Courtship and Visual Defects of cacophony Mutants Reveal Functional Complexity of a Calcium-Channel 1 Subunit in Drosophila

Corresponding author: Jeffrey C. Hall, Department of Biology, MS-008, Brandeis University, 451 South Street, Waltham, MA 02254-9110, hall{at}binah.cc.brandeis.edu (E-mail).

Communicating editor: C.-I WU

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We show by molecular analysis of behavioral and physiological mutants that the Drosophila Dmca1A calcium-channel 1 subunit is encoded by the cacophony (cac) gene and that nightblind-A and lethal(1)L13 mutations are allelic to cac with respect to an expanded array of behavioral and physiological phenotypes associated with this gene. The cac^S mutant, which exhibits defects in the patterning of courtship lovesong and a newly revealed but subtle abnormality in visual physiology, is mutated such that a highly conserved phenylalanine (in one of the quasi-homologous intrapolypeptide regions called IIIS6) is replaced by isoleucine. The cac^H18 mutant exhibits defects in visual physiology (including complete unresponsiveness to light in certain genetic combinations) and visually mediated behaviors; this mutant (originally nbA^H18) has a stop codon in an alternative exon (within the cac ORF), which is differentially expressed in the eye. Analysis of the various courtship and visual phenotypes associated with this array of cac mutants demonstrates that Dmca1A calcium channels mediate multiple, separable biological functions; these correlate in part with transcript diversity generated via alternative splicing.

CALCIUM channels are involved in cellular functions such as regulation of membrane excitability, neurotransmission, and generation of rhythmic or bursting potentials (HILLE 1992 ). Physiologically diverse channel subtypes are generated by multiple 1, 2, ß and subunit genes, alternative splicing of transcripts from a given gene, and combinatorial assembly (reviewed by HOFMANN et al. 1994 ; STEA et al. 1995 ). Six molecular classes of vertebrate 1 subunits have been cloned, and invertebrate 1 subunits have been cloned from Musca (GRABNER et al. 1994 ), Caenorhabditis elegans (SCHAFER and KENYON 1995 ; LEE et al. 1997 ), and Drosophila (ZHENG et al. 1995 ; SMITH et al. 1996 ). While much is known about molecular and physiological diversity of calcium channels, the biological role of calcium channel diversity is less well understood. Certain mammalian mutations (and an invertebrate one) offer hints at the biological functions of calcium channels (reviewed by MILLER 1997 ). However, the defects associated with these calcium-channel variants are in the main neuropathologies, e.g., migraines, ataxia, lethargy, and in one case muscular dysgenesis or myotonia; it is difficult to pit such mutant phenotypes against discrete, measurable behaviors exhibited by the wild-type humans, mice, or nematodes (cf. SCHAFER and KENYON 1995 ; LEE et al. 1997 ; MILLER 1997 ; and see the insightful discussion of this issue in GREENSPAN 1998).

We cloned the gene for the Drosophila Dmca1A calcium-channel 1 subunit and concluded from cDNA analysis that several variant isoforms were generated by alternative splicing (SMITH et al. 1996 ; PEIXOTO et al. 1997 ). The Dmca1A-encoding gene spans several chromosomal breakpoints that fail to complement the courtship song defective cacophony (cac) mutation, the visually defective nightblind-A (nbA) mutations, and the lethal(1)L13 mutations (LINDSLEY and ZIMM 1992 ), suggesting that Dmca1A might be encoded by this gene or these genes.

Mutant alleles originally isolated as nbA strains lead to defects in visually mediated behaviors (HEISENBERG and GOTZ 1975 ; BULTHOFF 1982 ; KULKARNI and HALL 1987 ) and light-induced responses of the visual system (HEISENBERG and GOTZ 1975 ; COOMBE 1986 ; COOMBE and HEISENBERG 1986 ; HOMYK and PYE 1989 ). These results stemmed from analyses of cac^H18 (née nba^H18) and cac^EE171 (née nba^EE171). An unpublished nbA mutant (now cac^P73, this article) was isolated in the lab of W. L. PAK (cf. PAK 1975 ) but was not extensively analyzed.

The courtship song of male Drosophila melanogaster is generated by wing vibration and has two components: sound "pulses" of 2–3 cycles each, repeated at approximately 35-msec intervals in trains of 2–30 pulses; and humming sounds (also called sine song, reviewed by HALL et al. 1980 , HALL et al. 1990 ; HALL 1994 ). The cac^S mutation causes song pulses to be polycyclic and higher amplitude, with longer-than-normal intervals between pulses. This is exemplified in Figure 5 (within the DISCUSSION), mentioning of which also previews a connection we will make between changes in song patterns and more than one gene's worth of ion-channel variations. cac^S causes a depression of locomotor activity, but only at high temperatures, and does not affect sine song or the adult's ability to fly (SCHILCHER 1976 , SCHILCHER 1977 ; KULKARNI and HALL 1987 ; WHEELER et al. 1989 ; PEIXOTO and HALL 1998 ).

View larger version (11K): In this window In a new window Download PPT slide   Figure 1. Sequence analysis of Dmca1A transcripts. The products of the cac gene are diagrammed, with locations of transmembrane domains of the Dmca1A polypeptide indicated by vertical black bars; untranslated regions (UTR) of the transcript are in gray. Mutually exclusive 118-bp alternative exons in the domain I-II loop (exons I-IIa and I-IIb) are expanded 10x; sequences from introns flanking exon I-IIa are indicated in gray. The horizontal line above the diagram indicates the probe used on Northern blots (Figure 4). Sequence fragments obtained from cac mutants by RT-PCR are indicated by horizontal lines underneath the diagram; diagonally descending sets derive from the same primary PCR. Unlabeled arrows indicate the positions of silent polymorphisms. Labeled arrows indicate nonsilent polymorphisms in the indicated mutant flies (H18, cacH18).

View larger version (32K): In this window In a new window Download PPT slide   Figure 2. Mutations affecting the Dmca1A calcium-channel 1 subunit. (A) cacS mutant flies have a TTC-Phe to ATC-Ile change; the arrow indicates the variant nucleotide in cacS; the protein sequence is aligned below the chromatographs, with the cacS variant in parentheses. (B) cacH18 mutant flies have a TGG-Trp to TAG-stop change in exon I/IIa; the arrow indicates the variant nucleotide in cacH18 mutant flies; the protein sequence is shown aligned below the chromatographs, with the cacH18-variant stop codon in parentheses. (C) The cacS mutation affects a conserved phenylalanine in transmembrane domain IIIS6; the Dmca1A calcium-channel sequence is aligned with representative calcium-channel, sodium-channel, and minK potassium-channel sequences; the phenylalanine affected by this Drosophila mutation is boxed; the channels in other species are as follows: 1A, rat brain class A (GenBank accession no. M64373; STARR et al. 1991); 1B, rat brain class B (GenBank accession no. M92905; DUBEL et al. 1992 ); 1E, rat brain class E (GenBank accession no. M94172; SOONG et al. 1993 ); 1C, rat brain class C (GenBank accession no. M67516; SNUTCH et al. 1991 ); 1D, human class D (GenBank accession no. M76558; WILLIAMS et al. 1992 ); 1S, rat skeletal muscle (GenBank accession no. X05921; TANABE et al. 1987 ). (D) The cacH18 mutation creates a stop codon in exon I/IIa; the location of the stop codon is indicated with an arrowhead; sequences known to be important in ß-subunit interaction (PRAGNELL et al. 1994 ) and in Gß-mediated channel modulation of vertebrate 1 subunits (ZHANG et al. 1996 ; DE WAARD et al. 1997 ; HERLITZE et al. 1997A , HERLITZE et al. 1997B ; ZAMPONI et al. 1997 ) are aligned below the exon sequences, with conserved amino acids boxed.

View larger version (11K): In this window In a new window Download PPT slide   Figure 3. Electroretinogram phenotypes of flies expressing various cac allelic combinations. Such alleles are indicated by superscripts only. ERGs were obtained by extracellular recordings from eyes of the indicated types; positive is up. Positive-going lights-on and negative lights-off transients (visible in "wild-type" traces in A and B) are caused by postsynaptic activity in the outer optic ganglia of the brain (COOMBE 1986 ; COOMBE and HEISENBERG 1986 ; PAK 1975 ). The sustained negative light coincident receptor potential (LCRP) is caused by depolarization of photoreceptor cells, and in wild type typically has a quick (<1/2 second) overshoot phase, followed by a sustained phase. The "kinetic" measure of the LCRP (see text; Table 3) is the time interval between light-on and the maximum absolute amplitude of the LCRP; in the case of cacP73, this corresponds to the time of lights-off. Downward-pointing arrows in the cacH18/cacL-6 type indicate the time of lights on and off; this time can be inferred in other traces by the onset and offset of the LCRP or by the onset and offset of transients; the interval between lights-on and -off is three seconds in all cases. (A) ERGs from wild-type and representative of those from the indicated cac mutants. (B) Representative examples of ERGs expressing a novel rebound component of the ERG; the second trace of each is expanded 5x in time. (C) Representative examples of ERGs expressing a novel light-refractory, wandering-baseline phenotype.

View larger version (118K): In this window In a new window Download PPT slide   Figure 4. Expression of Dmca1A transcripts. (A) Northern blots of head and body RNA from wild-type and mutant adult flies, detected with a Dmca1A-specific probe (see Figure 1). (B) Tissue-specific expression of exons I-IIa and I-IIb. The diagram shows the organization of the Dmca1A transcript and the relative positions (arrowheads) of primers used for isoform-specific semiquantitative RT-PCR. Agarose gels of RT-PCRs were ethidium-stained and photographed. The top lines of headers for the gels indicate (in the main) genotypes of the flies from which the head or body RNAs were extracted: WT, wild-type; cac, the song mutant cacS; 1, 3, and 4 the visual mutants cacH18, cacEE171, and cacP73; eya, eyes-absent; "eyes," RNA was extracted only from that WT tissue. Below the genotype or tissue-type headers are designations for the exon-specific primers used, i.e., A, B, C, or D. In this semiquantitative analysis, the amplification levels of the four PCR products are not comparable with each other, but amplification levels of each product within an experiment are comparable between individuals (see text).

View larger version (22K): In this window In a new window Download PPT slide   Figure 5. How the cacophony-encoded calcium-channel subunit could exert a key feature of courtship-song control. At the top, part of a train of song pulses is shown (occurring over about 0.25 sec) above a crude indication of voltage changes that accompany the function of a generic pacemaker cell (cf. HILLE 1992 ); six tone pulses—recorded from a diplo-X cacS/cacS fly transformed into a male by homozygosity for the tra mutation—are shown (thus, five interpulse intervals, whose timespans are ca. 40–45 msec); as is typified by this mutant's song, only the first two pulses exhibit cycle numbers in the normal range, with the others being abnormally polycyclic; the sawtooths below the song trace represent a hypothetical alignment of cyclic membrane potential changes in a song-controlling pacemaker cell with the repetitive bursts of courtship-song pulses; depolarized states, coinciding with the pulses, are caused by processes outlined in the bottom part of the figure: Here, a "leak" current in "bursting" cells causes a slow depolarization, eventually reaching a threshold and triggering a burst of action potentials (cf. HILLE 1992 ); this causes Ca2+ entry through voltage-dependent Ca2+-channels (whose subunits are suggested to be encoded by cac); this in turn activates K(Ca) channels (suggested to be encoded by the slo gene—mutant forms of which cause song disruptions; see text); this causes Ca2+-dependent inactivation of Ca2+ channels, which in turn leads to eventual hyperpolarization and cessation of action potentials; membrane potential and intracellular calcium return to basal levels, and the cycle can start again (cf. HILLE 1992 ).

Mutations of lethal(1)L13 (now cac^L-'s) cause late embryonic lethality in homozygotes (PERRIMON et al. 1989 ) with no apparent anatomical defects (PERRIMON et al. 1984 ). These lethal mutations fail to complement cac^S for song phenotypes; and most fail to complement cac^EE171 for visual phenotypes (KULKARNI and HALL 1987 ; HOMYK and PYE 1989 ), implying that the mutations are mutated at the same locus. However, one of these lethal mutations, cac^L-24, complements nbA mutations for all visual phenotypes but uncovers the song abnormalities associated with cac^S (this article). For its part, cac^S has not been demonstrably defective in visual functions; cac^EE171 was found to be normal for courtship song; and cac^S and cac^EE171 complemented each other for both song and visual phenotypes (KULKARNI and HALL 1987 ).

Nevertheless, the cytological localization, protein function, and transcript diversity of Dmca1A suggested that it might be the product of the cac gene. We have substantially extended and deepened the analysis of phenotypic interactions between cac alleles. Our results show that cac and nbA mutations are mutated with respect to the same function and indicate that the cacophony gene is involved in visual transduction as well as putative signal-transmission events that underlie the male's courtship song. In conjunction with these phenogenetic analyses, molecular characterization of mutations at this locus showed that cac^S carries a missense substitution at the site of a highly conserved phenylalanine residue; and cac^H18 (which affects visual but not song phenotypes) is a nonsense mutant, predicted to eliminate expression of an alternative exon that is differentially expressed in the eye.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Reverse transcription (RT)-PCR:
Total RNA for each genotype was isolated with TRIzol reagent (GIBCO BRL, Gaithersburg, MD) from intact adult flies. Total RNA (1.5 µg) was reverse-transcribed with random hexamer primers and AMV Reverse Transcriptase (Promega, Madison, WI) and then amplified for sequence analysis by nested PCR. All PCRs were in 100-µl volumes and included the following: 0.2 µM each primer; 0.2 mM each dNTP; 1.75 mM MgCL₂; 50 mM KCl; 10 mM Tris-HCl, pH 9.0; 0.1% Triton X-100; and 5 units Taq polymerase. Primary PCRs—designed to amplify 1.5–2-kb products, containing 2 µl of cDNA template (from a 1:100 dilution of the first-strand cDNA)—were denatured at 94° for 2 min and subjected to 20 amplification cycles (94°, 1 min; 59°, 1 min; 72°, 2 min), followed by 5 min at 72°. Secondary nested PCRs were prepared as above (except that they were designed to amplify 380–1250-bp products), contained 5 µl of the first-round PCR product as template, and were given 40 cycles (94°, 1 min; 59° 1 min; 72°, 1 min).

PCR from genomic DNA:
To extract genomic DNA for PCR, single flies were homogenized in 50 µl of "squishing buffer" (10 mM Tris/HCl, pH 8.2, 1 mM EDTA, 25 mM NaCl, 200 µg/ml Proteinase K), incubated 30 min at room temperature, and heated to 95° for 2 min. PCRs contained 2–4 µl of fresh extracted DNA, amplified 40 cycles (94°, 1 min; 59°, 1 min; 72°, 1 min), and were otherwise identical to those above. PCR primers were designed to genomic DNA flanking exon I/IIa, and had the sequences: (B5'-2) 5'-CCC AAA TTT TCG CCT GTT GC-3'; (B3'-2) 5'-GGT TGT GTT GTA TGA CGT TCG-3'. Control PCRs contained no template DNA.

Sequencing and analysis of resulting data:
PCR products were purified with the QIAquick PCR Purification Kit (Qiagen, Santa Clarita, CA) and eluted in 25 µl H₂O. Typically, 9.5 µl of eluant was sequenced using the ABI PRISM Dye-Deoxy Terminator Cycle Sequencing Kit (Perkin Elmer, Foster City, CA) and electrophoresed on an ABI 373A automated sequencer. Sequence analysis was done with GCG sequence analysis software (GENETICS COMPUTER GROUP 1991). Ambiguities were resolved by direct reference to the sequence chromatograph. At several sites, RT-PCR derived sequence did not allow unambiguous identification of nucleotide identity; nucleotides at these sites were considered not to be polymorphic if the chromatographic pattern from each of the different strains was similar and consistent with the published Dmca1A sequence. Silent third-base polymorphisms were detected at seven sites in the Dmca1A transcript; we interpret these as strain differences. All polymorphisms were confirmed by subsequent RT-PCR-sequencing analyses (n 3) from RNA independently purified from the relevant fly strain.

Mutants and other genetic variants:
All abnormal genotypes are listed in LINDSLEY and ZIMM 1992 . Most of these variants involve the X-chromosomal cacophony locus (see below). Others included the sightless no-receptor-potenital-A (norpA^P24) and eyes-absent (eya¹) mutants. cac^P73 (née nbA^P73) was obtained from W. L. PAK. Flies were maintained at 25°, 70% relative humidity, and those to be tested in behavioral or physiological experiments were collected under CO₂ anesthesia within 24 hr of eclosion and aged for 3–7 days before testing. All viable cac mutant and wild-type strains used for analysis were subjected to a five-generation outcross to a common stock marked with vermilion², garnet², and forked (v² g² f ). Several separately-maintained cac^S lines, initially expressing significantly different numbers of cycles per pulse (CPP) and interpulse internal (IPI) scores (see below), converged to common, mutant values during the outcrossing, indicating that genetic variability affecting these parameters had been removed. To guard against subsequent selection for genetic modifiers that might again cause phenotypic drift, the outcrossed viable mutants were crossed into and maintained in stocks with an attached-XY chromosome (FM7-Y; also known as Y^SX-Y^L, In(1)FM7: LINDSLEY and ZIMM 1992). The only fertile flies in these stocks are FM7-Y males and heterozygous females, which do not express recessive phenotypes, minimizing the accumulation of modifying factors that can degrade the severity of the behavioral phenotypes (cf. DE BELLE and HEISENBERG 1996 ). Lethal cac alleles were crossed to the v² g² f stock, then reisolated by crossing to a balancer stock [In(1)FM7] that had been outcrossed to the strain indicated above. Flies for visual-behavior or ERG tests (see below) were obtained by crosses from these stocks. Flies for courtship-song analysis were obtained by first crossing the relevant cac-mutant or cac⁺-containing stocks to flies from a single stock containing transformer (tra) mutation—which when homozygous transforms diplo-X, chromosomally female flies into phenotypic pseudomales that robustly perform male-specific behaviors (e.g., BERNSTEIN et al. 1992 ) and subsequently crossing the resulting progeny together to produce cac-heteroallelic diplo-X, homozygous tra pseudomales.

Courtship-song analysis:
The recording and analytical procedures were essentially as described by RENDAHL et al. 1992 . Males or pseudomales were stored individually for 3–7 days after eclosion. Each fly was then placed with one or two attached-X [C(1)DX, y f] females in the recording chamber of an Insectavox (GORCZYCA and HALL 1987 ). Five minutes of courtship were recorded on a Sony (Parkridge, NJ) Hi-8 audio-video recorder. The analog sound record was transferred at 2000 Hz via MacAdios II analog-digital converter to a Macintosh II or Quadra computer. Song-pulse locations were marked using LifeSong software (BERNSTEIN et al. 1992 ) and subsequently analyzed on a VAX 8650. The values associated with pulse-song parameters for a given genotype are the means among individuals, computed from the means determined for each individual's 5 min of recorded courtship. The parameters extracted from the analyses were IPIs (e.g., SCHILCHER 1976 , SCHILCHER 1977 ; RENDAHL et al. 1992 ; PEIXOTO and HALL 1998 ), CPPs (e.g., KULKARNI and HALL 1987 ; BERNSTEIN et al. 1992 ; RENDAHL et al. 1992 ), intrapulse (or carrier) frequencies (e.g., WHEELER et al. 1989 ; BERNSTEIN et al. 1992 ; RENDAHL et al. 1992 ), and amplitude of pulses (in arbitrary units; cf. PEIXOTO and HALL 1998 ). Recordings were obtained at temperatures between 20–25°; the IPI varies with temperature, with a slope of 1.74 msec/° (RITCHIE and KYRIACOU 1994 ); thus IPI values were corrected using this relationship to a standard temperature of 22°.

Visually mediated behaviors:
The walking optomotor assay was performed in a rotating arena essentially as described by RENDAHL et al. 1992 and references therein. Each fly was given six consecutive 1-min trials with alternating clockwise and counterclockwise rotation of the arena. The number of times a given fly crossed a quadrant line in the same direction as a rotating background (F), or in the opposite direction (B), were totaled. An "optomotor index" was calculated with the formula (F - B)/(F + B), yielding a score of one if the fly always followed the rotating background, and a score of zero if the fly was equally likely to move either direction.

A "countercurrent-regression" assay was devised to test the ability of flies to phototax under varying light intensities. A five-tube countercurrent phototaxis apparatus (BENZER 1967 ) was placed 10 cm from a light source (General Electric 18-W fluorescent bulb) equipped with light filters (Rosco N.6 2-stop neutral density). Approximately 35–50 flies were introduced into the proximal tube of the apparatus, tapped to the bottom, and allowed to distribute themselves between the proximal and distal tubes for 30 sec. The apparatus was then shifted and tapped to return the flies to the proximal tube. This was repeated five times, fractionating the flies into six groups based on the number of transitions toward light (0–5) by each. A "transition score" was calculated, representing for each group the number of actual transitions toward light as a fraction of the total possible. The procedure was repeated in the dark and with filtered light intensities of 0.1 f.c., 1.4 f.c., and 340 f.c. Parallel tests of wild-type and several mutant lines in constant darkness or constant 340-f.c. light detected no fatigue effects from the multiple trials. Transition scores were regressed against light intensity. Positive regression slopes indicate increased phototaxis in increasing light intensities, and negative slopes indicate decreased phototaxis.

Y-tube phototaxis was performed as described by KULKARNI and HALL 1987 . Flies were placed in a darkened start tube and then exposed for 120 sec to a choice point allowing access to either a dark tube or a tube with a light source at the distal end. The number of flies choosing the lighted (L) or darkened (D) tubes was counted, and a phototaxis index with values between minus one and one was calculated using the formula (L - D)/(total flies); positive scores indicate a preference for the lighted tube, and negative scores for the dark tube.

Electroretinograms (ERGs):
ERGs were recorded as described by RENDAHL et al. 1992 . All test recordings were preceded and followed by recordings of wild-type flies; only if both control recordings were normal was the test ERG accepted for analysis. The absolute amplitude in mV of the light-on and light-off transient spikes [from baseline and from the end of the light coincident receptor potential (LCRP), respectively] and the maximum amplitude of the LCRP were measured, as well as the time between light-on and the maximal absolute LCRP amplitude.

Statistics:
These analyses were performed by application of JMP software (SAS INSTITUTE 1994). Optomotor and Y-tube scores were transformed to arcsin[score], countercurrent-regression, ERG and song scores except CPP were transformed with 1/[score], and CPP scores with 1/sqrt[score], yielding homogeneity of variances (SOKAL and ROHLF 1995 ). ANOVA for each assay gave a P[equivalent means] of <<0.001. Genotypes were subsequently tested for equivalence to wild type (DUNNETT 1955 ), and also for equivalence to sightless norpA^P24 and eya¹ control flies (visual behavior) or to cac^S hemizygous mutant males (song). These parametric methods were supplemented with Wilcoxon/Kruskal-Wallis nonparametric pairwise comparison tests (see legends to Table 1 and Table 2).

Northern blotting:
Drosophila heads and bodies were separated by sieving in liquid nitrogen (LEVY and MANNING 1981 ). Total RNA was isolated using TRIzol reagent (GIBCO BRL). Poly(A)⁺ RNA was purified with the PolyATtract mRNA Isolation System (Promega). PolyA⁺ RNA (3.5 µg) was loaded in each lane, electrophoresed through a 1% agarose/formaldehyde gel, and transferred onto a Hybond-N⁺ membrane as described by the manufacturer (Amersham, Arlington Heights, IL). Probe (2 x 10⁶ cpm/ml) was generated by random-primed ³²P-labeling of a 682-bp PCR fragment generated from cDNA clone cSK53 (SMITH et al. 1996 ) with primers U857 (5'-CCG AAA TTG AAA GCC GTG TT-3') and L1522 (5'-ACT TCA TCA TCA GCA ATA GG-3'), corresponding to amino acids 889 (C-terminal end of IIIS4) through 1116 (C-terminal end of IVS1). The lower portion of each blot was removed and probed with 2 x 10⁶ cpm/ml of an rp49 probe (O'CONNELL and ROSBASH 1984 ) as a control for equal loading. The Dmca1A portion was exposed to film for 9 days and the rp49 portion, for 5 hr.

Tissue-specific RT-PCR:
Isolation of RNA and reverse transcription was performed as above, except that 0.25 µg of total RNA from eyes was reverse-transcribed, with reagents scaled proportionally. Eyes were isolated by freezing flies in liquid nitrogen, transferring to ethanol on dry ice, and "popping" the eyes off with a fine tungsten needle. A single-step PCR in 100 µl total volume was performed as above, with 2 µl of template cDNA, an annealing temperature of 61°, and amplification for 30 cycles. Primer sequences (applied in the experiments depicted in Figure 4) were designed to give products between 188 and 250 bp in length: (A) 5'-CCA TGT TTC AGA CAG CAA TGG-3'; (B) 5'-GTA CGA GAC CAT TGC TGT CTG-3'; (C) 5'-CCT AAA CTT AGA AGG CAG CAG C-3'; (D) 5'-CGA ATT CAC CAC TAA GGA CAC C-3', (5') 5'-TGA CCG TAT TCC AAT GTA TC-3', (3') 5'-CTT CCT CTT CCT CTG TAT-3'.

Preliminary titrations using cDNA dilutions of 1:1, 1:10, 1:100, and 1:1000 showed that dilutions of <1:10 were subsaturating for PCR amplification of each product (HUET et al. 1993 ); subsequent experiments were done with 1:100 dilutions for primer pairs A and B and 1:250 dilutions for primer pairs C and D. Each PCR (10 µl) was electrophoresed on 2% agarose gels, stained with Ethidium Bromide, and photographed under UV illumination. Representative examples of each size class were purified and sequenced as described in RESULTS.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Nucleotide substitutions in cac mutants:
The Dmca1A open reading frame (ORF) is diagrammed in Figure 1; it also depicts the basic structure of this calcium-channel subunit and its four quasi-repeated domains, each containing six transmembrane regions. We sequenced the Dmca1A ORF (by RT-PCR, as outlined in Figure 1) from four viable mutants: cac^S, cac^H18, cac^EE171, and cac^P73. Except as noted with respect to sites associated with certain cac mutations (Figure 1), the sequences obtained from these four previously unanalyzed strains were identical to the known informational content of Dmca1A (cf. SMITH et al. 1996 ; PEIXOTO et al. 1997 ).

One proviso that must accompany the statement just made involves post-transcriptional modification of certain adenosine residues within the Dmca1A ORF. We had found that certain sites exhibit heterogeneity (among cDNAs) in terms of the nucleotides present at these sites (SMITH et al. 1996 ; PEIXOTO et al. 1997 ). That the nucleotides were either adenosine or guanosine (and other arguments) suggested strongly that these sites are subjected to "A-to-G" RNA editing (see SMITH et al. 1996 , SMITH et al. 1998 , for evidence and discussion; as well as SIMPSON and EMESON 1996 , for background information). The current point is that the same kind of A/G heterogeneity was detected at the relevant sites in cDNAs derived from all four mutants (data not shown), as found in the pattern previously established for wild type.

The molecular etiology of two cac mutants seems to involve a straightforward missense and an intriguing nonsense mutation, respectively: A single transversion in sequence encoding transmembrane domain IIIS6 was the only nonsilent polymorphism detected in cac^S mutant flies (Figure 2A). This nucleotide substitution changes a phenylalanine codon (TTC) to an isoleucine one (ATC). Given the current knowledge of alternative splicing of Dmca1A's primary transcript (SMITH et al. 1996 ), the amino-acid substitution in cac^S would be expected to affect all of the mature transcript types. The affected phenylalanine is perfectly conserved among calcium channel 1 and sodium channel subunits (Figure 2C), and in the transmembrane domain region of the MinK subunit of I_kS potassium channels (WANG et al. 1996 ); this I_kS transmembrane region can be aligned with transmembrane domain IIIS6 of Na⁺ and Ca²⁺ 1 channel subunits but is otherwise quite diverged (TAKUMI et al. 1988 ; BARHANIN et al. 1996 ; SANGUINETTI et al. 1996 ). The cac^S-defined phenylalanine is not conserved in S6 transmembrane domains of six-transmembrane-domain potassium channels (see DISCUSSION).

Dmca1A has mutually exclusive alternative exons (I/IIa and I/IIb) that encode a portion of the intracellular loop between homologous repeats I and II (SMITH et al. 1996 ) (Figure 2D). Exon I/IIb of Dmca1A encodes a sequence motif that mediates interaction with calcium channel ß-subunits and is conserved in all 1 subunits (PRAGNELL et al. 1994 ). Exon I/IIb also encodes an overlapping sequence, present in a subset of 1 subunits, which mediates modulation by G-protein G_ßg subunits (ZHANG et al. 1996 ; DE WAARD et al. 1997 ; HERLITZE et al. 1997A , HERLITZE et al. 1997B ; ZAMPONI et al. 1997 ). The alternative Dmca1A exon I/IIa encodes a sequence that has poor conservation of the ß-subunit-interaction motif and does not have the G_ß interaction motif.

Exon I/IIb was readily amplified and sequenced from RT-PCR templates, but exon I/IIa was amplified at a reduced level. Exon I/IIa, along with 64 flanking 5' nucleotides and 56 flanking 3' nucleotides, was amplified and sequenced from genomic DNA. A single transition was found near the 3' end of exon I/IIa in cac^H18 (Figure 2B), altering a tryptophan codon to a TAG amber stop codon (Figure 2D). This nonsense substitution was the only nonsilent sequence polymorphism detected in the cac^H18 ORF. The nucleotide substitution would cause premature termination and eliminate expression of Dmca1A isoforms containing the relatively unconserved amino-acid sequence encoded by exon I/IIa.

We did not find molecular lesions in the ORFs of cac^EE171 or cac^P73 mutants. Possible explanations for this could be that these are mutations outside the ORF and might affect spatial or temporal regulation or alternative splicing of Dmca1A isoforms necessary for visual function. While we have no evidence for additional Dmca1A alternative exons or for alternative translation initiation sites (see SMITH et al. 1996 ; PEIXOTO et al. 1997 ), we cannot discount the existence of such molecular entities as being involved in visual system function. Given that cac^H18 is mutated in alternative exon I/IIa, that cac^H18 mutants are defective only in vision, and that cac^L-24 has no vision defects, we predicted that the cac^L-24 mutation might be in exon I/IIb. However, sequence analysis of this alternative exon from cac^L-24 mutants did not reveal any sequence polymorphisms (not shown). This could imply the existence of additional alternative Dmca1A transcripts that are not involved in visual processes.

To delve into those processes, and the courtship-song ones as well, we now turn to the phenogenetics of cacophony. The results that follow are sometimes less accessible than the molecular findings. Nevertheless, we believe that the behavioral and physiological defects exhibited by the several cac-mutant combinations—involving discrete as well as graded phenotypic impairments—are crucial for revealing both that cac mutations define an allelic series and how the channel subunit encoded by this gene participates in distinct features of CNS and PNS function.

Courtship-song defects are associated with a subset of cac mutants:
The original cacophony song mutant was found to generate anomalously polycyclic "tone pulses" (for example, Figure 5). To determine better whether further mutations at this locus cause these kinds of pulse abnormalities—or additional or other ones—we analyzed several song parameters beyond that involving number of cycles per pulse. The flies whose courtship wing vibrations were recorded represented all viable genotypic combinations of cac alleles, as well as those that are lethally mutated at or deleted of the cac locus (see MATERIALS AND METHODS). These diplo-X (chromosomally female) flies were rendered phenotypically male by use of the transformer mutation (cf. KULKARNI and HALL 1987 ; BERNSTEIN et al. 1992 ). The behavioral results were compared to those from wild-type flies and cac^S homozygotes (Table 1; summarized in Table 4). Elements of these phenogenetic results (here, and in the next three sections) are complex, but we hope the reader will bear with us.

The cac^S mutant types—carrying that allele as their only cac one, or cac^S heterozygous with a cac-lethal variant—exhibited defects in CPP, IPI, the pulse amplitude, and in the breadth of and number of peaks in the FFT-derived frequency spectrum (Table 1); but these flies were normal for intra-pulse frequency (a species-specific song character, e.g., BERNSTEIN et al. 1992 ). ANOVA revealed that significant differences existed between genotypes for each of these song parameters (P < 0.001). Post hoc comparisons of each genotype to both cac⁺- and cac^S-expressing types revealed no significant differences from the control groups for intrapulse frequency or the two frequency spectrum parameters ( = 0.01).

For both CPP and IPI song values, the cac genotypes fell into three groups: heteroallelic combinations of cac^S with lethal cac alleles or with the deletion were indistinguishable from mutant cac^S homozygotes; the cac^S/+ heterozygote was intermediate for these parameters; and the remainder of the genetic types were indistinguishable from wild type (Table 1; summary in Table 4). The CPP values form an essentially bimodal distribution, with a rather distinct break between mutant scores and ones that were like wild type, hence very few CPPs that could be termed intermediate. However, the IPIs computed from recordings of these many types led to a quasi-continuous distribution of interpulse intervals (as implied in Table 1), which is difficult to interpret. Nevertheless, the homozygous cac^EE171, cac^H18, and cac^P73 types, heteroallelic combinations involving these three viable mutations, and combinations of these three mutations with the cac^- deletion, all gave CPP and IPI values that are almost certainly normal (Table 4). Flies carrying these mutations exhibited singing defects (with values intermediate between mutant-like and normal) only when a given mutation was heterozygous with certain lethal alleles (Table 4). Thus, we hypothesize that the particular Dmca1A changes in these lethals interact with the viable mutations in special ways to produce relatively subtle song problems that are not observed in flies carrying two doses of the visual mutations or one dose of them and no cac gene (e.g., cac^EE171/Df).

For pulse amplitude, cac^S homozygotes and heteroallelic combinations with lethal cac alleles or the deletion (Df) gave mutant-like (high amplitude) scores, while cac^S/+ and Df/+ were intermediate. Although cac^P73 homozygotes have normal amplitude values, the heteroallelic combination of cac^P73 with cac^S led to high amplitudes. Combinations with the deletion and with several of the lethal alleles gave intermediate values, implying that the cac^P73 mutant is hypomorphic for a function required for normal pulse amplitude. One problem with interpreting this result in the narrow sense is that these amplitude values are highly variable among genotypes, including some unusually large values at the high end of the mutant range. Nevertheless, other genotypic combinations involving cac⁺, cac^H18, cac^EE171, and cac^P73 (that is, except for cac^P73/cac^S and cac^P73/lethal) gave low song-pulse amplitudes in the normal range (whereby we take that range to be a value of ca. 13 or less; Table 1).

In summary, the cac^S mutation is overwhelmingly "the" song variant at the genetic locus encoding Dmca1A (see summary Table 4, below). Yet—to resist an impulse to view a given cac allele as exquisitely "specific"—we reiterate that one song anomaly was revealed in several genetic combinations involving the "visual-only" allele cac^P73. Analogous findings are included in the next three sections, meaning that we have teased out subtle visual defects in flies carrying the so-called "song-only" allele cac^S.

The cac mutants define a phenotypic series for visually mediated behaviors:
nbA mutants exhibit certain subnormalities and anomalies in visual function, based on rather limited tests of visually-mediated behavior (see Introduction). We felt that a broader spectrum of visual tests should be applied to the full array of cac-locus variants, to better reveal which mutations are truly "visual-specific." The overall response of Drosophila to visual stimuli requires not only that photoreceptors function appropriately, but also that this signal be transmitted, integrated, and output to motor effectors; these behaviors constitute the basis for a global bioassay of these functions. We thus performed walking optomotor assays and a pair of phototaxis assays for all viable cac genotypic combinations and compared these results to those from wild-type controls and from flies known to be blind (Table 2).

The optomotor assay measures the ability of flies to respond to moving visual cues in the environment. All combinations involving cac⁺, cac^S or the lethal allele cac^L-24, which complements all visual phenotypes (see below) were normal in the optomotor assay (score >0.69, perfect score = 1.00). The heteroallelic combination cac^P73/cac^L-20 had an intermediate optomotor score (0.42). All other genotypes were optomotor-blind (<0.09) and in this sense behaved indistinguishably from the genetically blind or eyeless controls (Table 2; Table 4).

The countercurrent-phototaxis assay measures the effect of increasing light intensity on the ability of flies to phototax in a light gradient. cac^EE171 homozygotes and heteroallelic combinations with lethal alleles (except cac^L-24 ), all of which were optomotor-blind, exhibited a robust response in this assay but with reversed sign (score <-0.44); that is, as light intensity increased, the flies' phototaxis scores were reduced to below those obtained in the dark, confirming KULKARNI and HALL 1987 for the phenotype of cac^EE171. The other genotypes fell into two distinct groups (Table 2): those indistinguishable from blind flies (score between -0.11 and 0.22), and those indistinguishable from wild type (score >0.41). Every optomotor-normal genotype was phototaxis-normal in the countercurrent-regression assay. In addition, cac^H18 and cac^P73 homozygotes, which were optomotor-blind, were phototaxis-normal in this assay.

The Y-tube phototaxis assay asks flies to choose between a dark tube and one illuminated at its distal end. The genotypes fell into four distinct groups (Table 2): those involving cac^EE171 (except cac^EE171/cac^L-6, which gave a score not significantly different from the sightless controls) had negative scores (<-0.27); cac mutant types were indistinguishable from blind control flies (-0.13 and 0.09); two types yielded intermediate scores (cac^S/cac^L-6 and cac^P73/cac^L-6: 0.45 and 0.5); and a group of mutant types was indistinguishable from wild type (>0.69). Genotypes leading to nonblind countercurrent-regression scores gave similar results in this assay. The cac^H18 heteroallelic combinations with either cac^EE171 or cac^P73 were phototaxis-blind in countercurrent-regression but normal in the Y-tube.

With the caveat that cac^EE171 phototaxis scores are negative, the genotypes can be categorized as: behaviorally blind; Y-tube normal; countercurrent-regression- and Y-tube normal; or normal in all three assays (Table 2). This defines a gradient of defects in which relative severity of phenotypic consequence can be assigned (Table 4). The exceptions to this pattern are in the Y-tube assay and involve lethal allele cac^L-6. The behavior of cac^EE171/cac^L-6 is indistinguishable from blindness, and cac^S/cac^L-6 is intermediate between wild type and blind—although we predicted from the second data column of Table 2 that cac^EE171/cac^L-6 would be negatively phototactic and cac^S/cac^L-6 normally phototactic.

We uncovered special kinds of interactions between certain combinations of cac-locus variants—as opposed to a situation in which a given allele would always yield a certain phenotype whenever it is in combination with a cac-variant that falls within another particular category. Thus, for example, the cac^P73 allele is optomotor-defective when homozygous or when combined with lethal cac alleles (except cac^L-24); and cac^L-20 is optomotor-defective with either cac^EE171 or cac^H18, implying that both cac^P73 and cac^L-20 have defects in functions required for normal optomotor behavior. However, the cac^P73/cac^L-20 combination is not optomotor-blind, although it does have a reduced optomotor score; this implies that these alleles might be defective in partially separable functions, such that each can complement the defect of the other. Another such example involves flies homozygous for cac^H18 or cac^P73, which gave normal phototactic responses, and cac^H18/cac^P73 flies, which were phototaxis-blind in the countercurrent-regression assay. Furthermore, the pattern of lethal alleles that uncovers blindness is different between these two viable alleles: The cac^H18 phototaxis defect is uncovered by only three of the lethal alleles (cac^L-13, cac^L-6, and cac^L-20), while that of cac^P73 is uncovered by an overlapping but different set of lethal alleles (cac^L-13 and cac^L-10). This suggests that cac^H18 and cac^P73 are hypomorphic for separable functions, such that two copies of either are sufficient to normalize the phenotype, but one copy of each is insufficient.

cac mutants have genetically separable ERG defects:
ERGs measure the summed light-induced electrical activity from photoreceptors and optic ganglia (e.g., PAK 1975 ). We recorded ERGs from flies expressing all cac-associated genotypes and analyzed them for the presence and amplitude of lights-on and -off transients, as well as for the amplitude and kinetics of the LCRP (as defined in Figure 3A, Table 3). All genotypes involving cac⁺ or cac^L-24 led to quantitatively normal ERGs (Table 3), as did hemizygosity for cac^S (however, see next section for more on the latter mutant). Also, any cac-locus variant when heterozygous with cac⁺ resulted in normal ERGs (bottom of Table 3); this lack of dominant effects validates the mutant phenotypes observed in certain transheterozygotes.

ERGs from heteroallelic combinations involving cac^H18, cac^EE171, or cac^P73 with each other or with lethal cac alleles (except cac^L-24) had no lights-on and -off transients. cac^S by itself has been thought to have no visual system defects (KULKARNI and HALL 1987 ; HOMYK and PYE 1989 ). Yet, heteroallelic combinations of cac^S with cac^H18, cac^EE171, cac^L-6, cac^L-13, cac^L-20, or the cac^- deletion had lights-on transients that were significantly lower than normal in amplitude, and all of these genotypes except the combinations with cac^EE171 and cac^L-20 led to low-amplitude lights-off transients (Table 3). The nominal significance of these amplitude reductions notwithstanding (Table 3), we point out that the ERG transients measured in flies expressing the cac^S combinations (for example) were approximately 60% of normal, i.e., amplitude values from cac⁺-bearing controls or other types, such as cac^S/cac^S, that gave control-like values.

We found that several cac mutants have distinctive and easily recognizable LCRP aberrations (Figure 3; Table 3): cac^H18 had a normal-appearing LCRP with significantly reduced amplitude; cac^EE171 a low-amplitude LCRP with an initial reversal of polarity; and cac^P73 a low-amplitude LCRP with slow kinetics of onset and recovery (Figure 3A). Heteroallelic combinations of cac^H18, cac^EE171, or cac^P73 each produced LCRPs that fit one of these categories. The cac^H18 LCRP-shape phenotype is dominant to cac^EE171 and cac^P73 alleles, and cac^EE171 is dominant to cac^P73. With one exception, heteroallelic combinations with the lethal cac alleles (except cac^L-24) uncovered the LCRP shape associated with the relevant viable cac allele: cac^P73/cac^L-13 has a cac^EE171-like LCRP shape. Also, the cac^P73/cac^L-6 type gave a normal LCRP amplitude but defective kinetics, while cac^H18/cac^P73 exhibited a defective LCRP but normal kinetics. Thus, these two ERG components are separable.

We discovered an additional special kind of interaction involving the lethal mutation cac^L-6 (cf. previous section). Although cac^P73 hemizygotes had normal LCRP amplitudes, the cac^P73 heteroallelic combinations with cac^H18, cac^EE171, or with most lethal alleles had low amplitudes; this implies that cac^P73 is hypomorphic for this phenotype and unable to complement the amplitude defect of these alleles. The cac^L-6 heteroallelic combinations with cac^H18 and cac^EE171 nearly abolished the LCRP (Table 3), even though cac^H18 and cac^EE171 homozygotes and heteroallelic combinations with other lethal alleles or the deletion had intact LCRPs (albeit of reduced amplitude). However, the cac^P73/cac^L-6 type showed a normal LCRP amplitude. The cac^L-6 allele must retain a function for which cac^P73 is hypomorphic but would seem to be damaged in a (separate) function in which cac^H18 and cac^EE171 are defective.

In summary, the nbA mutations at this locus (by themselves, or when placed over most of the lethal alleles) lead to blatant ERG defects (Table 4). Combinations involving cac^S and two of the nbAs did produce smaller than normal transient spikes (Table 4). But this mild abnormality should not be taken to infer an all-out lack of complementation, especially inasmuch as other visual parameters are normal or close to it in these transheterozygotes (cac^S/nbs^H18 or cac^S/nbA^EE1) and other heteroallelic combinations involving cac^S and several of the genetic variations at the locus (Table 4). Conservatively, however, it is once again not warranted to claim that viable cac-locus mutants are completely without pleiotropies—by virtue of causing song defects and no visual ones at all, or vice versa. The next section includes some further suggestions that the so-called song mutation causes additional problems with visual-system functioning.

Novel ERG phenotypes of cac mutants:
In flies expressing several of the cac-variant genotypes, we observed a low-amplitude rebound component superimposed on the transition between the lights-on transient and the LCRP (Figure 3B; Table 3). Such an "extra" component seemed to us to be more pronounced than the subtle complexities that can accompany the electrical signals recorded (from Drosophila's visual system) when the lights go on or off (HEISENBERG 1971 ). Moreover, what we judge to be a rebound abnormality was seen only in certain genotypes: heteroallelic combinations of cac^S with vision-defective lethal alleles, with the deletion, or with cac^H18 and cac^P73, and also in cac^H18 and cac^P73 homozygotes; however, it was not seen in cac^S homozygotes (Table 4). A similar but much subtler break in the ERG trace was sometimes seen in cac^P73 heteroallelic combinations with lethal cac alleles, but the discontinuity was of much lower amplitude and not reliably scorable. The peak of the rebound component (Figure 3B) is temporally aligned with the peak of the corneal-positive component in cac^EE171 ERGs, implying that these might have a similar etiology. The appearance of a rebound component in some ERGs from cac^H18 and cac^P73 homozygotes, in which the transients are entirely missing, implies that this phenotype is separable from the lights-on transient defects.

An additional, novel ERG phenotype was sometimes observed in several cac^EE171/cac^P73 heterozygotes and in heteroallelic combinations of these with vision-defective lethal alleles (Table 4). These flies' visual systems produced a distinctive cyclic wandering of the baseline ERG potential, independent of any specific light stimulus (Figure 3C). The amplitude was variable but in no case exceeded the amplitude of normal LCRPs. In addition, these flies were refractory to light-induced responses, in that there was no apparent change in extracellular potential in response to specific light stimulation. Individual flies with detectable light-coincident responses were never observed to exhibit this cyclic wandering of the baseline ERG potential. Moreover, this baseline wandering/light-refractory phenotype was never observed in recordings from other mutant genotypes reported here or in over 250 recordings from wild-type control animals. These additional facts (that is, aside from the mutant result exemplified in Figure 3C) lead us to suggest that the wandering baseline phenotype is not an artifact (such as background noise, signals picked up from the brain, muscles, or heart, or unwanted movements of the specimen).

Genotypes involving cac^L-24 (which is also intact for all other visual functions assayed) did not exhibit either of these novel ERG phenotypes ("rebound," "wandering baseline"; Table 4). Flies expressing cac^H18 or cac^S did not exhibit the wandering/refractory phenotype; and genotypes involving cac^^EE171 did not cause the rebound phenotype, indicating that these visual defects are genetically separable from each other.

Dmca1A transcripts in cac mutants:
We analyzed mRNAs transcribed from the cac gene to ask whether there is a molecular corollary to the "vision-specific" mutations at this locus. Northern-blot analyses indicated, first of all, that the Dmca1A transcript is enriched in heads compared to bodies, and that there were no differences in Dmca1A transcript expression levels between males and females from a wild-type strain or from one carrying the cac^S mutation (Figure 4A). None of the viable cac mutations caused any substantial expression level or transcript size abnormality of the Dmca1A transcript in either body or head.

Given that the vision-defective cac^H18 mutant has a stop codon in exon I/IIa, we asked if transcripts containing either exon I/IIa or I/IIb were expressed in the eye or in other tissues. We designed 5' and 3' PCR primers internal to these exons and used these with primers specific to flanking exons to amplify 188–250-bp RT-PCR products from cDNA derived from body, head, or eyes of adult wild-type, cac, and eyes absent (eya¹) mutant flies (Figure 4B). Each of these four primer pairs yielded a product of the expected size. Southern blotting and hybridization with a Dmca1A probe, and lack of a detectable product in non-reverse-transcribed control reactions, confirmed that the products are specifically amplified from Dmca1A transcripts. Sequence analysis of a representative of each product and detection of the cac^H18 mutation in the appropriate product further confirmed the specificity of the RT-PCR products (data not shown). Amplification of a specific RT-PCR product using two different primer pairs for each of these exons, for each of the mutants, demonstrated that both exons I/IIa and I/IIb are expressed in heads, bodies, and eyes of wild-type and each of the viable cac mutant flies.

To allow semiquantitative analysis of differences in transcript expression levels, we used input quantities of cDNA that were subsaturating for our PCR conditions (HUET et al. 1993 ; see MATERIALS AND METHODS). Each of the four RT-PCR products (A, B, C, D; Figure 4) are from different primers, amplify different sequences, and use different input quantities of cDNA, so apparent quantitative differences between the four different RT-PCR products do not directly reflect underlying quantitative differences in RNA levels. However, the ratio of amplification of each product relative to the others is reproducibly consistent between experiments. In Figure 4B, note the pattern of amplification levels of A and B (specific to exon I/IIa) and C and D (specific to exon I/IIb) and, within a given tissue, the invariant nature of that pattern between wild-type and cac-mutant types. These data indicate that none of the mutants has any substantial effect on the relative expression levels of exons I/IIa and I/IIb.

Amplification from primer pairs A and B relative to C and D is reduced in body compared to head RNA, indicating that the proportion of transcripts containing exon I/IIa is likely to be relatively lower in body than in head. Similarly, reduced amplification from primer pairs A and B from eya¹ heads, which have no compound eyes, indicates that exon I/IIa expression is relatively low in these heads and therefore likely to be high in the eye. This was confirmed by analysis of RNA from isolated eyes, in which the relative amplification of exon I/IIa-specific products was similar to that in intact heads. This semiquantitative analysis gave no indication of the absolute levels of each transcript or of the absolute levels of the differences between them. However, the clear and reproducible differences in relative amplification levels of exon I/IIa- and I/IIb-derived RT-PCR products confirms that transcripts containing each exon are present in eyes, in heads, and in bodies and implies that expression of exon I/IIa is relatively enriched in the eye. These RNA-based results seems very likely to be related to the findings (presented above) about the apparent elimination of the I/IIa Dmca1A isoform by a cac-nbA mutation (Figure 2 and Figure 3) which leads exclusively to visual defects (Table 4).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The cac gene encodes the Dmca1A calcium-channel 1 subunit:
Phenogenetic analysis reveals that cac specifies an essential gene product that is involved in the operation of the visual system (cf. PAK 1975 ; ZUKER 1996 ) and thoracic neuromuscular systems (cf. HALL et al. 1990 ; HALL 1994 ) as well as being required for specific behavioral and physiological functions. The Dmca1A transcription unit maps to l(1)L13-associated chromosomal lesions (KULKARNI and HALL 1987 ; SMITH et al. 1996 ), and the cac^S and cac^H18 mutants have sequence polymorphisms that cause significant changes in the predicted Dmca1A protein.

It is still conceivable that this nbA (H18) mutation and the cac one (S) define two different functions (or even genes); but the fact that both are mutated in the same ORF, which encodes a protein that can be considered highly relevant to both song-ralated and visual-system functions (see below), increases the weight of evidence in favor of these two different kinds of mutants having identified the same molecular-genetic entity. Furthermore, the phenotypic interrelatedness of cac and nbA-defined functions has been boosted by elements of the current results. Thus, several genetic combinations involving cac^P73 (originally isolated on the basis of visual defects) give reduced courtship-song pulse amplitude, and several genotypes involving the cac^S mutation (identified initially with respect to an anomalous courtship song) have subtle ERG defects. The coupling of courtship-song and visual phenotypes, previously thought to be strictly separated between mutually exclusive classes of these interacting mutants, further suggests that all these mutations are allelic. We conclude that the cac gene encodes the Dmca1A calcium channel 1 subunit protein, that this protein is an important factor mediating behaviorally-related functions of excitable cells, and that mutations in this gene are responsible for the various phenotypes of cac mutants.

Calcium-channel function in the generation of courtship song:
The cac^S mutation does not lead to pathological abnormalities in the courtship song, but causes quantitative changes in elements of the song (KULKARNI and HALL 1987 ; WHEELER et al. 1989 ), leaving these acoustical signals nicely patterned. The cac^S mutation causes analogous (meaning nonpathological) changes in visual system physiology: it reduces the amplitude of the ERG transients (but does not eliminate them) and causes a novel but low amplitude aberration in the ERG. Also, the cac^S/cac^L-6 heteroallelic combination has defective Y-tube phototaxis. The cac^S mutation is in Dmca1A exon 19; this exon seems not to be subject to alternative splicing (PEIXOTO et al. 1997 ), so it would be expected to be included in all products expressed from the cac locus. These results imply that the cac^S mutation damages Dmca1A channel functions common to physiological processes underlying the generation of courtship song (Figure 5) and of a normal ERG, but not sufficiently (except in the one case just noted) to disrupt visually-mediated behaviors.

Mutations in ion-channel genes have often been associated with temperature-sensitive phenotypes such as paralysis (e.g., WU and GANETZKY 1992 ; GANETZKY 1996). In this context, an additional cacophony phenotype, temperature-sensitive convulsions, was recently discovered to be a feature of the "song allele" cac^S but not the "vision alleles" cac^H18 and cac^EE171 (PEIXOTO and HALL 1998 ). This raised the question of whether other temperature-sensitive ion-channel mutants would sing abnormally. Indeed, mutations at the slowpoke (slo) locus in D. melanogaster, which encodes a calcium-activated potassium channel (ATKINSON et al. 1991 ) cause severe song defects (PEIXOTO and HALL 1998 ). Calcium and potassium currents are involved in the function of pacemaker cells (HILLE 1992 ). Figure 5 presents a simple, speculative scheme of how the products of cac and slo could work together to form a pacemaker that would underlie the tone-pulse component of Drosophila's courtship song.

The lovesongs of these flies are thought to be involved in species recognition as well as stimulation of females to copulate and hence are hypothesized to be a component of prezygotic isolation during speciation (COYNE 1992 ; HALL 1994 ). It is intriguing that changes in courtship song caused by the cac^S mutation—specifically affecting the number of cycles per pulse, the interpulse interval, and the pulse amplitude (Figure 5)—are similar to the differences in song between several closely related species. The quasi-separability of cac phenotypes implies that it might be possible to "tune" the courtship song with relatively small evolutionary changes in this one gene. It will be interesting to examine the homologous channel protein from species known to differ in their song components (reviewed in HALL 1994 ) for differences that might contribute to evolutionary divergences of these species signatures. A gene such as slo may harbor song-related interspecific variations as well.

A phenylalanine (analogous to the cac^S-mutated residue) in the transmembrane domain of the minK subunit of the I_sK potassium channel (WANG et al. 1996 ) has been subjected to in vitro mutagenesis (WILSON et al. 1994 ). This protein change (F to C at residue 57) led to an I_sK potassium current that was normal in terms of half-maximal activation voltage and effect on membrane potential; that is, indistinguishable in these parameters from I_sK currents stemming from the expression of molecularly unaltered transcripts. However, the evolutionary and structural divergence of the minK protein makes it difficult to extrapolate these results to the Dmca1A calcium channel. Our results draw attention to the role of this transmembrane region (IIIS6)—and in particular to the cac-defined phenylalanine within it, which is now accessible for electrophysiological bioassay via analysis of calcium currents in cac