A strategy involving the use of high redundancy YAC subclone libraries facilitates the contiguous representation in cosmid and BAC clones of 1.7 Mb of the genome of the plant Arabidopsis thaliana

Ian Bancroft1*, Karina Love1, Elisabeth Bent1, Sarah Sherson2, Clare Lister1, Christopher Cobbett2, Howard M. Goodman3 and Caroline Dean1

1. John Innes Centre, Norwich Research Park, Colney, Norwich NR4 4UH, UK
2. Department of Genetics, University of Melbourne, Parkville, Victoria 3052, Australia
3. Department of Molecular Biology, Massachusetts General Hospital, Boston, Massachusetts 02114
* Tel. + 44 (0)1603 502200
Fax. + 44 (0)1603 505725
e-mail bancroft@bbsrc.ac.uk


A region of over 1.7 Mb of Arabidopsis thaliana chromosome 4 has been assembled into a detailed clone contig. Attempts to assemble contigs using only clones from genomic cosmid libraries left numerous gaps in coverage. To achieve complete coverage, we developed a strategy that involves the subcloning of gel-purified YACs into highly redundant cosmid libraries. Using such libraries, contig assembly and restriction mapping can be combined into a simple routine, involving little hybridization analysis. The cosmid vector used is designed such that the cloned DNA can be inserted into plant genomes by Agrobacterium-mediated transformation. Thus the contigs assembled represent a resource for systematic gene identification by complementation analysis. Other types of clones, such as Bacterial Artificial Chromosomes, can be integrated into the strategy. This strategy was used to provide the clones for the first large-scale sequencing of the Arabidopsis genome and should prove valuable for the detailed analysis of genomes for which YAC-based physical maps provide the most complete coverage.


The plant Arabidopsis thaliana (Arabidopsis) has been widely adopted as a model for molecular genetic analysis of plant development and metabolism. It has a small genome which, based on currently available data (1,2) we estimate consists of around 105 Mb of low copy DNA and 25 Mb of clustered repeat sequences. The low copy regions of two of the five chromosomes have been mapped into yeast artificial chromosome (YAC) contigs (2,3) and the YAC contig mapping of the remaining three chromosomes is almost complete. Cosmid-based maps provide much greater resolution than YAC-based maps due to the smaller sizes of the inserts in these clones. Detailed maps are required for gene cloning or transcript mapping experiments involving genomes with high gene densities. Large-scale sequencing of the genome of Arabidopsis is under way, and will provide the first complete genetic blueprint of a typical flowering plant. Although some preliminary investigations have been performed (4,5), it has not yet been demonstrated that sequencing templates can be derived directly from a YAC if completely contiguous sequence is required for megabase-scale regions at optimal cost efficiency. Thorough analysis of the collinearity of YACs with the genomic DNA of Arabidopsis, as for most organisms, is also extremely difficult. For sequencing, it is therefore preferable to convert the YAC-based clone maps of the Arabidopsis chromosomes to maps consisting of clones propagated in E. coli. A cosmid-based physical map of the genome of Arabidopsis, consisting of ca. 750 contigs is available (6). These contigs were estimated to contain 90 to 95% of the genome of Arabidopsis, but most of the contigs are not integrated with either the genetic or YAC-based maps. The investigation reported here demonstrates that this resource does not allow complete contiguity to be achieved using these clones, even when supplemented with further genomic cosmid clones. Other E. coli-based cloning systems have recently been used to propagate Arabidopsis DNA, including Bacterial Artificial Chromosomes (BACs, 7). In order to obtain complete coverage of large regions of the genome of Arabidopsis, we have developed strategies that have allowed us to construct a contig of over 1.7 Mb, composed mostly of cosmids, but including some integrated BAC clones. This contig corresponds to a well-characterised region of the YAC-based physical map of chromosome 4 and provided the tiling path of clones for genomic sequencing of this region (8).


Cosmid identification

Cosmids from the region to be represented were identified by three different strategies, all of which utilize the available YAC-based physical map. The preferred source of cosmids is the set of contigs already assembled by a conventional fingerprinting approach (6). The libraries assembled into contigs were constructed using vectors based on pLorist (9). We prefix these clones with "G" (after the Goodman laboratory, which supplied the clones). Most of the contigs in the target region had been identified during the assembly of the YAC-based physical map of chromosome 4 (2). We completed the search by hybridizing gel-purified YACs, selected from a 1.2 Mb region of the physical map, to dot-blot filters supplied as a resource to the Arabidopsis research community. In total, cosmid clones were identified corresponding to approximately 410 kb (34%) of the 1.2 Mb target region. The location of this region, along with regions subject to other analyses described later, is shown in Figure 1.

A second strategy was used to identify cosmids representing an 800 kb contiguous region (see Figure 1). Gel-purified YACs were used as hybridization probes to a complete library of Arabidopsis genomic DNA. This library was constructed using vector pLAFR3 (10). The clones from this library are prefixed "CC". Cosmids were also identified from a library constructed using vector pOCA18 (11) by hybridization with YAC end probes and subsequent chromosome walking. These clones are prefixed "cAt". Vectors pLAFR3 and pOCA18 are both based on a very large replicon, pRK290 (12) and produce libraries with very stable inserts. The sizes of inserts that can be cloned in these vectors (approximately 15 to 25 for pOCA18 and 20 to 30 kb for pLAFR3) is smaller than can be cloned in the pLorist-derived vectors (35 to 45 kb). Clones totalling 350 kb of coverage (44% of the region) were identified, in addition to 350 kb (44%) already identified in assembled G clone contigs. These clones assembled into eight contigs. Additional libraries were screened, including libraries of bacteriophage lambda clones, but contig links were not found.

In order to complete coverage of this 800 kb region, we found it necessary to subclone YACs selected from the physical map. To do this we developed a novel protocol that involved preparative gel-purification of the YACs, partial digestion with TaqI and cloning into a new cosmid vector called pCLD04541. This vector, shown schematically in Figure 2, is a binary T-DNA vector, so the inserts can be stably introduced into plant genomes using Agrobacterium-mediated transformation. Contigs assembled in this vector can therefore be used in complementation analyses as part of positional cloning experiments. The subclone libraries generated typically represent 20- to 60-fold redundant coverage of the subcloned YAC. Subclone cosmids completed the final 100 kb (12%) of the first 800 kb region assembled.

Initial cosmid contig assembly and restriction mapping strategy

The G cosmid clones were already assembled into contigs. The cAt and CC clones that had been identified by colony hybridization were confirmed by hybridization of Southern blots of these clones with gel-purified YACs. Contigs were assembled on the basis of matches of common-sized restriction fragments (CC clones) or by chromosome walking using restriction fragments of cosmids as hybridization probes to the original libraries (cAt clones). These contigs were confirmed and integrated with G cosmid contigs by hybridization to Southern blots.

Where gaps existed between the integrated G, CC and cAt clone-containing contigs, chromosome walking was performed within the YAC subclone libraries. This involved the gel-purification of restriction fragments from the inserts of terminal cosmids in the assembled contigs and hybridization of these to gridded array colony filters of the subclone library to identify overlapping clones. The whole insert, or selected restriction fragments from the inserts of these clones were then gel-purified and again hybridized to the subclone library filters. This process was repeated, from both ends of each gap, until the walks reached common clones and complete coverage was achieved.

Two cosmids, CC33M3 and CC16N19 had initially been thought to overlap as they hybridized to each other. Sequencing of both clones revealed that they did not overlap, but shared highly homologous sequences. These cosmids were used to identify several BAC clones, including IGF5D3 and IGF1P20, which overlap them both. (The IGF BAC library was obtained from Thomas Altmann, MPI fuer Molekulare Pflanzenphysiologie, Golm, Germany.)

For most applications it is necessary to construct restriction maps of the clones in the assembled contigs. This allows the accurate assessment of the overlaps between clones and the size of the contig, accurate placement of transcripts, probe fragments, etc. It also facilitates the analysis of the collinearity of the contig with the Arabidopsis genome. The restriction mapping of the clones was conducted using a combination of conventional approaches: permutations of multiple enzyme digestion, partial digestion strategies and hybridizations to Southern blots.

Modified cosmid contig assembly and restriction mapping strategy

Chromosome walking within YAC subclone libraries in order to fill gaps between contigs of genomic cosmid clones proved to be an effective but laborious process. Repeated sequences were encountered both during the identification of genomic clones and whilst walking within subclone libraries. These caused two problems. Repeats present in multiple genomic locations caused the identification of cosmids that were not from our target region and clustered repeats lead to contigs being incorrectly assembled by our hybridization-based approach. After completion of the first 800 kb contig, a new strategy was developed. Instead of making contigs of G, CC and cAt clones, followed by walking within YAC subclone libraries, contigs of subclone cosmids were assembled first and other clones identified were aligned with these later.

The YAC subclone libraries were gridded at low density and replicas hybridized with several different probes. These included gel-purified DNA of the subcloned YAC (usually TaqI partially digested DNA of a size smaller than that used for library construction) and fragments of the YAC cloning vector, pYAC4, from positions very close to the YAC cloning site in either the left or right vector arm ( to identify cosmid subclones containing the ends of the YAC insert and a part of the YAC vector). Since gel-purified YAC DNA is always contaminated with yeast DNA, pulsed field gel-recovered yeast DNA was also used as a hybridization probe. Cosmid clones could then be grouped by hybridization with gel-purified DNA from several YACs that overlap the subcloned YAC. This analysis also confirms that the subclone cosmids are genuinely from the target genomic region. If the subcloned YAC had been chimaeric or misidentified, some or all of the cosmid subclones would fail to hybridize with overlapping YACs. DNA was prepared from a sufficient number of cosmids to permit four- to 10-fold redundant coverage of the YAC. This DNA was digested with a combination of the restriction endonucleases HindIII and SalI to give a "fingerprint" of restriction fragments revealed by agarose gel electrophoresis and staining with ethidium bromide. Putative partial matches in fragment content were identified by visual inspection of the fingerprints and repeat digests run on agarose gels in predicted clone order for confirmation. Where there was insufficient fingerprint data to allow complete contig assembly, for example due to under representation of some regions, or too few restriction fragments, gel-purified inserts of clones from the ends of contigs were used as hybridization probes to the colony filters to identify overlapping clones. We used this procedure to investigate a ca. 400 kb region and obtained complete coverage in subclone cosmids.

For some applications, such as large-scale genomic sequencing, cosmids constructed using vector pCLD04541 are inappropriate as the major resource. This is because the vector is large (ca. 29 kb) and the inserts consequently relatively small (on average ca. 17 kb). Other types of clones can easily be aligned with the subclone contig via their HindIII plus SalI restriction digest fingerprints. An example of the alignment of a pLORIST-derived cosmid relative to cosmids representing a portion of the subclone contig is shown in Figure 3. The comparison of restriction digest fingerprints of clones derived independently from the genome allows also the analysis of collinearity of those clones with the genome. If the sets of fragments identified match exactly, it is highly likely that the clones both faithfully represent that portion of the genome. Differences would indicate a rearrangement in one of the clones.

The determination of a detailed restriction map for the clones contained in contigs assembled by restriction digest fingerprints is very efficient. Clones are selected, in contig order, that differ by one or very few fingerprint restriction fragments. Separate HindIII and SalI digests, along with a repeat HindIII plus SalI digest are performed and the fragments resolved by agarose gel electrophoresis in contig order. Ambiguities in the order of restriction fragments common to multiple clones are resolved by analysis of the order in which they appear and disappear when processing along the contig. The endonuclease HindIII cleaves the DNA of Arabidopsis relatively frequently (on average ca. every 1.5 kb) giving a detailed restriction map. SalI cleaves less frequently and the map for this enzyme is convenient to use as a contig-scale map.

Integrated BAC and cosmid contig assembly and restriction mapping

BAC clones were identified from two regions, of 400 kb and 100 kb, as shown in Figure 1, by hybridisation of gel-purified DNA of the YACs YUP16B9 and YUP24F4 to high density colony filters of the TAMU BAC library (7). Contigs were assembled and restriction maps were constructed for SalI mainly by hybridisation of BACs and cosmids to Southern blots of SalI-digested DNA of the BACs identified. The gaps between the BACs identified were filled using cosmid subclone libraries constructed from the YACs YUP16B9 and YUP24F4. These cosmids were identified using a subtractive hybridisation strategy. The BACs from the regions were used as hybridisation probes to the subclone libraries to allow the identification of cosmids from the regions they do not represent. These cosmids were assembled into contigs and restriction mapped using the restriction fragment fingerprint approach. BAC and cosmid contigs were integrated by hybridisation of BACs to Southern blots of the cosmids assembled into contigs. The 1.7 Mb contig that was constructed from part of Arabidopsis chromosome 4 using the approaches described, is shown in Figure 4 as displayed through AGR. A BamHI map is projected for the portion of the contig assembled using our initial approach, a SalI map is projected for the portions assembled using restriction digest fingerprint analysis.


Construction of cosmid vector pCLD04541

pCLD 04541 was constructed by the insertion of a fragment containing a cos site into the ScaI site of plasmid pSLJ1711 (13), which was derived from plasmid pRK290 (12).

YAC subcloning protocol

The following method was developed in order to optimize the key parameters for successful library preparation from gel-purified YAC DNA. The YAC DNA preparation protocol yields a high concentration of YAC DNA, with minimal degradation of endogenous yeast chromosomes. The DNA is resolved using a Pulsed Field Gel Electrophoresis (PFGE) system that is well-suited to preparative scale separations and is fragmented for cloning by a protocol that reproducibly yields sufficient partially digested material of the correct size for cloning into cosmid vector pCLD04541.

1. Yeast chromosome preparation

Cultures were grown at 30oC in 200 ml of YEPD medium (in 1 litre conical flask) until in stationary phase (2 to 3 days). Cells were collected by centrifugation at 3000 rpm for 10 min. in a Sorvall RC3C centrifuge. The cells were resuspended in 10 ml 1 M Sorbitol and collected in a 15 ml graduated screw-capped tube by centrifugation at 3000 rpm for 10 min in an MSE Centaur2 bench-top centrifuge. The cells were resuspended in 1 M Sorbitol to a total volume of 2.5 ml and 2.5 ml of 1% InCert agarose (FMC) made up in 1 M Sorbitol and cooled to 50oC was added. The cells and agarose were mixed thoroughly by gently pipetting, dispensed into 100 µl block formers (LKB) and left to set on ice for 10 min. The blocks were pressed out into 45 ml SCEM (1 M Sorbitol, 100mM Sodium Citrate pH5.8, 10 mM EDTA, 30 mM 2-Mercaptoethanol) in a screw-capped centrifuge tube. 0.5 ml of 100 mg ml-1 Novozyme 234 (Novo Enzyme Products) in SCE (filter sterilized after dissolving) was added and the tube placed on a cell mixer at room temperature for 4 hours. The blocks were then transferred to a 15 ml screw-capped centrifuge tube containing 7.5 ml ESP ( 1 mgml-1 Proteinase K in 100 mM EDTA pH8, 1% Sarkosyl). The tube was then filled with ES (100 mM EDTA pH8, 1% Sarkosyl) and incubated for 2 days with one change of ESP (containing 0.5 mg ml- of Proteinase K). The blocks were then transfered to a 50 ml screw-capped centrifuge tube and washed 2 x 1 hour with 1 mM EDTA pH8 containing 1 mM phenylmethylsulphonyl fluoride (PMSF) and 6 x 20 min with 1 mM EDTA pH8 and finally 1 x 20 min with TE. The blocks were stored at 4oC.

2. YAC gel-purification and fragmentation

Blocks containing yeast chromosome preparations were loaded, 24 per gel, and resolved by POE on a modified PHOGE PFGE apparatus (14, (15,(16). In all cases the gel used was 1% LMP agarose (BRL). POE gels were run at 320 V with an effective field angle of 130o. Runs were for 2 to 4 days using pulse times appropriate to maximize resolution around the size of the YACs to be purified.

After electrophoresis, the gels were stained with ethidium bromide and the DNA was visualized by exposure to long-wave UV. The YAC band was quickly excised in a minimum-sized gel slice (usually ca. 2 ml) which was then cut into 6 pieces and incubated overnight in 12 ml of ESP. They were then washed with PMSF and EDTA, as for the chromosome preparations, before storage overnight at 4 oC. The blocks were then
equilibrated with water, transfered to two microcentrifuge tubes and melted at 65 oC for 20 min. 300 µl aliquots were added to 300 µl aliquots of 2 x CA buffer pre-warmed to 65 oC (1 x CA buffer is 20 mM Tris pH7.5, 7 mM MgCl2, 100 mM KCl and 200 µg ml-1 BSA.) and incubation continued at 65oC for a further 10 min. Methylation buffer was prepared by adding 33 µl of New England Biolabs (NEB) buffer 4 and 3.3 µl of 20 mg ml-1 BSA (Boehringer) to 273 µl water. 8.3 µl of 32 mM SAM and 13.2 µl M.TaqI (NEB) was added and the solution quickly dispensed into 6 x 50 µl aliquots. To each aliquot, 5 µl of an appropriate dilution of alphaTaqI (NEB) was added to give final enzyme quantities of 0.64, 0.32, 0.08, 0.02, 0.005 and 0.00125 Units. Each partial digestion mix was added to an aliquot of DNA and incubated at 65 oC for 30 min. 35 µl of 0.5 M EDTA pH8 (pre-warmed to 65oC) was added to each tube, which was then transferred to a 37 oC waterbath to equilibrate for 10 min. 2 Units of -agarase I (Calbiochem or NEB) was added and incubation continued at 65 oC overnight. The tubes were then centrifuged briefly, heated to 65 oC for 20 min and equilibrated at 37 OC for 10 min. 1 Unit of -agarase I was then added and incubation continued at 37 oC for 4 hours. The tubes were heated to 65 oC and all six aliquots were pooled in a 15 ml centrifuge tube containing 4 ml of phenol and 0.4 ml of 3 M Sodium Acetate. 5 µl of 10 mg ml-1 carrier RNA was added. The tube was shaken and the phases resolved by centrifugation at 3500 rpm for 10 min in Sorvall Centaur2 centrifuge. The aqueous phase was transferred to a tube containing 4 ml of phenol/chloroform, shaken and phases were resolved as before. The aqueous phase was transferred to a 30 ml corex tube, 2.5 volumes of ethanol was added and placed at -20 oC overnight.

3. Size fractionation of YAC partial digest products and library plating

The TaqI partially digested YAC DNA was collected by centrifugation, dissolved in 200 µl TE and re-precipitated. It was then dissolved in 10 µl TE, run on a 1% LGT agarose gel and fragments >15 kb in size were recovered by agarase treatment. The products were co-precipitated with 1 µg of dephosphorylated vector DNA (pCLD04541 linearized by cleavage with ClaI and gel-purified). The precipitated DNA was air-dried and dissolved in 2 µl of water. Ligation was performed by adding 1 µl of 3 x ligation buffer and 0.2 µl of high concentration T4 DNA Ligase (NEB) and incubation at 4 oC for 3 to 4 days. The ligated DNA was packaged using Stratagene XL or Gold extracts and plated on a Tetracycline-sensitive derivative of Stratagene's "Sure" strain of E. coli. Colonies were picked into 384-well microtitre plates and stored at -80 oC

Other methods

Other methods used were essentially standard (17)


We initially attempted to use the pre-assembled G cosmid contigs as the primary resource for long-range contig assembly. These had been estimated to contain 90 to 95% of the Arabidopsis genome (6). Over the 1.2 Mb region screened for coverage in these clones, we identified approximately 410 kb of coverage in these clones (34%). The shortfall of coverage is probably the result of a combination of several factors. One is that the coverage provided by the contigs is less than originally estimated. Clones of DNA of Arabidopsis in pLorist vectors are often unstable and a proportion of the contigs may be missing from the "representative" filters available to us. Also some clones identified on these filters could not be recovered from the glycerol stocks stored in the laboratory of origin.

Further cosmid coverage was readily obtained by hybridization to the pOCA18- and pLAFR3-based genomic libraries. Added to the clones already identified in pLorist-based contigs, 88% of the first 800 kb region analyzed was represented. The clones assembled into eight contigs. We were able to bridge the gaps, leading to a single contig, only by identifying the missing 100 kb in high redundancy YAC subclone cosmid libraries.

As it proved necessary to subclone YACs to complete coverage in cosmids, we tested a strategy that involved the use of YAC subclone libraries from the beginning. Cosmids in individual YAC subclone libraries were confirmed and grouped by hybridization with DNA of YACs overlapping the one subcloned. DNA was prepared from a sufficient number of cosmids, randomly selected from each group, to permit four- to 10-fold redundant coverage of the YAC. Contigs were assembled mainly by analysis of restriction endonuclease digest fingerprints. This simple procedure greatly reduced the number of Southern blots and hybridizations required to assemble and map the contig. It also makes the approach less sensitive to problems caused by repetitive sequences. Only those sequences repeated within the region covered by the subcloned YAC can cause ambiguities. For our subclone libraries we chose a cosmid vector, pCLD04541, which includes the sequences necessary for the integration of the insert into plant genomes by Agrobacterium-mediated transformation. This allows the libraries to be used directly for complementation analyses in gene cloning experiments. The maximum size of the inserts is restricted to around 22 kb in this vector, which produces a very detailed physical map, but one requiring more clones for complete coverage than would be required if the inserts were larger. Map resolution is increased further by the construction of a detailed restriction map, based largely on the fingerprint data used to assemble the contigs.

For projects such as large-scale genomic sequencing, the pCLD04541-based cosmids are not ideal. Although very stable, the relatively small proportion of insert to vector (average 17 kb insert to 29 kb vector) makes them inefficient to sequence by the preferred "shotgun" strategy. In such cases, other types of clones can easily be integrated by comparison of their restriction endonuclease digest fingerprints with those of the assembled contig. In this way, any available clones that can be more efficiently sequenced, such as pLorist-based cosmids, can be exploited. Short contigs of BAC clones can also be linked together using YAC subclone cosmid contigs assembled by restriction digest fingerprint analysis.


We would like to thank Deverie K. Bongard-Pierce for supplying us with dot-blot filters representative of the pLorist-based contigs and for sending the clones requested, and Neil Olszewski for supplying the pOCA18-based genomic library, Renate Schmidt for making available to us, prior to publication, the chromosome 4 YAC contig data, including the identities of some integrated pLorist-based contigs, Thomas Altmann for supplying us with the IGF BAC library, Rod Wing for supplying us with the TAMU BAC library and the Sanger Centre for assistance in producing high density colony filters of the BAC libraries. This work was supported by the BBSRC, the EC as part of the European Scientists Sequencing Arabidopsis (ESSA) project and by a grant from Hoechst AG to HMG.


1. Dean C and Schmidt R: Plant genomes: a current molecular description. Annu. Rev. Plant Physiol. Plant Mol. Biol. 46 395-418 (1995).

2. Schmidt R, West J, Love K, Lenehan Z, Lister C, Thompson H, Bouchez D and Dean C: Physical map and organisation of Arabidopsis thaliana chromosome 4. Science 270 480-483 (1995).

3. Zachgo EA, Wang M-L, Dewdney J, Bouchez D, Camilleri C, Belmonte S, Huang L, Dolan M and Goodman HM: A physical map of chromosome 2 of Arabidopsis thaliana. Genome Research 6 19-25 (1996).

4. Chen EY, Schlessinger D and Kere J: Ordered Shotgun Sequencing, a strategy for integrated mapping and sequencing of YAC clones. Genomics 17 651-656 (1993).

5. Vaudin M, Roopra A, Hillier L, Brinkman R, Sulston J, Wilson RK and Waterston RH: The construction and analysis of M13 libraries prepared from YAC DNA. Nucl. Acids Res. 23 670-674 (1995).

6. Hauge BM and Goodman HM: In Methods in Arabidopsis research (ed. C. Koncz, N.-H. Chua and J. Schell), pp. 191-223. World Scientific Publishing Co. Pty. Ltd., Singapore (1992).

7. Choi SD, Creelman R, Mullet J and Wing RA: Construction and characterisation of a bacterial artificial chromosome library for Arabidopsis thaliana. Weeds World 2 17-20 (1995).

8. M. Bevan, I. Bancroft, E. Bent, K. Love, P. Piffanelli, H. Goodman, C. Dean, [54 members and group leaders of sequencing laboratories], H. Mewes, C. Scheuller, N. Chalwatzis, Sequence-level Organisation of a Contiguous 1.9 Mb Region from Chromosome 4 of the Plant Arabidopsis thaliana. Nature in press

9. Cross SH and Little PFR: A cosmid vector for systematic chromosome walking. Gene 49 9-22 (1986).

10. Staskawicz B, Dahlbeck D, Keen N and Napoli C: Molecular characterization of cloned avirulence genes from Race 0 and Race 1 of Pseudomonas syringae pv. glycinea. J. Bacteriol. 169 5789-5794 (1987).

11. Olszewski NE, Martin FB and Ausubel FM: Specialized binary vector for plant transformation: expression of the Arabidopsis thaliana AHAS gene in Nicotiana tabacum. Nucl. Acids Res. 16 10765-10783 (1988).

12. Ditta G, Stanfield S, Corbin D and Helinski DR: Broad host range DNA cloning system for Gram-negative bacteria: construction of a gene bank of Rhizobium meliloti. Proc. Natl. Acad. Sci. USA 77 7347-7351 (1980).

13. Jones JDG, Shlumukov L, Carland F, English J, Scofield SR, Bishop GJ and Harrison K: Effective vectors for transformation, expression of heterologous genes, and assaying transposition excision in transgenic plants. Transgenic Research 1 285-297 (1992).

14. Bancroft I and Wolk CP: Pulsed homogeneous orthogonal field gel electrophoresis (PHOGE). Nucl. Acids Res. 16 7405-7418 (1988).

15. Schwartz DC and Koval M: Conformational dynamics of individual DNA molecules during gel electrophoresis. Nature 338 520-522 (1989).

16. Bancroft I, Westphal L, Schmidt R and Dean C: PFGE-resolved RFLP analysis and long range restriction mapping of the DNA of Arabidopsis thaliana using whole YAC clones as probes. Nucl. Acids Res. 20 6201-6207 (1992).

17. Sambrook J, Fritsch EF and Maniatis T: Molecular cloning: A laboratory manual, 2nd. Edition. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY (1989).