B12 dictionary and geometry issues

  • A number of people have asked for a dictionary for vitamin B12 (cobalamin)
  • grade cannot (currently) handle such a metal containing compound.
  • But I have produced a dictionary by using grade plus manual data mining of the CSD
  • This work produced an important observation:

CCP4 dictionary and PDB chemical components definition of B12 are incorrect.

  • The problem is with atom "C19". Starting from the diagram at Ligand Expo http://ligand-expo.rcsb.org/reports/B/B12/B12_D3L0.gif the problem is:
    • B12_D3L0.gif
  • In the CCP4 dictionary distributed with release 6.3.0 atom C19 has a double bond to N24. There is no hydrogen atom attached to C19. The ideal bond angles for the three bonds around C19 are all 120 degrees. There is a plane C19-N24-C1-C18 holding C19 and its neighbors flat. This is wrong as C19 is actually tetrahedral and a chiral center with a hydrogen atom attached (see below). The B12 dictionary distributed at the refmac site is similar.
  • The PDB chemical components definition cif from http://ligand-expo.rcsb.org/reports/B/B12/B12.cif also lack a hydrogen atom "H19" and the chiral definition for C19. In contrast the model and ideal coordinates are identical and have atom C19 as tetrahedral. This is despite atom C19 lacking a hydrogen atom and the N24 to C19 bond being marked as a double bond. The coordinates for heavy atoms come from PDB entry 7req (that correctly models C19 as tetrahedral) but with incorrect hydrogen atoms. Something has gone wrong!
  • The correct B12 structure is provided by the small molecule crystal structure of adenosylcobalamin as determined by Luger and co-workers: http://pubs.acs.org/doi/abs/10.1021/jp902433x and available as CSD structure PAFBUV.
    • PAFBUV Schematic diagram provided by CCDC conquest PAFBUV.png
    • PAFBUV coordinates in PDB format as supplied by CCDC conquest PAFBUV.pdb
    • To help checking dictionaries I have produced PAFBUV adenosylcobalamin coordinates in pdb format marked up with PDB atomnames (solvent has been deleted) PAFBUV_pdbatomnames_inc5AD.pdb
  • Perhaps the best evidence is provided by looking at the chemical structure of B12 as determined by Dorothy Hodgkin see diagram available at http://www.nature.com/nrm/journal/v4/n11/fig_tab/nrm1243_F2.html#figure-title Hodgkin knew a thing or two about B12 structure!

PDB structures with poor C19 chirality.

  • I did at little search of the PDB structures containing B12 looking at the improper torsion angle C19-N24-C1-C18. This torsion angle is -31.7 degrees for PAFBUV. PDB remark 500 uses a limit of -25 degrees on improper torsions for chirality.
  • results: B12_check.tsv (tsv file suitable for excel or another spread sheet)
  • 19 of the 53 structures in the PDB have poor chirality for atom C19.
  • The codes are:
2xij 2xiq 2ycl 3bul 3ci1 3ci3 3gai 3gaj 3iv9 3iva 3kow 3kox 3koy 3koz 3kp0 3kp1 3kq4 3o0n 3o0o
  • One can note that the problem is with new structures. The earliest "bad" dates from 03-JAN-08 all older structures are OK.

B12 dictionary

  • After the PDB definition is corrected adding atom H19 and making C19 chiral (as per PAFBUV), a dictionary for B12 was produced using grade results for B12 with the cobalt atom removed and manual searches of the CSD for cobalt contacts. Note that the dictionary does not by default include a bond between the cobalt atom and atom N3B in the "tail" as this group does not coordinate in many protein structures. The dictionary includes instructions for how to edit this bond in.
  • The 5AD part of B12 can be treated using the normal grade dictionary:

example refinement with new B12 dictionary

  • As an example of the effect of using the new dictionary PDB entry 3ci1 (originally refined with refmac) was re-refined using BUSTER.
    • Before re-refinement the BUSTER density map has a blob of negative density (here contoured at 3.5 sigma) around C19. The improper torsion angle around atom C19 is -12.4 degrees indicating that density is fighting its planar assignment. The atom also lies at the edge of the 2Fo-Fc density contoured at 1.3 sigma:
    • 3ci1_00_maponly-C19.png
    • After refinement with BUSTER (including explicit hydrogen atoms) the geometry of the B12 is improved. C19 atom now sits happily in density and there is no difference density. The C19 improper torsion alters to -31.5 degrees. The density correlation coefficient for the B12 improves from 0.923 to 0.948.
    • 3ci1_02_plush-C19.png
  • If you want to look at models / density:

Report problem to ccp4bb and PDB

Page by Oliver Smart from 10 October 2012. Any questions regarding our software or this wiki should be directed to buster-develop@globalphasing.com