Maize Mu Transposons Are Targeted to the 5' Untranslated Region of the gl8 Gene and Sequences Flanking Mu Target-Site Duplications Exhibit Nonrandom Nucleotide Composition Throughout the Genome

Corresponding author: Patrick S. Schnable, Iowa State University, Ames, IA 50011., schnable{at}iastate.edu (E-mail)

Communicating editor: J. A. BIRCHLER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The widespread use of the maize Mutator (Mu) system to generate mutants exploits the preference of Mu transposons to insert into genic regions. However, little is known about the specificity of Mu insertions within genes. Analysis of 79 independently isolated Mu-induced alleles at the gl8 locus established that at least 75 contain Mu insertions. Analysis of the terminal inverted repeats (TIRs) of the inserted transposons defined three new Mu transposons: Mu10, Mu 11, and Mu12. A large percentage (>80%) of the insertions are located in the 5' untranslated region (UTR) of the gl8 gene. Ten positions within the 5' UTR experienced multiple independent Mu insertions. Analyses of the nucleotide composition of the 9-bp TSD and the sequences directly flanking the TSD reveals that the nucleotide composition of Mu insertion sites differs dramatically from that of random DNA. In particular, the frequencies at which C's and G's are observed at positions -2 and +2 (relative to the TSD) are substantially higher than expected. Insertion sites of 315 RescueMu insertions displayed the same nonrandom nucleotide composition observed for the gl8-Mu alleles. Hence, this study provides strong evidence for the involvement of sequences flanking the TSD in Mu insertion-site selection.

ABOUT a dozen families of maize transposons have been identified (NEVERS et al. 1985 ; PETERSON 1988 ; CAPY et al. 1998 ). Each family consists typically of two categories of transposons, autonomous and nonautonomous. Autonomous transposons encode all nonhost factors required for their own transposition. In contrast, the transposition of nonautonomous transposons is dependent upon factors encoded by autonomous transposons of the same family. The Mutator (Mu) family consists of the autonomous transposon MuDR (SCHNABLE and PETERSON 1986 ; HERSHBERGER et al. 1991 ; CHOMET et al. 1991 ; QIN et al. 1991 ; HSIA and SCHNABLE 1996 ) and at least seven classes of nonautonomous transposons: Mu1, Mu2, Mu3, Mu4, Mu5, Mu7/rcy, and Mu8 (for reviews see WALBOT 1991 ; CHANDLER and HARDEMAN 1992 ). Mu transposons contain 215-bp terminal inverted repeats (TIRs) that are highly conserved and are thought to be recognized by the MuDR-encoded transposase (BENITO and WALBOT 1997 ; RAIZADA and WALBOT 2000 ). In addition to these Mu transposons, a novel Mu TIR (GenBank accession no. AF231940), recently identified as part of a Mu insertion in the rf2 gene, was defined as the left TIR of Mu10 (X. CUI, A. HSIA, D. A. ASHLOCK, R. P. WISE and P. S. SCHNABLE, unpublished data).

Some transposons exhibit nonrandom patterns of insertion. For example, miniature inverted repeat transposable elements such as the Tourist and Stowaway described by BUREAU and WESSLER 1992 , BUREAU and WESSLER 1994 , have a preference for insertion into genic regions (ZHANG et al. 2000 ). In addition, they exhibit a preference for 2- to 3-bp A/T-rich target sites. Transposons and insertion sequences from bacteria show a wide range of insertion specificity ranging from recognition of specific target sequences to preferences for insertion into A/T-rich sequences of promoter regions (BERG and HOWE 1989 ). The most dramatic example of nonrandom insertion within genes is the preference of the P element of Drosophila to insert into the 5' untranslated regions (UTRs) of genes (reviewed by SPRADLING et al. 1995 ). Spradling compiled data for 56 insertions (from 49 different genes) for which sequence data were available and thereby demonstrated that all the insertions had occurred in the 5' halves of the affected genes and over one-half of all the insertions had occurred in the 5' UTRs.

As expected, Mu insertions that are responsible for mutations are located in genes—usually in exons, but in some cases in noncoding regions (reviewed by BENNETZEN 1996 ). However, even Mu transposons that were not preselected for being responsible for a mutation seem to be preferentially located in gene-like, low copy, hypomethylated regions of the genome. No obvious target site or secondary structure has been identified to explain this preference (BENNETZEN et al. 1993 ). More recently, PCR-based techniques such as amplification of insertion mutagenized sites (FREY et al. 1998 ) and the use of a transgenic modified Mu1 transposon, RescueMu (RAIZADA et al. 2001 ), have allowed for the large-scale isolation of DNA sequences that flank random Mu insertions. Analyses of such fragments have revealed that these sequences have significant levels of nucleotide identity to expressed sequence tags (ESTs) at frequencies higher than would be expected for random genomic DNA (J. VOGEL, personal communication; HANLEY et al. 2000 ; RAIZADA et al. 2001 ). This preference for Mu transposons to insert into genes, combined with their high transposition rate (ALLEMAN and FREELING 1986 ), results in mutation rates that are 50-fold higher than the spontaneous rate (ROBERTSON 1978 ) and makes Mu a powerful tool for tagging and cloning genes (for reviews see BENNETZEN et al. 1987 ; SHEPHERD et al. 1988 ; WALBOT 1992 ).

Despite the widespread use of Mu transposons for gene cloning, relatively little is known of Mu transposon insertion preference within genes. There is evidence to suggest Mu transposons insert nonrandomly within at least some genes (reviewed by BENNETZEN et al. 1993 ). For example, the mapping of 24 bz-Mu alleles via DNA gel blotting revealed that as many as 19 are clustered into an 600-bp region around intron 1 (TAYLOR et al. 1986 ; BROWN et al. 1989 ; HARDEMAN and CHANDLER 1989 , HARDEMAN and CHANDLER 1993 ; SCHNABLE et al. 1989 ; BRITT and WALBOT 1991 ; DOSEFF et al. 1991 ). Of these bz1 alleles, 5 were subsequently sequenced and 4 were found to have Mu insertions in a narrow region just 3' of intron 1 (SCHNABLE et al. 1989 ; BRITT and WALBOT 1991 ; DOSEFF et al. 1991 ; CHANDLER and HARDEMAN 1992 ). Indeed, three of these independent insertions occurred at exactly the same nucleotide position. Similarly, the 4 characterized adh1-Mu alleles have insertions in the first intron (BARKER et al. 1984 ; ROWLAND and STROMMER 1985 ; CHEN et al. 1987 ). Analysis of a collection of Mu-induced dominant alleles at the kn1 locus revealed that all 9 have Mu insertions within a 310-bp region of the third intron (GREENE et al. 1994 ). In addition, it has been suggested that Mu transposons may have a preference for insertion into promoters on the basis of the fact that a number of insertions have been isolated from promoter regions (BENNETZEN et al. 1993 ; BENNETZEN 1996 ).

Although some evidence suggests that Mu transposons may insert at preferred sites within genes, to date this hypothesis has been tested only via the analysis of relatively few mutant alleles that were generated from multiple, and often unrelated, Mu stocks. It is therefore difficult to draw firm conclusions regarding the specificity of Mu insertions from the extant data. In the current study, each member of a large collection of Mu-induced glossy8 (gl8) alleles generated from genetically related Mu stocks was characterized by PCR amplification to determine whether the locus was disrupted by a Mu insertion. Subsequently, sequence analyses of the resulting PCR products established the exact Mu insertion sites in 75 of the 79 gl8-Mu alleles. These data demonstrate that Mu transposons have a strong preference for inserting into the 5' UTR of the gl8 gene. Analysis of sequences flanking the 9-bp TSD has revealed a highly significant conservation of nucleotide composition in the positions directly flanking the 9-bp TSD. Analysis of insertion sites from 315 RescueMu transposons demonstrated that the nonrandom nucleotide composition flanking the gl8-Mu insertion sites is not unique to insertions in the gl8 gene.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Genetic stocks:
The Mu transposon stocks used to generate 75 of the Mu-tagged gl8 alleles have been described previously (STINARD et al. 1993 ). Of the remaining alleles, gl8-Mu 94-1480-26 and gl8-Mu 94-1641-25 were isolated from a different Mu stock maintained by the Schnable laboratory, and gl8-Mu 90-2940-A and gl8-Mu 90-3230-5 were isolated from F₂ Mu families provided by S. Briggs, when he was at Pioneer Hi-Bred International (Johnston, IA). Alleles gl8-Mu 89g-5348-24 and gl8-Mu 91-2145 (SCHNABLE et al. 1994 ) were found to have Mu8 insertions at the exact site and orientation as the Mu8 insertion allele gl8-Mu 88-3142. Because of the way the random-tagged alleles gl8-Mu 89g-5348-24 and gl8-Mu 91-2145 were isolated, it was not possible to confirm their independence from gl8-Mu 88-3142 and they were therefore not included in this study.

The gl8 locus was originally defined by a spontaneous mutation (EMERSON et al. 1935 ) that was designated as the reference allele (gl8-ref). The gl8-ref pr stock used in the directed-tagging experiment was provided by D. Robertson (Iowa State University; Schnable accession no. 552). The Aet/LineC genetic stock has the genotype Gl8 pr/Gl8 pr.

All inbred lines were maintained by selfing and/or sib mating. The inbred lines Q66 and Q67 (Schnable accession nos. 111 and 113, respectively) were originally obtained from A. Hallauer (Iowa State University). The inbred lines B77 and B79 (Schnable accession nos. 403 and 404, respectively) were originally obtained from D. Robertson (Iowa State University). The inbred line W64A (Schnable accession no.142) was provided by D. Pring.

Isolation and sequencing of the gl8 genomic clones:
A B73 genomic library constructed in Dash II (Stratagene, La Jolla, CA) by J. Tossberg (Pioneer Hi-Bred International) was screened by DNA hybridization (MANIATIS et al. 1982 ). A partial 0.8-kb gl8 cDNA was used as a probe in the initial library screen. In subsequent purification steps, a 140-bp fragment isolated from the 5' end of the 1.4-kb apparent full-length gl8 cDNA clone pgl8 (XU et al. 1997 ) was used as the probe (probe A, Fig 1). The gl8 genomic clone, 1512-38, was isolated and determined to contain a 6.8-kb HindIII fragment containing the entire gl8 gene. This 6.8-kb HindIII fragment was sequenced at the Iowa State University Nucleic Acid Facility on an ABI 373A automated DNA sequencer (Applied Biosystems, Foster City, CA). Sequence analyses were performed using the Sequencher Version 3.0 software package (Gene Codes, Ann Arbor, MI). The 3.6-kb HindIII/SacI fragment from the 5' half of the gl8 genomic clone 1512-38 was subcloned into pBSK to make the clone pgl83.6. Probe B (Fig 1) was obtained by PCR amplification of pgl83.6 with primers gl8a58 and gl8a51.

View larger version (5K): In this window In a new window Download PPT slide   Figure 1. gl8 gene structure. Coding regions are represented as solid boxes and 5' and 3' UTRs are shown as shaded boxes. The approximate positions and orientation of 16 gl8-specific primers used to characterize gl8-Mu alleles are indicated as arrows on a 6.8-kb HindIII fragment isolated from the B73 genomic clone 1512-38. Positions of probes A and B are shown below the genomic clone. Primers are designated as follows: 1, gl8a54; 2, gl8a61; 3, gl8a58; 4, gl8a62; 5, 8a2840; 6, gab457; 7, gab812; 8, g24he.p4; 9, 8a2637; 10, gl8ain; 11, gab869; 12, gab830; 13, xx022; 14, mcd696; 15, gl8a51; and 16, gl8a59. Indicated restriction sites: H, HindIII; S, SacI.

Isolation of genomic DNA:
F₂ families segregating for gl8-Mu alleles were grown in greenhouse sand benches for 7 days, at which time individual glossy plants (gl8-Mu/gl8-Mu) were identified by the "water-beading" phenotype (SCHNABLE et al. 1994 ). DNA was extracted from each of these plants using a version of the ROGERS and BENDICH 1985 hexadecyltrimethyl-ammonium bromide (CTAB) extraction protocol modified to allow a substantially higher throughput (eight 96-well plates can easily be completed in a single work day). In the high-throughput version of the protocol, sections of seedlings with fully expanded leaves one and two were harvested from just above the tip of the coleoptile to the top of the first leaf sheath. The first leaf sheath was removed from this section and the remaining tissue was inserted into 2-ml strip tubes in 96-well format and freeze dried. Approximately 20 1.7- to 2.5-mm glass beads (MO-Sci Corporation, Rolla, MO, catalog no. MS-302/GL-01915) were added to each tube and dried tissue was pulverized for 5 min using a paint shaker (Red Devil Equipment, Brooklyn Park, MN, model no. 5400). A total of 600 µl of CTAB extraction buffer [1% CTAB, 100 mM Tris (pH 7.5), 0.7 M NaCl, 10 mM EDTA, 100 mM 2-mercaptoethanol] was added to the resulting powder and incubated at 60° for 60 min. Samples were allowed to cool to room temperature before adding 300 µl chloroform/octanol (24:1). Tubes were then mixed by inverting for 5 min and centrifuged at 5000 x g for 10 min. The aqueous phase was withdrawn and mixed with an equal volume of isopropanol to precipitate the PCR-ready DNA.

For 11 of the gl8-Mu alleles derived from a directed-tagging experiment, F₂ seed was not available. For these alleles, 10 individual seedlings from crosses such as cross 2 or cross 3 (see RESULTS) were pooled and DNA was extracted via the method of DELLAPORTA et al. 1983 . DNA samples from the maize inbred lines Q66, Q67, B77, and B79 and a stock homozygous for the gl8-ref allele were extracted according to methods of SAGHAI-MAROOF et al. 1984 .

Mu transposons:
TIR sequences for Mu1, Mu3, Mu4, Mu5, Mu7/rcy, Mu8, and MuDR were obtained from GenBank accession nos. X13019, U19613, X14224, X14225, X15872, X53604, and M76978, respectively. The left and right TIRs were defined as the TIRs that were listed first and second, respectively, in the appropriate GenBank entry. The TIRs of Mu2 were obtained from TAYLOR and WALBOT 1987 . The single TIR from Mu10 was obtained from GenBank accession no. AF231940.

PCR amplification of Mu-flanking regions:
PCR was performed using a primer in the conserved region of the Mu TIR, primer Mu-TIR (5' AGA GAA GCC AAC GCC A(AT)C GCC TC(CT) ATT TCG TC 3'), in combination with individual gl8-specific primers to amplify the gl8 sequences flanking each Mu insertion. PCR amplification reactions were performed with a PTC-200 (MJ Research, Waltham, MA) thermal cycler with the following conditions: denature at 94° for 1 min, anneal at 62° for 1.5 min, and extend at 72° for 2 min for 40 cycles, followed by a final extension at 72° for 5 min. PCR products obtained by amplification with Mu-TIR and a gl8-specific primer upstream or downstream of the Mu transposon within the gl8 gene were called 5' or 3' products, respectively. To amplify the 3' product of Mu10 and Mu12 insertions, it was necessary to lower the annealing temperature to 55°. To compensate for the high GC content in the 5' half of exon 1 of the gl8 gene, DMSO was added to a final concentration of 10% for PCR reactions involving primers xx022, 8a2840, or mcd696. PCR products were purified using a QIAGEN (Valencia, CA) PCR purification kit (catalog no. 28104) and sequenced. PCR reactions were performed using the following gl8-specific primers. The approximate location of each primer is shown in Fig 1.

• gl8a54: 5' GCC ACC CGG ACT AAA ACC TG 3'

• gl8a59: 5' TAA TGG CCT CGC TGT CAC 3'

• gl8a61: 5' AGC AGC AGC GAT CAC CTC AG 3'

• gl8a51: 5' TGT GCC TGC CCC TGT GTC 3'

• gl8a58: 5' AAG AGT GTG GCG CGT GCT ATG 3'

• gl8a62: 5' AAG TGA GAA AGA AAG GTT GTC C 3'

• gl8a64: 5' TTT CGA ATA TTT GTC CTA CTG TTA G 3'

• 8a2840: 5' CCA CCC ACC ACC GGA TAT AGG TCA TG 3'

• mcd696: 5' CGC ACC TCG GGG ACC TTG G 3'

• xx022: 5' CGG ATC AGA AGG CAC GAC GGA G 3'

• gab457: 5' GGT GGA CGA GGA GCT GAT G 3'

• gab830: 5' CAT TGC ACA TCA ATA CCC TTG CTC TTG TAC TC 3'

• gab812: 5' TCA AGA TGC CTC TAT GTT GAG TAC AAG AGC AAG 3'

• gl8ain: 5' CTC AGG AGG TAA TGG TAG 3'

• gab869: 5' GCC AGC CCC TTC TTG CGG ATC TTA ATG 3'

• g24he.p4: 5' CCT ATG CTC GTG CTG CCG TTC GTC 3'

• 8a2637: 5' GTG GCG ACA AAG CTT GCA TCT ATC AGG AAG TCT 3'

Isolation of the 5' UTR region from Gl8 progenitor alleles:
A portion of the gl8 gene containing the 5' UTR region was sequenced from each of the Mu stock progenitors. This region of the Gl8-B77, Gl8-B79, and Gl8-Q67 alleles was PCR amplified with primers gl8a58 and mcd696 (GenBank accession nos. AF348367, AF348368, AF348369, respectively). Because the Gl8-Q66 allele could not be amplified with gl8a58, it was amplified using the primer pair gl8a62 and mcd696 (GenBank accession no. AF348370). The resulting PCR products were purified and cloned into the TOPO TA cloning vector (Invitrogen, Carlsbad, CA, catalog no. K4500-40) and a bulk of 10 individual clones from each gl8 allele was sequenced.

Genetic algorithm for alignment of sequences:
The genetic algorithm (GOLDBERG 1989 ) used to align the gl8-Mu and RescueMu sequences specifies a single alignment as a collection of either forward- or reverse-complement orientation for each insertion-site sequence. An initial random population of 2000 such alignments was used in each run of the genetic algorithm. The fitness of any given alignment was taken to be the sum, over columns and over the four bases, of the squared deviation of the nucleotide counts from the background distribution of C's, G's, T's, and A's of the sequence data. This function was maximized by selection. Selection was performed by shuffling the population randomly into 500 groups of four alignments. In each group of four alignments, the two lower-scoring alignments were discarded and the higher-scoring alignments were copied in their place. Suffixes of the copies, selected uniformly at random, were exchanged (this is a one point crossover) and then the copies were mutated. Uniform mutation with positional mutation probability of 1/n (the number of insertion sites being aligned) was used. Uniform mutation proceeds down the string of forward-reverse specifications that define an alignment and independently flips each sequence with a fixed probability, 1/n in this case. Each run of the genetic algorithm was permitted to continue for 250 selection steps (generations). For each collection of insertion sites, the maximized consensus sequence was found in a majority of 100 runs of the genetic algorithm, suggesting that it is the true optimum alignment. Each of the 100 runs used a different initial random population of alignments.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The gl8 gene structure:
The gl8 gene was previously cloned using a Mu-tagged allele (gl8-Mu 88-3142) that has a Mu8 insertion near the gene's start codon (XU et al. 1997 ). A B73 genomic clone, 1512-38, was isolated (see MATERIALS AND METHODS) and restriction analysis identified a 6.8-kb HindIII fragment that contained the entire gl8-coding region. This 6.8-kb fragment was completely sequenced (GenBank accession no. AF302098). Comparisons between this B73 genomic sequence and the apparent full-length gl8 cDNA clone (pgl8) described by XU et al. 1997 revealed that the gl8 gene contains two introns of 829 and 583 bp separated by a 70-bp exon (Fig 1). The sequenced 6.8-kb HindIII genomic gl8 fragment includes 2871 bp upstream of the 5' end of the gl8 cDNA clone and 1193 bp 3' of its polyadenylation site. Analysis of the gl8 genomic sequence revealed that within the first exon there is a stop codon 50 bp 5' of the start of the pgl8 sequence. As is true for many plant genes, there are no obvious TATA or CCAAT boxes in the 5' region of the gl8 gene.

Isolation of gl8-Mu alleles:
The gl8 gene product is the ß-keto acyl reductase component (XU et al. 2002 ) of the long-chain fatty acid elongase complex involved in the production of cuticular waxes. The inability of the gl8 mutant to produce a normal wax load gives the leaves of mutant seedlings a "glossy" appearance and permits the ready identification of mutants. As reported previously, (SCHNABLE et al. 1994 ; XU et al. 1997 ) genetic screens were used to isolate 58 gl8-Mu alleles. This collection included alleles generated via both random transposon tagging (10 alleles) and direct transposon tagging (46 alleles). Randomly tagged mutants were identified as new glossy mutant plants (gl*/gl*) in F₂ progenies of plants that carried an active Mu system. Subsequent allelism tests established which of these gl* alleles were allelic to gl8. Further analysis of the random-tagged gl* alleles described by SCHNABLE et al. 1994 , and additional alleles, has resulted in the identification of 5 more randomly tagged gl8-Mu alleles, bringing the total to 15. Further analysis of the progeny from the gl8 directed-tagging experiment previously described by SCHNABLE et al. 1994 and XU et al. 1997 has increased the number of directly tagged alleles isolated by the Schnable laboratory from 46 to 64. Hence, a total of 79 gl8-Mu alleles (randomly and directly tagged) were available for the current study. Directly tagged alleles were generated via cross 1 (in all crosses the female parent is listed first).

Cross 1: Mu Gl8 Pr/Gl8 Pr x gl8-ref pr/gl8-ref pr:
Ears from cross 1 were individually shelled and kernels from each ear were planted in greenhouse sand benches such that family structures were maintained. Rare glossy seedlings from this cross carried newly generated gl8-Mu alleles. Because each gl8-Mu allele was isolated from a different ear, each allele must necessarily represent an independent mutational event. The exceptional glossy seedlings, which had the genotype gl8-Mu Pr/gl8-ref pr, were transplanted to pots and crossed to a Gl8 pr stock (cross 2) to facilitate the genetic separation of the gl8-Mu and gl8-ref alleles.

Cross 2: gl8-Mu Pr/gl8-ref pr x Gl8 pr/Gl8 pr:
Kernels from cross 2 segregated one purple (Pr/pr):one red (pr/pr). Because pr is genetically tightly linked to gl8 (<1 cM; STINARD and SCHNABLE 1993 ), >99% of the purple kernels will have the genotype gl8-Mu Pr/Gl8 pr.

In some instances, glossy progeny from cross 1 were crossed onto inbred lines such as W64A (Gl8 Pr/Gl8 Pr) instead of the Gl8 pr stock (cross 3).

Cross 3: gl8-Mu Pr/gl8-ref pr x Gl8 Pr/Gl8 Pr (W64A or other inbreds):
Because all of the colored kernels from cross 3 were purple, it was not possible to use phenotypic selection to identify kernels that carried gl8-Mu alleles. Instead, in these families F₂ analysis was used to distinguish between gl8-Mu Pr/Gl8 Pr (F₂ will not segregate red kernels) and gl8-ref pr/Gl8 Pr (F₂ will segregate red kernels) progeny of cross 3.

Identification of Mu insertion sites:
To identify the Mu insertion site for each of the 79 gl8-Mu alleles, PCR was performed on each allele using a gl8-specific primer in combination with a primer located in the highly conserved TIRs of Mu transposons. The resulting PCR products that hybridized to the gl8 genomic sequence were purified and sequenced. For each gl8-Mu allele, as many as 16 gl8-specific primers (Fig 1) spanning a 6.0-kb interval containing the gl8 gene were used to identify a primer that in combination with the Mu-TIR primer would amplify the gl8/Mu-flanking DNA. Amplification products were obtained from 75 of the 79 gl8-Mu alleles analyzed. Although Mu-flanking PCR products were obtained from both sides of the Mu transposon for most of these alleles, sequence analysis of only one side was sufficient to determine the transposon insertion site. These analyses revealed that 62 of these 75 Mu insertions (>80%) had occurred in the 5' UTR to the gl8 gene and 52 (69%) of those occurred within an 60-nucleotide interval of the gl8 5' UTR (Fig 2). Ten positions within the 5' UTR experienced multiple Mu insertions. One position in particular was host to 15 independent insertion events. Each of the alleles associated with a multiple insertion site was reamplified and sequenced from DNA extracted from an independent batch of seedlings. In all cases the results from the second analysis were in agreement with those from the first.

View larger version (26K): In this window In a new window Download PPT slide   Figure 2. Mu transposon insertion sites in 75 independent gl8-Mu insertion alleles. The gl8 gene structure shown is based upon the Gl8-B73 allele with corresponding 5' UTR sequences provided for each of the four progenitor alleles that exist in the Schnable laboratory Mu stock. Triangles indicate Mu transposon insertion sites; the 62 insertion sites within the 5' UTR and the three insertions between position -11 and -4 are shown at the 5' side of the 9-bp TSD. MuDR insertions are indicated by shaded triangles and Mu1 or *Mu1 (see Table 1) are indicated by solid triangles. Open triangles designate transposon insertions other than Mu1 or MuDR. Solid circles indicate the positions of the right TIRs for those transposons for which orientation was determined. Numbers directly above or to the right of insertions identify alleles as listed in Table 1. Alleles with noncanonical insertion sites are indicated by asterisks and only their approximate insertion sites are indicated. Nucleotide positions are numbered with respect to the beginning of an apparently full-length gl8-B73 cDNA clone (XU et al. 1997 ). The translation start codon is shown in boldface type near position 140. Gaps (—) were inserted where necessary to optimize the alignment of the four sequences. These gaps represent IDPs and underlined bases designate single nucleotide polymorphisms. A silent substitution at nucleotide 199 introduces an additional polymorphism that was used to identify the progenitor alleles of the 5' insertion alleles. CACNG motifs in the 5' UTRs are boxed. A solid bar below position 20–80 indicates the region of highest Mu insertion activity.

Because the Mu stocks used to generate most of these mutant alleles were maintained via crosses to the F₁ hybrids B77 x B79 and Q67 x Q66 (STINARD et al. 1993 ), the gl8-Mu alleles were derived from four potential Gl8 progenitor alleles. The 5' UTR sequence of each of these four Gl8 progenitor alleles was obtained by PCR amplification with two gl8-specific primers (see MATERIALS AND METHODS). Several allele-specific insertion/deletion polymorphisms (IDPs) 5' of the start codon were identified among the four parental alleles (Fig 2 and data not shown). These IDPs, along with a silent substitution at position 199 of exon 1, permitted the identification of the progenitor alleles of each of the 65 gl8-Mu alleles that arose via a Mu insertion 5' of the gl8 start codon (Fig 2 and Table 1).

To ensure that the gl8-ref allele used in the directed-tagging experiments did not interfere with the PCR-based mapping of Mu transposons, gl8-ref was characterized by PCR amplification and sequence analysis. No gl8-hybridizing PCR products were obtained with DNA from gl8-ref/gl8-ref individuals using the Mu-TIR primer in combination with any of 16 gl8-specific primers (data not shown). In addition, no DNA sequence polymorphisms were found between the gl8-ref allele and the wild-type Gl8-B73 allele in the region defined by PCR primers gl8a58 to gl8ain, which includes the entire coding region.

Analysis of Mu TIR sequences:
The gl8/Mu-flanking PCR products contained 39 nucleotides of Mu TIR sequence terminal to the Mu-TIR primer annealing site. Comparing the sequence of these 39 nucleotides from each gl8-Mu allele to the left and right TIRs of the previously defined Mu transposons (Fig 3) identified most of the Mu transposons. Insertions corresponding to six Mu transposons (MuDR, Mu1, Mu2, Mu4, Mu8, and Mu10) were identified in this fashion. Several novel TIR sequences were also recovered. The novel TIRs identified from the 5' and 3' PCR products of gl8-Mu 91g211 define the left and right TIRs of the Mu11 transposon (GenBank accession nos. AF247740 and AF247741, respectively). An additional novel Mu transposon was identified from gl8-Mu 91g241 and was designated Mu12. Sequences from the 5' and 3' PCR product from gl8-Mu 91g241 revealed that the terminal 39 bp from the TIRs of Mu12 are perfect inverted repeats and were designated as the left and right TIRs of Mu12 (GenBank accession nos. AF247742 and AF302101, respectively). Analysis of the sequence of the 5' PCR product from gl8-Mu 91g209 identified the inserted transposon as Mu10. Because only the left TIR of Mu10 had previously been reported, the 3' PCR product from gl8-Mu 91g209 was sequenced to obtain the right TIR of Mu10 (GenBank accession no. AF302099).

View larger version (81K): In this window In a new window Download PPT slide   Figure 3. Comparison of the terminal 39 nucleotides of 22 Mu TIR sequences. Consensus1 was derived from the TIRs of Mu1, Mu2, Mu3, Mu4, Mu5, Mu7, Mu8, and MuDR only. Consensus2 was derived from all 22 TIR sequences. Nucleotides in the consensus lines are underlined if that position is invariant among the TIRs included in the consensus line. Nucleotides in the consensus line that are not underlined are well conserved (i.e., no other nucleotide is present more than twice at that position). Positions that do not have an invariant or a well-conserved nucleotide are indicated in the consensus line by their IUB ambiguity codes (M, R, S, V, W, and Y). Nucleotides present only one or two times at a position were not considered when assigning ambiguity codes. Nucleotides that do not conform to consensus1 are shaded. Asterisks indicate eight positions that are invariant in the TIRs of Mu1, Mu2, Mu3, Mu4, Mu5, Mu7/rcy, Mu8, and MuDR (consensus1) but are no longer invariant when the TIRs of the newly defined Mu10, Mu11, and Mu12 transposons are considered (consensus2).

Characterization of the four gl8-Mu alleles in which Mu insertions were not detected:
Mu insertions were not identified in 4 of the 79 gl8 alleles analyzed in this study. These 4 alleles were subjected to further analyses. On the basis of PCR amplification and sequence analysis it was determined that the progenitor of the directly tagged allele gl8-Mu 91g215 was Gl8-Q67 (data not shown). DNA gel blot analysis was then performed using probe B (Fig 1). Although a restriction fragment length polymorphism (RFLP) exists between gl8-Mu 91g215 and Gl8-Q67 (Fig 4, lanes 5 and 1, respectively), gl8-Mu 91g215 is indistinguishable from gl8-ref (Fig 4, lane 3) in this hybridization experiment. Therefore, it is likely that during cross 2 the gl8-Mu 91g215 allele was replaced by the gl8-ref allele as the result of a crossover between gl8 and pr.

View larger version (67K): In this window In a new window Download PPT slide   Figure 4. DNA gel blot analysis of six gl8-Mu alleles. Approximately 10 µg of HindIII-digested genomic DNA was loaded per lane and hybridized with probe B (Fig 1). Lanes 1, 2, and 10 contain DNA from inbred lines Q67, B73, and W64A, respectively. Lane 3 contains DNA from a stock homozygous for the gl8-ref allele. Lanes 4–8 contain DNA from stocks homozygous for the designated gl8-Mu alleles. Lane 9 contains DNA from gl8-Mu 93B227 isolated from a bulk of 10 progeny from cross 3.

DNA isolated from a plant homozygous for gl8-Mu 91g159 failed to hybridize to probe B (Fig 4, lane 4). Hybridization with an unrelated single-copy probe established that lane 4 in Fig 4 contains DNA of sufficient quality and quantity to yield hybridization signals to single-copy genes (data not shown). A subsequent experiment using the same blot and the 1.4-kb gl8 cDNA also failed to detect the gl8 gene in lane 4 (data not shown). This result suggests that gl8-Mu 91g159 consists of a deletion of the entire coding region of the gl8 gene.

The gl8-Mu 94-1641-25 allele was not generated from a Mu stock that contains Q66-, Q67-, B77-, and B79-derived alleles (see MATERIALS AND METHODS). Hence, it is not possible to determine with certainty its progenitor allele. However, the 5' UTR of gl8-Mu 94-1641-25 was PCR amplified and sequenced from plants homozygous for this allele and was found to be identical to Gl8-B73. In addition, DNA gel blot analysis using probe B revealed that gl8-Mu 94-1641-25 was indistinguishable from Gl8-B73 (Fig 4, lanes 6 and 2, respectively). These results suggest that gl8-Mu 94-1641-25 either contains a minor rearrangement not detectable via RFLP analysis or has an insertion (or other mutation) outside of the 6.8-kb HindIII fragment detected by the RFLP analysis.

Although it was not possible to identify a Mu insertion in gl8-Mu 93B227, RFLP analysis of this allele shed some light on the nature of its molecular lesion. Lane 9 of Fig 4 contains DNA from a pool of 10 progeny resulting from a cross between the inbred line W64 and a plant with the genotype gl8-Mu 93B227/gl8-ref (cross 3). Analysis with probe B revealed that this DNA sample contains two RFLP fragments. A comparison between lanes 3 and 10 revealed that the gl8-ref and Gl8-W64 alleles are indistinguishable in this hybridization experiment and account for the smaller RFLP signal in lane 9. Hence, the larger RFLP signal in lane 9 must be derived from gl8-Mu 93B227. This RFLP differs from each of the four possible progenitor alleles of gl8-Mu 93B227: Gl8-Q67, Gl8-66, Gl8-B77, and Gl8-B79 (lane 1 and data not shown). Hence, gl8-Mu 93B227 may contain a novel Mu transposon that could not be amplified because it contains divergent TIR sequences or because sequence divergence near the insertion site of gl8-Mu 93B227 relative to the Gl8-B73-derived primers may have impeded amplification. Alternatively, this allele may have resulted from another type of molecular rearrangement.

Noncanonical target-site duplications:
For most alleles, sequence analysis of the gl8/Mu PCR product derived from one side of the Mu transposon was sufficient to determine the Mu insertion site and Mu identity. Nonetheless, to better define a novel transposon or to confirm apparent sequence anomalies, the gl8/Mu-flanking PCR products were sequenced from both sides of the Mu transposons associated with 14 alleles. Analysis of these 14 sequences revealed that 4 alleles (gl8-Mu 91g168, gl8-Mu 91g169, gl8-Mu 91g213, and gl8-Mu 91g239) did not have the characteristic 9-bp TSD (Fig 5). Each of these alleles was reamplified and sequenced from an independent batch of seedlings. In all instances results from the second analysis were in agreement with the first set of analyses. The sizes of the 5' and 3' Mu-flanking PCR products from gl8-Mu 91g168 were not in agreement with the predicted product sizes from any of the 4 progenitor alleles, suggesting that this allele contained a deletion. Sequence analysis of both of these PCR products established the progenitor of gl8-Mu 91g168 as Gl8-Q66 and revealed an apparent deletion of 232 bp from the Mu insertion site (Fig 5A). However, the PCR-based characterization performed here could not distinguish between a deletion and two closely linked (232 bp) Mu1 insertions. Such a structure was observed in the hcf106-mum2 allele in which a second Mu1 transposon inserted 244 bp downstream of the original Mu1 insertion of the hcf106-mum1 allele (DAS and MARTIENSSEN 1995 ). To distinguish between these two possibilities, PCR was performed using primers Mu-TIR and gl8a64, which lies in the predicted deletion. The observation that this PCR reaction failed to amplify a gl8-hybridizing product provides evidence that the gl8-Mu 91g168 allele contains a 232-bp deletion rather than a double Mu insertion. Deletions adjacent to a Mu insertion have been observed to arise subsequent to initial insertion events at frequencies near 1% and are thought to result from aborted transposition events or illegitimate recombination between the transposon and the gene (LEVY and WALBOT 1991 ; reviewed by DAS and MARTIENSSEN 1995 ; RAIZADA et al. 2001 ). Given this frequency, during the propagation of 79 alleles over multiple generations, the recovery of such a deletion would not be unexpected.

View larger version (17K): In this window In a new window Download PPT slide   Figure 5. Insertion-site sequences of four noncanonical Mu insertions. Shaded triangles indicate MuDR transposons and the open triangle represents a Mu4 or Mu4-like transposon. Numbers above triangles indicate allele numbers as shown in Fig 1 and Table 1. Underlined text indicates aberrant TSDs. gl8-Mu 91g168 contains a 232-bp deletion from the gl8 5' region with no TSD. gl8-Mu 91g169 contains a seven-nucleotide deletion from the 3' TSD followed by an insertion of four C's, leaving only a 2-bp TSD. gl8-Mu 91g213 has a mismatch in the 9-bp TSD. The 3' duplication has an A-to-C transversion. gl8-Mu 91g239 has a single-nucleotide deletion from the 3' TSD, resulting in an 8-bp TSD.

Sequence data from one side of the Mu insertion of alleles gl8-Mu 91g169, gl8-Mu 91g213, and gl8-Mu 91g239 established polymorphisms relative to each of the four progenitor alleles. For these alleles, sequence data were also obtained from the other side of the inserted Mu transposon. The progenitor of gl8-Mu 91g169 could be identified as Gl8-B77 but contained a deletion of seven nucleotides from the 3' TSD followed by the insertion of four C's directly flanking the Mu insertion (Fig 5B). Analysis of the Mu-flanking sequence from gl8-Mu 91g213 identified the progenitor as Gl8-Q67, but the 3' flanking sequence contains an A-to-C transversion in the second nucleotide position of the TSD (Fig 5C). Sequence analysis of the 3' PCR product from gl8-Mu 91g239 identified the progenitor as Gl8-B79 and the Mu transposon as either Mu3 or Mu4. Sequence obtained from the 5' PCR product established that the inserted Mu transposon was Mu4. However, the TSDs on the two sides of this transposon were not identical due to an apparent deletion of a G from the 3' TSD (Fig 5D).

Statistical analysis of target-site sequences:
Mu insertions occurred most frequently in an 60-bp region between nucleotide positions 20 and 80 in the 5' UTR of the gl8 gene. Depending on the progenitor allele, and after removal of gaps that were inserted to allow alignment of the sequences, the actual length of this region varies from 47 to 51 nucleotides. A motif search of the B73-gl8 genomic sequence identified a five-base motif, CACNG, which appears frequently in this region of the 5' UTR that experienced the highest Mu insertion frequency. Also depending on the progenitor allele, the CACNG motif appears between four and six times in this 60-bp region with as many as five motifs arranged in tandem. Excluding the targeted Mu insertion region of the 5' UTR, the CACNG motif appears at the expected frequency (1/256 bp) in the 3045 nucleotides of the B73-gl8 sequence between primers 3 and 10 (Fig 1). Therefore the nonrandom distribution of CACNG motifs in the gl8 gene mirrors the nonrandom insertion pattern of Mu transposons in the gl8 gene.

A genetic algorithm was used to identify the sequence alignment that would provide the best consensus sequence for the Mu insertion sites identified in this study. Genetic algorithms create a population of solutions to a problem and use an iterative selection and variation process to identify good solutions. In this case, the goal is to determine, for each insertion site, which of the two possible orientations provides the best alignment with all other insertion sites. The genetic algorithm does this by determining for each insertion site the orientation that maximizes the divergence from the background distribution of nucleotides. This maximization is performed by selecting alignments with high divergence scores in pairs and then applying variation operations (crossover and mutation) to the alignment to generate similar and possibly superior alignments (see MATERIALS AND METHODS).

The 35 unique insertion sites as defined by the 10 nucleotides 5' of the TSD, the 9-bp TSD, and the 10 nucleotides 3' of the TSD were analyzed in the manner described above. Although each of the 71 Mu insertion sites listed in Table 1 is from an independent insertion event, to avoid biasing the data from sites with multiple insertion events, only a single instance of any given 29-bp insertion site was used in this analysis. Fig 6 shows the orientations of the sequence alignments as selected by the genetic algorithm.

View larger version (62K): In this window In a new window Download PPT slide   Figure 6. Alignment of the 35 unique insertion sites from Table 1. (A) The top sequence illustrates an example of an insertion site as present in a gl8-Mu allele. The underlined sequence indicates the 9-bp TSD and the triangle represents the Mu transposon. The bottom sequence illustrates the target site as present in the corresponding progenitor allele. The nine nucleotides in the TSD are designated as positions 1–9. The 9-bp TSD is separated from each set of flanking sequences by a double colon. The 10 nucleotides 5' to the TSD are designated positions -1 to -10 and the 10 nucleotides 3' to the TSD are designated as positions +1 to +10. (B) The DNA strand shown for each of the 35 independent insertion sites is the strand that gives the most nonrandom nucleotide composition as determined by the genetic algorithm (see MATERIALS AND METHODS). DNA strands oriented 5' to 3' with respect to the gl8 gene are indicated with an F and sequences in which the opposite strand was selected are indicated by R/C. The four positions identified in Table 2 as having significant deviations from the expected nucleotide frequency at the 99% confidence interval are shaded. The number to the left of each sequence indicates the number of gl8-Mu alleles with the identical 29-bp insertion site.

As shown in Table 2, the 35 insertion sites are GC rich and particularly low in T. This may be due, at least in part, to the GC richness of the gl8 gene. To normalize for this high GC content an expected nucleotide frequency was calculated based on the GC content of the combined insertion sites. Chi-square analyses were performed to identify positions that had significant deviations from the expected nucleotide composition. Within the TSD, T's, G's, and C's appeared at positions 2, 8, and 9, respectively, at significantly higher-than-expected frequencies. Position 4 has a lower-than-expected frequency of A. The conserved nucleotides at positions 2, 8, and 9 and the weak consensus nucleotides at the other TSD positions are consistent with the reverse complement of consensus sequences reported by CRESSE et al. 1995 and HANLEY et al. 2000 (Table 3).

Chi-square analyses of the 20 positions flanking the gl8 TSDs reveals a more significant nucleotide conservation than is observed within the 9-bp TSD itself. Five of the six positions directly flanking the 9-bp TSD have a nucleotide composition that differs from the expected nucleotide composition at the 99% confidence interval (Table 2 and Fig 6). These include conserved C's, A's, and G's at positions -2, +1, and +2, respectively, and significantly lower frequencies of Cs at positions -1 and +3. Two of these conserved nucleotides, the -2 C and +2 G, are particularly interesting given that they are complementary bases equidistant from the TSD. When this analysis was extended to include 50 nucleotides flanking either side of the Mu insertion, additional positions with significant deviations from the expected were not identified at rates higher than would be expected by chance (data not shown).

Analysis of RescueMu insertion sites:
To determine if the nonrandom nucleotide composition observed among gl8 insertion sites is a common feature of Mu insertion sites across the genome, a large collection of RescueMu insertion sites was analyzed as described above for the gl8-Mu insertion sites. A total of 369 independent RescueMu insertion-site sequences generated by the Walbot laboratory were recovered from GenBank by identifying forward and reverse sequences from individual RescueMu events. Eighteen of these RescueMu insertion sites have TSD lengths other than 9 bp (Table 4). An additional nine RescueMu insertion sites were recovered that contained mismatched TSDs (data not shown). Therefore, 315 RescueMu sequences remained after removal of these noncanonical RescueMu insertion sites and insertion sites that did not have at least 60 bp of sequence flanking each side of the TSD. These insertion sites (consisting of 60 bp flanking each side of the TSD plus the 9-bp TSD) were aligned using the genetic algorithm described previously. Table 5 contains the nucleotide composition and total chi-square values for the 10 positions flanking the TSD and the 9-bp TSD for the optimal RescueMu alignment. Within the TSD, six of the nine positions have nucleotides that differ from the expected nucleotide frequency at the 99% confidence interval and two of the remaining three positions have nucleotides that differ from the expected nucleotide frequency at the 95% confidence interval. The one remaining position (position 4) has a significantly lower-than-expected frequency of A's. Position 4 was therefore designated as B, in accordance with the International Union of Biochemistry (IUB) ambiguity code, in the RescueMu consensus TSD and in the combined consensus TSD. The 9-bp consensus TSD for the RescueMu insertion sequences is 5' CTCBCAGAC 3', which is strikingly similar to the consensus 9-bp TSD from the gl8-Mu insertion sites and the reverse complements of the CRESSE et al. 1995 and HANLEY et al. 2000 consensus TSDs. The four consensus sequences shown in Table 3 can be combined into a single 9-bp consensus TSD of 5' C-T-C-B-G/C-A/C-G/A-A/G-C 3'.

As was the case for the gl8-Mu insertion sites, the nucleotides with the highest deviations from the expected nucleotide composition were observed at the positions directly flanking the TSD. The three positions flanking either side of the TSD had a conserved nucleotide at the 99% confidence interval. These nucleotides are CCT at positions -1, -2, and -3, respectively, and AGG at positions +1, +2, and +3, respectively. Hence, the three positions immediately flanking either side of the TSD are conserved for the paired 3-bp inverted repeats CCT and ATT. Therefore, each of the four conserved nucleotides identified from the gl8-Mu insertions is also conserved at its respective position in the RescueMu insertion sites. The larger size of the RescueMu data set, however, apparently allowed the additional conserved nucleotides to be identified. A plot of the total chi-square at each of the 129 positions reveals that nearly all the positions with significant deviations from the expected occur within a 15-bp region centered on the TSD and that the positions with the highest chi-squares directly flank the 9-bp TSD, particularly the -2 and +2 positions (Fig 7).

View larger version (13K): In this window In a new window Download PPT slide   Figure 7. Chi-squares were calculated at each of the 129 positions for the 315 aligned RescueMu insertion sequences that consisted of the TSD and 60 bp flanking each side of the TSD. The symbol ** indicates the cutoff at the 99% confidence interval with 3 d.f. The 15-bp target site (TSD +6) is indicated by a shaded box and the -2 and +2 positions that have the highest chi-square values are indicated.

Analysis of the RescueMu sequence alignment also revealed a significant difference between the nucleotide composition of the sequences to the left and right of the TSD (Fig 8A). GC profiles of the gl8 and bz1 genes reveal that the preferred target site for Mu insertions also occurs at the interface between regions with low and high GC content (Fig 9). These data suggest that Mu transposons may have a preference for inserting into regions where DNA composition is transitioning from low to high GC content.

View larger version (27K): In this window In a new window Download PPT slide   Figure 8. Profiles of GC content (A), bendability (B), B-DNA twist (C), A-philicity (D), and protein-induced deformability (E) of the 315 RescueMu insertion sites. The average values of GC content and bendability were calculated using a window size of 3 bp and a shift increment of one; A-philicity, protein-induced deformability, and B-DNA twist were calculated using a window size of 2 bp and a shift increment of one as described at http://www.fruitfly.org/~guochun/pins.html. The solid line represents data from the RescueMu insertion sequences. The dashed line represents the average of 315 129-bp cDNA fragments that were randomly selected from maize ESTs. Solid squares indicate positions within the 15-bp target site that are significantly different from random DNA as determined by t-test analysis with P < 0.05.

View larger version (17K): In this window In a new window Download PPT slide   Figure 9. GC content profiles from gl8 and bz1 suggest that GC content may influence Mu insertion preferences. GC profiles were generated with the WINDOW algorithm of the GCG sequence analysis package. A window size of 100 bp and a shift increment of three were used. Each GC profile is overlaid with a schematic of the appropriate gene structure. The regions of each gene that have experienced high rates of Mu insertion are shaded. Of the 75 gl8 alleles analyzed in the study, 62 contain insertions in the shaded region. Mu transposons from each of the four bz1 alleles that have been positioned by sequencing inserted in the shaded region shown in the bz1 gene. Genomic sequence of bz1 was obtained from GenBank accession no. X135000 and 5' and 3' UTRs were determined from the bz1 cDNA (GenBank accession no. X13502).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Analysis of a collection of 79 gl8 mutant alleles derived from Mu stocks revealed that the vast majority (at least 75/79) contained Mu insertions. Of these 75 alleles, 62 had Mu insertions in the 140-bp 5' UTR of the gene. These 62 alleles include numerous independent insertions at exactly the same nucleotide positions of the 5' UTR. This provides strong evidence for the targeting of Mu transposons not only into the 5' UTR of the gl8 gene but also into specific positions within the 5' UTR. Although the genetic screen used to isolate the gl8-Mu alleles ensured that each allele arose via an independent transposition event, the independence of alleles that contain Mu insertions at the same nucleotide positions was further confirmed by their (1) distinctive transposon identities, (2) transposon orientations, and (3) the wild-type progenitor alleles. For example, of the 15 independent Mu insertions at position 67, 4 are MuDR and 11 are Mu1. Of the 11 Mu1 insertions, 6 are oriented left to right and 5 are oriented right to left. In addition, these 11 Mu1-induced alleles were derived from both Q67 and B79 progenitors. Hence, this collection of gl8 alleles provides convincing evidence for the preferential insertion of Mu transposons into specific regions of a maize gene.

Because each of the gl8-Mu alleles in this study was selected on the basis of the presence of a mutant phenotype, it is likely that the observed distribution of Mu transposons in this collection does not represent a random sampling of insertion events in the gl8 gene because insertions into introns or the 3' UTR may not have resulted in a mutant phenotype. However, this phenotypic selection cannot explain the clustering of multiple independent Mu insertions in the 5' UTR because Mu insertions elsewhere in the gl8-coding region also yield mutant phenotypes but were recovered at much lower frequencies. Analysis of Mu insertion alleles from a reverse genetics screen that did not depend upon phenotypic selection would provide a useful comparison. However, Mu-based reverse genetic screens typically do not yield a sufficient number of alleles per gene to provide evidence of targeting.

Another advantage of working with a defined set of independent insertion alleles generated from a related Mu stock is that alleles are not limited to ones that can be amplified by primers specific to known TIR sequences. If alleles that cannot be amplified with the Mu-TIR primer are present in the collection, insertions in such alleles can still be identified by DNA gel blot hybridization. Of the 79 gl8-Mu alleles analyzed in this study, only 1 allele failed to amplify with the Mu-TIR primer and was shown by gel blot analysis to have an insertion. This suggests that most, if not all, of the active Mu transposons present in at least the Schnable lab Mu stocks have now been identified. Of course, different Mu stocks may contain additional active Mu transposons.

In addition, because 72 of the 75 gl8-Mu alleles in which a Mu transposon was identified were derived from related Mu stocks maintained by the Schnable lab, the insertion frequency of each Mu transposon can be compared. Of the 72 insertions from the Schnable lab Mu stock, 41 were generated by insertions of Mu1, 18 by MuDR, 3 by Mu8, 1 by Mu2, 1 by Mu4, 1 by Mu10, 5 by Mu11, and 2 by Mu12. The overwhelming majority (59/72 or 82%) of these insertions were of the two transposons that are typically most common in Mu stocks, Mu1 and MuDR. The typically less common Mu2, Mu4, and Mu8 transposons were responsible for relatively few of the gl8-Mu alleles. Hence, insertions in the gl8 gene seem to occur at frequencies proportional to their abundance in typical Mu stocks. This is in contrast to observations involving other maize genes that appear to be preferentially targeted by specific Mu transposons. For example, all but 2 of the 24 Mu-induced bz1 alleles that have been characterized (TAYLOR et al. 1986 ; BROWN et al. 1989 ; HARDEMAN and CHANDLER 1989 , HARDEMAN and CHANDLER 1993 ; SCHNABLE et al. 1989 ; BRITT and WALBOT 1991 ; DOSEFF et al. 1991 ; CHANDLER and HARDEMAN 1992 ) contain Mu1 or Mu1-del insertions. Analyses of 8 Mu-induced sh1 alleles and 9 Mu-induced kn1 alleles suggest that these genes may be more susceptible to MuDR (HARDEMAN and CHANDLER 1993 ) and Mu8 (GREENE et al. 1994 ) insertions, respectively. It is of course possible that these apparent differences in susceptibility to insertion of particular classes of Mu transposons may be artifacts resulting from the small number of alleles analyzed or may reflect differences among the Mu stocks used to isolate these alleles.

The novel Mu transposons Mu10, Mu11, and Mu12 were found as new insertions from an active Mu line and have characteristics typical of Mu transposons such as conserved TIRs and the ability to generate 9-bp TSDs upon insertion. The partial TIR sequences from Mu10, Mu11, and Mu12 exhibit a high degree of identity to TIRs from known Mu transposons but are clearly distinct from each other and all previously identified Mu TIR sequences. Of the 39 bp of sequence obtained from these TIRs, 18 nucleotides are conserved among all Mu TIRs. Eight additional positions that are identical in the TIRs of Mu1, Mu2, Mu3, Mu4, Mu5, Mu7/rcy, Mu8, and MuDR are not conserved among the TIRs of Mu10, Mu11, and Mu12. Four of these eight divergent positions are part of the predicted MURA transposase-binding site that extends from nucleotides 25–56 (BENITO and WALBOT 1997 ). Due to the placement of the Mu-TIR primer used in this experiment only a small part (nucleotides 25–39) of the predicted MURA-binding site could be recovered and analyzed. A high degree of conservation in this predicted binding site is thought to be required for MURA binding because MURA fails to bind in vitro to the divergent left TIR of the apparently inactive Mu5 transposon (BENITO and WALBOT 1997 ). However, the identification of Mu10, Mu11, and Mu12 as new insertions from a Mu stock suggests that MURA can bind to the divergent TIRs of these Mu transposons. Isolation of the complete TIR sequences and the internal sequences of Mu10, Mu11, and Mu12 may significantly enhance our understanding of MURA-binding specificity and Mu activity in general.

The identification of only three previously uncharacterized Mu transposons in this study suggests that the number of active Mu transposon family members is relatively small. Previous studies using DNA gel blotting to compare hybridization patterns obtained with Mu TIR specific probes to patterns obtained from probes specific to the internal regions of the cloned Mu transposons suggested that only 50–60% of the Mu TIRs in the genome are associated with known Mu transposons (HARDEMAN and CHANDLER 1989 ). If Mu transposons insert at rates proportional to their abundance in Mu stocks, as the data presented in this study suggest, the three novel Mu transposons identified in this study would not account for the 40–50% of Mu TIRs not associated with previously cloned Mu transposons. Hence, there are probably a large number of Mu transposons that are not actively inserting into genes. Mu5 may be an example of such an inactive transposon because no Mu5 transposons have ever been found to have caused a mutation by inserting into a gene (TALBERT et al. 1989 ). Earlier reports suggested that Mu4 might also be inactive (TALBERT et al. 1989 ). The identification of a putative Mu4 insertion in gl8-Mu 91g239 suggests that Mu4 may indeed be active, although the possibility that a novel Mu transposon with the same TIR sequences as Mu4 inserted into gl8-Mu 91g239 cannot be excluded.

Small insertions, deletions, and single-nucleotide substitutions in TSDs, such as the ones associated with these gl8-Mu alleles, have been identified previously in revertant alleles resulting from transposon excision (reviewed by SAEDLER and NEVERS 1985 ) and in somatic excision products of Mu transposons (BRITT and WALBOT 1991 ; DOSEFF et al. 1991 ). However, the aberrant insertion sites of several gl8 alleles (Fig 5), to our knowledge, are the first reports of aberrant insertion sites (excluding deletions such as that of gl8-Mu 91g168) flanking resident Mu transposons. In addition to the noncanonical TSDs in the gl8-Mu insertion collection, 18 RescueMu insertion sites with TSDs other than 9-bp were identified from the 369 RescueMu insertion sites obtained from GenBank sequences (Table 4). The 6-, 7-, and 8-bp TSDs are likely the result of 1–3 nucleotide deletions from one of the original 9-bp TSDs. These insertion sites are therefore similar to the insertion site of gl8-Mu 91g239 that has an 8-bp TSD that appears to have resulted from a deletion of a G from the right TSD. The 10-bp TSDs could result from the addition of a single nucleotide at one of the TSD/TIR borders that, by chance, matched the corresponding nucleotide in the opposite flanking sequence. Alternatively, the 10-bp TSD could result from a 10-bp duplication instead of the typical 9-bp duplication. These two possibilities cannot be distinguished because the progenitor sequences are not available for the RescueMu insertions. In addition, sequencing errors cannot be ruled out as the cause of the noncanonical TSD associated with the RescueMu insertion sites. However, it is unlikely that all of the variation in RescueMu TSD length is a consequence of sequencing errors. This can be inferred from a higher frequency of recovery of TSDs with variations in length (18/369) than the frequency of recovery of insertion sites with mismatched TSDs (9/369), which provides an upper bound for sequencing errors. The recovery of noncanonical TSDs flanking both resident Mu and RescueMu transposons suggests that such TSDs can result from aberrant insertion events, although aborted transposition events cannot be ruled out. Variation in TSD length has been reported from Mu-like transposons in Arabidopsis (YU et al. 2000 ) and rice (TURCOTTE et al. 2001 ) but has not been previously reported for Mu transposons.

The tendency for Mu transposons to preferentially insert into genic regions has now become fairly well established (HANLEY et al. 2000 ; RAIZADA et al. 2001 ) yet the mechanism for this preference is not clear. Prior analyses of the 9-bp TSDs that occur at the site of Mu insertions have produced similar consensus sequences (CRESSE et al. 1995 ; HANLEY et al. 2000 ). This study reports two additional consensus sequences from independent collections of Mu insertion sites that are strikingly similar to those reported previously. Consistencies among these consensus sequences derived from independent collections of Mu insertion events suggest some biological significance to the consensus sequences. The most obvious characteristic of these consensus sequences is that they are GC rich. It is not yet clear, however, whether this GC richness is a cause or an effect of the tendency of Mu transposons to insert into genes, which are themselves typically GC rich (SALINAS et al. 1988 ; CARELS et al. 1998 ).

The composition of the sequences flanking TSDs ha