The enigma of LIM domains

GORDON N GILL

MINIREVIEW

The enigma of LIM domains LIM domains are two-zinc-finger structures found in proteins that have diverse functions. They are proposed to be protein dimerization motifs that assemble protein complexes necessary for growth, development and adaptive responses. Structure 15 December 1995, 3:1285-1289 LIM domains (LIMs) are protein motifs comprising approximately 55 residues, with the primary sequence CX2CX6-23 HX2CX 2CX 2 CX 16_23CX2C

(where X is

any amino acid). LIMs join an increasing number of sequence motifs, including SH2, SH3, phosphotyrosine binding, pleckstrin and leucine zipper domains, that are widely distributed amongst proteins of diverse function. These cysteine-rich LIM domains bind two atoms of Zn2 + , the tetrahedral coordination being S3 N and S4 [1]. The acronym LIM derives from the three homeodomain proteins first recognized as sharing this structural feature: lin-11, which functions in asymmetric division of Caenorhabditiselegans 20 vulvar blast cells [2]; isl-1, which binds the rat insulin I gene enhancer [3] but has its major function in motor neuron development; and mec-3, which is essential for proper differentiation of touch receptor neurons in C. elegans [4]. LIMs are widely distributed in nature, being found in plants, yeast, and a variety of metazoans. Most LIM proteins contain more than one LIM domain; particular LIMs within a protein are evolutionarily conserved as they are closer to the corresponding LIM in the same protein from other species than to other LIMs present within the same protein. This suggests specificity of function for individual LIMs and diversity of function within a protein. The mechanism of LIM function has not been established but the finding that proteins such as rhombotins (renamed LIM-only, LMO) and cysteine-rich proteins (CRP) contain very little sequence information that is not found within a LIM domain supports the hypothesis that LIMs act as protein-protein interaction motifs. Dimerization among LIMs and the existence of tyrosine-containing tight-turn structures within target proteins has been presented as evidence that LIMs recognize two different structural targets. Specificity in both the LIM domain and in its target for protein-protein interaction (and thereby its biological function) is necessary. Classes of LIM-domain proteins Using sequence comparison, Dawid, et al. [5] classified LIM domains into five groups. The N-terminal LIM of homeodomain proteins, LMO, and LIM kinase, belong to group 1, whereas the adjacent LIM is a member of group 2. The other groups are diverse, and are found in both LIM-only proteins and proteins containing LIMs at their C termini. LIM-domain proteins can be classified either by sequence comparison of their individual LIM domains or by the types of proteins in which LIMs are

found. A general scheme is shown in Figure 1. The first group consists of the LIM homeodomain proteins. In these proteins the LIM domains and homeodomain are located at the N-terminus and C-terminus, respectively. This is the largest and fastest growing group and contains homologous genes from C. elegans, Drosophila, Xenopus, zebrafish and vertebrates. The LIM-only proteins contain one to five LIMs and each protein may have multiple family members. Consistent nomenclature for this and other groups is under development. Proteins in the third group are similar only in the C-terminal location of LIMs, although most of the proteins are cytoplasmic. A fourth group contains any that do not easily fit into the above three. Such a general classification emphasizes that LIMs are indeed domains that have been conserved for protein interactions important in diverse functions. Structure of LIM domains LIM domains are signified by a specific spacing of cysteine residues in their primary sequence. The NMR structure of the C-terminal LIM of CRP indicated that two Zn 2+ atoms were bound independently in the Cys3His and Cys4 modules [6]. The core structure of this LIM domain is predominantly an antiparallel 3 sheet with the two Zn 2+-binding modules located at either end of a hydrophobic core. In Figure 2a the sequences of the three LIMs of enigma are aligned with LIM2 of CRP, with the secondary structural motifs highlighted above. Figure 2b shows features of the structure of CRP. Of the four sheets in CRP, 131 and 133 contain a rubredoxin-type turn characteristic of the metal chelating domains of iron-binding rubredoxins. The N-terminal Zn 2 +-binding module is formed by Cysll118, Cysl21, His139 and Cys142. Hisl39-Cysl42 forms a tight turn between 32 and 33. sheet, His139, at the C-terminal end of the second coordinates to Zn2 + through the N81 nitrogen. There is some variation in residues coordinating Zn2+ as LIM1 of enigma contains glutamine in place of the Cys142 of CRP and LIM2 of enigma contains aspartate in place of the cysteine at residue 169 of CRP. This pattern of variability is found in other LIMs and CRP with aspartate substituted for Cys169 continued to bind two Zn 2+ atoms [7]. LIM1 of enigma binds two atoms of Zn 2+ in a manner consistent with oxygen rather thaA sulfur coordination (D Winge, personal communication). Zn2 + binding in LIM domains is thus accomplished via S4, S3 N 1, S301, and S2 N 10 1 coordinations. Variability appears limited to the fourth and eighth cysteine positions as

© Current Biology Ltd ISSN 0969-2126

1285

1286

Structure 1995, Vol 3 No 12

Fig. 1. Types of LIM-domain proteins. Classification is adopted from [5] and references to the primary structures are contained therein. LM02, LIM only protein 2; CRP, cysteine-rich protein; SF3, sunflower 3, now named PLIM-1 (pollen LIM-1); LRG 1, LIM-Rho/rac GAP.

substitution of cysteine by histidine at the second position (Cysl21 in CRP) resulted in a molecule that bound only one Zn2 + atom in the C-terminal Cys4 module and substitution at the fifth cysteine position (Cysl42 in CRP) abolished all Zn2+ binding [7]. The N-terminal Zn 2 + module thus consists of CC (H,C,D) (C,H,E) and the C-terminal Zn2 + module of CCC (C,H,D). Metal binding is essential for protein structure and binding appears sequential, with the C-terminal module occupied first, followed by the N-terminal module. Hydrophobic residues constitute the core of LIM2 of CRP and similar hydrophobic residues are conserved in all LIM domains. The ring structures of Trp138 and Phe143 are aligned in parallel within the core of CRP; a ring structure is conserved in LIM domains at the Trp138 position but only conservation of a hydrophobic residue occurs at the position corresponding to Phe143. The two Zn 2+-binding modules differ in that the subsequent residues of the Cys4 module adopt an -helical conformation whereas the corresponding residues in the Cys3His module form a tight turn. The unique structure of LIMs does not readily explain how individual LIM

domains specifically interact with target molecules and therefore the structure of a LIM-domain-target complex is required. Specificity is likely to reside in the central portion of the LIM, that is stabilized by the two Zn 2 + modules at the ends. LIM-domain interactions Tetrahedral coordination of Zn 2+ is a reiterated occurrence essential for the tertiary structure of diverse protein motifs including zinc fingers, ring fingers and B boxes. The structure of the LIM domain of CRP differs significantly from that of zinc and ring fingers. Although the global fold of a LIM domain is unique, the C-terminal Cys4 module of CRP LIM2 resembles that of the GATA1 and glucocorticoid receptor transcription factors. Despite this homology, there is no evidence for LIM domains binding to specific DNA sequences and available evidence favors the cncept that LIM domains function as protein-protein interaction motifs for assembly of functional complexes. The existence of LIM-only proteins (i.e. proteins that lack other sequence motifs) provides one of the strongest arguments that LIM domains specify protein-protein interactions. Two structural

LIM domains Gill

Fig. 2. Structure of LIM domains. (a) Alignment of LIM2 of CRP with the three LIM domains of enigma. The cysteine and histidine residues that form the two Zn 2+ coordination modules are boxed and in yellow and the conserved residues of the hydrophobic core located between the two modules are in green. The orthogonally-arranged antiparallel p sheets are indicated by brackets. (b) The core structure of LIM2 of CRP, determined by NMR. Side chains of the Zn 2 + coordination residues and of the hydrophobic core are colored as in (a). Zn 2+ atoms are represented by red spheres. The figure was constructed on the basis of the data from [6].

targets have been identified for LIMs: other LIMs and tyrosine-containing tight turns, but proof of function is not available for either. Using gel overlay and GST-fusion binding assays, Schmeichel and Beckerle [8] demonstrated that zyxin, which contains three LIMs at its C terminus, bound to CRP. Interaction with CRP was specific for LIM1 of zyxin and neither LIM2 nor LIM3 of zyxin, nor the LIMs of mec-3, bound. Using similar techniques, ot-actinin, the other protein associated with zyxin in adhesion plaques, bound the N-terminal proline-rich region of zyxin, not the LIM domains. Human CRP was shown to homodimerize [9]; specificity was not apparent as interactions occurred with either of the two LIM domains of CRP. Although these studies indicate LIM-LIM interactions may occur, they do not easily explain cooperative interactions involving LIM-containing proteins that enhance transcription as the putative targets lack LIM domains. Enigma contains three LIMs at its C terminus [10]; it was discovered in a two-hybrid sreen using exon 16 of the insulin receptor (InsR), a 22 amino acid region characterized by two tyrosine-containing tight turns [11]. Only LIM3 of enigma bound specifically to the proximal tyrosine-containing tight turn that is essential for ligandinduced endocytosis of InsR. LIM3 of enigma specifically bound a short fragment of exon 16 with the

sequence GPLGPLYA and did not recognize endocytic and other trafficking codes present in other receptors, many of which consist of tyrosine-containing tight turns. More than 12 LIMs have now been tested and only LIM3 of enigma binds InsR via exon 16. Thus, there is specificity in both the LIM and the target, a feature expected to be necessary for in vivo function. It has recently been demonstrated that LIM2 of enigma recognizes a region containing a tyrosine-containing tight turn motif in a second receptor tyrosine kinase (R-Y Wu, KP Durick, SS Taylor and GN Gill, unpublished data). Specificity is high as there is no cross-reactivity of LIM2 and LIM3 of enigma with their respective targets. Function of LIM domains Given the initial recognition of LIMs as structural motifs present in homeodomain proteins identified by defective development in C. elegans it is not surprising that LIM proteins are also important regulators of development in vertebrates. Various LIM homeodomain proteins display unique patterns of expression during development and disruption of these genes disrupts normal development. In Xenopus gastrula embryos, Xlim-1 is expressed in the anterior-most dorsal mesoderm, within Spemann's organizer, a region that gives rise to the head and forebrain [12]. The mouse homolog LIM 1 (Lhx- in the revised nomenclature) shows a similar distribution as well as later

1287

1288

Structure 1995, Vol 3 No 12 expression in portions of the central nervous system and kidney, a pattern also exhibited by Xlim-1 [13]. Mice that are homozygous for disrupted Lhx-1 lack all head structures anterior to the hindbrain [14]. These striking results implicate Lhx-1 as essential to the organizer's function in the development of head structures. Body development was intact, although the few headless mice that survived to birth lacked kidneys and gonads consistent with a second function that parallels the later expression of Lhx-1 in mouse development. Lhx-1 is hypothesized to control transcription of target genes essential for cell specification and for recruitment of these cells into developing midbrain, forebrain and head structures. Isl-l, initially cloned on the basis of its interaction with the rat insulin I gene enhancer, is expressed in pancreatic islets, but more strikingly appears early in motor neuron development, before the expression of differentiated motor neuron markers [15]. Isl-1 expression is induced by sonic hedgehog (Shh), a peptide morphogen. Shh, produced by the notochord in the caudal portion of the developing neural tube, induces ventral floor plate and motor neurons in a concentration-dependent manner [16], the motor neurons being identifiable by the appearance of Isl-1. Shh is expressed along the entire ventral domain of the neural tube and induces Isl-1 positive cells in both the diencephalon and telencephalon. The appearance of these cells is distinct from that of motor neurons [17]. It is now known that Isl-1 is required for neuroepithelial cells to differentiate into motor neurons (S Pfaff, personal communication) but this may be only one of its multiple functions in development. LIM homeodomain proteins are not the only LIM proteins that control important developmental decisions. The LIM-only protein rhombotin 2 (LM02) is essential for erythroid development [18]. LMO1 and 2 were initially identified in chromosomal translocations into the T-cell receptor locus in acute T-cell leukemia [19-21]. Because these genes are not normally expressed in T-cells, this chromosomal translocation was proposed to deregulate their transcription. This hypothesis was confirmed, by targeting LMO 1 and LM02 expression via T-cell-specific promoters in transgenic mice [22,23]. Aberrant expression of LMO1 and 2 in T cells resulted in acute T-cell leukemia, although with a long latency period, suggesting that the accumulation of secondary defects may be necessary for full malignant transformation. LMO1 and LM03 are expressed primarily in the brain whereas LM02 is more widely expressed; fetal liver, the major site of erythropoiesis, has the highest level of expression. Targeted disruption of LM02 resulted in embryonic lethals due to failure of erythropoiesis [18]. Normal in vitro differentiation of ES cells to hematopoietic colonies was also abrogated by disruption of LM02 confirming a direct involvement for LM02 in erythroid development. Two transcription factors involved in erythroid development are the Tal 1 helix-loop-helix and the GATA 1 Zn2+-finger proteins. Targeted disruption of GATA 1

results in defective erythroid development with arrest occurring at a similar stage to that following disruption of LM02 [24]. Tal 1 is co-expressed with LM02 in erythroid cells and either, or both together, may be rearranged in acute T-cell leukemia. Co-immunoprecipitation experiments revealed an association between Tal 1 and LM02 [25]. In T-cell leukemia lines, similar experiments revealed that LMO1 associated with Tal 1 to a small extent and to a greater extent with a 46 kDa protein of unspecified function. Functional interactions between LMO1 and 2 and Tal 1 were demonstrated using two hybrid approaches in Jurkat cells [26]. Tal 1 showed specificity for the LIM domains of LMO1 and 2; CRIP, CRP and zyxin LIMs did not interact with Tal 1. LM02 interactions were restricted to Tal 1 and the related bHLH proteins Tal 2 and Lyl 1 and it did not interact with a variety of other bHLH proteins. Moreover, two hybrid analysis indicated that LM02 enhanced the transcriptional effects of the heterodimeric E47-Tal 1 complex that binds to DNA E-box sequences (CANNTG). These results support the hypothesis that LM02 (and LMO1) interacts with specific bHLH proteins to enhance transcription of target genes essential for developmental decisions. The Zn 2+-finger GATA1 transcription factor is a second target for LMO interaction [27]. LMO proteins that contain two LIMs might, in fact, assemble diverse transcription factors into unique combinations. A conceptually similar mechanism has been proposed for LIM homeodomain proteins as synergistic activation of the rat insulin I gene promoter by the LIM homeodomain protein lmx-1 and the bHLH E47 protein requires the LIMs of lmx-1 [28]. Removal of the LIMs of lmx-1 or substitution of the LIMs of Isl-1 for those of lmx-1 abolished the synergistic activation, attesting to specificity of LIMs for this reaction. Whether the functional synergy between E47 that binds to an E box and lmx-1 that binds to a distinct 'A' element in the promoter reflects direct or indirect protein interactions remains uncertain. The LIM homeodomain protein mec-3 and the POU homeodomain protein unc 86 cooperatively bind DNA sites in the mec-3 promoter in vitro [29]. Although this direct interaction does not appear to be dependent on the LIM domains of mec-3, deductions from analysis of mutant phenotypes suggest the cooperative interactions are functionally important. Lhx 3 (P-LIM), a LIM homeodomain highly localized to the developing anterior pituitary gland, synergizes in transcriptional activation with Pit 1, a POU homeodomain protein [30]. Because homeodomains appear to bind DNA better when trimmed of other sequences, DNA binding and enhanced function by LIM-deleted homeodomains must be interpreted with caution. Because LIMs function as protein-protein interaction motifs, it should be possible to identify interacting partners and to deduce pathways important for signal transduction, development, growth and cytoskeletal rearrangements. The precedence for this was established by SH2 and SH3 domains which provided the key to

LIM domains Gill Table 1. Proposed functions of LIM proteins. Development

LIM homeodomain LIM-only LMO2, MLIM

Cytoskeletal structure

paxillin, zyxin, CRP

Signalling and trafficking

LIM kinase, enigma

Growth control

transformation: LM01 and 2 anti-transformation: CRP, ril

assembling the pieces of a highly conserved mitogenic and developmental pathway [31]. Many interesting LIM proteins have been described and their roles in cellular function will clearly be diverse (Table 1). A number of gene knockouts are in the pipeline and will undoubtedly highlight the importance of LIM homeodomain proteins in development. The results of such studies underscores the importance of solving the enigma of how LIM domains carry out their recognition function in diverse proteins. References 1. Michelsen, .W., Schmeichel, K.L., Beckerle, M.C. & Winge, D.R. (1993). The LIM motif defines a specific zinc-binding protein domain. Proc. Natl. Acad. Sci. USA 90, 4404-4408. 2. Freyd, G., Kim, S.K. & Horvitz, H.R. (1990) Novel cysteine-rich motif and homeodomain in the product of the Caenorhabditis elegans cell lineage gene lin-11. Nature 344, 876-879. 3. Karlsson, O., Thor, S., Njorberg, T., Ohlsson, H. & Edlund, T. (1990). Insulin gene enhancer binding protein Isl-1 isa member of a novel class of proteins containing both a homeo- and a Cys-His domain. Nature 344, 879-882. 4. Way, J.C. & Chalfie, M. (1988). mec-3, a homeobox-containing gene that specifies differentiation of the touch receptor neurons in C. Elegans. Cell 54, 5-16. 5. Dawid, I.B., Toyama, R.and Taira, M. (1995). LIM domain proteins. C.R. Acad. Sci., Paris 318, 295-306. 6. Perez-Alvarado, G.C., Miles, C., Michelsen, J.W., Louis, H.A., Winge, D.R., Beckerle, M.C. & Summers, M.F. (1994). Structure of the carboxy-terminal LIM domain from the cysteine rich protein CRP. Nat. Struct. Biol. 1, 388-397. 7. Michelsen, J.W., Sewell, A.K., Louis, H.A., Olsen, J.l., Davis, D.R., Winge, D.R. & Beckerle, M.C. (1994). Mutational analysis of the metal sites in a LIM domain. J. Biol. Chem. 269, 1108-1113. 8. Schmeichel, K.L. & Beckerle, M.C. (1994). The LIM domain is a modular protein-binding interface. Cell 79, 211-219. 9. Feurerstein, R., Wang, X., Song, D., Cooke, N.E. & Liebhaber, S.A. (1994). The LIM/double zinc-finger motif functions as a protein dimerization domain. Proc. Nat. AcadSci USA 91, 10655-10659. 10. Wu, R.-Y. & Gill, G.N. (1994). The LIM domain recognition of a tyrosine-containing tight turn. J. Biol. Chem. 269, 25085-25090. 11. Backer, J.M., Shoelson, S.E., Weiss, M.A., Hua, Q.X., Chatham, R.B., Harig, E., Cahill, D.C. & White, M.F. (1992). The insulin receptor juxtamembrane region contains two independent tyrosine/p-turn internalization signals. J. Cell Biol. 118, 831-839. 12. Taira, M., amrich, M., Good, P.J.& Dawid, I.B. (1992). The LIM domain-containing homeo box gene Xlim-1 is expressed specifically in the organizer region of Xenopus gastrula embryos. Genes Devel. 6, 356-366. 13. Barnes, J.D., Crosby, J.L., Jones, C.M., Wright, C.V.E. & Hogan, B.L.M. (1994). Embryonic expression of Lim-1, the mouse homolog of Xenopus Xlim-1, suggests a role in lateral mesoderm differentiation and neurogenesis. Devel. Biol. 161, 168-1 78. 14. Shawlot, W. & Behringer, R.R. (1995). Requirement for Lim in head-organizer function. Nature 374, 425-530.

15. Ericson, J., Thor, S., Edlund, T., Jessell, T.N. & Yamada, T. (1992). Early stages of motor neuron differentiation revealed by expression of homeobox gene Islet-l. Science 256, 1555-1560. 16. Roelink, H., Porter, J.A., Chiang, C., Tanabe, Y.U., Chang, D.T., Beachy, P.A. & Jessell, T.M. (1995). Floor plate and motor neuron induction by different concentrations of the amino-terminal cleavage product of sonic hedgehog autoproteolysis. Cell 91, 445-455. 17. Ericson, J., Murh, J., Placzek, M., Lints, T., Jessell, T.M. and Edlund, T. (1995). Sonic hedgehog induces the differentiation of ventral forebrain neurons: a common signal for ventral patterning within the neural tube. Cell 81, 747-756. 18. Warren, A.J., Colledge, W.H., Carlton, M.B.L., Evans, M.J., Smith, A.J.H., & Rabbitts, T.H. (1994). The oncogenic cysteine-rich LIM domain protein Rbtn2 is essential for erythroid development. Cell 78, 45-57. 19. Boehm T., Baer R., Lavenir I., Forster A., Waters, J.J., Nacheva, E. & Rabbitts T.H. (1988). The mechanism of chromosomal translocation t(11 ;14) involving the T-cell receptor Cd2 locus on human chromosome 14g11 and a transcribed region of chromosome 11p15. EMBOJ. 7, 385-394. 20. McGuire, E.A., Hickett, R.D., Pollock, K.M., Bartholdi, M.F., O'Brien, S.J. & Korsmeyer, S.J. (1989). The t(11 ;14) (p15;ql 1) in a T-cell acute lymphoblastic leukemia cell line activates multiple transcripts, including Ttg-1, a gene encoding a potential zinc finger protein. Mol. Cell. Biol. 9, 21214-2132. 21. Royer-Pokora, B., Loos, U., Ludwig, W.-D. (1991). TTG-2 a new gene encoding a cysteine-rich protein with the LIM motif, is overexpressed in acute T-cell leukaemia with the t(11;14)(p13;q11). Oncogene 6, 1887-1893. 22. McGuire, E.A., Rintoul, C.E., Sclar, G.M. & Korsmeyer, S.J. (1992). Thymic overexpression of Ttg-1 in transgenic mice results in T-cell acute lymphoblastic leukemia/lymphoma. Mol. Cell Biol. 12, 4186-4196. 23. Larson, R.C., Fisch, P., Larson, T.A., Lavenir, I., Langford, T., King, G. & Rabbitts, T.H. (1994). T celltumours of disparate phenotype I mice transgenic for Rbtn-2. Oncogene 9, 3675-3681. 24. Pevny, L, Simon, M.C., Robertson, E., Klein, W.H., Tasia, S.-F., D'Agati, V., Orkin, S.H. & Costantini, F. (1991). Erythroid differentiation in chimaeric mice blocked by a targeted mutation in the gene for transcription factor GATA-1. Nature 349, 257-260. 25. Vale-Archer, V.E., Osada, H., Warren, A.J., Forster, A., Li, J., Baer, R. & Rabbitts, T.H. (1994). The LIM protein RBTN2 and the basic helixloop-helix protein TALl are present in a complex in erythroid cells. Proc. Natl. Acad. Sci. USA 91, 8617-8621. 26. Wadman, I., Li, J., Bash, R.O., Forster, A., Osada, H., Rabbitts, T.H. & Baer, R. (1994). Specific in vivo association between the bHLH and LIM proteins implicated in human T cell leukemia. EMBOJ. 13, 4831-4839. 27. Osada, H., Gruth, G., Axelson, H., Forster, A. & Rabbitts, T.H. (1995). Association of erythroid transcription factors: complexes involving the LIM protein RBTN2 and the zinc-finger protein GATA1. Proc. Natl. Acad Sci USA 92, 9585-9589. 28. German, M.S., Wang, J.,Chadwick, R.B. & Rutter, W.J. (1992). Synergistic activation of the insulin gene by a LIM-homeo domain protein and a basic helix-loop-helix protein: building a functional insulin mini-enhancer complex. Genes Devel. 6, 2165-21 76. 29. Xue, D., Tu, Y. & Chalfie, M. (1993). Cooperative interactions between the Caenorhabditis elegans homeoproteins UNC-86 and MEC-3. Science 261, 1324-1328. 30. Bach, I., Rhodes, S.J., Pearse, R.V., Heinzel, T., Gloss, B., Scully, K.M., Sawchenko, P.E. & Rosenfeld, M.G. (1995). P-LIM, a LIM homeodomain factor, is expressed during pituitary organ and cell commitment and synergizes with Pit-1. Proc. Natl. AcadSci USA 92, 2720-2724. 31. Pawson, T. & Gish, G.D. (1992). SH2 and SH3 domains: from structure to function. Cell 71, 359-362.

Gordon N Gill, Department of Medicine, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0650, USA.

1289