The canonical UCS (UNC-45/Cro1/She4p) protein, Caenorhabditis elegans UNC-45, was one of the earliest molecules to be shown genetically to be necessary for sarcomere assembly. Genetic analyses of homologues in several fungal species indicate that the conserved UCS domain functionally interacts with conventional type II and unconventional type V myosins. In C. elegans and other invertebrate species, UNC-45 and its orthologues interact with both sarcomeric and non-sarcomeric myosins whereas, in vertebrates, there are two UNC-45 isoforms: a general cell (GC) and a striated muscle (SM) isoform. Although the mechanism of action of UCS proteins is unknown, recent biochemical studies suggest that they may act as molecular chaperones that facilitate the folding and/or maturation of myosin.
Myosins are a large family of protein motors that contain at least 18 different classes. They interact with actin filaments to generate a broad spectrum of eukaryotic cell movements that includes phagocytosis, vesicle transport, cytokinesis and maintenance of cell shape in addition to their well-known role in muscle contraction (Berg et al., 2001; Reck-Peterson et al., 2000; Sellers,2000). All of the myosins share a common motor domain necessary for interaction with actin and the generation of movement but possess additional domains related to their specialized functions(Fig. 1).
The UCS domain family is a group of essential proteins necessary for a variety of myosin- and actin-dependent functions in eukaryotic organisms from fungi to nematodes (UNC-45 in Caenorhabditis elegans, CRO1 in Podospora anserina and She4p in Saccharomyces cerevisiae)and is also present in additional animal species from Drosophila to humans. They were originally identified through mutations that disrupt myosin-dependent processes. Recently, however, C. elegans UNC-45 has been shown to bind the well-known molecular chaperone Hsp90 and muscle myosin subfragment-1 (S1) and act as a molecular chaperone for myosin(Barral et al., 2002).
Here we discuss genetic, biochemical and developmental analyses of UCS proteins in fungi and C. elegans and the exciting new findings that specific isoforms of UNC-45 are present in humans and other vertebrates.
C. elegans UNC-45
The unc-45 gene was originally identified in C. elegansas a single recessive, temperature-sensitive mutant allele, e286(Epstein and Thomson, 1974). The mutant showed decreased body movement and disorganized myofilament arrays when at 25°C but an essentially wild-type phenotype when grown at 15°C. Temperature shifts reverse the phenotype in developing embryos or larvae, when most myofilament assembly occurs, but not in adults. The original hypothesis was that UNC-45 has a catalytic function necessary for formation of proper myofilament arrays (Epstein and Thomson, 1974).
A more detailed genetic analysis of unc-45 revealed recessive lethal as well as additional temperature-sensitive alleles, demonstrating that UNC-45 function is essential to C. elegans development(Venolia and Waterston, 1990). The lethal unc-45 alleles cause arrest at the two-fold embryonic stage with a failure to produce functional body wall muscle. This phenotype is similar to that of mutants lacking the essential myo-3 encoded myosin heavy chain A, one of the two myosins (A and B) found in C. elegansbody wall muscle thick filaments. Importantly, the temperature-sensitive alleles directly affect the function of the unc-54-encoded myosin heavy chain B. Suppression of the phenotype of temperature-sensitive unc-45 mutants by overproduction of myosin A required a background null for unc-54 (Venolia and Waterston, 1990). This result indicates that myosin B inhibits assembly of myosin-A-containing filaments in the background of an unc-45 mutant even though myosin A can functionally substitute for myosin B (Epstein et al.,1986). However, studies of the lethal unc-45 alleles clearly indicated that UNC-45 also interacts with myosin A and probably the pharyngeal muscle myosins C and D. The pharynges showed decreased pumping in all lethal unc-45 alleles whereas the maternally rescuable lethal allele unc-45(st604) phenotype is enhanced by myosin A overproduction in contrast to the usual suppression. The st604 mutation has not been characterized but is probably not a null allele, although the mutation dramatically reduces its activity. Overproduction of myosin A only enhances that phenotype by overwhelming the mutant UNC-45, thereby preventing it from functioning in other myosin-dependent processes. UNC-45 was therefore proposed to interact directly with all muscle myosins in C. elegans either to control myosin assembly, organization of thick filaments into arrays or myosin contractile activity (Venolia and Waterston, 1990).
Characterization of the genomic and cDNA sequences encoding UNC-45 provided significant insights into its function and the unc-45 mutations(Barral et al., 1998). UNC-45 contains three distinct regions: an N-terminal domain containing three tetratricopeptide (TPR) repeats, a central region showing only homology to other animal UNC-45-like proteins and the C-terminal UCS domain, which shares blocks of sequence identity with fungal UCS proteins and more extensive homology with the C-terminal regions of other animal UNC-45-like proteins. Three of the four known temperature-sensitive alleles are associated with missense substitutions in the UCS domain whereas two lethal alleles contain stop codons upstream in the central region. The latter produce null mutants because they prevent translation of much of the central region and the entire UCS domain and produce a phenotype identical to that observed in unc-45 knockouts generated by RNA interference(Venolia et al., 1999). Since the temperature-sensitive mutations produce unstable thick filaments containing scrambled A and B myosins, which normally assemble into distinct zones within the filament, the UCS domain is directly implicated in thick filament assembly. It is likely that the scrambled filaments are due to improperly folded myosin, which is consistent with its lower accumulation(Barral et al., 1998). However,the unc-45 mutation may have also uncovered an assembly function for UNC-45, and both these possibilities need to be explored further.
Localization of UNC-45 by either antibodies or GFP labeling(Ao and Pilgrim, 2000;Venolia et al., 1999) shows that it is present in both the body wall and pharyngeal muscles and in the cleavage furrows of early embryos in C. elegans. In the developing body wall muscle of early larvae, UNC-45 appears to be cytosolic, whereas in mature, adult muscle, the protein is clearly localized to the sarcomeric A-bands that contain thick filaments (Ao and Pilgrim, 2000). These results are consistent with the genetic and molecular studies; moreover, recent two-hybrid analysis suggests that, in addition to the muscle myosins, cytoskeletal type II and unconventional type V myosins interact with UNC-45 protein (W. Ao and D. Pilgrim, personal communication). This interaction with type II myosin is consistent with its localization to the cleavage furrow since type II myosins are necessary for the assembly of the contractile ring and its function in cytokinesis(Balasubramanian et al., 1998;De Lozanne and Spudich, 1987;Knecht and Loomis, 1987).
The terminal phenotype of an unc-45 null mutant is at a stage consistent with a lack of body wall muscle. This stage of arrest is later than would be expected considering the possible interactions of UNC-45 with myosins other than those found in body wall muscle and a possible role in cytokinesis(see S. pombe Rng3p in the next section). However, the null phenotype can be explained by maternal rescue as has been discussed for its co-chaperone Hsp90 (Birnby et al., 2000). Rescue probably occurs at the level of UNC-45 protein because RNA interference, which generally blocks maternal rescue at the RNA level, when targeted to UNC-45 produces a muscle-specific phenotype. The maternally contributed UNC-45 would be enough to fulfill its general cellular functions but would become overwhelmed when the embryos begin to form muscle —thus the arrest at the twofold stage.
The fungal UCS proteins
The realization that a family of proteins containing the UCS domain exists over a wide phylogenetic spectrum arose from the observation that the sequence of C. elegans UNC-45 is very similar to two fungal proteins: She4p(S. cerevisiae) and CRO1 (P. anserina). The subsequent findings that RNG3p (Schizosaccharomyces pombe) also contains a UCS-like region and that mutations in it produce altered assembly and function of the cytokinetic ring provided an important confirmation of the significance of these domains and their interactions with myosins(Fig. 2)(Wong et al., 2000).
She4p of S. cerevisiae
The gene encoding the 789-residue budding yeast protein She4p was discovered in two independent screens, one for expression of the HO endonuclease exclusively in mother cells but not buds (the SHE screen for Swi5p-dependent HO expression) and the other for endocytosis defects (the dim screen for defective internalization of membrane)(Jansen et al., 1996;Wendland et al., 1996). In each case, one mutant allele was isolated. Most of the other mutants isolated were in the SHE1 gene encoding Myo4p, an unconventional type V myosin. A deletion of the UCS homolog SHE4 leads to decreased growth and endocytosis, altered cell morphology and loss of actin cytoskeleton polarity (Jansen et al., 1996;Wendland et al., 1996). Likewise, the dim1 mutant shows temperature-sensitive loss of polarity of actin localization, defective secretion and constitutive rounding of cells. This broad phenotype is consistent with the idea that the mutations in She4p that occur in DIM1 and SHE4 mutants produce an intrinsic defect in the actin cytoskeleton and possibly in myosin motor activity.
CRO1 protein of P. anserina
The CRO1 protein of P. anserina is a 702-residue protein that shares 21% identity and 40% similarity with She4p. The cro1 gene was identified through a screen for defects in sexual sporulation. The cro1-1 allele identified in this screen is a null mutant owing to a premature chain termination caused by a frame shift mutation. It shows pleiotropic alterations: abortive meioses leading to polyploid nuclei, an inability to form septa between the daughter nuclei following mitotic division, and decreased filamentous growth. In wild type fungal filaments, the actin assembly is coordinated with microtubule disassembly. However, in the cro1-1 mutant, the syncytial cytoplasm becomes filled with multiple nuclei and the actin cytoskeleton becomes disorganized, which permits abundant microtubules to remain (Berteaux-Lecellier et al., 1998). In the absence of CRO1 function, myosins that interact with and organize the actin cytoskeleton may not be functional and as a result, the signaling pathway that regulates actin assembly and microtubule disassembly is disrupted. This pathway may be similar to one in S. pombe that monitors the integrity of the actin cytoskeleton and delays sister chromatid separation until the mitotic spindle is properly oriented(Gachet et al., 2001).
Rng3p of S. pombe
The RNG3 gene was identified in a screen for defective actomyosin ring assembly and cytokinesis in S. pombe cell division(Balasubramanian et al., 1998). Missense mutations in the UCS domain of the predicted 746-residue protein generate several temperature-sensitive mutants(Wong et al., 2000). When these were crossed specifically with myo2 mutants (myo2encodes the essential myosin heavy chain of the actomyosin cytokinetic ring),synthetic lethals resulted, which suggests at least functional interaction between the two proteins. The rng3 null mutant fails to undergo cytokinesis, generating spores with multiple nuclei. In addition, it exhibits defective actin organization and actomyosin ring assembly. Perhaps most significantly, the specific myo2-E1 mutant, in which there is a G345R substitution in the Myo2p motor domain, causes sequestration of wild type Rng3 protein in the defective actomyosin ring. None of the other myosin mutants or other cytokinetic-defective mutants does this, which suggests that there is a specific interaction between the E1 mutant motor domain of Myo2p and Rng3p. In interphase, myosin forms a `spot' that is the putative progenitor to the cytokinetic actomyosin ring. The maintenance of this spot requires the function of Rng3p and this spot is proposed to be a template upon which the actomyosin ring forms during cytokinesis(Wong et al., 2002). Rng3p may be necessary for maintaining the myosin in an assembly-competent state and therefore act in a manner similar to UNC-45 during the assembly of body wall muscle thick filaments.
The fungal UCS proteins SHE4, CRO1 and RNG3 all show sequence similarity in their C-terminal UCS domains to UNC-45, but the phenotypes of their mutants show few significant similarities. However, all three fungal UCS genes and unc-45 are linked by their common association with processes related to or requiring myosins (Fig. 2). These results provide additional evidence for UCS proteins acting not strictly on conventional myosins but on unconventional myosins as well and suggest that they may have a more general function in the cell.
Molecular studies of UNC-45
Unlike its fungal relatives, UNC-45 contains three predicted domains including the conserved UCS domain that interacts functionally with various myosins (Fig. 2)(Barral et al., 1998). The N-terminal TPR domain, not present in the fungal proteins, resembles those that interact with the molecular chaperones Hsp70 and Hsp90 while the central domain shows sequence similarity only to other animal UCS proteins. Several recombinant proteins have been constructed to test for protein-protein interactions.
Recombinant full-length UNC-45 protein (FL) forms a complex with Hsp70,Hsp90 and body wall muscle myosin (Fig. 3). Hsp70 and Hsp90 in Sf9 insect cells and C. eleganslysates and myosin in high salt extracts of C. elegans are pulled down by UNC-45 [(Barral et al.,2002); A. H. Hutagalung, J. M. Barral and H. F. Epstein,unpublished].
An UNC-45 construct lacking the TPR domain [TPR(-)] binds to myosin and possess general chaperone activity. Purified C. elegans myosin binds to purified Hsp90, full-length UNC-45 and TPR(-) and this interaction occurred at 30°C but not at 4°C (Barral et al., 2002). This is similar to the interaction of Hsp90 and its associated co-chaperones with client proteins such as the steroid receptors(Dittmar and Pratt, 1997;Kosano et al., 1998). The elevated temperature promotes hydrophobic interactions consistent with UNC-45 and Hsp90 acting as molecular chaperones for myosin.
The TPR domain of UNC-45 binds the conserved C-terminal MEEVD sequence of Hsp90 (Fig. 3). This sequence has been shown by crystallographic and binding studies to be sufficient in distinguishing between TPR domains that bind Hsp90 and those that bind Hsp70(Scheufler et al., 2000). In pull-down experiments, full-length UNC-45, but not the TPR(-) UNC-45 construct, associates with Hsp90. Furthermore, surface plasmon resonance experiments indicate that Hsp90 directly binds recombinant TPR domain, and this interaction is blocked most efficiently by peptides corresponding to the C-terminus of Hsp90 but not that of Hsp70 or non-specific peptides(Barral et al., 2002).
Both full-length UNC-45 and the TPR(-) construct, but not the TPR domain alone, show biochemical chaperone activity(Fig. 3) towards the general chaperone substrate citrate synthase (CS). They protect CS against thermally induced aggregation, which occurs at concentrations of UNC-45 that are substoichiometric to CS. In addition, they enhance the renaturation of thermally inactivated CS when the reactions are incubated with one of its substrates, oxaloacetate (Barral et al.,2002).
UNC-45 protein binds directly to myosin S1, or the myosin head, which contains the motor domain, and functions as a chaperone for S1 by preventing its thermally induced aggregation (Fig. 3) (Barral et al.,2002). This result thus explains the specific sequestration of Rng3p by the myosin motor domain mutation E1 in the myo2 cytoskeletal type II myosin heavy chain of S. pombe described above(Wong et al., 2000). Yeast two-hybrid experiments show that C. elegans UNC-45 can bind to C. elegans NMY-2 cytoskeletal and HUM-2 type V myosin heavy chains (W. Ao and D. Pilgrim, personal communication). These independent results suggest that the most likely targets of the UCS proteins are the shared motor domains of the different isoforms and classes of myosin and that facilitation of the folding of these domains, rather than purely assembly of myosins, is the most general function of UCS proteins.
Problems and successes in myosin folding
To date, expression of recombinant myosin, its constituent subunits and various fragments have produced an interesting and relevant mix of details. Functional myosin light chains can be expressed in and purified from bacteria(Saraswat and Lowey, 1991). When expressed in bacteria, by contrast, the rod and light meromyosin (LMM,Fig. 4) portions of myosin heavy chains produce insoluble inclusion bodies; however, they may be solubilized in 6.0 M guanidine hydrochloride and remain soluble in high salt(Sohn et al., 1997). Such recombinant myosin rod fragments can then form characteristic paracrystalline assemblies similar to those of the corresponding native fragments(Sohn et al., 1997). These results are reminiscent of experiments showing that myosin rods derived from native vertebrate skeletal muscle myosins can be denatured and renatured to form proper α-helical coiled coils(Lowey et al., 1991). If mixtures of the rods of myosin heavy chain isoforms that segregate to form myosin homodimers in muscle are similarly denatured and renatured, homodimeric coiled coils form (Kerwin and Bandman,1991). These results indicate that myosin light chains and the rod domains of myosin heavy chains are capable of spontaneously folding under experimental conditions and probably in their cells of origin.
Expression of myosin motor domains as either S1 or heavy meromyosin (HMM,Fig. 4) fragments leads to very different results, even with co-expression of the light chains. To date,neither muscle nor non-muscle (cytoskeletal) myosin heads can be expressed as functional proteins in bacteria (McNally et al., 1988; Mitchell et al.,1986). The ability to express recombinant proteins through baculovirus infection of insect cells has permitted expression of HMMs derived from cytoskeletal and smooth muscle type II, V and VI myosins that are soluble and exhibit motor activity (Sweeney et al., 1998; Wang et al.,2000; Wells et al.,1999). However, neither cardiac nor skeletal muscle sarcomeric myosin HMMs can be obtained in soluble form from these cells (H. L. Sweeney and J. R. Sellers, personal communication). Two hypotheses can explain these results. First, bacteria lack additional eukaryotic factors required for the proper folding of the motor domains of several classes of cytoskeletal myosins. Second, non-striated muscle eukaryotic cells lack but striated muscle cells produce additional factors specifically required for the proper folding of sarcomeric myosin heads. Indeed, the folding of chicken skeletal muscle HMM in rabbit reticulocyte lysates is enhanced by a muscle-derived extract and muscle cells, but not kidney epithelial cell lines, produce functional recombinant skeletal muscle myosin(Srikakulam and Winkelmann,1999; Chow et al.,2002).
Vertebrates express general cell and striated muscle UNC-45 isoforms
Recent studies in our laboratory have shown that both the human and mouse genomes contain two genes for UCS-domain proteins [see accompanying article in this issue (Price et al.,2002)]. Within each species, both predicted amino acid sequences show between 30-40% identity with each of the three regions of the C. elegans UNC-45 protein and its Drosophila homologue. Within each species, the two predicted isoforms share ∼55% identity whereas a cross-species comparison between mouse and human shows >90% identity between the most similar members. The two murine UNC-45 isoforms show very different patterns of expression both in development and among adult organs. One isoform is expressed in multiple adult organs, including brain, kidney and liver, whereas the other isoform is highly expressed only in heart and skeletal muscles. These observations led us to designate the two as general cell (GC) and striated muscle (SM) isoforms.
The complete genomes of C. elegans, Drosophila melanogaster and the mosquito Anopheles gambiae appear each to contain only one UNC-45 gene. The appearance of a second UNC-45 gene may have ancient roots in vertebrate radiation that occurred possibly during bony fish evolution(Fig. 5). We identified putative GC and SM UNC-45 gene products in the genome of the pufferfish Fugu rupbripies on the basis of their 64% and 72% identities to the respective verified mammalian isoforms.(Price et al., 2002). In common with their mammalian counterparts, the two potential fish isoforms share only 54% identity. As the complete genomes of other vertebrates and chordate ancestors become available, it will be possible to identify when the second gene evolved.
All animal UNC-45 proteins have the three-domain structure found in C. elegans UNC-45: an N-terminal TPR domain, a unique central region and a C-terminal UCS domain. Blocks of homology within each region are maintained throughout the entire vertebrate and invertebrate proteins. The largest conserved block, LVGLCK, is near the end of the central region. Such conserved blocks are even larger if one considers only vertebrate species. The sites of mutations demonstrated in C. elegans and S. pombe UCS domains are identical or conserved throughout all identified homologues(Barral et al., 1998;Wong et al., 2000).
The two mouse UNC-45 isoforms are differentially expressed in development. At 9-11 days of murine embryonic development, the GC isoform mRNA is most prominently seen in the branchial arches, and the SM isoform mRNA is chiefly expressed in the heart (Price et al.,2002).
In C2C12 myogenic cell development in vitro, only the GC isoform mRNA is detected in proliferating myoblasts whereas SM mRNA is first detected during cell fusion and becomes the predominant isoform during myotube maturation. Antisense oligonucleotides to the GC isoform inhibit myoblast proliferation and fusion whereas antisense to the SM isoform appears to have its greatest effect upon sarcomere organization (Price et al., 2002). These results suggest that the GC isoform plays a role predominantly in cytoskeletal motility and the SM isoform specifically functions in myogenic processes, including sarcomeric thick filament assembly. Whether the two isoforms have distinct intrinsic functions or whether their different roles are determined by their differential expression is not yet clear and further experiments will be required to answer this question.
Targeted chaperone systems
The UCS domain family of proteins, in its interactions with myosins,resembles the Cdc37/p50 family of Hsp90 co-chaperones/chaperones, which target specific protein kinases from a wide phylogenetic spectrum(Kimura et al., 1997;Stepanova et al., 1997;Hartson et al., 2000). These proteins appear to be necessary for the folding of protein kinases into active conformations. Similarly, the Dpit47 protein in Drosophila is necessary for the production of active DNA polymerase and acts as an Hsp90 and possible Hsp70 co-chaperone (Crevel et al.,2001). One common function of UNC-45 and these other proteins may be to target Hsp70 and/or Hsp90, molecular chaperones implicated in aiding distinct stages of folding, to key client proteins such as myosin-type protein motors, selected protein kinases and DNA polymerases. In addition, since several co-chaperones, including UNC-45, exhibit chaperone activity as purified proteins, they might function as chaperones in vivo. A model of how UNC-45 might function in a myosin folding cycle is based upon current views of the function of the co-chaperone Hop, which interacts with both Hsp70 and Hsp90 in the folding of steroid receptors and other systems(Fig. 6). Hop acts a scaffold for recruiting Hsp70, Hsp90 and their associated co-chaperones (Hsp40 and p23,respectively) to immature steroid receptors. Binding of Hsp90 and other co-chaperones such as immunophilins is favored as the receptor matures and becomes competent for binding steroid(Kosano et al., 1998;Morishima et al., 2000;Pratt and Toft, 1997).
Conclusions and perspectives
The concept of client-specific molecular chaperones is relatively new and supported by a handful of examples in the literature. That myosins may need such a chaperone is becoming increasingly apparent from work on several myosins and the various members of the UCS family of proteins. The myosin head, well-conserved amongst the different myosin classes, is a multi-domain structure that does not fold spontaneously from denaturant or become functional when expressed in naïve cells. Yet myosins are involved in and required for virtually all types of cell motility, as evidenced by the number of myosin classes identified. It is clear that UCS proteins interact with myosins belonging to class II and class V, and whether or not they interact with all myosins is still an open question. The connection between the myosin head and the UCS proteins UNC-45 and Rng3p is described above, and the evidence suggests that they function as myosin chaperones. In the case of UNC-45 and its orthologs, the additional TPR domain might recruit Hsp90 to facilitate folding of myosin. Co-localization of UNC-45 with thick-filament-containing-A-bands (Ao and Pilgrim, 2000) suggests that, as part of its chaperone function,UNC-45 monitors the folded state of myosin after it is assembled and ensures that the myosin is functional after repeated cycles of actin binding and release. In addition to functioning as chaperones, the possibility that some members may function as `assemblases' also exists. Muscle and non-muscle myosins assemble into filamentous structures to perform their functions, but their coiled-coil rods, although necessary, may not be sufficient to mediate formation of such assemblies (Epstein and Fischman, 1991). Additional factors must function in the cell to ensure the fidelity of thick filament assembly. The exact functions of the UCS proteins are under investigation, and exciting progress is expected in this area.
We are grateful to Jose M. Barral for his critical discussions. Research in our laboratory was supported by grants from the Muscular Dystrophy Association and the National Institute of General Medical Sciences.