An essential cell division gene of Drosophila, absent from Saccharomyces, encodes an unusual protein with tubulin-like and myosin-like peptide motifs

Abstract

Null mutations at the misato locus of Drosophila melanogaster are associated with irregular chromosomal segregation at cell division. The consequences for morphogenesis are that mutant larvae are almost devoid of imaginal disk tissue, have a reduction in brain size, and die before the late third-instar larval stage. To analyze these findings, we isolated cDNAs in and around the misato locus, mapped the breakpoints of chromosomal deficiencies, determined which transcript corresponded to the misato gene, rescued the cell division defects in transgenic organisms, and sequenced the genomic DNA. Database searches revealed that misato codes for a novel protein, the N-terminal half of which contains a mixture of peptide motifs found in α-, β-, and γ-tubulins, as well as a motif related to part of the myosin heavy chain proteins. The sequence characteristics of misato indicate either that it arose from an ancestral tubulin-like gene, different parts of which underwent convergent evolution to resemble motifs in the conventional tubulins, or that it arose by the capture of motifs from different tubulin genes. The Saccharomyces cerevisiae genome lacks a true homolog of the misato gene, and this finding highlights the emerging problem of assigning functional attributes to orphan genes that occur only in some evolutionary lineages.

At the present time, approximately a third of the genes in both Saccharomyces cerevisiae and the archaebacterium Methanococcus jannaschii can be classified as orphans—i.e., either they have no known biochemical properties or they have no obvious relatives in protein and DNA databases (1, 2). Many other proteins have been assigned to families on the basis of a single domain, and hence their biochemical attributes are, at best, only partially catalogued. Extrapolation of these results to other organisms implies that

(i) combinatorial shuffling in different evolutionary lineages has produced a pool of orphan genes which characterize individual lineages—e.g., the immunoglobulin loci of vertebrates, or

(ii) some orphans may be more apparent than real; proteins in different phyla are known to have similar three-dimensional structures, but their relationships sometimes remain unrecognized owing to extensive divergence at the sequence level—e.g., human interleukin-1β and soybean trypsin inhibitor (3).

In this report we focus on a particular orphan gene in Drosophila, which we have named misato (beautiful dawn). It codes for a protein with a surprising combination of peptide motifs, each of which is characteristic of a different part of a conventional tubulin. It also shares a motif with myosin heavy chain proteins. The gene occurs in a region of the fly X chromosome that has been well studied genetically (4), and it is a representative of a class of genes termed “late lethals,” a subset of which is almost diagnostic for essential cell cycle genes (5). Some mutant organisms carrying null alleles at this locus survive up until the early third-instar larval stage but are unable to complete morphogenesis because cell cycle defects lead to under-developed imaginal disks. The additional finding that the S. cerevisiae genome lacks a true homolog of this essential Drosophila gene highlights the limitations of comparative genomics. It reinforces the necessity for both conventional and unusual screens to analyze phenotypes in individual phyla (6–9) and the absolute necessity for crystallographic data at the protein level (10).

MATERIALS AND METHODS

Chromosome Preparations and Crossing Programs.

Neuroblast cells from larvae of different genotypes were prepared as previously described without the use of colchicine (11). The lethal alleles tested at the misato locus (previously designated as LB20), were LB20, LB20^C27, LB20^DA618, and LB20^SK1 (ref. 4; M.-T.Y., unpublished work); the lethal allele tested at the A112 locus was A112^17–62 (4). To determine the lethal stage of hemizygous or homozygous mutants, versus the lethal stage of mutant/deficiency individuals, LB20^SK1/Y male larvae were compared with LB20^SK1/JA117 female larvae. To examine the mitotic consequences of mutations at the misato locus, and to distinguish the sex chromosomes more easily from the large autosomes, mutant male larvae of the genotypes misato/0 and misato/Y, or their control sibs FM7/0 and FM7/Y, were generated from crosses of the type misato/FM7 females to XY^L.Y^S, y² sc cv v f/0 males, or from misato/FM7 females to FM7/Y males. Male larvae of the genotypes FM7/0 and FM7/Y have white malpighian tubules and can thus be readily distinguished from their mutant misato/0 and misato/Y sibs, which have yellow malpighian tubules. The mutants develop very slowly and die in the second or early third larval instars. They have almost no imaginal disk tissue and have significantly reduced neuroblast tissue and a nearly transparent body.

P-Element Constructs and Germ-Line Transformations.

Four parts of the misato region (denoted transforming fragments, A, B, C, and D; see Fig. 1) were cloned into the vector pW8 by previously described procedures (12–15) and used for transgenic analysis. Construct A consists of a 10.9-kb BamHI genomic fragment; construct B is a 4.8-kb SpeI–PstI fragment; construct C is an 11.6-kb BglII fragment; and construct D is a 15.3-kb EcoRI–XhoI genomic fragment used previously to rescue the sluggish gene. Germ-line transformations using constructs A, C, and D were carried out as previously described (12–14). Germ-line transformation using construct B was implemented using embryos of strain yw; +/+; Sb, P[ry⁺, Δ2–3]/TM6, Ubx (15). Three independent genomic insertions were obtained for fragment A, six for fragment B, three for fragment C, and three for fragment D.

Figure 1.

Genetic and molecular properties of genes, chromosomal deficiencies, and transforming fragments in the misato region. The genetic complementation groups within the region are as shown. The four transcription units are aligned on the genomic DNA; the intron–exon structure of only misato is shown. Restriction sites are E, EcoRI; P, PstI; Bg, BglII; S, SpeI; and B, BamHI. The breakpoints of the four chromosomal deficiencies are indicated by hatched boxes. The four transforming fragments A, B, C, and D are as indicated.

Molecular Protocols.

DNA and RNA extractions, blotting, hybridization, molecular isolation and cloning of cDNAs from standard Drosophila cDNA libraries, and genomic and cDNA sequencing were carried out as previously described (12–17). A number of misato cDNAs were incomplete at their 5′ and 3′ ends as determined from comparison to our genomic sequence. Reverse transcription (RT)–PCR analyses and the isolation of longer cDNAs (D. Slifka, A. B. Kasprzak, J. Cotsell, D. Hall, H. Campbell, and G.L.G.M., unpublished work), confirmed the position of the ATG codon and the position of the first intron, which was inferred from bioinformatic analysis of our genomic sequence data (S. Karlin, R. Maleszka, and G.L.G.M, unpublished work). Genomic and cDNA sequences were determined on both strands using Sequenase (United States Biochemical) and by manual as well as automated methods using the Pharmacia ALF DNA Sequencer. Database searches were carried out at the National Center for Biotechnology Information (NCBI) by employing the blast network service. clustal w and macaw were used to generate multiple sequence alignments.

RESULTS AND DISCUSSION

Genetic and Molecular Analysis of the misato Region.

The molecular genetics of the chromosomal segment under study is shown in Fig. 1. This region contains two essential loci, denoted lethals A112 and LB20, and is flanked distally by the sluggish locus and proximally by the tumorous head locus, neither of which is essential for organismal viability. These loci have all been ordered by deficiency analysis. Deficiency Q539 uncovers the mutant phenotypes of sluggish, A112, and LB20, but not tumorous head; deficiency JC77 uncovers A112, LB20, and tumorous head; deficiency HM44 uncovers sluggish and A112; and deficiency JA117 uncovers both A112 and LB20 (4).

We found four transcription units in this region, the most distal of which, sluggish, codes for proline oxidase, an enzyme involved in glutamate biosynthesis (14). The next transcription unit codes for a member of the DEAD-box family of ATP-dependent RNA helicases (13), and we have denoted it Helicase (Hlc). This is followed by the misato transcription unit and then by a transcription unit which we have called la costa (lcs), because it contains a leader sequence, a collagen-related repetitive motif, and a serine/threonine-rich motif found in many different proteins.

To align the genetic and molecular maps, we (i) determined the approximate molecular breakpoints of the four deficiencies; (ii) isolated all available cDNAs from the 35-kb region; (iii) sequenced the genomic DNA except for a small region near the proximal breakpoint of deficiency Q539; and (iv) analyzed the phenotypic consequences of transgenic organisms carrying one each of the wild-type portions of the regions denoted A, B, C, and D (see Materials and Methods).

The breakpoints of the four chromosomal deficiencies were determined by Southern analysis of the genomic DNA of individuals heterozygous for rearranged and normal chromosomes, using radiolabeled probes from across the entire region (data not shown). We found that deficiency Q539 has its breakpoint in the most proximal EcoRI–PstI fragment (Fig. 1), and this breakpoint anchors the centromeric end of the region, without interrupting any transcription unit. Deficiencies JC77 and HM44 both break in a 0.5-kb PstI–EcoRI fragment, but the deficiencies extend in opposite directions. At the molecular level, JC77 interrupts the Helicase transcription unit and deletes misato and la costa. HM44 also interrupts the Helicase transcription unit, and it deletes the sluggish transcription unit and further genes distal to this breakpoint. Unlike deficiencies Q539, JC77, and HM44, which extend for hundreds of kilobases, deficiency JA117 is a small deletion of approximately 1.5 kb which largely removes the 5′ regions of the Helicase and misato transcription units (Fig. 1). These deficiency-based data, in combination with the positions of the four transcription units and the genetic data described above, make it likely that the Helicase transcription unit corresponds to the lethal A112 locus and that the misato transcription unit corresponds to the lethal LB20 locus.

Rescue of the Mutant Phenotypes Associated with Alterations in the misato Region.

To obtain further evidence that the Helicase and misato transcription units correspond to the A112 and LB20 loci, respectively, we constructed transgenic organisms that carried different combinations of mutant alleles, deficiencies, and transforming fragments for the region (Fig. 1). Transgenic organisms containing a single copy of each of these fragments were crossed to individuals carrying mutant alleles at the A112, LB20, and sluggish loci, as well as to individuals carrying deficiency JA117, which interrupts both the Helicase and misato transcription units.

Three independent transgenic lines containing genomic fragment A, (which carries wild-type misato and la costa transcription units), each rescue the lethality caused by mutations at the LB20 locus, namely the mutations LB20 and LB20^C27. However, fragment A-containing individuals do not rescue a mutation at the A112 locus, namely the mutation A112^17–62, nor do they rescue the inviability of deficiency JA117/Y males.

The two tested transformant lines carrying fragment B (which carries a wild-type Helicase transcription unit) both rescue the lethality caused by a mutation at the A112 locus (namely the mutation A112^17–62). However, neither of these transformant lines rescues the mutations at the adjacent LB20 locus, namely the mutations LB20, LB20^C27, and LB20^SK1.

All three transformants carrying fragment C (which have wild-type copies of both the Helicase and misato transcription units), rescue the lethality of individuals carrying mutations at the A112 and LB20 loci. Fragment C-bearing transformants also rescue the lethality of deficiency JA117/Y males, a deficiency that uncovers lethal alleles at both the A112 and LB20 loci.

All three transformant lines carrying fragment D (which carries a wild-type sluggish transcription unit) rescue the sluggish phenotype but not the lethality caused by a mutation at the adjacent A112 locus, namely the A112^17–62 allele.

These transgenic, mutant, and deficiency combinations provide strong evidence that the A112 and LB20 loci correspond to the Helicase and misato transcription units, respectively, and henceforth we refer to them by these designations. Finally, since deficiency Q539 is not deficient for the tumorous head gene, the la costa transcription unit is unlikely to correspond to the tumorous head locus.

Cell Biology of Null Alleles at the misato Locus.

In attempting to determine the primary defect caused by lesions in the gene, we carried out germ-line clone analysis of three misato alleles (LB20, LB20^C27, and LB20^DA618), and we found that germ cells are unable to survive when homozygous for these alleles (4). The misato protein is clearly an essential one for normal functioning of the female germ line. We also examined homozygous mutant misato embryos derived from heterozygous misato/wild type mothers, and we found that none of the mutant embryos had any central or peripheral nervous system defects (4). It is likely therefore, that a significant maternal contribution of wild-type product from heterozygous mothers allows homozygous mutant embryos to develop apparently normally and that the lethality in the larval instars occurs as a result of the depletion of wild-type maternal product. Furthermore, the stage of lethality is similar whether larvae are homozygous for mutant misato alleles or are heterozygous for a mutant allele and deficiency JA117. Therefore the alleles used in this study are likely to be null, or nearly so.

In looking hard at the phenotype of mutant larvae, we found that they had very little imaginal disk tissue and significantly reduced neural tissue, a finding considered almost diagnostic for essential cell cycle genes (5). We therefore carefully examined cell division, chromosome numbers, and morphologies in larval brains of mutant misato males, and we found that more than 60% of the cells had aneuploid or polyploid chromosome complements. These findings are illustrated in Fig. 2: a normal diploid XY cell is shown in A, a tetraploid cell (4A XXYY) is shown in B, and cells with aneuploid complements, in which the chromosomes are typically hypercondensed, are shown in C and D. Except in cases of extreme chromatin condensation, we were able to score for the presence of the dot-like chromosome 4s, the X and Y chromosomes, and the number of large autosomes in a cell.

Figure 2.

Chromosome complements in neuroblast tissue of mutant misato males of the genotype LB20/Y. (A) Diploid cell. (B) Tetraploid cell. (C and D) Aneuploid complements with hypercondensed chromosomes. (All panels are at the same magnification.)

The distribution of chromosome numbers per cell in mutant individuals is illustrated in Fig. 3. Our data from 1,538 cells reveal that minimally 60% of the cells have aneuploid chromosome complements, and that the distribution of chromosome numbers is highly skewed above the diploid value. Any cell deficient for either an X chromosome or both copies of an autosome will be inviable. We presume that those cells that have a normal chromosome complement still retain sufficient maternal product for normal chromosomal segregation.

Figure 3.

Histogram of the variation in chromosome number in neuroblast cells from second- and early-third-instar mutant misato males. Pooled data from LB20/Y, LB20^C27/Y, LB20^DA618/Y, and LB20/O.

We also find from a detailed analysis of the chromosomal constitutions of individual cells that chromosome number (for any given chromosome, X, Y, 2, 3, or 4) increases approximately linearly with the number of chromosomes in a cell. Thus the absence of the normal misato product does not differentially affect the probability of nondisjunction of autosomes versus sex chromosomes, or of small versus large chromosomes. Finally, examination of metaphase and anaphase figures reveals that cytokinesis is normal and that we are dealing with a mitotic defect. The mutant phenotype at the cellular level is thus similar to a class of mutants affecting chromosome condensation previously described in Drosophila (5), but the underlying processes responsible for these condensation defects are unknown.

We also examined cell division in larval brains of transgenic organisms that carried a wild-type copy of misato on an autosome, together with a lethal allele of misato (LB20) on the X chromosome. These organisms exhibit normal cell division, have normal-sized imaginal disks and normal brain tissues, and reveal no evidence of aneuploidy (data not shown). It is clear that the consequences of being mutant at the misato locus are irregular mitotic segregation and hypercondensed chromatin, ultimately leading to cell death and the inability to complete morphogenesis. What, then, are some of the potential properties of the gene product that lead to these phenotypes?

The misato Gene and Its Predicted Protein.

We sequenced a number of misato cDNAs as well as the underlying genomic DNA, aligned the cDNAs on the genomic sequence, and deposited the cDNA and genomic sequences in public databases. The alignment revealed that misato contains two small introns of 55 and 60 bp. Database searches showed that the top 50 matches to the predicted misato product were tubulins, with sequence similarities being highest in the N-terminal half of misato, in two regions that we have have designated as segments I and III (Fig. 4a). In addition, database searches revealed a region with excellent sequence similarity to a motif found in myosin heavy chain proteins, which we have denoted segment II (Fig. 4 a and d).

Figure 4.

Amino acid sequence alignments. (a) misato and β-tubulin from Dictyostelium discoideum (SwissProt accession no. SP P32256P32256). (b) Segment I, N-terminal portion of misato, maize γ-tubulin [Protein Identification Resource (PIR) S43329], and human γ-tubulin (SP P23258). (c) Segment III of misato and α-tubulin from D. discoideum (SP P32225) and S. cerevisiae (PIR S50871). (d) Segment II of misato and myosin heavy chain sequences from D. melanogaster (PIR C35815), rabbit (SP P04460), and human (PIR S12458). Identical residues are in solid boxes, and conservative changes are in lightly shaded boxes.

A comparison of the predicted misato product with that of the highest scoring database relative, namely a β-tubulin from Dictyostelium discoideum, reveals a number of regions of excellent sequence similarity. In addition, when segment I of misato is compared with γ-tubulins from Zea mays and humans, there is 60% sequence similarity and nearly 50% amino acid identity (Fig. 4b).

When segment III of misato is compared with α-tubulins from D. discoideum and S. cerevisiae, there is 60% sequence similarity and 44% sequence identity (Fig. 4c). While the sequence similarity of segment III to the α-tubulins is striking, nothing is known of the properties of this segment in either the α- or β-tubulins. It lies between two highly conserved peptides which are presumptive components of the β-tubulin GTP-binding site, namely KGHYTEG and GGGTGSG (10), yet neither of these motifs is conserved within the misato sequence. This difference between misato and the β-tubulins extends to a peptide lying within segment I. In this tubulin peptide, a highly conserved β-tubulin cysteine (Cys-12) can be photocrosslinked to the guanine base of the bound GTP (10), yet this cysteine is not present in segment I of misato, and the marked corruption of the KGHYTEG and GGGTGSG motifs indicates that misato is not a GTP-binding protein. In addition, the MREI motif in segment I, which is known to contribute to the autoregulation of tubulin expression, is disrupted in the misato sequence, as it is in the γ-, δ-, and ɛ-tubulins. The δ-tubulin of Caenorhabditis elegans, for example, is missing the MREI motif altogether (18).

Segment II of misato is a major contributor to the difference in length between it and the various tubulins (Fig. 4a). The most significant sequence similarities to this 54-amino acid region are from a motif found in the heavy chains of myosins from different organisms (Fig. 4d). A comparison of segment II with the vertebrate myosin heavy chains reveals that it is homologous to a myosin peptide in the hinge region linking the S2 and LMM domains. Segment II also contains heptad repeats which are characteristic of the myosin tail α-helical coiled-coils. However, unlike the myosins, the misato sequence contains three helix-breaking proline residues, which are positioned toward the N and C termini and at the middle of segment II (Fig. 4d). This suggests that the tertiary structure of segment II consists of two short antiparallel α-helices separated by a proline-containing turn, and that they interact to form a short coiled-coil which projects from the tubulin-like domain.

The N-terminal half of the misato protein thus has a curious mixture of peptide motifs held in common with, and in the same order as, the conventional tubulins. This analysis is based upon a database of all available tubulins, now totalling approximately 400 sequences (R.G.B, unpublished work). This database has been inspected to determine the fraction of sequences with a specific amino acid at each residue position, and the misato sequence has been aligned with the α-, β-, and γ-tubulins. Plots of the summed fractional values at each residue position against the corresponding misato residue reveal the regions with the highest sequence similarities. These are all in the N-terminal 280 amino acids of each of the three tubulin sequences, and they are separated either by misato inserts or by peptides in which the misato sequence shares only weak sequence similarity with the tubulin sequences. By contrast, the C-terminal sequence of misato exhibits a weak sequence similarity of approximately 15% to the conventional tubulins.

Putative misato Relatives from Other Organisms.

We have searched the complete S. cerevisiae genome for relatives of misato at the sequence level and have found five predicted proteins, parts of which yield significant similarity scores. Four of these proteins are tubulins, namely TUB1, TUB2, TUB3, and TUB4, whereas the fifth, designated YM8261.05, is a predicted open reading frame which has weak sequence similarity to members of the tubulin superfamily. Since the blast score for YM8261.05 is higher than for each of the tubulins, we aligned misato and YM8261.05 to determine the regions of greatest sequence similarity. This alignment of the 574-amino acid misato sequence and the 479-amino acid YM8261.05 sequence is shown in Fig. 5. The highest density of identical residues is in the N-terminal half, in segments I, II, and III. By contrast, the C-terminal halves of these two proteins are totally unrelated, and in addition, there are no other predicted proteins in this yeast that contain both tubulin- and myosin-like motifs. We conclude that misato and YM8261.05 are not true homologs, and that S. cerevisiae lacks a comparable protein that contains both tubulin and myosin motifs. The presence of the tubulin motifs in both misato and YM8261.05 may indicate that the two predicted proteins share some functional properties. It is also possible that the S. cerevisiae genome has been streamlined by deletion, one of the losses being that of a misato homolog. Finally, it is entirely possible that misato is an invention restricted to some multicellular lineages. If this is the case, then appropriate information may come from other metazoa.

Figure 5.

Amino acid sequence alignment of misato and the YM8261.05 predicted protein of S. cerevisiae. The extents of segments I, II, and III are the same as in Fig. 4, and only the identical residues are highlighted.

We have found that four human expressed sequence tags (ESTs) show excellent sequence similarity to misato at its N terminus (R17341R17341, H12093H12093, and Z43579Z43579). They are 63% similar and 46% identical at the amino acid level, and their alignments involve no gaps. The fourth human EST (N23629N23629) is 46% similar and 32% identical, again without gaps, and its sequence extends into segment II of misato. These ESTs provide a starting point for analyses of human genes that may belong to the misato family.

The misato transcript is a relatively rare one as assayed by Northern blotting of embryos, larvae, pupae, and adults, and our reverse transcriptase–PCR analysis reveals that the highest levels of this message are in the female germ line. So far, however, we have been unsuccessful in producing antibodies to the misato protein and analyzing its cellular distribution, and hence in carrying out a phenotypic analysis in greater depth. Further transgenic and comparative analysis of misato relatives in vertebrates, using the available human ESTs, for example, may yield alternative approaches to assaying its biological significance, since the overlap at the gene level between Drosophila and vertebrates is large.

Homologies and Multifunctionalities.

The major stumbling block to understanding the properties of the tubulins, and in particular their tertiary structures, has been the inability to crystallize them (18, 19). This is compounded by the findings that the folding of the nascent β-tubulin molecule requires a multisubunit chaperonin, TCP1, as well as additional cofactors. Hence, in the absence of crystallographic data, almost all comparisons between tubulin and tubulin-like molecules have depended on site-directed mutagenesis and crosslinking analyses. To date, however, the mutagenesis approach has not yielded potential peptides that interact with the guanine base, and the crosslinking results have yet to provide a coherent picture of the properties of β-tubulin (10). Without crystallographic structures, inferences from sequence comparisons, site-directed mutagenesis, and crosslinking studies are limited.

In the case of sequence comparisons, for example, misato has a higher overall sequence similarity to tubulins than does the bacterial FtsZ protein, which has been proposed to be a tubulin homolog on the grounds that it is a GTPase, forms microtubule-like structures, and has some sequence similarity to tubulins (20). However, some enzymes can have properties strikingly similar to those of cytoskeletal proteins, even though they are not thought to be related to tubulins in an evolutionary sense. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) has as much sequence similarity to tubulins as does FtsZ, and GAPDH can bind GTP, actin filaments, microtubules, and colchicine. Furthermore, GAPDH can polymerize to form cytoskeleton-like oligomers (21). These data highlight the difficulties of assigning homologies, sensu strictu, to one of the best-characterized cytoskeletal protein families (22). They also put into perspective the emerging multifunctionality of proteins that have so far been considered to have only a single conserved property.

The existence of the misato gene, with its intriguing combination of motifs and potential secondary structures, highlights the paucity of functional data on unconventional proteins that go undetected when relatives are sought by means of hybridization, PCR assays, or database searches. The finding that the S. cerevisiae genome lacks a true misato homolog demonstrates that this yeast does not automatically provide the entire foundation for cell division processes in the metazoa, and it has yet to be demonstrated that cell division processes in this yeast are representative of those in other fungi. In addition, it is already clear that S. cerevisiae is missing other genes known to be present in Drosophila and in Homo (S. Karlin, R. Maleszka, and G.L.G.M., unpublished work). While the genome project data from the model organisms and humans will reveal the highly conserved core of proteins in different metazoan phyla (23), it is probable that the crystallographic analysis of orphan proteins will be essential to understanding the origins of evolutionary novelties in different lineages.