A Genome-Wide Collection of Mos1 Transposon Insertion Mutants for the C. elegans Research Community

Methods that use homologous recombination to engineer the genome of C. elegans commonly use strains carrying specific insertions of the heterologous transposon Mos1. A large collection of known Mos1 insertion alleles would therefore be of general interest to the C. elegans research community. We describe here the optimization of a semi-automated methodology for the construction of a substantial collection of Mos1 insertion mutant strains. At peak production, more than 5,000 strains were generated per month. These strains were then subject to molecular analysis, and more than 13,300 Mos1 insertions characterized. In addition to targeting directly more than 4,700 genes, these alleles represent the potential starting point for the engineered deletion of essentially all C. elegans genes and the modification of more than 40% of them. This collection of mutants, generated under the auspices of the European NEMAGENETAG consortium, is publicly available and represents an important research resource.


Introduction
The nematode worm Caenorhabditis elegans has long been a model of choice for many areas of biological research because of its powerful genetics.The worm continues to attract researchers thanks to the extensive community-generated resources that have been developed for genetic and functional genomic studies.Recent advances have now made it possible to engineer specific changes to the C. elegans genome through homologous recombination.Two techniques have become popular, MosTIC, for Mos1 excisioninduced transgene-instructed gene conversion, [1], and MosSCI, for Mos1-mediated single-copy insertion [2].Both methods rely on the availability of C. elegans strains carrying integrated copies of the heterologous Mos1 transposon at defined genomic addresses.Chromosomal breaks can then be generated at a single locus through the controlled excision of the Mos1 transposon.These breaks can be repaired through homologous recombination, using specifically designed transgenic templates, with homology arms that match the genomic sequence on either side of the Mos1 transposon insertion site.In the case of MosTIC, the repair template can be engineered to introduce a mutation at a specific locus [1].This method also makes it possible to ''knock-in'' reporter or affinity purification tags [3], thereby circumventing possible artifacts arising from transgene overexpression or chromatin-based position effects on gene expression.A third technique, MosDEL, can be used to generate Mos1-mediated targeted deletions of up to 25 kb [4], allowing null alleles to be generated with relative ease.A large collection of molecularly defined Mos1 insertion alleles would therefore be an extremely useful addition to the C. elegans toolkit.Previously, we documented the feasibility of generating such a collection [5] and detailed the implementation of a semiautomated high-throughput method for mutant production and screening [6].Here, we describe how three laboratories involved in the European NEMAGENETAG project [7] successfully produced and characterized a large collection of strains carrying Mos1 insertions.Theoretically, the collection of Mos1 insertion alleles obtained during this project is sufficient to permit essentially every gene in the C. elegans genome to be knocked-out and more than 40% of genes to be modified in a targeted fashion.We have also devised and implemented a simple web-interface (MosLocator) to help researchers identify potentially useful alleles.As the Mos1 strains are publicly available, they will be of general utility to the growing community of C. elegans researchers.

Generating a large collection of independent Mos1 alleles
In a pilot-scale experiment, we demonstrated the feasibility of generating and molecularly characterizing a large number of Mos1 mutant alleles [5].We subsequently described how the production of clonal lines of C. elegans with independent Mos1 transposon insertions could be semi-automated, leading the way to the creation of a genome-scale collection of Mos1 mutants [6].The procedure used a standard protocol [8], which was initiated by mating two strains of worms, EG1470 and EN547, each carrying an extrachromosomal transgenic array.EG1470 animals provided the Mos1 substrate on a transgene, and could be recognized because they also expressed a pharyngeal GFP marker.EN547 animals carried a heat-shock inducible Mos1 transposase transgene associated with expression of GFP in the coelomocytes.From the resulting cross-progeny, individual hermaphrodites carrying both transgenic arrays were manually identified on the basis of the expression of the two GFP markers.These animals were then subjected to a transient heat-shock in order to activate the expression of the transposase from one array, thus allowing the Mos1 substrate present on the other transgenic array to be mobilized.In a certain proportion of oocyte nuclei, the Mos1 substrate integrated into the genome.The F1 progeny of these animals were then sorted using the Union Biometrica COPAS Biosort on the basis of GFP marker expression.In order to maximize Mos1 insertions, and to prevent any further Mos1 transposition, only individual worms that retained the substrate array, associated with pharyngeal GFP expression, but not the transposase array, were retained.
The fourth generation progeny derived from these F1 worms were passed through the Biosort and single non-fluorescent offspring for each original F1 worm were isolated and individually placed in culture for an additional 2 generations.The progeny obtained from each of these individual worms were then tested with an automated PCR protocol for the presence of a chromosomally-integrated Mos1 transposon [6].
As previously reported, the transposition frequency and the final rate of obtaining Mos1 insertion mutants varied considerably from week to week [6].We noted that the frequency of Mos1 transposition declined relatively rapidly when doubly transgenic worms were maintained in culture over time (Figure S1A).Thus, in an effort to reduce this variability, we modified our strategy, and generated a large pool of early generation doubly transgenic worms that were cryogenically stored.We found that when a newly thawed batch of doubly transgenic worms was used for an interval limited to 5-8 weeks, this had a favorable impact on the efficiency with which Mos1 mutant strains were recovered at the end of the procedure.Typically 20-25% of F6 worms were found to harbor at least one Mos1 insertion (Figure S1B).
Despite this improvement, the manual sorting of large numbers of doubly transgenic worms was still laborious.We therefore produced a new strain (IG444) carrying a transgenic array composed of the Mos1 transposase and the pcol-12::dsRed marker construct [9], which is associated with bright red fluorescence in the epidermis.Introduction of this new fluorescent marker significantly improved the ease by which the doubly transgenic worms (Figure S1C) could be identified under a fluorescence binocular microscope.Given the decrease in Mos1 mobilization that had been observed previously, and the problem of resident transposons in the starting strain described below, we mated IG444 with an outcrossed strain derived from EG1470, then established a large stock of doubly transgenic animals, which were passaged for a minimal number of generations before being frozen in multiple aliquots.We found that the strain showed a consistently high rate of Mos1 transposition.When the two improvements were implemented, the yield of mutants increased markedly (Figure 1A) when compared to the overall 6.2% rate previously reported [6].At the height of production, 5,927 strains were generated in a single month.By the end of the program, this procedure had been repeated over 150 cycles, and more than 300,000 lines of worms had been cultured over multiple generations and individually subjected to automated PCR-based analyses.As a result, more than 55,000 independent mutant strains were generated (Figure 1B).
Worms from each Mos1-containing strain were robotically cherry-picked from liquid culture in 96-well plates to the standard nematode solid medium in a 24-well format [6].As a quality control step in the production process, a small aliquot of worms was removed from 28 randomly chosen wells in different plates (typically 24) and assayed by PCR for the presence of a Mos1 insertion.In 100% of tests (n.300 wells), all strains yielded the expected Mos1 amplicon.These plates were then sealed to prevent any cross-contamination between wells and sent on a weekly basis from Marseilles, France, to Lyon, France and to Bristol, UK.There they were molecularly characterized, in order to identify the site of insertion of the Mos1 transposon in the C. elegans genome.A total of more than 55,000 clonal strains were dispatched, 46,473 strains were sent to Lyon and 8,615 strains to Bristol (Table 1).

Molecular characterization of Mos1 alleles
When the weekly batches of 24-well plates arrived in Lyon and Bristol, they were processed in a systematic manner similar to that previously described [5].An aliquot of the worms was clonally transferred to a 96-well plate, and DNA was released by lysis.The DNA obtained in this manner was then subjected to digestion with one of two alternative endonucleases, MboI or HaeIII.Digestion was followed by ligation and an inverse PCR reaction was performed using a pair of nested primers [10].The resulting PCR amplicons were analyzed by agarose gel electrophoresis.When a unique band above a threshold size was obtained (see Material and Methods), it was purified and sequenced.When multiple bands were observed, generally the strongest one was purified, reamplified and again purified before sequencing.For 34.2% of the strains it was possible to obtain a sequence using this method.For the remaining strains, a second attempt was made to generate a specific amplicon using the alternative enzyme, and sequences were thus obtained for an additional 19.9% of the strains.In total, more than 90,000 digestions and 300,000 PCR amplifications were carried out, leading to the generation of 26,547 PCR amplicons that were sequenced.
The sequence obtained for each PCR amplicon was compared by BLASTN to the C. elegans genome to identify the site of the Mos1 insertion.The insertion site could be unambiguously identified for approximately 50% of the sequences.Overall, a Mos1 insertion site could be identified in 13,334 independent lines, roughly a quarter of the 54,440 strains processed by the two sites (Table 1, Table S1).For reasons that remain unclear, we observed that the sequencing quality around the Mos1-genomic DNA junction was sometimes poor, which made it difficult to determine the insertion site of some alleles with precision (,+/210 base pairs).Taking this into account, and eliminating redundant alleles (see below), we molecularly characterized 10,858 new independent mutant alleles (Table S1).

Archiving of Mos1 alleles
The archiving strategies differed between the two sites.Due to the higher number of strains that were handled in Lyon (Table 1), it was decided to characterize molecularly the strains before freezing.Only those strains yielding a unique or major PCR amplicon were conserved.After removing an aliquot from each well for the molecular study, the remaining worms were allowed to grow and reproduce in the 24-well plates until the food was exhausted.At this stage, which generally occurred around 3 weeks after reception, worms from wells that generated a unique or major PCR amplicon were cryopreserved.More than 25,000 strains were frozen in Lyon (Table 1).The Bristol team opted to freeze half of the worms in deep well 96 well plates soon after receipt, and to process the other half for DNA extraction and Mos1 PCR amplification (Table 1).

Analysis of the distribution of Mos1 alleles across the genome
Bioinformatics analyses were conducted to analyze the distribution of the complete collection of Mos1 alleles and to evaluate its potential usefulness.As in some cases multiple redundant insertions were obtained, it was important to try to address whether this reflected a real bias of the Mos1 transposon to insert at specific sites, or whether this was the consequence of experimental artifact.Of the 2,476 redundant alleles, 2,239 were found to have an insertion site that exactly matched that of an allele in the non-redundant set of 10,858 alleles (Table S1).When we looked at the allele numbers of such matching pairs, which reflects allele isolation history, we found that in almost 90% (1,972) of cases, the allele numbers differed by less than 25, indicating that they were derived from the same, or closest, 24-well plates.This strongly suggested, as discussed below, that the great majority of redundant alleles arose from experimental artifact.We therefore limited further analyses to the non-redundant set of 10,858 alleles.In our pilot study with 914 alleles, we reported a bias for insertions on chromosome I and against chromosome V [5]; with our new set of 10,858 alleles, this skewed distribution was not detected and the number of Mos1 insertions found on each chromosome was proportional to the chromosome length (Figure 2A).The previously observed imbalance presumably represents a sampling bias.The average distance between neighboring alleles was 9,230 bp, with 33%, 67% and 95% of gaps being less than 3.2 kb, 10 kb and 30 kb, respectively.The single largest gap between adjacent alleles was less than 100 kb (Figure 2B, Table S1).There were, however, a few local ''hot spots'' for Mos1 alleles.The extreme right end of chromosome III, and especially of chromosome I, for example, had a dense distribution (Figure 2C, results not shown).But otherwise, on the scale of each individual chromosome, the spread of Mos1 alleles was relatively uniform (Figure 3).Among the 10,858 new alleles spaced at least 10 bp apart, 6,345 Mos1 insertions were found within 4,948 different genes, of which 4,586 were protein coding (Table S1).To test for any overall bias for insertion into intragenic or intergenic regions, we calculated the percentage of Mos1 insertions contained within coding  sequences (CDS) relative to the genome as a whole.Overall, 50.4% of insertions were in CDS (Table S1), which can be compared to the figure of 57.3% that we calculated to be the proportion of the genome that is in CDS.Mos1 inserts at a TA dinucleotide [8].There is a slight under-representation of TA (or AT) dinucleotides in CDS (55% of the genomic total by our calculation in WS220).Taken together, this suggests that there is no particular overall preference for insertion in intragenic or intergenic regions.The subset of genes that are expressed in the germ line, and so expected to be in an open chromatin context at the time when Mos1 insertion occurs, might be predicted to be better targets.We used the overlap of 2 large-scale datasets [11,12] to define a high-confidence list of 373 germline-expressed genes.Of these genes, 74 (19.8%) contained a Mos1 insertion.As this is not significantly different (p.0.2, binomial test) from the number expected by chance, one can conclude that there was no enrichment for Mos1 insertions in germline-expressed genes.We next evaluated the number of new Mos1 insertions that fell into the coding regions of exons, which could provoke a loss of gene function.There are 1,816 such Mos1 insertions, corresponding to 1,739 different genes.Again, these are distributed throughout the genome (Table S1).Overall, therefore, there appears to be little if any bias in the distribution of the Mos1 insertions across the genome.
It has already been reported that the NEMAGENETAG collection allows essentially all C. elegans genes (99.4%) to be targeted by MosDEL [4].The MosDEL method can effectively generate a deletion using a Mos1 insertion located up to 25 kb away from the targeted gene.The MosTIC technique, on the other hand, permits genomic sites within 500 bp of a Mos1 insertion to be modified efficiently, and can be used for targets up to 1.5 kb away [1].We therefore calculated the number and proportion of protein-coding genes within 1.5 kb of a Mos1 allele, including those generated in the pilot study (a total of 14,300).More than 40% of all protein-coding genes in the C. elegans genome (close to 8,500) are potential MosTIC targets (Table 2).It should, however, be pointed out that for many purposes, such as introducing a particular point mutation or inserting a sequence for a fluorescent protein or an affinity tag at the 59 or 39 end of a gene, researchers will want to target a specific region of a gene.So, the real utility of the collection requires evaluation on a case-by-case basis.

MosLocator: an online tool to aid Mos1 allele identification
The complete set of unique Mos1 alleles has been entered into the community database Wormbase [13].Those alleles that fall within exons or introns of annotated genes can be readily found on the corresponding gene page.On the other hand, any allele that falls into an intergenic region is not associated with a gene.Nonetheless, such alleles are potentially useful as the starting point for targeted genome engineering using MosTIC or MosDEL.Although all insertions can be seen using the Wormbase genome browser, or found with WormMart [14], these methods are somewhat cumbersome.This prompted us to devise a simple, but flexible tool called MosLocator that allows all the Mos1 insertions within a set distance of a gene to be identified with ease (Figure 4).Because of the structure of the underlying database, MosLocator requires sequence or transcript names as an input.These identifiers can be found in Wormbase, or by using the online resource Wormbase Converter [15] that allows any of the common C. elegans identifiers to be converted to the gene sequence identifier (Figure 4A).Alleles associated with any number of genes can be found with MosLocator (Figure 4B).The tool also offers the possibility of returning only alleles that are found within or close to exons.When a Mos1 transposon is found, the results window displays a direct link to the relevant page on WormBase (Figure 4C) that includes details of the allele and provides access to a graphical representation of its physical environment (Figure 4D).MosLocator is freely available at www.ciml.univ-mrs.fr/applications/MosLocator.

Recovery of specific Mos1 alleles from the collection
The 1,297 Mos1 insertion alleles that were characterized in Bristol (numbered from ttTi50009 and upwards) are identified as such in Wormbase and can be obtained by request to PK (email: p.kuwabara@bristol.ac.uk).The strains characterized in Lyon were originally maintained and distributed by the Segalat and Gieseler laboratories.They are now kept and distributed by a dedicated facility, ''Biology of C. elegans'' (UCBL CNRS UMS 3421).There are 12,942 strains available; 12,037 strains generated by the NEMAGENETAG consortium (ttTi1 to ttTi46473) and 905 strains generated during the pilot study (cxTi8901 to cxTi10968) [5].Conditions of distribution are detailed at the website http:// ums3421.univ-lyon1.fr.Briefly, researchers have the choice of asking for strains to be verified before being sent, or not.When verification is required, a specific pair of primers is synthesized that can be used to determine by PCR the presence of a particular Mos1 insertion, and worms are sib-selected to derive a pure homozygous line [10].This is reflected in a higher price and longer delay (currently, 250J and up to 3 months, versus 35J and 2-4 weeks for non-verified strains).At the time of writing, 792 Mos1 strains had been requested from Lyon.A total of 451 strains (57% of requests) were sent without molecular verification.Among the other 341 requested strains, in 268 (79%) the Mos1 element was found at the expected site, in 16 cases (5%) a viable strain was not recovered after thawing, and in 57 cases, the expected Mos1 insertion was not found at the expected position.
We cannot exclude mistakes in labeling at the freezing and/or thawing steps, due to human error, as being a possible cause of not finding the expected Mos1 insertion in more than 15% of cases in the Lyon collection.Another hypothesis, however, is that worms moved from one well where food was exhausted to another with food during the weeks that the 24-well plates were kept before worms were frozen.
To test this hypothesis, we randomly chose 15 ''missing'' insertions and thawed all the available strains that had been harvested and frozen from the 15 original 24-well plates.We then used our PCR approach to look for the specific ''missing'' Mos1 insertions.In 14 out of 15 cases tested, the sought-after Mos1 insertion was found in one or more strains of worms that came from the same original plate.As an extreme example, in addition to the expected Mos1 insertion, a single well was also found to be PCR-positive for three independent Mos1 insertions originally  assigned to different wells on the same plate, while a single tested Mos1 insertion was also detected in worms from as many as 11 wells of the same plate (data not shown).Although, as discussed below, this suggests that movement of worms from well to well was sometimes extensive, we would still expect to be able to recover more than 97% of requested Mos1 insertions.

Discussion
The collaboration of three European laboratories, in the context of the NEMAGENETAG consortium, has led to the generation of an extensive collection of individual Mos1 alleles that should be of great use to the C. elegans research community.Because of the previously reported problems in recovering Mos1 strains from frozen stocks, this project adopted a relatively laborious procedure involving the clonal culturing of strains for six generations with the aim of driving worms carrying Mos1 insertions to homozygosity.The inverse PCR protocol does not allow one to distinguish between animals carrying homozygous or heterozygous insertions, but calculations suggest that only a small percentage of strains will not be homozygous [6].Although requesting laboratories are asked to provide information on whether the strain they received was homozygous or heterozygous, the response has been sparse.We have, however, generated relevant data for a set of 200 strains, by genotyping 10 individual animals revived from a thawed stock, each associated with a particular Mos1 insertion.With the caveat that this represents less than 2% of the total number of NEMAGENETAG Mos1 insertion strains, and may not be representative, in 100% of cases an amplicon from the expected Mos1 insertion was obtained in one or more of the tested animals.In only 53 cases (26.5%), however, were 10/10 tested animals positive for the expected Mos1 insertion.That 10/10 Mos1 positive worms were not always found may reflect the fact that the single worm PCR technique does not always work, leading to false negatives.Additionally, it is likely that the movement of worms between wells that we described above also contributed.
It is also important to note that only one Mos1 insertion per strain was characterized, although additional insertions could be present, and in many cases more than one PCR amplicon was obtained after the first round of PCR amplification (unpublished results).At an intermediate point in the project, a further step was added, which involved sending all of the 24-well plates from Marseilles to the laboratory of J.L. Bessereau in Paris.Here, a small volume of buffer was added to each well to facilitate removal of an aliquot of worms in order to construct a pooled library (J.L. Bessereau, personal communication) before the plates were redispatched to their final destinations, Lyon and Bristol.This pooled library, Moslib, comprises more than 40,000 independent strains and will be the subject of a future publication.Results of an analysis of Moslib have indicated that each haploid genome contained on average 2.0 Mos1 insertions (J.L. Bessereau, personal communication), which corresponds to roughly 80,000 independent insertions in the library since there is still every reason to believe that most frozen strains were homozygous for any Mos1 insertions they might contain.This suggests that Moslib might contain up to 65,000 non-characterized insertions.Indeed, preliminary results suggest that a substantial number of previously uncharacterized insertions can be recovered from Moslib (J.L. Bessereau, personal communication).With advances in technology, these can be identified using methods based on next generation sequencing [16].
It can be seen from Table 1 that the two sites were roughly equivalent in their success in going from attempting molecular characterization of a strain to assigning a genomic position for an allele (24.6% and 25.9% for Bristol and Lyon, respectively).They attained this level in very different ways.While Lyon had substantially better success at the PCR amplification step than Bristol (54.1% and 26.5%, respectively), Bristol had a much better success rate at the sequencing step than Lyon (92.6% and 47.9%, respectively).This may reflect differences in reagent suppliers, or the competence of sub-contracted sequencing companies.As the disparities are large, it is unfortunate that direct side-by-side tests were not carried out to identify the important factors that influenced the results.With hindsight, one can only regret that overall, the project did not have the best of both steps, as combined this would have meant an overall 50% success rate, which would have potentially translated into a set of alleles twice the size of the current one.
Intronic Mos1 insertions are unlikely to be mutagenic as they can be removed by splicing [8].Furthermore, most strains are expected to be stable, and would not be expected to exhibit any gross developmental defects, as this would have been counterselected during culture.It will be interesting to learn from requesting laboratories whether this is always the case, when an insertion is found within an exon of a gene previously associated with a visible phenotype.
Early in the project, multiple identical Mos1 insertions were found in worms that were derived from supposedly independent lines.Thus, 280 of the 423 redundant alleles characterized in Bristol corresponded to just 2 insertion sites.The most frequent was also characterized repeatedly in Lyon; more than 350 alleles all with the same insertion site (I:12456295..12456296, +/2 10 bp) were found in total.Similarly, 239 and 26 insertions at the second site (IV:1136537..1136538, +/2 10 bp) were characterized in Bristol and Lyon, respectively.As multiple strains carrying the same allele were found at both laboratories, the problem must have arisen in Marseilles.This could have occurred if the double transgenic starting strains accumulated Mos1 insertions at a low frequency even in the absence of forced expression of the recombinant transposase.Although this was not expected, and had not previously been reported (JL.Bessereau, personal communication), PCR analysis of non-transgenic progeny from the starting strain did reveal the presence of at least one resident Mos1 insertion.To prevent the problem from reoccurring, the non-transgenic progeny from the starting strain were systematically checked for the absence of Mos1 insertions before their transgenic siblings were subject to heat-shock.Once this technical (A) MosLocator (www.ciml.univ-mrs.fr/applications/MosLocator)finds Mos1 alleles using gene sequence or transcript names.For large lists of genetic gene names, the gene sequence or transcript names can be obtained using WormMart, or here, using WormBase Converter (www.ciml.univ-mrs.fr/applications/WB_converter)[15].In the example shown, the 23 ptr genes were used as input.(B) Screen grabs were captured to illustrate the use of MosLocator.Left panel: a list of sequence names was entered, and the search parameters were defined.Upper right panel: a display of the output for this search.Clicking on a non-zero number displayed in either of the last two columns, for example the ''2'' associated with the gene T21H3.2(ptr-16), generates the display shown in the inset.This is a list of the 2 Mos1 mutant alleles that are found within the gene T21H3.2.Each allele name is hyperlinked to Wormbase.(C) A partial view of the Variation report for the Mos1 allele ttTi21065 found on chromosome V at Wormbase (version WS225).(D) The genomic environment of the ttTi21065 allele is displayed.The figure is a screen-grab from Wormbase.doi:10.1371/journal.pone.0030482.g004problem had been overcome, any remaining minor redundancy was generally restricted to a single laboratory.Only 2 Mos1 insertions sites out of the 1278 characterized redundantly in Lyon were also found in Bristol.As this laboratory-specific redundancy was also generally confined to single 24-well plates, it is likely to have arisen from cross-contamination during molecular characterization of the worms, such as might result if worms moved from one well to another.This is expected to occur more often when worms are starved, and food can be found in an adjacent well.Otherwise, as in any large-scale project, sample-handling errors will have also contributed to this small problem.While the production procedure in Marseilles was almost fully automated [6], the downstream molecular characterization, sequence analysis, and archiving involved intensive human intervention.Given the amplitude of the project, dealing with hundreds of thousands of samples, and the typical rate of human error, which experience suggests is in the order of several percent, future endeavors would clearly benefit from investment in the development of fully automated procedures.This is especially true as towards the end of the project, production capacity in Marseilles exceeded the capacity of the other partners.
In our analyses, we chose to be conservative, and only considered alleles more than 10 bp apart to be unique.The figures presented may therefore be a slight under-estimate of the number of unique strains obtained.Regardless, the generation of more than 10,000 characterized transposon alleles will undoubtedly be a boon to the C. elegans research community.

Inverse PCR and sequence analysis
An aliquot of each well from the 24-well plates was removed by washing with 10 ml of worm lysis buffer (10 mM Tris-HCl, pH 8.3; 50 mM KCl; 2.5 mM MgCl 2 ; 0.5% Tween-20, and 100 mg ml 21 proteinase K) and transferred to wells in 96-well plates.Worm lysis was performed by freezing the plates at 280uC for at least 30 min, then incubating them for 60 min at 60uC, followed by 15 min at 95uC.Lysates were digested for 2 hours by one of the following enzymes: HaeIII or MboI.The enzymes were then heat inactivated and the digested DNA ligated using standard protocols.Then 3 ml of each ligation reaction was used as substrate in a PCR reaction with primers oJL103-oJL114 [8].A nested PCR was performed on 0.03 ml of the first reaction, using primers oJL115-oJL116 [8,10].PCR products were analyzed on a 1.8% agarose gel.When a unique band was seen, with a size greater than 270 bp for the MboI digestion protocol or 325 bp when using HaeIII, the product was purified.When multiple bands were seen, the strongest one was re-amplified, applying the same size limits as above to ensure that only amplicons containing a Mos1 insertion and flanking genomic DNA were sequenced.The sequence of each PCR product was compared to the C. elegans genome by BLASTN at the NCBI.Strains that gave PCR products for which the BLAST result was ambiguous, for example when the sequenced PCR fragment matched several genomic regions, were not characterized further.

Cyropreservation
Starved worms were washed off wells with M9 and transferred to 1.8 ml cryovials (Nunc).Three vials were prepared from each well.They were frozen following standard protocols [17].For permanent storage, one cryovial was kept in a 280uC freezer, and the other two were stored in liquid nitrogen.Alternatively, worms were mixed with freezing buffer and transferred to deep-well tubes in a 96-well format and frozen at 280uC in a Styrofoam container to slow the rate of freezing.

Bioinformatic analyses and MosLocator
Because of changes in database structure at mining.wormbase.org,it was decided to develop a stand-alone tool to locate Mos1 insertions relative to any given gene.A local MySQL database based on the publically available genome feature tables (from ftp://ftp.sanger.ac.uk/pub2/wormbase/releases/WS225/genomes/ c_elegans/genome_feature_tables/GFF2/c_elegans.WS225.gff.gz) was established.The graphic interface was coded in PHP and Javascript.Submitted gene sequence or transcript names were sent as MySQL queries to the database to obtain genomic coordinates (entire gene and exons, if appropriate).These were then in turn compared to the genomic coordinates of the complete set of Mos1 insertions present in a static file derived from WS225.The code is available on request.
To determine the potential genome coverage of the Mos1 alleles for MosTIC, using WS225, an R script was written to concatenate any contiguous sequence within 1.5 kb up-or down-stream of each allele.This was used to produce the graphic representation in Figure 3.A second R script counted the number of genes that were contained within, or overlapped with the Mos1-associated contigs.These scripts and the list of contiguous regions are available on request.

Supporting Information
Figure S1 Optimizing the yield of Mos1-containing strains.The percentage of F6 lines found to contain at least one Mos1 insertion by PCR varied as a function of the interval between the thawing of the starting double transgenic strain and the mobilization of the Mos1 transposon by heat shock.(A) A typical fluctuation for a batch of worms that was used for 8 consecutive weeks after thawing.(B) The results obtained with 3 successive batches (indicated by the different colors) that were used only between weeks 5 and 8 after thawing.(C) A doubly transgenic animal resulting from the cross of the strains IG358 and IG444; red and green fluorescence were visualized simultaneously.(PDF) Table S1 NEMAGENETAG Mos1 alleles.The complete list of the 13,334 alleles characterized during the NEMAGENETAG project.The genomic position of each insertion and its Wormbase reference (Var ID) are derived from the frozen release WS220.A set of 10,858 non-redundant alleles that are at least 10 bp apart is highlighted in color: in yellow when the Mos1 insertion is contained within a coding exon, and otherwise in green.For this non-redundant set, the number of genes that are directly targeted by each allele is given, together with a corresponding GeneID, derived from the frozen Wormbase release WS220, as well the distance to the neighboring allele (going in a 59 to 39 direction along each chromosome) and, in the last column an indication of whether an allele is contained within a gene.There are 940 cases where there is a lack of concordance between this last column and the columns ''Genes hit'' and ''Targeted Gene Name''; this reflects differences in data structure between the data sources used.Where multiple alleles exist for a given genomic position, the choice of allele was entirely arbitrary, and similarly for the 277 alleles that target more than one gene, the choice of the single GeneID displayed was arbitrary.(XLS)

Figure 1 .
Figure 1.Production of the Mos1 mutant collection.(A) A comparison of the efficiency of production of Mos1 mutant alleles, generated with independent heat-shocks and measured with successive weekly batches of strains, starting with worms obtained through the mating of EG1470 and EN547 (blue bars), or IG358, an outcrossed strain derived from EG1470, with the new transgenic strain IG444 (red bars).(B) Graph showing the cumulative total of strains produced over the last 10 months of the project.doi:10.1371/journal.pone.0030482.g001

Figure 2 .
Figure 2. Distribution of Mos1 alleles.(A) Graph showing the relationship between chromosome length (as a percentage of the whole nuclear genome) and the proportion of Mos1 alleles per chromosome reported in a previous study [5], and the 10,858 alleles obtained in the current project (black and red circles, respectively).The outliers, concerning chromosomes I and V, from the previous study are highlighted with lines.(B) Distribution of distances from one Mos1 allele to the next, in a 59 to 39 direction along each chromosome.The graph shows the cumulative percentage of alleles that are separated by less than the indicated distance.(C) Concentration of Mos1 alleles at the extreme right end of chromosome I (length 15,072,423 bp).The separation of the allele numbers indicates that almost all the alleles were generated independently, except in two cases (ttTi2276 and ttTi2284; ttTi13453 and ttTi13460), highlighted by an asterisk.This region was also preferentially targeted during the previous study as reflected by the presence of several cxTi alleles.doi:10.1371/journal.pone.0030482.g002

Figure 3 .
Figure 3. Genomic coverage of Mos1.Graphical representation of each C. elegans chromosome showing the regions of the genome that are potentially amenable to genome engineering using the publicly-available Mos1 alleles; it is assumed that any point up to 1.5 kb away from a transposon-insertion site can be targeted.The bottom line is a magnified view of the boxed region on chromosome X.doi:10.1371/journal.pone.0030482.g003

Figure 4 .
Figure 4. Finding Mos1 alleles with MosLocator.(A) MosLocator (www.ciml.univ-mrs.fr/applications/MosLocator)finds Mos1 alleles using gene sequence or transcript names.For large lists of genetic gene names, the gene sequence or transcript names can be obtained using WormMart, or here, using WormBase Converter (www.ciml.univ-mrs.fr/applications/WB_converter)[15].In the example shown, the 23 ptr genes were used as input.(B) Screen grabs were captured to illustrate the use of MosLocator.Left panel: a list of sequence names was entered, and the search parameters were defined.Upper right panel: a display of the output for this search.Clicking on a non-zero number displayed in either of the last two columns, for example the ''2'' associated with the gene T21H3.2(ptr-16), generates the display shown in the inset.This is a list of the 2 Mos1 mutant alleles that are found within the gene T21H3.2.Each allele name is hyperlinked to Wormbase.(C) A partial view of the Variation report for the Mos1 allele ttTi21065 found on chromosome V at Wormbase (version WS225).(D) The genomic environment of the ttTi21065 allele is displayed.The figure is a screen-grab from Wormbase.doi:10.1371/journal.pone.0030482.g004