Vitamin B12 and B12-Proteins (1999) B. Kräutler, D. Arigoni & B.T. Golding, eds, pp 335-347, Wiley-VCH, Weinheim, BRD.Strubilogo_small.jpg - 8059 Bytes

High-resolution Crystal Structures of Cobalamins

Karl Gruber, Gerwald Jogl, Gerd Klintschar & Christoph Kratky

Abteilung für Strukturbiologie, Institut für physikalische Chemie, Universität Graz

Heinrichstraße 28, A-8010 Graz, Austria


Summary: Crystallography of cobalamins is hampered by poor crystal quality, which is due a high content of partly disordered solvent molecules. Nevertheless, the crystal structures of several dozen B12 derivatives are known to date at medium to high crystallographic resolution. Recently, the use of synchrotron radiation and imaging plate detectors has significantly improved the accuracy of B12 crystal structures. Although the constitution of protein-bound B12 differs from cobalamins in solution or in crystals, the comparison of many crystal structures yields useful information about the flexibility of the B12 system.Notably, there exists a correlation between the length of the axial Co-N bond and the "upward" deformation of the corrin ring: short Co-N bonds induce upward folding by steric interaction between the corrin ring and the dimethylbenzimidazole base. For very short Co-N bonds, as observed e.g. in crystals of aquocobalamin, this steric interactions also decreases the flexibility of the B12 system, as revealed by thermal motion analysis. EXAFS data show that - as far as the geometry around the cobalt center is concerned - there is no detectable difference between dissolved and crystalline cobalamins.


1 Introduction

Crystallography of isolated B12 cofactors had its "golden days" more than 30 years ago, when several crystal structures from the laboratory of Dorothy Hodgkin revealed the molecular constitution and configuration of vitamin B12 [1], of coenzyme B12 [2] and of several other B12 derivatives [3]. Crystallographic structure analysis of the second biological B12 cofactor - methylcobalamin - emerged about two decades later from Jenny Glusker's laboratory [4]. Two reviews summarize the state of B12 crystallography of about ten years ago [5], at which time various speculations about possible biological implications of the (then known) crystallographic evidence were advanced. The number of crystal structure determinations of cobalamins published since then has been surprisingly small (see Table 1). This contrasts sharply with the vast number of structure determinations of "B12-model" complexes of various kinds [6]. (note that we use the term "cobalamin" here to include all B12 structures with a "complete" nucleotide loop which coordinates intramolecularly to the cobalt centre. Also included are compounds - such as Coß-cyanoimidazolyl-cobamide - which have a modified nucleotide base replacing the dimethylbenzimidazole).

Concerning the biological role of the B12 cofactors, much consideration was given to the remarkable intramolecular cobalt-coordination of the (biologically unusual) dimethylbenzimidazole (DMB) base. It therefore came as a big surprise, when two crystal structures of B12 dependent enzymes or enzyme fragments - of the B 12 binding domain of methionine synthase [7] and of methylmalonyl CoA mutase [8] - showed that the two enzymes bind B12 in a "base off" form, with a protein-derived imidazole replacing DMB from its cobalt coordination site.

While the above structure determinations of B12 binding proteins have thus undermined many conclusions derived from small-molecule (including cobalamin) crystallography, there is still much to be learnt from precise geometrical parameters obtainable by this technique. In the present contribution, we try to give an overview of the work of the Graz group in the field of cobalamin crystallography, and to show the potentials of modern crystallographic techniques for the B12 area. We also hint at possible ways for B12 crystallography - possibly in combination with related techniques, such as EXAFS - to yield structural information on B12 binding proteins, complementing the unique results of protein crystallography.


2 Accuracy of Crystallographic Results on B12

Crystal structures of cobalamins are inherently inaccurate as judged from the crystallographic reliability indices. This is due to a considerable number of - frequently disordered - solvent molecules (typically 10-20 per B12) within cobalamin crystals, and it manifests itself in large crystallographic residuals (typically around 10%) and large e.s.d.'s for derived geometrical parameters (typically exceeding 0.01 Å for a C-C bondlength).

Table 1 lists B12 crystal structure determinations, some of which have appeared in the literature; others were determined in our laboratory in recent years [9,10]. The above inaccuracy is evident from the crystallographic R-values; exceptions are only the structure determinations of aquocobalamin [11] and of cob(II)alamin [12], both of which were determined using synchrotron radiation.


3 Cobalamin Crystal Structures Tend To Be Similar

Figure 1 shows the superposition of 12 cobalamin molecules as observed in their respective crystal structures (Table 1). Evidently, these molecules superimpose very well, suggesting little molecular flexibility confined to the acetamide and propionamide sidechains. Moreover, analysis of the packing of B12 molecules within the crystal reveals recurrent packing motifs. The vast mayority of cobalamins crystallize in space group P212121 and can be assigned to one of three packing types, characterized by the cell axis ratios c/a and b/a. Each of these packing types results in a cluster in the corresponding scatterplot, as shown in Figure 2. What is more, the three types have an identical 2-dimensional arrangement of B12 molecules; they differ only in the way the 2-dimensional sheets of B12 molecules are stacked. This observation may thus (at least in part) explain the apparent lack of molecular flexibility (Figure 1), which could be the result of similar intermolecular contacts in the majority of known B12 crystal structures.


4 Solvent Structures in B12 Crystals

In view of this packing invariance of B12 crystals, it is particularly suprising that the distribution of solvent molecules (mainly water and acetone) around the B12 moieties differs considerably from one cobalamin crystal to the other. Only very few solvent sites - mainly within hydrogen bonding distance of one of the phosphate oxygen atoms - appear to be conserved between different B12 crystals [14]. The distribution of solvent molecules in different cobalamin crystals can thus be best described as a "glue", which fills voids between adjacent B12 molecules.

Analysis of the distribution of solvent molecules within cobalamin crystals was shown to be extremely useful for understanding the water distribution around biological macromolecules. This has been impressively demonstrated for the case of coenzyme B12 [15]. In analyzing solvent sites, a particularly powerful complement to X-ray diffraction analysis is neutron diffraction, which permits the direct observation of protons (or deuterons), which are poorly or not at all visible in electron density maps from X-ray diffraction data. To illustrate the latter point, Figure 3 shows a comparison of neutron [16] and X-ray [12] densities for the same section through the crystal structure of cob(II)alamin, B12r.


5 Structural Correlations in Cobalamins

It has been known for a long time [17] that a preferred mode of deformation of the corrin ring is a folding about an axis running "east-to-west", i.e. approximately bisecting the C1-C19 bond and passing through atom C10. This deformation was termed "upward folding", and it was suggested to be involved in enzyme-assisted cobalt-carbon bond homolysis (see e.g. [11]). The deformation is generally measured by a "fold angle", defined as the dihedral angle between best planes through rings A + B and rings C + D, respectively. Fold angles between about 10° and 20° were observed in cobalamin crystal structures, and there is the obvious question why some cobalamin structures show a larger fold angle than others. This question can be addressed by analysis of the available structural evidence.

Figure 4 shows a plot of observed fold angles versus the axial cobalt-nitrogen bond length observed in the same structure for the most accurate crystal structures of Table 1. We note that structures were included solely on the basis of their crystallographic residual; inclusion of the remaining structures does not alter the correlation, but introduces noise and makes the correlation thereby less evident.

Clearly - with the excpetion of one structure (see below) - the two quantities (fold angle and Co-N distance) correlate very well. According to the structure - correlation principle [18] such a parameter correlation maps a minimum energy path in the deformation potential hypersurface. The one point which does not follow the correlation - corresponding to the crystal structure of Coß-cyano-imidazolylcobamide - indicates the intramolecular interaction responsible for the observed correlation. Imidazolylcobamide is the vitamin B12 derivative with the DMB base replaced by (the much less bulky) imidazole. Thus, the corrrelation shown in Figure 1 is due to the bulk of the DMB base, which interacts with the corrin ring when the base is "pulled upwards" by a short Co-N distance.

The above results are illustrated and reiterated by Figure 5, which shows two superpositions of B12 structures. For each structure, the nucleotide base is represented by a lines drawing, whereas the corrin ring is represented by two lines within the least-squares planes through the "northern" and "southern" corrin ring halves (the dihedral angle between these planes is defined as the "fold angle"). Each part of Figure 5 shows the superposition of two B12 structures, projected into the plane through the nucleotide bases. The left side shows a superposition of coenzyme B12 (long Co-N distance) with aquacobalamin (short Co-N distance), and it illustrates the increase of the fold angle induced by a decrease of the axial Co-N distance. The right halve shows a superposition of vitamin B12 with Cob-cyanoimidazolylcobamide, illustrating the decrease in the fold angle as a result of a reduction of the bulkiness of the nucleotide base.


6 Thermal Motion Analysis

Recently, very accurate crystal structures became feasible with the advent of synchrotron radiation. Two structure determinations utilizing this technology have so far been carried out in our laboratory with the generous help of the staff of the EMBL outstation at DESY in Hamburg. They are the crystal structure determinations of aquocobalamin perchlorate [11] and the structure analysis of cob(II)alamin, B12r [12]. Notably, while the former structure shows a very short axial Co-N distance (1.93 Å) and a correspondingly large fold angle (19°), the latter is at the opposite end of the correlation of Figure 1 (Co-N: d=2.16 Å, fold angle 14.4°). In other words, in the aquocobalamin structure, the corrin ring and DMB base are much more "jammed together" than in B12r. We have performed a thermal motion analysis [19] with the crystallographically observed atomic displacement parameters of the two compounds. One might expect to see more intramolecular flexibility in the less strained B12r crystal structure than in the more strained aquocobalamin system.

To address this question, the corrin ring atoms and the atoms of the DMB base were subjected to a TLS analysis individually and together for both compounds. Figure 6 summarizes the result, listing the residual between observed and calculated a.d.p.s. While the isolated corrin rings and DMB bases appear to behave as rigid groups at a level of R~6-9%, together they can be simulated as a rigid group much better for aquocobalamin perchlorate (R~13%) than for Cob(II)alamin (R=20%). This is in excellent agreement with the view that the aquocobalamin system is more strained than B12r due to the shorter axial Co-N distance and therefore shows less flexibility between the corrin ring and DMB base.


7 B12 Crystallography and EXAFS Spectroscopy

EXAFS and the related XANES spectroscopy hold great promises for the elucidation of metal oxidation states and precise coordination geometries, in particular when applied to biological systems [20]. Since the experiment can be carried out with solutions, its results are not biased by crystal packing effects and compounds can be studied which resist attempts to obtain crystals suitable for diffraction purposes.Owing to theoretical and experimental difficulties (small number of correlated parameters, insufficient account of the contribution of higher shells and of multiple scattering, etc.) EXAFS experiments are most significant when carried out in tandem with high-resolution crystal structure analyses of suitable model complexes.

A few years ago, analysis of EXAFS spectra on a variety of cobalamins yielded geometric parameters of the cobalt-coordination [21] which were at variance with crystal structure results [13] or with expectations from coordination chemistry [11]. A particularly striking example was aquocoblamin, where a long axial Co-N distance of 2.14± 0.03 Å was deduced from EXAFs spectra [21]. Since coordination chemistry would predict a short distance trans to the weakly donating water ligand, we determined the crystal structure of aquocobalamin perchlorate and indeed found the shortest axial Co-N distance observed in a cobalamin so far (1.925± 0.002 Å). However, due to the strain between corrin ring and DMB ligand (both bulky, see above), the possibility could not be entirely ruled out that the difference of more than 0.2 Å between single-crystal and EXAFS results might be due to crystal packing effects. For that reason, we redetermined the EXAFS spectra on aquocobalamin (using the perchlorate salt), measuring the spectra of the compound both in solution and in the solid state. As Figure 7 shows, the two spectra agree within the experimental error. Since the crystal structure analysis of aquocoblamain perchlorate was determined with synchrotron radiation (R~4% !), we conclude that there is no experimental support for a difference in cobalt-coordination between crystalline and dissolved aquocobalamin, and that the value from the crystal structure (1.925 ± 0.002 Å) is the correct axial Co-N distance.


8 Biological Significance of Axial Co-N Distance

While the above example might be considered as a mere exercise on the relative merits of X-ray diffraction analysis versus EXAFS spectroscopy, it could also guide the way to experiments addressing the fundamental question of the mechanism of enzyme-induced cobalt-carbon bond homolysis. Recently, the crystal structure analysis of methylmalonyl - CoA-mutase (MMCoA mutase) was concluded [8], crowning many years of effort. This analysis showed an unexpectedly long axial cobalt-nitrogen bond to the (protein derived) imidazole. In this crystal structure, the B12 cofactor is present in its Co(II) state, i.e. with only one axial ligand to the cobalt centre [8]. Compared to the crystal structure of cob(II)alamin [13], the axial bond length differs by about 0.3 A.

There are no single crystal data available for a imidazolylcob(II)amide, i.e. the B12r derivative with imidazole as nucleotide base. However, the available structural evidence (summarized in Figure 8) argue for a very similar Co-N distance, with the distance to the imidazole base inherently even shorter (by ~ 0.05 A) than the one to DMB. Thus, the long axial Co-N bondlength observed in the MMCoA-mutase crystal structure appears to exceed the "expected" value by as much as ~ 0.35 A.

This is indeed a surprising result, which offers an attractive explanation for the mechanism of enzyme-assisted Co-C bond homolysis. As has been shown recently by structure correlation with cobaloximes as well as by theoretical considerations [22], there is an "anomalous" dependence between the bond lengths to the upper and lower ligands in B12 systems, where a lengthening of the distance to the lower ligand (which is a weak base) increases the distance to the upper ligand (being a very strong donor).

The crystal structure of MMCoA-mutase was determined at 2 A resolution, and the resulting bondlengths are subject to a corresponding experimental error. Also, the precise nature of the cobalamin cofactor could not be determined unambiquously. It appears that EXAFS data would ideally complement the crystallographic evidence on the long axial Co-N distance, particularly in combination with EXAFS spectra on cobalamins with known crystal structures. Experiments to that end are underway in our laboratory.



Several of the unpublished structures used in the correlations of Figures 2 and 4 were determined by Gerald Färber during his PhD work in our laboratory. We acknowledge a long and very enjoyable collaboration with Bernhard Kräutler and his group. This collaboration is one of the foundations of our work in the B12 field. The crystal structure of imidazolocobalamin quoted in Figure 8 was carried out in collaboration with Rudi van Eldik and Nicola Brasch. We also acknowledge the help of the EMBL staff in Hamburg (Keith Wilson, Zbysek Dauter, Viktor Lamzin) for the collection of the synchrotron data and of the ILL staff in Grenoble (Paul Langan, Sax Mason) for the neutron data. We received financial support from the Austrian Science Foundation FWF (Projects 8371, 9542 and 11599) and the Jubiläumsfonds der Österreichischen Nationalbank (Proj. 4991).




1. (a) D. C. Hodgkin, J. Lindsey, R. A. Sparks, K. N. Trueblood, J. G. White, Proc. R. Soc. London, Ser. A 1962, 266, 494-517. (b) C. Brink-Shoemaker, D. W. J. Cruickshank, D. C Hodgkin, M. J. Kamper, D. Pilling, Proc. R. Soc. London, Ser. A 1964, 278, 1-26.

2. P. G. Lenhert, Proc. R. Soc. London, Ser. A 1968, 303, 45-84.

3. (a) S. W. Hawkinson, C. L. Coulter, M. L. Greaves, Proc. R. Soc. London, Ser. A 1970, 318, 143-147. (b) H. Stoeckli-Evans, E . Edmond, D. C. Hodgkin, J. Chem. Soc., Perkin Trans.2 1972, 605-614.

4. M. Rossi, J. P. Glusker, L. Randaccio, M. F. Summers, P. J. Toscano, L. G. Marzilli, J. Am. Chem. Soc. 1985, 107, 1729-1738.

5. (a) J. P. Glusker in B12 (Ed.: D. Dolphin), Wiley, New York, 1982, Vol. 1, p. 23-106. (b) M. Rossi, J. P. Glusker in Molecular Structure and Energetics (Eds.: J. F. Liebman, A. Greenberg), VCH Publishers, Weinheim, FRG, 1988, Vol. X, p.1-58.

6. L. Randaccio, N. Bresciani-Pahor, E. Zangrando, L. G. Marzilli, Chem. Soc. Rev. 1989, 18, 225-250.

7. C. L. Drennan, S. Huang, J. T. Drummond, R. G. Matthews, M. L. Ludwig, Science 1994, 266, 1669-1674.

8. F. Mancia, N. H. Keep, A. Nakagawa, P. F. Leadlay, S. McSweeney, B. Rasmussen, P. Besecke, O. Diat, P. R. Evans, Structure 1996, 4, 339-350.

9. G. Färber, Dissertation, Universität Graz, 1993

10. K.Gruber, Dissertation, Universität Graz, 1994

C. Kratky, G. Färber, K. Gruber, K. Wilson, Z. Dauter, H.-F. Nolting, R. Konrat, B. Kräutler, J. Am. Chem. Soc. 1995, 117, 4654-4670.

12. The crystal structure of B12r was determined in our laboratory several years ago, using conventional diffraction data [13]. The synchrotron data from the same compound were collected recently, and have not yet been published.

13. B. Kräutler, W. Keller, C. Kratky, J. Am. Chem. Soc. 1989, 111, 8936-8938.

14. G. Klintschar, Diplomarbeit Universität Graz, 1993.

15. (a) H. F. J. Savage, Biophys. J. 1986, 50, 947-965. (b) H. F. J. Savage, Ibid, 1986, 50, 967-980. (c) F. Vovelle, J. M. Goodfellow, H. F. J. Savage, P. Barnes, J. L. Finney, Eur. Biophys. J. 1985, 11, 225-237. (d) H. Savage, A. Wlodawer, Methods Enzymol. 1986, 127, 162-183. (e) H. F. J. Savage, J. L. Finney, Nature 1986, 322, 717-720. (f) J. L. Finney, H. F. J. Savage, J. Mol. Struct. 1988, 177, 23-41. (g) J. P. Bourquiere, J. L. Finney, M. S. Lehmann, P.F. Lindley, H.F. J. Savage, Acta Crystallogr. B 1993, 49, 79-89.

16. G. Jogl, K. Gruber, P. Langan, C. Kratky, unpublished result.

17. V. B. Pett, M. N. Liebman, P. Murray-Rust, K. Prasad, J. P. Glusker, J. Am. Chem. Soc. 1987, 109, 3207-3215.

18. J. D. Dunitz, X-ray analysis and the structure of organic molecules, Cornell Univ. Press, 1979.

19. J. D. Dunitz, V. Shoemaker, K. N. Trueblood, J. Phys. Chem. 1988, 92, 856-867.

20. H. Bertagnolli, T. S. Ertel, Angew. Chem. 1994, 106, 15-37.

21. (a) I. Sagi, M. R. Chance, J. Am. Chem. Soc. 1992, 114, 8061-8066. (b) I. Sagi, M. D. Wirt, E. Chen, S. Frisbie, M. R. Chance, J. Am. Chem. Soc. 1990, 112, 8639-8644. (c) M. D. Wirt, I. Sagi, E. Chen, S. M. Frisbie,l R. Lee, M. R. Chance, J. Am. Chem. Soc. 1991, 113, 5299-5304.

22. D.J.A. De Ridder, E. Zangrando, H.-B. Bürgi, J. Mol. Struct. 1996, 374, 63-83.


Table 1: Crystal structure analyses of cobalamins

Each structure is represented by its name, the crystallographic residual and the year of publication. For structures which were determined more than once, the year of the first structure determination and the R-value of the most accurate structure determinations are given (i.e. the two values usually do not refer to the same structure analysis in these cases). Structure determinations marked with an asterisk (*) were carried out in our laboratory recently, but have not yet appeared in print.








13-epi-Cyanocobalamin (Neo-vitamin B12)



Vitamin B12-5'-phosphate















Cyanocobalamin-monocarboxylic acid






5'-Deoxyadenosyl-cobalamin ("Coenzym B12")



Cyanocobalamin (Vitamin B12)



















































Aquocobalamin perchlorate



Cob(II)alamin, B12r





Figure 1

Superposition of 12 cobalamin molecules as observed in the corresponding crystal structures. While not all compounds of Table 1 were included, the figure gives a representative sample of the known 3D structures of cobalamins.

Figure 2

Correlation diagram of c/a versus b/a for the cobalamin crystal structures of Table 1, excluding methylcobalamin and cyanocobalamin monocarboxylic acid. The triangle represents the crystal structure of aquocobalamin perchlorate.

Figure 3

Density section through the solvent region in crystals of cob(II)alamin, as computed from synchrotron X-ray data (a) and from neutron diffraction data (b). While both maps show the positions of oxygen atoms, the neutron data also reveal the location of deuterons.

Figure 4

Scatterplot of corrin-ring fold angle versus axial Co-N (DMB) bond length. The 16 most accurate structures of Table 1 were used. The point marked by a triangle corresponds to the crystal structure of Coß-cyano-imidazolylcobamide.

Figure 5

Superposition of cobalamin structures. A detailed explanation is given in the text.

Figure 6

Thermal ellipsoids of corrin ring and DMB base from the crystal structures of aquocobalamin perchlorate and cob(II)alamin. The R-values denote the agreement between observed a.d.p.s and those calculated from a rigid-body TLS analysis of the corresponding fragment(s).

Figure 7

Superposition of the k3 weighted EXAFS spectra of dissolved (broken line) and crystalline (solid line) aquocobalamin perchlorate

Figure 8

Summary of available evidence for the effect of an exchange of DMB by imidazole on the axial Co-N bondlength. Left: crystal structure of imidazolo cobalamin; centre: Cob-cyano-imidazolylcobamide; right: vitamin B12. According to these data, exchange of DMB by imidazole decreases the axial Co-N bondlength by about 0.05 A.