Tight Turns of Outer Membrane Proteins: An Analysis of Sequence, Structure, and Hydrogen Bonding

Abstract

As a structural class, tight turns can control molecular recognition, enzymatic activity, and nucleation of folding. They have been extensively characterized in soluble proteins but have not been characterized in outer membrane proteins (OMPs), where they also support critical functions. We clustered the 4 to 6 residue tight turns of 110 OMPs to characterize the phi/psi angles, sequence, and hydrogen bonding of these structures. We find significant differences between reports of soluble protein tight turns and OMP tight turns. Since OMP strands are less twisted than soluble strands, they favor different turn structures types. Moreover, the membrane localization of OMPs yields different sequence hallmarks for their tight turns relative to soluble protein turns. We also characterize the differences in phi/psi angles, sequence, and hydrogen bonding between OMP extracellular loops and OMP periplasmic turns. As previously noted, the extracellular loops tend to be much longer than the periplasmic turns. We find that this difference in length is due to the broader distribution of lengths of the extracellular loops not a large difference in the median length. Extracellular loops also tend to have more charged residues as predicted by the charge-out rule. Finally, in all OMP tight turns, hydrogen bonding between the side chain and backbone 2 to 4 residues away from that side chain plays an important role. These bonds preferentially use an Asp, Asn, Ser, or Thr residue in a beta or pro phi/psi conformation. We anticipate that this study will be applicable to future design and structure prediction of OMPs.

Keywords: beta barrels/β-barrels; loops; reverse turns; α-turns; β-turns.

Figure 1.. Comparison of distribution of loop lengths and turn lengths.

Loop data are shown in red bars, turn data are shown in blue bars. Distributions exclude incomplete structures and turns with extensive secondary structure. A) Left: distribution of strand connector lengths. Right: The structure of OmpA (PDB: 2GE4) is shown with the membrane-bound region approximated by the blue and red lines. B) The proportion of PDBs in our dataset that contain a strand connector of each length. C) The average number of residues in the strand connectors grouped by the number of strands per chain per barrel.

Figure 2.. Four residue strand connectors

A) Clustering of the 4-residue loops and turns of OMBBs. Loops are shown with diamond markers; turns are represented by circle markers. The plus symbols in the i+1 and i+2 plots mark four of the previously defined ideal turn types I, II, I’, and II’ as noted by Guruprasad and Rajkumar (2000). Green: type I; orange: type I’; purple: type II; yellow: type II’; red: unknown; black: no cluster. B) Representative example of each cluster, colored as in A. The red cluster has terminal positions spaced further apart than the other four canonical tight turn types. C and D) Sequence logos for the 4-residue strand connectors. Each position is shown as a column; the number of sequences in a cluster is shown in the subtitle. Amino acid groupings are described in the methods. C) Sequence logos for the 4-residue loops. D) Sequence logos for the 4-residue turns.

Figure 3.. Comparisons between 4-residue OMBB strand connectors and 4-residue soluble protein tight turns.

A and B) Comparison of the β-hairpins surveyed in Sibanda et al [24] compared to the hairpins of OMBBs. A) Proportion of each of the five different types of 4-residue strand connectors observed in the OMBBs compared to those of the analogous 4-residue tight turns of soluble hairpins. “Other” refers to tight turns that do not fall into any of the other canonical β-turn types. B) Kernel density estimator of the twist of the hairpins connected by 4-residue tight turns in Sibanda et al [24] and in OMBBs. C) Statistically significant differences in amino acid districbution between soluble tight turns as clustered by Guruprasad and Rajkumar ([1], abbreviated G&R) and the OMBB clusters. Statistical differences are based on a permutation test. White blocks indicate no significant difference; pale purple blocks indicate a significant difference of zero between the proportion of of amino acid in the OMBB vs G&R. The remainder of the colored blocks indicate a statistically significant difference between the two. Red blocks have a higher propotion in OMBBs, indicating a preference in OMBBs, while blue blocks are have a higher proportion in G&R, indicating a preference in G&R. The amino acid groupings are defined in the methods.

Figure 4.. 5-residue strand connectors

A) Clustering of the 5-residue loops and turns of OMBBs. Loops are shown with diamond markers; turns are represented by circles markers. The first panel shows the division of the Ramachandran plot into regions according to North et al. 2011. The other 5 plots show the phi/psi angles for each residue in the strand connector. B) Sequence logos for the 5-residue loops. C) Sequence logos for the 5-residue turns. Sequence logos are colored and grouped as in Figure 2.

Figure 5.. 6-residue strand connectors

A) Clustering of the 6-residue loops and turns of OMBBs. Loops are shown with diamond markers; turns are represented by circle markers. Because the i+3 residue of the green cluster straddles the boundary between the D and A regions, the green cluster is labeled AAAL in the loops and AADL in the turns. B) Sequence logos for the 6-residue loops. C) Sequence logos for the 6-residue turns. Sequence logos are colored and grouped as in Figure 2.

Figure 6.. Side chain backbone bonding in strand connectors.

A) Proportion of Asn, Thr, Ser, and Asp (NTSD) in any position compared to the proportion of sidechains donated to a side-chain:backbone hydrogen bond at that position. Positions that fall outside the B or P region of the Ramachandran plot are shown in grey. B) Turn 2 of PDB ID 3qra. The side-chain:backbone hydrogen bond between the side chain of i+1 Asn108 and the backbone of i+3 Tyr110 is shown in dotted lines. Glu109 is the i+2 residue and its side chain points directly outward. C) The proportion of sidechain:backbone hydrogen bonding at each pair of positions. The identity of the sidechain partner is shown along the x-axis; the identity of the backbone partner is shown along the y-axis. “Other” indicates hydrogen bonding to a backbone partner outside the loop. The radius of the circles is equivalent to the proportion of structures at that position involved in bonding. Circles are colored to match the clusters marked on the Ramachandran plots in Figs 2, 4 and 5. P-values for each position and each bond type are reported in the supplemental file HBondsByCluster_PValues.xlsx.

Figure 7.. Amino acid usage of the right side of the Ramachandran plot.

Proportion of Asp, Asn, and Gly in positions that fall into the G or L region of the Ramachandran plot. The remaining 17 amino acids are grouped into “Other”.

Figure 8.. Comparison of amino acid preferences of terminal and non-terminal positions.

Proportion of residues that fall into each category in the terminal or nonterminal positions. A) Distribution by chemical feature. Groups are colored as in Figure 2. B) Distribution by van der Waals volume. Small – G, A, S, C, P, D, T, N; medium – V, E, Q, H, I, L, M; large – R, K, F, W, Y.