Paramutation is an interaction between alleles that leads to a heritable change in the expression of one allele. In B′/B-I plants, B-I (high transcription) always changes to B′ (low transcription). The new B′ allele retains the low expression state in the next generation and paramutates B-I at a frequency of 100%. Comparisons of the structure and expression of B′ with that of a closely related allele that does not participate in paramutation demonstrated that transcription from the same promoter-proximal sequences is not sufficient for paramutation. Fine-structure recombination mapping localized sequences required for B′ expression and paramutation. The entire 110 kb upstream of the B′ transcription start site was cloned and sequenced and the recombination breakpoints were determined for 12 recombinant alleles. Sequences required for expression and paramutation mapped to distinct regions, 8.5-49 kb and 93-106 kb upstream of the B′ transcription start site, respectively. Sequencing and DNA blot analyses indicate that the B′ region required for paramutation is mostly unique or low copy in the maize genome. These results represent the first example of long-distance regulatory elements in plants and demonstrate that paramutation is mediated by long-distance cis and trans interactions.
PARAMUTATION was first described at r1 (Brink 1956) and b1 (Coe 1959) in maize. Subsequently, paramutation was observed at two other maize loci, pl1 (Hollicket al. 1995) and p1 (Sidorenko and Peterson 2001). All of these genes encode regulatory proteins required for flavonoid pigment synthesis. Paramutation and trans-interactions have also been described for both endogenous genes and transgenes in other plant species (reviewed in Brink 1973; Chandleret al. 2000) and in fungi (Colotet al. 1996; van Westet al. 1999). Although generally not meiotically heritable, allele interactions are involved in trans silencing in animals (reviewed in Henikoff and Comai 1998) and have been speculated to be involved in diabetes in humans (Bennettet al. 1997). Allele interactions also appear to be involved in a transposon excision-repair mechanism in Petunia (van Houwelingenet al. 1999). The mechanism is not understood for any system, although a heritable change in chromatin structure is a favored model for paramutation (reviewed in Hollicket al. 1997; Chandleret al. 2000).
In plants heterozygous for B-I and B′, the transcription of the B-I allele is always downregulated to a B′ transcription level (Pattersonet al. 1993). The 10- to 20-fold reduction in transcription causes a dramatic decrease in pigment in all plant tissues where B-I is normally expressed. B′ and heterozygous B′/B-I plants are lightly pigmented whereas B-I plants are darkly pigmented (Coe 1966; Pattersonet al. 1993). The change of B-I to B′ is heritable as only B′ is transmitted to progeny. These new B′ alleles are equivalent to the B′ parent; they retain the reduced transcription state and induce paramutation of B-I at a frequency of 100%. The paramutagenic (pg) B′ state is extremely stable; spontaneous changes from B′ to B-I have not been observed in ∼100,000 plants (Coe 1966; Pattersonet al. 1995), while the B-I state is very unstable; it spontaneously changes to B′ at frequencies of 1-10% (Coe 1966; Pattersonet al. 1995).
Comparison of B-I and B′ DNA sequences and genomic restriction maps generated with DNA-methylation-sensitive and -insensitive restriction enzymes revealed no differences in DNA sequences, DNA rearrangements, or DNA methylation patterns within 10 kb spanning the transcribed region (Patterson et al. 1993, 1995). However, the different transcription states are associated with a difference in chromatin structure. A DNaseI hypersensitive site was observed near the transcription start site that is more hypersensitive in B-I than in B′ (Chandleret al. 2000). Recombination mapping showed that the sequences required for paramutation are tightly linked to B′ and B-I (≤0.1 cM) and located upstream of the transcribed region (Pattersonet al. 1995).
We report here results from two approaches to identify where the sequences required for paramutation are located. First, we characterized a b1 allele, B-615, that does not participate in paramutation, but unlike all other known b1 alleles, it shares upstream sequences in common with B′ and B-I. Second, we used fine-structure recombination mapping to localize the B′ sequences required for paramutation. In addition, we identified a distinct region required for B′ expression.
MATERIALS AND METHODS
Plant stocks: In maize nomenclature (http://www.agron.missouri.edu/maize_nomenclature.html), a gene is designated with lowercase italics (b1). Specific alleles are indicated with an allele designation separated from the gene designation with a hyphen, dominant alleles are uppercase (B-I), and recessive alleles are lowercase (b-K55). Throughout the manuscript, a single allele listing indicates homozygosity, whereas heterozygous individuals are indicated with allele designations separated by a slash (/). All plant stocks contained dominant functional alleles for all of the anthocyanin biosynthetic genes required in vegetative plant and seed tissues. Stocks containing various b1 alleles have been maintained in the Chandler laboratory. They were originally obtained from various sources: E. H. Coe, Jr. (University of Missouri, Columbia) provided B-I (inbred W23 background), B′ (inbred K55 background), and gl2 b wt (inbred K55 background); M. G. Neuffer (University of Missouri, Columbia) provided B-Peru (inbred W22 background); and S. Evola (Ciba-Geigy, now Syngenta) provided B-615. The gl2 (glossy2) and wt (white tip) loci are morphological markers that flank b1. The B-615 allele was in a proprietary line (Selingeret al. 1998). We were using this line as a recipient for transformation and noticed that it carried a b1 allele with the same restriction map as B′, which was unusual. Because the b1 and r1 loci encode functionally duplicate genes, our stocks carried r1 alleles (r-g or r-r) that are not expressed in the seed or in the same juvenile and mature plant tissues as B-I, B′, and B-Peru.
Analyses of RNA levels and transcription rates: RNA was isolated and RNA levels assayed by RNase protection experiments as previously described (Dorweileret al. 2000). Transcription rates were assayed using nuclear run-on analyses as previously described (Dorweileret al. 2000).
DNA blot analyses: Standard blot analyses were done with ∼4 μg genomic DNA from leaves (Dellaportaet al. 1983) digested with several enzymes and enzyme combinations according to manufacturers’ specifications and size fractionated by electrophoresis in 0.5× TBE agarose gels followed by DNA blot analyses (Stamet al. 1997). For blots with large-molecular-weight DNA, high-molecular-weight DNA was isolated (Kaszas and Birchler 1998), the resulting plugs were incubated with the desired restriction endonuclease(s), and DNA was separated by pulsed-field gel electrophoresis (PFGE) using a CHEF-DRIII apparatus (Bio-Rad, Richmond, CA), in 0.5× TBE, 1% agarose gels [Sigma (St. Louis) A2929].
Recombinant screen: To isolate recombinant alleles between B′ and B-Peru, we used the phenotypic marker gl2, located 19 cM upstream (http://www.agron.missouri.edu/maps.html), and the tightly linked gene tmp (transmembrane protein), located 0.18 cM upstream of the b1 transcription start (Stamet al. 2000). The tmp alleles of the gl2 B′ and gl2 B-I stocks were identical, but polymorphic relative to the Gl2 B-Peru tmp allele in the third intron of tmp. Primers that amplified the third intron and 3′ exon sequences were developed, resulting in fragments of ∼685 bp (tmp-I) in B′ and B-I or ∼614 bp (tmp-P) in B-Peru (MS30, CACCATGCCTGGGCTGACC; Mtl-tmp, CCG GCAGCATCTCGAAGC). PCR conditions were the following: 1× PCR buffer (Sigma, product no. P2192), 0.1 mm dNTPs, 0.8 μm of each primer, 1 unit of Taq DNA polymerase (Sigma), reaction volume of 25 μl; 2 min at 94°, 30 cycles (1 min at 94°, 1 min at 62°, 3 min 30 sec at 72°), 10 min at 72°, 1 hr at 4°.
Plants carrying gl2 tmp-I B′ and Gl2 tmp-P B-Peru were crossed and the resulting F1 crossed with gl2 tmp-I B-I plants. To isolate recombinant B′ alleles, colorless seeds (B′/B-I genotype) were planted in trays and 10-day-old seedlings scored for their glossy2 phenotype (wild type vs. mutant have differences in wax content, easily visualized in seedlings). The recombinant seedlings (Gl2) were transplanted to the field and organized in grids containing 16 or 18 rows of 20 plants each. Leaf pieces were pooled for all plants in each row and in each column. DNA was isolated as described (Dellaportaet al. 1983), and PCR analyses were done on all pooled samples to identify candidate recombinants carrying the tmp allele linked to B-Peru.
To isolate recombinant B-Peru alleles, purple seeds (B-Peru/ B-I) were planted in trays. Recombinant seedlings (gl2) were transplanted to the field and vegetative pigment scored at maturity. Plants with lightly pigmented sheaths were carrying a paramutagenic B-Peru allele or the parental B-I allele underwent spontaneous paramutation. To identify recombinants every light plant was tested for loss of tmp-P. The nomenclature || used for recombinants is: n, neutral; pg, paramutagenic;, recombinant allele with the parental allele contributing sequences upstream of the breakpoint indicated to the left of the vertical bars and the parental allele contributing the coding region and promoter-proximal region indicated on the right. To follow recombinants in crosses, a PCR assay was used to specifically amplify a 427-bp fragment of tmp-P (primers: MS32, CTGTTGCTAACTAGTTTCACTGTTAA and MS33, GGCACA AATCGGTGAGAGTGAGAG). Primers amplifying the npi402 marker were used as an internal control (MS22, ACACGGCA CCTACGTATAAGG and MS23, CTGCACTCTCCGATAGAA TGG; 591-bp fragment that maps 1.1 cM upstream of b1; http://www.agron.missouri.edu/maps.html). PCR conditions were the following: 1× PCR buffer (Sigma, product no. P2192), 1 μl MgCl2 [for PCR, QIAGEN (Chatsworth, CA)], 0.2 mm dNTPs, 0.8 μm of MS32 and MS33, 0.2 μm of MS22 and MS23, 1 unit of Taq DNA polymerase (Sigma), reaction volume of 25 μl; 2 min at 94°, 35 cycles (1 min at 94°, 1 min at 62°, 1 min 20 sec at 72°), 10 min at 72° and 4°.
Isolation of pBACB′1 clone: To clone the ∼100-kb genomic MluI fragment, the pBeloBAC11 vector (Kimet al. 1996) was modified to contain a unique MluI site by cutting with BamHI and HindIII and ligating with a phosphorylated BamHI-MluI-AatII-HindIII linker (GATCCACGCGTGACGTCAAGCT), creating the pBeloBAC-Mlu vector. This vector was purified, cut with MluI, and dephosphorylated as in Birren et al. (1999). To remove the background of nondephosphorylated ends, the DNA was ligated in bulk (Osoegawaet al. 1998) and separated on a pulsed-field agarose gel, and linear DNA was isolated from SeaPlaque low-melting-point agarose using wizard PCR preps (Promega, Madison, WI).
Nuclei were isolated from B′ plants according to Kaszas and Birchler (1998). Small DNA was removed from agarose plugs with PFGE [6 V cm-1, 10-sec pulse (constant), 120° angle, 14°, 1% agarose, 0.5× TBE for 3 hr] using a CHEFDRIII apparatus (Bio-Rad). The plugs (0.1 ml, containing ∼10-20 μg of DNA) were incubated with the desired restriction endonuclease and PFGE was performed essentially as described (Osoegawaet al. 1998). Fragments smaller than ∼50 kb were removed by running the DNA toward the top of the gel [5 V cm-1, 15-sec pulse (constant), 120° angle, 14°, 0.5× TBE for 7.5 hr] and back into the wells using the same conditions. The remaining DNA was separated at 6 V cm-1, 2.7- to 20-sec pulse, 120° angle, 14°, 0.5× TBE for 16 hr. DNA (∼100 kb) was concentrated by running it into 4% agarose followed by electroelution for 41 hr at 4° as described by (Osoegawaet al. 1998).
The dephosphorylated vector and genomic DNA were ligated overnight (room temperature) in a 1:5 molar ratio (Birrenet al. 1999) followed by dialysis against sterile H2O and the DNA was concentrated by incubation in 30% PEG8000, 0.5× TE buffer. Escherichia coli DH10B cells were transformed according to Birren et al. (1999) using the electro cell manipulator ECM 600 (BTX). A total of 10,584 colonies were arrayed using the Biomek 2000 laboratory automation workstation (Beckman, Fullerton, CA) essentially as described (Birrenet al. 1999). The resulting filters were hybridized with a B′ promoter probe (SpeI/SalI, 1486-522 bp upstream of the start site) radioactively labeled using random hexamer priming (Feinberg and Vogelstein 1983). The results were visualized using a Storm 860 phosphorimager and ImageQuant software (Molecular Dynamics, Sunnyvale, CA). One positive colony, pBACB′1, was identified. DNA of this colony was prepared by alkaline lysis (Birrenet al. 1999) and verified to be the desired fragment by direct end sequencing.
Multiple restriction digestions using single or multiple enzymes were done to identify the approximate number of restriction sites present and their positions relative to each other. To order restriction fragments within the pBACB′1 clone, the insert was excised from the vector with NotI, followed by partial restriction with a dilution series of a second enzyme. DNA blots were hybridized with bacterial artificial chromosome (BAC) end probes.
Inverse polymerase chain reaction: Additional B′ sequence, upstream of the most 5′ MluI site in the pBACB′1 clone, was isolated by inverse polymerase chain reaction (IPCR). Genomic B′ DNA was cut with HinfI or BstYI followed by ligation (Snowden and Napoli 1998). IPCR was performed, in turn, on 200 ng of the HinfI or BstYI ligation mixtures. The resulting fragments were cloned into pGEM-T Easy according to the manufacturer’s recommendation (Promega) and sequenced by the Arizona GATC facility. Primer sets used were VC28: GAGCACCGAGCGGGGGTAGTACGTCG and VC29: AGTC GCAGCGCGCCATTGGAAGAGTC; VC41: GTCAAATGGAC GGTCGAACAAACATTCAG and VC42: CGAGCGTGAATAC TTTTTCTGATATCTAG; VC61: GATCTTGAAGTTAGCCTA AACAAGATC and VC62: CAAGTTTGAACTATATCTTTACC TTCACTC; VC89: GAAGCAACTTTACTGGATCGAGGG and VC90: GGATATTCATGCATAACCATTGCTTC. A total of 3.9 kb of additional upstream B′ sequences was obtained (accession no. AF475145).
Shotgun library construction for low-copy probe isolation: The pBluescript SKII+ vector was cut with EcoRV, dephosphorylated, self-ligated, and subjected to agarose gel electrophoresis, and the linear vector fragment was excised, isolated from agarose by phenol extraction, and resuspended in TE. pBACB′1 DNA was isolated according to Birren et al. (1999), purified by CsCl gradient ultracentrifugation, and sheared by nebulization for 2.5 min at 10 psi essentially as described (Wilson and Mardis 1997). Fragment ends were repaired by treating with T4 DNA polymerase, Klenow, and polynucleotide kinase. DNA fragments of ∼800 bp were gel purified (Osoegawaet al. 1998) and 6 ng were ligated to 5 ng of vector DNA, dialyzed, concentrated, and transformed into E. coli DH10B cells as described above.
A total of 631 white colonies were picked, arrayed, and replicated as described (Birrenet al. 1999) onto Hybond N+ filters using the Biomek 2000 laboratory automation workstation (Beckman). The filters were treated as described (Birrenet al. 1999) and hybridized with labeled Sau3AI, DdeI double-digested genomic maize B′ DNA. The 275 clones that gave no or a very weak hybridization signal were rearrayed onto five sets of three Hybond N+ filters. These filters were hybridized with all the previously identified single copy fragments from the regions 0-8 kb (Pattersonet al. 1995; Figure 2) and 94-106.6 kb (Figure 7) upstream of the B′ coding region. They were also hybridized with digested genomic maize B′ DNA a second time and with labeled pBeloBAC-Mlu DNA or DraI-cut genomic E. coli DNA. The empty vector clones were identified by restriction digestion and gel electrophoresis. To estimate the actual repetitiveness in the maize genome of the remaining 95 nonhybridizing clones, inserts were PCR amplified, radioactively labeled, and hybridized to filters containing BglII and EcoRI, SalI double-digested genomic B′ DNA. To position the fragments onto the pBACB′1, the plasmid DNAs of the 95 clones were sequenced by the Arizona GATC facility and aligned relative to the pBACB′1 sequence using BLAST 2 sequences (http://www.ncbi.nlm.nih.gov/gorf/bl2.html) and FAKtory, a sequence assembly program (http://bcf.arl.arizona.edu/faktory/).
Sequencing promoter-proximal region of B-615: The promoter-proximal region of B-615 was amplified with the primers B-I1086ds, CGGCGAATTCGTGAGATGGTAGATTGGTT GCGAC (anneals 1063 bp 5′ of transcription start) and BPSac-US, GGAGGAGGAGCTCACCTCCGCT (anneals 93 bp 5′ of transcription start). The PCR product was sequenced by the Arizona GATC facility using the primers B-I1086ds, BPSac-US, TMDDE1 (GCCGAATTCGACTAACCTTAGGCAAAGTG), and B-I Ddeus (CGGCGGATCCTGTCACACTTTGCCTAAGGTTA GTC). The accession number for this sequence is AF205792.
Sequencing of pBACB′1 clone: DNA from pBACB′1 was extracted using the large construct kit (QIAGEN) and sheared with a hydroshear device (Genemachines) to two different average sizes, 2 and 9 kb. The ends of the sheared fragments were made blunt with mung-bean exonuclease, extracted with a mixture of phenol and chloroform followed by chloroform extraction, and precipitated with isopropanol. The DNA was dephosphorylated with shrimp alkaline phosphatase, and “A” tails were added by incubation with Taq polymerase, run on a gel, and eluted using the gel elution kit (QIAGEN). Inserts were then ligated into pCR4-TOPO using the TA cloning system (Invitrogen, San Diego) and electroporated into E. coli DH10B electrocompetent cells (GIBCO BRL, Gaithersburg, MD). Subclones were sequenced from both directions using big dye terminator chemistry and run on an ABI3700 capillary sequencer. Base calling and quality assessment were done using PHRED (Ewing and Green 1998), the sequences were assembled by PHRAP, and the resulting contigs edited with CONSED (Gordonet al. 1998). To close gaps between the contigs and to go through GC-rich regions, a combination of different approaches was used: (1) sequencing using different chemistries, (2) sequencing using thermofidelase enzyme, (3) PCR amplification with primers flanking a gap, (4) shotgun sequencing of transposon-inserted subclones flanking a gap, and (5) direct sequencing of the BAC template. When gaps were associated with repetitive regions, subclones that start and end in unique regions but otherwise consist of repetitive DNA were assembled separately and inserted in the main assembly. The final error rate was estimated using CONSED. The accession number for the BAC sequence is AY078063.
Assembly programs suggested three tandem repeats in the upstream region, but restriction mapping indicated seven repeats. To obtain the complete sequence of this region, a 9-kb shotgun fragment spanning the repeats was cloned into pCR4-TOPO (Invitrogen) and a series of deletion derivatives were generated. The plasmid DNA was digested with SdaI, which linearized and created 3′ overhangs, and with PmeI, to create a blunt end. The DNA was subsequently digested with Exo-nuclease III, with aliquots removed at 2-min time points. The reactions were stopped by heat incubation, the DNA was treated with mung bean nuclease to create blunt ends, and the vector was self-ligated following the manufacturer’s instructions (Invitrogen). The DNA was used to transform E. coli (strain DH10B) and the resulting plasmids were sized on agarose gels. Clones with increasing deletions were then sequenced and the information assembled as described above.
Characterization of the B-615 allele suggests that sequences required for paramutation are >13 kb upstream of the coding region: The vast majority of b1 alleles do not participate in paramutation, but they have 5′ sequences that are distinct from those of B-I and B′ (Radicellaet al. 1992; Selinger and Chandler 1999). Initial DNA blot analyses suggested that the B-615 allele was an exception, because digestion with several enzymes revealed the same map as B′, B-I (the maps of B′ and B-I are identical) within at least ∼12 kb upstream of the coding region. This was intriguing because B-615 does not participate in paramutation; when B′/B-615 plants are crossed with B-I, dark (B-I/B-615) and light (B′/B-I) pigmented plants segregate 1:1, indicating that B-615 is insensitive (neutral) to paramutation (K. Kubo, M. Stam and V. Chandler, unpublished data). We hypothesized a more thorough investigation of its expression and DNA sequence relative to B′ might reveal differences, which would be candidates for requirements for paramutation.
Because paramutation involves an alteration in transcription, one possibility was that transcription from the same promoter-proximal sequences might be required for paramutation. In the original inbred genetic background, the B-615 allele gave rise to green plants, potentially because it is not transcribed. Alternatively, a very weak or null pl1 allele could be present (functional alleles of pl1 and either b1 or r1 are required to activate the anthocyanin biosynthetic pathway in plant tissues; Coe 1985). To investigate this, B-615 was introgressed into a b-K55 r-r Pl-Rh background by backcrossing six times. The r-r allele is expressed in the anthers, which allows scoring of the pl1 allele independently of B function by monitoring anthocyanin pigment levels in anthers. Pl-Rh is a fully functional pl1 allele (Coneet al. 1993), producing dark purple anthers in the presence of r-r. The resulting purple anthered plants, B-615/b-K55 r-r Pl-Rh, had weakly pigmented sheaths, similar to B′, suggesting that B-615 might be transcribed at a level similar to that of B′. To test this hypothesis, mRNA (Figure 1, A and B) and transcription levels (Figure 1, C and D) were compared in husk tissues from B-615 and B′. The promoter regions of these two alleles conferred similar transcription levels. Thus, the different paramutation properties of B-615 and B′ are not caused by differences in transcription level.
Extensive DNA blot analysis on B-615 and B-I genomic DNA, using multiple enzyme combinations (BamHI, BclI, BglII, EcoRI, EcoRV, HindIII, KpnI, PacI, SacI, XbaI, BamHI/BglII, BamHI/EcoRI, EcoRI/BglII, PstI/BglII, XbaI/EcoRI, XbaI/HindIII, and XbaI/BglII) and coding and upstream sequences as probes, detected no differences between the two alleles within ∼13 kb upstream of the transcription start site (map in Figure 2). An 1154-bp promoter-proximal fragment from B-615 was PCR amplified and sequenced (accession no. AF205792). The sequence was identical to that previously determined for B′ and B-I. Thus, either very subtle differences between B′, B-I, and B-615 in the 13-kb promoter-proximal region mediate paramutation or the sequences are located farther upstream.
Isolation of recombination-derived neutral B′ alleles: Previous studies had shown that the sequences required for paramutation were upstream of the coding region (Pattersonet al. 1995). Our comparison of B-615 with B′ suggested that the sequences required for paramutation might reside >13 kb upstream. To determine more precisely where the sequences required for paramutation mapped, we employed a fine-structure recombination mapping approach. To facilitate the isolation of recombinants we used B-Peru, which is insensitive (neutral) to paramutation and produces a nearly green plant and purple seed. In contrast, the paramutagenic B′ allele produces a lightly colored plant and colorless seed. The immediate upstream 2.5-kb promoter-proximal region of B-Peru contains a unique sequence that confers seed pigmentation (Radicellaet al. 1992; Selingeret al. 1998). The seed pigment and unique 2.5 kb in B-Peru relative to B′ provide visual and molecular markers to distinguish the promoter-proximal region of B′ and B-Peru in crosses.
We hypothesized that there are distinct upstream sequences in B′ required for paramutation and that these sequences are either missing or mutated in B-Peru. We further hypothesized that exchange of upstream B-Peru and B′ sequences via recombination would generate a neutral B′ allele, which we could easily identify as described below. Figure 3 shows the phenotypes and crossing strategy used. B′/B-Peru plants were crossed with plants carrying B-I, the allele sensitive to paramutation (colorless seeds and dark plants); colorless and dark seeds segregated 1:1. Planting colorless seeds selects for the B′ coding and immediate promoter-proximal sequences. Because B′ always paramutates B-I, all B′/B-I progeny should be lightly pigmented except when recombination separates B′ paramutation sequences from the B′ promoter-proximal region, creating an altered B′ allele that can no longer paramutate B-I. Such an allele would produce a dark plant when heterozygous with B-I.
In the initial test of this approach, ∼9000 colorless seeds were planted and the resulting ∼6500 plants scored for mature plant pigment. All but 2 plants were lightly pigmented. The 2 darkly pigmented plants were candidates for rare recombinants that had replaced the B′ sequences required for paramutation with the neutral sequences from B-Peru. A third recombinant was isolated in a separate experiment in which 1006 B′/B-I (colorless) seeds derived from a cross between B′/B-Peru and B-I/B-Peru plants were planted in the field. Among the ∼650 plants that survived to maturity, 1 dark plant was identified. PCR analyses using a polymorphic molecular marker 0.18 cM upstream of b1 (Stamet al. 2000; tmp) demonstrated that all 3 were recombinants, as they carried the upstream molecular marker linked to B-Peru and the B′ promoter-proximal and coding sequences (data not shown).
The recombinants were outcrossed with B′ and B-I to determine their ability to participate in paramutation and were self-pollinated and outcrossed to recessive b1 alleles to investigate their pigment phenotype. All three recombinant alleles were neutral, as all dark purple (B-I-like) plants (98) and no light purple (B′-like) plants (55) carried recombinant alleles; the three alleles were designated BP||B′-n1, BP||B′-n2, and BP||B′-n3, with n symbolizing neutral. The || indicates a recombination event, the allele on the right indicates the promoter-proximal and coding region, and the allele on the left indicates the sequences upstream of the recombination breakpoint.
Sequence polymorphisms between B′, B-Peru, BP B ′-n alleles, and B-615: As a first step toward determining where the recombination breakpoints occurred, sequence polymorphisms were mapped between B′, B-Peru, and the BP ||B′-n alleles. Genomic DNA was isolated from plants containing each allele, digested with different enzymes, and subjected to conventional gel electrophoresis or PFGE and DNA blot analysis using different probes. B-615 was included in the analysis to determine more precisely where differences between B′ and B-615 begin. A blot with MluI-digested DNA from B′, B-Peru, B-615, and the three B′-neutral recombinants illustrates one of the polymorphisms observed (Figure 4). We suspect the doublet seen with the recombinants and B′ reflect partial digestion caused by DNA methylation, as MluI is methylation sensitive. MluI cuts once in the B′ coding region, providing an anchor to position upstream sites. It was difficult to map restriction sites far upstream in the B-Peru parental allele for two reasons. First, the most upstream 5′G probe did not hybridize well to B-Peru DNA, and second, a large insertion of ∼30 kb is just upstream of the B-Peru promoter (Pattersonet al. 1995; Selingeret al. 1998), which contained sites for several of the restriction enzymes used for mapping. Thus, we primarily compared the recombinants relative to B′. The recombination breakpoints were assumed to have taken place between the most upstream conserved site and the first polymorphic site relative to B′. Restriction maps representing the key sites delimiting these recombination intervals are shown in Figure 5. Two neutral alleles retain the SwaI site 50 kb upstream of the B′ transcription start site, but instead of the ∼100-kb B′ MluI fragment, they have an ∼85-kb MluI fragment (Figures 4 and 5). This positions the recombination break-100 ′-n2 and BP||B′-n3 between 50 and points in BP B kb || upstream of the coding region. The recombination breakpoint in the BP||B′-n1 allele could be pinpointed more accurately because it occurred close to the B′ transcription start, between 8.4 and 8.9 kb upstream (Figure 5). The B-615 allele retained the B′ restriction map up to 22 kb upstream of the transcription start site, but lacked the SwaI site 50 kb upstream and gained a SnaBI polymorphism relative to B′. The observation that BP||B′-n2 and BP||B′-n3 retained 50 kb of B′ sequences, yet are neutral with respect to paramutation, indicates that the sequences up to 50 kb upstream of the B′ transcription start site are not sufficient for paramutation. The results with B-615 are also consistent with sequences farther upstream being required for paramutation.
Isolation of recombinant alleles delimiting the sequences required for paramutation: Isolation of the three neutral B′ alleles indicated that sequences required for paramutation could be further defined by isolating more recombinants. Additional mapping experiments were performed with B′ and B-Peru using the strategy outlined in Figure 6. Three recombination intervals can be distinguished between gl2 and the b1 coding region. Recombination between B′ and B-Peru in interval I, downstream of P (sequences required for paramutation), should generate B′ neutral alleles (as previously isolated) and B-Peru pg alleles. Recombination events in intervals II and III will have the parental paramutation phenotypes (colorless seeds will have paramutagenic alleles, and purple seeds neutral alleles), but they can be distinguished from each other using the tmp molecular marker (materials and methods).
A total of 21,027 colorless seeds were planted, seedlings were scored for recombination between gl2 and b1, and the resulting 3332 recombinants were assigned to one of the three intervals using the tmp molecular marker and genetic tests for paramutation. The molecular markers classified 32 recombinants to intervals I or II and 3300 recombinants to interval III. These numbers do not include the three previously isolated interval I B′ neutral alleles. All 32 candidates for intervals I or II showed a B′ phenotype when heterozygous with B-I, suggesting that they were paramutagenic and therefore interval II recombinants. However, a few light plants could result from B-I spontaneously paramutating to B′, which would obscure the presence of interval I recombinants. To test for paramutagenicity, each independent recombinant allele was either crossed directly to B-I or first self-pollinated and the resulting homozygous || recombinants were crossed to B-I. All 30 BP B′ recombinants that produced viable seed were paramutagenic (Table 1).
We used a slightly different strategy with the purple seed. From 19,267 purple seed that germinated, 2850 recombinants were identified between b1 and gl2. These were transplanted to the field and scored for dark purple vs. light purple pigment in the vegetative part of the plant. The lightly pigmented individuals (168) either were paramutagenic B′||BP alleles or were carrying a neutral B′||BP allele and were light because the B-I allele with which they were heterozygous had spontaneously changed to B′. Screening the light purple plants by PCR for the tmp marker identified 3 recombinants in intervals I or II, which were then tested for their paramutation properties by crossing with B-I. One allele, B′||BP-n1, was a neutral interval II recombinant, while the other two light B′||BP alleles were paramutagenic interval I recombinants.
When the sequences required for paramutation are transferred to B-Peru, this allele becomes paramutagenic; it is fully capable of paramutating B-I in trans. However, the aleurone color remains dark in the paramutagenic B-Peru alleles. Thus, although the B′||BP-pg alleles have light plant pigment and paramutate B-I, causing heritable, reduced expression throughout the plant, the seed expression directed by the B-Peru promoter is not detectably affected.
Cloning of an ∼100-kb fragment upstream of the B′ coding region: Additional upstream sequences were needed to map the recombination breakpoints more precisely. PFGE analyses on B′ DNA indicated that in genomic DNA MluI cuts in the B′ coding region and ∼100 kb upstream (Figures 4 and 5). This fragment was cloned as described in materials and methods. Direct end sequencing showed that one end of the BAC contained the B′ coding sequences upstream of the MluI site. Restriction digestions with eight different enzymes followed by PFGE showed the fragment sizes expected from genomic DNA blot analyses and the previously sequenced upstream B′ sequences (data not shown), indicating that the correct clone was isolated and that it was not rearranged or chimeric. Several different approaches were used to generate a restriction map of the pBACB′1 clone (materials and methods).
Identification of single- and low-copy-number regions in the pBACB′1 clone: To localize the breakpoints in the recombinant alleles, additional probes from the pBACB′1 clone suitable for DNA blot analyses, i.e., fragments of low copy number in maize genomic DNA, were isolated. Previous PFGE experiments (data not shown) showed that the most upstream region of the 100-kb clone was hypomethylated, suggesting that it might contain sequences that were either genic or low copy in the maize genome. The pBACB′1 clone was cut to completion with SwaI and religated, resulting in a clone containing 3.6 kb upstream of the MluI site in the B′ coding region and 12.6 kb downstream from the other end of the clone. Subclones were isolated and sequenced, and single copy regions were identified using DNA blot analyses (Figure 7). To identify additional single and low-copy regions throughout the pBACB′1 clone, a shotgun library containing fragments of ∼800 bp was made and clones containing sequences that were either unique or low copy number in the maize genome were identified (materials and methods).
Shotgun sequencing of pBACB′1 clone: To assist with all subsequent analyses, the pBACB′1 clone was completely sequenced. Shotgun library preparation, sequencing, finishing the complete sequence, and sequence analysis of pBACB′1 were done as described (Dubcovskyet al. 2001). Most of the sequences up to 80 kb upstream of B′ were related to transposons (Figure 7). This region contained five full-length long terminal repeat (LTR) retrotransposons (class I elements); Ji, Opie, PREM-2, Zeon, and Huck. Ji, Opie, and PREM-2 were inserted into a partial Ji element and an additional, partial, Opie fragment (Opie-p) was inserted into the full-length Opie element. The retrotransposons were flanked by sequences related to transposons that move via DNA intermediates (class II elements). These include a region homologous to the transposase of Doppia4 (accession no. AF187822) and a transposon homologous to the MuDR element of maize (D. Lisch, D. Selinger and M. Stam, unpublished results). Two small regions flanking the Zeon element were homologous to part of the maize starch synthase I precursor (accession no. AF036891). The most upstream 20 kb of pBACB′1 contained a 108-bp region homologous to part of an Arabidopsis phosphatidyl inositol 4-phosphate 5-kinase [T51821, 683 amino acids (aa); 68% identity], and a 530-bp region homologous to part of a hypothetical rice protein (accession no. AC084763, 513 aa; ∼35% identity; Figure 7). Upstream of the two hypothetical open reading frames (ORFs) was a region with seven tandem repeats, each containing a single MluI site. Farther upstream, a region of 299 bp was 100% identical to an expressed sequence tag (EST) from a stressed root cDNA library from Zea mays (accession no. AI649454). There was no evidence for any class I or class II elements in this 20-kb region.
The previously isolated subclones, which had been tested for their repetitiveness in the maize genome, were placed on the pBACB′1 restriction map by aligning their sequences with the pBACB′1 sequence (shown graphically in Figure 7). Most of the retrotransposons were not represented in these subclones, because they are highly repeated in the genome (SanMiguelet al. 1998) and were removed from the analyses by hybridization with total genomic DNA. Six of the high-copy-number regions that made it through the screen correspond to the Ji retrotransposon. The intermediate repetitive regions between 0-10 kb and 30-40 kb correspond to the MuDR-related element and the region between the Zeon and Huck elements, respectively. The low-copy-number regions are just upstream of B′, at the other end of the BAC and between 40 and 50 kb (Figure 7).
Recombination breakpoint mapping: To localize the breakpoints in the different recombinant alleles, genomic DNA was isolated from plants containing the various alleles, digested, and subjected to standard gel electrophoresis or PFGE, followed by DNA blot analyses using the unique or low-copy-number probes identified. These data, summarized in Table 1, localized the region required for paramutation to between ∼50 and ∼110 kb upstream of the B′ transcription start site, consistent with our B-615 results.
To define more precisely the breakpoints, the recombination regions were sequenced. For the seven BP B ′-pg alleles with a breakpoint between 107 and 110 kb|| upstream of the B′ transcription start site, additional B′ sequences were needed as this was upstream of the BAC clone. IPCR was used to amplify an additional 3.9 kb of upstream B′ sequence (materials and methods). PCR primers were designed using the B′ sequence and DNA fragments were amplified from B-Peru and the recombinant alleles in the regions surrounding the recombination breakpoints. All fragments were cloned and sequenced and the recombination breakpoints assigned to regions delimited by DNA sequence polymorphisms between B-Peru and B′ (summarized in Table 2 and Figure 8). These data revealed that the region of B′ between 93 and 106 kb upstream of the B′ transcription start site contains sequences required for paramutation. This 13-kb region in B′ is mostly unique or low copy in the maize genome and it does not contain any recognizable transposable elements, nor does it share homology with the promoter-proximal region. The region does contain seven tandem repeats of a sequence not represented elsewhere in the maize genome. DNA blots using probes from the shotgun library (Figure 7), in combination with the BAC sequence, showed that B-615 had the same map as B′ up to 44.3 kb upstream, with the first detectable difference in restriction sites occurring at ∼49 kb.
Pigment phenotype of recombinant alleles reveals an additional regulatory region: Examination of the pigment phenotypes of several recombinant alleles and B-615 revealed the location of sequences required for B′ expression (Table 2). The B′||BP-pg2 allele with a recombination breakpoint at ∼500 bp upstream gave rise to plants that looked like B′, demonstrating that it is possible to transfer sequences mediating B′ expression levels to the promoter-proximal region of B-Peru. Plants with B′||BP-pg1 (B-Peru sequences to ∼93 kb, B′ sequences upstream) looked like B-Peru, indicating that they did not gain the sequences required for B′ transcription. Plants containing the paramutagenic BP||B′-pg alleles 1-4, 19-21, and the neutral BP||B′-n2, -n3 alleles had the same pigment levels as B′ plants, indicating that they retained the sequences required for B′ expression. Of these alleles, the BP||B′-n2 allele retained the least amount of B′ sequences (up to 92 kb), demonstrating that these sequences are sufficient for B′ transcription. In contrast, BP||B′-n1 with a recombination breakpoint at ∼8.5 kb was essentially green, demonstrating the 8.5 kb upstream of B′ is not sufficient for B′ transcription levels. These data indicate that the regulatory sequences responsible for B′ expression are located between ∼8.5 kb (BP||B′-n1) and ∼92 kb (BP||B′-n2). Comparison of B′ and B-615 (same pigment levels and transcription rate) showed that they have essentially the same sequences until ∼49 kb upstream, suggesting that the sequences required for B′ expression lie between 49 and 8.5 kb.
We used recombination mapping to localize the sequences in B′ required for paramutation to a 13-kb region between 93 and 106 kb upstream of the transcription start site. Consistent with the recombination mapping results, the neutral B-615 allele shares very similar sequences (possibly the same sequences) with B′ until ∼49 kb upstream of the transcription start site. The fact that these two alleles are transcribed at the same rates demonstrates that transcription from the same promoter-proximal region is not sufficient for the ability to participate in paramutation.
Experiments have been performed to identify the sequences required for paramutation for two other genes, r1 and p1. The r1 paramutagenic alleles are complex, with three or four copies of the r1 gene and flanking DNA located in large tandem arrays (Egglestonet al. 1995; Panavaset al. 1999). The sequences required for paramutagenic activity were mapped using unequal crossing over among the repeats, revealing that no particular region was required for paramutagenicity, but the strength of paramutation correlated with the number of repeats (Kermicleet al. 1995; Panavaset al. 1999). At the p1 locus, a transgenic approach identified a 1.2-kb fragment normally located 5 kb upstream and 8 kb downstream of the p1 transcription start site, which, when fused to GUS and reintroduced into maize, was capable of paramutating the endogenous allele (Sidorenko and Peterson 2001). Once paramutated, the endogenous allele often is paramutagenic in the absence of the transgene, but the sequences required in the endogenous allele have not been defined. The paramutagenic transgene loci contained multiple copies of the transgene while the endogenous p1 gene is flanked by 5.2-kb direct repeats containing ∼1.5 or 2 copies of the 1.2-kb region, consistent with the involvement of repeated sequences. However, no transgene loci with a single copy of the p1 transgene were obtained, preventing a direct test of the requirement for repeats. A correlation between gene silencing and the presence of repeated sequences has also been observed in many, but certainly not all, examples of transgene-induced gene silencing (Hobbset al. 1990; Assaadet al. 1993; Dorer and Henikoff 1994; Stamet al. 1997; Muskenset al. 2000).
There is no sequence similarity between the 13-kb region from B′, the 1.2-kb region from p1, or the published r1 sequences. The 1.2-kb region of p1 contains moderately repetitive DNA as well as sequences unique to the p1 locus (Sidorenkoet al. 1999); 78 bp is 100% identical to the p1 transcript. At B′, the region required for paramutation is mostly unique or low copy in the genome and shows no sequence or structure characteristic of transposons, but does contain seven tandem repeats of a sequence not represented elsewhere in the maize genome. Intriguingly, DNA blot analyses indicate that two neutral alleles, B-615 and B-Peru, have approximately one copy of the sequence repeated in B′ (data not shown). Further analyses of the region required for b1 paramutation using additional recombination experiments should enable further delineation of the key sequences and determine if the tandem repeats in B′ are involved in paramutation.
All of the recombination events that we isolated occurred immediately upstream of the B′ coding region, ∼92-107 kb upstream, or even farther upstream. For the two recombinagenic regions sequenced, recombination occurred in gene-like, unique, or low-copy-number sequences that had a low level of DNA methylation compared to the sequences in between. For example, the methylation-sensitive enzymes MluI, SalI, and SnaBI cut only genomic DNA in the sequences within and immediately flanking the B′ coding region and 92-107 kb upstream; the multiple sites located in between were not cut and are thus presumably methylated. These results are consistent with the recent findings from studies of recombination around the bz1 locus in maize, which indicate that recombination takes place in maize genes and not in the intergenic regions (Fu et al. 2001, 2002), and earlier work noting the high frequency of recombination within maize genes (Civardiet al. 1994; Pattersonet al. 1995; Dooner and Martinez-Ferez 1997). One caveat is that most of the 80-kb region between the two recombinagenic regions in B′ consists of transposons. Because most of this region in B-Peru has not been cloned and sequenced, it is possible B-Peru and B′ do not share the same transposon sequences in this intergenic region, which would be expected to reduce recombination dramatically. It is also possible that the two alleles do share many of the same insertions, but recombination is suppressed in the regions containing repetitive DNA (Fuet al. 2002), potentially because of chromatin structure. Suppression of recombination between transposons would serve to decrease ectopic recombination and is likely to be a common feature of eukaryotic genomes. In Drosophila and Arabidopsis the rate of meiotic recombination is also reduced in transposon-rich regions of chromosomes (Charlesworthet al. 1994; CSHL/ WUGSC/PEB 2000). It will be necessary to clone and sequence the comparable region from B-Peru to address this hypothesis rigorously.
The sequences upstream of B′ consist of a large region containing primarily retrotransposons, flanked by a few DNA transposons. A similar large block of retrotransposons has been observed for the intergenic regions flanking several other maize genes such as adh1 (SanMiguelet al. 1996) and a1 (Chenet al. 1997). As seen with the Adh1-F allele, the highly repetitive Ji, Opie, and Huck elements that constitute a substantial portion of the maize genome (SanMiguelet al. 1996) are present in the intergenic region upstream of B′. As observed in other regions of the maize genome (SanMiguelet al. 1996; Fuet al. 2001) and other grass genomes (Shirasuet al. 2000; Wickeret al. 2001), these LTR retrotransposons are arranged as nested insertions. Upstream of B′, PREM-2, Opie, Opie-p, and a Ji element are all inserted into a partial Ji element. In contrast, the region around the bz1 locus is relatively gene rich (Fuet al. 2001).
Our studies demonstrate that classical fine-structure recombination mapping is an excellent approach to identify the key sequences required for paramutation. It will be important to use a similar approach to determine if the sequences required for B-I to participate in paramutation map to the same region as in B′. The ability to transfer the paramutation sequences to the neutral B-Peru allele, causing it to become paramutagenic, also suggests that a transgenic approach to further dissect the region required for paramutation should be feasible. Once the minimal sequences are defined in both alleles, it will be important to investigate whether there are sequence differences or differences in chromatin structure between B-I and B′.
Our recombination mapping experiments also identified an additional regulatory region required for B′ expression, located downstream of the sequences required for paramutation, but far upstream of the transcription start site. The observation that B-615 has the same transcription rate as B′ suggests that it retains these necessary regulatory sequences, which might be located within the region conserved with B′ (between ∼8.5 and ∼49 kb upstream of the B′ transcription start site). To our knowledge this is the first example in plants of key regulatory sequences being located so far upstream of a gene. There are hundreds of reports in the literature of transgene analyses in many different plant species, all of which demonstrate that key regulatory sequences controlling developmental and tissue-specific expression are usually near the transcription start site (for review, see Singh 1998 and references therein). While the number of regulatory regions characterized is lower in maize than in plant species with more efficient transformation methodologies, the theme is the same (for example, Russell and Fromm 1997; Hueroset al. 1999; Sidorenkoet al. 2000). Additional studies on more maize genes are needed, but it is tempting to speculate that our findings with B′ are related to its unique paramutation properties, which we have speculated involve boundary elements gone awry (Chandler et al. 2000, 2002). In contrast to most plant promoters, it is not uncommon for regulatory sequences to be located far from transcriptional start sites in animals (for example, Dillonet al. 1997; Siposet al. 1998; Bell and Felsenfeld 2000). The ability to do fine-structure recombination and large-scale mutagenesis experiments in maize makes B′ an excellent model system for further investigations of the mechanisms underlying long-distance cis and trans interactions that control transcription.
We are indebted to Pascal Lennertz and Zahra Mobasher for their expert technical assistance, which was crucial for identifying the recombinants, isolating the BAC clone, and the other molecular experiments described. We appreciate the assistance of Teresa Lavin with the initial PCR assays for recombinants and the initial characterization of B-615 by Ken Kubo. We thank Etienne Kaszas for protocols and advice regarding isolating HMW maize DNA and running pulsed-field gels, Ted Weinert and Mark Orbach for PFGE advice, David Frisch for BAC cloning advice, Yeisoo Yu for advice on shotgun library preparation, Hiroaki Shuzuya for providing pBeloBAC11, and Elizabeth Pierson for help using the Biomek 2000 laboratory automation workstation. This research was funded by a postdoctoral fellowship from the Human Frontier Science Program to M.S., by a postdoctoral fellowship from the National Science Foundation (NSF) to J.E.D. (BIR-9626082), by grants from the NSF to V.L.C. (MCB99-82447, MCB96-03638) and J.L.B. (DBI99-75618), and by a graduate fellowship from CNPq to C.B.
Communicating editor: V. Sundaresan
- Received May 23, 2002.
- Accepted July 16, 2002.
- Copyright © 2002 by the Genetics Society of America