JOURNAL OF ELECTRON MICROSCOk'Y TECHNIQUE 18:262-268 (1991)
Image Analysis of Mineralized and Non-Mineralized Type I Collagen Fibrils A. LARRY ARSENAULT Electron Microscopic Facility, Faculty of Health Sciences, MeMaster University, Hamilton, Ontario L8N 325 Canada
KEY WORDS
Collagen, Mineral, Calcification, Tendon, Electron microscopy
ABSTRACT
Turkey leg tendons at a n early stage of mineralization have been thin sectioned and imaged by electron microscopy. At this stage collagen-associated mineral apatite was found to be present within both the gap and overlap zones. The earliest apatite occurs in a microcrystalline form which gives a rather generalized and characteristic density to both the gap and overlap zones; with subsequent development larger defined apatite crystals arise which span gapioverlap zones. Fourier transformation of such images revealed the major 67 nm axial repeat of the gapioverlap zone plus four other maxima corresponding to repeat spacings of 22,16,13, and 11nm respectively. Computer imaging techniques were used to reconstruct images by using selected spatial frequencies from such transforms. In this manner the subperiodic distributions of mineral were visually enhanced. These subperiodicities are positioned in a n asymmetric fashion over the entire D unit repeat aligning with the molecular orientation of the fibril. Analyses of both negatively stained collagen and computer-generated maps of collagen hydrophobicity were compared to the mineral distribution of collagen. Densitometric comparisons showed a positional correlation between the axial banding patterns of mineralized fibrils and those of negatively stained non-mineralized fibrils. Comparable spatial frequencies were also present in transforms between hydrophobic maps and mineral distribution of collagen. These results suggest that the lateral clusterings of hydrophobic residues which span the fibril at specific sites in both the gap and overlap zones serve to prohibit early mineral deposition. This observed hydrophobic influence in combination with the gap space appear as contributing factors in the observed axial distribution of mineral within collagen.
INTRODUCTION Mineralized turkey leg tendons have been extensively used a s a n idealized model for the study of interrelationships between collagen and apatite crystals; this is due to the tendon's linear alignment of collagen fibrils as compared to the complex and varying patterns of those in bone and dentin. This is further extended by the fact that there is no evidence to indicate any molecular variation among these type I collagens. The axial periodicity of these mineralized fibrils has been well documented by using electron microscopy (Nylen et al., 1960; Meyers and Engstrom, 1965; Krefting et al., 1980; Landis, 1986; Weiner and Traub, 1986; Arsenault, 1988) and diffraction analysis (Meyers and Engstrom, 1965; Engstrom, 1966; Eanes et al., 1970; White et al., 1977; Berthet-Colominas et al., 1979; Bigi et al., 1988). These diffraction studies have reported the major axial banding repeat to be 66 to 67 nm which is the same as bone. Such banding repeats have been explained by mineral occupying the potential space present in the gap zone (Hodge and Petruska, 1963; White et al., 1977). This gap space is due to the axial ordering of collagen molecules (linear rods 300 x 1.4 nm) where the ends of each molecule are separated by a distance of 35 nm (Hodge and Schmitt, 1960; Hodge and Petruska, 1963; Hodge, 1970). The lateral stacking order is such that adjacent molecules are staggered by 67 nm spacing (234 amino acid residues); this spacing
D 1991 WILEY-LISS, INC
is commonly referred to as the D stagger or D unit. Thus, the gap zone is comprised of one gap space and four adjacent molecules of collagen, while each overlap zone contains a continuous lateral stacking of adjacent molecules. With the use of selected-area dark-field electron microscopy which enables the specific and direct localization of apatite, it has been illustrated that the major portion of apatite in collagen is indeed within the gap zone; however, a considerable amount of apatite is also present in the overlap zone even a t the early onset of mineral deposition (Arsenault, 1988). The relative amounts of apatite within the gap and overlap zones depend upon the stage or degree of mineralization (Arsenault, 1989). In further examination of this crystalcollagen interrelationship computer-generated arrays of type I collagen amino acid sequences (al-ol2-al)were constructed; such arrays can be aligned with electron microscopic images of mineralized leg tendon (Maitland and Arsenault, 1991).By such a comparison it was found that hydrophobic amino acid residues possessed a n inverse relationship with sites of early mineral deposition. Moreover, these hydrophobic domains occur as
Received J u n e 8, 1990; accepted in revised form August 29, 1990. Address reprint requests to Dr. Larry Arsenault, Electron Microscopic Facility, Faculty of Health Sciences, McMaster University, 1200 Main Street West, Hamilton, Ontario L8N 325, Canada.
TYPE I COLLAGEN
distinct axial crossbandings and their positions coincide with areas which are low in mineral content or have excluded mineral. Importantly, this work has demonstrated apatite to occur as axial crossbandings along the D unit in both the gap and overlap zones. Thus, in addition to the 67 nm repeats, mineral was distributed along the D unit as axial subperiodicities. These occur in an asymmetric fashion, and thereby, we have been able to determine the molecular polarity of unstained, mineralized collagen fibrils. This is particularly important considering that adjacent fibrils within the leg tendon may have opposite molecular orientations (Arsenault, 1991; Maitland and Arsenault, 1991). In order to enhance and clarify the presence of these subtle subperiodic bands Fourier transformation and computer imaging techniques have been utilized in the present study. Also, densitometric comparisons were made between the axial crossbandings of mineralized collagen and those produced by negative staining of non-mineralized collagen. To further characterize the correlation between early mineral deposition and hydrophobicity as previously reported, Fourier transformations were performed on mineralized collagen and computer-generated maps of collagen hydrophobic amino acid residues.
MATERIALS AND METHODS Electron Microscopy Portions of early mineralizing leg tendons from 14 week-old domestic turkeys were placed in 100% glycerol for 3 h. Areas of interest at the mineralization front were removed and placed in a graded series of glycerolimethanol and then embedded in Spurr’s resin. Thin sections were cut in longitudinal planes along the tendons with a diamond knife. Non-mineralized portions were examined after positive staining with aqueous 1%phosphotungstic acid (PTA) a t pH 3.2 followed by 1%uranyl acetate (UA) at pH 4.2 while mineralized portions were not stained. All sections were viewed in a Philips 300 operating at 80 kV by using a 20 pm objective aperture. Image Analysis Images were digitized with a Dage video camera (Dage-MTI, Inc., Michigan City, IN) at standardized settings for both magnification and contrast levels. Axial periodicities of non-mineralized and mineralized collagen were Fourier filtered by using a computer image analysis system (IBAS Kontron; Etching, West Germany). For this process 512 x 512 pixel matrices of original negatives (magnification of 19,000 times) were digitized and Fourier transformed. Masks were constructed directly over the transforms of both the stained, non-mineralized collagen and the mineralized collagen to insure the selection of specific spatial frequencies. The resulting masks were rod shaped with outward-curved ends and were large enough to fit over all of the Fourier maxima. These masks were then lowpass Gaussian filtered by using a matrix size of 5 x 5 pixels. For the stained, non-mineralized collagen the mask was contrast reversed to exclude the spatial frequencies which correspond to the axial repeat banding;
263
this was followed by inverse Fourier transformation to obtain an image in which the axial repeats have been removed. The Fourier transforms of mineralized collagen were filtered to either exclude (as above) or include the Fourier maxima. In the latter case, the inversed Fourier transform, having only axial repeats, was either subtracted or divided from the native image. The subtraction or division of these image sets gave qualitatively the same result. A computer-generated map of collagen fibril hydrophobicity as previously described (Maitland and Arsenault, 1991)was Fourier transformed for comparison with the transforms of mineralized collagen. Densitometric tracings of grey values were plotted along the D units of mineralized collagen and of negatively stained non-mineralized type I collagen. The negatively stained fibrils were taken from published micrographs (Chapman and Hulmes, 1984; Nimni and Harkness, 1988). In order to reduce minor localized variations in either the mineral or stain distribution the images were lowpass filtered by using a 5 x 2 pixel matrix. To achieve this the long axis of the filter was passed parallel to the lateral direction of the fibril.
RESULTS A longitudinally sectioned and stained type I collagen fibril from a non-mineralized portion of the turkey leg tendon is shown in Figure l a . In this image 12 crossbands are shown to axially repeat along the fibril; these bands have been labelled with the “e to a” nomenclature (Hodge and Schmitt, 1960)--“e” having two bands, “d” one band, “c” three bands, “b” two bands, and “a” four bands. The “e to a” direction corresponds to the amino- to carboxy-terminus orientation of collagen molecules; therefore in this image the N- to C-direction of collagen molecules is from left to right. The areas defined as the gap and overlap zones are also labelled. (The gap zone consists of “a3,2,1,e2,3rd, and c3” bands and the overlap zone consists of b2,1 and a4”.j A subtle ultrastructural feature of these type I fibrils is the presence of axially oriented striations in the gap zone. These striations are apparent in the gap zone (“a” and “e”) as lucent spaces; further into the gap zone the striations are less evident because of the generalized weaker staining regions bordering the d band. In order to enhance these subtle striations, the original image was Fourier filtered to remove the repeating axial crossbands (Fig. lb). In this computer-displayed image the aligned striations are clearly evident ranging from 5 to 8 nm in width. An unstained mineralized fibril is presented in Figure l c ; this image, like that of stained collagen, represents a map of electron transparency, the contrast of which is due to the electron-scattering ability of the specimen integrated through its thickness. For unstained mineralized collagen the inherent contrast results from the presence of apatite in both the gap and overlap zones (Arsenault, 1988). Characteristically, the mineral density in such overprojected images appears in early stages of mineral deposition as a punctate distribution which defines the gaploverlap zones and their interzone banding or in later stages as dense linear localizations which are not restricted to a par-
264
A.L. ARSENAULT
Fig. 1. a: A stained non-mineralized collagen fibril from the turkey leg tendon showing the characteristic axial crossbanding pattern which has been labelled with the e to a nomenclature; the gap (GZ) and overlap (OZ) zones are marked. A few o f the subtle axial striations present in the ,-e2,1 bands of the gap zone are labelled by arrowheads. The white arrowheads point out the banded alignments of these striations b: A Fourier filtered image in which the axial crossbandings o f a have been removed; this method serves to highlight the axial striations (arrowheads). The black arrowheads correspond to the same areas of the gap zone as the white arrowheads in a. c: An unstained mineralized collagen fibril characterized by alternating gap (GZ) and overlap (OZ) zones having an axial periodicity of
67 nm. The lucent spaces of the al-e2 bands (white arrowheads) are asymmetrically located nearer the N-terminal position of the gap zone; thus the N- to C-molecular orientation of the fibril is from left to right. Mineral density is found within both the gap and overlap zones. Two distinct qualities of mineral apatite are demonstrated: in the upper portion of the image the punctate distribution of apatite microcrystals clearly defines the gap and overlap zones and the interzone banding; in contrast, the lower portion contains, in addition to these microcrvstals. elongated auatite crvstals (arrows) which are not confined to" a particular zone b u t can "span the gapioverlap borders a,b, x 250,000; c, x 253,000.
TYPE I COLLAGEN
ticular zone. The axial periodicities of the mineralized sites are due to elevated levels of mineral accumulation in the gap zone as compared to the overlap zone; this is particularly evident in early stages of mineralization. In this image apatite is predominantly in the form of small microcrystallites. Collectively over the section thickness, these microcrystallites have a recognizable alignment which is specifically related to collagen. They are aligned in a spatial overprojection at different concentrations in the gap and overlap zones. Also, within the gap zone there is an asymmetrically placed lucent band at the a1-e2position indicating a reduction of mineral (Fig. l c ; white arrowheads), while the overlap zone mineral deposition is less dense along the borders with the gap zones (c,-c, and a4-a3).This is clearly illustrated by densitometric tracings of mineralized collagen (Fig. 4aj. The asymmetrically placed lucent band in the gap zone ( a l e 2 ) lies nearer the amino terminus side of the gap zone (Maitland and Arsenault, 1991) and thereby the N- to C-molecular polarity of the fibril can be determined. In this image (Fig. lc) the Nto C-orientation is from left to right. In addition, within these mineral densities larger and more distinguishable apatite crystals are observed; such crystals are believed to represent a more advanced stage of crystal growth. Figure 2a is the same specimen area as Figure l c but shown at a lower magnification and has been rotated 90" clockwise with the N- to C-orientation from top to bottom. This image has been used for Fourier analysis and subsequent computer image enhancements. Figure 2b is a Fourier transformation of Figure 2a showing the characteristic spatial frequencies of mineralized collagen. The first Fourier component through to the sixth represent the major axial periodicities of 67 nm and a series of subperiodicities 22, 16, 13, and 11 nm respectively. Absence of the second Fourier component is presumably due to the periodic square wave created by the high degree of contrast between the gap and overlap zones. These five Fourier components along with the zero spatial frequency were masked from the transform and were then inverse transformed to produce the image shown in Figure 2c. By this Fourier filtering process Figure 2c contains structural information which arises predominantly from the axial repeats caused by sites of microcrystal deposition. This image has been further computer enhanced in Figure 2d to show more clearly the major 67 nm banding and the coexisting subperiodicities. The random and non-periodic distribution of mineral in Figure 2a is readily appreciated by the subtraction of Figures 2a and c, the resultant image is Figure 2e showing the distribution of the larger apatite crystals which are not periodically arranged. When studying the structure of type I collagen fibrils it becomes apparent that mineralized fibrils and negatively stained, non-mineralized fibrils have similar axial periodicities. Comparisons by densitometric tracings of these fibrils are shown in Figure 3 where a is from an early stage of mineralization in the leg tendon, b from a negatively stained fibril by Chapman and Hulmes (1984; Fig. lob), and c is from another negatively stained fibril by Petruska published by Nimini
265
and Harkness (1988; Fig. 2). The important feature of these density profiles is not the actual grey-scale values for this depends upon several variables unrelated to the chemistry of collagen; the salient feature is, however, the similarity in location of peaks and troughs within the D unit. This corresponding spatial positioning of mineral and stain deposition may, in part, be influenced by hydrophobic interactions. Figure 4a is a computer-generated map of collagen hydrophobicity as described by Maitland and Arsenault (1991). The hydrophobic amino acid residues appear as white positional representations of number densities on a black background. Fourier transformations (Fig. 4b) of such a hydrophobic map are shown for comparison with a transform of mineralized collagen (Fig. 4c). The hydrophobic maps were scaled to align with images of mineralized collagen; this alignment was to insure the same distancelpixel between the maps and the mineralized collagen. The transform of the hydrophobic map (Fig. 4b) contained an increased number of spatial frequencies as expected; however, five of the major frequencies are in alignment with the spatial frequencies of mineralized collagen (Fig. 4c). This illustrates that their periodicities are to some degree comparable.
DISCUSSION Early sides of apatite deposition in type I collagen fibrils of the turkey leg tendons have been studied with the use of image enhancement techniques. Also, analysis of both negatively stained collagen and computergenerated maps of collagen hydrophobicity have been compared to the mineral distribution within collagen. This study indicates that: 1) the early deposition of apatite microcrystals within type I collagen occurs as distinct subperiodicities in addition to major 67 nm period; 2) these subperiodicities are positioned in an asymmetric fashion over the entire D unit repeat and not restricted to the gap zone; 3) similar crossbanding patterns occur in both mineralized fibrils and in those non-mineralized fibrils which have been negatively stained; and 4) corresponding subperiodic crossbandings are present in mineralized collagen and computergenerated maps of hydrophobic amino acid residues. The high degree of correlation among these results appears extraordinary when one considers the extremely varied conditions under which these comparisons have been made. For example, the axial periodicity of mineralized fibrils in routine thin sections (Fig. l a ) typically contain 40-50 collagen molecules stacked through the thickness of the sample; in contrast, negatively stained isolated fibrils contain a minimum of four times this amount, while the computer-generated maps of hydrophobicity are displayed as only one molecular layer. Also, the different operative factors of imaging can influence the basic data from which the comparisons were made; such factors are: sample preparation, degree of staining, extent of mineral accumulation, plane of section and tilt angle in the microscope, degree of focus, radiation damage, alignment, and stacking orders. However, these factors appear to have little effect given the consistent observations of corresponding subperiodicities throughout the D unit under the different study conditions. This reflects the high
A.L. ARSENAULT
266
Fig. 2. a: A computer-displayed image is of the same specimen area as Figure l c hut shown at a lower magnification and rotated by 90". The N C orientation is from top to bottom as indicated by the lucent a,-ez spaces (arrowheads). The aggregates of large apatite crystals are indicated by arrows. Non-mineralized lucent area (*). b A Fourier transformation of a showing the typical ordered spatial frequencies of mineralized collagen. The ordered frequencies 1 and 3 to 6 were used for the inverse transformation to obtain image (c).c: A
filtered Fourier image showing only the axial repeats of mineral in both the gap and overlap zones. d With further image enhancements of c by adjusting contrast levels and edge contours this figure more clearly shows the subperiodicities in both the gap and overlap zones. e: This image was formed by the subtraction of a and c; thus, by removing the axial repeats from the native image this figure illustrates the non-repeating mineral distribution contained within these fibrils. Aggregates of large apatite crystals (arrows). x 120,000.
degree of molecular ordering within collagen fibrils and its controlling influence over apatite placement particularly at early stages of mineralization.
The defined localization of mineral within both the gap and overlap zones implicates factors which are considered to be both architectural and chemical in na-
-f
267
TYPE I COLLAGEN Overlap zone
a
'
Hlnerallzed Leg Tendon
c]
b
4a
Gap zone
' c
Negatively Eitatned Type 1 Collagen
Fig. 3. Densitometric tracings across a single D unit in the N- to C-molecular direction of (a)mineralized collagen from turkey leg tendon and (b,c) two published examples of negatively stained type I collagen fibrils [18,19]. In the mineralized fibril (a) the peaks have been labelled and these correspond to specific bands, 1 = cz-cl, 2 = b,-b,, 3 = a*-a3,4 = a l e 2 ,5 = d-c,. Opacity relates to the increased deposition of either mineral or stain while lucency reflects their reduction. Each tracing has corresponding peaks and troughs a t spatially similar locations, with the exception of b having three peaks instead of two in the gap zone (a,cf; this extra peak in b labelled * corresponds to the e,-d location. Importantly, the asymmetrically positioned lucent space in the gap zone which corresponds to e2-a3 is present in both the mineralized and the stained fibrils.
ture. These factors may act in concert with one another to compartmentalize the fibril into a series of axial crossbandings which allow for varying degrees of mineral deposition throughout the gaploverlap zones. For example, the major known architectural feature of collagen fibrils which allows for the accommodation of mineral is the 35 nm gap space at the ends of linear molecules. As suggested and demonstrated by Hodge and co-workers (1960, 19631, the gap space is the most likely location for mineral accumulation based on the availability of potential space in which apatite crystals can fit. In fact, Hodge has postulated that the gap spaces are channels into which apatite crystals may be deposited (Hodge, 1970). This gap zone location for mineral has been generally supported by others using a variety of techniques. However, the fact that mineral is present in both the gap and overlap zones a t the earliest detectable stages precludes defining the gap zone as the sole site of apatite nucleation (Arsenault, 1988, 1989, 1991; Maitland and Arsenault, 1991). Functionally, the gap zone is an area which can accommodate more mineral than the bordering overlap zones and may facilitate in the diffusion of ions into collagen fibrils. Another aspect of our limited understanding about the gap zone is the extent of lateral alignment of adjacent gap spaces. In this regard, Hodge (1970) has constructed models of D staggered collagen molecules in which continuous channels are created by the alignment of gap spaces. However, the true nature and architecture of the resulting intrafibrillar channels are not known. It is tempting, however, to speculate that
e d c b a
Fig. 4. a: A computer-generated map of hydrophobic amino acid groups with the N- to C-molecular orientation from left to right as denoted by the e to a notation. This map is displayed in reverse contrast with white being the spatial representations of clustered crossbandings of hydrophobic residues on a black background. b,c These are Fourier transformations of hydrophobic computer-generated maps and mineralized collagen respectively. For this comparison the hydrophobic map was reduced so that the gapioverlap zones aligned with that of the mineralized collagen. The arrowheads in the transform of aligned hydrophobic maps (b) occur at the same spatial frequencies as those for mineralized collagen (c).
the lucent and aligned axial striations of the gap zone (Fig. la,b) may arise from the alignment of gap spaces; evidence for interconnections among these channels was not apparent. Localized mineral accumulations are noted laterally across the gap zone (Fig. lc); however, at present there is little evidence to suggest that such accumulations occur at sites analogous to the axial striations observed in positively stained fibrils. Ultrastructurally, the mineral apatite appears in two distinct forms depending upon the stage of deposition. In the early stages apatite is microcrystalline, having a fine granular t o punctate nature. The distribution of this mineral form defines the 67 nm period and the associated subperiodicities. With subsequent development larger apatite crystals arise within the areas occupied by the microcrystalline apatite and thereby occur in both gap and overlap zones (Fig. lc). In such overprojected images the true structural features of individual crystals are not revealed; however, their distribution and increased densities as related to mineral accumulation and crystal growth are clearly determined. From a chemical perspective, the distribution of charged and hydrophobic residues is not uniform along the collagen molecule; it has been shown that interactions between each group plus their combined interactions serve to give the preferred D unit (234 amino acid) stagger in collagen assembly (Hulmes et al., 1973). In this staggered alignment these groups are organized into clusters with distinct axial crossbanding patterns visualized at the electron microscopic level with either positive or negative staining. The axial banding pattern derived by the positive staining of charged amino acids has been correlated with sequence data (von der Mark et al., 1970; Meek et al., 1979; Tzaphilidou et al., 1982a,b). Similarly, the negative staining properties of collagen have been demonstrated to reflect the distribution of clustered hydrophobic res-
268
A.L. ARSENAULT
idues (Doyle et al., 1975; Kobayshi et al., 1986). In the present study the negative influence of hydrophobic residues on the distribution of apatite in mineralized collagen has been studied by comparative analyses with negatively stained non-mineralized type I collagen fibrils and hydrophobic computer-generated maps of type I collagen. Areas of reduced mineral density correspond positionally to axial bandings in non-mineralized fibrils which were negatively stained (Fig. 3); this similarity of crossbanding can be explained on the basis of hydrophobic repulsion of apatite and stain deposition. Mineral reduction or exclusion at the major sites of hydrophobic clustering is further substantiated by Fourier analysis with the comparable spatial frequencies for the hydrophobic maps and mineralized collagen (Fig. 4). Based on these studies of mineralized collagen, the gap zone like the overlap zone is viewed as becng compartmentalized by the lateral alignment of hydrophobic domains. As judged by the consistent reduction of mineral a t sites having a high degree of hydrophobicity, it appears that such domains exert a negative effect upon mineral deposition. These repulsive effects on mineral deposition are significant in the gap zone a t the a l e 2 site and in the overlap zone a t the c2-cl and a4-a3sites a t early stages of mineralization (Maitland and Arsenault, 1991). At these sites the hydrophobic domains completely span the width of the fibril. Other hydrophobic regions appear as small “islands” and therefore would have a reduced capacity to effect the visualized distribution of mineral. However, i t appears that with continued growth and deposition the apatite crystals can overcome, to varying degrees, this negative hydrophobic inf hence. This is evidenced a t more advanced stages of mineralization where longer individual crystals are found to course through such subperiodicities and can project from one zone into another (Fig. lc). In conclusion, the findings of this report together with our previous studies indicate that the distribution of mineral apatite in type I collagen is governed by both the architectural and chemical components of the fibril. At early stages of deposition the mineral is, to a high degree, restricted from specific axial crossbandings by interactions with clustered hydrophobic residues; this appears analogous to the mechanism of negative staining. The initial sites of deposition and subsequent sites of apatite crystal growth are not simply confined to the gap zone as is generally accepted but occur in both the gap and overlap zones as a function of the mineralization gradient.
ACKNOWLEDGMENTS This work was supported by the Medical Research Council of Canada. I sincerely thank Dr. Brian Robertson for his helpful discussions. REFERENCES Arsenault, A.L. (1988) Crystal-collagen relationships in the calcified turkey leg tendon visualized by selected-area dark field electron microscopy. Calcif. Tissue Int., 43:202-212.
Arsenault, A.L. (1989) A comparative electron microscopic study of apatite crystals in collagen fibrils of rat bone, dentin and calcified turkey leg tendons. Bone Mineral, 6:165-177. Arsenault, A.L. (19911Image analysis of collagen-associated mineral distribution in cryogenically prepared turkey leg tendons. Calcif. Tissue Int., 48:56-62. Berthet-Colominas, C., Miller, A,, and White, S.W. (1979) Structural study of the calcifying collagen in turkey leg tendons. J. Mol. Biol., 134:431-445. Bigi, A,, Ripamonti, A., Koch, M.H.J., and Roveri. N. (1988) Calcified turkey leg tendon as structural model for bone mineralization. Int. J . Biol. Macromol., 10:282-286. Chapman, J.A., and Hulmes, D.J.S. (1984) Electron microscopy of the collagen fibril. In: Ultrastructure of the Connective Tissue Matrix. A. Ruggeri and P.M. Motta, eds. Martinus Nijhoff, Boston, pp. 1-33. Doyle, B.B., Hukins, D.W.L., Hulmes, D.J.S., Miller, A,, and Woodhead-Galloway, J. (1975) Collagen polymorphism: its origins in the amino acid sequences. J. Mol. Biol., 91:79-99. Eanes, E.D., Lundy, D.R., and Martin, G.N. (1970) X-ray diffraction study of the mineralization of turkey leg tendon. Calcif. Tissue Res., 6:239-248. Engstrom, A. (1966) Apatite-collagen organization in calcified tendon. Exp. Cell Res., 43:241-245. Hodge, A. (19701Biology of Hard Tissues. ed. A.M. Budy. Gordon and Breach, New York, pp. 68-86. Hodge, A.J., and Petruska, J.A. (19631 Recent studies with the electron microscope on ordered aggregates of the tropocollagen macromolecule. In: Aspects of Protein Structure. G.N. Ramachandran, ed. Academic Press, New York, pp. 289-300. Hodge, A.J., and Schmitt, F.O. (1960) The charge profile of the tropocollagen macromolecule and the packing arrangement in nativetype collagen fibrils. Proc. Natl. Acad. Sci. U.S.A., 46:186-197. Hulmes, D.J.S., Miller, A., Parry, D.A.D., Piez, K.A., and WoodheadGalloway, J. (1973)Analysis ofthe primary structure of collagen for the origins of molecular packing. J. Mol. Biol., 79:137-148. Kobayshi, K., Ito, T., and Hoshino, T. (1986) Correlation between negative staining pattern and hydrophobic residues of collagen. J. Electron Microsc. (Tokyo), 35:272-275. Krefting, E.-R., Barckhaus, R.H., Hohling, H.J., Bond, P., and Hosemann, R. (1980) Analysis of the crystal arrangement in collagen fibrils of mineralizing turkey tibia tendon. Cell Tissue Res., 205: 485-492. Landis, W.J. (1986)A study of calcification in the leg tendons from the domestic turkey. J. Ultrastruct. Res., 94:217-238. Maitland, M.E., and Arsenault, A.L. (1991) A correlation between the distribution of biological apatite and amino acid sequence of type I collagen. Calcif. Tissue Int., 48:341-352. Meek, K.M., Chapman, J.A., and Hardcastle, R.A. (1979) The staining pattern of collagen fibrils: improved correlation with sequence data. J . Biol. Chem., 254:10710-10714. Meyers, H.M., and Engstrom, A. (1965) A note on the organization of hydroxyapatite in calcified tendons. Exp. Cell Res., 40:182-185. Nimni, M.E., and Harkness, R.D. (1988) Molecular structure and functions of collagen. In: Collagen, Vol. 1. M.E. Nimni, ed. CRC Press, Inc., Boca Raton, FL, Chp. 1, pp. 5. Nylen, M.U., Scott, D.B., and Mosley, V.M. (1960) Mineralization of turkey leg tendon. 11. Collagen-mineral relationships revealed by electron and x-ray microscopy. In: Calcification in Biological Systems. Sognnaes, R.F., ed. Amer. Assoc. Adv. Sci., Washington, DC, pp. 129-142. Tzaphilidou, M., Chapman, J.A., and Al-Samman, M.H. (1982a) A study of positive staining for electron microscopy using collagen as a model system-11. Staining by uranyl ions. Micron, 13:133-145. Tzaphilidou, M., Chapman, J.A., and Meek, K.M. (1982bl A study of positive staining for electron microscopy using collagen as a model system-I. Staining by phosphotungstate and tungstate ions. Micron, 13:119-131. von der Mark, K., Wendt, P., Rexrodt, F., and Kuhn, K. (1970) Direct evidence for a correlation between amino acid sequence and cross striation pattern of collagen. FEBS Lett., 11:105-108. Weiner, S., and Traub, W. (19861Organization of hydroxyapatite crystals within collagen fibrils. FEBS Lett., 206:262-266. White, S.W., Hulmes, D.J.S., Miller, A., and Timmins, P.A. (19771 Collagen-mineral axial relationship in calcified turkey leg tendon by x-ray and neutron diffraction. Nature, 266:421-425.
Image Analysis of Mineralized and Non-Mineralized Type I Collagen Fibrils A. LARRY ARSENAULT Electron Microscopic Facility, Faculty of Health Sciences, MeMaster University, Hamilton, Ontario L8N 325 Canada
KEY WORDS
Collagen, Mineral, Calcification, Tendon, Electron microscopy
ABSTRACT
Turkey leg tendons at a n early stage of mineralization have been thin sectioned and imaged by electron microscopy. At this stage collagen-associated mineral apatite was found to be present within both the gap and overlap zones. The earliest apatite occurs in a microcrystalline form which gives a rather generalized and characteristic density to both the gap and overlap zones; with subsequent development larger defined apatite crystals arise which span gapioverlap zones. Fourier transformation of such images revealed the major 67 nm axial repeat of the gapioverlap zone plus four other maxima corresponding to repeat spacings of 22,16,13, and 11nm respectively. Computer imaging techniques were used to reconstruct images by using selected spatial frequencies from such transforms. In this manner the subperiodic distributions of mineral were visually enhanced. These subperiodicities are positioned in a n asymmetric fashion over the entire D unit repeat aligning with the molecular orientation of the fibril. Analyses of both negatively stained collagen and computer-generated maps of collagen hydrophobicity were compared to the mineral distribution of collagen. Densitometric comparisons showed a positional correlation between the axial banding patterns of mineralized fibrils and those of negatively stained non-mineralized fibrils. Comparable spatial frequencies were also present in transforms between hydrophobic maps and mineral distribution of collagen. These results suggest that the lateral clusterings of hydrophobic residues which span the fibril at specific sites in both the gap and overlap zones serve to prohibit early mineral deposition. This observed hydrophobic influence in combination with the gap space appear as contributing factors in the observed axial distribution of mineral within collagen.
INTRODUCTION Mineralized turkey leg tendons have been extensively used a s a n idealized model for the study of interrelationships between collagen and apatite crystals; this is due to the tendon's linear alignment of collagen fibrils as compared to the complex and varying patterns of those in bone and dentin. This is further extended by the fact that there is no evidence to indicate any molecular variation among these type I collagens. The axial periodicity of these mineralized fibrils has been well documented by using electron microscopy (Nylen et al., 1960; Meyers and Engstrom, 1965; Krefting et al., 1980; Landis, 1986; Weiner and Traub, 1986; Arsenault, 1988) and diffraction analysis (Meyers and Engstrom, 1965; Engstrom, 1966; Eanes et al., 1970; White et al., 1977; Berthet-Colominas et al., 1979; Bigi et al., 1988). These diffraction studies have reported the major axial banding repeat to be 66 to 67 nm which is the same as bone. Such banding repeats have been explained by mineral occupying the potential space present in the gap zone (Hodge and Petruska, 1963; White et al., 1977). This gap space is due to the axial ordering of collagen molecules (linear rods 300 x 1.4 nm) where the ends of each molecule are separated by a distance of 35 nm (Hodge and Schmitt, 1960; Hodge and Petruska, 1963; Hodge, 1970). The lateral stacking order is such that adjacent molecules are staggered by 67 nm spacing (234 amino acid residues); this spacing
D 1991 WILEY-LISS, INC
is commonly referred to as the D stagger or D unit. Thus, the gap zone is comprised of one gap space and four adjacent molecules of collagen, while each overlap zone contains a continuous lateral stacking of adjacent molecules. With the use of selected-area dark-field electron microscopy which enables the specific and direct localization of apatite, it has been illustrated that the major portion of apatite in collagen is indeed within the gap zone; however, a considerable amount of apatite is also present in the overlap zone even a t the early onset of mineral deposition (Arsenault, 1988). The relative amounts of apatite within the gap and overlap zones depend upon the stage or degree of mineralization (Arsenault, 1989). In further examination of this crystalcollagen interrelationship computer-generated arrays of type I collagen amino acid sequences (al-ol2-al)were constructed; such arrays can be aligned with electron microscopic images of mineralized leg tendon (Maitland and Arsenault, 1991).By such a comparison it was found that hydrophobic amino acid residues possessed a n inverse relationship with sites of early mineral deposition. Moreover, these hydrophobic domains occur as
Received J u n e 8, 1990; accepted in revised form August 29, 1990. Address reprint requests to Dr. Larry Arsenault, Electron Microscopic Facility, Faculty of Health Sciences, McMaster University, 1200 Main Street West, Hamilton, Ontario L8N 325, Canada.
TYPE I COLLAGEN
distinct axial crossbandings and their positions coincide with areas which are low in mineral content or have excluded mineral. Importantly, this work has demonstrated apatite to occur as axial crossbandings along the D unit in both the gap and overlap zones. Thus, in addition to the 67 nm repeats, mineral was distributed along the D unit as axial subperiodicities. These occur in an asymmetric fashion, and thereby, we have been able to determine the molecular polarity of unstained, mineralized collagen fibrils. This is particularly important considering that adjacent fibrils within the leg tendon may have opposite molecular orientations (Arsenault, 1991; Maitland and Arsenault, 1991). In order to enhance and clarify the presence of these subtle subperiodic bands Fourier transformation and computer imaging techniques have been utilized in the present study. Also, densitometric comparisons were made between the axial crossbandings of mineralized collagen and those produced by negative staining of non-mineralized collagen. To further characterize the correlation between early mineral deposition and hydrophobicity as previously reported, Fourier transformations were performed on mineralized collagen and computer-generated maps of collagen hydrophobic amino acid residues.
MATERIALS AND METHODS Electron Microscopy Portions of early mineralizing leg tendons from 14 week-old domestic turkeys were placed in 100% glycerol for 3 h. Areas of interest at the mineralization front were removed and placed in a graded series of glycerolimethanol and then embedded in Spurr’s resin. Thin sections were cut in longitudinal planes along the tendons with a diamond knife. Non-mineralized portions were examined after positive staining with aqueous 1%phosphotungstic acid (PTA) a t pH 3.2 followed by 1%uranyl acetate (UA) at pH 4.2 while mineralized portions were not stained. All sections were viewed in a Philips 300 operating at 80 kV by using a 20 pm objective aperture. Image Analysis Images were digitized with a Dage video camera (Dage-MTI, Inc., Michigan City, IN) at standardized settings for both magnification and contrast levels. Axial periodicities of non-mineralized and mineralized collagen were Fourier filtered by using a computer image analysis system (IBAS Kontron; Etching, West Germany). For this process 512 x 512 pixel matrices of original negatives (magnification of 19,000 times) were digitized and Fourier transformed. Masks were constructed directly over the transforms of both the stained, non-mineralized collagen and the mineralized collagen to insure the selection of specific spatial frequencies. The resulting masks were rod shaped with outward-curved ends and were large enough to fit over all of the Fourier maxima. These masks were then lowpass Gaussian filtered by using a matrix size of 5 x 5 pixels. For the stained, non-mineralized collagen the mask was contrast reversed to exclude the spatial frequencies which correspond to the axial repeat banding;
263
this was followed by inverse Fourier transformation to obtain an image in which the axial repeats have been removed. The Fourier transforms of mineralized collagen were filtered to either exclude (as above) or include the Fourier maxima. In the latter case, the inversed Fourier transform, having only axial repeats, was either subtracted or divided from the native image. The subtraction or division of these image sets gave qualitatively the same result. A computer-generated map of collagen fibril hydrophobicity as previously described (Maitland and Arsenault, 1991)was Fourier transformed for comparison with the transforms of mineralized collagen. Densitometric tracings of grey values were plotted along the D units of mineralized collagen and of negatively stained non-mineralized type I collagen. The negatively stained fibrils were taken from published micrographs (Chapman and Hulmes, 1984; Nimni and Harkness, 1988). In order to reduce minor localized variations in either the mineral or stain distribution the images were lowpass filtered by using a 5 x 2 pixel matrix. To achieve this the long axis of the filter was passed parallel to the lateral direction of the fibril.
RESULTS A longitudinally sectioned and stained type I collagen fibril from a non-mineralized portion of the turkey leg tendon is shown in Figure l a . In this image 12 crossbands are shown to axially repeat along the fibril; these bands have been labelled with the “e to a” nomenclature (Hodge and Schmitt, 1960)--“e” having two bands, “d” one band, “c” three bands, “b” two bands, and “a” four bands. The “e to a” direction corresponds to the amino- to carboxy-terminus orientation of collagen molecules; therefore in this image the N- to C-direction of collagen molecules is from left to right. The areas defined as the gap and overlap zones are also labelled. (The gap zone consists of “a3,2,1,e2,3rd, and c3” bands and the overlap zone consists of b2,1 and a4”.j A subtle ultrastructural feature of these type I fibrils is the presence of axially oriented striations in the gap zone. These striations are apparent in the gap zone (“a” and “e”) as lucent spaces; further into the gap zone the striations are less evident because of the generalized weaker staining regions bordering the d band. In order to enhance these subtle striations, the original image was Fourier filtered to remove the repeating axial crossbands (Fig. lb). In this computer-displayed image the aligned striations are clearly evident ranging from 5 to 8 nm in width. An unstained mineralized fibril is presented in Figure l c ; this image, like that of stained collagen, represents a map of electron transparency, the contrast of which is due to the electron-scattering ability of the specimen integrated through its thickness. For unstained mineralized collagen the inherent contrast results from the presence of apatite in both the gap and overlap zones (Arsenault, 1988). Characteristically, the mineral density in such overprojected images appears in early stages of mineral deposition as a punctate distribution which defines the gaploverlap zones and their interzone banding or in later stages as dense linear localizations which are not restricted to a par-
264
A.L. ARSENAULT
Fig. 1. a: A stained non-mineralized collagen fibril from the turkey leg tendon showing the characteristic axial crossbanding pattern which has been labelled with the e to a nomenclature; the gap (GZ) and overlap (OZ) zones are marked. A few o f the subtle axial striations present in the ,-e2,1 bands of the gap zone are labelled by arrowheads. The white arrowheads point out the banded alignments of these striations b: A Fourier filtered image in which the axial crossbandings o f a have been removed; this method serves to highlight the axial striations (arrowheads). The black arrowheads correspond to the same areas of the gap zone as the white arrowheads in a. c: An unstained mineralized collagen fibril characterized by alternating gap (GZ) and overlap (OZ) zones having an axial periodicity of
67 nm. The lucent spaces of the al-e2 bands (white arrowheads) are asymmetrically located nearer the N-terminal position of the gap zone; thus the N- to C-molecular orientation of the fibril is from left to right. Mineral density is found within both the gap and overlap zones. Two distinct qualities of mineral apatite are demonstrated: in the upper portion of the image the punctate distribution of apatite microcrystals clearly defines the gap and overlap zones and the interzone banding; in contrast, the lower portion contains, in addition to these microcrvstals. elongated auatite crvstals (arrows) which are not confined to" a particular zone b u t can "span the gapioverlap borders a,b, x 250,000; c, x 253,000.
TYPE I COLLAGEN
ticular zone. The axial periodicities of the mineralized sites are due to elevated levels of mineral accumulation in the gap zone as compared to the overlap zone; this is particularly evident in early stages of mineralization. In this image apatite is predominantly in the form of small microcrystallites. Collectively over the section thickness, these microcrystallites have a recognizable alignment which is specifically related to collagen. They are aligned in a spatial overprojection at different concentrations in the gap and overlap zones. Also, within the gap zone there is an asymmetrically placed lucent band at the a1-e2position indicating a reduction of mineral (Fig. l c ; white arrowheads), while the overlap zone mineral deposition is less dense along the borders with the gap zones (c,-c, and a4-a3).This is clearly illustrated by densitometric tracings of mineralized collagen (Fig. 4aj. The asymmetrically placed lucent band in the gap zone ( a l e 2 ) lies nearer the amino terminus side of the gap zone (Maitland and Arsenault, 1991) and thereby the N- to C-molecular polarity of the fibril can be determined. In this image (Fig. lc) the Nto C-orientation is from left to right. In addition, within these mineral densities larger and more distinguishable apatite crystals are observed; such crystals are believed to represent a more advanced stage of crystal growth. Figure 2a is the same specimen area as Figure l c but shown at a lower magnification and has been rotated 90" clockwise with the N- to C-orientation from top to bottom. This image has been used for Fourier analysis and subsequent computer image enhancements. Figure 2b is a Fourier transformation of Figure 2a showing the characteristic spatial frequencies of mineralized collagen. The first Fourier component through to the sixth represent the major axial periodicities of 67 nm and a series of subperiodicities 22, 16, 13, and 11 nm respectively. Absence of the second Fourier component is presumably due to the periodic square wave created by the high degree of contrast between the gap and overlap zones. These five Fourier components along with the zero spatial frequency were masked from the transform and were then inverse transformed to produce the image shown in Figure 2c. By this Fourier filtering process Figure 2c contains structural information which arises predominantly from the axial repeats caused by sites of microcrystal deposition. This image has been further computer enhanced in Figure 2d to show more clearly the major 67 nm banding and the coexisting subperiodicities. The random and non-periodic distribution of mineral in Figure 2a is readily appreciated by the subtraction of Figures 2a and c, the resultant image is Figure 2e showing the distribution of the larger apatite crystals which are not periodically arranged. When studying the structure of type I collagen fibrils it becomes apparent that mineralized fibrils and negatively stained, non-mineralized fibrils have similar axial periodicities. Comparisons by densitometric tracings of these fibrils are shown in Figure 3 where a is from an early stage of mineralization in the leg tendon, b from a negatively stained fibril by Chapman and Hulmes (1984; Fig. lob), and c is from another negatively stained fibril by Petruska published by Nimini
265
and Harkness (1988; Fig. 2). The important feature of these density profiles is not the actual grey-scale values for this depends upon several variables unrelated to the chemistry of collagen; the salient feature is, however, the similarity in location of peaks and troughs within the D unit. This corresponding spatial positioning of mineral and stain deposition may, in part, be influenced by hydrophobic interactions. Figure 4a is a computer-generated map of collagen hydrophobicity as described by Maitland and Arsenault (1991). The hydrophobic amino acid residues appear as white positional representations of number densities on a black background. Fourier transformations (Fig. 4b) of such a hydrophobic map are shown for comparison with a transform of mineralized collagen (Fig. 4c). The hydrophobic maps were scaled to align with images of mineralized collagen; this alignment was to insure the same distancelpixel between the maps and the mineralized collagen. The transform of the hydrophobic map (Fig. 4b) contained an increased number of spatial frequencies as expected; however, five of the major frequencies are in alignment with the spatial frequencies of mineralized collagen (Fig. 4c). This illustrates that their periodicities are to some degree comparable.
DISCUSSION Early sides of apatite deposition in type I collagen fibrils of the turkey leg tendons have been studied with the use of image enhancement techniques. Also, analysis of both negatively stained collagen and computergenerated maps of collagen hydrophobicity have been compared to the mineral distribution within collagen. This study indicates that: 1) the early deposition of apatite microcrystals within type I collagen occurs as distinct subperiodicities in addition to major 67 nm period; 2) these subperiodicities are positioned in an asymmetric fashion over the entire D unit repeat and not restricted to the gap zone; 3) similar crossbanding patterns occur in both mineralized fibrils and in those non-mineralized fibrils which have been negatively stained; and 4) corresponding subperiodic crossbandings are present in mineralized collagen and computergenerated maps of hydrophobic amino acid residues. The high degree of correlation among these results appears extraordinary when one considers the extremely varied conditions under which these comparisons have been made. For example, the axial periodicity of mineralized fibrils in routine thin sections (Fig. l a ) typically contain 40-50 collagen molecules stacked through the thickness of the sample; in contrast, negatively stained isolated fibrils contain a minimum of four times this amount, while the computer-generated maps of hydrophobicity are displayed as only one molecular layer. Also, the different operative factors of imaging can influence the basic data from which the comparisons were made; such factors are: sample preparation, degree of staining, extent of mineral accumulation, plane of section and tilt angle in the microscope, degree of focus, radiation damage, alignment, and stacking orders. However, these factors appear to have little effect given the consistent observations of corresponding subperiodicities throughout the D unit under the different study conditions. This reflects the high
A.L. ARSENAULT
266
Fig. 2. a: A computer-displayed image is of the same specimen area as Figure l c hut shown at a lower magnification and rotated by 90". The N C orientation is from top to bottom as indicated by the lucent a,-ez spaces (arrowheads). The aggregates of large apatite crystals are indicated by arrows. Non-mineralized lucent area (*). b A Fourier transformation of a showing the typical ordered spatial frequencies of mineralized collagen. The ordered frequencies 1 and 3 to 6 were used for the inverse transformation to obtain image (c).c: A
filtered Fourier image showing only the axial repeats of mineral in both the gap and overlap zones. d With further image enhancements of c by adjusting contrast levels and edge contours this figure more clearly shows the subperiodicities in both the gap and overlap zones. e: This image was formed by the subtraction of a and c; thus, by removing the axial repeats from the native image this figure illustrates the non-repeating mineral distribution contained within these fibrils. Aggregates of large apatite crystals (arrows). x 120,000.
degree of molecular ordering within collagen fibrils and its controlling influence over apatite placement particularly at early stages of mineralization.
The defined localization of mineral within both the gap and overlap zones implicates factors which are considered to be both architectural and chemical in na-
-f
267
TYPE I COLLAGEN Overlap zone
a
'
Hlnerallzed Leg Tendon
c]
b
4a
Gap zone
' c
Negatively Eitatned Type 1 Collagen
Fig. 3. Densitometric tracings across a single D unit in the N- to C-molecular direction of (a)mineralized collagen from turkey leg tendon and (b,c) two published examples of negatively stained type I collagen fibrils [18,19]. In the mineralized fibril (a) the peaks have been labelled and these correspond to specific bands, 1 = cz-cl, 2 = b,-b,, 3 = a*-a3,4 = a l e 2 ,5 = d-c,. Opacity relates to the increased deposition of either mineral or stain while lucency reflects their reduction. Each tracing has corresponding peaks and troughs a t spatially similar locations, with the exception of b having three peaks instead of two in the gap zone (a,cf; this extra peak in b labelled * corresponds to the e,-d location. Importantly, the asymmetrically positioned lucent space in the gap zone which corresponds to e2-a3 is present in both the mineralized and the stained fibrils.
ture. These factors may act in concert with one another to compartmentalize the fibril into a series of axial crossbandings which allow for varying degrees of mineral deposition throughout the gaploverlap zones. For example, the major known architectural feature of collagen fibrils which allows for the accommodation of mineral is the 35 nm gap space at the ends of linear molecules. As suggested and demonstrated by Hodge and co-workers (1960, 19631, the gap space is the most likely location for mineral accumulation based on the availability of potential space in which apatite crystals can fit. In fact, Hodge has postulated that the gap spaces are channels into which apatite crystals may be deposited (Hodge, 1970). This gap zone location for mineral has been generally supported by others using a variety of techniques. However, the fact that mineral is present in both the gap and overlap zones a t the earliest detectable stages precludes defining the gap zone as the sole site of apatite nucleation (Arsenault, 1988, 1989, 1991; Maitland and Arsenault, 1991). Functionally, the gap zone is an area which can accommodate more mineral than the bordering overlap zones and may facilitate in the diffusion of ions into collagen fibrils. Another aspect of our limited understanding about the gap zone is the extent of lateral alignment of adjacent gap spaces. In this regard, Hodge (1970) has constructed models of D staggered collagen molecules in which continuous channels are created by the alignment of gap spaces. However, the true nature and architecture of the resulting intrafibrillar channels are not known. It is tempting, however, to speculate that
e d c b a
Fig. 4. a: A computer-generated map of hydrophobic amino acid groups with the N- to C-molecular orientation from left to right as denoted by the e to a notation. This map is displayed in reverse contrast with white being the spatial representations of clustered crossbandings of hydrophobic residues on a black background. b,c These are Fourier transformations of hydrophobic computer-generated maps and mineralized collagen respectively. For this comparison the hydrophobic map was reduced so that the gapioverlap zones aligned with that of the mineralized collagen. The arrowheads in the transform of aligned hydrophobic maps (b) occur at the same spatial frequencies as those for mineralized collagen (c).
the lucent and aligned axial striations of the gap zone (Fig. la,b) may arise from the alignment of gap spaces; evidence for interconnections among these channels was not apparent. Localized mineral accumulations are noted laterally across the gap zone (Fig. lc); however, at present there is little evidence to suggest that such accumulations occur at sites analogous to the axial striations observed in positively stained fibrils. Ultrastructurally, the mineral apatite appears in two distinct forms depending upon the stage of deposition. In the early stages apatite is microcrystalline, having a fine granular t o punctate nature. The distribution of this mineral form defines the 67 nm period and the associated subperiodicities. With subsequent development larger apatite crystals arise within the areas occupied by the microcrystalline apatite and thereby occur in both gap and overlap zones (Fig. lc). In such overprojected images the true structural features of individual crystals are not revealed; however, their distribution and increased densities as related to mineral accumulation and crystal growth are clearly determined. From a chemical perspective, the distribution of charged and hydrophobic residues is not uniform along the collagen molecule; it has been shown that interactions between each group plus their combined interactions serve to give the preferred D unit (234 amino acid) stagger in collagen assembly (Hulmes et al., 1973). In this staggered alignment these groups are organized into clusters with distinct axial crossbanding patterns visualized at the electron microscopic level with either positive or negative staining. The axial banding pattern derived by the positive staining of charged amino acids has been correlated with sequence data (von der Mark et al., 1970; Meek et al., 1979; Tzaphilidou et al., 1982a,b). Similarly, the negative staining properties of collagen have been demonstrated to reflect the distribution of clustered hydrophobic res-
268
A.L. ARSENAULT
idues (Doyle et al., 1975; Kobayshi et al., 1986). In the present study the negative influence of hydrophobic residues on the distribution of apatite in mineralized collagen has been studied by comparative analyses with negatively stained non-mineralized type I collagen fibrils and hydrophobic computer-generated maps of type I collagen. Areas of reduced mineral density correspond positionally to axial bandings in non-mineralized fibrils which were negatively stained (Fig. 3); this similarity of crossbanding can be explained on the basis of hydrophobic repulsion of apatite and stain deposition. Mineral reduction or exclusion at the major sites of hydrophobic clustering is further substantiated by Fourier analysis with the comparable spatial frequencies for the hydrophobic maps and mineralized collagen (Fig. 4). Based on these studies of mineralized collagen, the gap zone like the overlap zone is viewed as becng compartmentalized by the lateral alignment of hydrophobic domains. As judged by the consistent reduction of mineral a t sites having a high degree of hydrophobicity, it appears that such domains exert a negative effect upon mineral deposition. These repulsive effects on mineral deposition are significant in the gap zone a t the a l e 2 site and in the overlap zone a t the c2-cl and a4-a3sites a t early stages of mineralization (Maitland and Arsenault, 1991). At these sites the hydrophobic domains completely span the width of the fibril. Other hydrophobic regions appear as small “islands” and therefore would have a reduced capacity to effect the visualized distribution of mineral. However, i t appears that with continued growth and deposition the apatite crystals can overcome, to varying degrees, this negative hydrophobic inf hence. This is evidenced a t more advanced stages of mineralization where longer individual crystals are found to course through such subperiodicities and can project from one zone into another (Fig. lc). In conclusion, the findings of this report together with our previous studies indicate that the distribution of mineral apatite in type I collagen is governed by both the architectural and chemical components of the fibril. At early stages of deposition the mineral is, to a high degree, restricted from specific axial crossbandings by interactions with clustered hydrophobic residues; this appears analogous to the mechanism of negative staining. The initial sites of deposition and subsequent sites of apatite crystal growth are not simply confined to the gap zone as is generally accepted but occur in both the gap and overlap zones as a function of the mineralization gradient.
ACKNOWLEDGMENTS This work was supported by the Medical Research Council of Canada. I sincerely thank Dr. Brian Robertson for his helpful discussions. REFERENCES Arsenault, A.L. (1988) Crystal-collagen relationships in the calcified turkey leg tendon visualized by selected-area dark field electron microscopy. Calcif. Tissue Int., 43:202-212.
Arsenault, A.L. (1989) A comparative electron microscopic study of apatite crystals in collagen fibrils of rat bone, dentin and calcified turkey leg tendons. Bone Mineral, 6:165-177. Arsenault, A.L. (19911Image analysis of collagen-associated mineral distribution in cryogenically prepared turkey leg tendons. Calcif. Tissue Int., 48:56-62. Berthet-Colominas, C., Miller, A,, and White, S.W. (1979) Structural study of the calcifying collagen in turkey leg tendons. J. Mol. Biol., 134:431-445. Bigi, A,, Ripamonti, A., Koch, M.H.J., and Roveri. N. (1988) Calcified turkey leg tendon as structural model for bone mineralization. Int. J . Biol. Macromol., 10:282-286. Chapman, J.A., and Hulmes, D.J.S. (1984) Electron microscopy of the collagen fibril. In: Ultrastructure of the Connective Tissue Matrix. A. Ruggeri and P.M. Motta, eds. Martinus Nijhoff, Boston, pp. 1-33. Doyle, B.B., Hukins, D.W.L., Hulmes, D.J.S., Miller, A,, and Woodhead-Galloway, J. (1975) Collagen polymorphism: its origins in the amino acid sequences. J. Mol. Biol., 91:79-99. Eanes, E.D., Lundy, D.R., and Martin, G.N. (1970) X-ray diffraction study of the mineralization of turkey leg tendon. Calcif. Tissue Res., 6:239-248. Engstrom, A. (1966) Apatite-collagen organization in calcified tendon. Exp. Cell Res., 43:241-245. Hodge, A. (19701Biology of Hard Tissues. ed. A.M. Budy. Gordon and Breach, New York, pp. 68-86. Hodge, A.J., and Petruska, J.A. (19631 Recent studies with the electron microscope on ordered aggregates of the tropocollagen macromolecule. In: Aspects of Protein Structure. G.N. Ramachandran, ed. Academic Press, New York, pp. 289-300. Hodge, A.J., and Schmitt, F.O. (1960) The charge profile of the tropocollagen macromolecule and the packing arrangement in nativetype collagen fibrils. Proc. Natl. Acad. Sci. U.S.A., 46:186-197. Hulmes, D.J.S., Miller, A., Parry, D.A.D., Piez, K.A., and WoodheadGalloway, J. (1973)Analysis ofthe primary structure of collagen for the origins of molecular packing. J. Mol. Biol., 79:137-148. Kobayshi, K., Ito, T., and Hoshino, T. (1986) Correlation between negative staining pattern and hydrophobic residues of collagen. J. Electron Microsc. (Tokyo), 35:272-275. Krefting, E.-R., Barckhaus, R.H., Hohling, H.J., Bond, P., and Hosemann, R. (1980) Analysis of the crystal arrangement in collagen fibrils of mineralizing turkey tibia tendon. Cell Tissue Res., 205: 485-492. Landis, W.J. (1986)A study of calcification in the leg tendons from the domestic turkey. J. Ultrastruct. Res., 94:217-238. Maitland, M.E., and Arsenault, A.L. (1991) A correlation between the distribution of biological apatite and amino acid sequence of type I collagen. Calcif. Tissue Int., 48:341-352. Meek, K.M., Chapman, J.A., and Hardcastle, R.A. (1979) The staining pattern of collagen fibrils: improved correlation with sequence data. J . Biol. Chem., 254:10710-10714. Meyers, H.M., and Engstrom, A. (1965) A note on the organization of hydroxyapatite in calcified tendons. Exp. Cell Res., 40:182-185. Nimni, M.E., and Harkness, R.D. (1988) Molecular structure and functions of collagen. In: Collagen, Vol. 1. M.E. Nimni, ed. CRC Press, Inc., Boca Raton, FL, Chp. 1, pp. 5. Nylen, M.U., Scott, D.B., and Mosley, V.M. (1960) Mineralization of turkey leg tendon. 11. Collagen-mineral relationships revealed by electron and x-ray microscopy. In: Calcification in Biological Systems. Sognnaes, R.F., ed. Amer. Assoc. Adv. Sci., Washington, DC, pp. 129-142. Tzaphilidou, M., Chapman, J.A., and Al-Samman, M.H. (1982a) A study of positive staining for electron microscopy using collagen as a model system-11. Staining by uranyl ions. Micron, 13:133-145. Tzaphilidou, M., Chapman, J.A., and Meek, K.M. (1982bl A study of positive staining for electron microscopy using collagen as a model system-I. Staining by phosphotungstate and tungstate ions. Micron, 13:119-131. von der Mark, K., Wendt, P., Rexrodt, F., and Kuhn, K. (1970) Direct evidence for a correlation between amino acid sequence and cross striation pattern of collagen. FEBS Lett., 11:105-108. Weiner, S., and Traub, W. (19861Organization of hydroxyapatite crystals within collagen fibrils. FEBS Lett., 206:262-266. White, S.W., Hulmes, D.J.S., Miller, A., and Timmins, P.A. (19771 Collagen-mineral axial relationship in calcified turkey leg tendon by x-ray and neutron diffraction. Nature, 266:421-425.
