarily chosen for comparison here. When subdomains 1 (116 residues, 33–148) from each crystal structure were compared, 96 C^α atoms were superimposed with an r.m.s.d. of 0.3 Å. Similarly, for subdomain 2 (149–248, 105 residues) 80 C^α atoms were superimposed with an r.m.s.d. of 0.4 Å. Comparing the relative curvatures of the two structures showed that the HiOrt subdomain 2 must be rotated by 10° (with an 0.5 Å translation) about an axis passing approximately through the center of the structure to become well superimposed on subdomain 2 of HiMon (Fig. 5 ▸). This rotation coincided with an increase in the curvature of the domain and a narrowing of the concave cleft.

Figure 5.

Superimposition of HiMon (blue) on the previously reported HiOrt structure (cyan) accomplished by fitting their subdomains 1, and showing the 10° rotation axis necessary to then optimally superimpose their subdomains 2.

Previously, we described a positively charged pocket that was underneath helix H3 of HiOrt but was accessible from the concave cleft (Sathiyamoorthy et al., 2017 ▸). In that structure we identified a sulfate ion (SO₄) ligated by three arginine side chains (Arg71, Arg75 and Arg170) and a main-chain amide (Ala42) (Fig. 6 ▸ b). In the HiMon crystals, which were grown in the absence of sulfate, a chloride ion (Cl in Fig. 6 ▸ c) bound instead of sulfate, but was about 2.5 Å closer to Arg75, which assumed a different rotamer. A water molecule (black arrow in Fig. 6 ▸ c) occupied the position of the former sulfate. Also, Arg170 took on a different conformation, which allowed a closer approach to the chloride. This movement of Arg170 coincided with the movement of helix H9 of subdomain 2 closer to H3 of subdomain 1 by 1–2 Å (Fig. 6 ▸ a), which in turn narrowed the concave cleft formed by helices H3 and H9 above helix H7. These observations suggested that the binding of a natural ligand in the cleft of the N domain could affect the overall curvature of the domain.

Figure 6.

Superimposition of HiMon (yellow) on HiOrt (gray) suggesting that their binding of different anions can accommodate the different widths of their grooves. (a) Both structures are shown with the location of the sulfate bound in the groove of HiOrt and the chloride (green sphere) in HiMon. (b) Close-up view of sulfate (SO₄) ligated to arginine residues 75, 170 and 71 of HiOrt. (c) Close-up view of chloride (Cl) bound in the HiMon structure. Chloride was modeled as the bound anion in HiMon because the density was significantly larger than a water molecule, chloride was the only anion present in crystallization solutions and the distances to neighboring atoms were 3.2–3.4 Å.

4. Discussion  

The comparison of EcNMR and HiOrt was previously discussed by Sathiyamoorthy et al. (2017 ▸). Superimposing subdomains 1 or subdomains 2 of each molecule individually showed each of them to have similar C^α structures in the two molecules (r.m.s.d. values of 1.6 and 1.7 Å, respectively). However, superimposing the intact molecules based on fitting their subdomains 1 required a substantial additional rotation (45°) of subdomain 2 of one molecule to superimpose it optimally on subdomain 2 of the other molecule. In other words, the relationship between subdomains 1 and 2 in EcNMR was observed to be quite different from this relationship in HiOrt, which is reflective of EcNMR being relatively flat and lacking the superhelical twist observed in HiOrt.

To help judge whether the flat conformation of the N domain of EcLpoA in EcNMR was a necessary consequence of its amino-acid sequence or was rather, at least in part, a consequence of the solution conditions used, we also analyzed the EcTet crystal structure. The EcTet crystal structure showed more of a curved than a flat conformation, indicating that the flat conformation of the EcLpoA N domain observed in the EcNMR solution structure was not mandated by its sequence. Based on the calculated rotation angles to superimpose sub­domains 2, the EcTet crystal structure showed a subdomain 2 orientation intermediate between those of HiOrt and EcNMR. Judging from just these measurements, EcTet in the crystalline environment appeared to be closer in structure to HiOrt than to EcNMR. However, as shown in Fig. 7 ▸, the rotation axis relating subdomains 2 of EcNMR and EcTet was significantly different from that between HiOrt (or HiMon) and EcTet. To further compare the three structures, we measured an interatomic distance between helices H3 and H9 as an indicator of the width of the groove. Despite the large rotation angle relating subdomains 2 of EcTet and EcNMR, the two structures displayed similar groove widths of 19.5 and 19.7 Å, respectively (not shown; based on the distance between C^α of Lys75 and C^α of Ala176), while HiOrt displayed a narrower groove with a width of only 15.0 Å (measured between the structurally homologous residues Arg75 and Lys177). The similarity of the groove widths in the EcNMR and EcTet structures might be explained by the orientation of the rotation axis relating their subdomains 2 (Fig. 7 ▸ a) and its difference from the axis relating subdomains 2 of HiOrt and EcTet and from the axis relating subdomains 2 of HiMon and EcTet (Figs. 7 ▸ b and 7 ▸ c). The effect of the difference can also be observed by the relative positions of H11 and H12 in the two pairs of structures (compare Figs. 7 ▸ a and 7 ▸ b). Analysis of the high-resolution HiMon structure demonstrated that the domain can assume a more concave structure than previously observed in HiOrt. The groove of HiMon as defined above was measured to be 1.8 Å narrower than the HiOrt groove.

Figure 7.

End-on views of structures superimposed on EcTet (magenta). Also displayed are the rotation axes relating subdomains 2 between EcTet and the (a) EcNMR (gray), (b) HiOrt (cyan) and (c) HiMon (blue) structures. The orientations of these axes in the HiOrt and HiMon crystal structures were found to be similar to each other but significantly different from that in EcNMR.

All of the crystal structures presented here showed a more curved LpoA N domain than was observed in the EcNMR structure. These crystal and NMR structures taken together clearly indicate the N domain to be inherently flexible. Crystal-packing forces may have preferentially selected a more compact structure, but in solution the protein dynamics were less constrained, allowing the molecule to sample other conformations. The intrinsic flexibility of the N domain would be expected to affect the location of the C domain in the intact LpoA structure and its ability to interact with the PBP1A synthase transpeptidase domain in the periplasm, as discussed previously (Sathiyamoorthy et al., 2017 ▸).

Supplementary Material

PDB reference: amino-terminal domain of H. influenzae LpoA, 6dcj

PDB reference: amino-terminal domain of E. coli LpoA, 6dr3