Generation of a BAC-based physical map of the melon genome

BAC fingerprinting and contig assembly

The final physical map [Table 1] had 1,355 contigs and 441 singletons, with an estimated physical length of 407 Mb (0.9 × coverage of the genome). The longest contig was 3.2 Mb long; the average contig length, 300 kb; 40% of the contigs were made up of the assembly of more than 9 clones; 84% of the contigs contained between 3 and 24 clones, and 62 contigs contained less than the maximum number of Q-clones, 5%. This represents a small percentage of Q-contigs and was the result of the forced breaking of all contigs with more than 5% of Q clones in the first stage of construction. This was achieved by reducing the cutoff to values as low as 1e-99 where necessary. Although this increased the number of contigs and unassembled clones, the reliability of the resulting contigs was greatly improved.

These figures are comparable to those of other plant physical maps recently described. For example, the physical map of B. rapa, a species with a genome size of 550 Mb, similar to that of melon, was built by fingerprinting 67,468 BAC clones (×15 genomic equivalents). It has 1,428 contigs of average length 512 kb, 57% of which are made up of more than 9 clones, and the estimated genomic coverage of the map is 1.3 × (725 Mb) [34]. However, the number of singletons (14,001) is more than 30 times higher than that in our map, even though the iterated procedures used to build both maps were similar. The fact that the B. rapa library represents 2.6 times more genomic equivalents than the melon could partially explain the differences in singleton number. The analysis of the number of Q clones after each round of construction at different cutoffs suggests that clones remaining as singletons in the B. rapa map may not just be due to low quality fingerprints but may come from regions of low coverage in the BAC libraries used [34]. This means that the differences in number of singletons on these physical maps could also be due to differences in the representation of the B. rapa and C. melo genomes in the different libraries.

As another example, the physical map of papaya, a species with a genome size of 372 Mb, slightly smaller than that of melon, was built by fingerprinting clones representing 13.7 genomic equivalents. It has 963 contigs, 59% of which are made up of more than 9 clones, with an estimated genomic coverage of 0.96 × [36]. The difference in genomic coverage of the libraries could again explain the higher number of singletons, 4,358, ten times higher than in our map.

Regarding the internal structure of the C. melo FPC contigs, the comparison of ordered lists of contigs, based on the contig physical length or the contig size (number of clones belonging to the contigs), revealed a high proportion of 'stacked' contigs, that is, contigs containing regions where the depth far exceeds the estimated coverage of the library used (×5.7), possibly due to the non-randomness affecting all libraries constructed by one enzyme restriction of genomic DNA. This poses a problem in that many clones in these stacked contigs do not contribute to extending the physical length of the contig and their fingerprints carry no new information. Also, the visual inspection of contigs revealed some assembly artifacts affecting several contigs. For example, of three contigs containing more than 100 clones, the largest (Ctg148, 185 clones) most probably contains a large proportion of wrongly assembled clones. The sequence of several BAC-ends (BES) of clones from this contig revealed tandem repeat sequences of DNA in the form of ribosomal RNA genomic regions [data not shown]. Similar results have been obtained with other repetitive sequences, such as retrotransposons, analyzing BES from other problematic contigs.

Anchorage of the BAC library to the genetic map

The average number of BACs amplified per marker was 4.8. This is lower than expected based on the estimated genomic coverage of our library (5.7). This is a further indication of the absence of several genomic loci or overestimation of representation in the library used. A distribution of the number of positive BACs found per marker shows a two-zone distribution with about 79% of markers evenly distributed in the 1-6 BAC/marker range while the remaining markers are linked on average to 9-10 BAC/marker, probably reflecting the fact that about 20% of the primers used amplified duplicated or closely related genomic sequences [Figure 1]. In fact, while only 15% of the SNP markers were linked to contigs of more than six BAC clones, 26% of all anchored RFLPs were linked to between 7 and 14 clones. This indicates that the RFLP-derived primers were twice as prone to amplifying duplicated sequences as the SNP-derived primers. While an average of 5.4 clones were linked to RFLP markers, only 4.0 were linked to SNPs. Based on these results, it can be tentatively assumed that the genomic coverage of the library is somewhere in the 4-5 range.

Integration of the physical and genetic maps

The anchorage of BAC clones to the genetic map was used to establish a link between the fingerprint-based physical map and the genetic linkage map. To this end, information regarding anchored genetic markers and their chromosome locations was introduced in the FPC database for all anchored BAC clones successfully fingerprinted. As a result, 169 genetic markers were anchored to the physical map, with 157 of them linked to FPC contigs of several lengths while 12 linked only to singletons [Tables 1 and 2]. Figure 2 shows an example of an FPC contig anchored to a melon linkage group using information derived from the genetic map. One hundred and fifty-eight of the anchored FPC contigs were linked to just one marker, 11 to two markers, and two contigs to three markers. All but 30 genetic markers mapped a single FPC contig or singleton: 25 markers mapped two separate contigs/singletons, 4 markers were each linked to three FPC contigs/singletons and one marker to four contigs. Therefore, 18% of all markers anchored to the physical map were linked to more than one FPC contig/singleton, a figure in accordance to the above estimation of the number of primers amplifiying duplicated or closely related genomic regions.

In all, 2,215 fingerprinted BAC clones, distributed in 183 contigs/singletons and representing 55 Mb or 12.2% of the melon genome, have been positioned in unique loci of the genetic map (except for seven contigs/singletons associated to markers that map to two separate chromosome locations, and so cannot be assigned to a single locus). On average, 175 BAC clones, 15 contigs and 4.4 Mb have been anchored to unique loci for every C. melo linkage group. All information regarding contig estimated length, contig clone number and chromosome location of all FPC contigs or singletons anchored to genetic markers can be found in Additional file 2 Table S2. A representation of the anchored genetic map together with information on which markers are linked to FPC contigs is shown in Figure 3&4, while more detailed information regarding linkage group distribution of the type and number of genetic markers analyzed and linked to both physical and anchored genetic maps is given in Table 3. The resulting genetic-anchored melon physical map FCP database is available for download at http://melonomics.upv.es/static/files/public/physical_map/.

Validation of physical map contigs

The comparison between PCR and FPC contigs served as a quality control of the physical map building. Column 8 in Additional file 2 Table S2 shows the number of clones of every PCR contig anchored to any genetic marker while column 9 shows how many of those clones are also present in the FPC contigs anchored to the correspondent marker. Bearing in mind that an estimated 20% of all primer pairs tested produced positive BACs belonging to two or even three separate genomic regions and that, accordingly, 18% of all FPC-anchored markers are linked to more than one FPC contig or singleton, the degree of coincidence between PCR and FPC contigs has been computed using data from markers linked to a single FPC contig/singleton. As an average, 75% of those clones belonging to the same PCR contig are predicted to overlap according to the FPC information. Therefore, as 20% of the library clones were not successfully fingerprinted and so are absent from the physical map, we estimate that the degree of coincidence between PCR and FPC contigs is around 80%. However, when distinguishing duplicated or closely related genomic regions or gene families, the PCR screening procedure used to anchor the BAC library to the genetic map - i.e. PCR amplification of regions 100-1,000 bp long - should be much less efficient than fingerprinting whole BAC genomic regions. This can account for several of those clones that belong to PCR contigs but are absent from the correspondent FPC contig and, if so, the degree of coincidence between our physical and anchored genetic maps would be higher than the above estimation.

As another way to validate the physical map, the linkage group position of those markers linked to a common FPC contig was compared. As described above, of all FPC contigs/singletons linked to genetic markers, only 14 were anchored to more than one marker each. Of these, Ctg6, Ctg7, Ctg25, Ctg26, Ctg91, Ctg118, Ctg128, Ctg140 and singleton Cm19_G01 [Additional file 2 Table S2] are each linked to two or three markers that have the same position in the genetic map [Figure 3&4, markers in black-edged squares], which validates these FPC contigs as good assemblies. Two additional contigs, Ctg150 and Ctg147, were each linked to pairs of markers with conflicting genetic positions. An analysis of the construction of each contig revealed the merging of previous smaller contigs during the final steps of the autobuild process at greater cutoffs/lower stringency (1e-20 and 1e-15), which explains the presence of incompatible markers in these contigs. Another contig, Ctg898, is also linked to a set of two incompatible genetic markers. However, in this case, the clones belonging to the corresponding PCR contig coincide, and so they probably map two separate genomic regions that share some degree of sequence identity. The last two contigs, Ctg148 and Ctg 149, are also linked to conflicting sets of genetic markers (both markers linked to Ctg148 map to linkage group XI but at genetic distances much greater than the physical length of the contig), but this cannot be explained by low stringency automerge.

Source BAC library

A Bam HI BAC library from the double-haploid melon line 'PIT92' was previously developed in our laboratory [10]. With 23,040 BAC clones distributed in sixty 384-well plates, an average insert size of 139 kb and 20% empty clones, the library represents 5.7 genomic equivalents of the haploid melon genome.

Choice of genetic markers for PCR anchoring

Two-hundred and fifteen genetic markers from the PI 161375 × T111 melon genetic map were selected for anchoring to the physical map. Markers were evenly distributed along the 12 linkage groups. One hundred and seventeen RFLPs, 96 SNPs and 2 SSRs, previously described [20, 45–49], were selected (Additional file 1 Table S1).

BAC library 3D pooling and PCR anchoring

The 60 library 384-well plates were replicated in 96-well plates, with 4 clones in each well and 200 μl 2 × LB medium. After overnight growth, a sample of 50 μl was taken from each well from one plate, added to 2.5 ml LB containing chloramphenicol at 12.5 μg/ml and grown overnight at 37°C. DNA minipreps of the bacterial culture was as described by [50], resulting in 60 DNA pools, each containing 384 BAC clones. DNA from 5 pools was combined to give 12 DNA superpools. The remaining bacterial culture in the 96-well plates was used for row and column DNA pools for each plate. Each row pool contained 48 BAC clones while each column pool contained 32 BAC clones.

A pair of specific primers was designed for each genetic marker and a first round of PCR, using 0.5 μl BAC miniprep, was performed with the 12 DNA superpools, followed by a second with the DNA pools from the positive DNA superpools, and a third with the 12 column pools and 8 row pools from each positive plate pool. The clone coordinates were those of clones in the reduced 60 × 96-well library. An additional round of PCR established the final coordinates in the original BAC library for each positive. Positive controls using genomic DNA from the PIT92 parental lines (PI161375 and T111 'Piel de Sapo') were included in each round of PCR. The final amplified bands were sequenced using one of the primers used for PCR amplification and the sequences compared to that of the genomic markers analyzed to confirm the positives.

BAC DNA isolation and fingerprinting

Four μl of each BAC clone from a 384-well plate was inoculated into a 384-well plate containing 70 μl 2 × LB plus 12.5 μg/ml chloramphenicol. Plates were covered with adhesive gas permeable seals (Thermo Scientific) and incubated at 37°C, 250 rpm for 21-24 h. The following day, 15 μl of each BAC clone from the preinoculated 384-well plate were inoculated into four 96 deep-well plates, containing 1.2 ml LB plus chloramphenicol, and grown at 37°C, 300 rpm for 16 h. BAC DNA was obtained using a modified alkaline method followed by purification using the Microplate Unifilter+Uniplate Devices from Whatman Inc. DNA was resuspended in 30 μl of autoclaved milliQ water, with a typical final yield around 0.8-2 μg of DNA. Ten μl of the miniprep DNA was then digested using Bam HI, Eco RI, Nde I, Xba I and Hae III enzymes (New England Biolabs) for three hours at 37°C, and the mixture of restriction fragments labeled using the SNaPshot Multiplex kit (Applied Biosystems). The labeled fragments were precipitated by adding chilled 95% ethanol and resuspended in 10 μl of water prior to a final purification step using genClean plates (Genetix). A mixture of 9.95 μl HiDi™ (Applied Biosystems) and 0.05 μl 500 LIZ^® marker (Applied Biosystems) was added to each sample. The DNA was denatured by heating for 2 min at 95°C and then loaded in an ABI3730 DNA sequencer (Applied Biosystems) for fragment separation by capillary electrophoresis, using the Genescan LIZ500 marker as an internal standard for fragment sizing. The electrochromatograms were analyzed using GeneMapper 3.5 (Applied Biosystems).

Fingerprint data collection and physical map construction

A text file containing area, height and size data was exported from GeneMapper and background, poor quality fingerprints, vector bands and empty clones removed using the FPB software [44]. The FPB parameters were as follows: minimum bands, 40; tolerance, 0.4; multiplying factor, 30; peak width, 15; band sizes from 50 to 500; size per color, from 5 to 250; color background, 50; fixed threshold, 500; first and last values, 3 and 7; low index: 60; color offset, 0, 15,000, 30,000 and 45,000.

The FPC processed data were then assembled using the FPC v9.3 Software [[25]; http://www.agcol.arizona.edu/software/fpc/]. Tolerance was set at 12 (= 0.4 × 30) and the contigs assembled with the Best Contig parameter set at 100. Several preliminary assemblies were performed in order to determine the optimal cut-off value, using cut-off values of 1e-35, 1e-45, 1e-55 and 1e-65, minimizing the number of contigs without significantly increasing the number of Q-clones. The building of the physical map was according to a standard iterated procedure [32, 34]: initially at a sufficiently stringent cutoff to ensure the validity of the contigs, then gradually merging the contigs at successively greater cutoff values. Based on the results, the first automatic assembly was performed using a Sulston score of 1e-45. This was followed by breaking contigs containing more than 5% of Q-clones, applying the DQer function. The resulting contigs were end-merged using the 'End to End' function, setting the 'FromEnd' parameter to 50, 'Match' to 2 and the cutoff value to 1e-40. Finally, singletons were added to contig ends using the 'Singles to End' function. In several additional rounds of End to End/Singles to End the cutoff values were set to 1e-35, 1e-30, 1e-25, 1e-20 and 1e-15, with the DQer function applied when necessary after each round to break up those contigs containing more than 10% Q-clones.