INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The integrin family of cell surface receptors is important for adhesion between the extracellular matrix and the cytoskeleton (Yamada and Miyamoto, 1995; Dedhar and Hannigan, 1996). Binding of extracellular ligands and/or the formation of integrin complexes can trigger integrin association with a number of cellular components, including both cytoplasmic proteins and other membrane proteins. This association makes strong connections between the matrix and cytoskeleton, and also transmits signals that can regulate cellular functions such as proliferation, differentiation, and migration, often in cooperation with information from other cell surface receptors (Dedhar and Hannigan, 1996; Howe et al., 1998). Signals also can be transmitted in the opposite direction, as it is clear that events inside the cell can regulate the extracellular binding activities of the integrin heterodimers (Fernandez et al., 1998; Hughes and Pfaff, 1998).

The PS integrins of Drosophila are important for a variety of embryonic and postembryonic morphogenetic events (Stark et al., 1997; Brown et al., 2000). Like most vertebrate integrins, the PS1 and PS2 integrins are receptors for extracellular matrix components (Gotwals et al., 1994). To date, five different PS subunit genes have been identified: mew, encoding PS1; inflated, encoding PS2; scab, encoding PS3; and two other as yet poorly characterized PS3-like genes (Hynes and Zhao, 2000). At least the first three of these encode polypeptides that combine with PS subunits, encoded by the myospheroid (mys) gene, to generate PS1 (PS1PS), PS2 (PS2PS), or PS3 (PS3PS) integrins.

The primary sequences of integrins indicate a high degree of structural conservation. The PS integrins are no exception, and their sequences are as similar to vertebrate integrins as these are to one another (Gotwals et al., 1994). One structural constraint probably is associated with the interactions of  and  subunits that must be important in propagating conformational changes between the short cytoplasmic tails of the protein and the extracellular ligand binding domains (Humphries, 1996). In any case, integrin structure and function appear to be strongly conserved in evolution, and therefore basic information gleaned from studies of one integrin is likely to be applicable to others.

Much of what we know about integrin structure-function derives from studies involving site-directed mutagenesis followed by transfection into cultured cells for functional assays. Although very successful, this approach has limitations. For example, expression levels in transfected cells are often artificially high or unbalanced, and only phenotypes manifested by cultured cells can easily be examined. Also, it is impractical to sample more than a relatively small number of mutagenic changes. These limitations are ameliorated by the standard "forward" genetic approach of random mutagenesis followed by selection of mutants in situ, in developing animals. Mutations affecting required functions that are specific to certain cell types can be identified, and the animal tells the experimenter which mutations alter function, without regard to preconceived notions as to the functions of specific residues. Of course, "blind" genetic screens also have drawbacks; the important point is that this complementary approach has the potential to provide insights that would not readily be forthcoming from directed mutagenesis studies.

Genetic screens of this sort are not easily feasible with vertebrates, although  subunit mutations are revealed in human clinical syndromes, such as leukocyte adhesion deficiency (2) and Glanzmann thrombasthenia (3). Random mutagenesis screens can more easily be accomplished in cell culture (Baker et al., 1997), or with invertebrates such as Drosophila melanogaster and Caenorhabditis elegans. Drosophila provides a particularly good system for pursuing screens for integrin mutations, because the PS integrin genes are well characterized genetically and molecularly, and many different integrin-dependent functions have been defined during fly development (Brown et al., 2000).

Previously, a number of mutant alleles of myospheroid (PS) were generated and partially characterized (Wright, 1960, 1968; Costello and Thomas, 1981; Newman and Wright, 1981; Wieschaus et al., 1984; Leptin et al., 1989; Bunch et al., 1992; Zusman et al., 1993; Roote and Zusman, 1995). Some are null for protein function, others retain at least some protein function (hypomorphs), and one antimorphic allele (mys^XR04) has been described that, in complementation tests, is worse than a null allele. We have generated new strong alleles of myospheroid and have characterized these and existing alleles with respect to their molecular lesions and genetic properties. Most importantly, we have identified and analyzed additional alleles that behave genetically as antimorphs, suggesting that these proteins have dominant negative properties. The properties of these antimorphic mutants are compared with those of the human splicing variant 1B, which has molecular similarities and has been shown to have dominant negative properties when assayed in transfected tissue culture cells (Altruda et al., 1990; Balzac et al., 1994; Retta et al., 1998).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

General

All flies were grown on the food described in Condie and Brower (1989). Marker mutations not specifically referenced are described in Lindsley and Zimm (1992).

Screen for New myospheroid Mutants

Some of the new strong alleles described herein were by-products of a screen for mutations in mew, the gene encoding PS1, as described in Brower et al. (1995); both mew and myospheroid are on the X chromosome. Other alleles were from a similar screen set up specifically to identify myospheroid alleles (Figure 1). Briefly, males with the proximal FRT18A recombination site on the X chromosome (along with the cuticle markers yellow and forked^36a) were mutagenized with ethylmethanesulfonate (EMS) (Lewis and Bacher, 1968), and crossed to females with the same X chromosome FRT, as well as a heat shock-inducible FLPase on the second chromosome. Somatic recombination was induced by a heat shock of the F1 larvae to generate clones of cells homozygous for the mutagenized X chromosome. Animals with wing blisters from mutant clones (a known PS integrin phenotype) were selected, and those harboring myospheroid mutations were identified by complementation tests.

View larger version (58K): [in this window] [in a new window]   Figure 1.   (A) Mosaic screen to recover new strong mutations in the myospheroid gene. See text for details. (B) Wing blister resulting from clone of homozygous mysG1 mutant cells.

Phenotype and Genetic Assays

To score viability of various combinations of alleles, eggs were laid at the appropriate temperature, and vials were thinned to prevent overcrowding of larvae. Progeny were scored at least once a day, and if any animals from any single vial were scored, all subsequent progeny from that vial were counted, to guard against genotypic differences in developmental rates.

To quantitate the severity of dorsal herniation, eggs were collected from balanced stocks of the various mutants, and aged at 25^oC to allow wild-type animals to hatch. The mutant embryos were then dechorionated in bleach (Ashburner, 1989a) and scored blind under a dissecting microscope. For each genotype, 100-200 embryos were scored. For photography, embryos were mounted in Hoyers medium and photographed with phase contrast optics.

Mutant Sequencing and Reverse Transcription-Polymerase Chain Reaction (RT-PCR)

Heterozygous adults or hemizygous mutant embryos were homogenized and genomic DNA template was isolated using QIAGEN's QiaAmp tissue kit (Valencia, CA). PCR primers were designed to yield three overlapping PCR products, covering the myospheroid coding region (exons 2-7; Yee, 1993; Zusman et al., 1993). The first fragment began 139 base pairs before the initiating AUG and the third fragment continued 62 base pairs after the UAG stop codon. The introns were amplified but the largest (between exons 3 and 5) were not sequenced in their entirety. The resulting PCR fragments were purified using QIAGEN's QiaQuick PCR purification kit and sequenced directly by the University of Arizona Genomic Analysis and Technology Core (Tucson, AZ).

For RT-PCR of mys^XR04, 30 hemizygous mutant embryos were used to generate mRNA with QIAGEN's RNeasy mini kit. The complementary strand was generated using a myospheroid-specific primer and SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA). PCR primers were designed to then generate a 273-base pair fragment that included the intron between the sixth and seventh exons, as well as a potential downstream splice acceptor site. To examine the possible use of a potential downstream splice site, cDNAs from five developmental stages, embryo through adult, of wild-type flies were used for RT-PCR as described above. To maximize the likelihood of generating sufficient amounts of this novel potential fragment, amplified DNA was digested with HinfI at various times during the amplification procedure, taking advantage of a restriction site specific to the common, larger PCR fragment.

To search for the presence of a potential novel splice site in mys^P9, mRNA was isolated from 15 mys^P9/FM7i and y sn³ v FRT18A adults by using QIAGEN's RNeasy mini kit. The complementary strand was generated using QIAGEN's Omniscript reverse transcriptase and a specific primer located 62 base pairs after the UAG stop codon. Two sets of internal primers were used to generate a PCR fragment that spanned the splice site. The fragments from both mys^P9/FM7i and control DNA were examined on a 1.5% agarose 0.5× Tris borate-EDTA gel stained with ethidium bromide. The remainder of the PCR products was purified using QIAGEN's QIAquick PCR purification kit and sequenced.

Immunofluorescence and Protein Expression

For embryo staining, chromosomes with myospheroid alleles were balanced over an FM7c chromosome containing a P[w⁺, actin-lacZ] insert (Davis et al., 1996); simultaneous staining for -galactosidase and PS therefore allowed the unequivocal identification of hemizygous mutant embryos before overt myospheroid phenotypes became evident. Embryos were collected for 1.5 h at 25°C, and aged 23 h at 18°C, so that the majority of embryos were at stage 15-16, before muscles of the myospheroid embryos begin to detach. Embryos were then fixed and permeablized by using a protocol slightly modified from that of Tim Karr (Ashburner, 1989a). Embryos were incubated in three different antibody solutions, separated by washes. Blocking, washes, and antibody dilutions were in phosphate-buffered saline, pH 7, 1-10% fetal calf serum, 1% Triton-X 100. Antibody incubations were for 4 h at room temperature or overnight at 4^oC, with constant agitation. In order, the incubations were in rabbit anti--galactosidase (Harlan Sera-Lab, Crawley Down, Sussex, United Kingdom), the mouse monoclonal anti-myospheroid CF.6G11 (Brower et al., 1984), and finally a combination of goat anti-mouse fluorescein isothiocyanate (Jackson ImmunoResearch Labs, West Grove, PA) and goat anti-rabbit Texas Red (ICN Biomedicals, Cleveland, OH). In some experiments, an incubation in a mouse anti-Scr (Glicksman and Brower, 1988) was included as a control for antibody permeability. After the final wash, embryos (and the discs below) were mounted (in 30% 1 mM Tris pH 9.0, 70% glycerol, to which was added 2% n-propyl gallate to reduce bleaching) and examined in a standard Zeiss immunofluorescence or Leitz confocal microscope.

For staining wing imaginal discs, clones of homozygous myospheroid mutant cells were generated by somatic recombination in heterozygous animals. The procedure for clone induction was similar to that for the screen, except that known alleles were crossed to the FRT, FLPase-containing line, and heat shocks were applied daily to induce multiple clones per disk. (Clone induction with this protocol is typically so extensive that all of the animals of the correct genotype die as pupae or nonpupating larvae.) Wing discs were stained with the mouse monoclonal anti-myospheroid CF.6G11 and goat anti-mouse fluorescein isothiocyanate (Jackson ImmunoResearch Laboratories) as previously described (Brower et al., 1984).

RNAi Treatment of Cells

Production and use of double-stranded, interfering RNA (RNAi) was similar to that described in Clemens et al. (2000). To produce a DNA fragment containing 681 base pairs of 3'-untranslated sequence specific for the endogenous myospheroid gene (the 3'-untranslated sequence in the transgenes has been replaced with sequences from the Drosophila tubulin -1 gene; Bunch and Brower, 1992), genomic DNA from S2 cells was amplified by PCR with the primers mys3'd1 (CGGAAATCAGAAGGAACCC) and mys3'u2 (GTTAAGTATCCCAATTCTGAC). This fragment was then amplified using similar primers that also contained a 5' T7 RNA polymerase binding site (GAATTAATACGACTCACTATAGGGAGA). The PCR products were purified using a QIAquick PCR purification kit (QIAGEN) and used as templates to produce double-stranded RNA via the MEGASCRIPT T7 transcription kit (Ambion, Austin, TX). The double-stranded RNA was ethanol precipitated and resuspended in water, incubated at 65°C for 30 min followed by slow cooling to room temperature, and then stored at 20°C until use. Preliminary experiments showed similar effects when the final concentration of RNAi in the growth medium was 3-30 µg/ml. We used a final concentration of 15 µg of RNAi/ml medium in the experiments reported herein.

For treatment of cells with RNAi, 1 × 10⁶ cells were washed and suspended in 0.66 ml of HyQ-CCM3 serum-free medium (Hyclone Laboratories, Logan, UT) per well of a six-well culture dish. RNAi (30 µg) was added and the cells were incubated for 30 min at room temperature followed by the addition of 1.34 ml of M3 + 12.5% fetal calf serum + 0.2 µM methotrexate. Experiments were done 5 d or more after initial exposure to RNAi. Every 5 d cells were diluted in fresh medium containing 15 µg of RNAi/ml.

The effectiveness of RNAi treatment was tested on cells transfected with a gene expressing the PS2 subunit from a heat shock promoter (Bunch and Brower, 1992), without a transgene expressing the PS subunit. Therefore, expression of PS2 integrin is dependent solely on the endogenous PS gene. If these cells are heat shocked to induce high expression of PS2, RNAi treatment results in an 80-90% reduction in integrin expression levels, as measured by flow cytometry; thus, RNAi treatment does not eliminate steady-state PS integrin expression. However, under the conditions of our cell spreading and expression experiments, where cells are cleared of preexisting integrins during the heat shock by the addition of dispase/collagenase, and then analyzed after 3-4 h of recovery, RNAi treatment results in a virtually complete inhibition of integrin expression (our unpublished data). It should also be noted that under the experimental conditions used in this article, there are also high levels of competing PS subunits expressed from the heat shock-driven transgenes.

Cell Spreading Assays

Cell culture techniques and methods for transfection of cells have been previously described, as have Schneider's line 2 cells transfected with integrin transgenes under the regulation of the heat shock protein 70 promoter (Bunch and Brower, 1992; Zavortink et al., 1993). Some of the current transformed cell lines were transfected using the CellFECTIN technique (Invitrogen). PCR was used to generate mutations in the pHSPS plasmid that correspond to the protein encoded by the mys^G1 mutation and a cytoplasmic truncation that is missing exon 7 (mys⁷). In all experiments, the transgenes corresponded to the "c" isoform of PS2 and the "4A" isoform of PS (Graner et al., 1998).

The ligand used in cell spreading assays is RBB-Tig. RBB-Tig is a bacterial fusion protein that contains 53 amino acids of the Drosophila extracellular matrix protein tiggrin (residues 1964-2016, including the RGD sequence and 25 amino acids upstream and downstream), fused to a histidine tag, from the pTrcHisB vector (Xpress SystemTM; Invitrogen). This fusion protein is as active in promoting cell spreading as the previously described tiggrin fusion protein and tiggrin itself (Fogerty et al., 1994). Fusion protein was purified by affinity chromatography on Ni-NTA agarose (QIAexpress; QIAGEN). Ninety-six well tissue culture plates or slides were incubated with 500 ng/ml RBB-Tig diluted in phosphate-buffered saline (PBS) for either 1 h at room temperature or overnight at 4°C, blocked with 10% dried milk in PBS for 1 h at room temperature, and washed three times with PBS.

Standard cell spreading assays involved pretreatment with dispase/collagenase before integrin expression and assays for cell spreading in serum-free medium. These were done as previously described (Bunch and Brower, 1992; Zavortink et al., 1993) with one change. The dispase/collagenase treatment to remove existing integrins and extracellular matrix was done at 37°C at the same time as the heat shock, which induces expression of the integrin transgenes. Staining cells for surface integrins demonstrated that this treatment quantitatively removes the existing heterodimers from the surface of the cells. For RNAi-treated cells, RNAi was included in the cell-spreading medium (M3 + 2 mg/ml bovine serum albumin). The number of spread cells was determined by microscopy 3-4 h after heat shock and plating. Results of cell spreading assays are expressed as the averages of three experiments with SEs. For the "normal growth conditions" assay, cells were suspended and heat shocked without protease treatment, and replated into regular growth medium with serum. Spreading was assayed overnight on plates without RBB-Tig, or 3-4 h postinduction on plates coated with RBB-Tig.

To examine integrin expression levels, cells were incubated with biotinylated anti-PS2 integrin monoclonal antibodies (CF.2C7) and R-phycoerythrin-streptavidin (Molecular Probes, Eugene, OR) in M3 medium + 10% fetal calf serum on ice, followed by dilution with 1 ml of 2% formaldehyde in PBS. Cells were analyzed by flow cytometry at the Arizon Research Labs-Biotechnology Cell Sorting Facility (University of Arizona).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

mys^XR04 Is a Weak Dominant Negative In Situ

A number of strong mutant alleles of myospheroid have been isolated previously, and some of these have been shown to make little or no PS protein. By using these alleles, the myospheroid null embryonic lethal phenotype has been described in detail, along with some genetic interactions with other integrin mutants (Wright, 1960, Wieschaus et al., 1984; Newman and Wright, 1981; Wieschaus and Noell, 1986; Leptin et al., 1989; Bunch et al., 1992; Zusman et al., 1993; Roote and Zusman, 1995). In the course of these studies, it was noticed that one EMS-generated allele, mys^XR04, displays antimorphic (stronger than the null phenotype) genetic properties (Wilcox, 1990; Bunch et al., 1992; Brabant and Brower, 1993). Specifically, a double heterozygote of mys^XR04 and a null allele for the PS2-encoding gene (inflated) can be synthetically lethal. It is also true that the embryonic lethal phenotype of mys^XR04 hemizygous embryos (from heterozygous mothers) can be different from zygotic null embryos; most obviously, they often have a "tail up" phenotype, and the dorsal herniation characteristic of PS loss is often more severe. This phenotype has also been noted in embryos that lack both maternal and zygotic PS function (Wieschaus and Noell, 1986; Leptin et al., 1989; Roote and Zusman, 1995), consistent with the hypothesis that mys^XR04 could exert negative effects on the maternally contributed wild-type PS.

In addition to the above-mentioned phenotypes, we have characterized another indicator of the antimorphic properties of mys^XR04. Following on the findings of Bunch et al. (1992), we have analyzed the complementation behavior of dozens of weak (hypomorphic) alleles of myospheroid, and find that most show significant viability alone or in combination with null alleles, especially at low temperatures, but that these typically are completely lethal when trans-heterozygous with mys^XR04 (unpublished data). Table 1 gives complementation data for mys^b7, one of the weakest hypomorphic alleles.

Additional Myospheroid Alleles

We performed genetic screens that isolated additional mutant alleles of myospheroid. Male flies were fed the mutagen EMS and crossed to females to generate F1 animals. During larval development, somatic recombination was induced on the myospheroid-containing X chromosomes, to generate small clones of cells homozygous for the mutagenized X chromosome (Figure 1). The adult F1 flies were then screened for wing blisters, known to be caused by strong mutations in myospheroid. Potential mutations in PS were confirmed by complementation testing with other alleles of myospheroid, and, as described below, by sequencing. (See MATERIALS AND METHODS and Brower et al., 1995, for screen details.)

Although we have not characterized the full range of embryonic lethal phenotypes for all of the new alleles generated from this screen, they all show the dorsal herniation typical of null alleles of myospheroid. We have examined some alleles in more detail, including mys^G1, mys^G4, mys^G12, mys^M2, and mys^P9. No obvious deviations from the previously described null phenotype (Wright, 1960) have been noted, except for mys^G1 and mys^P9. These alleles can display the tail up and extreme herniation phenotype (Figure 2), similar to that described for mys^XR04. Because the phenotypes are variable, we sought to quantify the penetrance of the severe herniation. As shown in Figure 3, embryos of mys^G1 and mys^P9 were both significantly more severe than the null alleles.

View larger version (90K): [in this window] [in a new window]   Figure 2.   Embryo cuticle preparations from myospheroid mutants. The embryos show small, medium, and large dorsal holes (arrows) from the null mutant mysXG43 (A and B) and the antimorphic mysP9 (C) mutant. Frequencies of these phenotypes are quantitated in Figure 3. Note the anterior dorsal movement of the posterior spiracles (arrowheads) in the mysP9 embryo (C) relative to the wild type (D). Bar, ~0.1 mm.

View larger version (22K): [in this window] [in a new window]   Figure 3.   Dorsal herniation phenotype of myospheroid mutants. Hemizygous mutant embryos from the myospheroid alleles indicated, and the double mutant mysXB87 ifB4, were dechorionated and scored for the size of the dorsal hole. Larger holes are more frequent in the antimorphic and double mutants.

Embryos doubly mutant for null alleles of myospheroid and inflated (PS2) also were much more likely to exhibit a large dorsal hole (Figure 3). Because there is no maternally contributed inflated gene product (Bogaert et al., 1987; Roote and Zusman, 1995), the double mutant will be similar to a complete zygotic and partial maternal myospheroid mutant. The similarity of the double mutants and the myospheroid antimorphs supports the notion that the antimorphs lead to a reduction in the maternal wild-type myospheroid function.

All of the strong myospheroid alleles were tested for suggestions of dominant negative behavior in crosses with weak alleles (Table 1). Again, mys^G1 and mys^P9 consistently behaved similarly to the antimorphic mys^XR04 in these assays, whereas other alleles behaved similarly to known null mutations. Like mys^XR04, mys^G1 and mys^P9 also were synthetically lethal when in a double heterozygote with an inflated (PS2) null allele.

Dominant Negative Alleles Have Similar Molecular Lesions

We sequenced the coding regions of the genomic DNA for all of the new myospheroid alleles, as well as older alleles that displayed strong, apparently null phenotypes. As shown in Table 2, many of the alleles contain premature stop codons, deletions, frameshifts, or splicing mutations that would be expected to lead to a complete lack of functional PS protein, and for at least some of the alleles, this expectation has been confirmed by immunofluorescence (see below) or Western analysis (Leptin et al., 1989; Bunch et al., 1992). Only two alleles, mys^G4 and mys^G12, are characterized by missense mutations. One of these, mys^G4, alters the second serine of the essential DXSXS MIDAS motif, whereas mys^G12 changes the aspartate in a DYPS(hydrophobic) motif that is highly conserved in  subunits. One other myospheroid allele that is 100% lethal, mys^XN101, also results from a missense mutation (C627>S) in the extracellular stalk domain; however, the lethal phenotype of this allele indicates that it retains some wild-type function (our unpublished data).

The three alleles showing weak dominant negative effects all have molecular lesions midway through the cytoplasmic domain of the PS subunit. Both the mys^XR04 and mys^G1 chromosomes contain identical G>A mutations in the splice acceptor site before the seventh exon (Figure 4). Because these two mutations were generated independently in different genetic backgrounds (as verified by molecular polymorphisms in the two mutant strains), we can be confident that the unusual antimorphic phenotype seen in both lines results from this specific molecular lesion. The mys^P9 chromosome contains a nonsense mutation in exon 6, in the third codon upstream of the above-mentioned splice site.

View larger version (15K): [in this window] [in a new window]   Figure 4.   (A) Nucleotide sequence around the boundaries of the cytoplasmic intron of myospheroid. Exon sequence in caps, intron in lowercase. The G>A mutation in mysXR04 (and mysG1) is indicated; this results in the seventh exon beginning with the underlined G. The potential downstream splice acceptor site is indicated by double underlining. To search for mRNAs that might use this site, this entire region was amplified by RT-PCR. Amplification of products using the conventional splice site was inhibited by cutting at the HinfI site indicated in bold (see text). (B) Predicted amino acid sequence of the cytoplasmic domains of wild-type PS and that from mysXR04 and mysP9. The frameshift induced by the aberrant splicing in mysXR04 replaces the membrane-distal 21 residues with 25 essentially random amino acids (underlined). The first cytosine of exon 7 is a thymidine in mysG1; this polymorphism is silent in the parental strain, but changes to first residue encoded by the mutant exon 7 from alanine to valine.

Analysis of Dominant Negative Mutant mRNAs

The three cytoplasmic mutants all have mutations that have the potential to alter mRNA splicing, and so we characterized the transcripts that result from these lesions. The mutated guanosine in mys^XR04 and mys^G1 is required for proper mRNA splicing, and from examination of nearby sequences, three splicing outcomes seemed possible for the mutant mRNA: 1) The splice would shift one nucleotide to the next guanosine, creating a frameshift. 2) Splicing would occur at a nearby downstream sequence (24 nucleotides away) that resembles a consensus acceptor sequence (Figure 4); this would create an in-frame deletion of eight residues, centered on the first of the two NPXY motifs present in integrin  subunits (see DISCUSSION). 3) Splicing would not occur nearby, presumably resulting in a string of missense residues. To distinguish between these possibilities, we performed RT-PCR and sequence analyses of mRNA from mys^XR04 mutant embryos. PCR primers were designed to generate a 273-base pair fragment (in wild type) that includes the splice junction and the potential downstream splice acceptor site. Only a single gel fragment was visualized for mys^XR04. The sequence of the fragment indicated a splicing shift deleting one nucleotide of the mys^XR04 cDNA (corresponding to possibility 1 above). The frameshift introduced by the aberrant splice causes exon 7 to encode 25 novel residues (Figure 4). In mys^G1, a nucleotide polymorphism present in the parental strain, and which is silent in the normal reading frame, changes the alanine at the start of the new sequence to valine. Despite this minor difference, we will refer to the frameshifted PS subunits generically as the PS^XR04-G1 protein. This mutant protein is missing both NPXY motifs.

At first sight, the mys^P9 mutation appears to be a straightforward stop codon, leading to a protein truncated three residues before the mys^XR04-G1 frameshift. However, closer examination of the nucleotide sequence reveals the possibility of a cryptic splice donor sequence just upstream, and the mys^P9 mutation improves the potential intron consensus sequence at this site. To see whether mys^P9 might lead to significant use of this new potential splice site, we did RT-PCR analysis of mRNA from heterozygous mys^P9 animals. We found no trace of the smaller (263 vs. 273 base pairs) aberrantly spliced mRNA on agarose gels or upon direct sequencing of the PCR products. There also was no indication that expression level of the mutant mRNA was reduced. Thus, the mys^P9 mutation apparently leads to a straightforward truncation (Figure 4).

As mentioned above, examination of the nucleotide sequence of the wild-type gene indicated the potential for mRNA splicing that would precisely delete eight amino acids, centered on the first NPXY motif. We were curious to see whether the potential downstream splice site might be used in wild-type animals, at some restricted time or place in development. To examine this possibility, cDNA from five developmental stages, embryo through adult, of wild-type animals was used for RT-PCR as mentioned above. We anticipated that the potential alternatively spliced cDNA might be relatively rare, and PCR amplification can be a competitive process. To maximize the likelihood of generating sufficient amounts of the fragment in question, we digested the amplified DNA at various times during the amplification procedure, taking advantage of a HinfI restriction site present only in the common, larger PCR fragment (Figure 4). Several variations of this scheme were tried, but a new fragment size was never seen.

Expression of Mutant Proteins

We examined a number of the mutant alleles by immunofluorescence for expression of surface integrin. As expected, the nonsense mys^M2 mutant shows no surface PS integrin in late-stage embryos (our unpublished data); loss of PS integrins has previously been reported for the deletion and splicing mutants mys^XG43 and mys^XB87 (Leptin et al., 1989; Bunch et al., 1992). Both missense alleles, mys^G4 and mys^G12, express surface integrin at levels comparable to wild-type embryos (Figure 5), and show the typical accumulation of protein at embryonic muscle attachment sites. (The embryos shown in Figure 5 are fixed before the contractions that lead to the muscle detachment characteristic of the mutant phenotype.)

View larger version (99K): [in this window] [in a new window]   Figure 5.   Confocal immunofluorescence sections of the hypoderm and underlying muscles of stage 15-16 embryos. The animals are stained with an antibody against the PS subunit, which is seen to be concentrated in a metameric array of dots corresponding to muscle attachment sites. The wild type (wt) is a heterozygote for wild type over the nonsense allele mysM2; homozygotes for mysM2 show no staining. Bar, ~50 µm.

The frameshifted PS^XR04-G1 protein also appears to be expressed in embryos, and localized at muscle attachments. To look more thoroughly for possible expression reductions in these mutants, we examined surface integrin expression on the epithelium of larval imaginal discs containing clones of homozygous mys^G1 cells in a heterozygous background. These clones were generated similarly to those used in the mosaic screen for mutant alleles, except that the larvae were heat shocked more extensively (daily) to induce larger numbers of clones. As shown in Figure 6, control animals with the mys^M2 nonsense allele typically (13 of 15 wing discs) showed multiple patches of cells without detectable integrin expression. However, no patches of reduced expression were detected in mys^G1 discs grown and heat shocked at the same time (51 wing discs examined). Thus, the weakly dominant negative allele mys^G1 does not lead to large-scale reduction in integrin expression in the disk epithelium.

View larger version (130K): [in this window] [in a new window]   Figure 6.   Third instar wing imaginal discs stained with antibody against PS. These animals are heterozygous for myospheroid mutations, and have been treated to induce recombinant clones of homozygous mutant cells. (A) Heterozygote for the nonsense mutation mysM2; the clonal patches show no surface expression of PS integrins (asterisks). (B) Disk from mysG1 heterozygote cross, which shows no patches of reduced integrin expression. Bar, ~100 µm.

Dominant Negative Mutants Support Cell Spreading in Culture

To test the ability of mutant proteins to function in a simple cell spreading assay, we transformed Drosophila S2 cells with cDNAs for PS2 (the c splice variant) and PS (the 4A splice variant) proteins corresponding to the mys^G1 mutant and a mutant similar to mys^P9, but truncated precisely at the exon 6-7 splice boundary (designated mys⁷). This latter protein is three residues longer than that encoded by mys^P9. All transformed genes were under the control of a heat shock promoter, and for each the endogenous 3'-untranslated sequences were replaced with tubulin 1 3' sequence (Bunch and Brower, 1992; Graner et al., 1998).

S2 cells produce small but significant amounts of endogenous PS, and cells transformed with PS2 alone show a significant ability to spread on RBB-Tig, an RGD-containing fragment of the PS2 ligand tiggrin (Fogerty et al., 1994). This spreading is reduced by ~90% when the cells are treated with double-stranded RNAi (Carthew, 2001) against the 3'-untranslated region of the endogenous PS transcript (Figure 7). When cells are transformed with both PS2 and PS genes, and subjected to the protease clearing and heat shock induction protocol, RNAi has little effect, indicating that virtually all of the spreading is due to expression from the transformed genes. This RNAi control was used for all of the results reported herein.

View larger version (14K): [in this window] [in a new window]   Figure 7.   Cell spreading by mutant  subunits. S2 cells were transformed with PS2 and various forms of PS, and assayed for spreading on a fragment of the PS2 ligand tiggrin. For these experiments, cells were pretreated with protease to remove surface proteins and ECM before integrin induction. RNAi was added to inhibit expression of endogenous PS (filled bars). Untransformed S2 cells do not spread. Cells transformed with PS2 only can spread using the low levels of endogenous PS, but this is reduced by RNAi treatment. Spreading of cells transformed with wild-type or mutant PS is not significantly reduced by RNAi. The difference between wild-type and the two PS mutants reflects differences in expression level.

Both the mys^G1 and mys⁷ proteins were expressed well on S2 cells in combination with PS2. We tested the ability of the mutant proteins to mediate cell spreading under two different assay conditions. In one, cells are treated with protease to clear surface proteins and ECM, heat shocked to induce integrins, and allowed to spread on RBB-Tig. In this assay, both mys^G1- and mys⁷-containing integrins promote spreading that is comparable to that seen with wild-type PS (Figure 7). Although slight cell line variations were observed, these variations could be ascribed to minor differences in the expression levels of the integrins (our unpublished data).

Although the mutant integrins can mediate spreading in the above-mentioned assay, we noticed that these cells spread very poorly during regular culture, relative to cells expressing wild-type  subunits. To quantitate this, we performed a second assay similar to regular growth conditions, in which the cells were not pretreated with protease but were suspended, heat shocked to induce high-level integrin expression, and allowed to spread on plates overnight in regular growth medium containing serum. (Although not shown, qualitatively similar results were obtained if uncleared cells were spread on RBB-Tig for 3-4 h.) Under these conditions, the mutant integrins promote cell spreading very poorly, relative to wild type (Figure 8), despite the fact that these cells are expressing more integrin than is seen after protease clearing and induction (our unpublished data). Interestingly, in this assay the activity of the mutant PS subunits can be largely restored if combined with PS2 subunits containing a mutation of the conserved cytoplasmic GFFXR motif (GFFNR>GFANA) that promotes integrin activation (O'Toole et al., 1994; Hughes et al., 1996).

View larger version (17K): [in this window] [in a new window]   Figure 8.   Cell spreading in different assay conditions. Integrin induction in transformed S2 cells followed protease treatment as before (open bars) or was in the absence of protease (filled bars). In the latter case, the spreading assay was in normal growth medium, with serum, on uncoated tissue culture plates. activ-G1 and activ-truncation indicate cells that also express a mutant PS2 that promotes integrin activation. The mutant  subunits support spreading poorly except in conditions (cleared cells or activating PS2 subunits) that appear to promote artificially high activation. This is true in spite of the fact that the normal growth G1 and truncated cells express integrin at higher levels than any of the cleared or activating PS2 lines.

The morphology of spread cells on RBB-Tig may also be influenced by expression of the mys^G1- and mys⁷-containing integrins. Although a variety of morphologies are typical even in wild type, the mutants tend to display more regions of extreme peripheral thinning (Figure 9). These are characteristics also observed in S2 cells containing mutants that affect the ability of the cell to regulate integrin activity, such as the activating PS2 mutants (our unpublished data).

View larger version (93K): [in this window] [in a new window]   Figure 9.   Drosophila S2 cells that were cleared, induced to express wild-type or mysG1 mutant PS in association with PS2, and allowed to spread on high concentrations of a recombinant fragment of tiggrin. The mutant cells have a tendency to display more thin, clear areas at the periphery (asterisks).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The most interesting of the mutant alleles described herein are the three cytoplasmic mutants that display dominant negative properties in flies, and we will refer to them collectively as mys^DN alleles. Although they are viable as heterozygotes, these animals have reduced integrin function relative to animals heterozygous for null alleles. This is perhaps best illustrated by the synthetic lethality of the mys^DN inflated^null double heterozygotes; reducing PS2 expression by 50% is not sufficient to generate phenotypes in a myospheroid null heterozygote. Dominant negative qualities also are indicated by crosses with weak alleles; although mys^b7/mys^null supports viability very well at high temperatures, the mys^b7/mys^DN combination is strongly lethal. Finally, the embryonic lethal phenotypes of the mutants are consistent with the idea that these alleles compromise the functioning of the maternally contributed wild-type PS.

Similarity to Mammalian  Subunit Variants

The position of the cytoplasmic intron is conserved in most integrin  subunit genes (Schmitt and Brower, 2001), and just as mys^XR04 and mys^G1 create a frameshift at this site (caused by shifting the splice site one nucleotide), similar naturally occurring cytoplasmic splice variations have been described for human 1 and 3 (van Kuppevelt et al., 1989; Altruda et al., 1990; Languino and Ruoslahti, 1992; Kumar et al., 1997; Svineng et al., 1998; for review, see Melker and Sonnenberg, 1999). The 1C splice variant has been shown to result in loss of some cytoplasmic associations and signaling properties, and to affect cell proliferation (Languino and Ruoslahti, 1992; Fornaro et al., 1995, 1998, 2000; Meredith et al., 1995; Pfaff et al., 1998). Human 3B and 3C splice variants also generally show a decrease in cellular functions such as focal adhesion localization, focal adhesion kinase phosphorylation, and cell adhesion (Akiyama et al., 1994; LaFlamme et al. 1994; Kumar et al., 1997; Pfaff et al., 1998). Extensive analyses of one human isoform, 1B, have uncovered dominant negative properties similar to those we see, and we will compare human 1B with our Drosophila mutants in greater detail.

Like mys^XR04 and mys^G1, human 1B alters the protein sequence (relative to the common 1A isoform) distally from the splice site, but in this case as a result of a failure to splice and subsequent translation into the intron (Altruda et al., 1990). The 1B isoform is expressed primarily in skin and liver, where it constitutes a minority of the 1 protein (Balzac et al., 1993). To date, no function of 1B has been demonstrated in situ, but there have been a number of studies of the properties of the isoform in cultured cells (Balzac et al., 1993, 1994; Cali et al., 1998; Retta et al., 1998; Armulik et al., 2000). Overall, this work indicates that 1B is deficient in both "outside in" and "inside out" integrin signaling. For example, to mediate adhesion, the 1B-containing integrins typically require activation using nonphysiological treatments such as incubation in Mn²⁺ ions. And even upon adhesion or antibody cross-linking, 1B integrins are not efficient stimulators of cytoplasmic proteins such as focal adhesion kinase and paxillin.

Perhaps the most striking property of 1B is its ability to act in a dominant negative manner with respect to other cellular integrins. When transfected into cells, 1B has been reported to inhibit processes mediated by other 1 or 3 integrins, including cell spreading, motility, matrix assembly, and stimulation of specific cytoplasmic proteins (Balzac et al., 1994; Cali et al., 1998; Retta et al., 1998; see Armulik et al., 2000, for a contradictory view of the effects on v3 functions). Retta et al. (1998) further report that these dominant negative properties are dependent on the "intron-encoded" residues; proteins that are simply truncated at the splice site do not inhibit the functions of endogenous integrins.

All of the experiments showing dominant negative effects of 1B rely on cell transfection, and the mys^DN mutants are the first demonstration that replacement of the C-terminal portion of the  cytoplasmic domain can have dominant negative effects in situ. And because the Drosophila mutants are generated in otherwise normal chromosomes, and their expression is controlled by the wild-type regulatory machinery, it is by analogy reasonable to think that physiological levels of 1B expression could have significant repercussions for human cells in situ.

Why Dominant Negative?

How do the mys^DN mutants (and by extension, human 1B) exert their unusual properties? We can rule out some possibilities from these studies. It has recently been noted that 1B contains a new double lysine motif that can reduce surface expression through trapping of the protein in the endoplasmic reticulum (Kee et al., 2000). The PS^DN proteins contain no similar double lysine, and our expression studies in situ indicate that the mutant proteins reach the cell surface without great difficulty.

Another potential explanation is that the PS^DN proteins just get in the way, by adding a pool of nonfunctional integrins that compete on the cell surface with the wild-type proteins. The simplest version of this scenario does not appear to be true, because two missense alleles (mys^G4 and mys^G12) that make stable but probably nonfunctional (for ligand binding) protein do not show any dominant negative genetic properties. Competition models might be workable if one proposes that the cytoplasmic mutants also lead to greatly enhanced protein stability, so that the mutant proteins would eventually be present at much higher numbers than their wild-type competitors. However, when we follow the turnover of surface integrin after a single heat pulse of our transformed S2 cells, we see no indication that the PS^DN proteins are unusually stable (our unpublished data). And, in the analogous 1B system, expression of the variant can exert dominant negative effects on 1A functions when each is expressed at equivalent levels (Balzac et al., 1994). Also, 1B-expressing cells can display phenotypes such as reduced v expression (presumably in v3 heterodimers) in focal adhesions, without 1B actually displacing the v integrins directly (Retta et al., 1998).

Our genetic results are consistent with the previous proposal that the dominant negative mutants create a cellular signal that leads to inactivation of other surface integrins (Retta et al., 1998). This idea also is consistent with data on the activity of the dominant negative mutants themselves. In general, disruptions of  subunit NPXY motifs leads to reduced integrin activation and cytoskeletal associations, but 1B, for example, can mediate cell adhesion if activated artificially with Mn²⁺ (Retta et al., 1998; Armulik et al., 2000). In our S2 cell assays, we see evidence that the PS^DN proteins are only poorly able to mediate cell spreading during normal growth, but that a latent activity can be uncovered under conditions that are known (mutant  subunits) or suspected (protease pretreatment) to activate integrins. This also argues that a primary affect of the PS^DN proteins is related to the cells' ability to regulate integrin function.

The Drosophila results further allow some insights as to the structural requirements for the dominant negative properties of these cytoplasmic variants. The failure to detect inhibitory effects of subunits truncated at the splice site led Retta et al. (1998) to surmise that some contribution of the new residues of 1B is important. There is no sequence similarity between the C-terminal residues of the PS^XR04-G1 proteins and those of 1B, making it unlikely that a sequence-specific motif is involved. Moreover, we find that a truncation, mys^P9, has dominant negative properties in situ similar to the frameshift mutants, in contrast to the results of Retta et al. (1998) in cell culture. Thus, it appears that in general there is no absolute requirement for a particular sequence, or any sequence after the cytoplasmic splice site, to induce dominant negative effects. It is possible that 1B and PS are truly different in this respect, but it seems equally likely that if the human proteins were assayed in situ, where the entire range of integrin functions is required, dominant negative effects might emerge for the truncated 1 proteins.

Like 1B, the PS^DN proteins are missing both cytoplasmic NPXY motifs, which, along with neighboring residues, have been shown to be important for activation of extracellular ligand binding, intracellular signal transduction events, or processes such as adhesion or migration mediated by vertebrate integrins (Reszka et al., 1992; Filardo et al., 1995; O'Toole et al., 1995; Ylänne et al., 1995; Baker et al., 1997; Blystone et al., 1997; Chang et al., 1997; Tahiliani et al., 1997; Vignoud et al., 1997; Loo et al., 1998; Pfaff et al., 1998; Romzek et al., 1998; Sakai et al., 1998; Schaffner-Reckinger et al., 1998; Buttery et al., 1999; Kaapa et al., 1999; Mastrangelo et al., 1999; Sakai et al., 1999; Levy et al., 2000; Stroeken et al., 2000; Wennerberg et al., 2000; Boettiger et al., 2001; Ginsberg et al., 2001). Numerous studies have demonstrated requirements for specific residues in these motifs, although dominant negative effects have not generally been tested or noted. However, one anecdotal piece of evidence suggests that these motifs may be critical to the unusual genetics of our mutants and 1B. Grinblat et al. (1994) mutagenized various residues of the cytoplasmic tail of PS, and asked whether these mutants when transformed into flies could rescue myospheroid null phenotypes. (The mutant constructs contained endogenous myospheroid regulatory elements, and were generally expressed at levels equal to or below that of the endogenous myospheroid gene.) Like some others, they found that changing the tyrosines of the NPXY motifs to phenylalanines had relatively little effect. This alteration is expected to prevent potential phosphorylation of the motifs, but not alter their ability to make a predicted -turn structure (Haas and Plow, 1997; Ginsberg et al., 2001; Ulmer et al., 2001). However, when Grinblat et al. (1994) changed these tyrosines to alanines, they were unable to recover transformants that expressed the mutant proteins. Of course, this failure could have been due to some unknown technical glitch. However considering their success with numerous other constructs, Grinblat et al. (1994) proposed that the pair of Y>A mutations creates a toxic protein. Our results support this proposal, and combined, the studies suggest that the disruption of NPXY motifs is the critical requirement for creating these dominant negative  subunits. Finally, it appears that most of the proximal cytoplasmic tail must be intact for strong dominant negative properties. For example, Grinblat et al. (1994) were able to generate transformants with a cytoplasmic truncation that is only a few residues shorter than our dominant negative mys^P9 truncation.

Scarcity of Essential Residues of PS

Two missense mutations of myospheroid display the same strong phenotype seen for alleles that produce no functional PS. Both of the "null" missense mutations change oxygenated residues in the globular head domain of PS. mys^G4 (S196>F) changes the second serine of the conserved MIDAS sequence DXSXS, which is involved in the formation of a cation-binding pocket (Xiong et al., 2001), and this serine has been shown to be required for function or stability for vertebrate 2 and 3 (Bajt and Loftus, 1994; Bajt et al., 1995; Hogg et al., 1999). The mys^G12 chromosome contains two missense mutations, R5>K and D356>N. The latter change most likely is responsible for the mutant phenotype, because this residue is acidic in all sequenced  subunits, and has been shown to be essential for function in 1 (Puzon-McLaughlin and Takada, 1996). The arginine residue at position five is early in the PS signal sequence. The change to lysine is conservative, and analysis by the PSORT algorithm (Nakai and Kanehisa, 1992) predicts little effect on the ability of the altered domain to function as a signal sequence. Moreover, we find that mys^G12 (as well as mys^G4) subunits are expressed well at the cell surface in situ, arguing strongly that the signal sequence alteration is not responsible for the mutant phenotype.

Not surprisingly, the most common class of genetically strong mutations is nonsense mutants, represented by five alleles. Of the 14 null or antimorphic alleles, only two result from missense mutations. One other missense allele, mys^XN101, is known to be 100% lethal and have a dorsal herniation phenotype, but other studies suggest that it has some residual function (Wieschaus et al., 1984; Bunch et al., 1992; our unpublished data). Three of the strong mutations are in splice sites, and four alleles are associated with deletions or other rearrangements, mutagenic events that are expected to be relatively rare after EMS mutagenesis (Ashburner, 1989b). The most common mutation created by EMS is a G-to-A transition. If one assesses the potential results only of G-to-A transitions in the myospheroid gene, there are approximately 10 times as many sites that would create missense mutations as the sum of sites that would lead to nonsense mutations and changes that would eliminate correct splicing. The data are limited, but the results to date, in which the number of strong missense alleles is equaled by splice site mutations, suggest that relatively few of the 846 residues of PS are absolutely essential for function. On the other hand, experiments to be described elsewhere indicate that myospheroid missense mutations that alter PS function, but do not eliminate it, are relatively easy to generate, and just as whole animal genetics provides a sensitive assay for dominant negative properties, forward genetic screens should be useful for finer dissection of other integrin structure-function relationships.