Domains in proteins and proteoglycans of the extracellular matrix with

functions in assembly and cellular activities

Jiirgen Engel Department of Biophysical Chemistry, Biocenter o f the University o f Basel, Switzerland

Most proteins o f the extracellular matrix ( ECM ), such as the glycoproteins, collagens and proteoglycans, consist of man)' structurally autonomous domains that are often functionally distinct. Consequently these proteins are designated as mosaic proteins. Related domains are often Jound in several different ECM proteins. Domains which are o f importance for assembly have been identified by.fragmentation and other approaches. Triple-stranded coiled-coil domains in laminin and probably also in tenascin and thrombospondin are responsible

fi~r chain selection, a process which may be important.for the jormation o f tissue specific isoforms. Globular domains at the C-terminus o f collagenous domains are essential for the registration of the three chains and triple-helix Jbrmation. Fibrillar assemblies o f these triple helices with constituent globular domains serve important assembly.fimctions in man)' collagens including collagens IV and VI. Many other domains with more specialized .functions in assembly have been identified in laminin, fibronectin and other ECM proteins. Cys-rich domains with either distant or close homology with epidermal growth .[actor are repeated manifold in rod-like regions o f a number o f ECM proteins including laminin, tenascin and thrombospondin. They may serve as spacer elements but as suggested for laminin some domains o f this O'pe may also fimction as signals for cellular growth and d(ffbrentiation. Another important cellular.function common to many ECM proteins is" cell attachment. Several cell attachment sites have been localized in structurally unrelated domains o f the same or o f different ECM proteins.

Ke3words." Extracellar matrix; triple helices; globular domains

Introduction The important role of proteins of the extracellular matrix (ECM) in the development and maintenance of cellular organization has been recognized increasingly in the past decade ~-3. The ECM proteins have been classified as glycoproteins, collagens and proteoglycans. These definitions are, at best, operational in view of the fact that collagens (see for example collagen VI 4) share many non-collagenous domains with glycoproteins and that the core proteins of proteoglycans 5- 7 are often more closely related to glycoproteins 8'9 or even to collagens ~° than to other proteoglycans. Well studied examples of glyco- proteins are laminin 11, nidogen 3 (also called entactin), fibronectin ~ 2.13, tenascin 14, ~ 5 (also referred to as cytotactin, hexabrachion or J1) and thrombospondin. Interesting examples of complex collagens are collagen IV 3 and V P and for proteoglycans the low density basement membrane proteoglycan 9,17 (recently named perlecan) and the large cartilage proteoglycan s'ls. This list could be expanded with many other examples of ECM proteins. Some have

Presented at '1990 IUPAB Satellite Congress', 7-10 August 1990, Palmerston North, New Zealand

been less well characterized and others have not been isolated yet. Almost every month genes are being sequenced that appear to code for ECM proteins. The gap between cDNA derived sequence data and functional and structural information needed to interpret them is increasing.

Many ECM proteins are known as structural components of specialized forms of the ECM but already the existence of tissue specific isoforms and splicing variants as well as focal and transient expression at distinct stages of organ development 2' 13.15 suggest more specific biological activities. Indeed the most important functions of ECM proteins are their interactions with cells. This is reflected in the large number of cellular receptors for ECM proteins or individual domains which have been found 12'13'19. Members of the large family of integrin receptors acquire their specificity for individual domains in ECM proteins by the variable combination of a few fl subunits with a large number of ~ subunits z°.

Structural, and often functional, autonomous domains can occur as modular units in a range of different ECM proteins. Although these mosaic proteins are complex and their structures are only partly understood, some generalizations (and their limitations) are explored in the present discussion.

0141-8130/91/030147-05 © 1991 Butterworth-Heinemann Limited Int. J. Biol. Macromol., 1991, Vol. 13, June 147

Domains in E C M proteins." J. Engel

R e s u l t s and d i s cus s ion

Domains involved in assembly Schematic representations of the domain organizations

of three important ECM proteins namely laminin, tenascin and thrombospondin are shown in Figure 1. Only single chains are shown in the scheme but in reality all proteins are multimeric assemblies of several chains. Heptad repeat regions (Figure 1) which are capable of forming double and triple stranded coiled-coil structures are involved in their assembly. Only the A chain of laminin is shown in Figure 1 but the intact molecule consists of three chains: A, B! and B2 which assemble into a cruciform structure with three short (34, 34 and 48 nm) and a single long arm (76 nm) 11. Isoforms with genetically distinct A and B chains have been described 11, although molecules which lack an A chain may exist 2. The three chains of the heterotrimer are connected by a triple-stranded coiled-coil domain (see below). Similar domains are apparently responsible for the assembly of tenascin and perhaps thrombospondin. Tenascin exists mainly as a hexameric assembly with six 75-nm long arms (and is therefore also designated hexabrachion) but trimers are also known x4'15. The complex structure of thrombospondin is also formed by three chains 23. It is interesting to note that perlecan, which is closely homologous to the A chain of laminin, lacks the heptad repeat motif which is a prerequisite for the formation of a coiled-coil structure. In this case a different structural motif is present TM. Consequently this protein is a single chain structure aT. Fibronectin which is not represented in Figure 1, but which has been described in excellent texts 12'13 is composed of two arms (50 nm) in which small globular domains, designated type I, II and III are arranged like pearls on a necklace. Fibronectin chains are assembled primarily via C-terminal domains and disulphide bridges but other domains in fibronectin are apparently also involved in internal association 24 leading to a transition from an extended to a compact shape and to fibril formation 25.

Many ECM proteins possess the potential for further assembly to supramolecular structures. Some of the

domains which have been identified as being involved in assembly are listed in Table I. Laminin assembles with nidogen 3'11 (also called entactin), exhibits calcium- dependent self-assembly 26 and probably forms a nidogen

Table I A. Domains common to several ECM molecules with distinct functions in assembly (examples in brackets)

Coiled-coil domains." Chain selection registration and chain assembly. Spacer elements between functional sites (laminin ~ l,

tenascin 2'.22, thrombospondin'6).

Collagenous domains." Spacer elements Connection of monomers to multimers by microfibrillar

assembly (collagen VP, Clq 28, SP-A29). Fibre (collagens 1-11127) and network (collagen IV 26"32)

formations.

Globular domains C-terminal to a collagenous domain: Chain selection, registration and nucleation of triple-helix

folding (all collagens 3°'3' ). Assembly to collagen IV networks (dimerization of NCI

domains in collagen |V3'32'33).

Other globular domains in collagens." Control of fibril formation by selective cleavage (collagens I,

II, III34). Binding to collagen or other ECM components

(collagens IV, VI, VII, IX'*'26"27"32).

B. Examples for specialized domains involved in assembly

A region comprising type I and II domains in the N-terminal region of fibronectin 12'13 and the structurally different von Willebrand factor type A domains in collagen VI 5-7 are believed to be collagen binding domains.

Various different domains are involved in the internal association of fibronectin from an extended to a compact shape and in fibrillogenesis 12,13,24-.2 5.

Heparin binding sites in structurally different domains of laminin 3,,, and of fibronectin 12,13 may mediate interactions with heparan sulphate proteoglycans.

LAMININ

ss

~ 1 I I

1 2 3

( A - C h a i n )

8 G heptads G1- G5

T E N A S C I N (Cha in )

<fF-qO000000000000~ 4heptads 8-11 FN ~r Fibrinogen

T H R O M B O S P O N D I N (Chain)

s

heptads? Ca ++

O = EGF - like

Figure t Schematic representation of the domain organization of three ECM-proteins. Structurally autonomous domains are indicated by different symbols: a small open circle represents an EGF-like domain and a hexagon represents an FN domain. The latter is a type II1 domain first found in fibronectin 12'13. Details for laminin I l, tenascin14.ts and thrombospondin ~6 can be found in the indicated references

148 Int. J. Biol. Macromol., 1991, Vol. 13, June

mediated assembly with collagen IV j ' l 1. These collagen and laminin molecules are the major components of basement membranes. Collagen IV forms a network structure in contrast to typical fibril-forming collagen 3, like collagens I, II and 11127. For the formation of the network a specialized coUagenous domain at the N-terminus of collagen IV (7S domain) forms a short antiparallel microfibrillar assembly of four molecules. At the C-terminus non-collagenous globular domains (NC 1 ) dimerize thus leading to pairwise connections TM. This is a very specific process in which a disulphide rearrangement from intrasubunit to intersubunit bonds is essential 32'33. Lateral association between collagenous domains other than the 7S regions of different molecules are probably essential for the build up of the complex collagen IV network in basement membranes 26.

A second illustration of the involvement of collagenous and globular domains in assembly is provided by collagen VI. Dimers are formed by a staggered antiparallel lateral association of monomers which in turn assemble to tetramers 4. It was proposed that the von Willebrand type A domains found in the N- and C-terminal sequences of all collagen VI chains (el , ~2, ~t3) function as collagen binding domains 5-7 and are involved in this assembly process. This finally leads to microfibrils by association of tetramers via their terminal domains 4. For an overview of other domains involved in assembly see Table 1. Some are common to several ECM proteins and others appear to be unique to individual proteins. Since our knowledge is still incomplete more common principles may emerge from future research. In addition it should be noted that Table 1 contains well studied examples and is not an exhaustive list of all domains for which an assembly function has been proposed.

Coiled-coil formation and assembly Double stranded coiled-coil structures are abundant

structural elements in DNA binding proteins (jun, fos) 35-37, in cytoskeletal proteins (vimentin, myosin, tropomyosin), and in other systems (reviewed by Cohen and Parry39). Triple stranded coiled-coil structures have

Domains in ECM proteins: J. Engel

been found in virus proteins and fibrinogen 39. Recently a model of the macrophage scavenger receptor was proposed in which a triple coiled-coil domain is directly preceded by a collagen triple helix in a rod-like domain *°.

We were recently able to demonstrate that the formation of the coiled-coil domain in the long arm of laminin provides a specific mechanism of laminin assembly 41. Chain assembly was investigated using fragments E8 and C8-9 derived from the long arm of the molecule 11, whose rod-like domain consists of the e- helical regions of the A, B1 and B2 chains. Both fragments are visualized by electron microscopy as lollipop structures with rods of 35 nm (E8) and 78 nm (C8-9) length and a terminal globule which comprises the three globular subdomains G1-G3 of the A chain. Urea induced chain separation and unfolding were demonstrated using transverse urea/polyacrylamide gel electrophoresis (PAGE) and circular dichroism. Furthermore, A chain segments were separated from disulphide linked dimers of the B chains (B1-B2) by chromatography on heparin Sepharose. Reassembly was monitored by non-denaturing PAGE, circular dichroism and electron microscopy.

Results of published* 1 and unpublished work 41 a which were obtained for fragment E8 are summarized in Figure 2. The tail fragment reassembles readily from its component chains from 8 M urea via a triple-stranded coiled-coil. Complete refolding was also achieved after reductive cleavage of the single disulphide bond connecting the two B chains at their C termini. This bond can be correctly reformed by reoxidation under associative conditions. Fragment E8 B1-B2 alone refolds into a rod-like structure comprising a double stranded coiled- coil. Formation of a rod-like and probably double- stranded coiled-coil structure was also observed for a shorter fragment (25 K-fragment) .2 comprising the C-terminal halves of the E8 B1-B2 chains. E8 A chains alone do not reform coiled-coil structures. This is most convincingly demonstrated by electron microscopy. In the refolded material only the globular domains G1-G3 or aggregates linked by these domains can be detected but there is a lack of formation of rod-like structure.

E8 B1 -B2 B1

S-S HS

I

' \ : I I SH (SH 1 k. B 1 + B 2

B2 A

I +

.2

native

unfolded

(3-8M urea)

unfolded

and reduced

Figure 2 Schematic representation of reassembly products and refolding pathways of fragment E8 comprising the C-terminal region of the long arm of laminin. The chain segments B1 and B2 can form a stable double-stranded coiled-coil structure. The corresponding heptad region of the A chain can only associate to this structure thus forming a triple stranded coiled-coil

Int. J. Biol. Macromol., 1991, Vol. 13, June 149

Domains in E C M proteins." J. Engel

These findings may be related to the lower coiled-coil formation potential 39 of the heptad regions in the A chain in comparison with the B chains 11. This also agrees with the notion that isoforms of laminin and biosynthetic intermediates can exist as native molecules without an A chain2.11. The specificity of chain selection may be utilized in the formation of isoforms composed of specialized A and B chains. The specificity can be compared with the selective association of jun and fos by coiled-coil domain35-37. Preliminary experiments indicate that A chains refolded in this way can associate with refolded B chain segments to a native fragment 8 structure. This would provide an interesting example of a transition from a double-stranded to a triple-stranded coiled-coil by addition of a third chain under native conditions. We feel that such a process does not require a complete dissociation of the double-stranded structure and we are currently investigating the kinetics and mechanism of this process. Interestingly, it was possible to reconstitute the highly conformation dependent cellular activities of fragment E8 by mixing its A and B chains under native conditions 43 (see also Table 2 and next section).

Domains in E C M proteins with cellular functions

For most ECM proteins studied so far, important cellular functions have been identified. Numerous studies have been performed with cultured cells demonstrating that ECM components may promote, control or modulate cellular activities such as attachment, spreading, differen- tiation, proliferation, polarization and migration. For a few well studied examples including fibronectin, laminin and tenascin it was possible to assign cellular functions to individual domains. Here we shall focus on cell attachment sites and on growth factor-like activities.

Cell attachment sites have been identified in structurally unrelated domains of different ECM proteins (Table 2). In the pioneering work on fibronectin an Arg-Gly-Asp (RGD) site in one of the type III domains was identified as being essential for attachment and spreading 47. It was also demonstrated that binding of the fibronectin receptor (~5fll integrin) was inhibited by synthetic RGD peptides 48. When it was found that cell attachment by several other ECM proteins could be inhibited by these peptides (RGD dependent attachment) ~9'2° it was tempting to speculate that a general principle had been discovered. As a consequence many putative cell attachment sites were predicted from sequence data. It is now becoming apparent that many cell attachment processes are RGD-independent and that many major sites do not contain this sequence. Apparently even an RGD sequence which occurs in a type Ill domain of tenascin is not involved in attachment. A few examples which prove these points are listed in Table 2. An interesting example for an RGD-independent site is the terminal region of the long arm of laminin. This example also teaches us that cell attachment may depend strongly on the conformation of the multichain structure. The fact that single chains in randomly coiled conformation are inactive, strongly suggests that synthetic peptides tailored after sequences in this region cannot exhibit comparable activities in this case.

It is a striking observation that many ECM proteins, including the ones shown in Figure 1, exhibit Cys-rich domains with close (tenascin, thrombospondin) or distant (laminin, perlecan) homology with epidermal

Table 2 Cell attachment sites in different ECM proteins

Fibronectin t),pe I11 domains: One of the type III domains in fibronectin which contains an

RGD sequence is the major site for RGD dependent binding of the fibronectin receptor (alpha-5-beta-l-integrin). Other type III domains without RGD act synergistically 12'13

The tenth or eleventh type III domain in tenascin which lacks an RGD sequence is active but the third domain with an RGD is probably not.

Cell attachment to tenascin is not inhibited by RGD peptides 15.21.

RGD sites in collagenous domains." The triple helical domain of collagen VI contains many RGD

sequences which are believed to be involved in the attachment activities of alpha 2 and alpha 3 chains

Terminal region of the long arm of laminin consisting o['a coiled- coil domain and globular subdomains G1-G3: A corresponding fragment (T8) contains the major cell

attachment site 3'43 (recognized by alpha-6-beta-l-integrin) and is also active in neurite outgrowth '~3'45. The activity is highly dependent on correct conformation, intactness of disulphide bonds and presence of all constituent chains (T8A and T8 B1-B2). It can be reconstituted by reassembly from these chains 43.

Short arm structures c?[" laminin. The pepsin derived fragment P1 contains an attachment site which is latent in native laminin. It is believed to reside in an RGD-containing EGF-like domain of the A chain which becomes exposed after cleavage of the adjacent domain lVa 46.

growth factor EGF and transforming growth factor TGF~. The latter small diffusible growth factors are important in embryonic development. It is an appealing concept that some of the EGF-like domains in ECM proteins may exhibit localized growth factor activities which may act in a specific and vectorial way on adjacent cells 49. Indeed for laminin, thrombospondin and for tenascin (perlecan was not tested) growth promoting functions have been reported (reviewed in Ref. 49). For laminin, which is the first ECM molecule present in embryonic development, it was possible to localize this function s° in fragment PI which comprises short arm regions of the A and B chains with about 25 EGF-like repeats in toto. Certainly many of the EGF-like repeat units may have lost what was, perhaps, original cell signalling functions during evolution and may just serve a structural spacer function in a rod-like region between functional domains. Further work is needed to localize the active domains in the very large fragment P 1. It should also be mentioned that specific binding sites for small growth factors have been identified in several ECM proteins. Examples are complexes of TGFfl and fibro- nectin 5t and of fibroblast growth factor with heparan sulphate proteoglycans 52. This offers an alternative mechanism for presenting growth factors at specific sites to cells.

Conclusions

Alpha helical coiled-coil domains in a number of ECM proteins serve the functions of chain selection, registration and assembly. They also act as spacer elements between

150 Int. J. Biol. Macromol., 1991, Vol. 13, June

different f unc t i ona l d o m a i n s . E C M pro te ins are of ten ex tended since they m u s t med ia t e be tween d i s t an t sites as, for example , be t ween cells a n d mat r ix . S imi la r spacer a n d a s sembly func t ions are fulfilled by co l l agenous d o m a i n s . M a n y special ized d o m a i n s i n v o l v e d in a s sembly have been ident i f ied in va r ious E C M pro te ins b u t c o m m o n mot i fs or s t r u c t u r e - f u n c t i o n r e l a t i onsh ip are sparse. Th i s also ho lds t rue for the cell a t t a c h m e n t sites which reside in s t ruc tu ra l ly u n r e l a t e d reg ions in different E C M prote ins . The re are, however , i n d i ca t i o n s tha t some of the m a n i f o l d repea ted E G F - l i k e d o m a i n s which are found in a very wide r ange of E C M pro te ins m a y share a cell s igna l l ing func t i on with their g r o w t h factor homologues .

Acknowledgements This work was supported by the Swiss National Science Foundation. I am grateful to D. Noonan, J. Hassell and R. Deutzmann for communicat ing their results to me prior to publication.

References 1 Bernfield, M. Curr. Opinion Cell Biol. 1989, 1, 953 2 Ekblom, P. Faseb J. 1989, 3, 2141 3 Timpl, R. Eur. J. Biochem. 1989, 180, 487 4 Timpl, R. and Engel, J. in 'Structure and Function of Collagen

Types' (Eds R. Mayne and R. E. Burgeson), Extracell. Matrix, Academic Press, New York, 1987, p. 105

5 Koller, E., Winterhalter, K. H. and Trueb, B. EMBO J. 1989, 8, 1073

6 Chu, M.-L., Pan, T., Conway, D., Kuo, H.-J., Glanville, R.W., Timpl, R., Mann, K. and Deutzmann, R. EMBO J. 1989, 8, 1939

7 Bonaldo, P. and Colombatti, A. J. Biol. Chem. 1989, 264, 20235 8 Heinegard, D. and Oldberg, A. Faseb J. 1989, 3, 2042 9 Noonan, D. M., Horigan, E. A., Ledbetter, S. R., V6geli, G., Sasaki,

M., Yamada, Y. and Hassell, J. R. J. Biol. Chem. 1988, 263, 16379 10 Bruckner, P., Vaughan, L. and Winterhalter, K. H. Proc. Natl.

Acad. Sci. USA 1985, 82, 2608 11 Beck, K., Hunter, I. and Engel, J. Faseb J. 1990, 4, 148-160 12 Yamada, K. M. in 'Fibronectin' (Ed. D. F. Mosher), 'Biology of

Extracellular Matrix': A series (Ed. R. P. Mecham), Academic Press, 1989, p. 47

13 Hynes, R. O. 'Fibronectin' Springer Series in Molecular Biology (Ed. A. Rich), Springer Verlag, Berlin

14 Erickson, H. P. and Bourdon, M. A. Annu. Rev. Cell Biol. 1989, 5, 71 15 Chiquet, M. Dev. Neurosci. 1989, 11, 266-275 16 Lawler, J. and Hynes, R. O. J. Cell Biol. 1986, 103, 1635 17 Paulsson, M., Yurchenco, P., Ruben, G. C., Engel, J. and Timpl,

R. J. Mol. Biol. 1987, 197, 297

D o m a i n s in E C M p r o t e i n s : J. Enge l

18 Paulsson, M., M6rgelin, M., Wiedemann, H., Beardmore-Gray, M., Dunham, D., Hardingham, T., Heinegard, D., Timpl, R. and Engel, J. Biochemical J. 1987, 245, 763

19 Ruoslahti, E. Annu. Rev. Biochem. 1988, 57, 375 20 Ruoslahti, E. and Giancotti, F. G. Cancer Cells 1989, 1, 119 21 Spring, J., Beck, K. and Chiquet-Ehrismann, R. Cell 1989, 59, 325 22 Beck, K., Spring, J., Chiquet-Ehrismann, R., Engel, J. and Chiquet,

M. in ' Patterns in Protein Sequence and Structure' (Eds. W. Taylor and P. Argos) Springer Verlag (in press)

23 Lawler, J., Derick, L. H., Connoly, J. E., Chen, J.-H. and Chao, F. C. J. Biol. Chem. 1985, 260, 3762

23a Hassel, J. and Noonan, D. personal communication 24 Markovic, Z., Lustig, A., Engel, J., Richter, H. and H6rmann, H.

Hoppe Seyler's Z. Physiol. Chem. 1983, 364, 1795 25 H6rmann, H. and Richter, H. Biopolymers 1986, 25, 947 26 Yurchenco, P. D. and Schittny, J. C. Faseb J. (in press) 27 Burgeson, R. E. Ann. Rev. Cell Biol. 1988, 4, 551 28 Kilchherr, E., Hofmann, H., Steigmann, W. and Engel, J. J. Mol.

Biol. 1985, 186, 403 29 Voss, T., Eistetter, H., Sch~ifer, K. P. and Engel, J. J. Mol. Biol.

1988, 201, 219 30 D61z, R., Engel, J. and KiJhn, K. Eur. J. Biochem. 1988, 178, 357 31 Prockop, D. and Engel, J. Ann. Rev. Biophys. and Biophys. Chem.

(in press) 32 Martin, G. R., Timpl, R. and Kfihn, K., Adv. Protein Chem. 1988,

39, 1 33 Weber, S., Engel, J., Wiedemann, H., Glanville, R. W. and Timpl,

R. Eur. J. Biochem. 1984, 139, 401 34 Prockop, D. J., Berg, R. A., Kivirikko, K. I. and Uitto, J. in

'Biochemistry of Collagens" (Eds. G. N. Ramachandran and A. H. Reddi) Plenum Press, New York, 1976, p. 163

35 O'Shea, E. K., Rutkowski, R. and Kim, P. S. Science 1989, 243, 538 36 Gentz, R,, Rauscher III, F. J., Abate, C. and Curran, T. Science

1989, 243, 1695 37 Turner, R. and Tjian, R. Science 1989, 243, 1689 38 Cohen, C. and Parry, D. A. D. Tibs 1986, 11,245 39 Parry, D. A. D. Biosci. Rep. 1982, 2, 1017 40 Kodama, T., Freeman, M., Rohrer, L., Zabrecky, J., Matsudaira,

P. and Krieger, M. Nature, 1990, 343, 531 41 Hunter, I., Schulthess, T., Bruch, M., Beck, K. and Engel, J. Eur.

J. Biochem. 1990, 188, 205 42 Paulsson, M., Deutzmann, R., Timpl, R., Dalzoppo, D., Odermatt,

E. and Engel, J. EMBO J. 1985, 4, 309 43 Deutzmann, R., Aumailley, M., Wiedemann, H., Timpl, R. and

Edgar, D. Eur. J. Biochem. 1990, 191,513 44 Aumailley, M., Mann, K., Von der Mark, H. and Timpl, R. Exp.

Cell Res. 1989, 181,463 45 Edgar, D., Timple, R. and Thoenen, H. EMBO J. 1984, 3, 1463 46 Aumailley, M., Gerl, M., Sonnenberg, A., Deutzmann, R. and

Timpl, R. FEBS Lett. 1990, 262, 82 47 Pierschbacher, M. D. and Ruoslahti, E. Nature, 1984, 309, 30 48 Pytela, R., Pierschbacher, M. D. and Ruoslahti, E. Cell 1985, 40,

191 49 Engel, J. FEBS. Lett. 1989, 251, 1 50 Panayotou, G., End, P., Aumailley, M., Timpl, R. and Engel, J.

Cell 1989, 56, 93 51 Fava, R. A. and McClure, D. B. J. Cell Physiol. 1987, 131, 184 52 Rifkin, D. B. and Moscatelli, D. J. Cell Biol. 1989, 109, 1

Int . J. Biol. M a c r o m o l . , 1991, Vol. 13, J u n e 151