Structural Features and Methylation Patterns Associated With Paramutation at the r1 Locus of Zea mays

Corresponding author: Elsbeth L. Walker, Biology Department, 611 N. Pleasant St., University of Massachusetts, Amherst, MA 01003., ewalker{at}bio.umass.edu (E-mail)

Communicating editor: J. A. BIRCHLER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In paramutation, two alleles of a gene interact and, during the interaction, one of them becomes epigenetically silenced. The various paramutation systems that have been studied to date exhibit intriguing differences in the physical complexity of the loci involved. B and Pl alleles that participate in paramutation are simple, single genes, while the R haplotypes that participate in paramutation contain multiple gene copies and often include rearrangements. The number and arrangement of the sequences in particular complex R haplotypes have been correlated with paramutation behavior. Here, the physical structures of 28 additional haplotypes of R were examined. A specific set of physical features is associated with paramutability (the ability to be silenced). However, no physical features were strongly correlated with paramutagenicity (the ability to cause silencing) or neutrality (the inability to participate in paramutation). Instead, paramutagenic haplotypes were distinguished by high levels of cytosine methylation over certain regions of the genes while neutral haplotypes were distinguished by lack of C-methylation over these regions. These findings suggest that paramutability of r1 is determined by the genetic structure of particular haplotypes, while paramutagenicity is determined by the epigenetic state.

EPIGENETIC inheritance—"mitotically and/or meiotically heritable changes in gene expression that cannot be explained by changes in DNA sequence" (RIGGS et al. 1996)—is a widespread phenomenon that is involved in both normal developmental mechanisms and is a means of coping with repetitive and/or invasive sequences like transposons and viruses. Many epigenetic effects are recognized when expression of a gene is reduced or abolished following introduction of a homologous transgene and are described as "homology dependent gene silencing" (HDGS; reviewed recently by WOLFFE and MATZKE 1999 ). HDGS also operates naturally in repeated endogenous sequences such as transposable elements (FEDOROFF 1996 ; MARTIENSSEN 1996A ). The underlying mechanisms that lead to reduced gene expression are beginning to emerge (WOLFFE and MATZKE 1999 ), but the mechanisms that trigger HDGS remain elusive.

Paramutation is a special case of HDGS in which the homologous sequences that interact are alleles. Two key features define the phenomenon of paramutation (CHANDLER et al. 2000 ). First, two variants of the gene or locus interact in a heterozygote such that one of the variants becomes epigenetically silenced. Second, the silencing is meiotically heritable; i.e., it is maintained in subsequent generations. In paramutation of the r1 locus of maize, a paramutable haplotype, e.g., R-r:std, is made heterozygous with a paramutagenic r1 haplotype, e.g., R-st or R-mb (BRINK 1956 ; BRINK and WEYERS 1957 ). In the heterozygote, simple dominance of R-r:std is observed, producing dark, solid pigmentation of the aleurone layer of the kernel. However, when this heterozygote is subsequently crossed, expression from R-r:std is changed so that R-r:std confers a lighter, "mottled" pattern of anthocyanin deposition. Thus, paramutant R-r:std (designated as R-r') has been partially silenced. Silencing happens in virtually 100% of the kernels into which R-r:std is transmitted, but the level of silencing is quite variable, ranging from nearly complete silence to nearly complete expression. The pattern of expression of the paramutagenic (silencing) haplotype is unchanged following the interaction. Not all r1 variants are paramutable or paramutagenic. Some do not participate in paramutation and are referred to as "neutral."

r1 genes encode nearly identical myc-homologous, helix-loop-helix proteins (LUDWIG et al. 1989 ; PERROT and CONE 1989 ; CONSONNI et al. 1992 ;) that are capable of activating transcription from promoters of structural genes in the anthocyanin biosynthetic pathway (LUDWIG et al. 1989 ; GOFF et al. 1990 ). Because the r1 locus often contains multiple r1 genes, particular variants of the r1 locus are referred to as distinct r1 haplotypes with the designation, R-suffix, in which the suffix is usually based on a single striking phenotypic characteristic of the haplotype. For example, R-marbled (R-mb) confers a marbled pattern of pigmentation to the aleurone. A particular haplotype will comprise one to several individual r1 genes. These genes are not all identical; many alleles exist. These alleles are distinguished from one another without use of the R prefix. For example, the R-r:std haplotype contains four r1 genes with the distinct allele combination "P q S1 S2."

One approach to understanding why certain genes are particularly subject to epigenetic regulation is to examine the structural features that affect their ability to cause silencing or become silenced. Several distinct structural features have been associated with gene silencing: multiple gene repeats (ASSAD et al. 1993 ; ROSSIGNOL and FAUGERON 1994 ; PATTERSON and CHANDLER 1995 ; SIJEN et al. 1996 ; YE and SIGNER 1996 ; BENDER 1998 ; STAM et al. 1998 ), the presence of inverted repeats (QUE and JORGENSEN 1998 ; STAM et al. 1998 ; LUFF et al. 1999 ), the presence of transposon regulatory sequences (FEDOROFF 1996 ; MARTIENSSEN 1996B ), and the juxtaposition of low and high G/C regions (SCHLAPPI et al. 1993 , SCHLAPPI et al. 1994 ; JAKOWITSCH et al. 1999 ). Interestingly, r1 haplotypes that participate in paramutation share many of these features.

The physical structure of a single paramutable haplotype, R-r:std, has been the subject of intense investigation for many years (STADLER and NUEFFER 1953 ; STADLER and EMMERLING 1956 ; DOONER 1971 ; DOONER and KERMICLE 1971 , DOONER and KERMICLE 1974 ; KERMICLE 1980 ; ROBBINS et al. 1991 ; WALKER et al. 1995 ). Four distinct r1 genes are present in R-r:std (see Fig 1A): P (Plant color), a functional gene that colors coleoptiles, roots, and anthers; q, a nonfunctional gene fragment with strong sequence similarity to the promoter region of P; and two S (Seed color) genes, S1 and S2, that color the aleurone of the seed (ROBBINS et al. 1991 ; WALKER et al. 1995 ). The S1 gene is in an inverted orientation relative to the other r1 genes of the complex and appears to have arisen as a duplicate copy of the S2 gene. The P gene is located 190 kb proximal to the rest of the complex (which is referred to as the S-subcomplex) and is not strongly affected in paramutation (BRINK and MIKULA 1958 ; BROWN 1966 ). Derivative alleles of R-r:std that lack P are fully paramutable, indicating that instability is a function of the S-subcomplex (BROWN 1966 ).

View larger version (33K): In this window In a new window Download PPT slide   Figure 1. Schematic representations of three r1 haplotypes of known structure. (A) R-r:std (ROBBINS et al. 1991 ; WALKER et al. 1995 ). r1 sequences situated upstream of the Dop insertion site are shown in green; sequences downstream of this site are indicated in blue. Sequences derived from the transposable element Dop are indicated in purple and red with 3' Dop sequences in red and 5' Dop sequences in purple. Small arrowheads shown above and below indicate repeat motifs found in Dop end sequences. The white portion of corresponds to the rearranged region that is not derived directly from Dop. (B) R-st (EGGLESTON et al. 1995 ; MATZKE et al. 1996 ; CHANDLER et al. 2000 ; E. L. WALKER, unpublished results). The colors are as indicated for A. Additionally, the unique upstream portion of Sc is indicated in black. (C) R-mb (PANAVAS et al. 1999 ). The colors are as indicated for A and B. (D) Positions of primers used in PCR analysis. The color scheme is identical to that used for A–C. The positions of primers are indicated by numerals. Primers that direct synthesis toward the right are indicated above each gene; primers that direct synthesis toward the left are indicated below each gene. Primer 1, oLc-5'-1; primer 2, oR-5'-2; primer 3, oSc-5'; primer 4, oQ-3'; primer 5, oR-3'-3; primer 6, oR-3'-1, primer 7, oR-3'-2; primer 8, oS1-5'-1; primer 9, oS1-5'-2; primer 10, oS2-5'-1 and oS2-5'-2 (these primers overlap).

Within the S-subcomplex, the S genes, S1 and S2, are arranged in a head-to-head (5' ends together) orientation with only 381 bp separating them. This 381-bp sequence is called (sigma) and constitutes the promoter for both of the S genes (WALKER et al. 1995 ; MAY and DELLAPORTA 1998 ). Thus, the S1 and S2 genes of R-r:std can be thought of as having a novel promoter that is distinct from the promoters of other r1 alleles. The majority of is derived from the 3' end of a transposon called Doppia (Dop) and contains multiple copies of a sequence motif that can be bound by the Dop-encoded protein, DOPA (BERCURY et al. 2001 ). A second, very small fragment of the Dop 3' end is present just upstream of S1. This fragment contains no DOPA binding sites. Between the two Dop-derived portions of , a short "rearranged" region is present. The origin of the sequences in this region is not clear, but they appear not to be derived from the original Dop insertion. The q gene fragment contains upstream r1 sequences that are interrupted by a Dop 5' end (BERCURY et al. 2001 ). Thus, q is an r1 promoter with no adjacent coding sequences. Another set of DOPA binding sites is located in the Dop 5' sequences adjacent to q.

The simplest way to account for the structure of the S-subcomplex is to imagine that a Dop element inserted into r1 and then fractured to generate two entities: a q fragment with an adjacent Dop 5' end and an S2 fragment that included a Dop 3' end upstream of a complete r1 coding region. A few base pairs of the Dop 3' end, together with the entire coding region of S2, were duplicated, and these duplicated sequences inserted in reversed orientation upstream of the original S2 copy to form S1 (WALKER et al. 1995 ).

Detailed physical structures are available for two paramutagenic haplotypes. The canonical paramutagenic haplotype, R-st (see Fig 1B), contains four directly repeated genes designated Sc (Self color), Nc1, Nc2, and Nc3 (Near colorless; EGGLESTON et al. 1995 ; KERMICLE et al. 1995 ). Dop sequences are located at the 5' ends of each of the Nc genes of R-st (MATZKE et al. 1996 ). A second paramutagenic haplotype, R-mb, comprises three r1 genes: Sc, Lcm1, and Lcm2 (see Fig 1C). These are again arranged as direct repeats (PANAVAS et al. 1999 ). None of the genes in the R-mb haplotype contain Dop sequences. This rules out a role for Dop sequences in paramutagenicity. The only gene type that is common to both R-st and R-mb is Sc. It should be noted, however, that the Sc gene of either haplotype can be replaced by other r1 genes through unequal crossing over, and the resulting derivatives retain paramutagenicity (KERMICLE et al. 1995 ; PANAVAS et al. 1999 ). In studies of both R-st and R-mb, the strength of paramutagenicity was directly correlated with gene copy number (KERMICLE et al. 1995 ; PANAVAS et al. 1999 ). Thus, paramutagenicity appears to depend on the presence of multiple gene copies and seems not to be affected by the particular combination of genes found at the locus.

Here we present a survey of the genic compositions of a set of multigenic r1 haplotypes that includes paramutable, paramutagenic, and neutral types. A combination of PCR, genomic blotting, cloning, and sequencing was used to determine the gene types present in each haplotype. Paramutable haplotypes were found to be structurally very similar to one another, while paramutagenic haplotypes were much more diverse in their genic composition. Surprisingly, neutral haplotypes had no distinct structural features that distinguished them from paramutagenic haplotypes. Neutral haplotypes were, however, distinct from paramutagenic haplotypes in the pattern and level of DNA methylation present at sites in the upstream portions of their genes. These data suggest that paramutagenicity is dependent on the chromatin conformation of the locus rather than on the presence of particular sequence features. In contrast, the marked structural similarity of the paramutable haplotypes suggests that particular sequence level features may be most important for paramutability. Finally, the structural studies presented here shed light on the evolutionary history of these complex r1 haplotypes.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Genetic stocks:
All stocks were maintained in the W22 inbred background. All stocks are homozygous dominant for the a1, a2, c1, c2, bz1, and bz2 genes necessary for anthocyanin synthesis in aleurone and homozygous recessive for the pl1 and b1 genes; see DOONER et al. 1991 for descriptions. The R-n:geographic (VAN DER WALT and BRINK 1968 ), R-r:std (STADLER 1948 ; WALKER et al. 1995 ), and R-mb (WEYERS 1961 ; PANAVAS et al. 1999 ) haplotypes used in this study have been described previously. All but R-mb confer solid aleurone pigmentation. R-mb confers a pattern of aleurone color consisting of large pigmented sectors on a colorless background. The r allele confers a colorless phenotype and does not hybridize to r1 probes on genomic blots nor does it give PCR products with any combination of r1-specific primers that we have tested. The W23 strain was used as a colorless aleurone tester for genetic studies that did not include any molecular analyses.

Paramutation test:
The R-n:geographic haplotypes used in this study were kindly provided to us by Dr. J. Kermicle. Crosses of R-n:geographic haplotypes to r, R-r, and R-mb were performed during the summers of 1997 and 1998. Outcrosses to the W23 tester were made in the summer of 1999. The scoring of aleurone color was by visual inspection on whole ears. A minimum of four ears was scored for each haplotype. In cases where weak color changes were suspected, kernels were stripped from the ears, sorted, and scored on a seven-point color scale.

PCR:
DNA for PCR reactions was prepared in replicates from two sources: leaves of mature, flowering plants and ungerminated embryos from imbibed kernels (DELLAPORTA 1994 ). All PCR reactions were performed using the Expand High Fidelity PCR system (Boehringer Mannheim, Indianapolis) according to manufacturer's instructions. The PCR primers used in this study were as follows:

• oR-3'-1 (5' GGCATGCGTATGCTGGAAAGACGT 3')

• oR-3'-2 (5' TAGCTCCAGTTGATGCTCCTGGCG 3')

• oR-3'-3 (5' CCATGCGAAGGGTAGAGAAGAACCG 3')

• oR-5'-2 (5' AAAAGCAATCAGAAGCTAAAAACACGG 3')

• oLc-5'-1 (5' CTCCAAAAGGCTCAATTCTCCTCCCC 3')

• oQ-3' (5' CACATTTTCGTCGGTCACTCTTGCC 3')

• oSc-5' (5' TGCAAAGTATTCCCTTCTCTCCACCTCA 3')

• oS2-5'-1 (5' AAAATAAGTCGTTTTCGTCGGTACCGA 3')

• oS2-5'-2 (5' ATAAGTCGTTTTCGTCGGTACCGA 3')

• oS1-5'-1 (5' CGTGGGCCAAAAACCCACGAAA 3')

• oS1-5'-2 (5' CGTCGGTACCGACGAAAACGACT 3')

Inverse PCR:
DNA from leaves of the desired r1 geographic haplotypes was digested with HindIII and fractionated on a 1% agarose gel. The gel slice corresponding to the MW interval 2.5–4.5 kb was excised and the restriction fragments were purified using the Qiaquick gel extraction procedure (QIAGEN, Valencia, CA). The purified DNA was self-ligated for 2 hr at 22° at a concentration of 2 ng/µl. The self-ligated DNA fragments served as templates in PCR reactions using the Expand High Fidelity PCR system (Boehringer Mannheim) with oR-3'-1 and oLc3676_3702 (5' GGTGAGGCCCATCCAGATAACATAAGC 3') primers. The PCR fragments were cloned into the SmaI site of the pTZ19U plasmid and sequenced.

Genomic blotting:
Preparation of DNA for genomic blots was from leaves of mature plants as described previously (WALKER et al. 1997 ). The SAH, 1011, cDNA-B, and Scm5' probes have been described previously (WALKER et al. 1995 ; PANAVAS et al. 1999 ). The Lc1013 probe is derived from a PCR product extending from -470 to +71 bp relative to the end of Lc cDNA determined by LUDWIG et al. 1989 . The PCR primers to generate the Lc1013 fragment were oR61 (5' TCCCAACCATCATCAACTCGCTAGCCAAACACA 3') and oR-3'-3. Southern blotting was performed as described previously (ROBBINS et al. 1989 ).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

r1 geographic haplotypes:
The set of haplotypes used in this study was first described by VAN DER WALT and BRINK 1968 , who characterized a diverse collection of colored aleurone haplotypes, collected mainly from South, Central, and North America. Van Der Walt and Brink characterized each haplotype as being paramutagenic, paramutable, or neutral. The haplotypes are referred to generically as R-n:geographic. The collection, which had been introgressed into standard W22 background, was kindly provided to us by Jerry Kermicle. Paramutability was judged on the basis of reduction in aleurone pigmentation of R-n:geographic kernels in outcrosses from R-n:geographic/R-mb heterozygotes. Comparison was made to R-n:geographic kernels outcrossed from R-n:geographic/r_ heterozygotes, in which paramutation does not occur. Paramutagenicity was assessed by the capacity of each R-n:geographic haplotype to cause a reduction in pigmentation of R-r:std kernels in outcrosses from R-n:geographic/R-r:std heterozygotes. R-r:std kernels from outcrosses of r_/R-r:std heterozygotes were used as controls for full R-r:std pigmentation. It should be noted that paramutagenic strength and paramutability were not measured quantitatively for this study. Instead, whole ears were compared to appropriate control ears and scored qualitatively as "full color," "medium," "light," or "colorless." In a few cases, kernels were stripped from the ears, sorted, and scored on a seven-point color scale to resolve potential weak color changes, as described previously (PANAVAS et al. 1999 ). In every case, the suspected small changes related to haplotypes that were classified as neutral and, in every case, there was no statistically significant difference in color between control and experimental ears. No exceptions to Van Der Walt's previous characterization were noted.

PCR-based analysis of geographic r1 haplotypes:
Five distinct promoter types have been previously identified for particular genes in the R-r:std, R-st, and R-mb haplotypes (Fig 1; WALKER et al. 1995 ; MATZKE et al. 1996 ; PANAVAS et al. 1999 ). These five distinct structural features are referred to here as Lc-like, Sc-like, q-like, S1-like, and S2-like. The 5' portion of Lc-like alleles is similar at the level of sequence to the chromosomally displaced r1 gene, Lc. Members of this group include the P gene in R-r:std (Fig 1A; WALKER et al. 1995 ), the Lcm genes in R-mb (Fig 1C; PANAVAS et al. 1999 ), and the chromosomally displaced Lc and Sn genes (LUDWIG et al. 1989 ; TONELLI et al. 1991 ). Sc-like genes are identified on the basis of sequence similarity to the distinct upstream portion of the Scm gene of R-mb (PANAVAS et al. 1999 ). Sc-like genes are found in R-mb and R-st (Fig 1B and Fig C). Structures designated as q-like are similar to the q component of R-r:std, which has a Doppia transposable element 5' end located downstream of an Lc-like promoter sequence (Fig 1A). The designation q-like in this study does not necessarily imply a fractured Doppia element like the one found at q of R-r:std, but could in principle include an intact Dop element downstream of the r1 promoter region. S2-like genes are produced coincident with the formation of a q-like allele, as they are the right end of Doppia juxtaposed to an r1 coding region. The designation S2-like again does not imply a fractured Dop element. Examples of S2-like alleles include the S2 gene of R-r:std (Fig 1A) and the Nc genes of R-st (Fig 1B; MATZKE et al. 1996 ). S1-like alleles have r1 coding sequences in a position and orientation relative to a Doppia right end that is indicative of an r1 gene duplication and inversion similar to that found in R-r:std (Fig 1A). Thus, the presence of an S1-like gene indicates the presence of a gene inversion.

We designed PCR primers that allowed us to amplify each specific promoter. The positions of each of these primers are shown in Fig 1D. For Lc-like promoters, the primers were chosen so that the 5' primer, oLc-5'-1, is complementary to Lc in a region that is distinct from the Sc upstream sequence. It primes at a position 1.2 kb upstream of the Lc transcription start site. The 3' primer, oR-3'-1 was taken from the Lc sequence downstream of the position where Doppia is inserted in q and is common to all r1 genes that have an intact coding region. Thus, the primer combination oLc-5'-1 and oR-3'-1 is specific for Lc-like sequences and will not amplify a product from Sc or from q. Note that the oR-3'-1 primer also primes in both S1 and S2, so, theoretically this primer alone is sufficient to yield a PCR product from the S1/S2 inversion. In practice, however, such products were never observed using only the oR-3'-1 primer, probably because hairpin formation during the reaction prevents polymerase from successfully synthesizing such fragments. Sc-like promoters were amplified using the 5' primer, oSc-5', which is specific to Sc, in combination with the 3' primer oR-3'-3, which is complementary to the region common to Scm, Lc, and q. q-like structures were detected using the 5' primer, oR-5'-2, which is common to q and Lc, in combination with the 3' primer, oQ-3', which is specific to Doppia sequences adjacent to q of R-r:std. For both S1-like and S2-like genes, two redundant primer sets were chosen. The 5' primers for the sets (oS1-5'-1, oS1-5'-2, oS2-5'-2, and oS2-5'-1) were chosen from the Dop-derived portion of . The 3' primers (oR-3'-1 and oR-3'-2) were chosen from r1 transcribed sequences downstream of the Doppia insertion site.

PCR products were analyzed on agarose gels. In almost all cases, either a single band or a blank lane was observed. When more than one band was observed or when a product of unexpected size was observed, the fragments were isolated from the gel and sequenced to confirm their identity.

Table 1 summarizes the results of the PCR analysis. All paramutable haplotypes were positive for q, S1, and S2. Furthermore, because overlapping S1 and S2 products were amplified in all cases, we conclude that all paramutable haplotypes contain the duplication/inversion characterized by the S-subcomplex of R-r:std. Five of the paramutable haplotypes (R-r:India PI 166163, R-r:India PI 1210551, R-r:Kansas PI 222629, R-r:Missouri PI 222889, and R-r:Oklahoma PI 213748) had an Lc-like gene in addition to S-subcomplex structures. These five haplotypes also gave S1 and S2 amplification products identical to those of R-r:std (not shown); they appear to be very similar to R-r:std. Three other haplotypes (R-g:New Mexico PI 218148, R-r:Turkey PI 167989, and R-g:Argentina PI 162573) had S1 and S2 amplification products identical to those of R-r:std, but did not appear to contain an Lc-like component.

In 8 out of 16 paramutable haplotypes (R-g:Arizona PI 213729, R-g:Arizona PI 213738, R-g:Arizona PI 218175, R-g:Arizona PI 218178, R-g:Canada PI 214199, R-g:N. Dakota PI 213799, R-g:New Mexico PI 218170, and R-g:S. Dakota PI 213779) no S1-specific PCR products were generated using the primer set oS1-5'-1 and oR-3'-1. When the redundant primer set oS1-5'-2 and oR-3'-2 was used, S1-specific products were observed, but they were 0.4 kb smaller than expected based on the sequence of the S1 gene of R-r:std. We sequenced two of these smaller PCR products from haplotypes R-g:Arizona 213738 and R-g:New Mexico 218170. The two sequences were identical and had a 469-bp deletion relative to R-r:std. At the location of the deletion, a novel portion consisting of 91 bp of rearranged Doppia transposable element sequence was present. The deletion involves 166 bp of and 303 out of 306 bp of the 5' nontranslated region of S1. A GenBank search revealed a perfect match with the r1 haplotype R-d:Catspaw (GenBank U93178). Thus, we refer to these haplotypes as "Catspaw" type. We predicted that Catspaw types, which are missing a portion of the S1/S2 inversion owing to their truncated S1 gene, should be amplified by a set of R primers (oR3'-1 and oR3'-2). Amplification was predicted from the oR3'-2 site in S1 to the oR3'-1 site in S2 (see Fig 1) because, owing to the missing 5' portion of S1 in Catspaw types, no hairpin will be formed in such a product. Amplification of this fragment was successful for all Catspaw haplotypes. The amplification of these products also confirms the presence of a gene inversion in these haplotypes, which cannot otherwise be established for Catspaw types owing to the absence of amplification with oS15'-1 and oR3'-1.

None of the paramutagenic or neutral haplotypes contained either q- or S1-like genes, suggesting the absence of a gene inversion within these haplotypes. All paramutagenic haplotypes contained at least two genes, but there was no strict correlation of gene identity with paramutagenicity. All three of the other gene types (S2-like, Lc-like, and Sc-like) were detected in paramutagenic haplotypes. Sc-like genes were found in four of the seven paramutagenic haplotypes. R-g:Bolivia 724, R-g:Bolivia 568, and R-g:Chile 406 are the first naturally occurring paramutagenic haplotypes that we are aware of that do not contain an Sc-like gene.

To address whether multiple Lc-like genes might have been amplified from a single haplotype in which Lc-like products were detected, the PCR products were tested for NdeI restriction site polymorphisms. Such polymorphisms were detected between the Lcm1 and Lcm2 genes of R-mb (PANAVAS et al. 1999 ). When a homogeneous 1.5-kb Lc-specific PCR product is digested with NdeI, either two fragments (1.4 and 0.1 kb), e.g., R-g:Bolivia 1004, or three fragments (1.0, 0.4, and 0.1 kb), e.g., R-g:Peru San Miguel, are observed. The presence of four restriction fragments (1.4, 1.0, 0.4, and 0.1 kb) after NdeI digestion of Lc-specific PCR products indicates heterogeneity of the PCR product and suggests that amplification occurred from at least two different templates. The NdeI restriction analysis indicates that haplotypes R-g:Bolivia 724, R-g:Chile 406, and R-g:Peru 568 have at least two Lc-like genes.

All four neutral haplotypes tested had at least one Lc-like gene. One of the Lc-like genes of R-r:Ecuador 1172 has a 214-bp deletion (-265 to -53 relative to the putative transcription start site; LUDWIG et al. 1989 ) as revealed by sequence analysis of its unexpectedly small PCR product. This 0.2-kb size difference between two Lc-like fragments from R-r:Ecuador 1172 was confirmed by Southern blot analysis described below. R-r:Ecuador 1172, R-g:Bolivia 1004, and R-g:Peru Corongo ANC150 also have an Sc-like gene. In terms of gene number, the neutral haplotypes R-r:Ecuador 1172, R-g:Bolivia 1004, and R-g:Peru Corongo ANC150 are not strictly different from paramutagenic haplotypes, since they appear by this analysis to contain either three or two genes. Only one neutral haplotype in the collection (R-g:Bolivia 716-6759) has a single gene, making it distinct from the paramutagenic haplotypes analyzed here.

Genomic blot analysis of geographic r1 haplotypes:
Equal amounts of genomic DNA from each haplotype were digested with HindIII for Southern blot analysis. Each blot was hybridized sequentially with four different probes. The Lc1013 probe hybridizes to q-, Lc-, and Sc-like genes but not to S-like genes (green in Fig 1). The Scm5' probe (PANAVAS et al. 1999 ) hybridizes only to Sc-like genes and does not detect q-, Lc-, or S-like genes (black in Fig 1). The SAH probe is derived from transcribed regions and recognizes all r1 genes that have a coding region (blue in Fig 1). Thus, it recognizes all but q-like genes. The 1011 probe is specific for the rearranged part of at R-r:std (white in Fig 1). The Catspaw-type S-subcomplexes, which are missing this region, do not hybridize to 1011 probe nor do Lc-, q-, and Sc-like genes.

Sc-like genes are distinguished by hybridization to Scm5' probe and also are expected to hybridize to Lc1013 and SAH probes, presuming that downstream sequences are also present. All three probes are expected to hybridize to a single HindIII fragment from a given Sc-like gene and, thus, a band of a particular molecular weight that hybridizes to all three of these probes is judged to be Sc-like. Lc-like genes hybridize to Lc1013 and SAH, but not to the Scm5' probe. Again, both probes are expected to hybridize to the same HindIII fragment. q-Like structures are expected to hybridize only to Lc1013 probe since they have an Lc-like promoter but no coding region immediately downstream. The S1/S2 inversion is recognized using the 1011 and SAH probes, which detect a single HindIII fragment that contains the 5' ends of both S-like genes together with the intervening region. Note that if the rearranged part of is missing, as in haplotypes with Catspaw-type S-subcomplexes, no hybridization to 1011 will occur. Furthermore, the 1011 probe does not hybridize to the Nc genes of R-st (not shown) and thus is not expected to hybridize to any derivatives that lack the S1/S2 inversion. The S genes of Catspaw-type haplotypes should exhibit a HindIII fragment that is 0.4 kb smaller than that of R-r:std, when detected with the SAH probe. Finally, S2-like genes that are not part of an inversion complex will hybridize to the SAH probe, but not to any of the other probes used in this analysis.

Table 2 summarizes the Southern blot analyses. For each paramutable haplotype, the Southern analysis confirmed the gene compositions determined by PCR. In the Catspaw-type S-subcomplexes (R-g:Arizona 213729, R-g:Arizona 213738, R-g:Arizona 218175, R-g:Arizona 218178, R-g:Canada 214199, R-g:N. Dakota 213799, R-g:New Mexico 218170, and R-g:S. Dakota 213779) the S1/S2 fragment (detected by the SAH probe) is 0.4 kb smaller than that of the R-r:std types. This result is consistent with the PCR data showing a 0.4-kb deletion in these haplotypes. Furthermore, the Catspaw-type S-subcomplexes did not hybridize to the 1011 probe (as expected), while the standard type S-subcomplexes each had a 5.0-kb HindIII fragment that hybridized to this probe. The q fragment in each of these Catspaw-type haplotypes is 0.4 kb larger than that in R-r:std, which again indicates their similarity to each other and divergence from the R-r:std type of S-subcomplex. The HindIII fragments from Lc-like genes migrated either at 4.0 kb (similar to the P gene of R-r:std) or at 3.7 kb (similar to the Lcm1 and Lcm2 genes of R-mb). The S2-like genes that are not adjacent to an inverted S1 copy are represented by 3.5-kb HindIII fragments. The Sc-like genes were represented by 4.5-, 4.0-, or 3.3-kb bands. The 3.3 kb band observed in R-g:Peru San Miguel and R-r:Ecuador 1172 hybridized exclusively to Scm5' but not to the downstream probes Lc1013 and SAH. It is possible, but unlikely, given the lack of hybridization to the Lc1013 probe, that this fragment could represent an Sc promoter with a Dop 5' end adjacent. This arrangement could come about, for example, through recombination between a q gene and an Sc gene. (The relevant region of homology is shown in green in Fig 1.) To test this possibility, PCR reactions were performed using the primer combination oSc-5' and oQ-3', but no products were observed. This cannot conclusively demonstrate the absence of an Sc/Dop configuration, however, since we cannot be certain whether primer sites are present in the genes in question. We can state that R-g:Peru San Miguel and R-r:Ecuador 1172 appear to contain an additional Sc-like fragment that lacks downstream promoter and coding sequences and that was not detected using PCR.

In addition to determining gene identity, we evaluated the number of genes present in each haplotype. Since an equal amount of DNA was used in each lane, it was possible to judge the gene copy number even in cases with comigrating fragments. The intensity of the 3.7-kb band hybridizing to SAH and Lc1013 probes indicates that paramutagenic haplotypes R-g:Bolivia 724, R-g:Chile 406, R-g:Peru 568, and R-g:Peru San Miguel contained two Lc-like genes. We also found that R-g:Bolivia 1520 and R-g:Chile 370 have two S2-like genes. The presence of at least two S2-like genes in R-g:Chile 370 was confirmed by sequence polymorphism of the cloned S2 I-PCR fragments from this genotype (see description below).

I-PCR analysis of S2-like genes from paramutagenic haplotypes:
We used an inverse PCR (I-PCR) approach to investigate the upstream sequences in the S2 genes of paramutagenic haplotypes. Genomic DNA from R-g:Bolivia 724, R-g:Bolivia 1520, R-g:Chile 370, and R-g:Peru 1304-2993 was digested with HindIII and the fragments ranging from 2.5 to 4.5 kb were purified, self-ligated, and used as templates in I-PCR reactions. The primers were chosen so that the unknown 5' portion of S2 genes would be amplified. A band of the expected size (1.5 kb) was abundant in the I-PCR reactions from all four genotypes, and direct sequencing of the ends of these PCR products indicated that all four products are essentially identical in sequence. The I-PCR fragments from R-g:Bolivia 724 and R-g:Chile 370 were cloned for complete sequencing; however, this analysis was complicated by the presence of an 150-bp region with strong secondary structure, which was never successfully sequenced. However, 1021 bp of sequence from the 5' ends and 312 bp of sequence from the 3' ends were obtained. The Dop insertion site in the S2 genes from R-g:Bolivia 724 and R-g:Chile 370 is identical to that of the S1 and S2 genes of R-r:std. As in R-r:std, the terminal inverted repeat of Dop is missing its last 5 bp, suggesting that this sequence variation was present in the progenitor of all these haplotypes. Sequence similarity to the S2 gene of R-r:std continues for 221 bp of Dop-derived () sequence (Fig 2). The S2 genes of R-g:Bolivia 724 and R-g:Chile 370 contain additional Dop sequences that are not present in the S2 genes of either standard or Catspaw-type paramutable haplotypes. Additional Dop subterminal repeat element copies are present, and the terminal 269 bp of the transcribed portion of the Dop-encoded protein DOPA are also present. Upstream of the Dop sequences in R-g:Bolivia 724 and R-g:Chile 370, there is a region that shares similarity to upstream sequences at the maize adh1 locus (GenBank AF123535; TIKHONOV et al. 1999 ). In all, we estimate that the S2 genes of R-g:Bolivia 724 and R-g:Chile 370 contain 750 bp of sequence that is derived from the Dop 3' end. Thus, the S2 genes found in paramutagenic haplotypes have, in their 5' ends, additional Dop sequences not found in the regions of paramutable haplotypes, but do not appear to contain intact Dop elements.

View larger version (9K): In this window In a new window Download PPT slide   Figure 2. Extent of Dop sequences at S2-like alleles. r1 sequences are indicated in black; Dop sequences are represented in white; horizontal lines represent DNA that is not apparently derived from either r1 or Dop. Shown are schematic representations of the S1 and S2 genes from R-r:std and an S2-like gene from the paramutagenic haplotype R-g:Chile 370. The 5-bp terminal truncation of Dop sequences at the Dop/r1 junction is represented by the imperfect arrowhead. The 150-bp unsequenced region of the S2-like gene of R-g:Chile 370 is indicated by hash marks.

Structural summary:
The overall structural summary derived from combined PCR, Southern blot, and sequencing data is shown in Table 2. There is a strict correlation between paramutability and the presence of two structural features: a q-like gene and a gene inversion comprising S1- and S2-like genes. The presence or absence of an Lc-like gene did not correlate with paramutability, which is consistent with previous studies showing that the P gene of R-r:std is not necessary for its paramutability (BROWN 1966 ). The S-subcomplexes of paramutable haplotypes fall into two categories: some are like R-r:std in that they have two intact S genes, and others are similar to R-d:Catspaw in having a single intact S gene (S2) and a missing portion of the 5' noncoding sequences of the S1 gene.

The structure of paramutagenic haplotypes is more variable. Notably absent are q- or S1-like genes. Another common feature is that all paramutagenic haplotypes comprise at least two genes. Haplotypes containing up to five repeated components were identified. None of the paramutagenic haplotypes appeared to contain inverted gene copies, nor did any contain q-like genes. S2-like genes, however, were common.

Three of the four neutral haplotypes analyzed appeared to be structurally similar to paramutagenic haplotypes. These contained two or four genes that were either Lc-like or Sc-like. One neutral haplotype, R-g:Bolivia 716-6759, contained only a single Lc-like gene. None of the four neutral haplotypes had S2-like genes, nor did any contain features (gene inversion, q-like genes) associated with the S-subcomplex.

Cytosine methylation in r1 geographic haplotypes:
Because no physical structures were identified that distinguished paramutagenic haplotypes from neutral haplotypes, we examined whether there are differences in epigenetic features (i.e., chromatin level differences) that distinguish these two classes of haplotypes. To address this issue, we examined the pattern of C-methylation present in each r1 haplotype. We used the methylation-sensitive restriction enzyme HpaII to assess methylation of cytosine residues in CCGG sites within r1 genes. HpaII is sensitive to the methylation of either of the C residues within its recognition site. MspI, an isoschizomer of HpaII, was used to judge the presence of restriction sites. MspI is less sensitive to C-methylation and though it does not cut when the first C in its recognition site is methylated, it does cut when the second C in the CCGG site is methylated. HpaII/MspI sites were predicted on the basis of sequences of well-characterized r1 alleles from the R-r:std and R-mb haplotypes. The presence of each predicted HpaII/MspI site was confirmed in one of two ways. In many cases, MspI cleavage was detected, confirming the presence of a particular CCGG site. For those cases in which MspI did not cut, or when MspI cutting could not be detected on genomic blots due to the location of the probe, the relevant portion of the gene was amplified by PCR (or I-PCR, in the case of S2-like genes from paramutagenic haplotypes, see above) and digested to map these sites. The predicted sites were present in all cases (not shown). Genomic DNA of each geographic haplotype was digested with the methylation-insensitive enzyme HindIII and also with a combination of HindIII and HpaII or HindIII and MspI. The blots were hybridized sequentially with the SAH and Lc1013 probes.

Maps showing the HindIII and HpaII/MspI sites in the various r1 genes of the haplotypes are shown in Fig 3A, Fig 4A, and Fig 5A. There are two HpaII/MspI sites present close together in the large second intron of typical r1 genes. These are referred to as "intronic HpaII sites." In Lc-like genes there are two HpaII/MspI sites that lie just upstream (39 and 141 bp, respectively) of the transcription start site (LUDWIG et al. 1989 ; TONELLI et al. 1991 ; CONSONNI et al. 1992 , CONSONNI et al. 1993 ). Sc-like genes also contain these two sites, but their position relative to the start of transcription is not known, since the transcription start site of Sc has not been reported. In S genes, only one of the two HpaII/MspI sites is present, because sequences have replaced the upstream portion containing one of these two sites. Multiple transcription start sites have been mapped for the S1 and S2 genes of R-r:std (MAY and DELLAPORTA 1998 ), but because the transcribed portions of S1 and S2 are identical in this region, it is not possible to know whether S1, S2, or both genes initiate transcription at a particular site. These uncertain transcription start sites are indicated as gray arrows in Fig 4A. There is one start site within that must be used by the S2 gene. The HpaII/MspI sites at this position of all genes are referred to as "transcription start HpaII/MspI sites," regardless of whether the precise transcription start is known. Finally, there are HpaII/MspI sites that are present upstream of the transcription start HpaII/MspI sites, which are referred to as "promoter HpaII/MspI sites." It is important to note that, in contrast to the aforementioned HpaII/MspI sites, which are common to all r1 genes, the promoter HpaII/MspI sites may be at nonhomologous positions in the various r1 gene types.

View larger version (56K): In this window In a new window Download PPT slide   Figure 3. Cytosine methylation tests of the 5' ends of r1 genes from paramutagenic haplotypes. (A) Maps of the gene types (Sc-like, Lc-like, S2-like) found in paramutagenic haplotypes. M, HpaII/MspI; H, HindIII. The position and extent of the SAH probe is represented by black rectangles. Sizes of the predicted restriction fragments that result from complete digestion and sizes of the partial digestion products observed on genomic blots are indicated below the map. In the map of Sc-like genes, the solid line represents cloned sequences; the dashed line shows an uncloned region. Three sets of HpaII/MspI sites are indicated: promoter, transcription start, and intronic. (B) Southern blots of paramutagenic haplotypes. The genomic DNA was digested with HindIII alone (lanes 1, 4, 7, 10, 13, 16, 19, 22, and 25), HindIII + HpaII (lanes labeled H), or HindIII + MspI (lanes labeled M); separated on a 1% agarose gel; transferred to a membrane; and hybridized with the SAH probe. The molecular weights in kilobases of the fragments detected are indicated at the left and right. Contents of each lane are as follows: lanes 1–3, R-mb; lanes 4–6, R-g:Bolivia 724; lanes 7–9, R-g:Bolivia 1520; lanes 10–12, R-g: Chile 370; lanes 13–15, R-g:Peru 568; lanes 16–18, R-g:Peru San Miguel; lanes 19–21, R-g:Chile 406; lanes 22–24, R-g:Peru 1304-2993.

View larger version (62K): In this window In a new window Download PPT slide   Figure 4. Cytosine methylation tests of the 5' ends of r1 genes from paramutable haplotypes. (A) Maps of the gene types (Lc-like, S1/S2-like from R-r:std, and Catspaw-type S1/S2-like) found in paramutagenic haplotypes. The q-like genes of these haplotypes are not shown, as they are not detected by the SAH probe. M, HpaII/MspI; H, HindIII. The position and extent of the SAH probe is represented by black rectangles. Sizes of the predicted restriction fragments that result from complete digestion and sizes of the partial digestion products observed on genomic blots are indicated below the map. Three sets of HpaII/MspI sites are indicated: promoter, transcription start, and intronic. (B) Southern blots of paramutable haplotypes. The genomic DNA was digested with HindIII alone (lanes 1, 4, 7, 10, 13, 16, 19, 22, and 25), HindIII + HpaII (lanes labeled H), or HindIII + MspI (lanes labeled M); separated on a 1% agarose gel; transferred to a membrane; and hybridized with the SAH probe. The molecular weights in kilobases of the fragments detected are indicated at the left and right.

View larger version (68K): In this window In a new window Download PPT slide   Figure 5. Cytosine methylation tests of the 5' ends of r1 genes from neutral haplotypes. (A) Maps of the gene types (Lc-like, Sc-like) found in neutral haplotypes. M, HpaII/MspI; H, HindIII. The position and extent of the SAH probe is represented by black rectangles. The position and extent of the Lc1013 probe is represented by gray rectangles. One of the Lc-like genes in R-r:Ecuador 1172 bears a 200-bp deletion. The extent and position of the deleted region are indicated by parentheses. Sizes of the predicted restriction fragments that result from complete digestion and sizes of the partial digestion products observed on genomic blots are indicated below the map. Three sets of HpaII/MspI sites are indicated: promoter, transcription start, and intronic. (B) Southern blots of neutral haplotypes. The genomic DNA was digested with HindIII alone (lanes 1, 4, 7, and 10), HindIII + HpaII (lanes labeled H), or HindIII + MspI (lanes labeled M); separated on 1% agarose gel; transferred to a membrane; and hybridized with the SAH probe. The molecular weights in kilobases of the fragments detected are indicated at the left and right. (C) Southern blots of neutral haplotypes using the Lc1013 probe. DNA was prepared from a new set of individual plants and was digested with either HindIII + HpaII (lanes labeled H) or HindIII + MspI (lanes labeled M), separated on 1% agarose gel, transferred to a membrane, and hybridized.

Paramutagenic haplotypes generally showed heavy methylation of the HindIII fragments that include the promoters, the first two exons, the first intron, and 1.7 kb of the large second intron (Fig 3B). Haplotypes R-mb, R-g:Bolivia 724, R-g:Bolivia 1520, R-g:Chile 370, R-g:Peru San Miguel, and R-g:Chile 406 were very heavily methylated at all HpaII sites, judging by the presence of a major fraction of undigested 3.7- or 3.5-kb HindIII fragments that represent either Lc- or S2-like genes, respectively. The partial digestion observed with MspI demonstrates the presence of HpaII/MspI sites within these fragments. The larger HindIII fragments from Sc-like genes in haplotypes R-g:Bolivia 1520, R-g:Chile 370, and R-g:Peru San Miguel disappear after digestion with HpaII and MspI, indicating that they contain unmethylated sites. However, because complete sequence information is not available for Sc-like genes and because many Sc-like fragments comigrate with Lc- or S2-like fragments, it is not possible to assess the methylation status of individual HpaII/MspI sites in Sc-like genes.

In R-mb, R-g:Bolivia 724, R-g:Bolivia 1520, R-g:Chile 370, R-g:Peru San Miguel, and R-g:Chile 406, we detected very few or no HpaII-digested fragments using the SAH (Fig 3B) and Lc1013 (not shown) probes, which indicates that r1 genes in these haplotypes are heavily methylated at promoter, transcription start, and intronic sites. R-g:Peru 568 and R-g:Peru 1304-2992 are methylated less heavily. Very little or none of the undigested 3.7- or 3.5-kb HindIII fragment was observed after digestion with HpaII. However, the absence of 0.9-kb bands and the presence of 2.6- or 2.4-kb fragments following HpaII digestion indicate that the promoter and transcription start HpaII/MspI sites in the Lc-like and S2-like genes of these haplotypes are methylated. It is the intronic sites that are unmethylated in these two haplotypes. Thus, paramutagenic haplotypes are methylated at sites in their promoters and near the start of transcription. Downstream sites present in the large intron appear to be methylated in many paramutagenic haplotypes, but are clearly unmethylated in others.

Paramutable haplotypes are markedly less methylated than paramutagenic haplotypes (Fig 4B). There were no intact HindIII fragments following HpaII or MspI digestion of any of the paramutable haplotypes tested, indicating the absence of full methylation in the region of r1 detected by the SAH probe. The 0.9-kb and 1.0-kb HpaII fragments expected after complete digestion are present in all paramutable haplotypes except R-r:std, showing that the intronic and transcription start HpaII/MspI sites are hypomethylated in all cases. It should be noted that there is a discrepancy between the apparent molecular weight of full-length S1/S2 HindIII fragments on agarose gels (5.0 and 4.6 kb) and the actual molecular weights according to sequence (4.7 and 4.3 kb). The fragments consistently migrate larger than expected, perhaps because they consist of a large (2.2 kb) inverted repeat, which may form unusual secondary structures that cause aberrant migration.

Paramutable haplotypes that have Lc-like genes (R-r:std, R-r:India 166163, R-r:India 1210551, and R-r:Oklahoma 213748) are more methylated than are haplotypes without Lc-like genes. The 1.4-kb band observed in these Lc-like-containing haplotypes indicates that one of the transcription start HpaII/MspI sites is methylated in the HindIII fragment containing S1/S2. It is not possible to judge whether the methylated site is at S1 or S2. Because the intensity of the 1.4-kb bands in HpaII and MspI lanes of these haplotypes was very similar, the presence of both CCGG sites was confirmed by digesting S1- and S2-specific PCR products from these haplotypes with HpaII (not shown). The 2.1-kb fragment observed in R-r:std, R-r:India 166163, and R-r:Oklahoma 213748 indicates methylation of intronic HpaII sites. This 2.1-kb fragment could be derived from either Lc-like or S1/S2-like genes. R-r:std was somewhat more methylated than the other paramutable haplotypes, judging by the lack of a 0.9-kb band, increased intensity of the 2.1-kb band, and the presence of a 2.4-kb band. The latter represents an S1/S2 gene fragment in which methylation occurred at both transcription start HpaII/MspI sites.

All but one of the paramutable haplotypes that lack an Lc-like gene (R-g:Argentina 162573, R-g:Arizona 213738, R-g:New Mexico 218148, and R-g:New Mexico 218170) showed complete absence of methylation in the region tested. The exception is R-r:Turkey 167989, which exhibits a 1.1-kb band indicating methylation of one of the HpaII sites within the second intron. However, the transcription start HpaII/MspI sites in each of these haplotypes are unmethylated.

Absence of full-length HindIII fragments following HpaII digestion indicates that all four neutral haplotypes are hypomethylated (Fig 5B) relative to paramutagenic haplotypes. Some methylation was present at the intronic sites, as indicated by the presence of 1.1-kb fragments (Fig 5B, lanes 2, 5, and 11) or absence of the 0.9-kb band (Fig 5B, lane 8). The 1.1-kb bands (Fig 5B, lanes 2, 5, 11) could also result from methylation of both HpaII sites defining the 0.93-kb fragment, which would indicate methylation of at least one of the sites near the transcription start. However, when similar blots were probed with the Lc1013 probe (Fig 5C), no evidence for methylation near the transcription start site was observed. In particular, the lack of 0.8- or 1.8-kb bands in any of the lanes in Fig 5C demonstrates the absence of methylation of the transcription start HpaII/MspI sites in these haplotypes.

Blots were probed with the Lc1013 probe (Fig 5C) to indicate methylation of the promoter HpaII/MspI sites of three of the neutral haplotypes. The presence of the 1.5-kb band in HpaII and MspI lanes of all three neutral haplotypes tested shows methylation of promoter HpaII/MspI sites of Lc-like genes in these haplotypes. The 1.3-kb fragment in R-r:Ecuador 1172 represents promoter methylation of the second Lc-like gene, which has a 215-bp deletion. The weak 1.1-kb band in R-r:Ecuador 1172 and R-g:Peru Corongo ANC150 probably indicates some methylation of the promoter of the Sc-like genes in these haplotypes, as it cannot have come from Lc-like genes. That the Lc1013 probe does overlap the 100-bp HpaII fragment near the transcription start raises the concern that the 1.1-kb fragment arises from more 3' portions of the Lc gene. However, the overlap is small, 25 bp, and thus is not sufficient to allow detection on maize genomic blots under the stringency conditions employed in this study. The 0.7-kb bands in R-r:Ecuador 1172 and R-g:Peru Corongo ANC150 indicate that the promoter sites are unmethylated in a fraction of cells. The fourth neutral haplotype tested, R-g:Bolivia 1004, was unmethylated at all promoter sites, as indicated by the presence of a single, 0.7-kb band following hybridization with the Lc1013 probe (not shown).

Summarizing the methylation data for neutral haplotypes makes it clear that both CCGG sites near the putative transcription start sites in each gene are hypomethylated. The HpaII/MspI sites located upstream in the promoters tend to be hypermethylated in these neutral haplotypes, and the intronic sites are likewise frequently methylated. Thus, while the transcription start HpaII/MspI sites are generally hypermethylated in paramutagenic haplotypes, these same sites are always hypomethylated in neutral haplotypes. The methylation status of these sites near the transcription start is the most consistent molecular correlate for paramutagenicity vs. neutrality that we have identified.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have examined the composition of 26 complex haplotypes and 1 simple haplotype of the r1 locus of maize. Sixteen of the haplotypes examined are paramutable, and all 16 share a characteristic structure, described as an S-subcomplex, that includes the gene fragment q and two S genes that form a large, head-to-head inverted repeat. No neutral or paramutagenic complex has an S-subcomplex. Thus, paramutable r1 complexes are structurally quite similar to each other and are structurally quite distinct from nonparamutable complexes. Two distinct types of S-subcomplex were identified. One of these (the "standard" type) is indistinguishable from that of R-r:std. The other (the Catspaw type) is indistinguishable from the S-subcomplex of R-d:Catspaw. The region of Catspaw haplotypes lacks the non-Dop-derived "rearranged region" that is present in from R-r:std. This rearranged region contains a 162-bp segment bearing 85% sequence identity to the promoter regions of simple Lc-like r1 genes. Models in which this r1 promoter-homologous region of is critical for paramutability are thus ruled out by these findings.

Other conserved structural features, such as the presence of a large S1/S2 inversion or Dop-derived sequences at q or at , may be necessary for paramutability, but such an interpretation must be made with caution. The structural similarity of paramutable haplotypes may merely reflect a recent origin for paramutability, which arose in complexes that have this structure. The behavior of a mutant derivative allele of R-r called r-r:N1-3-1 argues against an important role for q or the large inverted repeat as determinants of paramutability. This derivative has a nearly precise deletion of but leaves q and the S1/S2 inversion intact (WALKER et al. 1995 ). The r-r:N1-3-1 derivative has lost both S gene promotors, which are normally contained within . This means that paramutability of this derivative haplotype cannot be judged by a simple color assay, but has instead been assayed by examining two additional changes normally associated with paramutability (KERMICLE 1996 ; WALKER 1998 ). The first such change is the acquisition of "secondary paramutagenicity," which refers to the fact that paramutant r1 haplotypes themselves become weakly paramutagenic. The second change that is typically associated with paramutability is acquisition of cytosine methylation during paramutation. The r-r-:N1-3-1 derivative is only weakly able to acquire secondary paramutagenicity and does not acquire C-methylation during paramutation. This means that loss of the region alone changes the response of this haplotype to paramutation. The impaired paramutability of the r-r:N1-3-1 derivative may result from lack of transcriptional activity of the genes or could reflect a role for transposon-derived sequences. Either way, paramutability was partially lost in spite of the presence of both q and the large S1/S2 inversion, indicating that these features alone do not explain paramutability.

The genetic compositions of paramutagenic and neutral haplotypes were much more variable. Paramutagenic haplotypes can include various combinations of three r1 gene types: Lc-like, Sc-like and S2-like. A variety of distinct gene combinations were found among paramutagenic and neutral haplotypes. The gene number in paramutagenic haplotypes was also variable, ranging from two to five. This variability is consistent with previous studies in which the gene number of paramutagenic haplotypes was experimentally manipulated through unequal crossing over (EGGLESTON et al. 1995 ; KERMICLE et al. 1995 ; PANAVAS et al. 1999 ). Both these features are most simply accounted for by the well-known tendency of complex r1 haplotypes to undergo unequal crossing over that can lead to expansion and contraction of the repeat array and can also lead to shuffling of gene types into and out of complexes (reviewed in ROBBINS et al. 1989 ). The variability of structure for the paramutagenic and neutral classes of r1 haplotypes suggests that primary structure is not the most critical feature for explaining their behavior in paramutation.

In contrast to the genic compositions, the epigenetic states, as reflected by the pattern of methylation, were fairly uniform within each class of haplotype. Most (five out of seven) of the paramutagenic haplotypes were hypermethylated at all HpaII sites tested, and the remaining two were both methylated at two HpaII sites near the transcription start, as well as at upstream (promoter) sites. This stands in contrast to the situation for paramutable and neutral haplotypes, which were almost all unmethylated near the transcription start. Methylation at promoter and intronic positions showed no such correlation with behavior in paramutation. These results suggest that the epigenetic signals—probably a particular chromatin conformation or composition—that cause paramutagenicity may lie close to the start of transcription.

The new structural data derived for this study also contribute to our understanding of r1 complex evolution (illustrated schematically in Fig 6). Structural analysis of R-r:std suggested that a Dop transposable element inserted into a duplicated copy of an Lc-like gene (WALKER et al. 1995 ). Formation of R-r:std occurred via breakage of this element, to form q and S2, and duplication of the S2 gene segment occurred to produce S1 and form the S1/S2 inverted duplication. It is now clear that the Catspaw-type haplotypes arose by a subsequent deletion within this original S-subcomplex type that removed the 5' portion of S1 and the rearranged portion of . Further evidence for common ancestry of the Catspaw-type paramutable haplotypes stems from the presence of a restriction fragment length polymorphism associated with q that differs between Catspaw and standard types, suggesting that these lineages were distinct prior to the deletion event.

View larger version (18K): In this window In a new window Download PPT slide   Figure 6. Model of the events that formed the known r1 genes. Following an initial duplication event, a Dop transposable element inserted into the distal copy. Two separate lineages were derived from this haplotype: one in which the 5' (proximal) end of Dop was lost that resulted in the S2-like genes in some paramutagenic complexes and a second in which a duplication and inversion occurred that gave rise to the S-subcomplex found in all known paramutable haplotypes.

The S-subcomplex-containing haplotypes represent only one of two modern lineages from the progenitor haplotype containing Dop at r1. The second lineage is represented by the S2-like genes of paramutagenic haplotypes. These genes cannot have been derived from an S-subcomplex, because they contain a far larger fragment of Dop than is found within . Yet, the S2-like genes from paramutagenic haplotypes contain the 3' end of a Dop transposable element inserted at a position identical to that of the S2 genes in paramutable haplotypes, and this Dop end has an identical 5-bp truncation in both. These similarities indicate a common ancestral r1 gene with a defective Dop element inserted. We hypothesize that the S2-like genes now found in paramutagenic haplotypes arose by a deletion of the 5' end of Dop together with adjacent r1 promoter sequences, leaving an r1 gene that is controlled by Dop sequences upstream, but is not part of a duplication/inversion.

	FOOTNOTES

¹ Present address: Department of Plant Pathology, University of Kentucky, Lexington, KY 40546.

	ACKNOWLEDGMENTS

We thank John Shaw and Kathryn Irenze for assistance in the field and lab. We thank Dr. Jerry Kermicle for providing the set of geographic haplotypes used in this study. This work was supported by the National Science Foundation (NSF9796088).

Manuscript received April 13, 2001; Accepted for publication August 9, 2001.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ASSAD, F., K. L. TUCKER, and E. R. SIGNER, 1993  Epigenetic repeat-induced gene silencing (RIGS) in Arabidopsis. Plant Mol. Biol. 22:1067-1085[Medline].

BENDER, J., 1998  Cytosine methylation of repeated sequences in eukaryotes: the role of DNA pairing. Trends Biochem. Sci. 23:252-256[Medline].

BERCURY, S., T. PANAVAS, K. IRENZE, and E. L. WALKER, 2001  Molecular analysis of the Doppia transposable element of maize. Plant Mol. Biol. 47:341-351[Medline].

BRINK, R. A., 1956  Change associated with the R locus in maize is directed and potentially reversible. Genetics 41:872-889[Free Full Text].

BRINK, R. A. and B. MIKULA, 1958  Plant color effects of certain anomalous forms of the R-r allele in maize. Z. Indukt. Abstammungs. Vererbungsl. 89:94-102.

BRINK, R. A. and W. H. WEYERS, 1957  Invariable genetic change in maize plants heterozygous for marbled aleurone. Proc. Natl. Acad. Sci. USA 43:1053-1060[Free Full Text].

BROWN, D. F., 1966  Paramutability of R-g and r-r mutant genes derived from an R-r allele in maize. Genetics 54:899-910[Free Full Text].

CHANDLER, V. L., W. B. EGGLESTON, and J. E. DORWEILER, 2000  Paramutation in maize. Plant Mol. Biol. 43:121-145[Medline].

CONSONNI, G., A. VIOTTI, S. L. DELLAPORTA, and C. TONELLI, 1992  cDNA nucleotide sequence of Sn, a regulatory gene in maize. Nucleic Acids Res. 20:373[Free Full Text].

CONSONNI, G., F. GEUNA, G. GAVAZZI, and C. TONELLI, 1993  Molecular homology among members of the R gene family in maize. Plant J. 3:335-346[Medline].

DELLAPORTA, S. L., 1994 Plant DNA miniprep and microprep: versions 2.1–2.3, pp. 522–525 in The Maize Handbook, edited by M. FREELING and V. WALBOT. Springer-Verlag, New York.

DOONER, H. K., 1971 Recombinational analysis of the R-r duplication in maize. Ph.D. Thesis, University of Wisconsin, Madison, WI.

DOONER, H. K. and J. L. KERMICLE, 1971  Structure of the R-r tandem duplication in maize. Genetics 67:427-436