A deletion-generator compound element allows deletion saturation analysis for genomewide phenotypic annotation

Methods

Results and Discussion

We have developed a technology (Fig. 1) that can rapidly generate nested deletion arrays starting with the insertion of a single-copy transgenic construct at any location in the genome. The general features of this system (Fig. 1A) are that it contains two independent mobile-element systems that are incapable of cross-mobilization. The outer “carrier” element contributes mobile-element termini sufficient to transport the entire transgenic construct to alternative genomic locations in the presence of the enzymes necessary to catalyze its mobilization. The inner “deleter” mobile element is one that has a high frequency of replicative mobilizations to nearby genomic sites (hereafter referred to as “local hops”) catalyzed by its mobilization enzymes. Further, when a local hop inserts in the same 5′-to-3′ orientation as the starting deleter element, then there is a high probability of recombination between the two copies of the element, leading to the production of genomic-deletion and tandem-duplication events. The genetic markers A and B bracketing the deleter element allow directional detection of the deletion events through the loss of one of the two genetic markers.

We have implemented this general strategy in D. melanogaster by constructing an element that we term P{wHy} (Fig. 1B) in which the P element serves as the carrier, the hobo element serves as the deleter, and the genetic markers w⁺ (the engineered derivative of a wild-type allele of the white eye-color gene) and y⁺ (the wild-type allele of the yellow body-color gene) provide the detection system for the directional detection of deletion events. The carrier P element can be mobilized only in the presence of a source of P transposase within the genome, and similarly, the deleter hobo element can be mobilized only in the presence of a source of hobo transposase. Based on work we carried out on hobo transposase-mediated mobilizations of a preexisting single defective hobo element located within the decapentaplegic (dpp) locus in Drosophila (ref. 10 and L.R.B., M.A.C., and W.M.G., unpublished data), deletion events in which one endpoint of the deletion falls exactly at the end of a hobo element can be obtained readily through introduction of a source of hobo transposase. The interpretation of the origin of these events is a two-step process that occurs at sequential times within the divisions of a single germ line: (i) local hops of hobo are generated at a high rate after exposure to hobo transposase followed by (ii) transposase-mediated intrachromosomal recombination between the original hobo and the transposed copy residing nearby in the same orientation (10, 18, 19). By bracketing the hobo element in P{wHy} with functional transgenes for w⁺ and y⁺, recombinational events following local hop of hobo in the same orientation into a nearby genomic location result not only in the loss of the intervening genomic material but also of the w⁺ or y⁺ marker gene on that side of the P{wHy} element (Fig. 1C). Thus, deletions can be detected purely on the basis of the markers internal to P{wHy} without regard to the extent of the genomic deletion.

For the P{wHy} system to be useful for global deletion analysis, three criteria need to be met: (i) distribution (deletions must occur over a broad size range), (ii) granularity (deletions typically must have different end points, reflecting a broad distribution of insertion sites produced by local hopping), and (iii) frequency (the system must be efficient enough to produce a large number of deletions with reasonable effort).

To investigate the ability of this system to meet these criteria, we have generated several single-copy P{wHy} autosomal insert lines. Of these, we chose for intensive analysis the first two lines we encountered that, at that time, were inserted into a large contiguous block of finished genomic sequence deposited in GenBank. Flies from these lines, P{wHy} CG12930^01D09 and P{wHy} 01D01, were crossed to a source of hobo transposase as described in Methods. The progeny of this cross containing P{wHy} and hobo transposase were test-crossed to y¹ w^67c23 flies. Their progeny were scored for the two phenotypic classes, w⁺y⁻ and w⁻y⁺, expected for deletion events. For each line, we recovered these two phenotypic classes. We will focus on the characterization of the two nested deletion series for which we have the most extensive data sets (Table 1). The hobo mobilization frequencies encountered in the two cases, 404/1,539 (26%) and 363/1,870 (19%), respectively, of the male germ lines tested (Table 1) are typical for other P{wHy} lines (ref. 21 and F.H., J.T.L., T. Martin, M.A.C., and W.M.G., unpublished data).

The recovered hobo mobilization derivatives were subjected to a three-step characterization procedure. The first step culled the hobo mobilization events in which the white or yellow marker genes are inactivated or deleted by events occurring solely within the P{wHy} element (Fig. 2B). Derivatives that pass this first step are classified as “genomic events” (Fig. 2A). Although the occurrences of P{wHy}-confined events are substantial (≈1/3 of the total transmitted events), they are detected readily and eliminated at an early stage of analysis and thus do not detract from the overall efficacy of the system.

The second step of characterization of the genomic events consisted of sequence mapping of the DNA adjacent to the relevant hobo terminus. By means of two primed genomic sequencing techniques (inverse PCR and universal fast walking; see Methods), the sequence flanking the terminus of the hobo adjacent to the genomic event was obtained for ≈70% of the genomic events (Table 1). Each flanking sequence then was aligned to the Drosophila genomic sequence (1) by using BLAST (20). “Candidate deletions” were selected on the basis of the following criteria: (i) the hobo-flanking genomic sequence is located in the vicinity of the original insertion site, (ii) the flanking sequence is on the expected side of the original insertion relative to the phenotypic marker lost, (iii) the flanking sequence is in the proper orientation relative to the hobo end, (iv) hobo is detected in only one copy, and (v) hobo is not flanking a transposon or repetitive sequence. Candidate deletions turned out to be the bulk of genomic events: 83% for P{3′wHy,w⁻y⁺}01D09Y (hereafter termed 01D09Y) deletions and 72% for P{5′wHy,w⁺y⁻}01D01W (hereafter termed 01D01W). The loss of one of the two phenotypic markers thus is an excellent predictor of deletions among genomic events.

Very few of the candidate deletion endpoints are identical in extent. Only 14/127 and 14/119 endpoints occur more than once in the 01D09Y and 01D01W series, respectively (Table 1). We conclude that multiple occurrences of the same deletion endpoints arise with a sufficiently low frequency that they do not hinder implementation of this system. In other words, local hopping produces many different hobo insertion sites and leads to a broad series of nested deletions.

If the candidates identified by these molecular analyses are true deletions and are of sufficient extent, they should uncover genetic markers within the deleted regions. Therefore, for the third step of characterization, complementation tests with molecularly localized recessive mutations were performed. Four recessive lethal P element insertion mutations were used: Uba1^s3484 and dap⁰⁴⁴⁵⁴ for the 01D09Y deletion set (Fig. 3A) and dev⁰¹⁷³⁸ and EP(2)2600 for the 01D01W deletion set (Fig. 3B). As expected, candidate deletions predicted to uncover these mutant transcription units fail to complement these mutants (Fig. 3). Only 1 exception of 114 01D09Y deletions and 5 exceptions of 105 01D01W deletions were encountered (Table 1). These exceptions may have been generated by a double event, a small deletion confined to the P{wHy} transgene, and a secondary local hobo insertion. We conclude that for the two sets of deletions, the vast majority of the candidate deletions behave as expected in the complementation tests and thus are validated deletions.

One of our key objectives was to assess the ability of this system to achieve “nested deletion saturation,” which produces deletion sets with endpoints staggered such that the transcription units in the flanking region are separable genetically. The molecular extent of recovered deletions ranges from 216 bp to 400 kb of adjacent genomic DNA. The overall size distribution (Fig. 4) of the validated deletions is consistent between the two reported sets (and among other less extensive sets, data not shown). The density of deletions falls off with distance, resulting in a higher number of deletions in the vicinity of the insertion site.

To define the region within which nested deletion saturation can be achieved, we compared the deletion boundary distribution and transcription unit locations in the vicinity of the two P{wHy} insertions. For the two sets of 100 and 113 validated deletions, it seems that ≈60 kb represents a reasonable range within which a high density of deletions is recovered (Fig. 5). The 01D09Y 60-kb region has only three putative genes in 60 kb (low gene density; ref. 1; Fig. 5A). All three transcription units in the 01D09Y 60-kb region are separated by deletion breakpoints. In the 01D01W 60-kb region, each transcription unit is separated by at least one deletion boundary from the neighboring transcription unit, which is true even within a genetically dense 23-kb subregion containing eight transcription units (Fig. 5B).

Conclusions

The P{wHy} technique should make significant contributions to the phenotypic annotation of the Drosophila genome. In the two regions that we investigated in detail here, we see features that are very encouraging regarding the general application of this methodology. Specifically, we find that hobo transposase-mediated P{wHy} deletions are recovered at sufficient frequencies to enable the detailed dissection of the molecular vicinity of an insertion. Within 60 kb of an insertion site, a collection of 100 deletions (a reasonable collection that could be generated from ≈1,000–2,000 initial mobilization crosses) would produce a nested deletion series in which endpoints would be staggered every 1–3 kb. Beyond this 60-kb distance, there would be occasional deletions as well, but the distances between adjacent endpoints would be considerably larger. Thus, for most regions of the genome, we would expect deletions to separate transcription units reliably within ≈60 kb on either side of the initial P{wHy} insertion site.

There are two obvious applications of P{wHy} deletions for phenotypic annotation of the genome. In one, nested deletions can be used for complementation of known recessive mutations that have been localized to a specific cytogenetic (or molecular) region. In the other, nested deletions from P{wHy} inserts in the vicinity of one another are generated, and through inter se crosses transheterozygous pairs of deletions are produced that remove both allelic copies of one or a cluster of transcription units. This approach would allow the association of phenotype with elimination of a specific genomic region in a systematic way. The same set of nested deletions could be used to assess the null phenotypes produced by loss of many transcription units in a given region, independent of the existence of other genetic reagents in that region. We have established that both the complementation test and overlapping deletion assay systems can be exploited effectively with the P{wHy} system (21).

The P{wHy} system has the additional feature that phenotypes are associated with the deletion of a specific set of base-pair locations in the genome; thus, even if the transcription unit models in a given genomic region are revised (as many inevitably will be over the next several years), these genomically defined phenotypic data will migrate easily forward to the new annotations. Further, this technique will allow the discovery of important functions in noncoding sequences in general and, in particular, noncoding transcripts and small transcripts such as micro-RNAs (22). Gene clusters of related function (23) can be eliminated, allowing studies of their aggregate contributions to phenotype.

Given the ability to recover finely staggered deletion endpoints within 60 kb of an insertion site, ≈2,000 P{wHy} insertions regularly spaced over the genome would be sufficient to produce a comprehensive set of deletions covering the entire euchromatic genome.

Finally, this technology should be applicable to a broad set of eukaryotic genomes for which carrier and deleter elements are already available. The carrier element that can be moved in single copy to many integration sites throughout the genome can be the Mos element in Caenorhabditis elegans (24), Minos and piggyBac in a variety of dipterans and other insects (25), Minos in human cells (26), transfer DNA in plants (27), gene-trap vectors (28) and the transposon Sleeping Beauty (29, 30) in mouse, and several retroviruses and transposons in zebrafish (31). The second essential component, the deleter hobo-like element, belongs to the hAT superfamily of transposable elements (18, 32) found in plants (Ac and Tam3), fungi (Tfo1), fish (Tol2), and mammals (Tramp). Ac is known to undergo mainly local transposition in Arabidopsis (33) and tobacco (34). Offset recombinational events were described first in maize as triggering the classical breakage–fusion–bridge cycles (35) and also have been illustrated in bacteria (36). All these data suggest that hobo or other hAT members can be used as deleter elements in non-Drosophila eukaryotes. Overall, because of the existence of comparable deleter and carrier elements, we anticipate that the deletion-generator strategy will have applications in many model organisms.