Intermediate filaments: new proteins,
some answers, more questions
Michael W Klymkowsky
U n i v e r s i t y of C o l o r a d o , Boulder, USA
The past year has seen significant progress in the characterization of
intermediate filament proteins. New proteins have been identified and
physiologically significant differences between known proteins have been
revealed. Changes in intermediate filament organization have been linked
to changes in cell behavior, and mutational analyses are beginning to reveal
the connection between intermediate filament expression, network formation,
cellular behavior and disease.
Current Opinion in Cell Biology 1995, 7:46-54
Introduction of characteristic and conserved motifs is a reliable guide
to identifying IF proteins. The presence of these motifs
A startling thing about intermediate filaments (IFs) is that is necessary, and sufficient, to classify a protein as an IF
they appear to be specific to multicellular animals. De- protein.
spite extensive searches, neither nuclear IFs 0amins) nor
cytoplasmic IFs have yet been identified in unicellular In writing this review I have attempted to avoid duplica-
eukaryotes. A recent report oflamins in plants, identifed tion o f previous reviews in Current Opinion in Cell Biol-
using immunocytochemical methods [1°], is exciting but ogy; interested readers should consult these [8,9], as well
remains to be confirmed by sequence data (see below). as the review by Lane in this issue (10p 118-125) and
The importance o f sequence data in the characterization a recent review elsewhere [10]. Furthermore, given the
o f IF proteins is illustrated by a number o f observations. substantial number of papers that were published dealing
The tektins, proteins associated with the shared micro- with IFs last year, I have had to omit a number of inter-
tubules of the flagellar axoneme wall, were once thought esting pieces of work and have discussed in detail only
to be IF-like, on the basis o f their immunological cross- relatively few papers. In particular, I have essentially ig-
reactivity; sequence analysis, however, indicates only that nored papers on the control o f cytoplasmic IF protein
they are composed of 0t-helical and non-helical domains, (clFp) expression and post-translational modification. I
with little else in the way o f similarity to true IF proteins beg forgiveness for any omissions, whether purposeful
[2]. Similarly, McConnell and Yaffe [3] reported that or inadvertent.
the yeast Saccharomyces cerevisiae protein Mdm-1 can be
assembled into IF-like structures in vitro; but sequence
analysis indicates that the protein bears little similar-
ity to known IF proteins, and does not contain the N e w IF proteins
highly conserved 'helix initiation' and 'helix termina-
tion' motifs characteristic o f IF proteins (see [4,5]). In It is generally assumed that the progenitor of the IF pro-
this light the observations of Geisler et al. ([6°], see also teins was a nuclear lamin (Fig. 1), although, until lamins
[7]) are informative: they found that short polypep- are identified in single-celled eukaryotes, this assump-
tides corresponding to the helix initiation and termi- tion remains unsubstantiated. That a lamin-like protein
nation regions of desmin can assemble into structures was the progenitor of the clFps is suggested by a body
that look very nmch like IFs under the electron mi- of work from Weber and co-workers, the latest install-
croscope; however, in these filaments the polypeptides ment of which [11 "°] is the analysis of the clFp-encoding
assume what appears to be an aberrant ~-sheet config- genes of the nematode Caenorhabditis elegans. Lamins are
uration. That such small polypeptides, - 2 0 residues in characterized by a -352 amino acid long central 'helical'
length, have the ability to form an 'IF' in vitro suggests domain composed largely ofheptad repeats. The clFps of
that filament formation is not a completely trustworthy C. elegans and other invertebrates are similar to the lamins
measure of a protein's relationship to true IF proteins. At in that their central domain is also -352 amino acids in
this point, only sequence data confirnfing the presence length; they differ from the lamins, however, in that the
Abbreviations
clF--cytoplasmic IF; clFp~cytoplasmic IF protein; GFAP~glial fibrillary acidic protein; IF--intermediate filament.
46 © Current Biology Ltd ISSN 0955-0674
, Intermediate filaments: new proteins, some answers, more questions Klymkowsky 47
characteristic nuclear localization sequence and CAAX to co-polymerize (Fig. 1). Vimentin, desmin, peripherin
box characteristic o f the lanfins are missing. C. elegans and glial fibrillary acidic protein (GFAP) are the most
has at least eight clFp genes [11"'], which is more than similar in terms of primary structure, and the genes
in any other invertebrate yet investigated. It remains to encoding them have the same basic intron/exon struc-
be seen how these proteins are expressed, but the genetic ture. In vitro, at least, all can form homopolymeric IF-like
accessibihty of C. elegans should be extremely useful not structures and all can co-polymerize with one another.
only in further defining the functions of cytoplasmic IFs, clFps assemble in a hierarchical manner, first as dimers,
but also in identifying proteins that interact, directly or then as dimer-dimers (tetramers), and then into higher
indirectly, with clFs. order structures. An analysis of vimentin-desmin assem-
bly suggests that these proteins form a tetramer com-
In the vertebrates, there appears to have been a relative
posed of a vimentin and desmin homodimers [12]. The
frenzy o f c l F p gene duplication (Fig. 1). On the basis of
formation of the tetramer appears to be driven by ionic
the intron/exon structure of clFp genes, it appears that
interactions (see [13] and references therein).
the vertebrate clFps are direct descendants of the inverte-
brate ones. Before the diversification of vertebrate clFps, A second group, the neuronal clFps, are related more
however, the progenitor gene appears to have lost a six in terms of intron/exon structure than primary se-
heptad repeat region of the helical region, as in the ver- quence. This group contains the rather divergent clF
tebrate clFps, this central helical region is - 3 1 0 amino proteins NF1, NFm, NFh, 0t-internexin and nestin. It
acids in length. Presumably the vertebrate clFp family appears that these proteins can all co-polymerize with
began to diversify only after this deletion event. Verte- vimentin and, from the point o f view of IF network
brate clFps can be grouped on the basis of sequence sim- formation, may be viewed as partners of vimentin.
ilarity, genomic intron/exon structure, and their abihty There is, however, a clear-cut distinction between the
Fig. 1. The IF protein family (based on
the work of Weber and co-workers [11 "'])
Progenitor(s) Assuming that the progenitor of the mod-
ern IF proteins was a nuclear lamin, it ap-
pears clear that the clFps of invertebrates
[ - 3 5 2 amino acid central s-helical domain arose through a gene duplication event(s)
Nuclear lamins -I Nuclear localization signal (NLS) followed by the loss of the nuclear lo-
j~ LCAAX box calization signal and the CAAX box in-
-NLS <2 volved in the isoprenylation of lamins. In
-CAAX the transition from invertebrate to verte-
brate, there was the loss of 6 heptad re-
Invertebratec lFps
peats from the central rod domain. The
gene encoding this shortened clFp ap-
~>-Six heptad repeats pears to have been duplicated to form the
members of the vimentin-like group. The
neurofilament proteins appear to have
arisen from a common progenitor, and
, VertebrateclFp progenitor ~ - ~ / ~ display a distinctive intron/exon pattern.
The founder neurofilament gene presum-
ably also underwent a series of gene du-
plications to produce the neuronal clFps.
Vimentin-like % Lossof intron/exon
J
Type I Type II Lens dFps
Throughout this process, the neurofila-
ments retained their ability to co-poty-
merize with the vimentin-like proteins.
clFps %% structure Keratins Keratins Filensin Somewhere during this period, however,
the keratin gene family diverged, and,
Vimentin ~2~ Neuronal clFps K9-K20 K1-K8 Phakinin
presumably early on became obligate
Desmin
GFAP NFI heteropolymers. The type l and type II
Peripherin NFm keratin progenitor genes then proliferated
NFh to form Ihe families we see today. Fi-
Plasticin? c~-internexin nally, a new family, the lens clFps, ap-
Nestin peared. The similarity of phakinin (CP47)
to type I keratin and of filensin to non-
XNIF?
keratin clFps suggest that these proteins
Gelfiltin?
Barrier to polymerization
t may have been 'hijacked' and special-
Tannabin?
ized to form the functionally unique clF
system of the lens. As new clFps are iden-
tified it should be possible to place them
© 1995 Current Opinion in Cell Biology in this scheme on the basis of their gene
structure and polymerization properties.
some answers, more questions
Michael W Klymkowsky
U n i v e r s i t y of C o l o r a d o , Boulder, USA
The past year has seen significant progress in the characterization of
intermediate filament proteins. New proteins have been identified and
physiologically significant differences between known proteins have been
revealed. Changes in intermediate filament organization have been linked
to changes in cell behavior, and mutational analyses are beginning to reveal
the connection between intermediate filament expression, network formation,
cellular behavior and disease.
Current Opinion in Cell Biology 1995, 7:46-54
Introduction of characteristic and conserved motifs is a reliable guide
to identifying IF proteins. The presence of these motifs
A startling thing about intermediate filaments (IFs) is that is necessary, and sufficient, to classify a protein as an IF
they appear to be specific to multicellular animals. De- protein.
spite extensive searches, neither nuclear IFs 0amins) nor
cytoplasmic IFs have yet been identified in unicellular In writing this review I have attempted to avoid duplica-
eukaryotes. A recent report oflamins in plants, identifed tion o f previous reviews in Current Opinion in Cell Biol-
using immunocytochemical methods [1°], is exciting but ogy; interested readers should consult these [8,9], as well
remains to be confirmed by sequence data (see below). as the review by Lane in this issue (10p 118-125) and
The importance o f sequence data in the characterization a recent review elsewhere [10]. Furthermore, given the
o f IF proteins is illustrated by a number o f observations. substantial number of papers that were published dealing
The tektins, proteins associated with the shared micro- with IFs last year, I have had to omit a number of inter-
tubules of the flagellar axoneme wall, were once thought esting pieces of work and have discussed in detail only
to be IF-like, on the basis o f their immunological cross- relatively few papers. In particular, I have essentially ig-
reactivity; sequence analysis, however, indicates only that nored papers on the control o f cytoplasmic IF protein
they are composed of 0t-helical and non-helical domains, (clFp) expression and post-translational modification. I
with little else in the way o f similarity to true IF proteins beg forgiveness for any omissions, whether purposeful
[2]. Similarly, McConnell and Yaffe [3] reported that or inadvertent.
the yeast Saccharomyces cerevisiae protein Mdm-1 can be
assembled into IF-like structures in vitro; but sequence
analysis indicates that the protein bears little similar-
ity to known IF proteins, and does not contain the N e w IF proteins
highly conserved 'helix initiation' and 'helix termina-
tion' motifs characteristic o f IF proteins (see [4,5]). In It is generally assumed that the progenitor of the IF pro-
this light the observations of Geisler et al. ([6°], see also teins was a nuclear lamin (Fig. 1), although, until lamins
[7]) are informative: they found that short polypep- are identified in single-celled eukaryotes, this assump-
tides corresponding to the helix initiation and termi- tion remains unsubstantiated. That a lamin-like protein
nation regions of desmin can assemble into structures was the progenitor of the clFps is suggested by a body
that look very nmch like IFs under the electron mi- of work from Weber and co-workers, the latest install-
croscope; however, in these filaments the polypeptides ment of which [11 "°] is the analysis of the clFp-encoding
assume what appears to be an aberrant ~-sheet config- genes of the nematode Caenorhabditis elegans. Lamins are
uration. That such small polypeptides, - 2 0 residues in characterized by a -352 amino acid long central 'helical'
length, have the ability to form an 'IF' in vitro suggests domain composed largely ofheptad repeats. The clFps of
that filament formation is not a completely trustworthy C. elegans and other invertebrates are similar to the lamins
measure of a protein's relationship to true IF proteins. At in that their central domain is also -352 amino acids in
this point, only sequence data confirnfing the presence length; they differ from the lamins, however, in that the
Abbreviations
clF--cytoplasmic IF; clFp~cytoplasmic IF protein; GFAP~glial fibrillary acidic protein; IF--intermediate filament.
46 © Current Biology Ltd ISSN 0955-0674
, Intermediate filaments: new proteins, some answers, more questions Klymkowsky 47
characteristic nuclear localization sequence and CAAX to co-polymerize (Fig. 1). Vimentin, desmin, peripherin
box characteristic o f the lanfins are missing. C. elegans and glial fibrillary acidic protein (GFAP) are the most
has at least eight clFp genes [11"'], which is more than similar in terms of primary structure, and the genes
in any other invertebrate yet investigated. It remains to encoding them have the same basic intron/exon struc-
be seen how these proteins are expressed, but the genetic ture. In vitro, at least, all can form homopolymeric IF-like
accessibihty of C. elegans should be extremely useful not structures and all can co-polymerize with one another.
only in further defining the functions of cytoplasmic IFs, clFps assemble in a hierarchical manner, first as dimers,
but also in identifying proteins that interact, directly or then as dimer-dimers (tetramers), and then into higher
indirectly, with clFs. order structures. An analysis of vimentin-desmin assem-
bly suggests that these proteins form a tetramer com-
In the vertebrates, there appears to have been a relative
posed of a vimentin and desmin homodimers [12]. The
frenzy o f c l F p gene duplication (Fig. 1). On the basis of
formation of the tetramer appears to be driven by ionic
the intron/exon structure of clFp genes, it appears that
interactions (see [13] and references therein).
the vertebrate clFps are direct descendants of the inverte-
brate ones. Before the diversification of vertebrate clFps, A second group, the neuronal clFps, are related more
however, the progenitor gene appears to have lost a six in terms of intron/exon structure than primary se-
heptad repeat region of the helical region, as in the ver- quence. This group contains the rather divergent clF
tebrate clFps, this central helical region is - 3 1 0 amino proteins NF1, NFm, NFh, 0t-internexin and nestin. It
acids in length. Presumably the vertebrate clFp family appears that these proteins can all co-polymerize with
began to diversify only after this deletion event. Verte- vimentin and, from the point o f view of IF network
brate clFps can be grouped on the basis of sequence sim- formation, may be viewed as partners of vimentin.
ilarity, genomic intron/exon structure, and their abihty There is, however, a clear-cut distinction between the
Fig. 1. The IF protein family (based on
the work of Weber and co-workers [11 "'])
Progenitor(s) Assuming that the progenitor of the mod-
ern IF proteins was a nuclear lamin, it ap-
pears clear that the clFps of invertebrates
[ - 3 5 2 amino acid central s-helical domain arose through a gene duplication event(s)
Nuclear lamins -I Nuclear localization signal (NLS) followed by the loss of the nuclear lo-
j~ LCAAX box calization signal and the CAAX box in-
-NLS <2 volved in the isoprenylation of lamins. In
-CAAX the transition from invertebrate to verte-
brate, there was the loss of 6 heptad re-
Invertebratec lFps
peats from the central rod domain. The
gene encoding this shortened clFp ap-
~>-Six heptad repeats pears to have been duplicated to form the
members of the vimentin-like group. The
neurofilament proteins appear to have
arisen from a common progenitor, and
, VertebrateclFp progenitor ~ - ~ / ~ display a distinctive intron/exon pattern.
The founder neurofilament gene presum-
ably also underwent a series of gene du-
plications to produce the neuronal clFps.
Vimentin-like % Lossof intron/exon
J
Type I Type II Lens dFps
Throughout this process, the neurofila-
ments retained their ability to co-poty-
merize with the vimentin-like proteins.
clFps %% structure Keratins Keratins Filensin Somewhere during this period, however,
the keratin gene family diverged, and,
Vimentin ~2~ Neuronal clFps K9-K20 K1-K8 Phakinin
presumably early on became obligate
Desmin
GFAP NFI heteropolymers. The type l and type II
Peripherin NFm keratin progenitor genes then proliferated
NFh to form Ihe families we see today. Fi-
Plasticin? c~-internexin nally, a new family, the lens clFps, ap-
Nestin peared. The similarity of phakinin (CP47)
to type I keratin and of filensin to non-
XNIF?
keratin clFps suggest that these proteins
Gelfiltin?
Barrier to polymerization
t may have been 'hijacked' and special-
Tannabin?
ized to form the functionally unique clF
system of the lens. As new clFps are iden-
tified it should be possible to place them
© 1995 Current Opinion in Cell Biology in this scheme on the basis of their gene
structure and polymerization properties.
