The kinemage graphics file (Davis_Backrub_kinSup.kin) of the Supplement to Davis, Arendall, Richardson, & Richardson (2006) shows 3D examples that animate between alternate-conformation pairs from ultra-high-resolution crystal structures, with electron density maps, hydrogen bonds, deposited models, and BACKRUB-fitted models for both sidechain and backbone. The *.kin file is a text file that can be read and displayed by the free, cross-platform programs Mage or KiNG (1-3) available from Alternatively, to view the supplement kinemages without software download, go to the MolProbity service (4) of the above URL. Rather than choosing or uploading a coordinate file, upload the supplement kin file; it will appear in the list of your working files within MolProbity, where you can choose to view it in KiNG directly on line.

A kinemage display has a scrollable text window with explanatory information (which is reproduced here, below) and a 3D graphics window with associated menus and buttons. The basic kinemage viewing functions are:

Table of Contents for Davis et al. Backrub Supplement kin file

The *.pdb files in the supplementary material contain PDB-format coordinates for the BACKRUB-refit alternate conformations shown in the kinemages, plus several more serine examples. Only the atoms that shifted in the refitting are provided, plus the Calphas at each end; other coordinates can be taken from the deposited structures. The two conformations are labeled as the A and B alternates, with occupancies taken from the original crystal structure. Most of these pdb coordinates were generated by a repeat BACKRUB fitting independent of the kinemage illustrations, and as such may differ very slightly from them. This should reinforce that these models are only approximations and have not been re-refined against the experimental data.


Text-window explanations of the kinemages in Davis_Backrub_kinSup.kin:

kinemage1 image *{Kinemage 1}*, views 1 & 2, illustrate alternate conformations for Ser a15 of PDB file 1DY5, as in Figure 3b of the paper, with electron density contours at 1.2 sigma (gray) and 3 sigma (purple). Serines constitute 25% of the residues that show backrub motions, and this is a very typical example. The sidechain changes H-bond partners and changes rotamer between m (-50 chi1, here) and p (+65), and the Cbeta shifts by just under 0.5Å. The white backbone and cyan sidechains show the deposited coordinates (display controlled by the 'orig Ser15' checkbox on the button panel), which were fit with a common backbone and Calpha position, since there is only minor anisotropy in the backbone electron density. [Turn off both 'brubA Sa15' and 'brubB Sa15' to see the original model more clearly.] That produces severe distortion of bond angles - tetrahedral angles at the Calpha vary from 94 to 133 degrees. You can measure them yourself by turning on Measures on the Tools menu. [Afterward turn off the white marker lines with the 'm' key.]

The BACKRUB tool in KiNG was used to refit the two clear sidechain conformations, to see whether a small backrub shift in the backbone can fit the density well without significant bond angle or peptide distortion. Animate (with the button, or the 'a' key) to switch between the BACKRUB models for the two alternates; turn off the original models, and animate or turn on both backrub models together. Especially, by animating in View2 without the density, you can see the primary backrub motion that rotates Ser 15 and the two peptides around an axis between Calphas 14 and 16; a very slight secondary counter-rotation was used in this case for the 15-16 peptide, to keep the 15 CO shift small. The 15 Calpha shifts 0.3Å and other backbone atoms shift 0.2Å or less. The tau angles (N-Ca-C) change by 0 at Calpha 14 and by 2 degrees at Calpha 16, staying within less than one standard deviation of ideal. Each atom is very close between the original and the backrub models, as must be true if they both fit the density well, but the consequences for ideality of covalent geometry (and therefore for energy cost) are very different. Turn on one of the map contours and both backrub models, and turn the image in various directions to see that the modest elliptical anisotropy of the backbone density peaks is elongated in the same direction as the shift between the two backrub models, confirming that this does indeed represent the real motion taking place in the molecule.

In this example it is easy to see what forces produce the conformational changes between alternates. The B alternate is quite unconstrained, with one H-bond to a water, good van der Waals contacts with the surrounding structure, and presumably a locally relaxed backbone position. The A alternate has better H-bonding (to the 49 backbone CO and to a more-occupied water), but positioning the sidechain oxygen from the altB Calpha and Cbeta would require either a very bad chi1 (about -20) or a bad clashing distance rather than H-bonding distance to the 49 CO. Thus the Cbeta of altA is shifted outward by a relatively low-energy backrub motion in order to allow good H-bonding in a good rotamer.

View3 illustrates another serine example from the same protein, Ser b 50, also with "T-shaped" electron density. Ser b 50 is the N-cap of an alpha-helix, and the two alternate conformations let it make N-cap H-bonds with the two different peptide NH's, at positions Ncap+2 and Ncap+3 in the first turn of the helix. The alt-B conformation is rather crowded, which produces a backrub shift relative to alt-A.

In these, and in all the backrub alternate-conformation examples, our overall claim is that the BACKRUB models provide an equally good crystallographic fit to the electron density as the original deposited coordinates, with fewer adjustable parameters and much more ideal geometry. Therefore we conclude from this experimental data that the backrub motion provides an improved general description of how these residues shift backbone conformation to accommodate sidechain changes.

kinemage2 image *{Kinemage 2}* illustrates alternate conformations for Lys 100 of PDB file 1US0, as in Figure 3c of the paper, with electron density contours initially visible at 3 sigma (purple); the lower contour level can be turned on when desired. As in Kin. 1, white & cyan designate the deposited models and peach & orange the BACKRUB-fit models; animate between the latter (button or 'a' key) or turn them both on together with the button-panel checkboxes. Cbeta moves about an Angstrom, and the sidechain rotamer changes between mppt and mtmm. Lys 100 is a Schellman-type C-cap for the alpha-helix, in Lalpha conformation and with a capping sidechain H-bond from Nz to the n-2 CO. The alternate conformations combine sidechain and backbone movements in such a way that all 3 H-bonds to the Lys Nz are preserved (green dotted lines).

The backrub model was started in the altA conformation, and altB was modeled by changing just 2 of the backrub parameters plus sidechain chi angles, fitting the exceptionally clear electron density just as well as the deposited altB coordinates. Atoms Cbeta-Cdelta show pairs of separate peaks, while Cepsilon, Calpha, N, and both carbonyl O's show strong anisotropy whose directionality is well fit. Note that Ser 97 (lower right) has Cbeta alternates; that residue can be fit with BACKRUB models (not shown) whose CO positions shift in such a way as to help preserve the backbone H-bonds with Lys 100.

kinemage3 image *{Kinemage 3}* illustrates alternate conformations for Ile 47 of PDB file 1N9B, as in Figure 3d of the paper, with electron density contours at 1.2 sigma (gray). In this case, only the two BACKRUB models are shown for the Ile, for clarity (the deposited coordinates are extremely close for the sidechain, with one intermediate model for the backbone). Separate density peaks for all sidechain Cgamma and Cdelta atoms make the switch between tt and mm rotamers clear, and allow unambiguous BACKRUB fit of an ideal-geomatry backbone model to fit each alternate (orange). Lenses of pale green dots show the overlap for the beta-sheet H-bonds, all 4 of which are conserved in both conformations. The carbonyl O atoms each shift about 0.25Å, while the Cbeta moves nearly 0.6Å and the sidechain Cgamma1 moves 2.1Å.

[Note: Counter-rotation of the individual peptides with the BACKRUB tool could have kept the CO shifts even smaller, but that was not done here because all the carbonyl O densities are somewhat anisotropic, most notably the CO of residue 58 (see View2). These are presumably correlated motions within the sheet, connected by the H-bonds. However, the relative assignment of altA vs altB for the 268 strand and the 58 strand was matched backwards (i.e., B to A and A to B); animate to see this. One of the two strand-alternate assignments could be reversed, because all relative occupancies are near 1:1.]

kinemage4 image *{Kinemage 4}* shows Leu 30 of 1N9B, one of the few cases with a Cbeta shift >0.2Å (0.26Å, here) that were dropped from the list of clear alternates analyzed in this study, because the Cbeta shift is actually not needed at all. Here, animation switches between the original pair of models and the BACKRUB pair of models. The electron density contours are at lower levels of 0.5 and 1.1 sigma. The original models are not a good fit to the rather confusing density shape, and altB (in gold) has a very poor sidechain rotamer with both chi angles nearly eclipsed. Refitting in KiNG using the two best Leu rotamers (mt and tp) from a single ideal Cbeta position gave a very good fit to the density, and this seems a preferable interpretation. This Leu is solvent-exposed on a helix, and the low sidechain density may indicate that it is actually a mixture of even more than two conformations.

kinemage5 image *{Kinemage 5}* shows alternate conformations of Tyr 378 of PDB file 1GWE, with highly anisotropic density for the carbonyl O. As before, animate (button, or 'a' key) to switch between the two BACKRUB-fitted models. View2 is a closeup to show how the 378-9 peptide rotation in this case adds in the same direction as the primary BACKRUB motion (around Calphas i-1 to i+1), producing the large movement of the 378 CO which also switches between water H-bond states (turn on 'H-bonds'). One low-occupancy water in grayed out for altB, since it is too close to the altB CO. View3 looks down the non-crystallographic 2-fold axis of the molecular dimer, to show the reason why alternate positions are necessary for Tyr 378. The extra density is the Tyr 38 ring of the B chain, which clearly overlaps altB of Tyr a 378; thus only one chain can use the altB position at a time, and the other one must go elsewhere (altA).

kinemage6 image *{Kinemage 6}* illustrates the alternate conformations produced by reduction of the 41-57 disulfide in PDB file 1PQ7. The electron density (contoured at the higher levels of 1.5 and 6.0 sigma here) shows clearly that Cys 41 changes rotamer when the SS bond breaks, while Cys 57 does not. Animate to see that the orange BACKRUB models fit the rotamer change of Cys 41 very well. Both the density shape for Cys 57 Sg and the distance to Cys 41 Sg show that Cys 57 must actually move somewhat when the SS breaks, but an alternate was not fitted originally, perhaps because the motion cannot be accomplished with a simple rotation of chi1. The animation shows that a small backrub motion can shift Cys 57 Sg in the required direction. There seem to be no tight contacts that would keep Cys 57 from just rotating chi1 (a low-energy change) to let the unbonded S atoms move apart. However, backbone conformations are often slightly strained to pull together into a disulfide, and the backrub change seen here could well be a relaxation back to a preferred position.

[Note: for Cys 41, the original backbone had quite distorted peptides (omega changing by 30 degrees between A and B alternates), and so our refitting started from a somewhat idealized version of the original A conformation.]