J. Mol. Biol. (1990) 213, 1-5
COMMUNICATIONS
Cross-linkage Sites in Type I Collagen Fibrils Studied by Neutron Diffraction T. J. Wess, A. Miller and J. P. Bradshaw Department of Biochemistry University of Edinburgh Medical School Hugh Robson Building, George Square Edinburgh EH8 9XD, Scotland
(Received 30 August 1989; accepted 20 December 1989) Cross-links in tendon collagen are essential for the biomechanical strength of healthy tissue. The nature and position of these cross-links has long been a subject for conjecture. We have approached this problem in a non-destructive manner, by studying neutron diffraction from collagen fibrils that have been specifically deuterated by reduction at keto-amine and Schiff base groups with sodium borodeuteride (NaB2HH,). The intensities of the first 23 meridional reflections were recorded for both native and reduced tendons. These data were used to calculate the neutron-scattering density profile of the 67 nm (D) repeat of type I collagen fibrils in rat tail tendon. This approach not only succeeds in determining the location of the cross-linkage sites with respect to the fibril structure, as projected onto the fibre axis, but also presents a novel form of the isomorphous derivative solution to the phase problem.
Collagen fibres gain mechanical strength from specific post-translational reactions that result in cross-linking between molecules. The process of cross-linking is initiated by the oxidative deamination of e-NH 2 groups of specific residues of lysine or hydroxylysine. The resultant aldehydes can condense together, or with other lysine and hydroxylysine residues, to form intra and intermolecular crosslinks. The specific nature of the cross-link is dictated by the adjacent residues in each collagen trimer (intra-molecular cross-links) and the staggered packing of molecules in fibrils (inter-molecular cross-links). Several types of crosslink have been isolated (Eyre el al., 1984; Eyre, 1987): one type is characterized by its reducibility by reagents such as sodium borohydride, NaBH4, and its deuterated and tritiated forms. A process of age-related cross-link maturation has been proposed, in which the reducible form is regarded as an intermediate (Robins & Bailey, 1975; Bailey & Skimokamaki, 1971). The reduction of collagen using NaBZH4 has been shown to incorporate deuterons into a limited number of residues, these primarily being the crosslinks between molecules, the deaminated precursors of these cross-links and hexitol lysine derivatives. These all contain keto, keto-amine or Schiff bases that react with NaBI-I 4 (Robins & Bailey, 1977). However, the localization of these sites by biochemical separation techniques has been hindered by degradation of the cross-links during the hydro0022-2836/90/090001-05 $03.00/0
lytic process involved in analysis, and thus it may not be possible to detect all reducible sites. R a t tail tendon is reported to contain allysine-derived crosslinks (Bailey & Peach, 1968); more recent investigations have postulated the existence of hydroxyallysine-mediated cross-links (Nakamura, 1987). There is a general consensus that in rat tail tendon there are cross-links between residues in laterally adjacent telopeptides (these being the non-helical domains at the N and C termini of the molecule). Another cross-link has been shown to link a telopeptide residue with a residue in the main triple helix domain of the molecule (Barnard et al., 1987). The existence of further cross-links, often of a tri or tetra-functional nature, has been debatable; however, evidence for the structure and formation of a stable histidine-based trifunctional cross-link in skin collagen has been reported (Yamauchi el al., 1987), as has evidence for a hydroxylysine-derived pyridinoline cross-link that is not an artifact of acid hydrolysis (Fujimoto, 1980). To date, the cross-links shown to exist in type I collagen from rat tail tendon can account only for the stabilization of sheets of connected molecules (~d]ley et dl., 1980; Miller, 1982), and not for the three-dimensional structure that is suggested b y X-ray studies (North et al., 1954; Miller & Wr~y, 1971; Fraser et al., 1983, 1987). Chemical analysis has provided evidence in periodontal ligament ~or an intermolecular cross-link that supports tl~eedimensional molecular packing (Yamauchi et al., ~) 1990 Academic Press Limited
2
T.J.
Wess et al.
1986). It is therefore possible that in collagen crosslinks may exist that have not been located by degradative chemical analysis and may provide the link between topologically linked sheets of molecules and the three-dimensional crystalline domains found in rat tail tendon. The specific nature of collagen deuteration by NaB2H4 permits an examination of the axial structure of tendon collagen. The position of reducible groups can also be determined by neutron diffraction. This is because the introduction of deuterons into a structure in a specific manner will alter the neutron-scattering density profile of the molecule and therefore also the diffraction image. The coherent neutron-scattering cross-section of deuterons is large and when deuterons are incorporated into the molecule, the position of the meridional reflections in the diffraction pattern remains fixed, while the intensity of the reflections change. This allows the deuterated derivatives of rat tail tendon to be treated as isomorphous derivatives. The methods of using multiple isomorphous replacement/addition to obtain phase information are standard in protein X-ray crystallography and have been used to determine the axially projected electron density profile of a rat tail tendon (Bradshaw et al., 1989). The use of isomorphous derivatives in structure factor phase determination allows the phase to be determined unambiguously a n d therefore is not dependent on the constraints of model building, as was the case with the axial neutronscattering density profile reported by Hulmes et al. (1980). The neutron-scattering density profile of collagen and its deuterated derivatives are presented here. The deuterated isomorphous derivatives for neutron diffraction were prepared using a method developed from that reported by Bailey & Peach (1968). One gram (wet weight) of rat tail tendon was obtained from four to six-month-old Wistar rats for each derivative and for the native sample. The length of deuteration with NaB2H4 was varied in an attempt to cause differences in labelling position and density. Due to the limited time allocated for neutron diffraction, the number of derivative types had to be limited to two. Reaction conditions were therefore 20 milligrams of NaB2H4 in phosphate/ borate-buffered Ringer solution (pH 7"5), to which one gram of rat tail tendon was added. The times of exposure of rat tail tendon to the reducing agent were six minutes and 30 minutes, respectively, Success of cross-link reduction could be judged by observing the lack of solubility of reduced collagen in 0"5 ~-acetic acid. The axial packing of the collagen in reduced samples was shown not to have been altered from the native state, as judged by the X-ray fibre diffraction patterns obtained from the samples. The assumption of isomorphism is therefore valid. Approximately 150 tendon fibres (1 g wet weight) of average individual diameter approximately 200 micrometres were clamped close together on a frame S° as to produce a parallel array of collagen fibrils.
Experiments were conducted at the Institut Laue Langevin, Grenoble. Two diffractometers were used: D1 l, a low-angle diffraction apparatus (Ibel, 1976) and D16, a four-circle stepping diffractometer. For experimentation on D1 l, the frame was sealed in a Teflon cell with quartz windows. For experimentation on D16, the frame was mounted and sealed in an aluminium can. Diffraction experiments were conducted on D l l for lower axial orders of diffraction (1 to 7) and on the diffractometer D16 for the higher orders (5 to 23). Experiments on Dll were carried out with the collagen fibres perpendicular to the beam, and with the axis of the fibres vertical. Two sample-todetector distances were used in order to collect a low-angle dataset. To record the first three orders and the transmitted intensity on the 64 centimetre by 64 centimetre area detector, neutrons of wavelength 0"9 nanometre were selected and the sampleto-detector distance set at 5'6 metres. In order to record orders 2 to 7, the same wavelength neutrons were used and the sample-to-detector distance shortened to 2.53 metres by moving the detector. Collimation consisting of a circular cadmium aperture (diameter l0 mm) was used in order to prevent stray neutrons hitting the sample cell. Transmitted beam intensities were measured for a fixed time interval after attenuation of the beam at source with a graphite moderator. Data were collected for the transmitted beam intensities at a number of 2H20/H20 ratios, to allow the scaling of the intensities between datasets to be determined. The beam stop was removed to record the flux of the through beam. The D l l data were initially corrected for detector response efficiency using the isotropic incoherent scattering of H20. Higher orders of axial reflections were measured on the four-circle diffractometer D16 using 0)/20 scans with the beam axis horizontal and the collagen fibres in the horizontal plane. The experimental parameters used were as in Hulmes et al. (1980). Integrated intensities for each reflection obtained using instrument D11 were determined by using a suite of programs written for FORTRAN 77. It was therefore possible to merge the data from the two different camera lengths by using standard programs, as described by Bradshaw et al. (1989). The intensity of the second and third orders in each D l l dataset were used to determine a scale factor for merging the data. In order to relate these intensities to the true structure factor for each reflection, it was necessary to correct for specimen disorientation, beam divergence, wavelength spread and the intersection of each reflection with the Ewald sphere (Lorenz factor). The magnitude of these corrections was determined experimentally by Hulmes et al. (1980). The corrections were applied to each D l l dataset. D16 data were analysed using programs written also for :FORTRAN 77. This allowed the intensity of each reflection to be estimated by the relative height of each peak. The intensities could then be
Communications Table 1 Rat tail tendon neutron diffraction data Amplitudesf Order
Native
6 min
30 min
Phases:~ (native)
1 2 3 4 5 6 7 8 9 l0 11 12 13 14 15 16 17 18 19 20 21 22 23
10,000 1138 4522 1212 2584 2204 1144 1759 1487 1263 1229 1466 760 889 1045 654 527 1059 949 1109 1096 1265 1030
10,000 1129 4797 1233 2582 1896 1208 1525 1375 1023 1121 1491 830 1050 934 1029 1119 780 687 1084 1156 1124 559
10,000 1077 4359 1086 2767 2578 1680 2283 1893 1354 1563 1620 633 t078 t079 566 704 964 945 1097 874 1131 463
3"37 0"95 0"66 5-65 3.71 0'71 2-52 2 73 4-87 5'03 4'03 3"54 3-97 0"73 3-00 4"30 2"05 4"49 5"05 1-74 0"59 0"45 2-43
Corrected and scaled to a Ist order value of 10,000. :~In radians. corrected for local background scattering and a Lorenz factor applied to each reflection. These d a t a were then scaled to the corrected intensities of D11 d a t a b y using the fifth, sixth and seventh orders common to both datasets. The resulting intensities were all scaled to a first order value of 10,000 and are shown in Table 1. As with the determination of three-dimensional molecular structure using X - r a y diffraction, the method of multiple isomorphous addition also required information relating to the position of label in the unit cell. This information was obtained from two sources, the difference Patterson function resulting from derivative and native diffraction intensities and known amino acid sequence information. Autocorrelation functions of the proposed labelling site were compared with the difference Patterson maps. The difference Patterson functions created showed discrete peaks indicating deuteration at specific sites within the axial unit cell. The n u m b e r of peaks was judged to be sufficiently small to allow the determination of deuteron label position and hence the phases. The position and relative strengths of these peaks were used to construct and modify models of the labelling sites within the molecule. Initial autocorrelations used only the crosslinkage sites at the telopeptides for positions of deuteration. Compared to the difference P a t t e r s o n maps, this produced a map with insufficient peaks. The possibility of other sites being d e u t e r a t e d was investigated. Many molecular species have been isolated, from collagen, t h a t can be reduced b y NaBH4. These include the cross-linkage groups in
3
various forms, hexitol lysine residues and the processed precursors of cross-links, the aldehyde derivatives of lysine and hydroxylysine (Bailey et al., 1974; Robins, 1982). I t is possible t h a t these components can be deuterated in r a t tail tendon and therefore contribute to the difference in neutronscattering between native and derivative samples. A number of trial distributions of deuteration in the axial unit cell were tested. These distributions were dictated by the position of lysine and h y d r o x y lysine residues in a D repeat. The autocorrelation functions tested began to correspond well with the difference P a t t e r s o n function produced when hydroxylysine residues were used as sites of deuteration. P r e d o m i n a n t deuteration sites corresponding to lysine residues in the gap were also postulated (Lys434 alpha l, Lys885 alpha 1, alpha 2), since this significantly improved the match between the autocorrelation and difference Patterson function. The combined contribution of deuteration at the teiopeptide regions, hydroxylysine-rich regions and a lysine-rich region in the gap provided an initial map of label only. The significant differences in the six minute and 30 minute NaB2H4-1abelled derivative difference Patterson maps could be produced by altering the relative occupancy at each labelling site. This gave a starting point t h a t allowed phase determination. The process of phase determination and refinement was the same as t h a t used b y Bradshaw et al. (1989). An iterative series of cyclic calculations used the intensity of native and derivative neutron diffraction peaks to determine the phase c o m p o n e n t of each structure factor for both native and derivative tendon samples. The relative intensity of diffraction from native and derivative structures used values initially derived from transmission data. However, alterations of up to 15~/o of the scaling factors were used to optimize the refinement. This process was continued until the autocorrelation of the label in the neutron-scattering density profile showed good agreement with the difference Patterson map, taking into account the differences due to protein-label vectors in the difference P a t t e r s o n map. The neutron-scattering profile of native rat tail tendon (Fig. 1) has been determined to a resolution of ±0"043 D (the best to date). I t shows general agreement with t h a t deduced by model building (Hulmes et al., 1980). I t has a gap to overlap ratio of 0"49 : 0"51, which compares well with the most accurate value given by X - r a y diffraction (Bradshaw et al., 1989). The profile shows a n u m b e r of peaks and troughs superimposed on the g a p - o v e r l a p function. These features show some correlation with the banding pattern of positively stained electron micrographs, b u t the fundamental differences between the imaging characteristics of these two techniques mean t h a t profiles obtained b y them are not directly comparable. T h e profile presented here has telopeptide regions t h a t appear as peaks a t the end of each gap region. The C-terminal peak is
4
Wess et al.
T.J.
6 min
PI
v
over,°°
¢¢
0 X 0 t¢-
I
0
0.5 D
g
I.O
Figure 1. The neutron-scattering profile of rat tail tendon collagen projected onto the fibre axis. The horizontal scale is distance through the unit cell along the fibre axis and the vertical scale is neutron-scattering density/unit axial translation in arbitrary units. N and C denote N and C termini. higher than that at the N terminus. Although there will be underlying differences in scattering projected onto this region by the other segments of the axial structure, it could also be related to the difference in structure of the C and N-terminal telopeptides. The difference in the axial translation of residues in the two telopeptides as proposed by Hulmes et al. (1980) may contribute to the size of the peaks, since it was reported that the C-terminal telopeptide is more contracted per amino acid translation than is the N-terminal telopeptide. The difference profiles (Fig. 2) show a series of discrete peaks that can be related to the axial distribution of the reducible components in collagen. Two of the main peaks are situated approximately 0"4 D apart. The position of these peaks corresponds to the telopeptide positions in the native profile. Chemical analysis has revealed that cross-linking occurs in these regions (Light & Bailey, 1979), suggesting that either the cross-links or their precursors are being labelled here. This region of the profile also contains the majority of hydroxylysine residues, which may have been converted to hydroxyallysine by deamination and therefore are reducible. This creates a problem, for the cross-linking at the N terminus may be masked by other hydroxyallysine residues in this region. The resolution of the profile is sufficient to allow the separation of the cross-link(s) from a group of hydroxylysine residues in the gap region before the telopeptide. These latter residues appear to be reduced only after 30 minutes of exposure. This may be due to one of two causes: either steric blocking of the labelling in this region, which is unlikely, or the intrinsically different rates of chemical labelling for cross-links and precursor groups. The large number of reducible components in the N-terminal region of the profile may indicate that more cross-links exist than had been previously thought.
X~
J,elopept,de
8m
o--o--
Lys434 z
L y s 8 ~ 1~---~ ~ - ~ C telopeptide / , ~ r 0.1/0:2 0:3 0.4 0:5 0.6 0:7 0:8 0:9 1.0 o
alpha I chains Figure 2. The location of reducible sites along the fibre axis of the rat tail tendon collagen fibril unit cell, as revealed by subtraction of native fibril neutron-scattering density from derivative fibril neutron-scattering density; phased by isomorphous addition of deuterium. The axes are as Fig. 1. The lowest part of the Figure shows positive staining bands found in the collagen D repeat by electron microscopy. These correspond to parts of the structure containing positive charge and therefore lysine residues. (O) hydroxylysine; (0) lysine residues thought to be involved in cross-links.
Around the C terminus there are fewer reducible components. The telopeptide lysine residues in this region correspond laterally with three hydroxylysine residues (Hyl87 alpha 1 × 2, alpha 2). These residues probably comprise the C-terminal crosslinks, since the relative difference in peaks height after six and 30 minutes of labelling is not great. This implies deuteration of cross-links at-this position, since the deuteration appears to occur at a rate similar to the labelling of N-terminal cross-link(s). The size of the difference peak corresponding to both the C and N termini should be similar, since both contain the same reducible components. The difference in peak size may be due to the degree of contraction of the C-terminal telopeptide, which may have a differential effect on the accessibility of the two regions to the reducing compound, with. slower diffusion to the more compact C terminus. It can be seen also that both telopeptide regions are laterally aligned with hydroxylysine residues in both the alpha 1 and alpha 2 chains of collagen, The
Communications
similarity of these two regions indicates t h a t hydroxylation of lysine m a y be essential for crosslinking in this region. The maps also contain a series of peaks t h a t cannot readily be related to known cross-links. These are more predominant in the 30 minute map than the six minute map, and appear to be larger when they occur in the less stericaily restrictive gap region. The large peak in the gap region may result from the non-enzymic glycosylation of lysine residues, which would enable them to be reduced by NaB2Hd. The discrete nature of this peak indicates that, if it is caused by deuteration of a sugar linkage, then the glycosylation process may be more specific than is to be expected from the observations made by Le Pape et al. (1984), where non-enzymic glucosylation is found to be distributed t h r o u g h o u t the collagen molecule in diabetic tissue. The position of the peak corresponds to the axial position of lysine residues 434 of the alpha 1 chain (numbering does not include the telopeptides). The other peak t h a t appears in the gap region after 30 minutes of labelling corresponds to the lysine residues at position 855 in both alpha 1 and alpha 2 sequences. This is the first experimentally determined phasing of the meridional neutron diffraction pattern of collagen and has provided identification of the sites t h a t have been modified in an enzymic or non-enzymic manner to produce reactive groups t h a t can be labelled by specific deuteration. The authors would like to thank all the staff at the Institut Laue Langevin, especially the local contact Peter Timmins, and also Kevin Duff for technical assistance. References
Bailey, A. J. & Peach, C. M. (1968) Biochem. Biophys. Res. Commun. 33, 812-819. Bailey, A. J. & Skimokamaki, M. (1971). F E B S Letters, 16, 86-88. Bailey, A. J., Robins, S. P. & Balian, G. (1974). Nature (London), 251, 105-109.
5
Bailey, A. J., Light, N. D. & Atkins, E. D. T. (1980). Nature (London), 288, 408-413. Barnard, K., Light, N. D., Sims, T. J. & Bailey, A. J. (1987). Biochem. J. 244, 303-309. Bradshaw, J. P., Miller, A. & Wess, T. J. (1989). J. Mol. Biol. 205, 685-694. Chapman, J. A. & Hulmes, D. J. S. (1984). In Ultrastructure of the Connective Tissue Matrix (Ruggeri, A. & Motta, P.M., eds), pp. 1-33, Martinus Nijhoff, Amsterdam. Eyre, D. R. (1987). Methods Enzymol. 144, 115-140. Eyre, D. R., Paz, M. A. & Gallop, P. M. (1984). Annu. Rev. Biochem. 53, 717-748. Fraser, R. D. B., MacRae, T. P., Miller, A. & Suzuki,, E. (1983). J. Mol. Biol. 167, 497-521. Fraser, R. D. B., MacRae, T. P. & Miller, A. (1987). J. Mol. Biol. 193, 115-125. Fujimoto, D. (1980). Biochem. Biophys. Res. Coramun. 93, 948-953. Galloway, D. (1982). In Collagen in Health and Disease (Weiss, J. B. & Jayson, M. I. V., eds), pp. 528-557, Churchill Livingstone, Edinburgh. Hulmes, D. J. S., Miller, A., White, S. W., Timmins, P. A. & Berthet-Colominas, C. (1980). Int. J. Biol. Macromol. 2, 338-345. Ibel, K. (1976). J. Appl. Crystallogr. 9, 630-643. Le Pape, A., Guitton, J. D. & Muh, J.-P. (1984). F E B S Letters, 170, 23-27. Light, N.. D. & Bailey, A. J. (1979). F E B S Letters, 97, 183-188. Miller, A. (1982). Trends Biochem. Sci. 7, 13-18. Miller, A. & Wray, J. S. (1971). Nature (London), 230, 437-438. Nakamura, Y. (1987). Int. J. Biol. Macromol. 9, 281-290. North, A. C. T., Cowan, P. M. & Randall, J. T. (1954). Nature (London), 174, 1142-1143. Robins, S. P. (1982). Methods Biochem. Anal. 28, 329-379. Robins, S. P. & Bailey, A. J. (1975). Biochem. J. 149, 381-385. Robins, S. P. & Bailey, A. J. (1977). Biochem. J. 163, 339-346. Yamauchi, M., Katz, P. Z. & Mechanic, G. L. (1986). Biochemistry, 25, 4907-4913. Yamauchi, M., London, R. E., Guenat, C., Hashimoto, F. & Mechanic, G.L. (1987). J. Biol. Chem. 252, 11428-I1434.
Edited by V. Luzzati
COMMUNICATIONS
Cross-linkage Sites in Type I Collagen Fibrils Studied by Neutron Diffraction T. J. Wess, A. Miller and J. P. Bradshaw Department of Biochemistry University of Edinburgh Medical School Hugh Robson Building, George Square Edinburgh EH8 9XD, Scotland
(Received 30 August 1989; accepted 20 December 1989) Cross-links in tendon collagen are essential for the biomechanical strength of healthy tissue. The nature and position of these cross-links has long been a subject for conjecture. We have approached this problem in a non-destructive manner, by studying neutron diffraction from collagen fibrils that have been specifically deuterated by reduction at keto-amine and Schiff base groups with sodium borodeuteride (NaB2HH,). The intensities of the first 23 meridional reflections were recorded for both native and reduced tendons. These data were used to calculate the neutron-scattering density profile of the 67 nm (D) repeat of type I collagen fibrils in rat tail tendon. This approach not only succeeds in determining the location of the cross-linkage sites with respect to the fibril structure, as projected onto the fibre axis, but also presents a novel form of the isomorphous derivative solution to the phase problem.
Collagen fibres gain mechanical strength from specific post-translational reactions that result in cross-linking between molecules. The process of cross-linking is initiated by the oxidative deamination of e-NH 2 groups of specific residues of lysine or hydroxylysine. The resultant aldehydes can condense together, or with other lysine and hydroxylysine residues, to form intra and intermolecular crosslinks. The specific nature of the cross-link is dictated by the adjacent residues in each collagen trimer (intra-molecular cross-links) and the staggered packing of molecules in fibrils (inter-molecular cross-links). Several types of crosslink have been isolated (Eyre el al., 1984; Eyre, 1987): one type is characterized by its reducibility by reagents such as sodium borohydride, NaBH4, and its deuterated and tritiated forms. A process of age-related cross-link maturation has been proposed, in which the reducible form is regarded as an intermediate (Robins & Bailey, 1975; Bailey & Skimokamaki, 1971). The reduction of collagen using NaBZH4 has been shown to incorporate deuterons into a limited number of residues, these primarily being the crosslinks between molecules, the deaminated precursors of these cross-links and hexitol lysine derivatives. These all contain keto, keto-amine or Schiff bases that react with NaBI-I 4 (Robins & Bailey, 1977). However, the localization of these sites by biochemical separation techniques has been hindered by degradation of the cross-links during the hydro0022-2836/90/090001-05 $03.00/0
lytic process involved in analysis, and thus it may not be possible to detect all reducible sites. R a t tail tendon is reported to contain allysine-derived crosslinks (Bailey & Peach, 1968); more recent investigations have postulated the existence of hydroxyallysine-mediated cross-links (Nakamura, 1987). There is a general consensus that in rat tail tendon there are cross-links between residues in laterally adjacent telopeptides (these being the non-helical domains at the N and C termini of the molecule). Another cross-link has been shown to link a telopeptide residue with a residue in the main triple helix domain of the molecule (Barnard et al., 1987). The existence of further cross-links, often of a tri or tetra-functional nature, has been debatable; however, evidence for the structure and formation of a stable histidine-based trifunctional cross-link in skin collagen has been reported (Yamauchi el al., 1987), as has evidence for a hydroxylysine-derived pyridinoline cross-link that is not an artifact of acid hydrolysis (Fujimoto, 1980). To date, the cross-links shown to exist in type I collagen from rat tail tendon can account only for the stabilization of sheets of connected molecules (~d]ley et dl., 1980; Miller, 1982), and not for the three-dimensional structure that is suggested b y X-ray studies (North et al., 1954; Miller & Wr~y, 1971; Fraser et al., 1983, 1987). Chemical analysis has provided evidence in periodontal ligament ~or an intermolecular cross-link that supports tl~eedimensional molecular packing (Yamauchi et al., ~) 1990 Academic Press Limited
2
T.J.
Wess et al.
1986). It is therefore possible that in collagen crosslinks may exist that have not been located by degradative chemical analysis and may provide the link between topologically linked sheets of molecules and the three-dimensional crystalline domains found in rat tail tendon. The specific nature of collagen deuteration by NaB2H4 permits an examination of the axial structure of tendon collagen. The position of reducible groups can also be determined by neutron diffraction. This is because the introduction of deuterons into a structure in a specific manner will alter the neutron-scattering density profile of the molecule and therefore also the diffraction image. The coherent neutron-scattering cross-section of deuterons is large and when deuterons are incorporated into the molecule, the position of the meridional reflections in the diffraction pattern remains fixed, while the intensity of the reflections change. This allows the deuterated derivatives of rat tail tendon to be treated as isomorphous derivatives. The methods of using multiple isomorphous replacement/addition to obtain phase information are standard in protein X-ray crystallography and have been used to determine the axially projected electron density profile of a rat tail tendon (Bradshaw et al., 1989). The use of isomorphous derivatives in structure factor phase determination allows the phase to be determined unambiguously a n d therefore is not dependent on the constraints of model building, as was the case with the axial neutronscattering density profile reported by Hulmes et al. (1980). The neutron-scattering density profile of collagen and its deuterated derivatives are presented here. The deuterated isomorphous derivatives for neutron diffraction were prepared using a method developed from that reported by Bailey & Peach (1968). One gram (wet weight) of rat tail tendon was obtained from four to six-month-old Wistar rats for each derivative and for the native sample. The length of deuteration with NaB2H4 was varied in an attempt to cause differences in labelling position and density. Due to the limited time allocated for neutron diffraction, the number of derivative types had to be limited to two. Reaction conditions were therefore 20 milligrams of NaB2H4 in phosphate/ borate-buffered Ringer solution (pH 7"5), to which one gram of rat tail tendon was added. The times of exposure of rat tail tendon to the reducing agent were six minutes and 30 minutes, respectively, Success of cross-link reduction could be judged by observing the lack of solubility of reduced collagen in 0"5 ~-acetic acid. The axial packing of the collagen in reduced samples was shown not to have been altered from the native state, as judged by the X-ray fibre diffraction patterns obtained from the samples. The assumption of isomorphism is therefore valid. Approximately 150 tendon fibres (1 g wet weight) of average individual diameter approximately 200 micrometres were clamped close together on a frame S° as to produce a parallel array of collagen fibrils.
Experiments were conducted at the Institut Laue Langevin, Grenoble. Two diffractometers were used: D1 l, a low-angle diffraction apparatus (Ibel, 1976) and D16, a four-circle stepping diffractometer. For experimentation on D1 l, the frame was sealed in a Teflon cell with quartz windows. For experimentation on D16, the frame was mounted and sealed in an aluminium can. Diffraction experiments were conducted on D l l for lower axial orders of diffraction (1 to 7) and on the diffractometer D16 for the higher orders (5 to 23). Experiments on Dll were carried out with the collagen fibres perpendicular to the beam, and with the axis of the fibres vertical. Two sample-todetector distances were used in order to collect a low-angle dataset. To record the first three orders and the transmitted intensity on the 64 centimetre by 64 centimetre area detector, neutrons of wavelength 0"9 nanometre were selected and the sampleto-detector distance set at 5'6 metres. In order to record orders 2 to 7, the same wavelength neutrons were used and the sample-to-detector distance shortened to 2.53 metres by moving the detector. Collimation consisting of a circular cadmium aperture (diameter l0 mm) was used in order to prevent stray neutrons hitting the sample cell. Transmitted beam intensities were measured for a fixed time interval after attenuation of the beam at source with a graphite moderator. Data were collected for the transmitted beam intensities at a number of 2H20/H20 ratios, to allow the scaling of the intensities between datasets to be determined. The beam stop was removed to record the flux of the through beam. The D l l data were initially corrected for detector response efficiency using the isotropic incoherent scattering of H20. Higher orders of axial reflections were measured on the four-circle diffractometer D16 using 0)/20 scans with the beam axis horizontal and the collagen fibres in the horizontal plane. The experimental parameters used were as in Hulmes et al. (1980). Integrated intensities for each reflection obtained using instrument D11 were determined by using a suite of programs written for FORTRAN 77. It was therefore possible to merge the data from the two different camera lengths by using standard programs, as described by Bradshaw et al. (1989). The intensity of the second and third orders in each D l l dataset were used to determine a scale factor for merging the data. In order to relate these intensities to the true structure factor for each reflection, it was necessary to correct for specimen disorientation, beam divergence, wavelength spread and the intersection of each reflection with the Ewald sphere (Lorenz factor). The magnitude of these corrections was determined experimentally by Hulmes et al. (1980). The corrections were applied to each D l l dataset. D16 data were analysed using programs written also for :FORTRAN 77. This allowed the intensity of each reflection to be estimated by the relative height of each peak. The intensities could then be
Communications Table 1 Rat tail tendon neutron diffraction data Amplitudesf Order
Native
6 min
30 min
Phases:~ (native)
1 2 3 4 5 6 7 8 9 l0 11 12 13 14 15 16 17 18 19 20 21 22 23
10,000 1138 4522 1212 2584 2204 1144 1759 1487 1263 1229 1466 760 889 1045 654 527 1059 949 1109 1096 1265 1030
10,000 1129 4797 1233 2582 1896 1208 1525 1375 1023 1121 1491 830 1050 934 1029 1119 780 687 1084 1156 1124 559
10,000 1077 4359 1086 2767 2578 1680 2283 1893 1354 1563 1620 633 t078 t079 566 704 964 945 1097 874 1131 463
3"37 0"95 0"66 5-65 3.71 0'71 2-52 2 73 4-87 5'03 4'03 3"54 3-97 0"73 3-00 4"30 2"05 4"49 5"05 1-74 0"59 0"45 2-43
Corrected and scaled to a Ist order value of 10,000. :~In radians. corrected for local background scattering and a Lorenz factor applied to each reflection. These d a t a were then scaled to the corrected intensities of D11 d a t a b y using the fifth, sixth and seventh orders common to both datasets. The resulting intensities were all scaled to a first order value of 10,000 and are shown in Table 1. As with the determination of three-dimensional molecular structure using X - r a y diffraction, the method of multiple isomorphous addition also required information relating to the position of label in the unit cell. This information was obtained from two sources, the difference Patterson function resulting from derivative and native diffraction intensities and known amino acid sequence information. Autocorrelation functions of the proposed labelling site were compared with the difference Patterson maps. The difference Patterson functions created showed discrete peaks indicating deuteration at specific sites within the axial unit cell. The n u m b e r of peaks was judged to be sufficiently small to allow the determination of deuteron label position and hence the phases. The position and relative strengths of these peaks were used to construct and modify models of the labelling sites within the molecule. Initial autocorrelations used only the crosslinkage sites at the telopeptides for positions of deuteration. Compared to the difference P a t t e r s o n maps, this produced a map with insufficient peaks. The possibility of other sites being d e u t e r a t e d was investigated. Many molecular species have been isolated, from collagen, t h a t can be reduced b y NaBH4. These include the cross-linkage groups in
3
various forms, hexitol lysine residues and the processed precursors of cross-links, the aldehyde derivatives of lysine and hydroxylysine (Bailey et al., 1974; Robins, 1982). I t is possible t h a t these components can be deuterated in r a t tail tendon and therefore contribute to the difference in neutronscattering between native and derivative samples. A number of trial distributions of deuteration in the axial unit cell were tested. These distributions were dictated by the position of lysine and h y d r o x y lysine residues in a D repeat. The autocorrelation functions tested began to correspond well with the difference P a t t e r s o n function produced when hydroxylysine residues were used as sites of deuteration. P r e d o m i n a n t deuteration sites corresponding to lysine residues in the gap were also postulated (Lys434 alpha l, Lys885 alpha 1, alpha 2), since this significantly improved the match between the autocorrelation and difference Patterson function. The combined contribution of deuteration at the teiopeptide regions, hydroxylysine-rich regions and a lysine-rich region in the gap provided an initial map of label only. The significant differences in the six minute and 30 minute NaB2H4-1abelled derivative difference Patterson maps could be produced by altering the relative occupancy at each labelling site. This gave a starting point t h a t allowed phase determination. The process of phase determination and refinement was the same as t h a t used b y Bradshaw et al. (1989). An iterative series of cyclic calculations used the intensity of native and derivative neutron diffraction peaks to determine the phase c o m p o n e n t of each structure factor for both native and derivative tendon samples. The relative intensity of diffraction from native and derivative structures used values initially derived from transmission data. However, alterations of up to 15~/o of the scaling factors were used to optimize the refinement. This process was continued until the autocorrelation of the label in the neutron-scattering density profile showed good agreement with the difference Patterson map, taking into account the differences due to protein-label vectors in the difference P a t t e r s o n map. The neutron-scattering profile of native rat tail tendon (Fig. 1) has been determined to a resolution of ±0"043 D (the best to date). I t shows general agreement with t h a t deduced by model building (Hulmes et al., 1980). I t has a gap to overlap ratio of 0"49 : 0"51, which compares well with the most accurate value given by X - r a y diffraction (Bradshaw et al., 1989). The profile shows a n u m b e r of peaks and troughs superimposed on the g a p - o v e r l a p function. These features show some correlation with the banding pattern of positively stained electron micrographs, b u t the fundamental differences between the imaging characteristics of these two techniques mean t h a t profiles obtained b y them are not directly comparable. T h e profile presented here has telopeptide regions t h a t appear as peaks a t the end of each gap region. The C-terminal peak is
4
Wess et al.
T.J.
6 min
PI
v
over,°°
¢¢
0 X 0 t¢-
I
0
0.5 D
g
I.O
Figure 1. The neutron-scattering profile of rat tail tendon collagen projected onto the fibre axis. The horizontal scale is distance through the unit cell along the fibre axis and the vertical scale is neutron-scattering density/unit axial translation in arbitrary units. N and C denote N and C termini. higher than that at the N terminus. Although there will be underlying differences in scattering projected onto this region by the other segments of the axial structure, it could also be related to the difference in structure of the C and N-terminal telopeptides. The difference in the axial translation of residues in the two telopeptides as proposed by Hulmes et al. (1980) may contribute to the size of the peaks, since it was reported that the C-terminal telopeptide is more contracted per amino acid translation than is the N-terminal telopeptide. The difference profiles (Fig. 2) show a series of discrete peaks that can be related to the axial distribution of the reducible components in collagen. Two of the main peaks are situated approximately 0"4 D apart. The position of these peaks corresponds to the telopeptide positions in the native profile. Chemical analysis has revealed that cross-linking occurs in these regions (Light & Bailey, 1979), suggesting that either the cross-links or their precursors are being labelled here. This region of the profile also contains the majority of hydroxylysine residues, which may have been converted to hydroxyallysine by deamination and therefore are reducible. This creates a problem, for the cross-linking at the N terminus may be masked by other hydroxyallysine residues in this region. The resolution of the profile is sufficient to allow the separation of the cross-link(s) from a group of hydroxylysine residues in the gap region before the telopeptide. These latter residues appear to be reduced only after 30 minutes of exposure. This may be due to one of two causes: either steric blocking of the labelling in this region, which is unlikely, or the intrinsically different rates of chemical labelling for cross-links and precursor groups. The large number of reducible components in the N-terminal region of the profile may indicate that more cross-links exist than had been previously thought.
X~
J,elopept,de
8m
o--o--
Lys434 z
L y s 8 ~ 1~---~ ~ - ~ C telopeptide / , ~ r 0.1/0:2 0:3 0.4 0:5 0.6 0:7 0:8 0:9 1.0 o
alpha I chains Figure 2. The location of reducible sites along the fibre axis of the rat tail tendon collagen fibril unit cell, as revealed by subtraction of native fibril neutron-scattering density from derivative fibril neutron-scattering density; phased by isomorphous addition of deuterium. The axes are as Fig. 1. The lowest part of the Figure shows positive staining bands found in the collagen D repeat by electron microscopy. These correspond to parts of the structure containing positive charge and therefore lysine residues. (O) hydroxylysine; (0) lysine residues thought to be involved in cross-links.
Around the C terminus there are fewer reducible components. The telopeptide lysine residues in this region correspond laterally with three hydroxylysine residues (Hyl87 alpha 1 × 2, alpha 2). These residues probably comprise the C-terminal crosslinks, since the relative difference in peaks height after six and 30 minutes of labelling is not great. This implies deuteration of cross-links at-this position, since the deuteration appears to occur at a rate similar to the labelling of N-terminal cross-link(s). The size of the difference peak corresponding to both the C and N termini should be similar, since both contain the same reducible components. The difference in peak size may be due to the degree of contraction of the C-terminal telopeptide, which may have a differential effect on the accessibility of the two regions to the reducing compound, with. slower diffusion to the more compact C terminus. It can be seen also that both telopeptide regions are laterally aligned with hydroxylysine residues in both the alpha 1 and alpha 2 chains of collagen, The
Communications
similarity of these two regions indicates t h a t hydroxylation of lysine m a y be essential for crosslinking in this region. The maps also contain a series of peaks t h a t cannot readily be related to known cross-links. These are more predominant in the 30 minute map than the six minute map, and appear to be larger when they occur in the less stericaily restrictive gap region. The large peak in the gap region may result from the non-enzymic glycosylation of lysine residues, which would enable them to be reduced by NaB2Hd. The discrete nature of this peak indicates that, if it is caused by deuteration of a sugar linkage, then the glycosylation process may be more specific than is to be expected from the observations made by Le Pape et al. (1984), where non-enzymic glucosylation is found to be distributed t h r o u g h o u t the collagen molecule in diabetic tissue. The position of the peak corresponds to the axial position of lysine residues 434 of the alpha 1 chain (numbering does not include the telopeptides). The other peak t h a t appears in the gap region after 30 minutes of labelling corresponds to the lysine residues at position 855 in both alpha 1 and alpha 2 sequences. This is the first experimentally determined phasing of the meridional neutron diffraction pattern of collagen and has provided identification of the sites t h a t have been modified in an enzymic or non-enzymic manner to produce reactive groups t h a t can be labelled by specific deuteration. The authors would like to thank all the staff at the Institut Laue Langevin, especially the local contact Peter Timmins, and also Kevin Duff for technical assistance. References
Bailey, A. J. & Peach, C. M. (1968) Biochem. Biophys. Res. Commun. 33, 812-819. Bailey, A. J. & Skimokamaki, M. (1971). F E B S Letters, 16, 86-88. Bailey, A. J., Robins, S. P. & Balian, G. (1974). Nature (London), 251, 105-109.
5
Bailey, A. J., Light, N. D. & Atkins, E. D. T. (1980). Nature (London), 288, 408-413. Barnard, K., Light, N. D., Sims, T. J. & Bailey, A. J. (1987). Biochem. J. 244, 303-309. Bradshaw, J. P., Miller, A. & Wess, T. J. (1989). J. Mol. Biol. 205, 685-694. Chapman, J. A. & Hulmes, D. J. S. (1984). In Ultrastructure of the Connective Tissue Matrix (Ruggeri, A. & Motta, P.M., eds), pp. 1-33, Martinus Nijhoff, Amsterdam. Eyre, D. R. (1987). Methods Enzymol. 144, 115-140. Eyre, D. R., Paz, M. A. & Gallop, P. M. (1984). Annu. Rev. Biochem. 53, 717-748. Fraser, R. D. B., MacRae, T. P., Miller, A. & Suzuki,, E. (1983). J. Mol. Biol. 167, 497-521. Fraser, R. D. B., MacRae, T. P. & Miller, A. (1987). J. Mol. Biol. 193, 115-125. Fujimoto, D. (1980). Biochem. Biophys. Res. Coramun. 93, 948-953. Galloway, D. (1982). In Collagen in Health and Disease (Weiss, J. B. & Jayson, M. I. V., eds), pp. 528-557, Churchill Livingstone, Edinburgh. Hulmes, D. J. S., Miller, A., White, S. W., Timmins, P. A. & Berthet-Colominas, C. (1980). Int. J. Biol. Macromol. 2, 338-345. Ibel, K. (1976). J. Appl. Crystallogr. 9, 630-643. Le Pape, A., Guitton, J. D. & Muh, J.-P. (1984). F E B S Letters, 170, 23-27. Light, N.. D. & Bailey, A. J. (1979). F E B S Letters, 97, 183-188. Miller, A. (1982). Trends Biochem. Sci. 7, 13-18. Miller, A. & Wray, J. S. (1971). Nature (London), 230, 437-438. Nakamura, Y. (1987). Int. J. Biol. Macromol. 9, 281-290. North, A. C. T., Cowan, P. M. & Randall, J. T. (1954). Nature (London), 174, 1142-1143. Robins, S. P. (1982). Methods Biochem. Anal. 28, 329-379. Robins, S. P. & Bailey, A. J. (1975). Biochem. J. 149, 381-385. Robins, S. P. & Bailey, A. J. (1977). Biochem. J. 163, 339-346. Yamauchi, M., Katz, P. Z. & Mechanic, G. L. (1986). Biochemistry, 25, 4907-4913. Yamauchi, M., London, R. E., Guenat, C., Hashimoto, F. & Mechanic, G.L. (1987). J. Biol. Chem. 252, 11428-I1434.
Edited by V. Luzzati
