Two modes of transvection at the eyes absent gene of Drosophila demonstrate plasticity in transcriptional regulatory interactions in cis and in trans

Abstract

For many genes, proper gene expression requires coordinated and dynamic interactions between multiple regulatory elements, each of which can either promote or silence transcription. In Drosophila, the complexity of the regulatory landscape is further complicated by the tight physical pairing of homologous chromosomes, which can permit regulatory elements to interact in trans, a phenomenon known as transvection. To better understand how gene expression can be programmed through cis- and trans-regulatory interactions, we analyzed transvection effects for a collection of alleles of the eyes absent (eya) gene. We find that trans-activation of a promoter by the eya eye-specific enhancers is broadly supported in many allelic backgrounds, and that the availability of an enhancer to act in trans can be predicted based on the molecular lesion of an eya allele. Furthermore, by manipulating promoter availability in cis and in trans, we demonstrate that the eye-specific enhancers of eya show plasticity in their promoter preference between two different transcriptional start sites, which depends on promoter competition between the two potential targets. Finally, we show that certain alleles of eya demonstrate pairing-sensitive silencing resulting from trans-interactions between Polycomb Response Elements (PREs), and genetic and genomic data support a general role for PcG proteins in mediating transcriptional silencing at eya. Overall, our data highlight how eya gene regulation relies upon a complex but plastic interplay between multiple enhancers, promoters, and PREs.

Fig 1. Models for transvection via enhancer action in trans and pairing-sensitive silencing.

A, mutant alleles of yellow can be placed into at least three classes based on their molecular lesions: Class A includes deletions of enhancers and insertions of insulator elements between the enhancers and the promoter; Class B includes point mutations and deletions within the core promoter and transposon insertions in either the promoter or the 5’ UTR; Class C includes point mutations and indels in the coding region. Class A and Class B alleles complement when paired, indicating robust enhancer action in trans, whereas complementation between Class A and Class C is either non-existent or too weak to be detected, indicating poor enhancer action in trans. B, a PRE fused to a mini-white transgene typically permits expression in a hemizygous transgenic insertion, but silences expression when insertions are homozygous, reflecting pairing-sensitive silencing.

Fig 2. Classical alleles of eya.

A, top, alleles affecting regulatory sequences mapped to a schematic of the locus. Transcription of eya initiates from two major promoters, eya-B and eya-A, which are separated by approximately 10 kb. Several eye-specific enhancers are located upstream of the eya-B promoter; E is active in early instar discs and is required for eye development, whereas 1 (also known as IAM) and 2 drive expression in later disc tissues [–26]. Other fragments identified as candidate late enhancers in other regions of the locus are omitted for simplicity. Alleles eya¹, eya², and eya^cs delete eye-specific enhancer sequences, and eya³ and eya⁴ carry retrotransposon insertions into the 5’ UTR of exon 1B. Bottom, alleles with point mutations or indels mapped to the Eya protein structure. Alleles in red had their molecular lesions characterized in this work (see S1 File). The eya^D3 allele carries an indel in exon 5 where a TT doublet is substituted by CCCCCCCCCG, creating a frameshift at F737 and resulting in a C-terminus of 48 random amino acids in place of the usual 23. B, example of transvection between eya² (Class A) and eya⁴ (Class B), first demonstrated by Leiserson et al. [27]. Homozygosity of either allele results in eyeless flies, but trans-heterozygotes produce an eye of approximately ¾ size via enhancer action in trans.

Fig 3. Transvection via enhancer action in trans is supported by diverse alleles of eya.

A, eyes of representative flies carrying the Class A allele indicated at left and the Class B or Class C allele indicated above. B, quantification of eye development for flies carrying either eya¹ (left graph) or eya² (right graph) and the Class B or Class C alleles indicated below. For each genotype, approximately 20 eyes from 10 flies were scored for the number of ommatidia.

Fig 4. Transvection at eya via enhancer action in trans requires somatic homolog pairing.

A, eyes of representative flies carrying the allele indicated at left in trans to the Class B or Class C allele indicated above. ETD2.2 is a transvection-disrupting rearrangement of a second chromosome carrying eya². B, quantification of eye development for flies carrying the indicated alleles. Approximately 20 eyes from 10 flies were scored for each genotype. Note that the data for eya² crosses are identical to those presented in Fig 3.

Fig 5. Plasticity in promoter use for enhancer action in trans.

A-F, third instar eye-antennal discs from wild type (A, B), eya⁴/eya² (C, D), or eya⁴/eya¹ (E, F) larvae. Discs were subjected to in situ hybridization using probes specific to exon 1B (A, C, E) or exon 1A (B, D, F). Arrowheads indicate the approximate position of the morphogenetic furrow. Robust staining is seen for exon 1B in all cases; Exon 1A shows weak staining in wild type, is generally undetectable above background staining in eya⁴/eya², and appears prevalent in eya⁴/eya¹. Schematic diagrams (above) indicate likely promoter usage in each genotype.

Fig 6. Deletion of the eya-B promoter causes a switch in specificity of eye-specific enhancers to the eya-A promoter.

A, strategy for sgRNA design to delete the core promoter and first exon of the eya-B transcript using CRISPR/Cas9. B-C, eyes of representative flies homozygous (B) or hemizygous (C) for the allele eya^2m23a, which carries the deletion indicated in part A. Eyes are nearly identical to wild type. D-G, third instar eye-antennal discs from wild type (D, E), or eya^2m23a/eya^2m23a (F, G) larvae subjected to in situ hybridization using probes specific to exon 1B (D, F) or exon 1A (E, G). Loss of exon 1B in eya^2m23a flies leads to robust upregulation of the eya-A transcript. H, quantitative RT-PCR on cDNA isolated from wild type (left) or homozygous eya^2m23a (right) third instar eye-antennal discs using primers specific for exon 1B or exon 1A. Deletion of exon 1B results in a roughly 3.5-fold increase in transcripts from the A promoter.

Fig 7. The hypomorphic Class A allele eya^cs can participate in enhancer action in trans.

A, eyes of representative flies carrying the indicated genotypes. ETD4.3 is a transvection-disrupting rearrangement of a second chromosome carrying eya⁴. B, quantification of eye development for flies carrying the indicated alleles in trans to eya^cs. Consistent with enhancer action in trans, increased eye development is observed when eya^cs is placed in trans to Class B or Class C alleles relative to eya^cs homozygotes, and the increase is disrupted by the eya⁴ rearrangement ETD4.3 (compare column 3 to column 6).

Fig 8. Pairing sensitive silencing between Class A alleles reflects a direct role for PcG proteins in regulating eya.

A-D, eyes of representative flies carrying the indicated genotypes. ETD2.2 is a transvection-disrupting rearrangement of a second chromosome carrying eya². E, quantification of eye development for flies carrying the indicated alleles in trans to eya^cs. Consistent with pairing-sensitive silencing, decreased eye development is observed when eya^cs is placed in trans to other Class A alleles relative to eya^cs homozygotes, and the decrease is disrupted by the eya² rearrangement ETD2.2 (compare column 4 to column 5). Furthermore, 98.8% of eya^cs/eya² flies are completely eyeless (n = 462; see Panel K), but all (n = 120) eya^cs/ETD2.2 flies develop scorable eye tissue. F, genomic features of the eya locus (note that genomic maps are reversed relative to the reference sequence.) Top, Hi-C contact map showing TAD structure, with chromatin color map below. The eya locus occupies a TAD that is primarily blue chromatin, indicative of Polycomb silencing. Below, ChIP-seq peaks for K27-trimethylated histone H3 and individual PcG proteins as assayed from larval disc tissues. Six putative PREs (red boxes) are indicated by the pattern of peaks. G, schematic showing positions of eye-specific enhancers of putative PREs upstream of the eya B promoter. Bars below the schematic indicate the deletions carried by Class A alleles. H-I, eyes of representative flies carrying the indicated genotypes. J, quantification of eye development for flies carrying the indicated genotypes. Completely eyeless flies were not scored in this analysis. Significant increases in the ommatidia count are observed for eya^cs/eya¹; E(z)^S1/+ (p = 0.03, Mann-Whitney test) and eya^cs/eya¹; Pc⁴/+ (p = 0.04) relative to eya^cs/eya¹; +/+. K, scoring of flies that have at least one eye of any size in the indicated genotypes. Asterisks indicates significant difference relative to eya^cs/eya²; +/+ (p = 0.008, Mann-Whitney U test) or eya^cs/eya¹; +/+ (p = 0.03). “N” indicates the number of flies scored, with the number of separate vials in parentheses. L, Models for interactions between eye-specific enhancers and putative PREs. For simplicity, only the E enhancer (required for initiating eye development) and its neighboring PRE are shown; we expect that other local PREs behave similarly. In wild type, we propose that the enhancer has dual roles to block silencing by PREs through an unknown mechanism and to activate transcription primarily from the eya-B promoter. The eya^cs allele retains partial enhancer activity, whereas the eya² allele has no detectable enhancer activity; in eya^cs/eya² flies, pairing between homologous PREs (dual arrows) increases their capacity to silence, resulting in an eyeless phenotype in the paired state, but a partial eye when unpaired by chromosomal rearrangements. Since eya¹ deletes some of the putative PREs, pairing-dependent silencing is reduced, and the eya^cs/eya¹ phenotype is less severe than that of eya^cs/eya².