Intermediate filaments (IF), actin and microtubules constitute the bulk of the cytoskeleton in most eukaryotic cells. However. unlike the other cytoskeletal components, little is known about the assembly pathway(s) or the structural basis of the interactions between subunits in native IFs. A series of recent have used directed mutagenesis to address these problems and several key aspects of the molecular interactions between IF proteins are beginning to emcrge . IF proteins form an extensive multigene family which can be subdivided into six classes (Table 1) on the basis of tissue specific expression and sequence comparisons (reviewed in ref. 5 ) . Three primary sequence features unite IF proteins (Fig. 1). All possess an extensive central a-helical coiled-coil rod domain. flanked by Nand C-terminal globular domains of variable size and sequence. The rod domain is punctuated at conserved points with three short non-a-hclical sections, so that it can be subdivided into helix IA, IB and helix IIA, IIB. Although the principal molecular interaction in IFs is between rod domains, the globular domains also contribute to filament assembly by both lateral and endto-end interactions. Earlier studies were quite successful in identifying the role of each domain in assembly and their contribution to filament stability as well as interfilament interactions, but were unable to pinpoint specific residues important in molecular interactions. Although some information has been obtained on the relative molecular positions in two-molecule tetramers, an important intermediate in IF assembly(‘), the specific residues that determine this interaction have
not been identified. Indeed, there is a paucity of structural data to direct the use of recombinant methods to dissect the pathway of filament assembly and the higher order molecular interaction geometries in filaments. Expression and mutagenesis of IF cDNAs in both eukaryotic and prokaryotic systems have allowed a more precise investigation of IF molecular interactions to be initiated. In particular, these experiments have drawn attention to small regions of the sequence near the ends of the rod domain that appear to exert a major influence on IF assembly (Fig. 1). One approach has been to introduce specific point mutations at sequence positions that are absolutely conscrved betwecn different IF classes. Hatzfeld and Webed’) monitored the effect on in v i m assembly of point mutations at the C-terminus of helix IIB in keratins 8 and 18. Their data showed that changing such residues had a profound effect upon filament assembly, altering those molecular interactions important for the control of filament length as well as inter-filament interactions. Some effects could be partially reversed by varying the conditions of in vitro assembly and this raises the possibility that several mechanisms exist for filament assembly. Filament assembly in v i m does not require all of the C-terminal globular domain, as this can be reduced to those 9 residues immediately adjacent to the rod(2). These data complement results obtained with pr~teolytic(~) and recombinant desmin fragmend8).A further illustration of the importance of regions near the ends of the rod to filament assembly is the observation that phosphorylation by the cdc2 kinase of serine 16 in chicken lamin B2 appears to depolymerise the nuclear lamina(4). Phosphorylation is also thought to be associated with the assembly and disassembly of other 1Fs and it has been proposed that the cdc2 kinase is also responsible for the cell cycle chan es in cytoplasmic intermediate filament architectures?91 . Other approaches have deleted whole domains in IF proteins or produced chimeric proteins from different IF classes and then monitored the effect on native filament networks by cell transfection. However. interpreting the results of these studies is not always straight-forward because the level of expressed protein
Table 1. Clusses of intermediate filament proteins
Sequence class
Proteins
Tissue specificity
I
Acidic keratins Basic,/neutral keratins Desrnin Vimentin GFAP Peripherin Neurofilaments (XF) (NF-H. NF-M, NF-L in vcitebrates) Nuclear lamins (often A-, B- and C-) Nestin
Epi tlielia Epithelia Muscle Mesenchyme Astrocytes, some glia Peripheral neurones Neurones
I1 111
IV V VI
Nuclear envelopes Neuroepithclial stem cells
-
GLOBULAR DOMAIN
4
I
Assembly critical domain N-Terminal domain
HELIX I
HELIX II
v
I /
a
Coiled-coilforming domains
”
IA
’/
/
’
7
LI
IB
’
L
-
GLOBULAR DOMAIN
1
--b
Assembly influencing domain
I
C-Terminal domain
-
Fig. 1. Schematic illustration of the functional domains of intermediate filament protein sequences. The sequences of IF proteins divide into three distinct domains. The central rod domain is thought to have a mainly cu-hclical conformation, punctuated by Thort stretches (or ‘linker regions’ - LI, LII and LI-11) in which the &-helicalconformation may be lost. Therefore, the rod can be divided into two major helices (IsiTT), and subdivided into helices IA, IB, IIA and IIB. The central rod domain dimerizes to produce a coiled-coil molecule (often refcrrcd to as a ‘dime?). The rod domain is flanked by a Y- and C-terminal globular domains, which are important for corrcct filament assembly, albeit to varying degrees. Two particularly important sites in the protein sequence have been indicated, namcly the cdc2 kinase pliosphorylation site in the N-terminal globular domain lamins, and the highly conserved C-teminal sequence at the end of helix IIB. As illustrated by a range of expcriments using engineered mutant IF proteins (see text), both sites are critically involved in TF assemhly.
and the phenotypic background of the host cells can influence the outcome. Although such variables limit this approach to asking very broadly defined questions, these studies have been particularly successful in helping to delineate the minimum length of protein sequence that permits or disrupts filament assemblies. Chin et al.(’) investigated the influence of N- and Cterminal deletions in neurofilament proteins on the endogenous vimentin filament network of cultured fibroblasts by transient transfection. The two most extensive deletions, which removed all but the first 127 and 139 residues, still caused the collapse of the endogenous vimentin IF network. In both cases, only the N-terminal globular domain and rod helix 1A (the smallest a-helical segment) remained. Conversely, expression of an N-terminal deletion of NF-L, which only left 70 residue? of helix 11B and the globular Cterminal domain, had no deleterious effect on the endogenous filament network. Chin et al. examined a range of deletion mutations and concluded that deletions involving the C-terminus of the rod domain were far more detrimental to filament structure than Nterminal deletions. This is consistent with a range of proteolysis and other experiments (reviewed in ref. 5 ) , although the C-terminal domain may be more important in the nuclear l a m i n ~ ( ~ , ’Small ~ ) . deletions at the rod C-terminus produced dominant negative mutations in keratins(”) and removal of the conserved RLLEGE sequence abolished filament assembly and resulted in the complete disassembly of the endogenous filament network in transfected cells. Similar conclusions were reached using chimeric keratin-vimentin constructs,
which underline the importance of helices IB and IIB in the correct sorting of diffcrcnt 1F p r ~ t e i n s ( l ~ This . ~ ~ )is. consistent with the molecular interaction geometry proposed for tetramerd6) and confirms the importance of helices IA,B and the sequence at the rod C-terminus. Assessing the contribution of different domains to filament integrity in keratin IF assembly is more complicated because it requires an equimolar ratio of type I and type I1 IF proteins. Moreover, the interaction of type 1 and I1 keratins is hierarchical and reflects their preferential expression in different cell types(12). These additional levels of complexity permit several special questions to be asked of this IF assembly system. Lu and Lane(’3) expressed mutated keratin cDNAs in cultured fibroblasts to study keratin pairing and filament formation and showed that, in any pair of interacting keratins, there must be one complete Nterminal and C-terminal globular domain. This requirement was found even when keratin 19, a naturally occurring keratin which lacks a globular C-terminal domain, was paired with a mutated type I1 keratin lacking the same domain. However, this latter result is not entirely consistent with in vifro experiments where filaments could be formed from such a combination under certain conditions(2); this discrepancy indicates the value of concomitant in vitro and in vivo studies. Although these results indicate that an interaction between the N- arid C-tcrminal globular domains of the same protein type in different molecules is necessary for filament formation, it is clear that these domains contribute in different ways to filament structure. Coulomb et aZ.(”) showed that keratin 14, lacking both
these domains, was able to form filaments in vivo and also in vitro with appropiate type I1 partners but the morphology of these filaments was altered, being less regular in diameter. This suggests that there are complementary interactions involving both apposed globular domains in keratin molecules. It is possible that part of this difference could be attributed to the different keratin pairs being studied, given that the keratin 5/14 complex is one of the most stable pairs known, whereas keratins 5/18 are one of the least stable pairs with respect to urea-mediated dissociation. I n another approach, Lu and Lane(13) removed some of the rod sub-domains to show that helix IIA and the linker sequences flanking it were not necessary for complex formation, but were important for the correct cellular distribution of the filamentous aggregates. Another construct in which part of the rod domain spanning residues 139-340 of keratin 18, which includes most of helix IB and IIB, was deleted apparently gave no heterocomplex or filament formation. Heterocomplex formation is requisite for keratin filament assembly, but so far the operative mechanisms that enable mutated type I or 11 keratin to permeate the filament network are unclear. The rules governing the exchange of the various individual protein chains within the protein molecule of two chains as well as the assembly competent form (the molecular dimer) must be delineated. Once these inter-relationships are better understood, we should be nearer to explaining the apparent discrepancies between some of the in vitro and in vivo experiments. Because of the absence of detailed structural data regarding molecular arrangements in lFs, the design and the interpretation of in vivo transfection experiments is sometimes not completely straightforward. It is, for example, difficult to be certain at which lcvel of organisation a mutant IF protein manifests itself. Does the mutation influence the formation of two-chain molecules, four-chain tetramers or any of range of higher assembly aggregates? The generation of dominant negative mutations in many of the transfection studies indicated that at least the phenotypes observed were unlikely to have completely trivial explanations. The potency of these mutations was often very high and some were able to influence filament assembly when present at less than 1 % of the total IF protein concentration("). This indicates that these mutations might affect important and sub-stoichiometric, higherorder interaction geometries needed for filament integrity. Similar observations have been made with the neurofilamen t system(14). Such dominant negative mutations should prove invaluable for studying the function of IFs in vivo. Another noteworthy feature of the transfection studies is that they lend support to the idea that intermediate filament assembly may be dynamic. Although it is possible to explain the reorganisation of the endogenous IF network following tranfection through fragment association with the filament walls and a subsequent destabilisation of the
polymer, the presence of a soluble pool in equilibrium with the filaments appears to be an inescapable conclusion. Another problem is the thorny issue of subunit association versus integration when interpreting the cell transfection data: it is quite possible that the apparently smooth surface of intermediate filaments belies a substructure that is composed of a fascia of ropes that perhaps not only stop and start along the length of the filament but in which the individual subunits are also freely accessible to a soluble pool. Despite the limitations of many of the experiments using IF mutants. a number of firm conclusions are emerging. It is clear that short regions near the ends of the m-helical rod domain are vital for filament assembly and changing these sequences can produce dominant negative mutants; the latter are certain to prove effective tools for investigating IF function and assembly in vivo. These sequences. as well as portions of the globular domains that flank them. not only influence the assembly but also the extent of filament aggregation and filament girth. It is also clear that the N-terminal globular domain is more important than the corresponding C-terminal domain to filament integrity. So far, most experiments have addressed filament assembly as an all-or-nothing process, while more subtle changes have not been investigated in great detail. One way to extend the work along these lines would be to use antibody probes to follow specific epitopes. Mutations that alter filament morphology and architectures should identify sequences important for different interaction geometries within IFs and sequences associatcd with interfilament interactions. Finally, the extent and function of the soluble subunit pool, which is tacitly assumed in many experiments, must become a focus for more attention to extend initial s t ~ d i c s ( ' ~ , ' ~in) this pivotal area of cytoskeletal structure. References 1 CiiiN, S. S. M., MACIOCE, P.AND LEIM,R. K.H. (1991). Effects of truncatcd
neurotilament proteins on thc cndogcnous intermediate filamcnts in tran5fected fibroblasts. I. Cell Sci. 99. 335-350. 2 HATZFELD. M. AND WEBER,K. (1991). Modulation of keratin intermediate filament assembly by single aminn acid euclianges in the curisensub sequence at the C-tcrniinal end of the rod domain. J. C-dl Sci. 99, 351-362. 3 Mom, R. D., DOKALDSON. A. D. AND STEWAKT. M. (1991). Expression in EschPrichiu coli of human lamins A and C: influence of head and tail domains on assembly properties and paracrystal formation. J. Cell Sci. 99. 363-372. 4 PETER, M., NAKAG4WA. J.. DOREE, M., LAERF., J. c. AND NIGG,E. (1991). In vitm disasscmbly of thc nuclear lamina arid M-phase specific phosphorylation of lamins by cdc2 kinase. Cell 61, 591-602. 5 STFWART,M . (1990). Intermediate filaments: structure. assembly and molecular interactions. C'urr. Opinion Cell Bid. 2, 91-100. 6 STEWAKI..M.. QLIINLAN,R. A . AND MOIR. R. D. (1989). Molecular interactions in paracrystals in a fragniciit corre%ponding to the helical rod domain of glial fibiillary acidic protein: evidence for an antiparallel packing of molecules and polymorphism rclated to intermediate filament structure. J. Cell B i d . 109, 225-234. 7 K41JFM4NX, E., WERER,K. m u GLISL~R, N . (1985). Intermediate filament forming ability of desniin derivatives lacking either the amino-terminal 67 or the carboxyl-terminal 27 residues. J. Molec. Biol. 189, 733-742. 8 KAATS, J . M. H., PIEPEK, F. R., VKEEEGBERTS, W. 7. M.. VFRRUP, K. N., RAMAEKLRS, F. C . S . A N D BLoeMEhm41, H. (1990). Assembly of aminoterminally deleted desmiu in vimentin-free cclls. J. Cell Brol. 111, 1971-1985. 9 C H O U , Y. H.. BISCOFF,J . R., REAC'H, D. A N D GOLDUN,K. D. (1990).
Intermediate filament reorganisation during mitosis is mediated by p34cdc2 phmphorylation of vimentin. Cell 62, 1063-1071. 10 MOIR,R. D., QUINLAN, R. A. AND STEWART, M. (1990). Exprewion arid characterisation of human lamin C. FEBS Lett. 268, 301-305. 11 COULOMB, P. A., CHAN.Y-M.,ALBERS. K. AND FUCHS, E. (190). Deletions in epidermal keratins leading to alterations in filament organisation in vivo and in iiitcrmediate filament assembly in virro. J. Cell Bid. 111. 3039-3064. 12 FRANKE. W. W., SCHILLER. D . L.. HATZFELD, M. AND WIXTER.S. (1983). Protein complexes of intermediate dzcd filaments: mclting of cytokcratin coniplexes in urea reveals different polypeptide scpamtion eharactcristics. Proc. Nnrl Acad. Sci. USA. 80. 7113-7117. 13 Lu, X. AND LANE.E. B. (1990). Retrovirus-mcdiated transgenic keratin cxprcssion in cultured fibroblasts: specific domain functions in kcratin stahilisation and filamcnt formation. Cell 62, 681-696. 14 WONG,P. C. AND CLW~LAND, D. W. (1990). Charecterization of dominant and recessive assembly-defective mutations in mouse neurofilament NF-M. J. Cell Biol. 111, 1987-2003.
15 MCCORMICK, M. B.. CouLom, P. A. AND PUCHS, E. (1991). Sorting out IF networks: consequences of domain swapping 011 IF recognition and assembly. J . Cell Biol. 113, 1111-1124. 16 SOELLNER, P., QDINLAN, R. A. .4ND FRAKKE, w. w. (1985). Identification of a distinct soluhle pool of an intermediate filament protein: tetranicric vimentin from living cells. Pmc. Not1 Acad. Sci. USA 82. 7929-7933. 17 MILLER,R. K., VIKSTROM, K . AND GOLOMAN, R. D. (1991). Filament incorporation into intermediate filament networks is a tapid process. J. Cdl Biol. 113, 843-855.
Biochemistry, University of Dundee, Dundee DD1 4HN, Scotland, UK and Murray Stewart is at the MRC Laboratory of Molecular Biology, Cambridge
not been identified. Indeed, there is a paucity of structural data to direct the use of recombinant methods to dissect the pathway of filament assembly and the higher order molecular interaction geometries in filaments. Expression and mutagenesis of IF cDNAs in both eukaryotic and prokaryotic systems have allowed a more precise investigation of IF molecular interactions to be initiated. In particular, these experiments have drawn attention to small regions of the sequence near the ends of the rod domain that appear to exert a major influence on IF assembly (Fig. 1). One approach has been to introduce specific point mutations at sequence positions that are absolutely conscrved betwecn different IF classes. Hatzfeld and Webed’) monitored the effect on in v i m assembly of point mutations at the C-terminus of helix IIB in keratins 8 and 18. Their data showed that changing such residues had a profound effect upon filament assembly, altering those molecular interactions important for the control of filament length as well as inter-filament interactions. Some effects could be partially reversed by varying the conditions of in vitro assembly and this raises the possibility that several mechanisms exist for filament assembly. Filament assembly in v i m does not require all of the C-terminal globular domain, as this can be reduced to those 9 residues immediately adjacent to the rod(2). These data complement results obtained with pr~teolytic(~) and recombinant desmin fragmend8).A further illustration of the importance of regions near the ends of the rod to filament assembly is the observation that phosphorylation by the cdc2 kinase of serine 16 in chicken lamin B2 appears to depolymerise the nuclear lamina(4). Phosphorylation is also thought to be associated with the assembly and disassembly of other 1Fs and it has been proposed that the cdc2 kinase is also responsible for the cell cycle chan es in cytoplasmic intermediate filament architectures?91 . Other approaches have deleted whole domains in IF proteins or produced chimeric proteins from different IF classes and then monitored the effect on native filament networks by cell transfection. However. interpreting the results of these studies is not always straight-forward because the level of expressed protein
Table 1. Clusses of intermediate filament proteins
Sequence class
Proteins
Tissue specificity
I
Acidic keratins Basic,/neutral keratins Desrnin Vimentin GFAP Peripherin Neurofilaments (XF) (NF-H. NF-M, NF-L in vcitebrates) Nuclear lamins (often A-, B- and C-) Nestin
Epi tlielia Epithelia Muscle Mesenchyme Astrocytes, some glia Peripheral neurones Neurones
I1 111
IV V VI
Nuclear envelopes Neuroepithclial stem cells
-
GLOBULAR DOMAIN
4
I
Assembly critical domain N-Terminal domain
HELIX I
HELIX II
v
I /
a
Coiled-coilforming domains
”
IA
’/
/
’
7
LI
IB
’
L
-
GLOBULAR DOMAIN
1
--b
Assembly influencing domain
I
C-Terminal domain
-
Fig. 1. Schematic illustration of the functional domains of intermediate filament protein sequences. The sequences of IF proteins divide into three distinct domains. The central rod domain is thought to have a mainly cu-hclical conformation, punctuated by Thort stretches (or ‘linker regions’ - LI, LII and LI-11) in which the &-helicalconformation may be lost. Therefore, the rod can be divided into two major helices (IsiTT), and subdivided into helices IA, IB, IIA and IIB. The central rod domain dimerizes to produce a coiled-coil molecule (often refcrrcd to as a ‘dime?). The rod domain is flanked by a Y- and C-terminal globular domains, which are important for corrcct filament assembly, albeit to varying degrees. Two particularly important sites in the protein sequence have been indicated, namcly the cdc2 kinase pliosphorylation site in the N-terminal globular domain lamins, and the highly conserved C-teminal sequence at the end of helix IIB. As illustrated by a range of expcriments using engineered mutant IF proteins (see text), both sites are critically involved in TF assemhly.
and the phenotypic background of the host cells can influence the outcome. Although such variables limit this approach to asking very broadly defined questions, these studies have been particularly successful in helping to delineate the minimum length of protein sequence that permits or disrupts filament assemblies. Chin et al.(’) investigated the influence of N- and Cterminal deletions in neurofilament proteins on the endogenous vimentin filament network of cultured fibroblasts by transient transfection. The two most extensive deletions, which removed all but the first 127 and 139 residues, still caused the collapse of the endogenous vimentin IF network. In both cases, only the N-terminal globular domain and rod helix 1A (the smallest a-helical segment) remained. Conversely, expression of an N-terminal deletion of NF-L, which only left 70 residue? of helix 11B and the globular Cterminal domain, had no deleterious effect on the endogenous filament network. Chin et al. examined a range of deletion mutations and concluded that deletions involving the C-terminus of the rod domain were far more detrimental to filament structure than Nterminal deletions. This is consistent with a range of proteolysis and other experiments (reviewed in ref. 5 ) , although the C-terminal domain may be more important in the nuclear l a m i n ~ ( ~ , ’Small ~ ) . deletions at the rod C-terminus produced dominant negative mutations in keratins(”) and removal of the conserved RLLEGE sequence abolished filament assembly and resulted in the complete disassembly of the endogenous filament network in transfected cells. Similar conclusions were reached using chimeric keratin-vimentin constructs,
which underline the importance of helices IB and IIB in the correct sorting of diffcrcnt 1F p r ~ t e i n s ( l ~ This . ~ ~ )is. consistent with the molecular interaction geometry proposed for tetramerd6) and confirms the importance of helices IA,B and the sequence at the rod C-terminus. Assessing the contribution of different domains to filament integrity in keratin IF assembly is more complicated because it requires an equimolar ratio of type I and type I1 IF proteins. Moreover, the interaction of type 1 and I1 keratins is hierarchical and reflects their preferential expression in different cell types(12). These additional levels of complexity permit several special questions to be asked of this IF assembly system. Lu and Lane(’3) expressed mutated keratin cDNAs in cultured fibroblasts to study keratin pairing and filament formation and showed that, in any pair of interacting keratins, there must be one complete Nterminal and C-terminal globular domain. This requirement was found even when keratin 19, a naturally occurring keratin which lacks a globular C-terminal domain, was paired with a mutated type I1 keratin lacking the same domain. However, this latter result is not entirely consistent with in vifro experiments where filaments could be formed from such a combination under certain conditions(2); this discrepancy indicates the value of concomitant in vitro and in vivo studies. Although these results indicate that an interaction between the N- arid C-tcrminal globular domains of the same protein type in different molecules is necessary for filament formation, it is clear that these domains contribute in different ways to filament structure. Coulomb et aZ.(”) showed that keratin 14, lacking both
these domains, was able to form filaments in vivo and also in vitro with appropiate type I1 partners but the morphology of these filaments was altered, being less regular in diameter. This suggests that there are complementary interactions involving both apposed globular domains in keratin molecules. It is possible that part of this difference could be attributed to the different keratin pairs being studied, given that the keratin 5/14 complex is one of the most stable pairs known, whereas keratins 5/18 are one of the least stable pairs with respect to urea-mediated dissociation. I n another approach, Lu and Lane(13) removed some of the rod sub-domains to show that helix IIA and the linker sequences flanking it were not necessary for complex formation, but were important for the correct cellular distribution of the filamentous aggregates. Another construct in which part of the rod domain spanning residues 139-340 of keratin 18, which includes most of helix IB and IIB, was deleted apparently gave no heterocomplex or filament formation. Heterocomplex formation is requisite for keratin filament assembly, but so far the operative mechanisms that enable mutated type I or 11 keratin to permeate the filament network are unclear. The rules governing the exchange of the various individual protein chains within the protein molecule of two chains as well as the assembly competent form (the molecular dimer) must be delineated. Once these inter-relationships are better understood, we should be nearer to explaining the apparent discrepancies between some of the in vitro and in vivo experiments. Because of the absence of detailed structural data regarding molecular arrangements in lFs, the design and the interpretation of in vivo transfection experiments is sometimes not completely straightforward. It is, for example, difficult to be certain at which lcvel of organisation a mutant IF protein manifests itself. Does the mutation influence the formation of two-chain molecules, four-chain tetramers or any of range of higher assembly aggregates? The generation of dominant negative mutations in many of the transfection studies indicated that at least the phenotypes observed were unlikely to have completely trivial explanations. The potency of these mutations was often very high and some were able to influence filament assembly when present at less than 1 % of the total IF protein concentration("). This indicates that these mutations might affect important and sub-stoichiometric, higherorder interaction geometries needed for filament integrity. Similar observations have been made with the neurofilamen t system(14). Such dominant negative mutations should prove invaluable for studying the function of IFs in vivo. Another noteworthy feature of the transfection studies is that they lend support to the idea that intermediate filament assembly may be dynamic. Although it is possible to explain the reorganisation of the endogenous IF network following tranfection through fragment association with the filament walls and a subsequent destabilisation of the
polymer, the presence of a soluble pool in equilibrium with the filaments appears to be an inescapable conclusion. Another problem is the thorny issue of subunit association versus integration when interpreting the cell transfection data: it is quite possible that the apparently smooth surface of intermediate filaments belies a substructure that is composed of a fascia of ropes that perhaps not only stop and start along the length of the filament but in which the individual subunits are also freely accessible to a soluble pool. Despite the limitations of many of the experiments using IF mutants. a number of firm conclusions are emerging. It is clear that short regions near the ends of the m-helical rod domain are vital for filament assembly and changing these sequences can produce dominant negative mutants; the latter are certain to prove effective tools for investigating IF function and assembly in vivo. These sequences. as well as portions of the globular domains that flank them. not only influence the assembly but also the extent of filament aggregation and filament girth. It is also clear that the N-terminal globular domain is more important than the corresponding C-terminal domain to filament integrity. So far, most experiments have addressed filament assembly as an all-or-nothing process, while more subtle changes have not been investigated in great detail. One way to extend the work along these lines would be to use antibody probes to follow specific epitopes. Mutations that alter filament morphology and architectures should identify sequences important for different interaction geometries within IFs and sequences associatcd with interfilament interactions. Finally, the extent and function of the soluble subunit pool, which is tacitly assumed in many experiments, must become a focus for more attention to extend initial s t ~ d i c s ( ' ~ , ' ~in) this pivotal area of cytoskeletal structure. References 1 CiiiN, S. S. M., MACIOCE, P.AND LEIM,R. K.H. (1991). Effects of truncatcd
neurotilament proteins on thc cndogcnous intermediate filamcnts in tran5fected fibroblasts. I. Cell Sci. 99. 335-350. 2 HATZFELD. M. AND WEBER,K. (1991). Modulation of keratin intermediate filament assembly by single aminn acid euclianges in the curisensub sequence at the C-tcrniinal end of the rod domain. J. C-dl Sci. 99, 351-362. 3 Mom, R. D., DOKALDSON. A. D. AND STEWAKT. M. (1991). Expression in EschPrichiu coli of human lamins A and C: influence of head and tail domains on assembly properties and paracrystal formation. J. Cell Sci. 99. 363-372. 4 PETER, M., NAKAG4WA. J.. DOREE, M., LAERF., J. c. AND NIGG,E. (1991). In vitm disasscmbly of thc nuclear lamina arid M-phase specific phosphorylation of lamins by cdc2 kinase. Cell 61, 591-602. 5 STFWART,M . (1990). Intermediate filaments: structure. assembly and molecular interactions. C'urr. Opinion Cell Bid. 2, 91-100. 6 STEWAKI..M.. QLIINLAN,R. A . AND MOIR. R. D. (1989). Molecular interactions in paracrystals in a fragniciit corre%ponding to the helical rod domain of glial fibiillary acidic protein: evidence for an antiparallel packing of molecules and polymorphism rclated to intermediate filament structure. J. Cell B i d . 109, 225-234. 7 K41JFM4NX, E., WERER,K. m u GLISL~R, N . (1985). Intermediate filament forming ability of desniin derivatives lacking either the amino-terminal 67 or the carboxyl-terminal 27 residues. J. Molec. Biol. 189, 733-742. 8 KAATS, J . M. H., PIEPEK, F. R., VKEEEGBERTS, W. 7. M.. VFRRUP, K. N., RAMAEKLRS, F. C . S . A N D BLoeMEhm41, H. (1990). Assembly of aminoterminally deleted desmiu in vimentin-free cclls. J. Cell Brol. 111, 1971-1985. 9 C H O U , Y. H.. BISCOFF,J . R., REAC'H, D. A N D GOLDUN,K. D. (1990).
Intermediate filament reorganisation during mitosis is mediated by p34cdc2 phmphorylation of vimentin. Cell 62, 1063-1071. 10 MOIR,R. D., QUINLAN, R. A. AND STEWART, M. (1990). Exprewion arid characterisation of human lamin C. FEBS Lett. 268, 301-305. 11 COULOMB, P. A., CHAN.Y-M.,ALBERS. K. AND FUCHS, E. (190). Deletions in epidermal keratins leading to alterations in filament organisation in vivo and in iiitcrmediate filament assembly in virro. J. Cell Bid. 111. 3039-3064. 12 FRANKE. W. W., SCHILLER. D . L.. HATZFELD, M. AND WIXTER.S. (1983). Protein complexes of intermediate dzcd filaments: mclting of cytokcratin coniplexes in urea reveals different polypeptide scpamtion eharactcristics. Proc. Nnrl Acad. Sci. USA. 80. 7113-7117. 13 Lu, X. AND LANE.E. B. (1990). Retrovirus-mcdiated transgenic keratin cxprcssion in cultured fibroblasts: specific domain functions in kcratin stahilisation and filamcnt formation. Cell 62, 681-696. 14 WONG,P. C. AND CLW~LAND, D. W. (1990). Charecterization of dominant and recessive assembly-defective mutations in mouse neurofilament NF-M. J. Cell Biol. 111, 1987-2003.
15 MCCORMICK, M. B.. CouLom, P. A. AND PUCHS, E. (1991). Sorting out IF networks: consequences of domain swapping 011 IF recognition and assembly. J . Cell Biol. 113, 1111-1124. 16 SOELLNER, P., QDINLAN, R. A. .4ND FRAKKE, w. w. (1985). Identification of a distinct soluhle pool of an intermediate filament protein: tetranicric vimentin from living cells. Pmc. Not1 Acad. Sci. USA 82. 7929-7933. 17 MILLER,R. K., VIKSTROM, K . AND GOLOMAN, R. D. (1991). Filament incorporation into intermediate filament networks is a tapid process. J. Cdl Biol. 113, 843-855.
Biochemistry, University of Dundee, Dundee DD1 4HN, Scotland, UK and Murray Stewart is at the MRC Laboratory of Molecular Biology, Cambridge
