Following the procedure illustrated in Figure 1, a total of 6,816 new, independently derived insertions were isolated in the F₁ generation and of these, 5,657 new insertion lines were successfully tested for lethality/sterility. 589 potentially homozygous lethal lines were identified in the first round of F₃ crosses, of which 421 (i.e. 7.4% of 5,657 insertions) were confirmed to be homozygous lethal in the second round (Table 1; for details on the two rounds of screening F₃ crosses please see Methods). A subset of the viable insertion lines, those producing fewer homozygotes than expected, was tested for semi-lethality. Insertion lines were designated as potentially semi-lethal if homozygosity of a parent was indicated for no more than one single-pair mating in the first round of F₃ crosses, or less than four single-pair matings after the second round. This was true for 236 insertions (out of the subset of 2,940 insertions analyzed in Göttingen and Erlangen) after the first round, of which 75 remained in this category after the second round. Hence, 2.5% (75/2,940) of all insertions tested for semi-lethality met the criteria for semi-lethality. This somewhat relaxed scoring criterion reduced the likelihood of missing or overlooking lethal or semi-lethal mutations.

Potentially homozygous sterile insertions lines were identified by evaluating the single-pair matings: Whenever two or more of the initial single-pair F₃ self-crosses (round one, Figure 1E) failed to produce offspring (although the parents were alive and healthy), the line was classified as potentially sterile. This was the case for 160 insertions (Table 1). We used either of two methods to confirm or refute a tentative diagnosis of recessive sterility. In the first method, we set up a second round of single-pair self-crosses bringing the total number of F₃ crosses to 20. The diagnosis was considered to be corroborated when the number of single-pair matings not producing any offspring increased to four or more. Using this definition, 124 potentially sterile lines were reduced to 21. However, further testing of these presumably sterile insertion lines showed that this criterion was not always reliable (see below). In the second method we set up 10 male and 10 female outcrosses. The diagnosis of recessive sterility was considered to be corroborated if the crosses failed to reveal either a fertile, homozygous male or a fertile, homozygous female. Out of 36 potentially sterile lines tested by the second method, only eight lines fulfilled this definition of sterility. Since the second follow-up test appeared to be more rigorous than the first, we retested 11 of the 21 apparently sterile lines from the former test using the more rigorous criterion. All 11 lines proved to be fertile in both sexes. It seems to be clear that most sterile lines found by using the first criterion are false-positives. Hence, we suggest using the stricter test for recessive sterility, which has the added benefit of identifying the affected sex.

3xP3-driven EGFP expression is typically seen only in the eyes and central nervous system 3. We analyzed all new insertion lines for additional, i.e. enhancer-dependent EGFP expression, and detected novel patterns at all developmental stages. Although we observed a bias for certain patterns (i.e. certain central nervous system patterns, segmentally-repeated stripes in embryos, or small dots at the hinges of extremities in larvae and pupae), we identified 505 unique enhancer-trap patterns. The bias for certain patterns might be caused by a favored expression in certain tissues due to the paired-class homeodomain binding sites in the 3xP3 element of the transformation marker 15. For a random subset of about 200 of all newly-identified insertions, we also dissected pupae and adults to look for EGFP expression in internal organs that might not be visible without dissection. Such expression patterns (e.g. a spermatheca enhancer) were found only rarely. Examples of enhancer-trap lines are shown in Figure 2A-H. Descriptions and/or photographs of all enhancer-trap lines together with information about their chromosomal locations (when known) are available in GEKU-base (http://www.geku-base.uni-goettingen.de; see Methods).

We analyzed the embryonic cuticle phenotypes of many lines identified as lethal and found a number of distinct cuticular abnormalities (Figure 2I-L). For example, line G08519 displays a phenotype similar to the proboscipedia ortholog maxillopedia in that maxillary (grey arrows) and labial (white arrow) palps are transformed to legs (Figure 2J); 1617. Indeed, this insertion is located in the first intron of maxillopedia. In addition, many lethal lines showed a high proportion of embryos that died prior to cuticularization, indicating early embryonic lethality.

To test whether the semi-lethal lines are false positives or true lethals with occasional escapers, we checked what portion of these lines (Göttingen subset) produce lethal L1 cuticle phenotypes (at least two cuticles with similar strong defects in one preparation when scoring at least 10 individuals) and compared it to the percentage of cuticle phenotypes produced by the other classes. A total of 25.8% (8/31) of a random selection of lines complying with the strict definition of lethality showed such phenotypes. Of lines with one or two single-pair matings (out of 20) indicating homozygosity (semi-lethality), this portion was 16.6% in each case (5/30 and 3/18, respectively). Lines with three single-pair matings indicating homozygous viability gave rise to cuticle phenotypes in only 6.25% (1/16). Thus analyzing semi-lethal lines led to the identification of additional cuticle phenotype-inducing mutations.

We determined the chromosomal location of 400 piggyBac insertions by BLAST analysis of amplified flanking sequences against the Tribolium genome (see Methods). These insertions included lethal, semi-lethal and sterile as well as viable lines that showed an enhancer-trap pattern. The distribution of 280 homozygous lethal insertions on the linkage groups is shown in Figure 3. The lethal insertions appear to be distributed randomly among the linkage groups, showing a range from 1.1 insertions per Mb for linkage group 10 up to 2.2 insertions per Mb for linkage group 4 (Table 2). Superimposed on the generally random pattern of insertion site locations, there appear to be insertion hotspots and coldspots, the most evident example being the hotspot for local reinsertion near the donor site on linkage group 3.

Based on a pilot screen published in Lorenzen et al. 11 we have performed the first high-efficiency, large-scale insertional mutagenesis screen in an insect species outside the genus Drosophila, and we have established a crossing scheme that circumvents the need for balancer chromosomes or embryo injections. From our experience, we estimate that using the procedure presented here, one person should be able to establish 150 lethal strains per year. While the GEKU screen has identified many interesting enhancer traps and lethal phenotypes, genome-wide saturation would be difficult to achieve at the current level of efficiency. The most time-consuming step is setting up and evaluating 20 single-pair matings for each new insertion line to detect recessive lethality. For this reason we set up a small number of single-pair matings first, as most viable insertions can be identified by evaluating just a few crosses from each subset. However, also for insertions recognized as viable it was important to assess the fertility of all remaining single-pair matings in order to ensure that recessive sterile insertions were detected.

We found that lethal lines were readily detected by single-pair matings. Based on the frequency with which semi-lethal lines produced strong L1 larval cuticle phenotypes, we suggest defining lines as potentially lethal when only one or two out of 20 single-pair matings indicate homozygosity. However, our definition of sterility proved to be too lax in the beginning, since most potentially sterile lines turned out to be false-positives in more detailed analysis.

The efficiency of generating lethal mutations by piggyBac-based insertional mutagenesis in Tribolium (7.4%) is similar to equivalent screens in Drosophila based either on piggyBac 918 or P elements 192021. Whether the efficacy of such screens can potentially be doubled by the inclusion of splice acceptor sites or insulator sequences within the mutator element - as has been shown in Drosophila 22 - still has to be determined in Tribolium.

The enhancer detection rate within this large-scale insertional mutagenesis screen was also 7.4%. This is actually higher than in a comparable Drosophila screen where enhancer detection without a suitable amplification system was about 2% 9. Only after including a GAL4-based amplifier system could Drosophila enhancer detection be raised to 50% 9. However, such directed expression systems still need to be further developed and assayed in Tribolium before they can be used in insertional mutagenesis screens.

In 14% of all lethal insertions, piggyBac had clearly jumped into the coding sequence of a gene. However, the majority of lethal insertions (61%; see Table 3) were located in introns, apparently disrupting transcription or splicing of the affected gene. One possibility is that the SV40 UTR in the transposon, which serves as a terminator of transcription in both directions, causes early transcriptional termination of the host gene. The tendency of piggyBac to insert into intronic sequences had already been observed in Drosophila insertional mutagenesis screens 1822.

In the described scheme, when new crosses were set up, one had to switch between fluorescence (to detect the transformation marker) and normal light (to determine sex) several times, which was a time-consuming procedure. To improve this situation considerably, we constructed and are testing new donors that use rescue of eye color by vermilion⁺ as an indication of transformation 2324. The use of such a system will also facilitate stock-keeping.

Another way to enhance efficiency in future screens might be the establishment of donors that include an artificial maternal-effect selfish element, e.g. Medea [2526, see also Methods]. Such elements induce the death of all offspring of a female (maternal-lethal effect) except for those that have inherited the element (zygotic rescue). For example, a modified piggyBac donor element could incorporate a Medea element in tandem with the 3xP3-EGFP-marker. This modified donor element would be inserted at a chromosomal location tightly linked to an easily-scored recessive marker, such as the body-color mutation black 27. In the P₁ cross this donor strain (homozygous for wild-type body color) would be mated with a helper strain (homozygous mutant for black). The resulting P₂ animals would carry one copy each of the helper, the donor and the mutant black allele. Moreover, the latter two would be located in trans at similar positions on homologous chromosomes. Such P₂ hybrids would be mated with beetles that were trans-heterozygous for black and the Medea-containing donor element, or, if the P₂ hybrid is female, they could instead be mated to homozygous black (non-helper, non-Medea) males. F₁ offspring with black body-color then would arise only if they inherit the black chromosome from both parents. Because the Medea-tagged donor is arranged in trans with black, such black offspring do not carry a donor, and hence lack zygotic rescue activity of the Medea element. This leads to their death by the maternal-lethal effect of the element. Only if the donor has been remobilized to another genomic location can offspring carry both black alleles as well as the rescuing donor. Hence, black body-color in the offspring indicates a remobilization event. This design would be an elegant means to enhance the detection of new insertion lines by obviating the need for fluorescence detection. It would also simplify the stock-keeping of lethal insertions, since a Medea element tightly linked with the lethal insertion would constitute a type of balanced lethal.

The donor strain used in this screen, Pig-19, carries a 3xP3-EGFP marked piggyBac element, pBac [3xP3-EGFPaf], that confers both, insertion-site-independent eye-specific EGFP expression, and donor-site-dependent muscle-specific EGFP expression 3. We previously demonstrated that remobilization of the Pig-19 insertion results in G₁ beetles lacking muscle-specific expression, but retaining eye-specific expression 311. Thus, the loss of muscle-specific expression can be used to detect remobilization events. The jumpstarter/helper strain used in this screen, M26 11, carries an X-chromosomal insertion of a 3xP3-DsRed marked Minos element (pMi [3xP3-DsRed; Dm'hsp70-pBac]) 9. Both strains are in a white-eyed pearl mutant background to facilitate detection of eye-specific fluorescence.

We used a P₁, P₂ and F₁ to F₄ scheme to comply with the nomenclature of standard Drosophila F₁ and F₃ genetic screens, respectively (Figure 1). Donor remobilization occurred in the germline of the P₂ generation, while new insertions and mutant homozygotes first appeared in the F₁ and F₃ generations, respectively. All crosses were carried out at 30-32°C. Virgin females were collected as pupae and stored at 23°C for up to six weeks prior to use. Insertional mutagenesis is described in detail in 11. In summary, P₁ mass-matings were set up between donor males and helper females (Figure 1A) and subcultured at intervals of two to three weeks. P₂ offspring were collected as pupae and examined to verify the presence of both piggyBac-based donor (EGFP marker) and Minos-based helper (DsRed marker) constructs. Individual P₂ virgin females were outcrossed to three pearl males each to ensure insemination (Figure 1B). The piggyBac donor element can be remobilized by piggyBac transposase activity in the germ line of the hybrid. New insertions were recognized in the F₁ progeny by the loss of donor-site-dependent EGFP expression (i.e. muscle fluorescence) coupled with retention of insertion-site-independent EGFP expression (i.e. eye fluorescence). For each P₂ outcross, a single F₁ beetle carrying a new insertion was outcrossed once again to pearl to check for single insertions (based on 50% Mendelian segregation of the new insert) and to generate families for subsequent analysis. For stability of the new insertions, only individuals carrying a new insertion and lacking the helper element (i.e. DsRed negative) were chosen (Figure 1C). Additionally, depending on the new chromosomal location of piggyBac, a new insertion might show a novel enhancer-trap pattern. Even when a P₂ cross produced multiple EGFP-positive offspring, only one F₁ beetle was chosen for continued study in order to ensure independent origin of each new insertion. This was necessary because several offspring carrying the same insertion could appear within a P₂ family as a result of a premeiotic remobilization event. For each F₁ outcross, five female and three male F₂ siblings were crossed to each other to establish new insertion strains and to enable testing for homozygous viability (F₂ cross; Figure 1D). To accomplish the latter, we performed a number of single-pair F₃ matings (Figure 1E) and analyzed their progeny for the presence of the donor element (see below).

If an insertion mutant were homozygous viable, then (after positive marker selection) the progeny of the F₂ cross would consist of a 1:2 ratio of homozygous to heterozygous beetles. Under the assumption of random sib-mating, 11.1% (1/3 × 1/3) of all F₃ single-pair matings would be between two homozygous beetles, 44.4% [2× (1/3 × 2/3)] between one homozygous and one heterozygous beetle, and 44.4% (2/3 × 2/3) between two heterozygous beetles. This implies that about 55.5% (11.1% + 44.4%) of the single-pair matings (given a fully-viable insertion) would produce only EGFP-positive progeny (because at least one parent would be homozygous for the insertion). The remaining 44.4% would produce mixed progeny (i.e. approximately 75% EGFP positive and approximately 25% EGFP negative) because both parents would be heterozygous for the insertion. In contrast, for recessive lethal insertions, no homozygous beetles would be present in the F₃ generation so all F₃ crosses would produce mixed progeny. Thus, the presence of even a single EGFP-negative beetle in the F₄ generation indicates heterozygosity of both parents, and the complete absence of EGFP-negative progeny indicates homozygosity of at least one parent. Depending on the distribution of the above-mentioned phenotypes, each single-pair mating was scored and assigned to one of five categories (see Figure 1F and Table 4 for details).

Since more than 40% of all single-pair matings were expected to produce mixed progeny (even if the insertion was fully viable) we analyzed a total of 20 single-pair matings before concluding that an insertion was lethal. On the other hand, since viable insertions were usually identified after evaluating just a few single-pair matings, we split the 20 crosses into two consecutive rounds to maximize throughput. The second round of single-pair matings would be set up only if an insertion were not clearly identified as viable after evaluating the first round (Table 5).

The following potential errors could occur using this method to test for recessive lethality: (1) A homozygous-viable insertion mutant could be falsely judged homozygous lethal because all single-pair matings produced mixed progeny. This could occur if, by chance, all single-pair matings consisted of heterozygous beetles. The probability of such an occurrence is (²/₃)ⁿ (n = number of beetles tested), because two-thirds of all EGFP-positive F₃ beetles carrying a viable insertion are heterozygous. For eight single-pair matings (number of test beetles = 16), this probability equals 0.15%. For 20 single-pair matings, the probability that all (40) test beetles selected from a homozygous-viable line are heterozygous, is only 9.0 × 10^-6. Thus, evaluating 20 single-pair matings is sufficient to exclude false-positive lethal lines with a very high level of confidence. (2) A homozygous-lethal insertion (all F₂ progeny are heterozygous) could be falsely identified as homozygous viable if, by chance, no EGFP-negative progeny are observed from a single-pair mating, even though 25% are expected. The probability of this happening when 20 progeny are analyzed is about 0.3% (0.75ⁿ; n = number of progeny screened). Because the probability of misdiagnosing a lethal insertion rises if fewer progeny are analyzed, single-pair matings yielding less than 20 progeny were not used to make inferences about the lethality of the insertion (= 'uninformative single-pair mating' in Table 4) unless some progeny were EGFP negative.

The fact that the helper insertion used in this work is X-linked imposed restrictions on the design of our crossing scheme. X-chromosomal insertions that were homozygous lethal or sterile could be obtained only if the following is considered: Because only new transformants that segregated away from the helper element were selected, hybrid females had to be used to set up P₂ crosses in order to avoid bias against new X-linked insertions. Additionally, males with a new hemizygous X-linked lethal insertion would not survive and ones hemizygous for a new X-linked sterile insertion would be useless for generating a new stock. Hence, one could obtain X-linked lethal and sterile insertions only if female beetles carrying the donor element were used to set up the P₂ as well as the F₁ crosses. Therefore, we selected only female hybrids and used females carrying new insertions whenever possible.

At least one new insertion was detected in about 30% of all P₂ crosses when about 20 offspring were screened. The percentage of P₂ crosses that yield new insertions can be greatly increased by screening a larger number of progeny per P₂ cross. For a subset of P₂ crosses we screened 100 progeny per cross, and found at least one new insertion in every case. In practice, about 10 - 30 P₂ pupae were present when the P₂ progeny were screened for new insertions. The decision to discard the larval P₂ offspring when a new insertion could not be detected in the first attempt represented a compromise between the need to find at least one new insertion in each family and the aim to obtain a large number of independent insertions with limited resources in time and space.

The genomic location of an insertion was determined by sequencing flanking DNA obtained by one of the following three polymerase chain reaction (PCR) -based methods: inverse PCR 28, universal PCR 329, or vectorette PCR 30. The procedure for inverse PCR including primer design was adapted from 'Inverse PCR and Sequencing Protocol on 5 Fly Preps', Exelixis Pharmaceutical Corp (South San Francisco, California, USA) 22. Following DNA isolation, approximately 1 μg of DNA was digested with Sau3A1, BfUC1, or Ase1 (for 5' iPCR) or HinP1 (for 3' iPCR). Approximately 100 ng of digested DNA was then self-ligated to obtain circular DNA fragments, followed by two rounds of nested PCR. DNA templates (PCR products and/or cloned PCR products) were sequenced by Seqlab (Göttingen, Germany), Macrogen (Seoul, Korea), or using an ABI 3730 DNA sequencer (Sequencing and Genotyping Facility, Plant Pathology, Kansas State University, Manhattan, Kansas, USA). Data analysis was performed using Vector NTI^® software (Invitrogen, Carlsbad, California, USA). After trimming vector sequences, flanking DNA sequences were then searched (BLASTN) against Tribolium castaneum genome sequences at HGSC, Baylor College of Medicine http://www.hgsc.bcm.tmc.edu/projects/tribolium/, NCBI http://www.ncbi.nlm.nih.gov/genome/seq/BlastGen/BlastGen.cgi?taxid=7070 or BeetleBase http://beetlebase.org/. If the insertion was in a predicted gene (GLEAN set), a transcription unit (EST or cDNA) or region indicated by Drosophila BLAST or other gene prediction method as a potential gene, the predicted Tribolium gene was examined by BLAST analysis at FlyBase for the top Drosophila hit, and NCBI (nr database) to identify other potential orthologs. Insertion site sequences were deposited to NCBI (for accession numbers see Additional File 1) and also put - including the retrieved information - into GEKU-base (see below).

When hybrid females and pearl males (P₂ generation) were crossed severe segregation distortion was observed: 98% of the progeny were EGFP positive, rather than the expected 50%. The DsRed marker however showed the expected 1:1 ratio (i.e. segregated independently of the EGFP marker). The unusual segregation of EGFP has been shown 11 to be the result of close cis-linkage (approximately 2 cM) of the maternally acting selfish gene Medea 25 with the Pig-19 donor insertion 3 on LG3. However, the segregation ratios of new insertions were affected only when the piggyBac element reinserted near the original donor insertion (representing a local hop).

All available information about the analyzed insertion lines can be found at a web-based database called GEKU-base http://www.geku-base.uni-goettingen.de. Information provided includes (if available) photographs and descriptions of enhancer traps and phenotypes, flanking sequences and chromosomal location, affected genes and their orthologs. GEKU-base also provides information on how to obtain insertion lines.

Marker-gene fluorescence was detected using a Nikon fluorescence stereomicroscope SMZ1500 (Nikon GmbH, Düsseldorf, Germany) at Göttingen and Erlangen, an Olympus SZX12 fluorescence stereomicroscope (Olympus Corporation, Tokyo, Japan), or a Leica MZ FLIII fluorescence stereomicroscope (Leica Microsystems Inc., Wetzlar, Germany). The filter sets used for EGFP expression were: [Göttingen: 470/40 nm excitation filter, 500 nm LP emission filter, and 495 nm beamsplitter], [Erlangen: 480/40 nm excitation filter, 510 nm emission filter, and 505 nm beamsplitter], [KSU: 480/40 nm excitation filter and 535/50 nm emission filter], [USDA: GFP Plus filter set (excitation filter: 480/40 nm, barrier filter: 510 nm)]. The filter sets used for DsRed expression were: [Göttingen: 546/12 nm excitation filter, 605/75 nm emission filter, and 560 nm beamsplitter], [Erlangen: 565/30 excitation filter, 620/60 nm emission filter, and 585 nm beamsplitter], [KSU: 545/30 excitation filter and 620/60 emission filter], [USDA: TXR TEXAS RED filter set (excitation filter: 560/40 nm, barrier filter: 610 nm)]. To detect enhancer-trap patterns in embryos, we dechorionated embryos derived from F₃-crosses.

Gene names refer to respective Drosophila orthologs. The line E00321 is homozygous lethal and carries an insertion in lethal (2) giant larvae (Figure 2A). The line E00713 is homozygous viable and carries an insertion 149-bp upstream of the 5' end of GLEAN_03347, Glutatione S transferase, (Figure 2B). The homozygous viable line G01004 carries an insertion near Ultrabithorax (Figure 2C). The homozygous viable line G04717 carries an insertion near lame duck (Figure 2D). The line KT1539 is homozygous lethal and the insertion site is near the gene pointed (Figure 2E). The homozygous lethal line KS030 bears an insertion in an intron of lozenge (Figure 2F). The KS406 line is homozygous viable and carries an insertion in an intron of GLEAN_00277 which shows similarity to genes encoding protein tyrosine phosphatases. Other genes in the vicinity of this insertion are Fgf8 or Or48 (Figure 2G). The homozygous viable line MH30a has an insertion near female sterile (2) Ketel (Figure 2H). The line E00916 is homozygous lethal and carries an insertion in an exon of GLEAN_08270 (Drosophila ortholog: Cyclin D) (Figure 2I). The G08519 insertion is located in the first intron of proboscipedia (Figure 2J). The KT1096 insertion is in an intron of the pecanex ortholog (Figure 2K). The E03501 insertion is in an intron of the Tribolium ortholog of Ftz-F1 (Figure 2L).