An Analysis of Transvection at the yellow Locus of Drosophila melanogaster

Corresponding author: Chao-ting Wu, Department of Genetics, Harvard Medical School, 200 Longwood Ave., Boston, MA 02115., twu{at}rascal.med.harvard.edu (E-mail)

Communicating editor: J. A. BIRCHLER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Studies of a wide variety of organisms have shown that homologous sequences can exert a significant impact on each other, resulting in changes in gene sequence, gene expression, chromatin structure, and global chromosome architecture. Our work has focused on transvection, a process that can cause genes to be sensitive to the proximity of a homologue. Transvection is seen at the yellow gene of Drosophila, where it mediates numerous cases of intragenic complementation. In this article, we describe two approaches that have characterized the process of transvection at yellow. The first entailed a screen for mutations that support intragenic complementation at yellow. The second involved the analysis of 53 yellow alleles, obtained from a variety of sources, with respect to complementation, molecular structure, and transcriptional competence. Our data suggest two ways in which transvection may be regulated at yellow: (1) a transcriptional mechanism, whereby the ability of an allele to support transvection is influenced by its transcriptional competency, and (2) a structural mechanism, whereby the pairing of structurally dissimilar homologues results in conformational changes that affect gene expression.

THE structure and function of a segment of DNA can be profoundly affected by the presence of homologous sequences (reviewed by BESTOR et al. 1994 ; HENIKOFF and COMAI 1998 ). In some cases, the impact of homologous sequences is revealed through the introduction of transgenes bearing sequence homology to endogenous genes. Such transgenes can induce sequence alterations and de novo methylation in fungi (reviewed by SELKER 1997 ) and gene silencing in plants (reviewed by MEYER and SAEDLER 1996 ; DEPICKER and VAN MONTAGU 1997 ; JORGENSEN et al. 1998 ; MATZKE and MATZKE 1998 ), insects (DORER and HENIKOFF 1994 ; PAL-BHADRA et al. 1997 ), and mammals (GARRICK et al. 1998 ). Diploidy and polyploidy provide more natural states where genes are present in multiple copies, and in these contexts, homologous sequences can also influence gene expression. Here, examples are provided by paramutation in plants (reviewed by HOLLICK et al. 1997 ), imprinting in mammals (reviewed by LALANDE 1996 ; BARTOLOMEI and TILGHMAN 1997 ; REIK and WALTER 1998 ), and sex chromosome and autosomal dosage compensation (reviewed by CLINE and MEYER 1996 ; HEARD et al. 1997 ; LUCCHESI 1998 ; also HIEBERT and BIRCHLER 1994 ; BIRCHLER 1996 ). In fact, ploidy itself can alter gene expression (MITTELSTEN SCHEID et al. 1996 and references therein). Taken together, these "homology effects" attest to a diverse set of mechanisms whose evolution may have been in response to the appearance of homologous sequences, such as must have happened during transitions from haploidy to higher ploidies (for related interpretations, see BESTOR 1990 ; FLAVELL 1994 ; BESTOR and TYCKO 1996 ; MATZKE et al. 1996 ; MITTELSTEN SCHEID et al. 1996 ; BINGHAM 1997 ; HENIKOFF and MATZKE 1997 ; YODER et al. 1997 ; JORGENSEN et al. 1998 ; KUMPATLA et al. 1998 ; MATZKE and MATZKE 1998 ; references therein). In this article, we explore the manner in which homologous sequences can interact by examining the phenomenon of transvection.

In Drosophila, the state of diploidy is accompanied by the feature of somatic homologue pairing (METZ 1916 ; FUNG et al. 1998 and references therein). The expression of some genes can be modulated by proximity to a homologue, and these genes are said to exhibit transvection effects (LEWIS 1954 ). While transvection was defined in Drosophila, there is evidence for its occurrence, and the occurrence of related phenomena, elsewhere (JUDD 1988 ; TARTOF and HENIKOFF 1991 ; WU 1993 ; HENIKOFF 1997 ; HENIKOFF and COMAI 1998 ; references therein). Recent studies in a wide variety of systems further suggest that the impact of homologue pairing may be quite broad. For example, homologue pairing may influence DNA accessibility in yeast (KEENEY and KLECKNER 1996 ), may be associated with the process of parental imprinting in mammals (LASALLE and LALANDE 1996 ), may allow nonintegrated plasmids to transinduce a chromosomal gene in mammalian cell lines (ASHE et al. 1997 ), and may play a general role in methylation transfer (for example, BESTOR and TYCKO 1996 ; COLOT et al. 1996 ; FORNE et al. 1997 ). The mechanisms by which homologous sequences influence each other are not well understood, and many models have been proposed (reviewed by HENIKOFF 1997 ; HENIKOFF and COMAI 1998 ; references therein; also see PETERSON et al. 1994 ; JUDD 1995 ; GOLDSBOROUGH and KORNBERG 1996 ; DONALDSON and KARPEN 1997 ; MORRIS et al. 1998 ; SIPOS et al. 1998 ). Some of the underlying mechanisms may be common among different processes, while others may be process, locus, or allele specific.

We have asked how the proximity of homologous genes can influence gene expression and have focused our attention on the Drosophila X-linked yellow gene (y, 1–0.0), which shows transvection effects (GEYER et al. 1990 ). The yellow gene is required for pigmentation of larval and adult cuticular structures and its pattern of expression is under the control of tissue-specific enhancers located in the upstream 5' regulatory region and in the single intron (GEYER and CORCES 1987 ; MARTIN et al. 1989 ). Interestingly, there exist several recessive mutant alleles which, on their own, reduce pigmentation in wings and body, but which, in certain combinations, complement each other to give flies with nearly wild-type pigmentation in these structures (STONE 1935 ; FRYE 1960 ; GREEN 1961 ; NASH 1976 ; GEYER et al. 1990 ). These cases of intragenic complementation can be explained by transvection (GEYER et al. 1990 ).

One model for transvection at yellow suggests that enhancers of one allele act in trans on the promoter of the other allele when the two alleles are in close proximity (GEYER et al. 1990 ; MORRIS et al. 1998 ). This model can be illustrated by the complementation seen between the y² and y^1#8 alleles (Figure 1A; GEYER et al. 1990 ). The y² allele produces flies with mutant pigmentation of the wings and body, but wild-type pigmentation of other structures. The mutant phenotype is caused by the insertion of a gypsy retrotransposon between the promoter and the two upstream enhancers responsible for pigmentation of the wings and body (reviewed by CORCES and GEYER 1991 ). The gypsy element inhibits communication between these enhancers and the promoter because it has binding sites for the suppressor of Hairy-wing [su(Hw)] protein, which, when bound, establishes a chromatin insulator that prevents the enhancers on one side from communicating productively with the promoter on the other side (reviewed by DORSETT 1996 ; GDULA et al. 1996 ; GERASIMOVA and CORCES 1996 ; GEYER 1997 ). The y^1#8 allele causes fully mutant pigmentation of all larval and adult structures (GEYER et al. 1990 ). It is a derivative of a P-element allele and consists of a 0.8-kbp deletion that removes the promoter and retains minimal P-element sequences. Although neither y² nor y^1#8 flies show significant wing or body pigmentation, these alleles complement each other and y²/y^1#8 flies show nearly wild-type wing and body pigmentation. This intragenic complementation can be explained by the action of the wing and body enhancers of y^1#8 in trans on the y² promoter when the two alleles are in close proximity (Figure 1A).

View larger version (20K): In this window In a new window Download PPT slide   Figure 1. Model for trans-acting enhancers and regulation by cis-preference (GEYER et al. 1990 ). (A) The y1#8 allele complements y2 by the trans action of its enhancers. (B) The y1 allele does not complement y2. W, wing enhancer; B, body enhancer; Br, bristle enhancer; T, tarsal claw enhancer; black rectangles, exons; arrow above first exon, transcription in wings and body, not intended to indicate status of transcription in other tissues; white rectangle, gypsy long-terminal repeat; arrow above white rectangle, transcription of gypsy; stippled square, su(Hw) binding region.

Transvection is a regulated process at the yellow gene. Some alleles with intact wing and body enhancers, such as y^1#8, complement y², while others that carry these enhancers do not. For example, the y¹ allele, which has intact wing and body enhancers, does not complement y² (Figure 1B; GEYER et al. 1990 ). This allele is an A to C transversion in the ATG translation initiation codon (GEYER et al. 1990 ). As the y² promoter is the only promoter in the y²/y^1#8 and y²/y¹ genotypes that can give rise to functional transcripts, issues regarding the control of transvection can be reduced in these cases to asking why the y² promoter is activated to complementing levels in the former, but not the latter, genotype. Differences in the number of wing and body enhancers present cannot account for the differences in complementation because both genotypes are identical in this respect.

This article is concerned with the manner in which transvection is regulated at yellow. It has been proposed that a prerequisite for the trans action of an enhancer at yellow is the disruption of its own promoter in cis (GEYER et al. 1990 ). For example, the release of the enhancers of y^1#8 to act upon the y² promoter may result from deletion of the promoter of y^1#8. By contrast, the intact promoter of y¹ may preclude such trans interactions. Cis-preference of regulatory elements for their own promoter may be a general modulator of transvection because correlations between changes in the promoter region and the ability of a gene to support transvection have also been noted at Ultrabithorax (Ubx, MARTINEZ-LABORDA et al. 1992 ; CASARES et al. 1997 ) and Abdominal-B (Abd-B, SIPOS et al. 1998 ).

Our studies have extended the promoter-based model. We began by asking whether there are regions outside the promoter that control transvection. Evidence for such regions exists at Abd-B (HENDRICKSON and SAKONJU 1995 ; HOPMANN et al. 1995 ; SIPOS et al. 1998 ). We took two experimental approaches. Importantly, neither was biased toward the promoter. First, we carried out mutageneses to identify elements that control transvection. Second, we characterized 53 yellow alleles, obtained from a variety of sources, according to their complementation patterns, molecular structures, and transcriptional competence. The data emphasize that the promoter region is important in the regulation of transvection and suggest two models for how this region may exert its influence.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Drosophila stocks:
Mutations not listed in Table 1 are described in LINDSLEY and ZIMM 1992 . The chromosome bearing y^3c3 carries a new allele of echinus and the following X chromosomes also carry markers in addition to those at yellow (a superscript ? denotes ambiguity in the identity of the allele): y^1like sc^? w^a, In(1)FM7, y^31d sc⁸ w^a B, y^50k w^bf61d5, y^62a sc^? cv, y⁶⁹ w spl sn³, y^76-1 ras² v m, y^76-2 su(w^a) w^a, y^76-3 f^s su(f), y^79h ac, y^86K1.1 sc^? w^bl, y²⁰¹ sc^? w^a sd^88g22, y⁴¹⁷¹ sn³ lz^50e, y^a77 ras^a77, y^bg ct⁶ car, y^c4 sc^S1 B, y^Ky w^k. The Df(1)y^- ac^- w¹¹¹⁸ chromosome lacks all yellow sequences (GEYER and CORCES 1987 ).

Culture conditions:
Flies were cultured at 25° ± 1° on standard Drosophila cornmeal, yeast, sugar, and agar medium with p-hydroxybenzoic acid methyl ester as a mold inhibitor. In general, three females were mated with three or more males in vials and brooded daily. Temperature and crowding were carefully monitored as both affect pigmentation.

Mutageneses:
Mutageneses using ethylmethane sulfonate (EMS, M0880; Sigma, St. Louis) and diepoxybutane (DEB, Sigma D7019) were performed as follows: males were collected and aged 2–4 days, desiccated 12 hr, fed 25 mM EMS or 5 mM or 7.5 mM DEB in 10% sucrose for 24 hr, allowed to recover on standard food for 6 hr, and then mated overnight to females that had been aged 2–4 days. Males were then discarded and the females placed in bottles in groups of 15–25. Bottles were brooded every day or every other day. X-ray mutageneses were done similarly except that males were not desiccated and males and females were allowed to mate for up to 4 days. Males were irradiated with 3800 or 4000 rads of X rays using a Phillips X-ray machine at a dose of 465 rad/min.

The screen for complementing alleles produced 1 complementing allele and 33 derivatives of y² of which 3 were obtained from mutageneses using EMS, 14 from mutageneses using DEB, and 16 from mutageneses using X rays. The EMS control screen produced 24 F₁ flies showing mutant pigmentation. One was fully mutant and transmitted a null phenotype. The remaining 23 had a patterned phenotype in that some structures showed mutant pigmentation while others were wild type. Of these, 7 transmitted a null phenotype, 2 transmitted patterned phenotypes, 4 did not transmit the mutant phenotype, 7 were sterile, 2 died before mating, and 1 was due to a mutation extragenic to yellow. The DEB control screen produced 5 F₁ flies with a patterned phenotype. Of these, 1 transmitted a null phenotype, 1 did not transmit a mutant phenotype, 2 were sterile, and 1 was due to a mutation extragenic to yellow.

Genetic analysis of y^3c3:
Confirmation that y^3c3 is an allele of yellow was carried out with linkage and recombinational studies as described below; complementation analysis was not useful because y^3c3 was induced in a y¹ background. In the following discussion, the asterisk represents the lesion responsible for complementation. Linkage to the second, third, and fourth chromosomes was ruled out by mating y²/y¹ * females to y²/Y; Bl vg/CyO or y²/Y; st Ch^V red Tb/TM3 or y²/Y; ey^D/ci^D males, collecting complementing females carrying the CyO second chromosome balancer, the TM3 third chromosome balancer, or ey^D, mating these females to y²/Y males, and noting random segregation of the complementing phenotype with respect to the balancers and ey^D. Linkage to the X chromosome as well as to y¹ was revealed by singly mating 10 male progeny, showing the y¹ phenotype, of complementing y²/y¹ * females to y²/y² females and noting that all female progeny showed the complementing phenotype. Recombinational mapping then placed the lesion responsible for complementation between sc at 0.0 and w^h at 1.5, which is the interval within which yellow lies. In these experiments, y¹ */y¹ * females were mated to w^h cv/Y or sc ec cv ct⁶ v g² f/Y males, and y¹ */w^h cv and y¹ */sc ec cv ct⁶ v g² f females were collected in the next generation. These females were mated to wild-type males and 6 recombinant y¹ (*) w^h cv/Y and 13 recombinant y¹ (*) cv ct⁶ v g² f/Y male progeny were recovered and singly mated to y²/y² females. All female progeny showed complementation. As the degree of complementation observed with these recombinant chromosomes was identical to that seen with the original y^3c3 chromosome, our data indicate that the complementing phenotype is due to a single hit, most likely at yellow. Subsequent molecular analysis proved y^3c3 to be associated with a lesion within yellow.

Pigmentation scores and complementation tests:
Pigmentation phenotypes were determined by examining 1- to 3-day-old flies on a white cold stage (WU and HOWE 1995 ). The scores were determined independently by two people and based upon at least 100 females from each of two or more independent crosses. A five-point pigmentation scale (see RESULTS) was used for the wings and body, where "wings" refers to the wing blades and not the wing bristles, and "body" refers to the abdominal stripes and not the interstripe abdominal cuticle or the thoracic cuticle. This scale was developed independently of scales described in other studies, and therefore comparisons between scales should not be made. There can be small differences in pigmentation among flies scored at a particular value.

Southern analysis:
Genomic DNA was isolated from adult flies as described by ASHBURNER 1989A , digested with restriction enzymes (Boehringer Mannheim, Indianapolis), and separated on a 0.8% agarose gel. Subsequent steps were performed as described by SAMBROOK et al. 1989 . GeneScreen hybridization membranes were used according to the recommendation of the manufacturer (NEN) and probe DNA was labeled with the Random Primed DNA Labeling Kit (Boehringer Mannheim).

Reverse-transcriptase-PCR assay:
Total RNA was isolated using the Ultraspec RNA Isolation System (Biotecx, Houston) from homozygous and/or hemizygous 6- to 8-day-old animals at the pupal stage. Approximately 2 µg of total RNA was used in a 20-µl reverse transcriptase (RT) reaction using oligo(dT) as a primer and SuperScript RT (Gibco/BRL, Gaithersburg, MD). PCR was done using 2 µl from the reverse transcriptase reaction in a 50-µl reaction using Taq DNA polymerase (Boehringer Mannheim). PCR parameters were 93° for 5 min, 30 cycles of 94° for 30 sec, 70° for 30 sec, and 72° for 1 min, followed by an extension period of 72° for 5 min. All alleles were analyzed using two sets of primers in separate reactions. Two different forward primers in the first exon and a common reverse primer in the second exon were used. The first forward primer had the sequence 5'AGCCGAAGGCTAGAGAAGAACCCCCTATAGCTG beginning at +51 and the second forward primer had the sequence 5'CAGCTTAGAGCTAAGTGCAATGTTCC beginning at +153. The reverse primer had the sequence 5'CATCCACTTTAATGCGGTAGGCAGTGGTAA beginning at +3272. Wild-type RNA produces a product of 503 bp when the first forward primer is used and a product of 401 bp when the second forward primer is used. The y⁴¹⁷¹ allele failed to give a product even though it produces pigmentation in bristles, and the y^Ky and y^RBK alleles yielded a faint product of the expected size and a darker product that was slightly smaller in size regardless of which primer pair was used. We do not have explanations for these results. The y^m1 allele is associated with a deletion that is likely to remove the region where the reverse primer anneals (data not shown), and therefore its analysis was carried out with another reverse primer, 5'CGTTGTGCTGGTTGAAAATATAGGC, beginning at +4379. In combination with either forward primer, this reverse primer yielded products of sizes that were predicted by Southern analysis.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A screen for mutations that promote transvection at yellow:
The proposal that the yellow promoter region controls transvection was based on a study indicating that alleles capable of complementing y² are also transcriptionally compromised (GEYER et al. 1990 ). The data, however, could not clearly define the features of the promoter region that are critical for promoting transvection. All alleles that supported transvection, such as y^1#8, carried a deletion of, or an insertion into, the promoter region, leaving open the question of whether the critical aspect of these alleles is the disruption of transcription, the disruption of structural integrity regardless of the consequences on transcription, or the alteration of elements that are near to, but separable from, the promoter. In addition, the complementing alleles may represent only a subset of lesions capable of supporting transvection because they had all been isolated on the basis of their mutant pigmentation phenotypes and not because of their ability to support transvection. For example, it may be possible to generate lesions that can support transvection without interfering with pigmentation, but these would not be expected to be among the alleles studied. Finally, interpretations of transvection based on the alleles used are complicated by the association of the alleles with foreign sequences. Foreign sequences may bring in heterologous promoters, DNA-binding proteins, and other associated factors.

To address these issues, we undertook a mutagenic screen to recover new yellow alleles entirely on the basis of their ability to support transvection. Specifically, we screened for mutations that promote complementation between two otherwise noncomplementing yellow alleles. Importantly, our screen did not depend on the new mutations having a mutant yellow phenotype, had no presumed bias for the promoter region, and did not make use of transposable elements.

The two noncomplementing alleles chosen for the screen were y² and y¹ (Figure 1B and Figure 2A). We mutagenized y¹/Y males, mated them to y²/y² females, and examined the y²/y¹ progeny for complementation as indicated by increased wing and/or body pigmentation (Figure 3A). We wished to optimize our ability to generate both structurally normal alleles, such as point mutations, as well as structurally altered alleles. To this end, some mutageneses were carried out with EMS, which generally causes point mutations, while others were carried out with DEB or X-irradiation, both of which can produce small deletions (ASHBURNER 1989B ).

View larger version (141K): In this window In a new window Download PPT slide   Figure 2. Complementing and noncomplementing phenotypes. (A) The y1 allele does not complement y2, and the wings and body of the fly show an extreme mutant pigmentation. (B) The y3c3 allele complements y2 to give a fly with nearly wild-type wing and body pigmentation. (C) The y59b allele complements y2 to a degree that is comparable to that seen in the y2/y3c3 fly.

View larger version (11K): In this window In a new window Download PPT slide   Figure 3. Screens. (A) Scheme used to isolate mutations that complement y2. (B) Scheme used to isolate mutations that disrupt yellow expression. (*), Mutagenized chromosome.

The y² allele was used because it has been extensively characterized and appears in all well-documented cases of intragenic complementation at yellow (STONE 1935 ; FRYE 1960 ; GREEN 1961 ; NASH 1976 ; GEYER et al. 1990 ). On a scale of 1 to 5, where 1 represents the null or nearly null state and 5 the wild-type or nearly wild-type state with respect to pigmentation, y²/y² females and y²/Y males give scores of 1 in the wings and 1–2 in the body, and hemizygous y²/Df(1)y^- ac^- w¹¹¹⁸ females give scores of 1 in the wings and body.

The y¹ allele was chosen for three reasons. Because the exceptions in our screen were to be identified by their ability to complement y² as revealed by increased pigmentation, we wanted to begin with an allele that does not produce pigmentation. The y¹ allele met this requirement. It is a null allele (GEYER et al. 1990 ) that produces scores of 1 in wings and body when homozygous or hemizygous, as well as when heterozygous with y². Then, because we wanted to address the possible role of gene structure in the region of the promoter, we required a structurally normal allele. The y¹ allele, being a single base pair change, also satisfied this criterion. Finally, to address the possibility that disruption of transcription is a prerequisite for transvection, we sought an allele that is not compromised for transcription. Again, the y¹ allele met this requirement because it supports wild-type levels of transcription (GEYER et al. 1990 ).

One complementing yellow allele was isolated:
Two types of mutations were recovered from 89,000 EMS-treated, 78,000 DEB-treated, and 91,000 X-ray-treated X chromosomes. First, we recovered one EMS-generated derivative of y¹. This allele, called y^3c3, is a 3.6-kbp intragenic deletion that removes the promoter region (MORRIS et al. 1998 ). Flies homozygous for y^3c3 and those bearing y^3c3 in trans to y¹ or Df(1)y^- ac^- w¹¹¹⁸ are fully mutant in wings and body and produce scores of 1 in both tissues. In contrast, flies bearing y^3c3 in trans to y² show a nearly wild-type phenotype (Figure 2B); y²/y^3c3 females produce scores of 4 in wings and body. The second class of mutations consisted of 33 partial and complete phenotypic revertants of y². These derivatives of y² were obtained in all three mutageneses and presumably were spontaneous in nature as y² was on the nonmutagenized X chromosome.

We wished to compare the rate of generating a complementing allele to the rate of generating alleles that alter yellow pigmentation. To this end, we conducted control screens in which alleles were isolated based on their mutant pigmentation (Figure 3B). These screens were modeled after an EMS mutagenesis described by NASH 1976 . In our study, wild-type y⁺/Y males were mutagenized with EMS or DEB and mated to FMA3/Y females, where FMA3 is an attached-X chromosome. The F₁ patriclinous males were then screened for mutant pigmentation.

Our control screens produced 10 yellow alleles from a total of 28,300 EMS-treated X chromosomes and 1 yellow allele from a total of 18,600 DEB-treated X chromosomes (Table 1). None of these alleles complemented y². Our data are comparable to those of NASH 1976 who screened 80,000 EMS-treated X chromosomes for mutations giving a mutant pigmentation phenotype and recovered 40 yellow alleles, none of which complemented y². These data indicate that the rate of isolating EMS-induced yellow alleles based on their mutant pigmentation phenotype alone is significantly greater than that of isolating EMS-induced yellow alleles based on their ability to complement y². Our data from mutageneses using DEB show a similar trend.

Complementation analyses of a collection of yellow alleles defined four classes of alleles:
The second approach that we took in our analysis of the regulation of transvection involved the characterization of 53 yellow alleles, including the alleles generated in this study (Table 1). We classified these alleles according to their ability to promote complementation and then carried out structural and transcriptional analyses (Table 2 Table 3 Table 4). The ability of each allele to support transvection was assessed by determining its ability to complement two yellow alleles for wing and body pigmentation. These two alleles, y² and y^59b, complement each other (Figure 2C). The y^59b allele is a derivative of y² that lacks the yellow promoter and part of gypsy, including the su(Hw) binding sites (GEYER et al. 1990 ). It is a null allele and produces scores of 1 for wings and body in homozygous females, hemizygous males, and hemizygous females. Females heterozygous for y² and y^59b give scores of 4 in wings and body. For the complementation tests, two alleles were considered to complement each other if, as a trans-heterozygote, scores in the wings or body were at least 1 point darker on the pigmentation scale than the scores of analogous tissue in females homozygous for either allele (criterion 1), females heterozygous for either allele and y¹ (criterion 2), and females heterozygous for either allele and Df(1)y^- ac^- w¹¹¹⁸ (criterion 3). In a few cases, the phenotype of females homozygous for the yellow allele could not be determined because homozygosity for the X chromosome bearing the yellow allele resulted in lethality. In these cases, satisfaction of criteria 2 and 3 constituted a positive score for complementation.

The complementation tests with y² and y^59b defined four classes of alleles (Table 2 and Table 3). Class A alleles complement y^59b but do not complement y². There are three members of this class, including y², which is the prototype. Class B alleles complement y² but do not complement y^59b. There are 17 alleles in this class, including y^59b, which is the prototype. Class C alleles complement neither y² nor y^59b. There are 27 Class C alleles, and y¹ is their prototype. The Class C alleles are further divided into two subgroups, C1 consisting of 24 alleles, and C2 consisting of 3 alleles. Justification for this subdivision rests upon additional complementation data that are discussed at a later point. The fourth class of alleles, Class D, was not expected because the two members of this class complement both y² and y^59b. Finally, 4 alleles could not be classified because their pigmentation in the wings and body approaches wild-type levels, making complementation tests uninformative.

In some cases, the number of alleles in a class may be an overestimate. Complementation tests and structural and transcriptional analyses (below) show striking resemblance between several alleles (noted in Table 1), with small differences in complementation possibly attributable to differences in genetic background. In fact, some alleles that are similar to each other have names that are also similar. Examples of such alleles are the two Class D alleles, y²³⁶⁴ and y²³⁷⁴. In spite of these similarities, we have considered all alleles separately because it has not been possible to resolve ambiguities in nomenclature.

After the four classes of alleles were defined by complementation analyses, we tested the generality of the groupings by crossing the two other Class A alleles and the two Class D alleles to the entire set of alleles. The results of these tests are presented in Table 4. We found that the Class A y², y^62a, and y^82f29 alleles do not complement each other in any pairwise combination. Furthermore, we found that y^62a and y^82f29 behave for the most part like y². They complement the majority, although not all, of the Class B alleles (cases of noncomplementation noted c in Table 4), and fail to complement all of the Class C (C1 and C2) alleles. This similarity of the three Class A alleles validates our classification system. It argues that the four classes of yellow alleles define common features within each group and that these features are relevant to transvection.

The two Class D alleles, y²³⁶⁴ and y²³⁷⁴, also lend support to our classification system. Class D alleles complement both y² and y^59b. By their ability to complement y^59b, Class D alleles resemble Class A alleles, and additional complementation tests revealed the resemblance to be extensive. Just as all pairwise combinations among the Class A alleles failed to show complementation, all pairwise combinations among the Class A and D alleles, with one exception, failed to show complementation. The exceptional circumstance, of course, is the ability of Class D alleles to complement y². Furthermore, Class D alleles complemented all Class B alleles and failed to complement all C1 alleles. In contrast to the Class A alleles, however, the Class D alleles complemented the three C2 alleles. In fact, it was this complementation that defined the C2 Class. In sum, the Class D alleles resemble the Class A alleles except in two situations: Class D alleles complement y² from among the Class A alleles and C2 alleles from among the Class C alleles. These situations have proven informative (below).

All Class A alleles appear structurally normal in the promoter region and are transcriptionally competent:
After classifying the yellow alleles by complementation analysis, we characterized them with respect to structure and transcriptional competence to find the signature features that would unify members of one class and also distinguish them from members of other classes. The promoter-based model for the regulation of transvection predicts that the status of the promoter region should correlate with the classification of a given allele. For example, alleles that can complement y² would be expected by the model to be disrupted in the region of the promoter and transcriptionally compromised.

The structural studies involved Southern analysis of genomic DNA digested with BamHI and HindIII and probed with sequences complementary to the entire yellow gene (Figure 4). The 0.7-kbp fragment produced in this protocol contains the minimal promoter (GEYER and CORCES 1987 ). The transcriptional capacity of the alleles was assessed in two ways. We visually determined the level of pigmentation in the wings, body, bristles, and aristae of homozygous females and hemizygous males, and considered the presence of pigmentation in any structure as evidence for a functional promoter. Alleles with a functional promoter were scored as transcriptionally competent. We also carried out RT-PCR assays (MATERIALS AND METHODS) for yellow transcripts using RNA obtained from the pupal stage, the time of maximal yellow transcription. An allele was judged to have at least a minimally functional promoter if RT-PCR analysis produced bands of the expected sizes.

View larger version (51K): In this window In a new window Download PPT slide   Figure 4. Top, Southern analysis of yellow alleles used in this study. Genomic DNA was digested with BamHI and HindIII. The Southern pattern for y25 is identical to that of y2S (data not shown). The Southern patterns for ym15, ym16, ym19, ym24, ym25, ym27, ym32, ym33, and ym36 are identical to that of y+ (data not shown). The Southern pattern for y2364 is identical to that of y2374 (data not shown). Arrow, 0.7-kbp band encompassing the promoter region; (*) 4.1-kbp band common to alleles that are structurally similar to y2. Bottom, map of the wild-type yellow gene indicating extent of the probe and location of the gypsy insertion present in y2. Symbols as described in legend to Figure 1; B, BamHI; G, BglII; H, HindIII; S, SalI.

We first consider the three alleles of Class A. As these alleles complement y^59b, which lacks a promoter, we predicted that the promoter of each Class A allele would be intact and functional. This is the case. Although each allele is associated with structural changes, the promoter region remains intact. Southern analysis of each allele produced the 0.7-kbp BamHI-HindIII promoter band (Figure 4; Table 3). Furthermore, all three alleles are transcriptionally competent. They produce pigmented structures and give rise to the expected bands when characterized by RT-PCR (Table 3).

Class B alleles are all structurally altered in the promoter region but are heterogeneous with regard to transcriptional competence:
All 17 Class B alleles are associated with structural alterations in the promoter fragment. For 13 of these alleles, the promoter band is the only band that is altered in Southern analyses (Figure 4; Table 3). In contrast, the 17 Class B alleles are heterogeneous with respect to transcriptional competence (Table 3). Four gave no evidence of transcription either by the visual assay or by RT-PCR, while the remaining 13 produced at least some level of pigmentation in one or more structures. Of these 13, 12 yielded the expected bands by RT-PCR.

All except one allele of Class C1 appear structurally normal in the promoter region and are transcriptionally competent:
The 24 C1 alleles are strikingly different in structure from the alleles of Class B. All but one, y^76-1, appear structurally normal in the region of the promoter, and only 8 of the remaining alleles show rearrangements elsewhere (Figure 4; Table 3). Consistent with these data, all C1 alleles except y^76-1 are also transcriptionally competent as assayed by RT-PCR (Table 3). Six of these also scored positively in our visual assay. These results suggest that the majority of C1 alleles are mutant in pigmentation because of lesions that affect post-transcriptional processes.

C2 alleles are related in structure to y² and are heterogeneous with respect to transcription:
The C2 alleles distinguish themselves from the C1 alleles by their ability to complement the Class D alleles. One C2 allele, y⁶⁹, is a derivative of y² that retains the entire su(Hw) binding region but lacks the yellow promoter and coding region (GEYER et al. 1990 ). As expected, y⁶⁹ is transcriptionally silent (Table 3). Our data show that the other two C2 alleles, y^1like and y²⁰¹, are also structurally related to y².

The y²⁰¹ allele is structurally disrupted in the promoter region and transcriptionally silent in our assays, while y^1like is structurally unaltered in the promoter region and transcriptionally competent (Figure 4; Table 3). Interestingly, Southern analyses showed that both y^1like and y²⁰¹ lack the wild-type 1.3-kbp HindIII fragment and instead have a 4.1-kbp HindIII fragment that is characteristic of the gypsy insertion in y² and y⁶⁹ (Figure 4; Table 3). This observation suggested that a gypsy element is inserted in the same position in all four alleles. We tested this interpretation using PCR analysis with primers homologous to sequences in yellow and gypsy. Both y^1like and y²⁰¹ gave a band that was similar in size to that of a band produced by both y² and y⁶⁹ (data not shown). Sequence analysis of these PCR products confirmed the presence of gypsy sequences at -700, which is the position at which gypsy is inserted in y² and y⁶⁹.

Class D alleles are structurally altered in the region of the promoter but remain transcriptionally competent:
The Class D alleles are transcriptionally competent in that they produce pigmented structures and give rise to the expected bands when characterized by RT-PCR (Table 3). They are also disrupted in the promoter region and fail to produce the promoter band (Figure 4; Table 3). In these ways, they possess characteristics resembling those of both the Class A and B alleles.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A screen for mutations that support transvection points to the promoter region as a key regulator of transvection:
Our first approach to the study of the regulation of transvection involved a screen for mutations that allow complementation between y¹ and y². A single mutation, y^3c3, was recovered, and the association of y^3c3 with an intragenic deletion removing the promoter region supports the promoter-based model for the regulation of transvection. In light of the design of our mutagenesis, which did not require mutations to disrupt yellow gene expression, it is particularly significant that y^3c3 is disrupted in the promoter region. Our results also show that it is more difficult to recover EMS-induced yellow alleles on the basis of their ability to promote intragenic complementation than it is to recover EMS-induced yellow alleles on the basis of pigmentation phenotype. This finding suggests that the elements controlling transvection are relatively rare, small in size, recalcitrant to mutagenesis by EMS, DEB, and X rays, functionally redundant with other elements, or recessive when mutant. Alternatively, lesions that promote transvection may need to satisfy structural requirements that are difficult to meet with the mutagens we used.

Our screen also had the potential for identifying elements that are extragenic to yellow and that, when mutant, can promote transvection at yellow. Extragenic modifiers of yellow expression have been reported (GEORGIEV and GERASIMOVA 1989 ; BEZBORODOVA et al. 1997 ), and our screen had the capacity to identify those that act at the level of transvection. Extragenic elements, however, were not identified. It may be that mutations in such elements are recessive and therefore could not be identified in our screens where only dominant or haplo-insufficient mutations could have been noted.

Complementation studies define four classes of yellow alleles:
Our second approach to studying the regulation of transvection involved the characterization of 53 yellow alleles. These alleles were analyzed for their ability to complement two tester alleles, y² and y^59b, and defined four classes (Table 2 and Table 3). Class A alleles complement y^59b but not y², Class B alleles complement y² but not y^59b, Class C alleles complement neither y² nor y^59b, and Class D alleles complement both y² and y^59b. We addressed the concern that our classifications might reflect choice of tester alleles, rather than the predominant mechanisms of transvection, by determining the ability of all alleles to complement y^62a and y^82f29, two alleles belonging to the class into which y² had been placed, Class A. The results confirmed the generality of our classification scheme. Although we found that choice of tester alleles can influence the classification of some alleles, by and large, the behavior of y^62a and y^82f29 proved strikingly similar to that of y². The observation of multiple alleles in Class A also supports our finding that transvection at yellow does not require the y² allele and that it is an inherent property of the gene and not the special feature of an unusual allele (MORRIS et al. 1998 ). Before these studies, the only reported cases of intragenic complementation at yellow that did not involve y² were those involving the y^3P allele (NASH 1976 ). Published reports of y^3P complementation patterns are conflicting (GREEN 1961 ; NASH 1976 ) and this allele was not available for our study.

Although y^62a and y^82f29 resemble y², they are weaker in their complementation profile. They failed to complement some Class B alleles and, compared to y², frequently produced a lesser degree of complementation (Table 4). It is possible that the weaker phenotype reflects differences in genetic background, in the structures of y², y^62a, and y^82f29, or in the mechanisms of transvection involving these alleles (see MORRIS et al. 1998 regarding y^82f29). On the other hand, the failure to observe complementation with some genotypes could result from technical limitations. For instance, because our tests for complementation rely on visual assays, changes in transcript levels that do not translate into detectable changes in pigmentation would not be noted. This limitation may be aggravated by the association of some Class B alleles with a significant amount of pigmentation that could further compromise our ability to detect complementation. Alternatively, our definition of complementation may be too stringent, leading us to incorrectly judge some complementing genotypes as noncomplementing. In general, we concluded that two alleles complement only if females bearing the two alleles as trans-heterozygotes are darker than all three control genotypes: females homozygous for either allele, females heterozygous for either allele and y¹, or females hemizygous for either allele. However, because transvection involves transcription from just one allele in some situations, comparison with control homozygous females, carrying two doses of an allele, may not always be appropriate. In these cases, a truer test of complementation might be through comparison to control genotypes in which only one allele is expressed, that is, the latter two control genotypes. If we conduct our comparisons in this way, the number of Class B alleles that complement y^62a or y^82f29 increases.

Class D alleles are surprising in their ability to complement both y² and y^59b. By their complementation of y^59b, Class D alleles resemble Class A alleles, and consistent with this interpretation, we found that they complement all Class B alleles. In addition, the Class D alleles failed to complement all but three of the 27 Class C alleles. On the basis of this finding, we divided the Class C alleles into two subgroups, C1 and C2, where C2 alleles complement Class D alleles while C1 alleles do not. Interestingly, Class D alleles also defined two subgroups of Class A alleles in that they complemented y² but failed to complement the other two Class A alleles, y^62a and y^82f29. An explanation for these complementation patterns may lie in the molecular structure of the C2 and y² alleles (below).

The structural and transcriptional profiles of the Class A, B, and C1 alleles suggest two models for the regulation of transvection:
All 3 Class A alleles are transcriptionally competent, and while they are also structurally altered, the alterations fall outside the promoter region as defined by the 0.7-kbp BamHI-HindIII fragment. In contrast, all 17 Class B alleles show structural changes in the region of the promoter. Of these, 4 appear transcriptionally silent, while the remaining 13 support transcription by visual assays or RT-PCR analyses. With respect to structure, C1 alleles are nearly the opposite of Class B alleles. Of the 24 C1 alleles, only 1 shows an alteration by Southern analysis in the region of the promoter. This allele, y^76-1, is also the only C1 allele for which we were unable to obtain a product by RT-PCR.

The structural differences between the Class B and C1 alleles support the proposal that the promoter region plays an important role in the regulation of transvection. Our data support two models. One, a transcriptional model, suggests that the ability of an allele to promote transvection when paired with y² is determined by its transcriptional status and that disruption of transcriptional competency allows transvection by, for example, releasing enhancers to act in trans (Figure 5A; GEYER et al. 1990 ). The second, a structural model, emphasizes the impact of gene structure on transvection and proposes that participation of an allele in transvection can be governed by the structure of its promoter region independent of effects on transcription (Figure 5B).

View larger version (25K): In this window In a new window Download PPT slide   Figure 5. Transcriptional and structural models. (A) Transcriptional model. This model proposes that transcriptional competency regulates transvection at yellow. The circular motif represents a transcriptionally competent promoter complexed with transcriptional machinery and associated factors and can be interpreted to be the preinitiation complex or any factor or factors involved in the later steps of transcription. Absence of the motif indicates compromised transcription. According to this model, y1 does not complement y2 because y1 is fully competent for transcription. In contrast, y1#8 and yh12 can complement y2 because they are transcriptionally compromised. (B) Structural model. This model suggests that structural rearrangement in the region of the promoter facilitates, or may even be a prerequisite for, transvection at yellow. The yellow alleles are drawn as they might appear when paired, with regions of homology shown apposed, and regions of nonhomology shown as loops. According to this model, y1 is unable to complement y2 because the two promoters are fully paired. The vertical lines emphasize pairing in the promoter region. In contrast, alleles that are structurally dissimilar to y2 in the region of the promoter lead to complementation by aligning with y2 in a way that unpairs the y2 promoter. Pairing of y2 with an allele lacking promoter sequences, such as y1#8, results in a looping out of the y2 promoter. Pairing of y2 with an allele carrying an insertion in the promoter region, such as yh12, causes a "breathing" at the insertion site. Symbols as described in legend to Figure 1; triangle, insertion of foreign sequence. The status of transcription from the yh12 yellow promoter in y2/yh12 flies is not known.

The most extreme version of the transcriptional model predicts that all Class B alleles are incapable of supporting any level of transcription. The data do not support this version because 13 of the 17 Class B alleles produce transcripts (Table 3; also GEYER et al. 1990 ). The data also do not support a less extreme version where the promoter of Class B alleles needs only be nonfunctional in tissues exhibiting intragenic complementation. In fact, most of the transcribed Class B alleles are pigmented to some degree in the wings and/or body and therefore produce RNA in these tissues (Table 4). The data, however, are consistent with a transcriptional model proposing that a partial reduction of transcription is sufficient to generate a Class B allele.

Our assessment of the transcriptional model assumes that the RNA produced in transcriptionally competent Class B alleles arises from the yellow promoter. It is possible, however, that the transcripts are driven by cryptic or foreign promoters brought in by transposable elements (for example, GEORGIEV et al. 1997 ). If so, our data would not contradict the transcriptional model. Furthermore, in cases where transvection involves the release of enhancers to act in trans, our observations would suggest that, compared to the wild-type yellow promoter, the cryptic or foreign promoters of a Class B allele are less able to capture cis enhancers (also CASARES et al. 1997 ).

The second model proposes that disruption of structural integrity in the region of the promoter is a key determinant in the generation of a Class B allele. The data are consistent with this model in that all Class B alleles are structurally altered in the promoter region (Figure 4; Table 3). How might structure play a role in the regulation of transvection? One explanation evokes pairing-mediated topology effects (TOPEs) and suggests that structural alterations of the promoter region arising from the pairing of dissimilar alleles can influence gene expression. Figure 5B illustrates theoretical paired structures for one noncomplementing genotype and two complementing genotypes. In the noncomplementing y²/y¹ genotype, the promoter regions of both alleles are intact and completely paired. In contrast, uniform pairing throughout the promoter regions is not possible for the two complementing genotypes. In the case of y^1#8, the deletion causes an unpairing and looping out of the y² promoter region. In the case of y^h12, which carries an insertion in the promoter region (GEYER et al. 1990 ; Table 5), the presence of a foreign sequence creates a region of nonhomology that may lead to "breathing" or unpairing of the promoter region. Once again, the theoretical paired structure of a complementing genotype places the y² promoter region into an unpaired loop.

We suggest that the unpaired and/or looped state of the promoter region can be a key feature in promoter activation. An obvious candidate for a genetic element within the promoter region that responds to the unpaired state would be the promoter itself, although our data do not rule out the involvement of other elements. Presence of the promoter in a loop could increase its accessibility to transcription factors or regulatory input. The looped structure could also be subject to torsional constraints that lead to abnormal regulation of the promoter. Alternatively, in genotypes where one allele carries a deletion, a looped gene structure may promote transvection by altering the proximity of genetic elements to each other (WU and HOWE 1995 ; CASARES et al. 1997 ; MORRIS et al. 1998 ; SIPOS et al. 1998 ). For instance, due to the deletion nature of y^1#8, pairing of y^1#8 with y² may draw the y² promoter closer to the enhancers of both y² and y^1#8 (Figure 5B; also MORRIS et al. 1998 ). Although our illustrations highlight activation of the y² promoter, pairing-mediated TOPEs may also influence the transcriptional status of any promoter-bearing allele that is paired with y². Increases in productive yellow transcripts from this allele would contribute to the complementing phenotype.

While the structural model can explain the disruptions of the promoter region in all Class B alleles, we note two circumstances unrelated to transvection that may also contribute to the propensity of Class B alleles to have structurally disrupted promoter regions. First, if the promoter region of yellow is a hot spot for the insertion of mobile elements, then Class B alleles will have a high probability of carrying insertions in their promoter region. Second, if the function of the yellow promoter region is resilient to small changes in sequence such as single base pair mutations, then, with respect to the promoter region, isolation of yellow alleles based on pigmentation phenotype will select for gross structural changes. Nevertheless, the structural model warrants consideration. In a separate study of transvection at yellow, we found that a structural model could explain an unexpected facet of complementation in the y²/y^3c3 genotype (MORRIS et al. 1998 ).

Our data are also consistent with transvection being regulated both by transcription as well as by structure. The relative contributions of these two features may vary from one allele pair to another, or disruption of either feature alone may be sufficient to permit allelic interactions, but disruption of both is necessary to achieve the levels of transvection that are required for detection in our assays. For example, transvection may require a degree of transcriptional disruption that can only be achieved by substantial structural change. Finally, we suggest that disruption of transcription may produce a change in gene structure. This interpretation stems from a study indicating that transcription of intergenic spacer repeats in the rDNA locus of Drosophila plays an important role in the pairing of the X and Y chromosomes during male meiosis (reviewed by MCKEE 1996 , MCKEE 1998 ), and this interpretation proposes that transcription machinery may participate in homologue pairing (JORGENSEN 1990 ; COOK 1997 ). If transcription is a key step in pairing at yellow, then transcriptionally compromised yellow alleles may support transvection because they cause localized unpairing in the promoter region.

In our survey, only one allele from Classes B and C1 cannot be explained by either the transcriptional or structural model. This allele, y^76-1, appears transcriptionally silent and is structurally altered in the promoter region, yet it fails to complement y². It is possible that the complementation afforded by y^76-1 is too subtle to be noted in our assays or that y^76-1 is transcriptionally competent but its transcripts are unstable or cannot be detected by our RT-PCR protocol. Alternatively, the size or nature of the inserted sequences or the genetic background of y^76-1 may not be permissive of transvection. The y^76-1 also emphasizes the probability that our definition of the promoter region by restriction enzyme recognition sites does not accurately represent the region of yellow that regulates transvection; the structural disruption of y^76-1 may fall within the BamHI-HindIII fragment yet not disrupt the domain that affects transvection. The features of y^76-1 are reminiscent of two other alleles, y^76d28 and y^1#7, that were shown to carry an insertion in the promoter region but failed to complement y² (GEYER et al. 1990 ).

The behavior of the C2 and D alleles may reflect the impact of gypsy sequences:
Several studies have proposed that gypsy may influence transvection (E. B. LEWIS as cited in BABU et al. 1987 ; GEYER et al. 1990 ; CORCES and GEYER 1991 ; HENIKOFF 1994 ; MORRIS et al. 1998 ). Recently, this view has been advanced by reports that su(Hw) protein, when bound to one chromosome, can act in trans on another chromosome (GEORGIEV and CORCES 1995 ; GEORGIEV and KOZYCINA 1996 ; MORCILLO et al. 1996 ; SIGRIST and PIRROTTA 1997 ). Our interpretation of the C2 and D alleles in light of the transcriptional and structural models is also supportive of the potential of gypsy to modulate transvection.

The C2 alleles are members of Class C because they fail to complement y², and they form a special subclass of Class C because of their ability to complement Class D alleles. Molecular analysis proved the three C2 alleles to be structurally similar to each other and to y²; all three carry gypsy sequence inserted at -700. Of the three, y^1like appears structurally normal in the promoter region and is competent for transcription, so its failure to complement y² is consistent with either the transcriptional or structural model. In contrast, the behavior of the other two alleles, y⁶⁹ and y²⁰¹, is not easily explained by either model. They are structurally disrupted in the promoter region and are not competent for transcription, yet they fail to complement y². Particularly puzzling is their similarity to the Class B y^59b allele that does complement y². Like C2 alleles, y^59b carries gypsy sequences at -700 (GEYER et al. 1990 ). One explanation may lie in the fact that the gypsy insertion in y⁶⁹ retains binding sites for the su(Hw) protein, while the insertion in y^59b lacks these sites. The presence of binding sites in y⁶⁹ may inhibit transvection (GEYER et al. 1990 ). Consistent with this proposal, another allele that resembles y⁶⁹ but carries fewer binding sites, called y^88d, was found to complement y², but to a level that is less than that seen in y²/y^59b flies (GEYER et al. 1990 ). We do not yet know whether y²⁰¹ retains binding sites, but, on the basis of observations of y⁶⁹ and y^88d, would predict that it does. This interpretation suggests an alternative explanation for the failure of y^1like to complement y²; the y^1like allele may also carry binding sites.

Class D alleles complement both y² and y^59b, and therefore resemble both the Class A and Class B alleles. This dual nature can be explained by the molecular profiles of the two Class D alleles and is consistent with either the transcriptional or structural model. Both alleles carry insertions in the promoter region but remain capable of transcription. By their disrupted promoter region, Class D alleles resemble Class B alleles, and therefore their ability to complement y² is not surprising. On the other hand, as Class D alleles remain capable of transcription, they are able to respond to y^59b and other Class B alleles by becoming transcriptionally activated.

Class D alleles are also remarkable in their ability to complement C2 alleles. Based on the structural similarities between y² and the C2 alleles, we suggest that Class D alleles can be considered a special type of Class A allele, one that can complement y² and alleles that are structurally related to y². By their Class A status, Class D alleles would therefore be expected to complement Class B, but not C1, alleles, and, by their special relationship with y² and related alleles, Class D alleles would be expected to complement y² and the C2 alleles. This is the case.

The influence of pairing-mediated TOPEs in gene regulation:
Pairing-mediated TOPEs were first predicted by H. J. MULLER 1935 (cited by EPHRUSSI and SUTTON 1944 ). He proposed that the forces of pairing, when "acting between unlike genes, should tend to deform them, and this deformation might well affect the nature and the quantity of the gene products which they form" (MULLER 1941 , pp. 157–158). Pairing-mediated TOPEs have since been considered in the context of transvection (for example, E. B. LEWIS as cited in BABU et al. 1987 ; GUBB et al. 1990 ; HAZELRIGG and PETERSEN 1992 ; MARTINEZ-LABORDA et al. 1992 ; WU 1993 ; PETERSON et al. 1994 ; WU and HOWE 1995 ; GOLDSBOROUGH and KORNBERG 1996 ; CASARES et