X-ray and neutron diffraction studies of the crystal and molecular structure of the predominant monocarboxylic acid obtained by mild acid hydrolysis of cyanocobalamin. Part III. Neutron diffraction studies of wet crystals

Prec. Indian Aead. Sci. (Chem. SCi.), Vol. 93, No. 3, April 1984, pp. 235-260.

9 Printed in India.

X-ray and neutron diffraction studies of the crystal and molecular structure of the predominant monocarboxylic acid obtained by mild acid hydrolysis of cyanocobalamin. Part III. Neutron diffraction studies of wet crystals

F H M O O R E , B H O ' C O N N O R * , B T M W I L L I S t a n d D O R O T H Y C R O W F O O T H O D G K I N : ~

Neutron Diffraction Group, Australian Institute of Nuclear Science and Engineering, Sutherland, NSW 2232, Australia

* School of Physics and Geosciences, Western Australian Institute of Technology, Kent Street, Bentley, Western Australia 6102

t Materials Physics Division, AERE, Harwell, Oxon OXI 10RA, UK

:~ Chemical Crystallography Laboratory, Oxford, UK: Mailing address: Crab Mill, Ilmington,

$hipston-on-Stour, Warwickshire, UK

Abstract. The molecular structure of the predominant monocarboxylic acid, E2, obtained by mild acid hydrolysis of cyanocobalamin has been determined by neutron diffraction with some support from x-ray diffraction. The undried crystals, formula probably C63H87OIsN13PCo.16H20, have unit cell parameters a = 14.91(1)A, b = 17-49(1)A, c = 16-41(1) A, fl = 104"11(5)~ space group P21, Z = 2.

The analysis was carried out in two stages with data extending to 1.3 A (1531 terms) and to 1"0 A (2993 terms) respectively. It was initiated by the use of coordinates of 84 atoms from the parallel x-ray analysis to phase the first Fourier series. The atomic positions derived from the 1-3 A data set appeared in good agreement with the chemical evidence both on the corrin structure and on the position of the acid group at e. However the analysis by Fourier and least squares methods on the extended data defined the atomic positions much more dearly and showed that diffraction ripples had distorted some of the hydrogen atom positions in the low resolution map. The acid group appeared clearly placed in the higher resolution map at the b position. The positions of disordered atoms in the e chain and some water molecules were checked with the parallel x-ray analysis.

It seems most likely to us therefore that this acid is ~-(5,6-dimethylbenzimidazolyl)cobamic acid a,c,d,,q,e pentamide cyanide.

Keywords. a monoacid of vitamin B12; neutron diffraction; identification of the propionic acid position; Fourier resolution.

1. Introduction

T h e m o n o c a r b o x y l i c acid E2 first o b t a i n e d as the p r e d o m i n a n t i s o m e r by the mild acid hydrolysis o f c y a n o c o b a l a m i n a p p e a r e d i m m e d i a t e l y to be a very attractive material for a n e u t r o n diffraction study. T h e crystals were large, c a 4 m m a n d m o r e across, a n d beautifully f o r m e d as s h o w n in figure 1. It was clear that a n e u t r o n diffraction study o f these crystals o u g h t to show, n o t only the p o s i t i o n o f the single acid a m o n g the a m i d e side chains b u t also, by the definition o f all the h y d r o g e n a t o m s present, it s h o u l d complete the d e t e r m i n a t i o n by diffraction m e t h o d s o f the chemical structure o f the

* To whom all correspondence should be addressed.

Proc (Chem. Sci.) - 4

235

236 F H Moore et al

\

0TI f

,/

TTO /

~b

Figure 1. Diagrams illustrating the habit of the Btz monocarboxylic acid E2 crystals.

vitamin. Figure 2 illustrates the problem and shows the structure eventually preferred.

Table 1 shows the scattering factors for the atoms present for neutrons compared with those for x-rays.

After a preliminary feasibility study, the analysis was carried out in two stages. The first stage was based on 1531 reflections collected to a spacing limit o f 1-3 A and led to a number of conclusions we now think are wrong, particularly that the acid group was at e as suggested by Bernhauer and others. The extended analysis, based on the data collected to 1 A spacing, indicated position (b) as that of the acid group in the molecule.

The parallel x-ray analyses of the monoacid were useful at two stages in the neutron analysis. At the beginning, 84 atomic positions derived from the wet crystal study initiated the structure factor and Fourier calculations from the neutron diffraction data. Towards the end, a comparison with the positions reached in the x-ray analysis helped in the final selection of the positions o f disordered atoms and some water molecules.

2. Experimental

2.1 Crystal data

The crystals used were air-dried crystals given by Dr E Lester Smith and wet crystals given by Dr F Wagner. The crystal data may be summarised as follows:

Part III. Neutron diffraction studies of wet crystals 237

(_b) coo.<,,~c.: , t ~ NtI~-CO'Cil.

CtI'c/(

/ "1r CIt

NHaCO-CtI-CII 1) I

(a)

c,f.-i'.

.~.,'--o~

1:1

N=c (a~

( 'c,,

\,~-c:

tl C ~c~C"3

-C~..C~ CH3

~r

;ol

6 3 ~ ! " ,16

S 8 ~ 5 4 ~53 4~ 49b'~4~51

Pr 1~"459 ^^

B4

. s . - $ " ~ (b)

g5

Figure 2. (a) Structure found for B~ monoacid E2; (b) Numbering scheme used.

Table 1. Neutron scattering factors compared with the corresponding x-ray values.

Neutron X-ray (at sin 0/2 = 0)

Cobalt 2.50f 27.00 electrons

Carbon 6.61 6.00

Nitrogen 9-40 7.00

Oxygen 5.77 8.00

Phosphorus 5-30 15-00

Hydrogen - 3-78 1-00

Monocarboxylic acid E2 from vitamin B12

"C63Ha7OlsNI3P"

1356-38 + 16H20; monoclinic, {110}, {011} {li0} {0il} and {10i}; {10i} tending to predominate. The cleavage is marked in the (10i) plane, dry crystals cleave more easily than do wet crystals.

Unit Cell parameters: a = 14.91(1) A, b = 17.49(1) A, c = 16-41(1) A, fl = 104.11(5) ~ derived from least squares fit of intensity profiles during measurement o f neutron data with d > 2 A; space group P2~, Z = 2, density measured by flotation in a mixture o f benzene and methyl iodide, 1-339; calc. for 16 H 2 0 , 1.316;/~ = 2.38 c m - 1 (for thermal neutrons).

2.2 Neutron diffraction measurements

A dry crystal of the acid was used first for the initial feasibility measurements. All peaks observed were broad, ,,, 65' wide at half height and often doubled. The substantial background due to incoherent scattering from hydrogen was reduced by installing a receiving slit at the aperture o f the tube. Unfortunately the original dry crystal was accidentally destroyed and it was decided to change to wet crystals.

The neutron intensity data were measured using the A.E.R.E. Pluto reactor as a source of thermal neutrons and, initially, a manual three-circle goniometer. Following

238 F H Moore et al

the first tests, reflection profiles were investigated within the range 0 = 3"5-10 ~ using the automated Ferranti Mark I neutron diffractometer (Arndt and Willis 1962) and a wavelength of 1"533 A. Extinction was found to be low, resolution sufficient to allow a wide scan of peak and background. The very high background levels (1500 counts per minute at 0 = 5 ~ were reduced to 700 counts per minute with the slit system at the counter aperture.

2.3 The 1"3 A data collection

The crystal of E2 was sealed in a quartz tube, with sufficient mother liquor to keep the crystal wet and was supported by a small wad of quartz wool saturated with mother liquor. It was held to the side of the tube by surface tension. The intensity profiles for all reflections with d > 1"3 A with the exception of (100) and (001) were recorded in three shells, viz 0 = 3"5-20 ~ 20-30 ~ and 30-36 ~ The time spent on counting was 30 minutes per reflection; the standard reflections, 040, 302, 40Tand 701 were counted after every tenth reflection.

In reducing each profile to an estimate of F0 the first and the last N/4 points of the total N in the profile were taken as background values B 1 and B2 respectively and the central N/2 points were taken as peak counts. The integrated intensity is therefore I = P - B 1 - B 2 with an associated variance 0"2(1)= P + B I + B 2. After the data reduction was complete, all reflections for which tr(Fo)/Fo was greater than 1/3 were rejected i.e. a confidence level slightly less than 90 ~o was chosen. No corrections were applied for absorption effects or the 2/2 components of the incident beams.

2.4 The 1"0 A data collection

Following the 1"3 A measurements, the copper monochromator was replaced with a beryllium crystal in order to intensify the flux at the specimen position and to extend the area of effective beam uniformity. An incident beam, free from 2/2 contamination, was obtained by optimising the orientation of the (301) planes of the monochromator crystal in order to reflect neutrons of wavelength 1.448 A. The resultant increase in incident beam intensity for the 1.0 A collection was offset by diminished reflectivity due to the slight decrease in wavelength, with the net result that the intensity levels for the 1"0 A and 1.3 A collections were almost identical.

The reflection intensity profiles were recorded using the measuring procedure described above for the 1-3 A data collection, with a total measuring time of 30 minutes per profile div~ted equally between peak and background. Analysis of the intensity statistics showed clearly that there is a virtual data cut-off near d = 1.0 A for the given experimental conditions. The final set of high-angle data consisted of 2167 reflections and only 1351 of these had I >! a(I).

During the course of the high-angle data collection the crystal (viz the same sample used for the 1.3 A measurements) began to cleave and had to be replaced by another wet crystal. It was therefore decided to measure the intensities of a set of thirty medium- level 1"3 A reflections at 2 = 1-448 A for the two crystals in order to place all measurements on the same scale. The final correlated set of reflections comprised 4200 reflections of which 2993 had I >>.a(1). The complete set of IFol 2 and the corresponding e.s.d's are listed in table 2.

Part I l L Neutron diffraction studies of wet crystals 239 3. N e u t r o n s c a t t e r i n g d e n s i t y c a l c u l a t i o n s

3.1 The 1-3 A resolution analysis

The 1.3 A study was based entirely on Fourier methods. Although it was computed, the Patterson distribution, as expected, was uninterpretable; the highest scattering atoms were 13 nitrogen atoms. The solution therefore was based initially on positions for 84 of the 93 atoms in the molecule found in the parallel x-ray analysis, II. F r o m the first electron density and difference distributions, these positions were partly refined and employed as coefficients for the next series. Positional shifts were derived from the density gradients in each Ap synthesis at the proposed nuclear site and the individual isotropic parameters were found by a trial and error method according to the residual density in Ap at each nuclear site. In all seven rounds of alternate structure factor and Fourier series were computed, as in flow sheet 1, new atomic positions being added in the phasing calculations as discovered. By the seventh series 205 atomic positions had been derived, 93 vitamin non-hydrogen atoms (expected 93) 87 vitamin hydrogen atoms (expected 87), 14 water oxygen a t o m s (expected 16) and 11 water hydrogen atoms (expected 32).

In p 7 the skeleton as a whole of the molecule seemed well-defined, apart from the longer side chains and a few hydrogen atoms which were not well-placed. However the derivation of the positions o f the atoms in the e chain had given difficulty and so had many o f the water molecules; the subsequent extended study has shown these were largely incorrect. The relative heights o f the atoms, particularly in the propionic residues, were very inaccurate; peaks nearby thought to be hydrogen atoms have vanished since, particularly two near phosphate oxygen, P5 and P4, one near a t o m 0 3 4 and two near a t o m 045. These peaks had led us to assignments we now believe

Flow sheet 1. The course of the low angle (I.3A data) analyses.

Neutron structure analysis.

Feasibility analysis Unit cell refinement

Data collection and reduction 1531 reflections at 99 % confidence level

X-ray analysis -~ 84 atoms, structure factors, R = 0-42 pl, Apl,

refinement of 84 atoms, R = 0-38 p2, Ap2

159 atoms, including 75H R = 0.33 p3 - 6 Ap3 - 6

refinement and addition of atoms p7, Ap7

93 monoacid non-hydrogen atoms + 87 monoacid hydrogen atoms 14 water oxygen atoms + 11 water hydrogen atoms.

240 F H Moore et al

erroneous, that the group b, 33, 34 was an amide, that the amide 44 and 45 was disordered, that there was disordered hydrogen attached to the phosphate group and that e is probably the acid group (Moore et al 1967).

Table 3 summarises the relative heights of the terminal oxygen and nitrogen atoms at this stage.

3.2 The 1"0 A resolution analysis

The extension of the refinement with data measured out to the 1.0 A limit substantially improved the resolution of the Fourier representation of the neutron scattering density.

Unexpectedly it also changed aspects of our model and particularly the interpretation of the propionamide and propionic acid side chains. The first stages, (i), of the

Table 3. S u m m a r y o f peak neutron scattering-densities ( x 10f/A 3) for the amide chains following the low-angle (1.3A data) refinement. The S and B are site-occupancy and thermal parameters, respectively, and p~ and p~ are, respectively, the m a x i m u m densities in the vicinity o f each site when the listed a t o m s are included in or omitted from the structure factor calculation.

Group identification A t o m identification S B p~ pc" p~176 p~/p~

a-acetamide C27 1.0 4.7A 2 55 29

0 2 9 1.0 5" 1 42 23 0-76 0-79

N28 1.0 5'1 88 49 1"60 1.69

N28H1 1.0 5.4 - 4 0 - 2 9 - 0 . 7 3 - 1 . 0 0

N 2 8 H 2 1-0 5.4 - 34 - 23 - 0.62 - 0.79

c-acetamide C38 1.0 4.6 46 27

0 3 9 1 0 5.2 51 27 1.10 1.00

N 4 0 1 "0 5.2 94 52 2-04 1-92

N40H1 1"0 5"5 - 3 7 - 2 3 - 0 - 8 0 -0"85 N 4 0 H 2 1"0 5-5 - 3 4 - 1 8 - 0 ' 7 4 - 0 - 6 7

g-acetamide C61 1.0 4.5 54 31

0 6 2 1.0 4.7 46 22 0-94 0.71

N63 1.0 5-2 86 45 1-60 1.45

N 6 3 H I 1.0 5.2 - 5 0 - 3 0 - 0 - 9 3 - 0 . 9 7

N 6 3 H 2 1-0 5-8 - 3 0 - 2 1 - 0 . 5 6 - 0 . 6 8

b-propionamide* C32 1.0 4.9 64 29

0 3 3 1.0 5.4 45 21 0.70 ff72

N34" 1.0 5.7 62 32 0.97 1.11

N 3 4 H I 1"0 6-5 - 4 6 - 37 - 0 . 7 2 -- 1.27

N34H2** 1.0 6-5 - 2 4 - 1 5 - 0 - 3 7 - 0 . 5 2

disordered C43 1.0 5-2 46 26

d-propionamide* A44"# 1.0 6.0 53 29 1-15 1.12

A45*t 1.0 5-7 54 25 1.18 0-96

A44H 1 * * 0. 5 7.0 - 29 - 24 - 0.63 - 0-92

A44H2** 0.5 7.0 - 2 5 - 1 8 - 0 - 5 4 - f f 6 9

A45H1 0-5* 7.0 - 2 8 - 2 1 - 0 - 6 1 - 0 - 8 1

e-propionic acid* A45H2 0-5* 7-0 - 2 2 - 18 - 0 - 4 8 - 0 - 6 9

C50 1"0" 5'8 46 25

O51" 1.0" 6.9 37 13 0-80 0.52

0 5 2 * * 1.0 5-0 36 12 0-78 ff48

* Amended during the I ' 0 A analysis; ** Rejected during the 1.0A analysis; * Site disordered betwoen oxygen and nitrogen.

Part I l L Neutron diffraction studies of wet crystals 241 refinement were carried out by p and Ap syntheses as before, followed by least squares calculations, (ii). In the final stages of definition of disordered atomic positions, (iii), a comparison was made with the conclusions of the parallel x-ray analysis of the same crystal (compare II).

The course of the 1.0 A analysis is summarised in flow sheet 2.

1. In order to gain an impression of progressive improvement in resolution the first maps of this series p8, Ap8 were calculated on data extending to 1"1 A only. Phase angles were computed on 168 only of the atoms defined from p7; the less well-defined atoms omitted included hydrogen atoms on the terminal atoms of the b and d chains, all atoms at the end of the e chain from 49 on, B4H and B5 and all the water molecules. The map showed a marked relative change in the peak heights of the d chain amide, which had been weighted as oxygen atoms, together with only two negative peaks, adjacent to

Flow sheet 2. The course of the high-angle (1.0A data) analyses.

Experimental

Data collection and reduction

(2167 reflections measured, 1351 with 1 > a (I))

Data correlation (4200 correlated data, 2993 with I > a (I)) Structure analysis

Initial model (168 atoms selected from I'3A model) R = 0-290

p8, Ap8 (1.1A data)

p9-13, Apg-13

Refinement of positional and thermal parameters b-chain identified as propionic acid

d-chain identified as ordered propionamide e chain disordered, difficult to characterise

186 atoms established R = 0-210

preliminary least squares, 6 cycles R = ~193

p14, Apl4

e chain placed as disordered propionamide least squares, 3 cycles, expt. weights

R = 0.183 p15, Apl5

10 water molecules + disordered sites of e chain added least squares cycles

p16, Apl6 p17, Apl7

X-ray data compared, removed e chain and 10 water molecules p18, Apl8

added two positions e chain and 8 water molecules, two disordered, p19, Apl9

242 F H M o o r e et al

the heavier peak, 45, indicating this was an amide nitrogen atom. Small adjustments were suggested for other atoms. The following maps, p9, Ap9, were screened for peaks which might represent water molecules. Eight out of 45 possible peaks were chosen;

these 8 all appeared in good hydrogen-bonded positions. It was realised that the e chain was almost certainly disordered and these positions continued undefined. On the other hand only one hydrogen atom appeared attached to atom 34. That this was the second oxygen of an acid group was strongly indicated by the following pl0 and Apl0. In pl0, atoms 33 and 34 appeared nearly equal in height, while a negative peak appeared at 34, weighted as nitrogen, in Ap 10. Some refinement occurred in the following three rounds but no substantial improvement in the definition of the e chain. Least squares refinement was therefore initiated.

2. Since the least squares refinement was to be performed for approximately 800 parameters with only 2993 observations it began cautiously, using the block diagonal approximation. Initially only the 120 atoms which satisfied the criterion (p~ 3)/(Ap~ 13)

> 10 were refined, least squares weights of unity were assigned and the thermal parameters constrained to those from Ap 13. The convergence in the initial two cycles was very encouraging and a further four cycles in which the positional and thermal parameters of all atoms were refined, followed. There was however very slow convergence for 11 hydrogen atoms and the calculation of p 14, Ap 14 was undertaken to reposition these and the atoms of the e side chain.

In p 14, Ap 14 better evidence was obtained for the positions of the omitted hydrogen atoms and two plausible alternative positions could be derived for the disordered atoms of the e chain. It seemed possible to assign weights to the peaks that appeared corresponding to half weight oxygen and nitrogen atoms.

The positional and thermal parameters of all 191 atoms were next refined; three cycles reduced all shifts below one estimated standard deviation. To cross check the identification of the amide nitrogen and oxygen atoms six further least squares cycles were then performed in which the identities of the adjacent nitrogen and oxygen atoms were reversed and atoms 33 and 34 were both weighted as nitrogen. The only parameters refined were the thermal parameters of these particular atoms. The results, shown in table 4 strongly support the proposed new identifications, allowing for the still imperfect definition of the e chain. At this point it seemed clear that the problem of the molecular structure was essentially solved, though a number of water molecules still remained unplaced. A calculation of p15 and Apl5 was carried out to conclude this stage.

3. p15 and Apl5 were used next to attempt to complete the crystal structure determination by placing the remaining water molecules. Ten further plausible positions were selected among the difference peaks and also a possible third position of the e chain. Additional least squares cycles further reduced R; they were interleaved with two electron density maps which suggested removing certain water molecules and reducing the occupancy of others.

At this point it became possible to compare the refined atomic positions derived with those obtained in the parallel x-ray analysis of the monocarboxylic acid crystals (II).

There was very close agreement between the positions of the atoms in the molecule except over the e chain, which had been interpreted in terms of three different conformations in the two analyses. In addition only eight of the water molecules had been assigned even approximately similar parameters. In each series therefore new electron density and difference maps were calculated, omitting once again the atoms 49,

Part III. Neutron diffraction studies of wet crystals 243

Table 4. Least squares shifts in the thermal parameters (after six cycles) upon interchanging

the identities of oxygen and nitrogen.

B B AB

(final model) (identities interchanged)

029 6"3 12"5 6'2

N28 4'1 4"2 0.1

039 5' 1 11'7 6"6

N40 4-7 5"2 0-5

062 4"5 9"2 4"7

N63 5"9 6"0 ffl

033 4-8 lff5 5'7

034 6-2 12"4 6"2

044 11-5 22"5 11"0

N45 7-7 7"9 0-2

052* 14-0 (diverged) > 20-0

N51* 10-5 14.9 4.4

052* 12-2 (diverged) > 20-0

N51* 12.1 14.9 2"8

* The latest solutions contain no atoms near the sites of the oxygen atoms but still atoms near

the sites of the proposed nitrogen atoms.

50, 51 and 52 and all but the 8 similar water molecules. In these maps it was possible to

see two conformations o f the e chain which agreed with one another and also evidence for seven additional water molecules near closely similar positions. One further possible water molecule site was recognised in the next round o f calculations. The appearance of the additional water molecules indicate that several of these occupy more than one adjacent site; others are low and spreading corresponding with space-filling difficulties. There is evidence for quantitative differences in the relative site occupancies observed in the x-ray and neutron maps. These are probably real corresponding with differences in crystal growth, which are indicated by the small differences in the recorded lattice constants. The atomic positions finally reached are recorded in table 5 and the observed and calculated structure factors on which they are based in table 2*.

Even within the molecule the estimates o f error are quite large. F o r example the individual C--C distances in the corrin nucleus have ESDS of 0" 1 A. N o great significance can therefore be attached to differences in distances within the amide groups. F o r the record, they are shown in figure 3.

4. T h e d e f i n i t i o n o f t h e m o l e c u l a r s t r u c t u r e

The neutron scattering density observed is shown in a s u m m a r y f o r m in figures 4a and b. Almost all the non-hydrogen atoms are resolved at half height or better and all are clearly resolved at peak maximum. However it should be realised that the peaks are superimposed on a background of diffraction ripple, shown particularly by the elongation of most o f the hydrogen peaks. In spite o f this effect the positions o f the

* can be obtained from authors on request.

244 F H M o o r e et al

~g 8

.=,

0

=o

"a

0

•

0

• .

&.=.

r

~ ' ~ t ' ~ . l o'1 r -.~r t"q r . ~ 1 ' , ~ : 1 - ~ " ~ ~ e ~ e ~ ' ~ l -

0 0 0 0 ~ 0 0 0 0 0 O 0 ~ Z ~ 0 0 0 0 0 0 0 0 0 0 0 0 0

' , ~ t ' ~ l "~ ~ ~ r ~ , , ~ - ~ * - ~ . t ~ ~ . . . ~ , ~ , ~ i".--t"-.-~"-~r

~ r,.) r ~ r r r,.) r,.) ~ 0 0 ~ r,.) r,.) r,.)

Part III. Neutron diffraction studies of wet crystals 245

O 0 0 0 0 0 0 0 0 0 0 0 ~ 0 Z Z ~

o o o o z z z o o o o o o o

246 F H Moore et al

[ -

q " ~ " ~ " r r162 r ~ ' ~ - " r 1 6 2 r r ~'~l r ~l" r r r t ' ~ r t",4 r ~ .~l- ,~l" o ' - ~ ' , ~ o o r-~ t-q

Part III. Neutron diffraction studies of wet crystals 247

~d

[-

2 4 8 F H Moore et ai

observed hydrogen atoms clearly define the existence of the resonating bonds in the corrin nucleus and complete by diffraction methods the determination of the chemical structure of the nucleus.

The geometry of the molecule is shown in figure 5. It differs from that found in cyanocobalamin by a small change in the relative positions of the benzininazole sugar- phosphate chain and the corrin nucleus, which is described in I and II. To the information already collected there on the heavier atom positions, we can now add details of the hydrogen atom distribution. All the methyl groups appear fixed, not

33 ?4

29 28 oJ

%:^ _+,+~ ,3r ₃₅

36 3 7

6 3

pJO

y

4 5

58 ~

5 9

53

5211

B3

N59 B2 ~e

Prl R7 -

BI B5 BIO

Pr3 t 56 Z ~ B6

P| : " II

P5 ~ ~ R4

133

R5 R8

Figure 3. (u)

Part III. Neutron diffraction studies of wet crystals 249

,.I ~,

z

o

z ~ u _{9 i} 2

"o

z ~

/ =

,i ~. r

o _ o

=

t,,t,.

9

.. I

9

= o

ca.

o

" o

8

0

r

0

.=

I

.o

8 a a

250 F H M o o r e et al

rotating; both the methyl groups of the benziminazole ring and 35 and 53 attached to the corrin ring have conformations in which one CH bond is in the plane of the ring.

Other methyl groups, e.g. 46 and 47 have generally staggered conformations.

The characteristics of the acid and amide groups are illustrated in figures 6, 7 and 8 and table 6. In each figure the appearance of the atoms is shown in maps computed with and without the atoms illustrated included in the phasing calculation. All the acetamide groups are very well defined. In all three (table 6) there is excellent agreement between the mean observed acetamide O : C, N: C and H : C peak height ratios 0-80, 1"37 and -0-54 respectively and the expected values, 0-87, 1-42 and -0.54. The diffraction

f \

[..-~ /

\ .

F - - ! i p l i ~ \ \ i \

i - ' M ~ , r ~ . , \ -- i T f4 -"5-l/fr:.,.~,--'~4"-) I

C

. - . l~-',,,~.l({r(~ ~ ( "

/ ( 7 - " " ...

t / /

\ /1 t t ~,.~ , /I

(,2 ,Iii I;" "

I

( ~ ~ L ~ = ' ~ ~ ) - L / , - - - . - - - - , " ~

~ " ~'~'~ ^{t ' "} . . . . . . I

\ ( \ ~ , ,,

iC'< \

k t I ~ / /

/(

L )

Figure 4u. Summary of the neutron scattering density synthesis for the corrin nucleus projected on the best plane for N21, N22, N23 and N24. One conformation only of the e chain is shown. The first contour is at 2.5f/A 3 and the contour interval 2-5f/A 3.

Part III. Neutron diffraction studies of wet crystals 251

\ c ~---sj4 "

",\ ~,~)1/

/

//,'--'~

~

, : \ \

,J

f~

l ~ a r e 4b. Summary of the neutron scattering density synthesis. The propanolamine, nuclcotide and cyanide projected on the mean plane of the bcnzimidazolr ring.

9 ^{N=t roge~}

0 carbon

0 | ly~Jrooort Fignre 5. View o f the molecule.

Proc (Chem. Sci.) --5

252 F H Moore et al

Figure 6. The neutron scattering-density synthesis at 1-0A resolution in the best plane for atoms (a) C26, C27, 029, N28, N28H 1, N28H2, (b) C37, C38, 039 and N40, N40H1, N40H2, (c) C60, C61, 062 and N63, N63H1, N63H2. The diagram on the left is the Fourier representation when all atoms are included in the phase calculation, and that on the right is the corresponding representation when the terminal atoms are omitted from the phase calculation.

The molecular skeleton represents the model at p 15, with hydrogen atoms indicated by dots.

The first level is at l'0f/A 3 and the contour interval is l'5f/A 3.

ripples which were c o n f u s i n g in the 1.3 A m a p s are relatively m u c h less p r o m i n e n t here a n d less easily m i s t a k e n t b r h y d r o g e n a t o m positions.

T h e p r o p i o n i c acid at b also appears well-defined; the single O H g r o u p is clearly h y d r o g e n b o n d e d to the water molecule, 101. T h e r e is still a second b u t n o w very small

Part III. Neutron diffraction studies of wet crystals 253

_b-chn~ ta'~x'n~mn I tO~ D ~

/ z / _ . \ ~ 0 ~ 3

t l .

i t ; t t

5 , . - ' . ' d

(i)

;<, 9

/ . f - ~ \

/" ~\ f

I t-~,, \\

(b)

i:2-chn~ terminatEon l 1.0~ IData

~ . ~ ;.- _.,._. , x ~

I I \ \

..I ii I i "1 ~ 0

(el F~-~.t~,~

,2

0

Figure 7. The neutron scattering-density synthesis at 1.0A resolution in the best plane for atoms C31, C32, 033, 034, O34H. The diagram on the left is the Fourier representation when all atoms are included in the phase calculation, and that on the right is the corresponding representation when the terminal atoms are omitted from the phase calculation. The molecular skeleton represents the model at p 15, with hydrogen atoms indicated by dots. The first level is at 1-0f/A 3 and the contour interval is 1-5f/A 3. (c) and (d) show the density over the corresponding atoms in the Fc and difference density map.

negative peak near 034; as this also appears on Fcalc maps where it was not introduced, it is certainly a diffraction effect (figure 7). Similar very small peaks occur in the d chain maps where oxygen and nitrogen are nevertheless well differentiated (figure 8). The e chain is certainly the least clear. Two of the atoms in the two conformations overlap which makes the relative peak heights uncertain; figure 9 shows the electron density over the centre of each peak in the latest maps. It seems most likely that 152 and 251 are nitrogen atoms but there may be additional disorder involving rotation of the amide groups (figure 9). All but one of the outside contacts made with the atoms of this group are with water molecules. All the amide groups show a close approach to planarity; the plane includes C--C, O, N and the two amide hydrogen atoms, (table 7).

254 F H M o o r e et al

- c h e ~ n Z , r r r ~ t ~ / 1.0~

(,(5-_4-J' 3 ) ~

F~ f ]

;

,_-"-) (27 o4'4 f, ~ - - "

]

/ J

~t/ i;/

Figure g. The neutron scattering-density synthesis at 1.0A resolution in the best plane for atoms C42, C43, 044, N45, N45HI, N45H2. The diagram on the left is the Fourier representation when all atoms are included in the phase calculation, and that on the right is the corresponding representation when the terminal atoms are omitted from the phase calculation.

The molecular skeleton represents the final I'0A model, with hydrogen atoms indicated by dots. The first level is at 1.0f/A 3 and the contour interval is l'5f/A 3.

13 13

A B

Figure 9. The neutron scattering density over the e chain.

5. The crystal structure

The arrangement o f the molecules in the crystals is illustrated in figure 11. It is very different f r o m that in cyanocobalamins. The molecules lie with the corrin nucleus planes of adjacent molecules approximately at right angles to one another. They are linked together by six hydrogen bonds directly involving active groups:

N H 2 2 8 . . . O 7 2 , 0 2 9 . . . O H 81, O 3 3 . . . N H 2 2 8 , N H 2 4 5 . . . O 6 9 , 0 4 4 . . . 152(?NH~) N H 2 4 5 . . . N C 6 5 .

All but one of these are bonds characterised by well-defined hydrogen atoms.

The water molecules lie between the monocarboxylic acid molecules roughly along the (10]-) plane, an arrangement which accounts for the pronounced crystal cleavage in this plane. Some o f them, well ordered, take part in linking together active groups on adjacent molecules. An example is illustrated in figure 10. One, 104, links N59 internally within one molecule to 062, as in many other B12 crystals. But several o f them form somewhat disordered chains as in figure 12. The definition o f these molecules in the

Part III. Neutron diffraction studies of wet crystals 255

Table 6. Summary of peak neutron scattering-densities ( x 10f/A 3) for the amide chains following the high-angle (1-0A data) refinement. The S and B are site-occupancy and thermal parameters, respectively, and p~ and p~ are, respectively, the maximum densities in the vicinity of each site when the listed atoms are included in or omitted from the structure factor calculation.

Group identification Atom identification S B p~ pO, p~176 p~'/p~'(C)

a-acetamide C27 1-0 3"9 118 73

029 1-0 6-3 77 54 0-65 ff74

N28 1.0 4.1 153 97 1.30 1.33

N28H1 1-0 4.5 - 6 0 - 4 7 -if51 -0-64

N28H2 1.0 5.5 - 5 6 - 4 0 -0.48 -0-55

c-acetamide C38 1"0 4-4 92 55

039 1"0 5.1 78 46 0-85 0.84

N40 1'0 4-7 152 107 1-65 1.94

N40H1 1-0 3.2 - 7 0 - 4 4 -0.76 -0-80

N40H2 1.0 6.7 - 4 3 - 3 3 -0-47 -0-60

o-acetamide C61 1.0 3"3 100 58

062 1"0 4.5 90 60 0.90 1-03

N63 1-0 5'9 117 73 1-17 1"26

N63HI 1'0 6"1 - 5 6 - 3 3 -0-56 -ff57

N63H2 1.0 9' 1 - 45 - 39 - 0.45 - 0-67

b-propionic acid C32 1.0 3-5 122 76

033 1.0 4.8 76 51 0"62 0-67

034 1-0 6-2 77 53 0"63 0-70

O34H 1-0 6.4 - 57 - 32 - 0-47 - 0.42

d-propionarnide C43 1.0 7.0 77 47

044 1.0 11.5 43 20 0.56 (}43

N45 1.0 7.7 91 49 1-28 1.04

N45HI 1-0 9.9 - 4 7 - 3 4 -0"61 -ff72

N45H2 1-0 11.9 - 2 6 - 17 -0.34 -0-36

n e u t r o n study is less easy t h a n in the x-ray study, p r o b a b l y t h r o u g h some o v e r l a p p i n g o f positive a n d negative peaks a n d the lower relative scattering lengths o f the oxygen atoms.

6. Conclusions

It is difficult f r o m this analysis n o t to c o n c l u d e that the single carboxyl g r o u p o f this m o n o a c i d is at the b position. This w o u l d seem a priori very likely f r o m the s t r u c t u r e o f c y a n o c o b a l a m i n itself. T h e A ring is geometrically different f r o m the B a n d C rings t h r o u g h the presence w i t h i n it o f the tetrahedrally linked a t o m at C19. T h e small change in the ring g e o m e t r y that results c a n t h r o w the p r o p i o n i c side c h a i n further into the solvent in the extended c o n f o r m a t i o n t h a n is the case with the d d n d e chains, which tend to pack against the benziminazole ring. T h e particular role played by the acid in biosynthesis a n d as a n a n t i m e t a b o l i t e d o n o t at this stage present a p r o b l e m ; they p r e s u m a b l y are related to p a r t i c u l a r e n z y m e i n t e r a c t i o n s involved in the synthesis a n d a c t i o n o f the vitamin, a n d can o n l y be u n d e r s t o o d w h e n the e n z y m e structures are k n o w n .

2 5 6 F H Moore et al

Table 7. Distances o f a t o m s from the least squares plane defined by the terminal atoms o f the side chains, following the high-angle refinement. The degree o f plauarity it given by av~, the ESD o f the distances from the plane.

a-chain b-chain

C26 0-00A

C27 - 0-02

0 2 9 0.01

N28 0"08

N 2 8 H I - 0 " 0 5

N28H2 - 0 - 0 2

(%~ = 0-04A) c-chain

C37 - 0 - 0 1 A

C38 - 0 - 0 3

0 3 9 0"03

N40 - 0.01

N40H1 - i f 0 1

N40H2 0"03

(O'pl

#-chain

C60 - 0.03A

C61 - 0 - 0 4

0 6 2 0.03

N63 0.13

N63H 1 - 0.08

N63H2 - 0 - 0 1

(Ppl = 0-07A)

C31 - f f l l A

C32 - ff22

0 3 3 if00

0 3 4 ff23

O 3 4 H - 0 - 1 8

(OpX = 0-14A) d-chain

C42 0.11A

C43 - 0 - 0 6

0 4 4 - 0-06

N45 - 0-02

N 4 5 H 1 - 0-08

N 4 5 H 2 0-13

(op~ = 0-08A)

i\~ L - ' > / . / W 1 0 9 A

. . . . .. I N 2 8 H 2

"... /_-~\

".. o33 (/i-c- ~

i l I F', ' I I I \ \ \ \ " / - - / / I I H ~ - - ~

-

tT{t:::,/?l

Ltttt )? J ..,

" ~ . . ~---~ ,, o34

9 1 i \(~-/ ] #

Figure 10. Hydrogen bonding to 0 3 3 and 0 3 4 seen in the neutron scattering density.

Part III. Neutron diffraction studies of wet crystals 257

029 ~ ~AN20

0 3 9 ~

/ /

0 3 3

\

NA

,, \',

\ \

_s- i N-

\ /'

O62

/ / /

Figure 11. Projection of the atomic positions on the b plane.

On the other hand, it still is difficult to understand the reactions studied by Professor Bernhauer's group; the effect of the acid on the acidity of the phosphate group in cobalamin and the complex series of hydrolyses carried out through the intervention of cerium(III) hydroxide following permanganate oxidation of the methyl groups at C5 and C15 (Rapp et al 1973). Both these are easily explicable if the acid group in E2, cMsl, is at e. Professor Bernhauer has checked for us that the crystals we studied are indeed identical with his cMsl.

Since effectively our conclusion rests on the non-observation of one hydrogen atom combined with the relative peak heights of the two terminal atoms at the b position in this complicated molecular structure, we must hope for further study of the problem to make assurance sure. Additional evidence might be reached either by the neutron diffraction analysis of deuterated samples of the mono-acid or by x-ray diffraction of derivatives of the related acid, factor VIb, which may crystallise more easily than those of the mono-acid of cobalamin itself.

258 F H Moore et al

o~, / c , s ~cs6

N.~3r ~CS7

CSOA ~ -02 W

0 S I S ~ ~A ~ 1 1 ~ ~ 0 ~ N.~

. . . . . . <

7o .---NFo ?'

c,, 6' -0; ~

/t W ~ x 033 |

4.,, ^{.. . .}

/ ' " c : r ~ o~

coc~ ..t '~

x " ~ N 2 4 C~9

Figmre 12. Density over some disordered water molecules. Heavy line neutron scattering density; fine line, electron density. (Hodgkin et al 1984).

A c k n o w l e d g e m e n t s

We thank Dr F Wagner and Dr E Lester Smith for the large crystals used in this study.

The neutron diffraction measurements were carried out at Harwell; we are grateful to the U.K. Atomic Energy Authority for the facilities and to many of the staff for assistance, particularly Mr T M Valentine. The early calculations were carried out on the Science Research Council Atlas computer, the later ones at the University of Western Australia and at the Australian Institute for Nuclear Science and Engineering.

We are grateful to Dr E N Maslen and Dr J R Cannon in Perth for very useful discussions and support, and to Mr S Kraemar for assistance one summer in Oxford in sorting out water molecule positions. Fellowships were awarded by the Medical Research Council (F.H.M.), the U.K.A.E.A. Research Group (B.H.O'C) and the Queen .Elizabeth II fund (B.H.O'C).

Appendix I. Fourier resolution

The most significant aspect of the present study, in relation to the feasibility of neutron diffraction analysis o f large molecules, is that of Fourier resolution. The limiting effect of series-termination on peak definition and on hydrogen atom identification is of particular interest. The importance of series-termination effects in neutron diffraction structure analysis has been summarised by Bacon (1962). In general terms the Fourier synthesis with neutron diffraction data has more spurious detail than that with x-ray data since the atoms act as thermally-smeared point scatterers for neutrons and as smeared, spatially-extended scatterers for x-rays.

The definition of the amide group terminations at 1.0 A and 1.3 A resolution has already been discussed in the description of the structure analysis. In order to look at

Part III. Neutron diffraction studies of wet crystals 259

Figure 13. Neutron scattering density synthesis in the best plane for atoms C26, C27, 029, H28, N28H 1, N28H2 calculated for d-spacing limits of 1-0A, 1.3, 2-0 and 2.5 A. The first level is i at l-0f/A 3 and the contour interval is at 1.0f/A a.

the question of resolution in a systematic manner it is instructive to examine the Fourier representation of the scattering density in the best plane for the a-chain at 1.0, 1.3, 2.0 and 2.5 A resolution levels. These density maps are shown in figure 13.

At the 1-0 A level (2993 Fourier coefficients) resolution is achieved at half-height along the C - O and C - N bonds, and there is virtually complete resolution along the longer C - C bond. The hydrogen atoms bonded to N28 are easily distinguished from the background of spurious peaks. However, the hydrogen distributions are distorted due obviously to the influence of diffraction ripple from N28. Some of this distortion may be due to thermal anisotropy, although the presence of almost identical features in the corresponding Fc-synthesis shows that anisotropy is a second-order effect. Each positive peak is surrounded by a distribution of spurious negative peaks which are easy to recognise as such since they are much less diffuse than the hydrogen atom distributions. Positive diffraction ripple from the hydrogen atoms is virtually absent due to the relatively weak scattering power of hydrogen together with its relatively high degree of thermal motion.

There is some loss in resolution at the 1.3 A level (1678 coefficients) although the identification o f the group is unambiguous in this particular example. However the diffraction ripple features are more diffuse than at 1-0 A resolution, to the extent that these features can have the general appearance of hydrogen atoms. Clearly some of the features, stlr, h as that immedia4ely above C27 in figure 13, can be rejected on chemical grounds. It is noteworthy that the most prominent spurious peaks are found adjacent to the nitrogen atoms in the corrin nucleus which are the strongest scattering centres in the molecule.

260 F H M o o r e et al

At the 2.0 A level (522 coefficients) it is impossible to chemically assign the atoms, even though some course structure can be recognised. Also it is debatable that hydrogen density can be distinguished from diffraction ripple at this resolution when it is remembered that the hydrogen positions are included in the phase calculation for the given example.

Finally, at the 2.5 A level (269 coefficients) little information is obtained. The negative density has virtually disappeared. Clearly an x-ray diffraction study would be preferable at this resolution level.

References

Arndt U W and Willis B T M 1966 Sinole crystal diffractometry (Cambridge: University Press) Bacon G E 1962 Neutron Diffraction (Oxford: Clarendon Press) Ch. VIII

Moore F H, Willis B T M and Hodgkin D C 1967 Nature (London) 214 130

Nockolds C E, Ramaseshan S, Waters J M, Waters T N M and Hodgkin D C 1967 Nature (London) 214 129 Rapp P, Bozler G and Fridrich E 1973 Hoppe-Seyler's Z. Physiol. Chem. 354 970-3