J. Mol. BioZ. (1991) 220, 111-123

Morphology of Sheet-like Assemblies of pN-collagen, PCcollagen and Procollagen Studied by Scanning Transmission Electron Microscopy Mass Measurements David F. Holmes-f, A. Paul Mould and John A. Chapman Departments of Medical Biophysics and Biochemistry/Molecular Biology University of Manchester, Oxford Road, Manchester Ml3 9PT, England (Received 12 November 1990; accepted 14 March 1991) At high concentrations, type I pN-collagen, PC-collagen and procollagen (the first 2 generated from procollagen by enzymic cleavage of C-propeptides and N-propeptides, respectively) can all be made to assemble in vitro into thin D-periodic sheets or tapes. Scanning transmission electron microscopy mass measurements show that the pN-collagen sheets and procollagen tapes have a mass per unit area corresponding to that of approximately 68 monolayers of close-packed molecules. PIN-collagen sheets are extensive and remarkably uniform in mass thickness (fractional S.D. 9035); procollagen tapes are neither as extensive nor as uniform in thickness. The mean thickness of PC-collagen tapes is less and the variability is greater. In pN-collagen sheets, the overlap : gap mass contrast in a D-period is increased from 5 : 4 (the ratio in a native collagen fibril) to 6 : 4, showing that the N-propeptides do not project into the gap but are folded back over the overlap zone. Assuming all N-propeptides to be constrained to the two surfaces of a sheet, their surface density can be found from the mass thickness of the sheet. In a lateral direction (i.e. normal to the axial direction where the spacing is D-periodic), the N-propeptide domains are calculated to be spaced, centre to centre, by 2.23( -fi@l) nm on both surfaces. This value (approx. 1.5 x the triple-helix diameter) implies close-packing laterally with adjacent domains in contact. Sheet formation and the “surface-seeking” behaviour of propeptides can be understood in terms of the dual character of the molecules, evident from solubility data, with propeptides possessing interaction properties very different from those displayed by the rest of the molecule. The form and stability of sheets (and of first-formed fibrils assembling in viva) could, it is suggested, depend on the partially fluid-like nature of lateral contacts between collagen molecules.

Keywords: collagen;

procollagen;

self-assembly; sheet-like STEM mass measurements

1. Introduction

to whom

all

correspondence

should

assemblies;

this control is absent and only broad distributions of diameters are observed (Bard & Chapman, 1973). Although collagen fibrillogenesis in vitro is a classic example of an entropy-driven self-assembly process in which the information required for the formation of D-periodic fibrillar structures is intrinsic to the collagen molecule (Kadler et al., 1987), it is evident that additional cell-mediated factors must be superimposed in vivo to regulate lateral accretion and hence fibril size. Current evidence points to the participation of propeptides in this regulatory process. Fibrillogenesis of type I collagen in viwo is accompanied by the enzymic removal of the N and C-terminal propeptide domains from procollagen,

Regulation of size and form is a central issue in developmental biology. This regulation is evident in the growth of collagen fibrils in who. It is most apparent in embryonic and neonatal connective tissues where the cylindrical fibrils tend to be uniform in diameter over extended populations (Parry & Craig, 1984), implying that lateral growth is in some way synchronized throughout the population. When collagen monomers (without propeptides) are made to reconstitute into fibrils in vitro, ‘f Author addressed.

macromolecular

be

111 002%%36/91/130111-13

$03.00/O

0

1991 Academic

Press Limited

112

D. F. Holmes

the biosynthetic precursor of collagen (Tanzawa et al., 1985; Hojima et al.; 1985; Kessler et al., 1986). Immunolocalization studies by Fleischmajer et al. (1983, 1985, 1987a,b) show that N-propeptides are associated with the surface of small-diameter collagen fibrils in developing connective tissue (C-propeptides have been similarly localized on the surface of larger diameter fibrils). Such observations have led to the proposal that uncleaved N-propeptides are confined to the fibril surface where they block accretion of molecules; further lateral growth requires enzymic cleavage of the propeptides. In t’his way, growth regulation is cellmediated through the release of N-proteinase. This proposal forms the basis of two models suggesting possible mechanisms for regulating collagen fibril diameters in wivo (Hulmes, 1983; Chapman, 1989). Both seek to account for the quantization of fibril diameters reported in a wide range of connective tissues (Parry & Craig, 1979; Eikenberry et al., 1982a). The Hulmes (1983) model depends on a quasi-crystalline packing of molecules; the other calls for no specific lateral packing and treats lateral contacts between non-cross-linked molecules as having a partially fluid-like character (Chapman: 1989). Both models recognize that fibrils growing in uivo are constrained, by some as yet unknown mechanism, to circular cross-sectional outlines. The separat,ion and purificat,ion of procollagen from cultured fibroblasts and the isolation of N and C-proteinases, the specific enzymes that cleave the S and C-propeptides, have led to the development of a cell-free system in which the assembly of partially and fully processed procollagen can be studied (Kadler et al., 1987, 1988). This has allowed the possible regulatory roles of the propeptides and their cleavage enzymes to be investigated in vitro. These cell-free studies have shown that persistence of the propeptides on a proportion of the collagen molecules can profoundly affect the morphology of the D-periodic assemblies (Hulmes et al., 1989a,b). Fully processed collagen assembles intro D-periodic fibrils, near-circular in cross-section, and never, to our knowledge, into thin sheets. Tn complete contrast, the studies of Hulmes et al. (1989a,b) show that pN-collagent forms distinctive sheet-like assemblies, also D-periodic; mixt,ures of fully processed collagen and pN-collagen give D-periodic assemblies intermediate in form, noncircular and multi-lobed in cross section and with perimeter : area ratios increasing with the proportion of pN-collagen. The presence of the N-propeptides has the effect, therefore, of increasing the surface area of a hybrid fibril. These and other that, observations support the idea the t Abbreviat,ions used: pn’-collagen and pC-collagen. intermediates in the conversion of procollagen to collagen containing, respectively, the K-propeptide domain only and the C-propeptide domain only; STEM. scanning transmission electron microscopy; TEM, transmission electron micropseopy; SLS, segment long spacing.

et al.

-

N-propeptides are preferentially located on rhe surface of the assembly where they regulate !aterat growth (Hulmes, 1983; Fleischmajer & al.; 198iF: Chapman, 1989). Unlike fibrils growing in oivo: however, pN-collagen-containing assemblies in t,he cell-free system a,re not constrained to circular out’lines

during

growth.

At sufficiently high concentrations, procollagen (Mould et aZ., 1990) and pC-collagen (Hulmes et nl., 19896) also form D-periodic assemblies; most take the

form

of narrow

tapes

rather

than

Dhe extensive

sheets formed from pN-collagen. Estimates of t’he thickness of the sheets and tapes have been made by conventional TEM on embedded and sectioned pellets of centrifuged specimens. The pN-collagen sheets appear uniform in thickness wit’h a mean thickness (after dehydration, embedding and sectioning) of 7.9(* 1.3) nm (s.D., L! = 104): i.e. the fractional standard deviation is approximately 16% (Hulmes et al., 1989a). Comparable values are report’ed for prorollagen t,apes (Mould et al.. 1990). Quantit’ative mass measurements by STEM now make it possible to obtain accurate quant’itative data about mass t,hicknesses and mass distributions in suitable macromolecular assemblies. The sheets and tapes formed in t,he cell-free procollagenprocessing system are particularly suited to this approach. In the work described here, we have used a conventional hot-filament instrument with a STEM attachment, modified to provide image a,nalysis and other facilities. This has been used to measure the mass per unit area of the three types of D-periodic assembly (p;“\‘-collagen, procollagen and pC-collagen). Preliminary data are also presented on the axial distribution of mass wit,hin a D-period of a pN-collagen sheet.

(a) Preparation

of ~wbstrutes

Type 1 procollagen was purified from suspefisioi; cuitures of chick embryo tendon fibroblasts as described (Mould et al.. 1990). Type I pC-collagen was purified from overnight cultures of chick embryo tendon fibroblasts by precipitation with ammonium sulphate and ion-exchange chromatography (Hoffman et ni., 1976; Jlould et al., 1990). Type I procollagen (I-prot~einase (for the preparation of pK-collagen) was purified from organ cultures of chick embryo tendons (Hojima, et al.. 1985). (b)

Porrrlatiorb

uf D-periodic

sheets and tupen

Procollagen (3.5 mg/ml) was incubated in phosphatebuffered saline for 24 h at 37°C. D-periodic structures were separated from other aggregates (Mould et al.. 1990) and resuspended in phosphate-buffered saline. pN-collagen “sheets” were formed as described by Hulmes et al. (1989a). except that chick procollagen was used as the substrate. Procollagen (360 pg/ml) was incubated with 40 units of C-proteinase/ml for 24 h at, 37Y’. PC-collagen (1 mg/ml) was incubated in approx. ti6 x standard st,rength phosphate-buffered saline for 24 h at 37°C. (Low ionic strength was found to favour the D-periodic, assembly of pC-collagen.) Tn all cases. t,he

Collagen

final D-periodic structures were diluted with the buffer used for assembly, or with 0.2 M-ammonium acetate, to a concentration suitable for application to carbon-coated grids and giving low background contamination.

(c) Electron (i) Use of a conventional

microscopy

instrument

in the STEM

mode

The basic instrument was a JEOL 1200EX transmission electron microscope equipped with a JEOL ASIDlO scanning unit and the standard bright field and annular dark field detectors, both of the scintillator/ photomultiplier type. The STEM attachment was subsequently modified by interfacing it with a microcomputer to provide optimal facilities for quantitative operation. The modifications included digital scanning, digital data acquisition and remote control of electron beam blanking and image shift. The digital system was based on an Amstrad PC1512 fitted with a Citadel PC-30 data acquisition card. The latter contained 2 12.bit D/A converters used for scan raster generation, an &bit A/D converter used for digitizing the detector signal and an &bit D/A converter to provide a signal for the hard-copy device. The interface link with the STEM included provision for integration and scaling of the detector signals. The interface device and control software were constructed to detailed specifications by Tom Brown Developments, Bury, Lanes. Scan formats could be set up to a maximum of 1024 x 1024 points and the time per point could be set from 180 ~LS upwards. The standard image consisted of 512 x 512 points with 200 ~LS per point giving an image acquisition time of 50 s. Image data were transferred online to a Matrox PIP1024 framestore in an Olivetti M28 microcomputer for display and subsequent analysis (see below). Rigorous testing indicated no loss of image quality (by distortion or loss of resolution) in the digital mode compared with the internal analogue mode. Linearity of the dark field detector response was established for the relevant range of gain settings and incident electron beam intensities. The STEM was operated at 120 kV with the standard lens settings. The collection angle of the annular dark field detector ranged from 25 x 10m3 to 75 x 10m3 radians; the effective camera length in the STEM mode was determined from the diffraction pattern of an evaporated nluminium film. The signal from the annular dark field detector was linearly dependent on carbon film mass thickness up to approx. 50 kDa nrn-‘.

(ii) Sample

preparation

Carbon films were prepared by evaporation onto freshly cleaved mica, using carbon fibre (Balzers) as a source; the coating unit was turbomolecular pumped (Nanotech). Carbon film thicknesses were typically 2 to 2.5 nm, as measured by electron scattering in STEM. The films were float’ed onto a clean water surface and collected on 600mesh copper grids (Gilder Grids), previously ultrasonically cleaned for 15 min in acetone. The carbon films were used freshly prepared without further conditioning steps. A drop of the sample was left on the filmed grid for approx. 15 s; flushed with 4 or 5 drops of buffer to remove excess unaggregated collagen, then flushed with approx. 6 drops of water and dried in air. All water used for the flotation of carbon films and for washing specimens was from a commercial ultra pure water ion-exchange system (Still Plus HP model from Purite).

-

Sheets

(iii)

113

Mass

measurements

Techniques used for mass measurements were similar to those developed for a dedicated STEM instrument with a field emission gun (Engel, 1978, 1982; Engel et al., 1981; Engel & Reiehelt, 1984; Freeman & Leonard, 1981). Image analysis was implemented on an Olivetti M28 (AT compatible) fitted with a Matrox PIP1024 framestore and co-processor. Specific routines for mass measurement were constructed within the the MicroSemper software package (Synoptics, Cambridge). Micrographs used for mass per unit area measurements were taken using a medium spot size (approx. 3 nm) and exposing the specimen to a low electron dose A diffraction grating replica (< lo2 e nme2). (2160 lines/mm) was used for magnification calibration; this calibration was estimated to be accurate to better than 2%. At the instrumental magnification set’ting of 25,000 x used here, the pixel size was 8.5 nm. Specimens were at room temperature during electron microscopy. All mass measurements described here are based on the use of tobacco mosaic virus (TMV) as a mass standard (mass per unit length = 131 kDa nm-‘; Kaper, 1986). The probe current was stable to approx. 1 T/, over 15 min intervals and the relative intensity of the probe could be adequately monitored between dark field images using the bright field detector with the probe over a hole in the carbon film. Grids with deposited TMV particles could then be used as external standards to check the system response (and, if necessary; to correct the calibration). This procedure proved to be more accurate than one in which fd phage was mixed with the sample to act as an internal mass standard. For the mass per unit length measurements, the relative S.D. was approx. 4% on TMV and approx. 7% on fd phage, in agreement with the data of Freeman & Leonard (1981); the measured ratio of mass per unit length of TMV to that of fd phage was 6.5, again in agreement with data from dedicated STEM systems.

3. Results (a)

Morphology

of aggregates

All three samples gave rise to D-periodic sheetlike or tape-like aggregates. The pN-collagen sheets were the most striking in appearance by virtue of their size and uniformity (Fig. l(a)). Many were highly extended, laterally and axially; widths up to approximately 20 pm and axial lengths up to approximately 50 pm were observed. Gap/overlap zoning was sharply defined. Conventional electron micrographs of the sheets after negative staining and

in

section

have

been

presented

(Hulmes

et al.,

1989a,b). STEM images of the procollagen tapes (Fig. l(b)) displayed the same morphological features as those observed after conventional electron microscopy (Mould et al.: 1990). Tapes were variable in width (approx. 0.05 to 1 pm) with some regions flat, others twisted

and

folded.

The

D-periodic

variation

in elec-

tron scattering in STEM micrographs of procollagen was characterized by strong scattering of rather illdefined outline from one region in each D-period. The conventional TEM studies reported by Mould et al. (1990) showed, in negatively stained procollagen

114

D. F. Holmes

et al.

micrographs of typical D-periodic assemblies fro-m (a) pK-collagen, showing Figure 1. STEM dark-field extended sheet, (b) procollagen (some filamentous fd phage is present in this sample as an internal mass and (cl) PC-collagen. In (c) the p&collagen occurs as tape-like assemblies with multiple folding. The narrow edges of the assemblies (indicated by an arrow) are typical of those selected as non-overlapping regions for area measurements. (d) A pC-collagen assembly taking the form of a very thin and laterally extended sheet sub-structure suggestive of a looser packing of molecules. The D-period is 68 nm (65 nm in (d)).

part oi an standard), (c) regions at the mass per unit with a diffuse

Sheets

Collagen

the extended tapes or sheets to the carbon film substrate before dehydration (see section (d), below). The more loosely packed sheet form of PC-collagen did show a slightly reduced D-period, with a mean value 4.5% lower than that for the tape form (Table 1). Shrinkage in directions normal to the electron probe is important in STEM mass determinations because it will affect the apparent mass per unit area of an assembly (see section (d), below).

tapes, a prominent stain-excluding region, also illdefined in outline. End-on views revealed globular projections, located in the vicinity of the C-terminal ends of the molecules. After positive staining, pronounced additional uptake of stain occurred in the same region (between the a, and b, bands). This region, the most conspicuous feature in a procollagen D-period, was tentatively identified by immunolabelling studies as the region in which the C-propeptide domains were located. It seems likely, therefore, that the strongly scattering region in STEM images of procollagen tapes also arises from the C-propeptides. The PC-collagen sheets were the most heterogeneous. The morphology ranged from compact aggregates of tapes showing multiple folding and superposition (Fig. l(c)) to extended diffuse, sheetlike structures, seemingly much more loosely packed (Fig. l(d)). Measurements on these pC-collagen assemblies were separated into two groups, those made on the narrow unfolded regions at the edges of the more compact tapes (see the arrow in Fig. l(c)) and those made on the extended diffuse sheets. As with procollagen tapes, the most conspicuous feature in a D-period was an ill-defined region exhibiting strong electron scattering, presumably from the C-propeptides. (b) Measurement

of the

115

(c)

Mass

per unit

area

measurements

Typical histograms of mass per unit area data are shown in Figure 2 for the three forms of collagen. Means and standard deviations are presented in the right-hand columns of Table 1. The relative spread of values about the mean (i.e. the fractional S.D., here expressed as a percentage of the mean) increases in the order pN-collagen, procollagen and PC-collagen compact tapes, PC-collagen diffuse sheets. With pN-collagen, where sheets were large enough for multiple measurements to be made on the same sheet, there were no significant differences between data obtained from intra-sheet measurements and those from inter-sheet measurements. Negatively stained or sectioned pN-collagen sheets imaged by conventional TEM appear remarkably uniform in thickness (Hulmes et al., 1989a). This uniformity is amply confirmed by STEM mass measurements. Whereas Hulmes and co-workers obtained (from dehydrated, embedded and sectioned material) fractional S.D. values of approximately 16% in their thickness measurements, direct mass per unit area determinations have a fractional S.D. of only 3.5%. The much smaller spread in the measurements also demonstrates the far greater precision attainable in the STEM-based procedure. The re-expression of the mass per unit area (4.29 kDa nrn-’ for pN-collagen sheets) as a geometric thickness will be considered in the Discussion. For the moment we consider only how the mass per unit area results of Table 1 can be expressed in terms of the “molecular thickness”, i.e. the number, N,, of molecules that make up the sheet

D-period

D-period measurements are summarized in the left-hand columns of Table 1. Unfixed and unstained pN-collagen sheets, procollagen tapes and the compact tape form of pC-collagen all exhibit, within experimental error, the same D-periodicity in the STEM images; mean values are close to 68.3 nm (with S.D. values in the range 0.5 to 08 nm). This result can be compared with that from X-ray diffraction data; meridional reflections from native tendon fibrils maintained in a moist state reveal an axial D-periodicity of 67 to 68 nm (Miller, 1982, 1985). There is no indication, therefore, that the tape and sheet samples have suffered axial shrinkage during dehydration for electron microscopy. Shrinkage is probably prevented by adherence of

Table 1 D-period

measurements

and

mass

per

unit

area

pC-collagen D-period

data

for

pN-collagen,

procollagen

and

assemblies (mu)

Mass per unit

area (kDa

nr~~~)

Assembly

Mean

S.D.

N

Mean

S.D.

S.D. (%)

INI

pN-Collagen Procollagen pC-Collagen pC-Collagen

68.24 6838 6830 6521

0.55 0.48 084 1.70

23 15 21 10

4.29 564 454 373

0.15 @69 071 1.20

3.5 12.2 156 32.3

69 74 43 27

(tapes) (sheets)

Each D-period measurement was made by averaging 10 to 20 consecutive D-periods and this was repeated at N different locations in the sample (i.e. the number of D-periods contributing to the final average is much greater than N). The mass per unit area measurements were made over sample areas of 40 x 100 nm, apart from those for the pC-collagen tapes, where the restricted width of the regions available for measurement reduced the sample area to 40 x 50 nm.

D. F. Holma

116

et al Tabk 2 Summary of molecular mcnsses jwr type I collagen components used in calculations of the ma8.s per unit area data and in interpreting axial mass distributians Mass Collagen domain

01’ molecule

(kDa)

- .-__

N-propeptide domain (1) C-propeptide domain (2) Collagen molecule (3)

35 100 290

pN-collagen Procollagen pC-collagen

325 425 390

N=69

(1 + 3) (1 + 2 + 3) (2 + 3)

Molecular mass values for the N and C-propeptide domains and the collagen molecule come from data published by: (1) Timpl & Glanville (1981), (2) Olsen (1981), (3) Traub & Piez (1Q’il). The molecular masses of pN-collagen, procollagen and PC’-collagen are derived by the a,ppropriate additions, as indicated.

total mass per unit area equal to that of the sheet. D, the axial periodicity, is 68 nm (65 nm for p&x% lagen sheets, see Table 1) and a: the diameter of the collagen triple helix, is taken to be 1.52 nm (1.5% nm is the mean centre to centre separa,tion between quasi-hexagonally packed molecules in mature rat tail tendon collagen fibrils maintained in a hydrat,ed state (Hulmes & Miller, 1979)). Table 2 shows published molecular mass values for type I colla.gen and for the N and C-propeptides. Values of M for p&-collagen, PC-collagen and procollagen have hem calculated from these molecular masses and are shown in Table 2. Values of the molecular thickness, N,, derived from these parameters and from the observed mass per unit area data, appear in Table 3. The molecular thicknesses of the pN-collagen sheets and the procollagen tapes; both slightly less than seven molecules thick, are near identical (to within O.6o/o). Values for the PC-collagen assemblies are lower.

(b) 15

LO

5

0

I

2

3 Mass

4

5

per umtorea

6

7

a

(kDo/nmz)

(c)

Figure 2. Histograms of mass per unit area (kDa) nm -‘) for STEM measurements made on (a) pN-collagen sheets, (b) procollagen tapes and (c) pc-collagen. (In Fig. 2(c), the stippled area is from diffuse sheets, the filled area from the edges of tape-like assemblies). thickness. N, will be taken to be unity for a sheet consisting of a single monolayer of collagen molecules in D-staggered close-packed array. Such a monolayer would have a mass per unit area of &Z/5 Da (kDa nmP2), where J4 is the molecular mass of a single molecule (in kDa) and a is its diameter (in nm). (Although the molecular length is approx. 45 D, each molecular mass must be regarded as occupying, on average, an area 5 Da of the monolayer.) For a sheet in which the experimentally observed mass per unit area is m,: J, = 5Da(m,/M). Defined in this way, number of hypothetical cular layers that, when

(1)

can be regarded as the close-packed monomolesuperimposed, would have a

IV,

Number, thickness thickness)

Table 3 Xr, of molecular layers contributing to t& QJ” the sheets and tapes {i.e. the molecula.r calculated from th,e mass per unit area data of Table I

Assembly pN-collagen Procollagen pC-collagen @Y-collagen

(tapes) (sheets)

Mean

53.0.

682 686 ml2 494

cl.24 0.84 094 1-m

The molecular thickness, IV{, is the number of close-packed monomolecular layers that would give the same total mass per unit area, m,, as that of the sheet. It is calculated from: Art = !iLla(m,jlW)

(1) where VL, is the measured mass per unit area of the sheet 01’ t,ape in kDa nm -’ (from Table l), M is the molecular mass (in kDa) of t,he component motecule (from Table 2). and n ( = I.52 nm) is the diameter of the collagen triple helix (see Results, section (c)).

Collagen

(d) Estimation

of errors

A small S.D. in the mass per unit area measurements (Table 1) implies a very small error mean. For pN(s.D.IJN) in m,, the calculated collagen sheets: m, = 4.29( fO.02) kDa nm-‘, i.e. an error of approximately @5% due to random errors in measurement. N,, the molecular thickness of a pN-collagen sheet (and b, the mean lateral spacing of N-propeptide domains on the sheet surface, see Discussion) will exhibit similar random percentage errors. Systematic errors can be expected to be greater than this. The magnification calibration has been estimated to be accurate to better than 2% (see Materials and Methods). Another possible source of error is specimen shrinkage. If this occurs in the plane of the sheet and the supporting film, the apparent mass per unit area of an assembly will increase. (Shrinkage in thickness, normal to the supporting film, will not affect mass measurements.) Any shrinkage taking place in the direction of the molecular axes in an assembly can be detected from the change in D-period. For the pN-collagen sheets, procollagen tapes and the more compact PC-collagen tapes prepared in unstained form for STEM studies and examined at low electron probe intensities, the D-period was, within experimental error, identical to the 68 nm D-periodicity in hydrated fibrils (Table 1). Only the more loosely packed PC-collagen sheets showed a small reduction (4.5%) in the D-period. With this one exception, therefore, none of the samples suffered axial shrinkage. We cannot be equally sure from our data that lateral shrinkage (in the plane of the sheet and the supporting film) is also absent. Nevertheless, several considerations point to such shrinkage being small, probably no more than a few percent. (1) Shrinkage would have to take place remarkably uniformly throughout all sheets in the pN-collagen samples to maintain a fractional S.D. of 3.5% in the mass per unit area measurements. (2) Sheets are known to be “sticky”, adhering strongly in the incubating solution to glass and other surfaces; similar adherence to the carbon supporting film is probably responsible for the occurrence of such large areas of undistorted sheet (up to 20 pm in lateral extent in the case of pN-collagen); this extensive contact with the supporting film should stabilize sheets against axial (and presumably lateral) shrinkage in the plane of the film. (3) The electron probe intensity (=$ lo2 e nmw2) used to obtain the quantitative mass data from which sheet parameters (N, and b) were derived was below the level at which significant mass loss (leading to possible beam-induced shrinkage effects) is detectable (Engel, 1982). (e) Intraperiod

mass

The characteristic collagen fibrils arises of molecules and between D and the

contrast

in pN-collagen

sheets

overlap/gap structure of native from the regular D-staggering the non-integral relationship molecular length, L (Hodge &

Sheets

117

“1

Overlap

O

GQP

I.0

0.5 Axial

distance

I.5

0

(D-periods)

Figure 3. The STEM-derived axial distribution of mass in 2 D-periods of a pN-collagen sheet. The distribution has been averaged from approx. 25 scans, each of axial length 1 D-period (68 nm) and height 100 nm. The horizonal broken lines show the maximum and minimum values used to derive the ratio of 6 : 4 for the overlap : gap mass contrast.

Petruska, 1963). As L N 4.5 D (for a fully processed collagen molecule), overlap zones and gap zones are roughly comparable in axial extent, with five molecules traversing an overlap zone for every four traversing adjacent gap zones. In a single D-period, therefore, the distribution of mass along the direction of the fibril axis (the axial mass distribution) is, to a first approximation, a rectangular step function in which the ratio of the two step heights is approximately 5 : 4. STEM images of unstained sheets and tapes formed from pN-collagen, procollagen and pc-collagen all display strong contrast in their D-periodic banding (Fig. 1). Figure 3 shows at low resolution, the STEM-derived axial distribution of mass (averaged from many observations) in two D-periods of a pN-collagen sheet. In each D-period, two fairly sharply defined zones of different mass thickness are present. The contrast ratio (the ratio of the mean amplitude in the high mass zone to that in the low mass zone) is about 6 :4. STEM and conventional TEM observations on pN-collagen sheets weakly stained with heavy metal ions (data not shown) confirm that the axial positions of the high mass and low mass zones are roughly those of the overlap and gap zones of native collagen fibrils; the high mass zone extends over the c,, b,, b, and ad-a1 staining bands and the low mass zone over the e,; e,, d and c2 bands. Higher-resolution data on axial mass distributions in pN-collagen sheets and procollagen and PC-collagen tapes will be presented in a later publication. The mass contrast ratio of approximately 6 : 4 per D-period in pN-collagen sheets implies that the N-propeptide domain does not extend into the gap zone but must be folded back so that it contributes most, if not all, of its mass to the overlap zone. In native collagen fibrils, the “gaps” in the gap zone

118

D. F. Holmes C

N

C

N

PN c-

C

PN c=-

!a) C

c-

Cbi

4 L QVWlWl Gap Figure 4. A representation of the axial relationships between molecules in (a) a collagen fibril (based on Hodge 1963), and (b) a pPj-collagen sheet. & Petruska, Collagencollagen interactions cause molecules to associate in staggered array; each molecule axially displaced wit,h respect to its neighbours by D (68 nm) or integral multiples of D. 4s the molecular length, L, is approximately 450, each D-period comprises an overlap zone and a gap zone of roughly comparable axial extent. In collagen fibrils, for every 5 molecules in an overlap zone, 4 will traverse a gap zone, giving a 5:4 overlap:gap mass ratio. The observed mass contrast ratio of approx 6 : 4 in pR--collagen sheets suggests that the conformation of the

Kpropeptide

domain

approximates

more

folding-back of the R’-propeptides over the collagen molecule, as indicated in (b), rather extension of the ?;-propeptides into the gap.

closely

to a

end of the than to an

occur because the molecules are approximately 4.5 D long, leeving spaces of axial extent 0.5 D. The gaps correspond, therefore, to the absence of roughly l/9 of the mass of a collagen molecule. Using the molecular mass of collagen given in Table 2, this “absent mass” is approximately 32 kDa. Also from Table 2: the molecular mass of the N-propeptide domain is approximately 35 kDa. Extension of this additional mass into the gap zone would largely eliminate the overlap: gap mass contrast in each D-period. On the other hand, a “folded-back” conformation of the N-propeptide domain, adding its mass to the overlap zone, would, as indicated in Figure 4, increase the contrast ratio from approximately 5 : 4 to approximately 6 : 4, the observed value. The axial extent of the N-propeptide domain is uncertain but available data indicate that its N-terminal “Co1 1” subdomain and its triple-helical sub-domain are both about 14 nm long (Engel et rcl., 1977; Timpl & Glanville, 19Sl), giving an over-all axial length similar to that of an overlap zone.

4. Discussion (a) pN-collagen sheets Two new observations stand out in our results. The first is that, although N and C-propeptides both

et al.

have the effect of limiting lateral aeeretion in vitro and bringing about the assembly of sheet or tapelike structures, this limitation is much more precise when accretion is restricted by the presence of X-propeptides. pN-collagen sheets are remarkably uniform in mass per unit area (and hence, presumably, in thickness). This uniformity is less when both N and C-propeptides are present (in procoliagen tapes) and is reduced further when snly C-propeptides are present (in pC-collagen tapes). The second observation is that an N-propeptide domain in a sheet appears to have a fairly welldefined conformational relationship with respect to its parent molecule, a relationship in which the domain is folded back over the end of the parent molecule and contributes mass to the overlap zone rather than projecting into the gap. It is difficult, to see how the intra-period mass contrast In a nK-collagen sheet can otherwise be explained. A partial folding back of t’he S-propeptides has been proposed by Morris et al. (1979) to account for t’he apparent resistance of the t’riple-helical region of the N-propeptide domain to att,aek by bacterial coliagenase. Others from observations on abnormal collagen fibrils in dermatosparactie tissues. have suggest,ed t’hat) the K-propeptides are partly in the gap zone (Fjolstad & Helle, 1974; Mosler et al.. 1986); linkage of the N-propeptide chains to the collagen triple helix ma,y. however, be incomplete in these fibrils. Conventional TEIM provides strong evidence for a folded-back conformation. Rotary shadow-ing of individual molecules of procollagen and pX-collagen has demonstrated the presence of a “kink” (a flexible region) at a position corresponding to the junct’ion of the collagen triple helix with t,he N-propeptide domain (Bichinger et al., 1982; Hofmann et al., 1984). Mould & Hulmes (1937) report that a kink is not always observed but, instead, the shadowing reveals a short thickening at this end of the molecule, suggesting that the N-propeptide domain is folded back along the collagen triple helix; some 60% of molecules in freeze-dried preparations have this appearance. Other evidence comes from immunogold labelling of N-propeptides in procollagen assemblies (Mould et nE., 1990); subsequent staining and electron microscopy reveal a fairly broad distribution of tbe label an a D-period (the resolution of the method is only 20 nm) but the maximum is clearly located on the overlap side of the gap/overlap junction. (Early TEM studies, however, gave no hint of a foldedback conformation, with N-propeptides fully extended in p&I-collagen SLS (segment long spacing) aggregates (Dehm et al., 4972; Veis et al., 1973); SLS formation requires ATP ions and a low pH, creating a, non-physiological electrostatic environment (Bowden et al., 1968); which could give rise to an abnormal propeptide conformation .) Secondary structure predictions from sequence data point to a “hairpin-loop” conformation of the N-t’elopeptide domain with a ,&turn occurring at t,he central lysyl residue (the cross-linking site) in the

Collagen

telopeptide sequence (Helseth et al., 1979; Jones & Miller, 1987; Dombrowski & Prockop, 1988). This turn could be that involved in the folded-back conformation of the N-propeptide domain. N-propeptides, it was noted earlier, are thought to regulate fibril diameters in vivo because they have properties that constrain them to the surface of a growing assembly where they restrict further lateral growth (Hulmes, 1983; Fleischmajer et al., 1987b; Chapman, 1989; Hulmes et aZ., 1989a). Regulation is possible, it is claimed, because accretion will everywhere be blocked when the surface density of N-propeptides is sufficiently high. This prevents further growth until the propeptide chains have been enzymically cleaved, allowing cellmediated control of fibril growth via enzyme release. The folded-back conformation of an N-propeptide domain, causing it to be positioned where it is most likely to impede further accretion, is consistent with this view. The surface-seeking behaviour of propeptide domains can also account for sheet formation. When all assembling molecules have an intact propeptide domain, a sheet provides the greatest possible surface area to accommodate the domains (see Discussion, section (c)). The number of domains present on the two surfaces will depend on the thickness of the sheet. A very thin sheet will have a low surface density of domains and vice versa. We now show that an accurate assessment of the surface density of N-propeptide domains on a pN-collagen sheet can be obtained from our mass data. The only assumption made is that all domains are surface-located. In an axial direction, domains are arranged D-periodically (on overlap zones), i.e. their nearestneighbour centre to centre spacing in this direction can only be D, regardless of the sheet thickness. In a lateral direction in the sheet surface (i.e. in a direction normal to the molecular axes in the surface), the mean centre to centre spacing will vary with the thickness of the sheet, and hence with m,, its mass per unit area. It is convenient to refer to this mean lateral spacing of the domains, determined from m,, as b (nm). Imagine a pN-collagen sheet subdivided into rectangular blocks, each of axial length D, lateral width b, and each extending through the full thickness of the sheet. A single block will have, on average, one N-propeptide domain on each outer surface, i.e. two N-propeptide domains in all. Since each N-propeptide domain is associated with one pN-collagen molecule, the total mass in each block will, on average, be ZM, where M is the mass of a single pN-collagen molecule. It follows that the mass per unit area, m,, will be 2MjDb. Hence: b = ZMIDm,

(nm).

(2) Using values for m, and M quoted previously, this lateral spacing of the N-propeptide domains on the sheet surface is calculated to be b = 2.23( f0.1) nm, i.e. about 1.5 times the diameter (a = 1.52 nm) of the triple-helical body of the collagen molecule. The

Sheets

119

error (approx. 4.5%) in the value of b is an estimate based not on random errors in measurement (which were small) but on possible systematic errors (magnification, shrinkage, see Results, section (d)). The picture that emerges is one in which N-propeptides cannot be widely separated on the sheet surface but must sit closely side by side in lateral directions, probably with adjoining domains in contact. A precise value for the lateral width of an N-propeptide domain (known to be elongated in an axial direction) is not yet available but in the light of current knowledge of amino acid sequences and axial dimensions (Engel et al., 1977; Timpl & Glanville, 1981) a value that is approximately 1.5 times that of the collagen triple helix is not unreasonable for the overall width of the domain together with that part of the parent molecule over which it is folded back. This result provides quantitative support for the view that N-propeptides can regulate collagen fibril growth by blocking further accretion when their surface density reaches a critical value. On pN-collagen sheets, this critical surface density occurs when the N-propeptide domains have a lateral centre to centre separation of b = 2.23( 10.1) nm and are presumably closepacked in this direction. A representation of such a sheet in transverse section is shown in Figure 5. A sectioned N-propeptide domain, together with that part of the parent molecule over which it is folded back, is represented as a stippled circle of large diameter. Smaller circles indicate where other (triple-helical) parts of the pN-collagen molecules have been intersected; for every large circle there must, on average, be four smaller ones. The larger circles are all surface-located, with centre to centre separation = b. If, as shown in the diagram, they are closely packed laterally (confining smaller circles to the interior of the sheet), b is also the diameter of a larger circle. The value calculated for this lateral spacing, b, does not, however, depend on the assumption of a specific cross-sectional outline for an N-propeptide domain; circles are shown merely to simplify the representation. It is stressed, moreover, that no specific intermolecular lateral packing arrangement of the smaller circles, regular or otherwise, need be assumed. In the axial direction, normal to the plane of the cross-section, molecules are staggered by D, or multiples thereof; i.e. the packing is D-periodic in this direction. An accurate value for the geometric thickness, t (nm), of a sheet (or tape) is less readily deduced from the molecular parameters. To a first approximation, the thickness of a hydrated sheet is the product of the number of molecular layers, N,, and the molecular diameter, a, giving an estimated thickness of 10.4 nm. This takes no account, however, of the extra thickness of the N-propeptide domains on the surface. It also ignores the extent to which molecules are close-packed in the direction of the sheet thickness. Applying corrections to take these factors into account (corrections that are bound to be crude) leads to rather higher values oft,

D. F. Holmes

120

Figure 5. A representation of a pi%-collagen sheet (i.e. a sheet consisting exclusively of pPj-collagen molecules) in transverse section through an overlap zone. A sectioned PJ-propeptide domain, plus that part of its parent molecule over which it is folded back, is represented as a large stippled circle; smaller circles are sections through other (triple-helical) parts of the pPu’-collagen molecules. The I,: D ratio of 4.5 means that, for every large circle there will, on average, be 4 smaller ones. All K-propeptide domains are asumed to be located on 1 or other of the 2 surfaces of the sheet. Mass per unit area measurements show t,hat, in a lateral direction on the surface (i.e. in the piane of the section), the centres of the domains are mutually spaced, on average, by b = 2.23( +O.l) nm, roughly 15 times the diameter of the collagen triple helix. This result implies that the domains are close-packed laterally (shown here by placing the larger circles in contact). It is to be noted that no specific intermolecular lateral packing arrangement of the smaller circles, regular or otherwise need be assumed.

in the region of 11.0 to 11.5 nm. It is emphasized that this refers to the overall thickness of a hydrated pN-collagen sheet (through an overlap zone and the N-propeptide domains) and not to its actual thickness during electron-optical examination. As shrinkages of 25 to 30% in collagen fibril diameters can occur on dehydration, embedding and sectioning (Eikenberry et al., 1982b; Craig et al., 1986), these mass-based estimates of approximately 11 nm for the geometric thickness of a hydrated pN-collagen sheet are not inconsistent with the much smaller value of 7.9 ( f 1.3) nm measured by Hulmes et al. (1989a) on sectioned sheets. (b) Procollagen

and pC-collagen

sheets

As noted earlier, the sheet/tape thickness (here the molecular thickness, N,, Table 3) is least uniform when N-propeptides are absent and only C-propeptides are present; the morphology and compactness of the pC-collagen tapes are also much more variable. When N-propeptides are present (in pN-collagen sheets and procollagen tapes), the

et al.

molecular thickness is the same, whether the C-propeptides are present, or not (Nt = approx. 6.8 layers of molecules). The C-propeptides merely have the efFect of increasing the thickness variability (i.e. the fractional s.D.). Seen in toto, these observat’ions support the view that the N-propeptides could be playing a key st,ructural role in regulating latera,l accretion in viuo. Existing STEM data from sheets or tapes incorporating C-propeptides cannot yet be accounted for in terms of a specific and clearly defined spatial relationship (viewed in axial projection) between the C-propeptide domain and it,s parent molecule, as can tentatively be done in the case of the K-propeptide domain. Axial mass distributions from procollagen and PC-collagen tapes (analogous to that for pN-collagen displayed in Fig. 3) show high contrast but have hitherto failed to reveal a well-defined axial location for the bulky C-propeptide domain (approx. 3 x the mass of an N-propeptide domain), largely because t,he mass distribut’ion profiles cannot be aligned accurately with the collagen overlap/gap zoning pattern. Studies using conventional TEM techniques suggest that the C-propeptide domains project out sideways from the procollagen tapes at the C-terminal overlap/gap junctions (Mould et al., 1990). Bruns et al. (1979) and Hulmes et aZ. (1983) have reported the occurrence, in the cult’ure media of chick embryo tendon cells a.nd in homogenates of t,hese cells, of large numbers of procollagen aggregates in which the molecules appear in zero-D array (analogous to the packing of collagen molecules in SLS crystallites). A striking feature of many of the aggregates is the presence of “globules on stalks” extending axially from one end (occasionally both ends) of an aggregate. A more recent study of rotary-shadowed and negatively-stained procollagen aggregates by Mould & Hulmes (1987) confirms that the globules are aggregates of C-propeptide domains. These observations show that C-propeptides can exist in an extended conformation, projecting axially from the parent molecules. Nothing can be inferred from the micrographs in the two earlier papers about the conformation of the N-propeptides in the aggregates. ic)

Sheet

formation

Why pN-collagen, procollagen and pkmllager~ assemblies are sheet-like in form, rather than fibrous, remains to be considered. What, we now ask, are the properties of collagen and its propeptides that bring about such a radical change in assembly behaviour? The sheets have certain features in common with soap films. It has long been known that the thinnest soap films (“black” films) have a thickness of the order of 12 nm and that, “variations in thickness of the black portions of the films are but a small fraction of that thickness” (Reinold Bi Riicker, 187’7; 1883). The stability of these films depends on the dual character of a soap molecule; typically part

Collagen

121

Sheets

Table 4 Solubility

data for

type

I collagen,

pN-collagen,

procollagen

and pC-collagen

Critical logical Solute Collagen pN-collagen Procollagen pC-collagen

concentration in physiobuffer at 37°C (pg ml-’ (+S.D.))

Gibbs free energy change, AC”; on assembly? (kJ mol-‘)

942 (kO.10) (a) 175 (+25) (b) 1250 (& 250) (c) 1000(,200) (d)

-53 -37 -33

-33

-

(a) Data from collagen prepared de ROWJ by the enzymic cleavage of propeptides from type I procollagen purified from the medium of cultured human fibroblasts (Kadler et al., 1987). (b) Data from pN-collagen-prepared by the enzymic cleavage of C-propeptides from type I procollagen purified from the medium of cultured human fibroblasts (Hulmes et aE., 1989b); similar data were obtained from chick _. pN-collagen (Mould, unpublished results). (c) and (d) Data from material prepared from type I procollagen extracted from chick embryo metatarsal tendon (Hulmes et al., 1989b). In all cases, the critical concentrations in solution (in equilibrium with fibrils or sheets at 37°C in a physioiogical buffer). are those quoted in these publications. i The critical concentration; c,:, (expressed as a molarity), is related to the equilibrium constant, K,,, by e.,,, = l/K,, for monomer addition to fibrils (Lauffer, 1975; Oosawa & Asakura, 1975). This allows AG”, the change in Gibbs free energy on assembly, to be calculated from AC” = RT ln(c&.

hydropilic (a carboxyl group at one end) and part hydrophobic (a hydrocarbon chain). Hydrogen bonding and ionic interactions (between hydrophilic groups, water molecules and metal ions) are maximized when the non-interacting hydrophobic chains are excluded and are confined to a layer at the film/ air interface. The film is stable at a certain thickness because any local extension and thinning of the film leads to migration of hydrophilic groups and water molecules into the surface layer, so reducing attractive interactions and increasing the free energy of the system. Conversely, a local thickening and contraction of the film forces hydrophobic groups from the surface into the bulk phase, introducing energetically unfavourable contacts between hydrophilic groups and hydrophobic groups and so to another increase in the overall free energy of the system. A soap film of a certain thickness is, therefore, in a state of minimum energy, its thickness depending on the number of water molecules present and the nature of the soap molecules themselves (Adams, 1941). pN-collagen molecules form stable films of comparable thickness and uniformity, but in an aqueous environment rather than in air. Like a soap molecule, pN-collagen has a dual character with the N-propeptide domain possessing interaction properties quite different from those displayed by the rest of the molecule (the triple helix and telopeptides). This is evident from solubility data (Table 4). Fully processed collagen is highly insoluble and readily precipitates into fibrils at low concentration (for references, see the legend to Table 4). In the absence of propeptides, therefore, collagen-collagen interactions are associated with a substantial drop in free energy (AGO = - 53 kJ mol-’ at 37°C). The solubility of pN-collagen is much higher (approx. 400. fold), so the reduction in free energy on assembly is less (AGO = - 37 kJ mall’ at 37°C). This implies that the presence of N-propeptides is associated with energetically unfavourable interactions that

oppose collagencollagen interactions. On assembly, greatest reduction in AGassembly will occur when favourable intermolecular interactions (collagencollagen contacts) are maximized and unfavourable interactions (involving N-propeptides) are minimized. By analogy with soap films, this can be expected to lead to the preferential location of Pu’-propeptides at the interface between the assembly and its environment, now an aqueous environment. If all molecules are pN-collagen, the minimum energy configuration is a thin film of uniform thickness with the N-propeptides forming a continuous layer on both surfaces. Again the film is stable because departures from this thickness lead to fewer intermolecular contacts that are energetically favourable and to an increase in contacts that are not. Similar considerations presumably apply to C-propeptides. Procollagen and pC-collagen are both highly soluble, and assembly is accompanied by a reduction in free energy slightly smaller than that for pN-collagen (Table 4). (d) Implications

for Jibrillogenesis

in vivo

Although there is no evidence that collagen precursors assemble into sheets other than in vitro, the phenomenon does have certain implications for our understanding of fibrillogenesis in vivo. We have seen, using a soap film as an analogy, that sheet formation is most readily accounted for (on energetic grounds) in terms of a layer of propeptides on both surfaces. This adds weight to the argument that a similar location is energetically preferred by uncleaved propeptides on collagen fibrils growing in vivo. Confined to the fibril surface in this way, the propeptides could then play a regulatory role by restricting further lateral growth. A second implication, also prompted by the soap film analogy, concerns the nature of intermolecular contacts in sheets and first-formed fibrils. Soap films

D. F.

122

Holmes

are certainly liquid films. Their stability depends on fluid-like intermolecular contacts, allowing continual molecular re-arrangement and freedom to adopt an overall minimum energy configuration. The same should be true of sheets assembled from collagen precursors; their stability might also be expected to depend on intermolecular contacts that are continually changing. It is difficult to escape the conclusion that, as in soap films, the molecules are continually undergoing mutual re-arrangement. In the case of collagen, however, this fluid-like behaviour of intermolecular contacts can only be partial. The regular D-periodic structure of the sheets indicates that the mutual alignment of molecules in an axial direction must everywhere be conserved; molecular re-arrangements cannot occur to any significant extent in this direction. Fluidity in intermolecular contacts is, therefore, restricted to directions normal to the long axes of the molecules. There is no reason to suppose that such fluid-like behaviour (in 2 dimensions) should be confined to collagen structures formed in. vitro. Indeed, it is already known that intermolecular contacts in native collagen fibrils grown in viva have a partially fluid-like character. This is well documented by the nuclear magnetic resonance studies reported by Torchia and co-workers; 13C and ‘H spectra show that molecules undergo azimuthal fluctuations of considerable amplitude, greatest in non-crosslinked fibrils but still appreciable after crosslinking (Jelinski & Torchia, 1979; Jelinski el al., 1980; Sarkar et al., 1983, 1985). Such fluctuations must mean that a multiplicity of varying intermolecular contacts in dynamic equilibrium exists in lateral directions in a fibril. Seen together, our observations and those of Torchia and his co-workers are entirely consistent with a dynamic model for a first-formed collagen fibril in which, prior to crosslinking, liquidlike molecular re-arrangements are continually taking place in directions normal to the fibril axis. Partial liquid-like behaviour of this character need not preclude the existence of preferred mutual orientations of adjacent molecules and some shortrange lateral order, extending to long-range order when crosslinking occurs. Liquid-like behaviour in lateral directions could also provide a simple explanation for the circular outline of fibrils. We thank 1Mr T. Brown (Tom Brown Developments) and Mr S. M. W. Grundy for expert technical assistance with the instrumental modifications and construction of the STEM/computer interface. We are grateful to Dr D. J. S. Hulmes for valuable discussions and Mrs C. Cummings for assistance with the electron microscopy. Our thanks are due to the Arthritis and Rheumatism Council for extended financial support. Funds for the purchase of the electron microscope also came in part from the Medical Research Council. References Adam, R’. K. (1941). The Physics and Chemistry of Surfaces, 3rd edit., Oxford University Press, London. Bachinger, H. P.: Doege, K. J., Petsehek, J. P., Fessler,

al.

et

I. & Fessler, J. II. (1982). j. Biol. Chem. 25’7. 14590-14592. Bard, J. B. L. & Chapman, J. A. (1973). Nature Lvezr:Bioi. 246, 83-84. Bowden, J. K., Chapman, J. A. & Wynn, C. H. (1968). Biochim. Biophys. Acta, 154, 19&195. Bruns, R. R., Hulmes, D. J. S. Therrien; S. F. & Gross, J. (1979). Proc. Nat. Acad. Sci., U.S.A. 76, 313-317. Chapman, J. A. (1989a). Riopolymers, 28, 136771382. Chapman, J. A. (1989b). Biopolymers, 28, 2201-2295 (erratum). Craig, A. S., Robertson, J. G. & Parry, D. A. D. (1986). J. Ultrastruct. Mol. Struct. Res. 96, 172-175. Dehm; P., Jimenez, S. A.; Olsen B. R. & Prockop, D. J. (1972). Proc. Xat. Acad. Sci., U.S.A. 69, 60-64. Dombrowski, K. E. & Proekop. D. J. (1988). j. BioE. Chem. 263, 1654516552. Eikenberry, E. F., Brodsky, B. & Parry, D. A. D. (1982a) L.

Int.

J. Biol.

Macromol.

4: 322-328.

Eikenberry, E. F., Brodsky, B.? Craig, A. S. & Parry, D. A. D. (1982b). Int. J. Biol. Macromol. 4, 393-398. Engel, $. (1978). Cltramicroscopy, 3, 273-281. Engel, A. (1982). :Micron. 13, 425-436. Engel, A., Reichelt R. (1984). J. Vltrastruct. Res. 88, 105120. Engel, J.; Bruckner, P., Becker. C., Timpl, R. & Rutschmann, B. (1977). Biochemistry, 16, 402G-4033. F. & Michel, B. (1981). Engel, A., Christen, Ultramicroscopy, 7, 45-54. Fjolstad, M. & Helle, 0. (1974). J. Pathol. 112, 183%I88. Fleischmajer, R., Olsen, B. R., Timpl, R., Perlish, J. S. & Lovelace, 0. (1983). Proc. Nat. Acad. Sci., C’.S.A. SO, 3354-3358. Fleischmajer, R.; Perlish, J. 8. & Timpl, R. (1985). Ann,. N. Y. Acad.

Sci.

460,

24e-257.

Fleischmajer, R.? Perlish, J. S. & Olsen, B. R. (1987a). Cell Tiss. Res. 247, 105-109. Fleischmajer, R.: Perlish. J. S. & Olsen, B. R. (19876). J. Invest. Dermatol. 89; 212-215. Freeman R. & Leonard, K. R. (1981). J. Microsc. 122, 275-286. Helseth, D. L., Lechner, J. H. & Veis. A. (1979). Biopolymers,

18. 3005-3014.

Hodge,

A. J. & Petruska, J. A. (1963). In Aspects of Protein Structure (Ramachandran, G. E., ed.), pp. 289300, Academic Press, London & Pjew York. IIoffmann, H.-P., Olsen, B. R.: Chen. H-T. & Proekop, D. J. (1976). Proc. Nat. Acad. Sci., U.S.A. 73, 4304-4308. Hofmann, H., Voss, T., Kuhn, K. & Engel. J. (1984). J. Mol.

Biol.

172,

325-343.

IIojima?

Y., van der Rest, M. & Prockop, D. J. (1985). J. Biol. Chem. 260, 15996-1600X Hulmes, D. J. S. (1983). CoElagen Rel. Res. 3, 317-321. Hulmes, D. J. S. & Miller, A. (1979). Xature (Lon,donj. 282, 878880. Hulmes, D. J. S., Bruns, R. R. & Gross, J. (1983). Proc. Nat.

Acad.

Sci.,

U.S.A.

80,

388-392.

Hulmes, D. J. S., Kadler, K. E., Mould, A. P., Hojima. Y., Holmes, D. F., Cummings, C.. Chapman, J. A. 8; Prockop, D. J. (1989aj. J. Mol. Biol. 210> 337-345. Hulmes, D. J. S., Mould, il. P., Kadler, K. E., Chapman, J. A. & Prockop, D. J. (1989b). In Cytoskeletal and Extrucelldar Proteins (Aebi, U. & Engel, J., eds). pp. 292-301. Springer-Verlag. Berlin. Jelinski, L. W. & Torchia, D. A. (1979). J. &Iol. Biol. 133. 45-65. Jelinski, L. W.. Sullivan. C. E., Batchelder, L. S. & Torchia, D. A. (1980). Biophys. J. 32, 515-529.

Collagen

Jones,

E. Y. & Miller, A. (1987). Biopolymers, 26, 463480. Kadler, K. E., Hojima, Y. & Prockop, D. J. (1987). J. Biol. Chem. 262, 15696615701. Kadler, K. E., Hojima, Y. & Prockop, D. J. (1988). J. Biol. Chem. 263, 10517-10523. Kaper, J. M. (1986). In Molecular Basis of Virology (Fraenkel-Conrat, H., 1-133, ed.), PP. Van-Nostrand-Reinhold; ru’ew York. Kessler; E., Adar, R., Goldberg, B. & Niece, R. (1986). Collagen Rel. Res. 6, 2499266. Lauffer, M. A. (1975). Entropy-driven Processes in Biology, Chapman & Hall, London. Miller, A. (1982). Trends Biochem. Sci. 7; 13-18. Miller, A. (1985). In Fibrosis, Ciba Foundation Symposium 114, pp. 65-79, Pitman, London. Morris, N. P., Fessler, L. & Fessler, J. H. (1979). J. Biol. Chem. 254, 11024-l 1032. Mosler, E., Folkhard, W.; Geercken, W., Helle, O., Knorzer, E., Koch, M. H. J., Lap&e; C. M.; Nemetschek-Gansler, H., Nusgens, B. & rjemetschek, Th. (1986). 2. Naturforseh. 41c, 4899492. Mould, A. P. & Hulmes, D. J. S. (1987). J. Mol. Biol. 195; 543-553. Mould, A. P., Hulmes, D. J. S., Holmes, D. F.; Cummings, C., Sear, C. H. & Chapman, J. A. (1990). J. Mol. Biol. 211, 581-594. Edited

by D.

Sheets

Olsen,

123

B. R. (1981). In Cell Biology of the Extracellular Matrix (Hay, E. D., ed.), pp. 139-177, Plenum Press, New York. Oosawa, F. & Asakura, S. (1975). Thermodynamics of the Polymerization of Protein, Academic Press, London & Kew York. Parry; D. A. D. & Craig, A. S. (1979). Nature (London), 282, 213-215. Parry, D. A. D. & Craig, A. S. (1984). In Ultrastructure of the Connective Tissue Matrix (Ruggeri, A. & Motta, P. M., eds); pp. 34-64, Martinus Nijhoff, Boston, MA. Reinold, A. W. & Riicker, A. W. (1877). Proc. Roy. SC~. London, 26, 334345. Reinold, A. W. & Riicker, A. W. (1883). Phil. Trans. Roy. Sot. London, 174, 645-662. Sarkar, S. Ii., Sullivan, C. E. & Torchia, D. A. (1983). J. Biol. Chem. 258, 9762-9767. Sarkar, S. K., Sullivan, C. E. & Torchia, D. A. (1985). Biochemistry, 24, 234882354. Tanzawa, K.; Berger, & Prockop, D. J. (1985). J. Biol. Chem. 260, 1120-1126. Timpl, R. & Glanville, R. W. (1981). Clin. Orthoped. 158, 224-242. Traub, W. & Piez, K. A. (1971). Advan. Protein Chem. 25, 243-353. Veis, A.; Anesey, J., Yuan, L. & Levy, S. J. (1973). Proc. Nat. Acad. Xci., U.S.A. 70, 1464-1467. DeRosier