VOL. 14, 1581-1595 (1975)
BIOPOLYMERS
Refinement of the Structure of @-Chitin K. H. GARDNER* and J. BLACKWELL,** Department of Macromolecular Science, Case Western Reserve University, Cleveland, Ohio 441 06
Synopsis T h e structure of 0-chitin has been refined by rigid-body least-squares methods, based on the intensity data for highly crystalline specimens from the pogonophore Oligobrachia i ~ a n o u i . T h e structure consists of an array of poly-N-acetyl-D-glucosamine chains all having the same sense, which are linked together in sheets by N-H- - -O=C hydrogen bonding of the amide groups. In addition to the 0-3'-H- - - 0 - 5 intramolecular hydrogen bond, analogous to that in cellulose, the CH20H side chain forms an intrasheet hydrogen bond to the carbonyl oxygen on the next chain. This structure shows considerably better agreement between observed and calculated intensities than that possessing an intersheet hydrogen bond, as had been proposed previously. The structure is consistent with the swelling properties of &chitin and can also be seen to be analogous to that of native cellulose.
INTRODUCTION Chitin is a linear polysaccharide consisting of p-( 1+4)-linked residues of 2-deoxy-2-acetamido-D-glucose (N-acetyl-D-glucosamine), and serves in nature as the skeletal material in lower animals. Three naturally occurring polymorphic forms have been recognized, known as a-, @-, and y-chitins.' Detailed crystallographic investigations have been reported for the and /3-f0rms.~-~Basic to the proposed structures for all three systems is the presence of sheets of parallel chains (i.e., having the same sense) linked by N-H-O=C hydrogen bonds through the amide groups. The three forms differ in the sense of the chains in successive sheets. In @-chitinthe sheets are all arranged in a parallel m i n ner, whereas in the a-form successive sheets are antiparallel. For the less well characterized y-form, a repeating structure with two sheets parallel followed by one antiparallel has been proposed.' In the present paper we will report a reinvestigation of the structure of the @-form. This form was first distinguished from the much more common a-chitin by Lotmar and Picken,s who observed a new X-ray pattern for deproteinized pens from the squid, Loligo. Dweltz4 indexed this X-ray pattern in terms of a unit cell containing only one chitin chain, which necessitated a parallel chain structure. * Present address: H. H. Wills Laboratory of Physics, University of Bristol, Bristol BS8 ITL, England. **,To whom all correspondence should be addressed. 1581 0 1975 by John Wiley & Sons, Inc.
1582
GARDNER AND BLACKWELL
Fig. 1. Model for the disaccharide repeat of chitin with atomic numbering and the potentially refinable parameters labeled. [The acetyl amide group has been rotated about the C-2-N bond to facilitate labeling.]
In later work, highly crystalline P-chitin was found in pogonophore tubesg and the spines of certain marine diatoms.lOJ1 Such specimens show a ribbonlike morphology analogous to algal cellulose (e.g., from Valonia), and are especially suitable for crystallographic study. The unit cell is monoclinic, with dimensions a = 4.85 A,b = 9.26 A, c = 10.38 A (fiber axis), and y = 97.50.11 The space group is P21 and the unit cell contains two sugar residues related by the twofold screw axis. The structure was refined by Blackwell,6 using trial and error methods. The fiber repeat is the same as that for cellulose, and thus an 0-3'-H- - - 0 - 5 intramolecular hydrogen bond (see Figure 1 for numbering of the atoms) was incorporated in the chain. The chain was rotated first about its fiber axis and then the CH20H side chain was allowed to rotate about the C5-C6 bond. The final proposed structure had approximately straight N-H-O=C bonds but the best agreement ( R = 0.302) between observed and calculated intensities positioned the CH20H groups so that they were not involved in hydrogen bonds. In separate work, Dweltz et al.7 proposed a structure in which the CHzOH
STRUCTURES OF B-CHITIN
1583
groups formed intersheet hydrogen bonds to the C=O group of the next chain along the a b diagonal of the unit cell. Neither of these structures were acceptable in terms of the infrared data. The polarized spectral show two 0-H stretching bands with parallel dichroism; there was no evidence for unbonded 0-H groups, or for the perpendicular dichroism predicted for the structure with an intersheet hydrogen bond. A t this point it could be concluded that the X-ray data were inadequate to solve the structure by trial and error methods. A similar situation prevailed until recently for the structure of cellulose I. However, we have recently ~ h o w n that ~ ~ the J ~ rigid-body leastsquares refinement techniques14 could be used to distinguish between the various possible structures, which has led to determination of the relative chain polarities and elucidation of the hydrogen bonding network. We have now applied the same techniques to the @-chitindata, and the results are described below.
RESULTS Intensity Data The reinvestigation of the structure of @-chitinwas based on the unit cell and intensity data for specimens from the pogonophore Oligobrachiu iuanoui, as described by Blackwell.6 The only systematic absences were for the odd-order 001 reflections. The reported intensity data consist of 61 nonmeridional reflections, indexed by 105 potentially reflecting planes. The 61 reflections included 5 which were “unobserved,” and for these the intensity was set at zero (they comprise only a small fraction (-8%) of the total data). The observed data in Ref. 6 had been corrected for Lorentz and polarization effects and were listed as F2. Scaled derived observed structure factors, F(obs), and the respective Miller indexes are shown in Table I.
Atomic Coordinates for the Isolated Chitin Chain The model for the chitin chain was based on that used in our refinement of c e l l ~ l o s e . Averaged ~~ D-glucose residues15 are arranged with @-(1-+4)-glycosidiclinkages on a 21 screw axis, repeating in 10.38 A. This chain is the same as that used for the refinement of cellulose, except that the 0-2-H group is replaced by the trans planar NHCOCH3 group of N - acetyl-a-D-glucosamine.l6 The disaccharide residue has a glycosidic bond angle of 114.8’ and glycosidic torsion angles of # = 23.0’ and $ = 23.7’ (using the convention followed by Sundararajan and Rao17). An 0-3’-H- - -0-5 intramolecular hydrogen bond is present, with length 2.75 8. In this conformation, the chitin chain is completely rigid, except for the following possible rotations of the side chains, which are also shown in Figure 1:
GARDNER AND BLACKWELL
1584
TABLE I Observed and Calculated Structure Factor Amplitudes for 8-Chitin models: (A), Nonbonded Model; (B), Proposed Structure for p-Chitin Incorporating Intrasheet 0-6-H---0-7; Hydrogen Bonds; (C), Model Containing 0-6--H--0-7'd Intersheet Hydrogen Bonds Miller indexes of potentially reflecting planes
Observed structure amptitudesa
010 100 020 iio 110 120 120 030
J
z}
A
B (proposed model)
C
41.2 70.7
35.2 81.0
37.6 78.9
27.1 66.0
92.2
96.8
97.9
119.6
41.2 31.6 24.5 20.0
36.9 21.8 26.9 22.0
32.8 22.6 21.4 22.6
25.8 4.4 6.1 16.5
50.0
51.5
49.3
53.3
23.4
21.1
21.0
27.3
24.5 33.2 10.0 14.1 20.0 31.6 31.6 24.5 14.1 10.0
23.4 31.9 21.0 11.9 18.9 27.2 23.1 14.0 11.0 7.6
28.4 30.6 23.0 13.4 18.2 26.6 25.3 12.2 9.5 4.5
17.5 26.4 15.2 11.3 8.1 14.2 21.0 21.0 3.8 17.0
22.4
21.4
20.3
21.7
33.2
24.4
37.4
28.3
0.0
17.2
15.6
20.5
20.0
22.3
22.1
18.8
26.5 33.2 22.4 10.0 10.0 22.4
20.7 28.2 30.2 16.6 15.6 7.8
18.5 32.2 25.9 16.4 15.0 10.8
24.1 15.8 21.3 17.6 21.2 19.8
14.1
17.8
23.4
11.1
10.0
11.1
9.2
9.0
10.0
17.9
17.9
19.8
210
22 0 140 01 1 101 021
iii 111
121 121 03 1
131 131
\
210 211
E} 22 1 121
\
23 2211 012 102 022 112 112 122 032 122 132 202 132
I 1
STRUCTURES OF @-CHITIN
1585
TABLE I (Continued) Miller indexes of potentially reflecting planes
Observed structure amplitude+
04 2 22 2 142 013 103 032 ii3 113 123
1
033 183 133 213 3431
223 143 014
ii4 114 124 034 184
B A
(proposed model)
C
22.4
24.6
21.0
29.5
17.3
17.9
17.7
14.6
10.0 64.0 22.3
20.3 43.8 18.7
18.1 40.9 19.5
12.9 39.8 30.9
43.6
44.2
35.9
43.8
8.7 13.4
14.2 14.3
0.0 0.0
1.1 8.6
10.0
25.4
33.1
19.5
30.0
8.0
9.3
18.5
24.5
32.1
26.8
27.7
17.3
18.0
20.6
15.4
24.5
14.6
17.4
9.2
26.5
20.2
21.2
16.8
26.5 0.0
23.2 4.3
23.6 5.6
18.7 2.9
20.0
19.5
23.1
11.7
17.3
13.0
7.7
13.2
14.1
19.5
17.7
24.0
20.0
16.5
15.9
14.9
20.0
19.3
18.2
20.3
26.5
12.8
9.2
17.2
24.5
17.4
19.6
24.8
22.4
22.3
20.7
16.7
14.1
15.0
16.6
17.5
2i4
224 144
144 015
ii5 125
135
GARDNER AND BLACKWELL
1586
TABLE I (Continued) Miller indexes of potentially reflecting planes
Observed structure amplitudesa
A
B (proposed model)
C
17.3
17.7
18.0
16.1
225 016
17.3
6.3
8.4
10.1
ii6
24.5
13.6
17.1
13.0
04 5
”‘} 116
aBased o n intensities in Ref. 6.
(i) x, rotation of 0-6 about the C-5-C-6 bond. x = 0’ when C-6-0-6 is cis to C-4-C-5, and increases as 0 - 6 is rotated anticlockwise when viewed along C-5-C-6. (ii) x’,rotation of the entire amide side chain about the C-2-N bond. x‘ = 0 when N-C-7 is trans to C-1-C-2 and increases as C-7 is rotated anticlockwise when viewed along C-2-N. In the initial model, x’ = 120°, such that N-C-7 was 180’ away from the gauche-gauche conformation relative to C-1-C-2 and C-2-C-3, i.e., the hypothetical C-2-H and N-H bonds would be trans. (iii) x”, rotation of the COCH3 group about the N-C-7 bond. X” = 0 when C-7-0-7 is cis to N-C-2, and increases as 0 - 7 is rotated anticlockwise when viewed along N-C-7. A nonzero value for X” represents a deviation from planarity in the amide group.
Packing Models The single polymer chain was arranged in the unit cell with its molecular axis through (O,O,O) and parallel to c. Translation along the c axis is arbitrary and was chosen such that the glycosidic oxygen 0-1had z = 0. The only packing parameter remaining is a, the orientation of the chain about the helix axis. No useful convention exists to define 6, = 0, and an origin was assigned arbitrarily. Increase in corresponds to anticlockwise rotation of the chain about the c axis. Since the chitin chain is polar and y # 90°, there are two possible parallel chain structures for P-chitin: one with the chain oriented “up” in the unit cell (i.e., 20-5 > Z C . ~ ) ,and the other with the chain oriented “down” in the unit cell (i.e., 20.5 < zc-5). Calculation of R values for otherwise similar “up” and “down” models easily demonstrated that the
STRUCTURES OF @-CHITIN
1587
structure must contain “up” chains, as was the case for c e l l ~ l o s e . ~ ~ Furthermore all “down” structures that could form reasonable N-H-. O=C hydrogen bonds were stereochemically unacceptable due to short contacts between the sheets. All calculations described hereafter refer to refinement of models containing “up” chains.
Refinement Method The possible structural models for 0-chitin were refined using the “linked-atoms’’ least-squares method of Arnott and W ~ n a c o t t . ’ ~By adjusting the refinable parameters, using a least-squares process, the function minimized is 0, where
A(Fm2) is the difference between the squares.of the observed (F,(obs)) and calculated (F,(calc)) structure factor amplitudes for the rnth of M reflections. The values of F , (calc) incorporated scale ( K ) and isotropic temperature (B) factors derived in the least-squares procedure. w, is the weight to be applied to the mth observation. In the course of this investigation, a Cruickshank weighting schemels was used, with constants determined by least-squares. In cases where more than one hkl plane contributed to an observed reflection, the calculated structure amplitude used for comparison is given by
where the F,. (calc) value is the calculated structure amplitude for the r t h of R planes contributing to the observed reflection. In the minimization of 8, the explicit variables are the molecular torsion angles x,X I , x N ,packing parameter and K and B. In order to examine potential structural models containing specific hydrogen bonds, constraints were incorporated, which limited certain interatomic distances to ranges within specified maximum and minimum values. These constraints G, were applied with respect to the Lagrange multipliers A, in the equation for 8. In order to consider alternative structural models, the R values were compared using the statistical tests due to Hami1t0n.l~ For each model, R” is computed where
+,
The more common crystallographic residual
is also quoted but is not used in the quantitative comparisons.
GARDNER A N D BLACKWELL
1588
Refinement The structure was refined initially by varying five parameters: two molecular parameters, x and x'; one packing parameter, @; and two crystallographic parameters, the isotropic temperature factor, B, and the overall isotropic scale factor, K . These parameters were refined simultaneously to give the best least-squares fit between the observed and calculated structure amplitudes. The preliminary model produced had R = 0.244 and R" = 0.274. Inspection of this model, hereafter referred to as the unconstrained model, showed that the value of x(= -81.7') oriented the CHzOH group so as to produce a marginally short contact: 0-5-0-6 = 2.61 A. To remove this deficiency, a constraint was incorporated such that 0-5-.0-6 1 2.80 A. Refinement of this constrained model led to residuals of R = 0.250 and R" = 0.288. According to the statistical tests of Hamilton,lg for least-squares refinement of 5 parameters against 61 observations, this difference in R" is significant only a t the 50% level, i.e., either model could be chosen if selected only on the basis of R values. Thus in terms of crystallographic and stereochemical criteria, the model with the 0-5-0-6 constraint is selected in preference to the unconstrained alternative. A defect of this model is that the 0-6-H group is not involved in donor hydrogen bonding. For comparison with later models, this structure will be referred to as the nonbonded model. The calculated structure amplitudes for the nonbonded model are compared with the observed data in Table I. The final values of the refined parameters, their estimated standard deviations, and the R values are given Table 11. @, which defines the orientation of the chains about the c axis, has a value of 110.lo,with an estimated standard deviation of ua = 1.3', indicating that the X-ray data defines the value of this parameter with a high degree of accuracy. (The actual value of has no significance, except for comparison with later refinements). The refined values of x and x' are -60.7' and 107.4', respectively; the estimated standard deviations associated with these parameters are o x = 9.7' and ox = 4.0'. Since a small change in x' moves a large group of atoms (the planar TABLE €1 Refined Crvstal Parameters for p-Chitin Models Parameters
Nonbonded model
X
-60.7" (9.7")s 107.4" (4.0") 110.1" (1.3") 85.0 (6.7") 8.2 (1.8")
Intrasheet hydrogen bond
Intersheet hydrogen bond ~~
X
0
K B ~~~
~~~~~~
~
(4.9") (3.6") (1.3") (7.2") (1.9")
142.6" 116.3" 108.9" 99.4 6.7
(8.8") (3.3") (1.6") (9.9") 12.4")
~
Special constraints on refining model
R R'
-54.8" 103.0" 110.4" 95.7 7.1
0-5
.
. . 0-6 > 2.80 A
0.250 0.288
0 - 5 . . . 0 - 6 > 2.80 A
0 - 5 . . . . 0 - 6 > 2.80 A
2.65 A G 0-6; ---0-7 G 2.90 A
2.65 A
0.267 0.,302
0-6---0-7;1 0.342 0.369
aThe numbers in parenthesis are the estimated standard deviations.
2.90
a
STRUCTURES OF &CHITIN
1589
amide group) it is not unreasonable that the data is more sensitive to x’ than to x and that cX’is smaller than cx. The value of x‘ places the planar amide group 12.6’ away from its initial trans orientation. The refined value of x(= -60.7’) is very close to the gg position (xgg = -6OO). x positions the -CH20H group approximately 180’ away from the position that would be necessary to form an intersheet hydrogen bond of the type incorporated in the Dweltz model.7 The nonbonded model contains a N-H- - -0 -7 distance of 2.70 between residues along the a axis. The only other intermolecular contact of less than 3.20 is a 3.14 separation distance between the 0 - 7 oxygen and 0-6’,. (To facilitate the description of alternative intermolecular hydrogen bonding possibilities in the @-chitin structure, the following nomenclature is used: numbered atoms (unprimed, unsubscripted) comprise the first residue of the chain (coordinates (x,y,z)); atoms in the second residue of the same chain (-x,-y,z $) are primed (see Figure 1); atoms in the molecule related to the origin molecule by a translation of -a (i.e. (x - l,y,z) and (-x-1,-y,z $)) are denoted by the subscript a (e.g., 0-7, and 0-7’,); and finally, atoms in the molecule related to the origin molecule by translations of +a and + b (i.e. (x l,y 1,z) and (-x 1,-y 1, z K ) are indicated by the subscripted (e.g., 0-7d and 0-7’d). Further refinement incorporated the possibility of nonplanarity of the amide group. The resulting structure had x‘’ = -6.3O and R” = 0.285. This improvement in R” (as compared to R” = 0.288 for the model containing planar amide groups) is not statistically significant and thus inclusion of nonplanarity of the amide group was accordingly not justified. The nonbonded model gives good agreement between the observed and calculated structure amplitudes, but the CH2OH groups are not hydrogen bonded. Two possibilities for hydrogen bonding present themselves: (i) an intrasheet hydrogen bond to the carbonyl oxygen in the next chain along the a axis: 0-6’, -H- - -0-7; and (ii) an intersheet hydrogen bond to the carbonyl oxygen in the chain along the a b diagonal: 0-6-H- - -0-7’d. In view of the stereochemistry of the chain and the unit cell dimensions, these are the only two hydrogen bonds which are possible for the CHpOH groups in this structure. Since from infrared evidence the CH20H groups must be hydrogen bonded, the model was constrained further to form first the intra- and then the intersheet bond, and the five parameters were refined in each case. The results for these two refinements are discussed in turn.
+
+
+
+
+
Intrasheet 0-6‘, -H-
+
+
- - 0 - 7 Hydrogen Bond
Refinement of the structure of @-chitinwas performed for the model constrained such that the 0-6’,-0-7 distance was in the range 2.65-2.90 A. Projections of the resulting structure are shown in Figure 2. The
GARONER AND BLACKWELL
1590
0
b
Fig. 2. Projections of the proposed model for p-chitin. The structure contains N-H -O=C and 0-6',-H- - - 0 - 7 intrasheet hydrogen bonds.
fractional coordinates of the model are found in Table 111 and the refined parameters and their associated standard deviations are given in Table 11. T h e residuals for this structure are R = 0.267 and R" = 0.302. T h e calculated structure factor amplitudes are given in Table I. @ refined to a value of = 110.4' (a+ = 1.2O); this is 0.3" (0.25aa) away from the value found in the nonbonded model. x and x' refined t o values of -54.8' and 103.0°, respectively. The new x value represents a rotation of 5.2" away from the gg position. T h e lower estimated standard deviation of this parameter (a, = 4.9') is due t o the presence of the 0-7--0-6', nonbonded constraint which severely limits the allowed value of x. T h e value of x' (-103.0') indicates that in the constrained model the planar amide group is rotated an additional 4.4' away from the initial trans position of the group. T h e reorientation is presumably to allow formation of the desired 0-6', -H- - -0-7 hydrogen bond. This reorientation (and, to a lesser degree, the small rotation of the chain) has increased the N-H- - -0-7 hydrogen bond length t o 2.76 A. Through the additional constraint, the structure contains a n 0-6', -H- - -0-7 intrasheet hydrogen bond of length 2.89 A. T h e structure contains no other intermolecular contacts of less than 3.2 A. T h a t w o hydrogen bonds are indicated in Figure 2.
STRUCTURES O F p-CHITIN
1591
TABLE I11 Fractional Coordinates for the N-Acetyl-D -glucosamine Residue in the Refined Structure of p-Chitin which contained the Zntrasheet 0-69-H--0 - 7 Hydrogen Bonda Atom
c-1 c-2 c-3 c-4
c-5
C-6 c-7 C-8 0-1 0-3 0-5 0-6 0-7 N
x/a
Y/b
z/c
- .010 -0.146 -0.007 -0.002 0.124 0.114 -0.366 -0.321 -0.161 -0.155 -0.026 -0.164 -0.589 -0.131
-0.038 -0.150 -0.130 0.027 0.132 0.290 -0.377 -0.527 -0.051 -0.227 0.105 0.315 -0.330 -0.299
0.386 0.291 0.159 0.114 0.218 0.183 0.392 0.438 0.500 0.068 0.337 0.153 0.401 0.399
aThis is the final proposed structure for p-Chitin.
Intersheet 0 - 6 -H-
- -0-7'd
Hydrogen Bond
The &chitin structure was refined with the model containing an additional constraint such-that the 0-6-0-7'd distance was in the range 2.65-2.90 A. The refined model containing this intersheet hydrogen bond is shown in Figure 3. The refined parameters and associated estimated standard deviations are given in Table 11. The residuals for this model are R = 0.342 and R" = 0.369. Comparison of observed and calculated structure factor amplitudes is made in Table I. @J, the packing parameter, refined to a value of = 108.9 with an estimated standard deviation of u = 1.6O. @J is approximately 0 . 7 5 ~from the value found in the nononded model. x and x' have values of 142.6" and 116.3", respectively, with estimated standard deviations of ux = 8.8" and ux' = 3.3". The CHzOH groups are oriented 37.4" away from the gt position; the 0-6-H- - -0-7', bond length is 2.84 A. The amide group is rotated 9.3O from the position found in the unconstrained model. The 0-7- - -N, intrasheet hydrogen bond has length 2.68 A. Other nonbonded contacts of less than 3.2 A involve the amide group to which the 0 - 6 is hydrogen bonded: 0 - 6 4 - 7 d is 2.81 A and 0-6- *C-8d is 2.67 A in length. Inclusion of a single constraint which removed both short contacts resulted in a small increase in R". Figure 4 shows a plot of R" against x for the unconstrained model where x was varied in increments of 5", with all the parameters held constant. Regions where unacceptable chemical contacts occur are also shown. The R" curve shows two minima, the lowest corresponding to the unconstrained model a t x = 258.3", and the second a t x N 125" and
GARDNER AND BLACKWELL
1592
0
m
0
b
Fig. 3. Projections of the rejected model which contained an 0-6-H- - - 0 - 7 ' d intersheet hydrogen bond. The structure also contains the 0-7-N, intrasheet hydrogen bond.
higher R". The x values for the refined unconstrained, nonbonded, intrasheet hydrogen bond and intersheet hydrogen bond models are indicated by A, B, C, and D, respectively. It is interesting that the two minima fall in the only two regions which are stereochemically allowed. The subsidiary minimum was not sufficiently deep to prevent refinement to the true minimum, i.e., removal of the hydrogen-bonding constraint from model D led to refinement to model B (or to model A if the 0-5-0-6 constraint was relaxed).
DISCUSSION In summary, three models for P-chitin have been considered. The structure with the lowest R values, R = 0.250, R" = 0.288, is stereochemically acceptable but does not provide for hydrogen bonding involving the -CHzOH group. A second structure, containing an 0-6', -H- - - 0 - 7 hydrogen bond, was found to be stereochemically acceptable and has R values of R = 0.267 and Rf' = 0.302. A third structure, containing an intersheet 0-6-H- - -0-7'd hydrogen bond (as has been previously proposed), was found to have a short 0-6-C-8d contact and has R values of R = 0.342 and R" = 0.369.
STRUCTURES OF @-CHITIN
1593
c
0
60
i0
X
Fig. 4. A plot of R” as a function of x, which defines the rotational conformation of the -CH*OH group about the C-5-C-6 bond. The angular ranges for x where there are intra- and intermolecular short contacts are indicated by the boxes. The numbers in the boxes refer to the atom in contact with 0-6. Open boxes indicate contacts in the range 2.6-2.7 A. x values for the refined models considered in this paper are indicated by the vertical lines: (A) unconstrained model with the lowest R and R”; (B) nonbonded model with no bad contacts but with nonhydrogen bonded -CH20H group; (C) intrasheet hydrogen bond model constrained to form the 0-6’,-H- - -0-7 bond; and (D) intersheet hydrogen bond model constrained to form 0-6-H- - -0-7‘d bond.
By all criteria, the model containing the 0-6-H- - -0-7’d intersheet hydrogen bonding can be rejected. Firstly, the structure contains an unacceptable short nonbonded interaction and, secondly, application of Hamilton’s test for least-squares refinement of 5 parameters against 61 observations shows that there is much less than 1 chance in 200 that the structure containing the intersheet hydrogen bond is as good a representation of the true structure as either the nonbonded model or the structure containing the intrasheet hydrogen bond. The choice between the remaining two models is not as straightforward as the rejection of the intersheet hydrogen bond model. Neither of the remaining models contains unacceptable intermolecular nonbonded contacts, and although the nonbonded model has a slightly lower R” value (0.288 as compared to 0.302), application of Hamilton’s
1594
GARDNER A N D BLACKWELL
test shows t h a t this difference is not significant and there is a 50-50 chance t h a t either structure is correct. However, the basic difference between the two models is t h a t one allows for hydrogen bonding involving the -CH20H group while the other does not, and on this basis the nonbonded model is rejected and the structure containing the 0-6‘, H- - -0-7 intrasheet hydrogen bond is proposed as the structure for 0chitin. For trial and error refinement of &chitin, BlackwelF reported poor agreements for four individual reflections: i.e., the 130, 120, 121, and 102 planes. He found that the intensities for these four sets of planes were very sensitive to the position of the -CH20H group, but t h a t variation of x did not improve the overall agreement. In the present investigation the same discrepancies are observed in the model containing the intersheet hydrogen bond, but refinement of x produced models (nonbonded and intrasheet hydrogen bond) that give good agreements for these reflections combined with superior overall agreement. In the proposed intrasheet hydrogen bond structure the 0-3-H and 0-6-H bonds are oriented such t h a t both should give rise to parallel dichroism in the 0 - H stretching region of the infrared spectrum. In contrast, the 0-6-H- - -0-7’d intersheet hydrogen bond should give rise to perpendicular dichroism. @-Chitinis easily swollen in water: two crystalline hydrate structures have been recognized6 and suggested t o contain one and two water molecules per residue. In the hydrated structures the hydrogen-bonded sheets remain intact but are moved apart t o include the water molecules. T h e intrasheet hydrogen bond model is consistent with this behavior whereas a n intersheet bond would tend to hold the sheets together and stabilize the anhydrous form. The intrasheet 0-6’,- - -0-7 bond can be seen as an additional stabilizing factor for the sheets of chitin chains, also linked by the amide N-H- - -O=C bonds. I t is possible t o see numerous points of analogy between the structures of 0-chitin and cellulose I. Both structures contain extended chains with the same (parallel) sense. Depending on the source, specimens show different degrees of crystallinity, but the large crystalline fibrils are seen as arrays of smaller “elementary fibrils” in electron microscope structures of deformed samples.20,21 A t the molecular level, both structures are seen as a n array of hydrogen-bonded sheets. In cellulose I, the 0-2-H group forms an intramolecular 0-2’-H- - -0-6 bond, and since there is also an 0-3-H- - -0-5’ intramolecular bond, the ribbonlike chain has approximately parallel hydrogen bonds on both sides of the glycosidic bridge. In 0 chitin, the same 0-3-H-0-5’ intramolecular bond is present, but the N-H is involved in bonding with the amide group on the next chain. Bonding of the type 0-6-H- - -N’ would not be expected: such acceptor bonds are apparently not formed by amide groups and in any case, this bond is stereochemically unacceptable. However, rotation of the CH20H group from the t g position of cellulose I t o the gg position in P-chitin allows formation of the 0-6’,-H- -0-7
STRUCTURES OF @-CHITIN
1595
bond giving a similar chain structure with parallel hydrogen bonds on both sides of the chain. On conversion of cellulose I to cellulose 11, the 0-2’-H-0-6 intramolecular bonds are broken and 0-2-H- - - 0 - 2 d bonds are formed between antiparallel chains.22 This is analogous to the conversion of P-chitin to the a-form and it will be interesting to see if a-chitin contains additional intersheet hydrogen bonds. This work was supported by N.S.F. Grant No. GH 34227 and N.I.H. Research Career Development Award No. AM 70642 (to J.B.)
References 1. Rudall, K. M. (1963) Adu. Insect Physiol. 1,257-313. 2. Carlstrom, D. (1957) J . Biophys. Biochem. Cytol. 3,669-683. 3. Dweltz, N. E. (1960) Biochem. Biophys. Acta 44,416-435. 4. Ramakrishnan, C. & Prasad, N. (1972) Biochem. Biophys. Acta 261,123-135. 5. Dweltz, N. E. (1961) Biochem. Biophys. Acta 51,283-289. 6. Blackwell, J . (1969) Biopolymers 7,281-298. 7. Dweltz, N. E., Colvin, J . R. & McInnes, A. G. (1968) Can. J . Chem., 46, 1513-1521. 8. Lotmar, W. & Picken, L. E. R. (1950) Experientia 6,58-59. 9. Blackwell, J., Parker, K. D. & Rudall, K. M. (1965) J . Mar. B i d . Assn. U.K. 45, 659-661. 10. Falk, M., Smith, D. S., McClachlan, J . & McInnes, A. G. (1966) Can. J . Chem. 44, 2269-2286. 11. Blackwell, J., Parker, K. D. & Rudall, K. M. (1967) J . Mol. Biol. 28,282-385. 12. Gardner, K. H. & Blackwell, J . (1974) Biochem. Biophys. Acta 342,232-237. 13. Gardner, K. H. & Blackwell, J . (1974) Biopolymers 13,1975-2001. 14. Arnott, S. & Wonacott, A. J . (1966) Polymer 7,157-166. 15. Arnott, S. & Scott, W. F. (1972) J . Chem. SOC.Perk. Trans. 11,324-335. 16. Johnson, L. N. (1966) Acta Cryst., 21,885-891. 17. Sundararajan, P. R. & Rao, V. S. R. (1969) Biopolymers 8,305-312. 18. Cruikshank, D. W. J . & Pilling, D. E. (1961) Computing Methods and the Phase Problem i n X - R a y Crystal Analysis, p. 46, Oxford, Pergamon Press. 19. Hamilton, W. C. (1965) Acta Cryst. 18,502-510. 20. Gardner, K. H. & Blackwell, J. (1971) J . Ultrastruct. Res. 36,725-731. 21. Gardner, K. H. & Blackwell, J. (1971) J . Polym. Sci. C36,327-340. 22. Kolpak, F. J., & Blackwell, J. (1975) Submitted to Macromolecules.
Received September 27,1974 Accepted March 12,1975
BIOPOLYMERS
Refinement of the Structure of @-Chitin K. H. GARDNER* and J. BLACKWELL,** Department of Macromolecular Science, Case Western Reserve University, Cleveland, Ohio 441 06
Synopsis T h e structure of 0-chitin has been refined by rigid-body least-squares methods, based on the intensity data for highly crystalline specimens from the pogonophore Oligobrachia i ~ a n o u i . T h e structure consists of an array of poly-N-acetyl-D-glucosamine chains all having the same sense, which are linked together in sheets by N-H- - -O=C hydrogen bonding of the amide groups. In addition to the 0-3'-H- - - 0 - 5 intramolecular hydrogen bond, analogous to that in cellulose, the CH20H side chain forms an intrasheet hydrogen bond to the carbonyl oxygen on the next chain. This structure shows considerably better agreement between observed and calculated intensities than that possessing an intersheet hydrogen bond, as had been proposed previously. The structure is consistent with the swelling properties of &chitin and can also be seen to be analogous to that of native cellulose.
INTRODUCTION Chitin is a linear polysaccharide consisting of p-( 1+4)-linked residues of 2-deoxy-2-acetamido-D-glucose (N-acetyl-D-glucosamine), and serves in nature as the skeletal material in lower animals. Three naturally occurring polymorphic forms have been recognized, known as a-, @-, and y-chitins.' Detailed crystallographic investigations have been reported for the and /3-f0rms.~-~Basic to the proposed structures for all three systems is the presence of sheets of parallel chains (i.e., having the same sense) linked by N-H-O=C hydrogen bonds through the amide groups. The three forms differ in the sense of the chains in successive sheets. In @-chitinthe sheets are all arranged in a parallel m i n ner, whereas in the a-form successive sheets are antiparallel. For the less well characterized y-form, a repeating structure with two sheets parallel followed by one antiparallel has been proposed.' In the present paper we will report a reinvestigation of the structure of the @-form. This form was first distinguished from the much more common a-chitin by Lotmar and Picken,s who observed a new X-ray pattern for deproteinized pens from the squid, Loligo. Dweltz4 indexed this X-ray pattern in terms of a unit cell containing only one chitin chain, which necessitated a parallel chain structure. * Present address: H. H. Wills Laboratory of Physics, University of Bristol, Bristol BS8 ITL, England. **,To whom all correspondence should be addressed. 1581 0 1975 by John Wiley & Sons, Inc.
1582
GARDNER AND BLACKWELL
Fig. 1. Model for the disaccharide repeat of chitin with atomic numbering and the potentially refinable parameters labeled. [The acetyl amide group has been rotated about the C-2-N bond to facilitate labeling.]
In later work, highly crystalline P-chitin was found in pogonophore tubesg and the spines of certain marine diatoms.lOJ1 Such specimens show a ribbonlike morphology analogous to algal cellulose (e.g., from Valonia), and are especially suitable for crystallographic study. The unit cell is monoclinic, with dimensions a = 4.85 A,b = 9.26 A, c = 10.38 A (fiber axis), and y = 97.50.11 The space group is P21 and the unit cell contains two sugar residues related by the twofold screw axis. The structure was refined by Blackwell,6 using trial and error methods. The fiber repeat is the same as that for cellulose, and thus an 0-3'-H- - - 0 - 5 intramolecular hydrogen bond (see Figure 1 for numbering of the atoms) was incorporated in the chain. The chain was rotated first about its fiber axis and then the CH20H side chain was allowed to rotate about the C5-C6 bond. The final proposed structure had approximately straight N-H-O=C bonds but the best agreement ( R = 0.302) between observed and calculated intensities positioned the CH20H groups so that they were not involved in hydrogen bonds. In separate work, Dweltz et al.7 proposed a structure in which the CHzOH
STRUCTURES OF B-CHITIN
1583
groups formed intersheet hydrogen bonds to the C=O group of the next chain along the a b diagonal of the unit cell. Neither of these structures were acceptable in terms of the infrared data. The polarized spectral show two 0-H stretching bands with parallel dichroism; there was no evidence for unbonded 0-H groups, or for the perpendicular dichroism predicted for the structure with an intersheet hydrogen bond. A t this point it could be concluded that the X-ray data were inadequate to solve the structure by trial and error methods. A similar situation prevailed until recently for the structure of cellulose I. However, we have recently ~ h o w n that ~ ~ the J ~ rigid-body leastsquares refinement techniques14 could be used to distinguish between the various possible structures, which has led to determination of the relative chain polarities and elucidation of the hydrogen bonding network. We have now applied the same techniques to the @-chitindata, and the results are described below.
RESULTS Intensity Data The reinvestigation of the structure of @-chitinwas based on the unit cell and intensity data for specimens from the pogonophore Oligobrachiu iuanoui, as described by Blackwell.6 The only systematic absences were for the odd-order 001 reflections. The reported intensity data consist of 61 nonmeridional reflections, indexed by 105 potentially reflecting planes. The 61 reflections included 5 which were “unobserved,” and for these the intensity was set at zero (they comprise only a small fraction (-8%) of the total data). The observed data in Ref. 6 had been corrected for Lorentz and polarization effects and were listed as F2. Scaled derived observed structure factors, F(obs), and the respective Miller indexes are shown in Table I.
Atomic Coordinates for the Isolated Chitin Chain The model for the chitin chain was based on that used in our refinement of c e l l ~ l o s e . Averaged ~~ D-glucose residues15 are arranged with @-(1-+4)-glycosidiclinkages on a 21 screw axis, repeating in 10.38 A. This chain is the same as that used for the refinement of cellulose, except that the 0-2-H group is replaced by the trans planar NHCOCH3 group of N - acetyl-a-D-glucosamine.l6 The disaccharide residue has a glycosidic bond angle of 114.8’ and glycosidic torsion angles of # = 23.0’ and $ = 23.7’ (using the convention followed by Sundararajan and Rao17). An 0-3’-H- - -0-5 intramolecular hydrogen bond is present, with length 2.75 8. In this conformation, the chitin chain is completely rigid, except for the following possible rotations of the side chains, which are also shown in Figure 1:
GARDNER AND BLACKWELL
1584
TABLE I Observed and Calculated Structure Factor Amplitudes for 8-Chitin models: (A), Nonbonded Model; (B), Proposed Structure for p-Chitin Incorporating Intrasheet 0-6-H---0-7; Hydrogen Bonds; (C), Model Containing 0-6--H--0-7'd Intersheet Hydrogen Bonds Miller indexes of potentially reflecting planes
Observed structure amptitudesa
010 100 020 iio 110 120 120 030
J
z}
A
B (proposed model)
C
41.2 70.7
35.2 81.0
37.6 78.9
27.1 66.0
92.2
96.8
97.9
119.6
41.2 31.6 24.5 20.0
36.9 21.8 26.9 22.0
32.8 22.6 21.4 22.6
25.8 4.4 6.1 16.5
50.0
51.5
49.3
53.3
23.4
21.1
21.0
27.3
24.5 33.2 10.0 14.1 20.0 31.6 31.6 24.5 14.1 10.0
23.4 31.9 21.0 11.9 18.9 27.2 23.1 14.0 11.0 7.6
28.4 30.6 23.0 13.4 18.2 26.6 25.3 12.2 9.5 4.5
17.5 26.4 15.2 11.3 8.1 14.2 21.0 21.0 3.8 17.0
22.4
21.4
20.3
21.7
33.2
24.4
37.4
28.3
0.0
17.2
15.6
20.5
20.0
22.3
22.1
18.8
26.5 33.2 22.4 10.0 10.0 22.4
20.7 28.2 30.2 16.6 15.6 7.8
18.5 32.2 25.9 16.4 15.0 10.8
24.1 15.8 21.3 17.6 21.2 19.8
14.1
17.8
23.4
11.1
10.0
11.1
9.2
9.0
10.0
17.9
17.9
19.8
210
22 0 140 01 1 101 021
iii 111
121 121 03 1
131 131
\
210 211
E} 22 1 121
\
23 2211 012 102 022 112 112 122 032 122 132 202 132
I 1
STRUCTURES OF @-CHITIN
1585
TABLE I (Continued) Miller indexes of potentially reflecting planes
Observed structure amplitude+
04 2 22 2 142 013 103 032 ii3 113 123
1
033 183 133 213 3431
223 143 014
ii4 114 124 034 184
B A
(proposed model)
C
22.4
24.6
21.0
29.5
17.3
17.9
17.7
14.6
10.0 64.0 22.3
20.3 43.8 18.7
18.1 40.9 19.5
12.9 39.8 30.9
43.6
44.2
35.9
43.8
8.7 13.4
14.2 14.3
0.0 0.0
1.1 8.6
10.0
25.4
33.1
19.5
30.0
8.0
9.3
18.5
24.5
32.1
26.8
27.7
17.3
18.0
20.6
15.4
24.5
14.6
17.4
9.2
26.5
20.2
21.2
16.8
26.5 0.0
23.2 4.3
23.6 5.6
18.7 2.9
20.0
19.5
23.1
11.7
17.3
13.0
7.7
13.2
14.1
19.5
17.7
24.0
20.0
16.5
15.9
14.9
20.0
19.3
18.2
20.3
26.5
12.8
9.2
17.2
24.5
17.4
19.6
24.8
22.4
22.3
20.7
16.7
14.1
15.0
16.6
17.5
2i4
224 144
144 015
ii5 125
135
GARDNER AND BLACKWELL
1586
TABLE I (Continued) Miller indexes of potentially reflecting planes
Observed structure amplitudesa
A
B (proposed model)
C
17.3
17.7
18.0
16.1
225 016
17.3
6.3
8.4
10.1
ii6
24.5
13.6
17.1
13.0
04 5
”‘} 116
aBased o n intensities in Ref. 6.
(i) x, rotation of 0-6 about the C-5-C-6 bond. x = 0’ when C-6-0-6 is cis to C-4-C-5, and increases as 0 - 6 is rotated anticlockwise when viewed along C-5-C-6. (ii) x’,rotation of the entire amide side chain about the C-2-N bond. x‘ = 0 when N-C-7 is trans to C-1-C-2 and increases as C-7 is rotated anticlockwise when viewed along C-2-N. In the initial model, x’ = 120°, such that N-C-7 was 180’ away from the gauche-gauche conformation relative to C-1-C-2 and C-2-C-3, i.e., the hypothetical C-2-H and N-H bonds would be trans. (iii) x”, rotation of the COCH3 group about the N-C-7 bond. X” = 0 when C-7-0-7 is cis to N-C-2, and increases as 0 - 7 is rotated anticlockwise when viewed along N-C-7. A nonzero value for X” represents a deviation from planarity in the amide group.
Packing Models The single polymer chain was arranged in the unit cell with its molecular axis through (O,O,O) and parallel to c. Translation along the c axis is arbitrary and was chosen such that the glycosidic oxygen 0-1had z = 0. The only packing parameter remaining is a, the orientation of the chain about the helix axis. No useful convention exists to define 6, = 0, and an origin was assigned arbitrarily. Increase in corresponds to anticlockwise rotation of the chain about the c axis. Since the chitin chain is polar and y # 90°, there are two possible parallel chain structures for P-chitin: one with the chain oriented “up” in the unit cell (i.e., 20-5 > Z C . ~ ) ,and the other with the chain oriented “down” in the unit cell (i.e., 20.5 < zc-5). Calculation of R values for otherwise similar “up” and “down” models easily demonstrated that the
STRUCTURES OF @-CHITIN
1587
structure must contain “up” chains, as was the case for c e l l ~ l o s e . ~ ~ Furthermore all “down” structures that could form reasonable N-H-. O=C hydrogen bonds were stereochemically unacceptable due to short contacts between the sheets. All calculations described hereafter refer to refinement of models containing “up” chains.
Refinement Method The possible structural models for 0-chitin were refined using the “linked-atoms’’ least-squares method of Arnott and W ~ n a c o t t . ’ ~By adjusting the refinable parameters, using a least-squares process, the function minimized is 0, where
A(Fm2) is the difference between the squares.of the observed (F,(obs)) and calculated (F,(calc)) structure factor amplitudes for the rnth of M reflections. The values of F , (calc) incorporated scale ( K ) and isotropic temperature (B) factors derived in the least-squares procedure. w, is the weight to be applied to the mth observation. In the course of this investigation, a Cruickshank weighting schemels was used, with constants determined by least-squares. In cases where more than one hkl plane contributed to an observed reflection, the calculated structure amplitude used for comparison is given by
where the F,. (calc) value is the calculated structure amplitude for the r t h of R planes contributing to the observed reflection. In the minimization of 8, the explicit variables are the molecular torsion angles x,X I , x N ,packing parameter and K and B. In order to examine potential structural models containing specific hydrogen bonds, constraints were incorporated, which limited certain interatomic distances to ranges within specified maximum and minimum values. These constraints G, were applied with respect to the Lagrange multipliers A, in the equation for 8. In order to consider alternative structural models, the R values were compared using the statistical tests due to Hami1t0n.l~ For each model, R” is computed where
+,
The more common crystallographic residual
is also quoted but is not used in the quantitative comparisons.
GARDNER A N D BLACKWELL
1588
Refinement The structure was refined initially by varying five parameters: two molecular parameters, x and x'; one packing parameter, @; and two crystallographic parameters, the isotropic temperature factor, B, and the overall isotropic scale factor, K . These parameters were refined simultaneously to give the best least-squares fit between the observed and calculated structure amplitudes. The preliminary model produced had R = 0.244 and R" = 0.274. Inspection of this model, hereafter referred to as the unconstrained model, showed that the value of x(= -81.7') oriented the CHzOH group so as to produce a marginally short contact: 0-5-0-6 = 2.61 A. To remove this deficiency, a constraint was incorporated such that 0-5-.0-6 1 2.80 A. Refinement of this constrained model led to residuals of R = 0.250 and R" = 0.288. According to the statistical tests of Hamilton,lg for least-squares refinement of 5 parameters against 61 observations, this difference in R" is significant only a t the 50% level, i.e., either model could be chosen if selected only on the basis of R values. Thus in terms of crystallographic and stereochemical criteria, the model with the 0-5-0-6 constraint is selected in preference to the unconstrained alternative. A defect of this model is that the 0-6-H group is not involved in donor hydrogen bonding. For comparison with later models, this structure will be referred to as the nonbonded model. The calculated structure amplitudes for the nonbonded model are compared with the observed data in Table I. The final values of the refined parameters, their estimated standard deviations, and the R values are given Table 11. @, which defines the orientation of the chains about the c axis, has a value of 110.lo,with an estimated standard deviation of ua = 1.3', indicating that the X-ray data defines the value of this parameter with a high degree of accuracy. (The actual value of has no significance, except for comparison with later refinements). The refined values of x and x' are -60.7' and 107.4', respectively; the estimated standard deviations associated with these parameters are o x = 9.7' and ox = 4.0'. Since a small change in x' moves a large group of atoms (the planar TABLE €1 Refined Crvstal Parameters for p-Chitin Models Parameters
Nonbonded model
X
-60.7" (9.7")s 107.4" (4.0") 110.1" (1.3") 85.0 (6.7") 8.2 (1.8")
Intrasheet hydrogen bond
Intersheet hydrogen bond ~~
X
0
K B ~~~
~~~~~~
~
(4.9") (3.6") (1.3") (7.2") (1.9")
142.6" 116.3" 108.9" 99.4 6.7
(8.8") (3.3") (1.6") (9.9") 12.4")
~
Special constraints on refining model
R R'
-54.8" 103.0" 110.4" 95.7 7.1
0-5
.
. . 0-6 > 2.80 A
0.250 0.288
0 - 5 . . . 0 - 6 > 2.80 A
0 - 5 . . . . 0 - 6 > 2.80 A
2.65 A G 0-6; ---0-7 G 2.90 A
2.65 A
0.267 0.,302
0-6---0-7;1 0.342 0.369
aThe numbers in parenthesis are the estimated standard deviations.
2.90
a
STRUCTURES OF &CHITIN
1589
amide group) it is not unreasonable that the data is more sensitive to x’ than to x and that cX’is smaller than cx. The value of x‘ places the planar amide group 12.6’ away from its initial trans orientation. The refined value of x(= -60.7’) is very close to the gg position (xgg = -6OO). x positions the -CH20H group approximately 180’ away from the position that would be necessary to form an intersheet hydrogen bond of the type incorporated in the Dweltz model.7 The nonbonded model contains a N-H- - -0 -7 distance of 2.70 between residues along the a axis. The only other intermolecular contact of less than 3.20 is a 3.14 separation distance between the 0 - 7 oxygen and 0-6’,. (To facilitate the description of alternative intermolecular hydrogen bonding possibilities in the @-chitin structure, the following nomenclature is used: numbered atoms (unprimed, unsubscripted) comprise the first residue of the chain (coordinates (x,y,z)); atoms in the second residue of the same chain (-x,-y,z $) are primed (see Figure 1); atoms in the molecule related to the origin molecule by a translation of -a (i.e. (x - l,y,z) and (-x-1,-y,z $)) are denoted by the subscript a (e.g., 0-7, and 0-7’,); and finally, atoms in the molecule related to the origin molecule by translations of +a and + b (i.e. (x l,y 1,z) and (-x 1,-y 1, z K ) are indicated by the subscripted (e.g., 0-7d and 0-7’d). Further refinement incorporated the possibility of nonplanarity of the amide group. The resulting structure had x‘’ = -6.3O and R” = 0.285. This improvement in R” (as compared to R” = 0.288 for the model containing planar amide groups) is not statistically significant and thus inclusion of nonplanarity of the amide group was accordingly not justified. The nonbonded model gives good agreement between the observed and calculated structure amplitudes, but the CH2OH groups are not hydrogen bonded. Two possibilities for hydrogen bonding present themselves: (i) an intrasheet hydrogen bond to the carbonyl oxygen in the next chain along the a axis: 0-6’, -H- - -0-7; and (ii) an intersheet hydrogen bond to the carbonyl oxygen in the chain along the a b diagonal: 0-6-H- - -0-7’d. In view of the stereochemistry of the chain and the unit cell dimensions, these are the only two hydrogen bonds which are possible for the CHpOH groups in this structure. Since from infrared evidence the CH20H groups must be hydrogen bonded, the model was constrained further to form first the intra- and then the intersheet bond, and the five parameters were refined in each case. The results for these two refinements are discussed in turn.
+
+
+
+
+
Intrasheet 0-6‘, -H-
+
+
- - 0 - 7 Hydrogen Bond
Refinement of the structure of @-chitinwas performed for the model constrained such that the 0-6’,-0-7 distance was in the range 2.65-2.90 A. Projections of the resulting structure are shown in Figure 2. The
GARONER AND BLACKWELL
1590
0
b
Fig. 2. Projections of the proposed model for p-chitin. The structure contains N-H -O=C and 0-6',-H- - - 0 - 7 intrasheet hydrogen bonds.
fractional coordinates of the model are found in Table 111 and the refined parameters and their associated standard deviations are given in Table 11. T h e residuals for this structure are R = 0.267 and R" = 0.302. T h e calculated structure factor amplitudes are given in Table I. @ refined to a value of = 110.4' (a+ = 1.2O); this is 0.3" (0.25aa) away from the value found in the nonbonded model. x and x' refined t o values of -54.8' and 103.0°, respectively. The new x value represents a rotation of 5.2" away from the gg position. T h e lower estimated standard deviation of this parameter (a, = 4.9') is due t o the presence of the 0-7--0-6', nonbonded constraint which severely limits the allowed value of x. T h e value of x' (-103.0') indicates that in the constrained model the planar amide group is rotated an additional 4.4' away from the initial trans position of the group. T h e reorientation is presumably to allow formation of the desired 0-6', -H- - -0-7 hydrogen bond. This reorientation (and, to a lesser degree, the small rotation of the chain) has increased the N-H- - -0-7 hydrogen bond length t o 2.76 A. Through the additional constraint, the structure contains a n 0-6', -H- - -0-7 intrasheet hydrogen bond of length 2.89 A. T h e structure contains no other intermolecular contacts of less than 3.2 A. T h a t w o hydrogen bonds are indicated in Figure 2.
STRUCTURES O F p-CHITIN
1591
TABLE I11 Fractional Coordinates for the N-Acetyl-D -glucosamine Residue in the Refined Structure of p-Chitin which contained the Zntrasheet 0-69-H--0 - 7 Hydrogen Bonda Atom
c-1 c-2 c-3 c-4
c-5
C-6 c-7 C-8 0-1 0-3 0-5 0-6 0-7 N
x/a
Y/b
z/c
- .010 -0.146 -0.007 -0.002 0.124 0.114 -0.366 -0.321 -0.161 -0.155 -0.026 -0.164 -0.589 -0.131
-0.038 -0.150 -0.130 0.027 0.132 0.290 -0.377 -0.527 -0.051 -0.227 0.105 0.315 -0.330 -0.299
0.386 0.291 0.159 0.114 0.218 0.183 0.392 0.438 0.500 0.068 0.337 0.153 0.401 0.399
aThis is the final proposed structure for p-Chitin.
Intersheet 0 - 6 -H-
- -0-7'd
Hydrogen Bond
The &chitin structure was refined with the model containing an additional constraint such-that the 0-6-0-7'd distance was in the range 2.65-2.90 A. The refined model containing this intersheet hydrogen bond is shown in Figure 3. The refined parameters and associated estimated standard deviations are given in Table 11. The residuals for this model are R = 0.342 and R" = 0.369. Comparison of observed and calculated structure factor amplitudes is made in Table I. @J, the packing parameter, refined to a value of = 108.9 with an estimated standard deviation of u = 1.6O. @J is approximately 0 . 7 5 ~from the value found in the nononded model. x and x' have values of 142.6" and 116.3", respectively, with estimated standard deviations of ux = 8.8" and ux' = 3.3". The CHzOH groups are oriented 37.4" away from the gt position; the 0-6-H- - -0-7', bond length is 2.84 A. The amide group is rotated 9.3O from the position found in the unconstrained model. The 0-7- - -N, intrasheet hydrogen bond has length 2.68 A. Other nonbonded contacts of less than 3.2 A involve the amide group to which the 0 - 6 is hydrogen bonded: 0 - 6 4 - 7 d is 2.81 A and 0-6- *C-8d is 2.67 A in length. Inclusion of a single constraint which removed both short contacts resulted in a small increase in R". Figure 4 shows a plot of R" against x for the unconstrained model where x was varied in increments of 5", with all the parameters held constant. Regions where unacceptable chemical contacts occur are also shown. The R" curve shows two minima, the lowest corresponding to the unconstrained model a t x = 258.3", and the second a t x N 125" and
GARDNER AND BLACKWELL
1592
0
m
0
b
Fig. 3. Projections of the rejected model which contained an 0-6-H- - - 0 - 7 ' d intersheet hydrogen bond. The structure also contains the 0-7-N, intrasheet hydrogen bond.
higher R". The x values for the refined unconstrained, nonbonded, intrasheet hydrogen bond and intersheet hydrogen bond models are indicated by A, B, C, and D, respectively. It is interesting that the two minima fall in the only two regions which are stereochemically allowed. The subsidiary minimum was not sufficiently deep to prevent refinement to the true minimum, i.e., removal of the hydrogen-bonding constraint from model D led to refinement to model B (or to model A if the 0-5-0-6 constraint was relaxed).
DISCUSSION In summary, three models for P-chitin have been considered. The structure with the lowest R values, R = 0.250, R" = 0.288, is stereochemically acceptable but does not provide for hydrogen bonding involving the -CHzOH group. A second structure, containing an 0-6', -H- - - 0 - 7 hydrogen bond, was found to be stereochemically acceptable and has R values of R = 0.267 and Rf' = 0.302. A third structure, containing an intersheet 0-6-H- - -0-7'd hydrogen bond (as has been previously proposed), was found to have a short 0-6-C-8d contact and has R values of R = 0.342 and R" = 0.369.
STRUCTURES OF @-CHITIN
1593
c
0
60
i0
X
Fig. 4. A plot of R” as a function of x, which defines the rotational conformation of the -CH*OH group about the C-5-C-6 bond. The angular ranges for x where there are intra- and intermolecular short contacts are indicated by the boxes. The numbers in the boxes refer to the atom in contact with 0-6. Open boxes indicate contacts in the range 2.6-2.7 A. x values for the refined models considered in this paper are indicated by the vertical lines: (A) unconstrained model with the lowest R and R”; (B) nonbonded model with no bad contacts but with nonhydrogen bonded -CH20H group; (C) intrasheet hydrogen bond model constrained to form the 0-6’,-H- - -0-7 bond; and (D) intersheet hydrogen bond model constrained to form 0-6-H- - -0-7‘d bond.
By all criteria, the model containing the 0-6-H- - -0-7’d intersheet hydrogen bonding can be rejected. Firstly, the structure contains an unacceptable short nonbonded interaction and, secondly, application of Hamilton’s test for least-squares refinement of 5 parameters against 61 observations shows that there is much less than 1 chance in 200 that the structure containing the intersheet hydrogen bond is as good a representation of the true structure as either the nonbonded model or the structure containing the intrasheet hydrogen bond. The choice between the remaining two models is not as straightforward as the rejection of the intersheet hydrogen bond model. Neither of the remaining models contains unacceptable intermolecular nonbonded contacts, and although the nonbonded model has a slightly lower R” value (0.288 as compared to 0.302), application of Hamilton’s
1594
GARDNER A N D BLACKWELL
test shows t h a t this difference is not significant and there is a 50-50 chance t h a t either structure is correct. However, the basic difference between the two models is t h a t one allows for hydrogen bonding involving the -CH20H group while the other does not, and on this basis the nonbonded model is rejected and the structure containing the 0-6‘, H- - -0-7 intrasheet hydrogen bond is proposed as the structure for 0chitin. For trial and error refinement of &chitin, BlackwelF reported poor agreements for four individual reflections: i.e., the 130, 120, 121, and 102 planes. He found that the intensities for these four sets of planes were very sensitive to the position of the -CH20H group, but t h a t variation of x did not improve the overall agreement. In the present investigation the same discrepancies are observed in the model containing the intersheet hydrogen bond, but refinement of x produced models (nonbonded and intrasheet hydrogen bond) that give good agreements for these reflections combined with superior overall agreement. In the proposed intrasheet hydrogen bond structure the 0-3-H and 0-6-H bonds are oriented such t h a t both should give rise to parallel dichroism in the 0 - H stretching region of the infrared spectrum. In contrast, the 0-6-H- - -0-7’d intersheet hydrogen bond should give rise to perpendicular dichroism. @-Chitinis easily swollen in water: two crystalline hydrate structures have been recognized6 and suggested t o contain one and two water molecules per residue. In the hydrated structures the hydrogen-bonded sheets remain intact but are moved apart t o include the water molecules. T h e intrasheet hydrogen bond model is consistent with this behavior whereas a n intersheet bond would tend to hold the sheets together and stabilize the anhydrous form. The intrasheet 0-6’,- - -0-7 bond can be seen as an additional stabilizing factor for the sheets of chitin chains, also linked by the amide N-H- - -O=C bonds. I t is possible t o see numerous points of analogy between the structures of 0-chitin and cellulose I. Both structures contain extended chains with the same (parallel) sense. Depending on the source, specimens show different degrees of crystallinity, but the large crystalline fibrils are seen as arrays of smaller “elementary fibrils” in electron microscope structures of deformed samples.20,21 A t the molecular level, both structures are seen as a n array of hydrogen-bonded sheets. In cellulose I, the 0-2-H group forms an intramolecular 0-2’-H- - -0-6 bond, and since there is also an 0-3-H- - -0-5’ intramolecular bond, the ribbonlike chain has approximately parallel hydrogen bonds on both sides of the glycosidic bridge. In 0 chitin, the same 0-3-H-0-5’ intramolecular bond is present, but the N-H is involved in bonding with the amide group on the next chain. Bonding of the type 0-6-H- - -N’ would not be expected: such acceptor bonds are apparently not formed by amide groups and in any case, this bond is stereochemically unacceptable. However, rotation of the CH20H group from the t g position of cellulose I t o the gg position in P-chitin allows formation of the 0-6’,-H- -0-7
STRUCTURES OF @-CHITIN
1595
bond giving a similar chain structure with parallel hydrogen bonds on both sides of the chain. On conversion of cellulose I to cellulose 11, the 0-2’-H-0-6 intramolecular bonds are broken and 0-2-H- - - 0 - 2 d bonds are formed between antiparallel chains.22 This is analogous to the conversion of P-chitin to the a-form and it will be interesting to see if a-chitin contains additional intersheet hydrogen bonds. This work was supported by N.S.F. Grant No. GH 34227 and N.I.H. Research Career Development Award No. AM 70642 (to J.B.)
References 1. Rudall, K. M. (1963) Adu. Insect Physiol. 1,257-313. 2. Carlstrom, D. (1957) J . Biophys. Biochem. Cytol. 3,669-683. 3. Dweltz, N. E. (1960) Biochem. Biophys. Acta 44,416-435. 4. Ramakrishnan, C. & Prasad, N. (1972) Biochem. Biophys. Acta 261,123-135. 5. Dweltz, N. E. (1961) Biochem. Biophys. Acta 51,283-289. 6. Blackwell, J . (1969) Biopolymers 7,281-298. 7. Dweltz, N. E., Colvin, J . R. & McInnes, A. G. (1968) Can. J . Chem., 46, 1513-1521. 8. Lotmar, W. & Picken, L. E. R. (1950) Experientia 6,58-59. 9. Blackwell, J., Parker, K. D. & Rudall, K. M. (1965) J . Mar. B i d . Assn. U.K. 45, 659-661. 10. Falk, M., Smith, D. S., McClachlan, J . & McInnes, A. G. (1966) Can. J . Chem. 44, 2269-2286. 11. Blackwell, J., Parker, K. D. & Rudall, K. M. (1967) J . Mol. Biol. 28,282-385. 12. Gardner, K. H. & Blackwell, J . (1974) Biochem. Biophys. Acta 342,232-237. 13. Gardner, K. H. & Blackwell, J . (1974) Biopolymers 13,1975-2001. 14. Arnott, S. & Wonacott, A. J . (1966) Polymer 7,157-166. 15. Arnott, S. & Scott, W. F. (1972) J . Chem. SOC.Perk. Trans. 11,324-335. 16. Johnson, L. N. (1966) Acta Cryst., 21,885-891. 17. Sundararajan, P. R. & Rao, V. S. R. (1969) Biopolymers 8,305-312. 18. Cruikshank, D. W. J . & Pilling, D. E. (1961) Computing Methods and the Phase Problem i n X - R a y Crystal Analysis, p. 46, Oxford, Pergamon Press. 19. Hamilton, W. C. (1965) Acta Cryst. 18,502-510. 20. Gardner, K. H. & Blackwell, J. (1971) J . Ultrastruct. Res. 36,725-731. 21. Gardner, K. H. & Blackwell, J. (1971) J . Polym. Sci. C36,327-340. 22. Kolpak, F. J., & Blackwell, J. (1975) Submitted to Macromolecules.
Received September 27,1974 Accepted March 12,1975
