Mu killer Causes the Heritable Inactivation of the Mutator Family of Transposable Elements in Zea mays

Corresponding author: Damon Lisch, Berkeley, CA 94720., dlisch{at}uclink.berkeley.edu (E-mail)

Communicating editor: D. VOYTAS

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mutations in a number of genes responsible for the maintenance of transposon silencing have been reported. However, the initiation of epigenetic silencing of transposable elements is poorly characterized. Here, we report the identification of a single dominant locus, Mu killer (Muk), that acts to silence MuDR, the autonomous regulatory transposon of the Mutator family of transposable elements in maize. Muk results in the methylation of MuDR TIRs and is competent to silence one or several active MuDR elements. Silencing by Muk is not dependent on the position of the MuDR element and occurs gradually during plant development. Transcript levels of the MuDR transposase, mudrA, decrease substantially when Muk is present. The other transcript encoded by MuDR, mudrB, also fails to accumulate in the poly(A) RNA fraction when MuDR and Muk are combined. Additionally, plants undergoing MuDR silencing produce small, mudrA-homologous 26-nt RNAs, suggesting a role for RNA-directed DNA methylation in MuDR silencing. MuDR elements silenced by Muk remain silenced even in plants that do not inherit Muk, suggesting that Muk is required for the initiation of MuDR silencing but not for its maintenance.

MUCH of the maize genome consists of a vast number of quiescent transposable elements (SANMIGUEL et al. 1996 ). This inactivity is due to a variety of processes that have evolved to keep these potentially potent mutagens transcriptionally and transpositionally silent (KUMAR and BENNETZEN 1999 ). A number of genes have been identified in a variety of organisms that, when mutant, result in the activation of otherwise silenced transposons (for review see OKAMOTO 2001). The nature of these genes suggests that the initiation and maintenance of transposon silencing is a complex process that involves both transcriptional and post-transcriptional gene-silencing pathways. For example, in Caenorhabditis elegans, mutations in the RNaseD mut-7 (KETTING et al. 1999 ) and the Argonaute family rde-1 gene (TABARA et al. 1999 )—both of which are involved in small RNA-based post-transcriptional silencing—activate otherwise silent transposons. Likewise, in Chlamydomonas reinhardtii a mutation in the DEAH-box RNA helicase Mut6 can result in transposon reactivation (WU-SCHARF et al. 2000 ). In these cases the mutations relieve RNA-based post-transcriptional silencing and reverse previously established transposon inactivation. Mutations in other genes, such as the Arabidopsis thaliana SWI2/SNF2 chromatin remodeling factor decrease in DNA methylation 1 (DDM1), reactivate silenced transposable elements not because they are necessary for RNA-based silencing, but because they are involved in the chromatin-based maintenance of the silenced state (HIROCHIKA et al. 2000 ; MIURA et al. 2001 ; SINGER et al. 2001 ). Similarly, a mutation in the WD-repeat-containing gene Mut11 in C. reinhardtii also reactivates silent transposons (JEONG et al. 2002 ). Similarities between Mut11 and fungal transcriptional corepressors suggest a direct role for Mut11 in transcriptional regulation of transposons (ZHANG et al. 2002 ).

Nearly all screens designed to detect genes involved in transposon silencing have sought mutations that affect previously established silenced states (OKAMOTO 2001). Thus, the genes identified to date are specifically those involved in the maintenance (rather than in the establishment) of silencing. Very little is known about the initiation of the silenced state. Further, the transposon reactivation observed in both Arabidopsis and C. reinhardtii refers to populations of transposons without knowledge of the direct effects of the mutations on specific autonomous elements.

Three genes are known to affect the maintenance of the silenced epigenetic state of Mutator (Mu) and Mutator-like element (MULE) transposons. Mutants of DDM1 in A. thaliana result in global and heritable cytosine hypomethylation (KAKUTANI et al. 1999 ). Previously silenced hypermethylated MULEs are transcriptionally reactivated in a mutant ddm1 inbred line (SINGER et al. 2001 ). However, the DDM1 gene does not specifically target MULE elements. DDM1 targets other genomic sequences such as centromeric tandem repeats, retrotransposons, and some single-copy genes (KAKUTANI et al. 1996 ; VIELLE-CALZADA et al. 1999 ; HIROCHIKA et al. 2000 ). Also in Arabidopsis, mutations in the PIWI/PAZ domain protein Argonaute4 (Ago4) abolish RNA-based post-transcriptional silencing and result in transcriptional activation of silenced MULE elements (ZIBERMAN et al. 2003 ). In maize, modifier of paramutation 1 (mop1) mutant plants exhibit hypomethylation of previously silenced Mutator elements (LISCH et al. 2002 ). After several generations of inbreeding in a mop1 mutant background, Mutator elements become transpositionally active (LISCH et al. 2002 ). In addition to its effects on Mutator transposons, the mop1 mutation can prevent paramutation from taking place at three different loci (DORWEILER et al. 2000 ).

The initiation of epigenetic silencing has long been a recognized feature of the highly mutagenic family of Mutator transposable elements in maize (reviewed in LISCH 2002 ). Robertson found that self- or sibling-crossed active Mutator lines could maintain high levels of activity for several generations, but activity was abruptly lost in the fifth generation of inbreeding (ROBERTSON 1983 ). To explain the loss of activity, he postulated the activity of a dominant negative regulatory mechanism, perhaps triggered by an increased copy number of the transposon. Similarly, others have since postulated the presence of a negative regulatory factor to explain enigmatic Mutator silencing results (BENNETZEN 1994 ). In one such report, multiple MuDR elements became inactivated simultaneously (MARTIENSSEN and BARON 1994 ). This inactivation was associated with methylation of cytosine residues within the termini of the MuDR regulatory transposons. The simultaneous inactivation of multiple functional MuDR elements was consistent with the activity of a dominant negative regulator, but it was unclear what factors were important for inducing the silencing process. Analysis was complicated by the multiple MuDR elements segregating at various positions, as well as multiple deletion derivatives of MuDR. Consequently, any number of factors, including "poisoning" by deletion derivatives, MuDR elements at certain positions, and overall copy number may have been contributing to the silencing.

Given the complexity of most Mutator lines, identification of factors involved in Mutator regulation can be problematic. To reduce this complexity we have developed a simplified Mutator system. This minimal Mutator line (CHOMET et al. 1991 ) is a W22-derived inbred line completely permissive for Mutator activity. The minimal line has only one functional MuDR element and a lone nonautonomous Mutator element, a Mu1 insertion in the recessive reporter allele a1-mum2. Spotted kernels are generated by the MuDR-catalyzed excision of the Mu1 element at a1-mum2 late in somatic development, restoring A1 function and pigment production. The single active MuDR element in the minimal line is located on the long arm of chromosome 2L and is designated MuDR(p1). In a cross between a plant with an active hemizygous MuDR(p1) element in the a1-mum2 minimal line and an a1-mum2 minimal line plant with no MuDR, the progeny segregate nonspotted [no MuDR(p1)] and heavily spotted [MuDR(p1)] kernels in a 1:1 ratio.

In previous studies using the minimal Mutator line, three mechanisms for loss of Mutator activity were outlined (LISCH and FREELING 1994 ; LISCH et al. 1995 ). The first occurs when MuDR is absent due to genetic segregation. The second is the result of internal deletions in functional MuDR elements, which can be produced both somatically and germinally. The third involves the loss or reduction of MuDR activity due to position effects. All losses of activity observed to date in the minimal line can be traced to one of these three mechanisms. Spontaneous epigenetic silencing of a MuDR(p1) element has never been observed in the minimal Mutator line over the course of many generations and several hundred test crosses. Only when the minimal line was crossed to a different genetic background has silencing of the MuDR(p1) element in the minimal line been observed (LISCH and FREELING 1994 ).

Here we describe the identification and characterization of a new locus, Mu killer (Muk), which acts in a dominant manner to silence full-length MuDR elements at stable active positions. This silencing is associated with the absence of steady-state poly(A) MuDR transcripts, methylation of Mutator element terminal inverted repeats (TIRs), and the loss of germinal and somatic Mutator activity. Small 26-nt sense and antisense RNAs homologous to the 5' region of the mudrA transcript have also been identified in plants carrying both MuDR and Muk. Genetic analysis reveals that Muk is necessary for the initiation of Mutator silencing, but not for the maintenance of the subsequent silenced state.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Generation of lines:
The a1-mum2 minimal Mutator line and the a1-mum2 minimal line tester: The W22-derived minimal Mutator line was previously generated and described (LISCH et al. 1995 ). It contains one full-length functional MuDR element, as well as none of the other nonautonomous Mutator elements with the exception of one Mu1 element in the allele a1-mum2 (O'REILLY et al. 1985 ; CHOMET et al. 1991 ). When an active MuDR element is present, the Mu1 element excises out of A1 late in somatic development, creating characteristic Mutator spotting in the kernel. When no MuDR activity is present, the Mu1 element interrupts the A1 gene and the kernel has no spotting. A hemizygous MuDR element on the long arm of chromosome 2 named MuDR(p1) (CHOMET et al. 1991 ) is the single active element in the minimal Mutator line. This MuDR(p1) element has never been observed to become spontaneously epigenetically silenced in the minimal Mutator line. When the MuDR(p1) element is not present in the minimal Mutator line, the minimal line is referred to as the minimal Mutator line tester.

Although other MuDR-homologous sequences (hMuDR elements) are present in this (and all) maize backgrounds (CHOMET et al. 1991 ; RUDENKO and WALBOT 2001 ), these sequences do not contribute to Mutator activity in the minimal line (CHOMET et al. 1991 ; LISCH et al. 1995 ; this report).

Mu killer in the minimal Mutator line: We have previously described the epigenetic silencing of a single MuDR element from the minimal Mutator line when crossed to a full-color line (B-I, R-r, Pl-Rh) from a mixed genetic background (LISCH and FREELING 1994 ). The MuDR element in this particular minimal Mutator line individual was a transposed copy of the original MuDR element first cloned from the minimal line [MuDR(p1); LISCH and FREELING 1994]. Because it was not at the original position on the long arm of chromosome 2L, it was possible that the new chromosomal position was responsible for the silencing in this line. To test this, a hemizygous active MuDR(p1) element in the minimal Mutator line was crossed to an individual from the family that was exhibiting silencing. To exclude the possibility that previously silenced MuDR elements were necessary for inactivation, we crossed only silencing-line individuals that lacked full-length MuDR elements. Three of the nine progeny that carried MuDR(p1) had inactive hypermethylated Mu1 TIRs as judged by HinfI-digested Southern blots probed with Mu1. When test crossed, these plants gave rise to all nonspotted or very weakly spotted progeny kernels. In this and all subsequent families, Mu killer (Muk) activity is functionally defined as the presence of a previously active MuDR element that has become inactivated. Inactivation results in the hypermethylation of Mutator element TIR HinfI sites while the MuDR element remains full length and at the same position.

The Muk locus has been introgressed into the minimal Mutator line for three generations. All Muk-carrying individuals in this report are in the minimal Mutator line background without any full-length MuDR elements and with only one Mu1 element at a1-mum2.

Transposon-tagging line: The active Mutator transposon-tagging line used in this report was generated as in CHOMET 1994 . The estimate of eight active MuDR elements in this line is based on analysis of the number of unique EcoRI fragments and the intensity of the SacI internal fragments on Southern blots probed with a mudrA probe, as well as genetic segregation results.

DNA extraction and Southern blotting:
DNA preparation and genomic Southern blotting were performed as previously described (DORWEILER et al. 2000 ). A total of 10 µg of maize genomic DNA was digested for >4 hr with an excess of 20 units of restriction enzyme. Mutator restriction sites used in this report are shown in Fig 1. Mature leaf tissue of leaf 8 was used for DNA extraction unless otherwise noted.

View larger version (13K): In this window In a new window Download PPT slide   Figure 1. Representation of MuDR at position 1 (p1) and Mu1 at a1-mum2. Shaded boxes represent Mutator TIRs while open boxes are Mutator internal sequences that differ between MuDR and Mu1. Restriction sites used in this report are indicated. The mudrA and mudrB transcripts are shown just below the MuDR element as lines, with arrows indicating the direction of transcription. Exons are solid boxes while lines that angle down represent introns. Probes used in this report are indicated below the transcripts. Primers used in this report are shown as arrowheads.

RNA extraction and Northern blotting:
Total and Poly(A) RNA was isolated from the tips of immature ears using the Trizol reagent and manufacturer's directions (Invitrogen, San Diego). A total of 10 µg of RNA was run through a 1.5% agarose gel containing 2.2 M formaldehyde. The RNA was transferred to uncharged Hybond nylon filters (Amersham, Arlington Heights, IL) in 20x SSC and fixed by UV crosslinking (Stratagene, La Jolla, CA) as specified by the manufacturer. Hybridization was performed in 0.5 M sodium phosphate pH 7.2, 7% SDS, 1 mM EDTA, at 65° overnight. The most stringent wash was in 1x SSC, 0.1% SDS at 58° for 1 hr. Blots were stripped of radioactivity by washing with 1 liter of 10 mM Tris-Cl pH 7.4, 0.2% SDS at 75° for 1 hr.

Poly(A) RNA extraction:
Poly(A) RNA was extracted from total RNA samples (see above), using the Oligotex mRNA mini kit (QIAGEN, Valencia, CA). A total of 2 µg of poly(A) RNA was used in Northern analysis as described above.

Reverse transcription-PCR analysis of mudrA:
The same immature ear total RNA samples used for Northern blotting were treated with DNase I (Invitrogen) and then reverse transcribed using Superscript II reverse transcriptase (Invitrogen) and an oligo(dT) primer. Samples were amplified for 29 cycles using the primers 5'AF2, 5' ATCCGGCATTGGGCGAAACA and 5'AR2, 5' TTGTCCGTATCCAAACTTCCCT (see Fig 1 for primer locations) with an annealing temperature of 56°. PCR products were electrophoresed on a 1.5% agarose gel. Amplification of mudrA RNA generates a band of 241 bp, while amplification of the DNA gives a 386-bp band.

Reverse transcription-PCR analysis of mudrB:
The same immature ear total RNA samples used above were treated with DNase I (Invitrogen) and then reverse transcribed using Superscript II reverse transcriptase (Invitrogen). Both the oligo(dT) primer and the mudrB-specific primer B1020r (5' CCCATCACCAAGTTCATCATCA) were used to prime cDNA synthesis (see Fig 1). Samples were amplified for 20 cycles using the primers mudrBRTF, 5' ATCTTGCCACCTTGTACCTCTGGA and mudrBRTR, 5' AGATGCGCGGTATTTGTTGCTGAG (see Fig 1 for primer locations) with an annealing temperature of 59°. PCR products were electrophoresed on a 1.5% agarose gel and blotted to a nylon membrane as described above. The blot was hybridized with the 5' mudrB probe as seen in Fig 1. Amplification of mudrB RNA generates a band of 241 bp, while amplification of the DNA produces a band of 314 bp.

Reverse transcription-PCR analysis of ubiquitin transcripts:
The same oligo(dT)-primed cDNA used in the reverse transcription (RT)-PCR analysis of the mudrA and mudrB transcripts was amplified with primers specific for the ubiquitin transcript (MORENO et al. 1997 ) to ensure equal starting amounts of RNA. Amplification was done for either 29 cycles or 20 cycles followed by blotting to a nylon membrane (described above) and hybridizing (described above) with a ubiquitin probe (MORENO et al. 1997 ).

Assay to determine the presence of MuDR(p1):
The presence of a full-length MuDR element was assayed by Southern blots using DNA digested with SacI and probing with any internal region of MuDR (see below for generation of probes). If a full-length MuDR element is present, a fragment of 4684 bp is visualized. To determine if the MuDR element was at the p1 position, two different methods were used. First, PCR between the MuDR(p1) element and the p1 flanking sequence was used. The MuDR primer p1F, 5' ACCACATTCGATGAGGCCTT and the p1 flanking primer p1R, 5' GGATGTCGGGGGCGCAGAGA (see Fig 1 for primer locations) were used in the following PCR program: 94° for 3 min, 94° for 45 sec, 55° for 45 sec, 72° for 1 min, repeated for 30 cycles; and 72° for 5 min. An amplification product of 837 bp signifies that MuDR(p1) is present. Alternatively, an EcoRI-digested Southern blot was prepared and probed with a p1 flanking probe (see below for generation of probes; Fig 1). If MuDR(p1) is present, a 3.9-kb fragment is produced, while if MuDR(p1) is not present, a 6.3-kb band is produced. To show that the MuDR(p1) was full length, the EcoRI-digested Southern blots were probed with a mudrA-specific probe (see below for generation of probes), which results in a 6.8-kb fragment for full-length MuDR(p1). If both the flanking and the internal probes result in the expected fragment sizes, the MuDR element was presumed to be at the correct position and full length.

Mutator TIR methylation assay:
Mutator activity can be followed by the methylation status of the HinfI restriction site present in all Mu element TIRs (LISCH et al. 1995 ). Mutator-active individuals have hypomethylated Mu TIRs and will produce a 1.3-kb band when digested with HinfI and probed with the internal region of Mu1. Individuals without MuDR or with silenced MuDR elements have hypermethylated Mu1 TIRs that are not digested by the methyl-sensitive HinfI restriction enzyme, producing Mu1 restriction fragments >1.3 kb (LISCH et al. 1995 ). The exact size of the inactive Mu1 restriction fragment is dependent on the position of the hypermethylated Mu1 element. In the allele a1-mum2 the size of this fragment is 2.1 kb. Additional fragments that are the result of hybridization of the Mu1 probe to MRS-A, a maize gene that is homologous to Mu1, can also be observed (CHANDLER et al. 1986 ). The HinfI sites in this gene, which lacks Mu TIRs, are not affected by the presence or absence of MuDR.

The methylation and activity status of MuDR(p1) TIRs can also be assayed by restriction digestion using methyl-sensitive restriction enzymes HinfI and SacI. When using a MuDR TIR probe (see below for generation of probes), digestion of MuDR(p1) with HinfI produces a 311-bp fragment when hypomethylated and a larger fragment of 497 bp when hypermethylated. Any MuDR internal probe can be used to assay the SacI methylation status of MuDR(p1). SacI digestion of an active hypomethylated MuDR(p1) produces a fragment of 4684 bp and a larger fragment when hypermethylated.

Generation of probes:
Mu1 probe: The plasmid that carries the probe for the internal region of Mu1 has been previously described (TALBERT and CHANDLER 1988 ). The Mu1 internal probe is generated by gel isolating an internal AvaI/BstEII fragment.

5' mudrA probe: The 5' mudrA probe was generated by PCR amplification from a MuDR(p1)-containing minimal Mutator line individual. The probe was sequenced to ensure that it was identical to MuDR(p1). The 5' mudrA probe was amplified using the primers 5' ATCGCCAAAACAGAAAGGTGACAG and 5' GCATGGACCAAAGGCACAAAAGAA. The touchdown PCR cycle used was 96° for 15 sec, 95° for 5 min, 95° for 45 sec, 64° - 0.5° per cycle for 30 sec, 72° for 2 min, back to 95° for 45 sec 19 times, 95° for 30 sec, 54° for 30 sec, 72° for 2 min + 1 sec/cycle, back to 95° for 30 sec 29 times, and 72° for 10 min. The PCR product was cloned using the TOPO TA cloning kit (Invitrogen).

3' mudrA probe: The 3' mudrA probe was also generated by PCR from a MuDR(p1)-containing Mutator minimal line individual. Again the probe was sequenced to ensure that it was identical to MuDR(p1). The 3' mudrA probe was amplified using the primers 5' CATGCCCGATAGTGTGATTGAGAT and 5' CTTTTCTTGGGGGTGATTTTCTTC. The same touchdown PCR program as above was used except the first-round annealing temperature was from 66° to 56° and the second-round annealing temperature was 55°. The PCR product was cloned using the TOPO TA cloning kit (Invitrogen).

5' and 3' mudrB probes: The 5' and 3' mudrB probes were digested from a plasmid (pBMP1.3) carrying the entire mudrB gene from MuDR(p1). pBMP1.3 is a BamHI clone that includes the mudrB portion of MuDR(p1) as well as 4 kb of p1 sequence flanking that element (LISCH et al. 1995 ). The 5' mudrB probe was created by digesting pBMP1.3 with SalI and EcoRI. The resulting 503-bp mudrB-specific fragment was then gel purified. The 3' mudrB probe was generated by digesting pBMP1.3 with EcoRI and EcoRII followed by gel purification of the resulting 861-bp fragment.

MuDR TIR probe: The TIR probe was generated by amplification of the pBMP1.3 plasmid. The PCR primers used were 5' GAGATAATTGCCATTATGGA and 5' GATGTCGACCCCTAGAGC. The PCR product was cloned using the TOPO TA cloning kit (Invitrogen).

p1-flanking probe: The p1-flanking probe was generated by PstI digestion of the pBMP1.3 plasmid. The 800-bp p1-specific fragment hybridizes to a single-copy sequence in the maize genome.

All DNA probes in this report were gel isolated and prepared by the random priming method using a Prime-It II kit (Stratagene) and ³²P-radiolabeled dCTP (Perkin-Elmer, Norwalk, CT). All blots were exposed to a Molecular Dynamics (Sunnyvale, CA) phosphor imaging screen, saved as TIFF files, and processed using Adobe Photoshop or Deneba Canvas programs.

Small RNA Northern analysis:
Total RNA from seedling second leaves was extracted using RNAwiz (Ambion, Austin, TX). Total RNA was run on a 15% polyacrylamide gel containing 7 M urea. The gel was electroblotted at 100 V to charged Zeta-Probe blotting membrane (Bio-Rad, Richmond, CA). The hybridization conditions are the same as in HAMILTON and BAULCOMBE 1999 . Single-stranded sense or antisense RNA probes were generated by cloning the probes described above behind the T7, T3, or SP6 promoter. Run-off transcription was performed using a Maxiscript in vitro transcription kit (Ambion), and the hybridization was done overnight at 40°. The Northerns were washed twice for 15 min at 50° in 2x SSC, 0.2% SDS. Sizes of the hybridizing bands were estimated using single-stranded RNA oligos of known length homologous to MuDR. These control RNA oligos also served as a positive hybridization control.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mu killer segregates as a single locus unlinked to MuDR(p1):
In experiments described that employ an active MuDR(p1)-containing plant, the same MuDR(p1) male was also test crossed to the a1-mum2 minimal Mutator line tester to ensure that the MuDR(p1) element did not epigenetically silence in the absence of Muk. In no case was such silencing observed.

To study the pattern of Muk inheritance, we crossed 16 plants that were heterozygous for Muk but lacked full-length MuDR (Muk/-; a1-mum2) to male minimal Mutator line plants hemizygous for MuDR(p1) (MuDR(p1)/-; a1-mum2). The progeny segregated two nonspotted kernels to one weakly spotted kernel to one heavily spotted kernel (see kernel phenotypes, Fig 3A). This ratio is consistent with the 1:1 segregation of a single dominant locus associated with weak spotting (Table 1). For one particular cross (cross 1 in Table 1), a subset of individuals from all kernel phenotypic classes was assayed for the presence of a full-length MuDR(p1) element and for methylation of the HinfI site in Mu1 TIRs (generation 1 in Table 2). Of the progeny analyzed, 16 of 17 plants grown from nonspotted kernels lacked full-length MuDR(p1), 63 of 63 individuals grown from heavily spotted kernels had a full-length MuDR(p1), and 61 of 61 individuals grown from weakly spotted kernels also had a full-length MuDR(p1). Next we compared the Mu1 TIR methylation status in plants grown from the heavily spotted kernels to that of plants grown from weakly spotted kernels (a subset of this data is presented in Fig 3C). A total of 58 out of 63 heavily spotted individuals had hypomethylated Mu1 TIRs, while 55 of 61 weakly spotted individuals had hypermethylated Mu1 TIRs (Table 2). Overall, 60 of 124 MuDR(p1) elements (48.4%) examined in this family were inactive. The 1:1 segregation of silenced to active MuDR(p1) elements demonstrates that Mu killer segregates as a single dominant Mendelian locus.

View larger version (14K): In this window In a new window Download PPT slide   Figure 2. Crossing scheme as described in RESULTS.

View larger version (77K): In this window In a new window Download PPT slide   Figure 3. Mu killer results in weakly spotted kernels and hypermethylated Mu1 TIRs. (A) Nonspotted (left), heavily spotted (center), and weakly spotted (right) kernel phenotypes associated with the segregation of MuDR(p1) and Muk from the cross Muk/-; a1-mum2 x MuDR(p1)/-; a1-mum2. All kernels are homozygous for a1-mum2. (B) HinfI-digested Southern blots probed with Mu1. In the presence of an active MuDR element (heavily spotted kernel), Mu1 is hypomethylated and produces a characteristic 1.3-kb band (arrow). In the absence of MuDR (nonspotted kernel) or with a Muk-silenced MuDR element, Mu1 is hypermethylated and produces a HinfI restriction fragment >1.3 kb. (C) Data from a typical Muk-segregating family generated from the cross Muk/-; a1-mum2 x MuDR(p1)/-; a1-mum2. The heavily spotted kernels have an active MuDR(p1) element and no Muk, while the weakly spotted kernels have MuDR(p1) and Muk.

A total of 28 active MuDR(p1) plants from heavily spotted kernels and 15 silenced MuDR(p1) plants from weakly spotted kernels that were analyzed by Southern blot were then test crossed as female to the a1-mum2 minimal Mutator line tester without MuDR(p1) (see Fig 2). In each instance, the active MuDR(p1) plants generated ears segregating 1:1 heavily spotted and nonspotted kernels (control generation in Table 7), while the inactive MuDR(p1) plants generated 90% nonspotted kernels and 10% weakly spotted kernels (generation 1 in Table 7).

If Muk segregates as a single locus unlinked to MuDR(p1), then half of the progeny that lacked MuDR(p1) from the cross of a female Muk heterozygote with homozygous a1-mum2 (Muk/- ; a1-mum2) to a plant hemizygous for MuDR(p1) with homozygous a1-mum2 (MuDR(p1)/- ; a1-mum2) would be expected to carry Muk. To test this, 25 nonspotted kernels from the above cross were planted. Twenty-three of the resulting plants [none of which carried MuDR(p1)] were crossed as females to active MuDR(p1)/- ; a1-mum2 individuals (Table 3). Thirteen of these crosses resulted in progeny exhibiting a 1:1 ratio of heavily spotted to pale kernels, consistent with the lack of Muk (crosses 1–13 in Table 3). The other 10 families gave a ratio of 2 nonspotted kernels to 1 weakly spotted kernel to 1 heavily spotted kernel, consistent with the presence of Muk in the female parents (crosses 14–23 in Table 3). The segregation of Muk to 43.5% of nonspotted kernels further suggests that Muk is unlinked to MuDR(p1). A subset of the progeny from these crosses was then analyzed for Mu1 TIR methylation (generations 2A and 2B in Table 2). In a family that lacked Muk (segregating 1:1 for heavily to nonspotted kernels), 12 of 12 nonspotted individuals tested lacked MuDR(p1) and had hypermethylated Mu1 TIRs, while 30 of 30 heavily spotted individuals had hypomethylated Mu1 TIRs (generation 2A in Table 2). When the same MuDR(p1)-donating male parent was crossed to the nonspotted kernels that carried Muk/- (generation 2B in Table 2), 20 of 20 nonspotted kernels lacked MuDR(p1), 15/15 heavily spotted kernels had a full-length MuDR(p1) and hypomethylated active Mu1 TIRs, while 12 of 13 weakly spotted individuals tested had a full-length MuDR(p1) and hypermethylated Mu1 TIRs.

We have subsequently followed Muk inheritance for three additional generations from the direct progeny of the initial Muk segregation crosses described here. Muk has continuously segregated as a single Mendelian locus unlinked to MuDR(p1), which can silence MuDR(p1) in a reproducible manner. These data demonstrate that Mu killer is a single locus unlinked to MuDR(p1), which can silence Mutator activity in a dominant fashion.

Mu killer silences Mutator activity when inherited from either the male or the female parent:
When inherited from the female parent, the dominant Muk silences MuDR and causes the weakly spotted a1-mum2 phenotype characteristic of a silencing Mutator system (Fig 3A). However, when Muk is inherited from the male parent in the cross MuDR(p1)/-; a1-mum2 x Muk/-; a1-mum2, heavily spotted and nonspotted kernels segregate in a 1:1 ratio, and no weakly spotted kernels are observed (Table 4). To determine if Muk acts only if inherited from the female parent, we reciprocally crossed Muk/-; a1-mum2 plants with MuDR(p1)/- ; a1-mum2 plants. Progeny of the reciprocal crosses were analyzed by Southern blot and test crossed as females to the a1-mum2 minimal Mutator line tester (Table 5).

Progeny of the cross in which Muk was inherited from the female parent were separated into nonspotted, weakly spotted, and heavily spotted kernel phenotypic classes. From this cross, six of six plants grown from nonspotted and six of six plants grown from weakly spotted kernels had hypermethylated Mu1 TIRs (Table 5). When crossed to the minimal Mutator line tester, all six individuals carrying inactivated MuDR(p1) transmitted only weakly and nonspotted kernels. In contrast, five of five plants grown from heavily spotted kernels showed hypomethylated Mu1 TIRs and yielded at least 50% heavily spotted progeny kernels upon crossing to the a1-mum2 minimal Mutator line tester (Table 5).

Progeny from the reciprocal cross, in which the Muk/- parent was male, were divided into heavily spotted and nonspotted kernel phenotypic classes; there were very few weakly spotted kernels. Of the 19 heavily spotted kernels tested, 11 had hypermethylated Mu1 TIRs. When test crossed, all 11 yielded very few to no heavily spotted kernels and <10% weakly spotted kernels (Table 5). The eight progeny that had hypomethylated Mu1 TIRs produced ears with near 50% heavily spotted kernels when crossed to the a1-mum2 minimal Mutator line tester.

The observation that roughly half (57.9%) of the MuDR(p1) elements in a family in which Muk is inherited from the male are silenced suggests that Muk can silence MuDR when inherited from the male as well as the female parent. The poor correlation of a1-mum2 spotting to Mutator activity when Muk is inherited from the male parent may be due to the dosage of Muk and MuDR(p1) in the triploid aleurone layer of the endosperm. When Muk is inherited from the female, the aleurone layer has the genotype Muk/Muk/- ; -/-/MuDR(p1), while when Muk is inherited from the male parent the aleurone genotype is -/-/Muk ; MuDR(p1)/MuDR(p1)/-. Thus, it is likely that this imbalance in dosage of Muk and MuDR is responsible for the nonreciprocal a1-mum2 kernel-spotting phenotypes.

Muk silencing of Mutator transposons is not dependent on number or position of MuDR elements:
To test if Muk silencing of MuDR is specific to MuDR(p1), we crossed a Mu-tagging line containing an estimated eight active MuDR elements as a male to both an a1-mum2 minimal Mutator line tester and a related plant heterozygous for Muk (Fig 4). Analysis of the transposon-tagging line showed that the MuDR(p1) element was not present in this line. For each cross 22 individual progeny were assayed by Southern blot and test crossed as female to the a1-mum2 minimal Mutator line. From the control cross between the multiple-MuDR line and the a1-mum2 tester, 19 of 22 (86.4%) individuals had active hypomethylated Mu1 TIRs and produced heavily spotted kernels upon crossing to the a1-mum2 minimal Mutator line tester (Table 6). The three individuals that were inactive all had at least one full-length MuDR element, based on Southern blots of DNA digested with SacI and probed with an internal portion of MuDR (data not shown). The 13.6% frequency of spontaneous silencing in this line is typical of a standard multiple-MuDR Mutator line (BENNETZEN 1996 ).

View larger version (84K): In this window In a new window Download PPT slide   Figure 4. Muk silences MuDR elements independent of their position or copy number. An active multiple MuDR transposon-tagging line individual was crossed as a male to both an a1-mum2 minimal Mutator line tester and a Muk heterozygote. Progeny were scored by digesting with HinfI and probing with the internal region of Mu1. The 1.3-kb fragment (arrow) is characteristic of hypomethylated Mu1 TIRs and an active Mutator system. When crossed to the a1-mum2 minimal Mutator line tester, 13.6% of progeny with at least one MuDR element were silenced, while when crossed to Muk/-, 63.6% were silenced.

In the cross between the female Muk heterozygous plant and the same Mutator active individual, only 8 of 22 individuals had hypomethylated Mu1 TIRs and produced heavily spotted kernels when crossed to the a1-mum2 minimal Mutator line tester (Table 6). Fourteen of the 22 individuals had hypermethylated Mu1 TIRs and produced only nonspotted kernels when crossed to the a1-mum2 minimal Mutator line tester (Table 6). This 63.6% frequency of Mutator silencing is above the 50% expected for the segregation of Muk. However, if the 13.6% of spontaneous inactivation found in the cross of the transposon-tagging line to the a1-mum2 minimal Mutator line tester is taken into account, the number of individuals silenced by Muk is approximately half.

The 1:1 segregation of silencing in the progeny of Muk/- crossed to an active transposon-tagging line without MuDR(p1) demonstrates that Muk can silence MuDR elements independent of their copy number or position in the genome.

Muk-induced hypermethylation of Mutator elements occurs gradually:
The Muk-induced weakly spotted kernel phenotype provides an excellent marker for determining which individuals have a silenced MuDR element. However, if the MuDR element was completely silenced by Muk in F₁ kernels, we would expect them to lack excisions altogether. To test how complete MuDR silencing is in individuals grown from weakly spotted kernels, we grew weakly spotted Muk/-; MuDR(p1)/-; a1-mum2 F₁ kernels from the cross Muk/-; a1-mum2 x MuDR(p1)/-; a1-mum2. DNA from 11 seedling L2 leaves, which are initiated early in embryogenesis, were digested with HinfI and probed with the internal region of Mu1. In all cases the Mu1 at a1-mum2 in plants grown from weakly spotted kernels was only partly methylated—the digests produced both a hypomethylated 1.3-kb Mu1 band and a larger hypermethylated Mu1 band (Fig 5). Later in vegetative development, at leaf L6, we again isolated DNA from the same individuals, digested with HinfI, and probed with Mu1. At this later developmental stage, the Mu1 TIRs are completely methylated (Fig 5). The progressive inactivation of Mutator elements observed here is similar to that observed previously in more complex Mutator lines in which increasing portions of the plant tissue carried silenced MuDR elements as the plants developed (MARTIENSSEN and BARON 1994 ).

View larger version (75K): In this window In a new window Download PPT slide   Figure 5. Mu elements in Muk/-; MuDR(p1)/-; a1-mum2 individuals grown from weakly spotted kernels become increasingly methylated through somatic development. Southern blots of DNA digested with HinfI and probed with the internal region of Mu1 show that L2 embryonic leaves are partially methylated, while the L6 leaf of the same individual has completely methylated Mu1 TIRs. When crossed as female to the minimal Mutator line tester, progeny from the same Muk/-; MuDR(p1)/- individual with MuDR(p1) and without Muk are fully methylated (F2). All plants are homozygous for the a1-mum2 allele.

We also investigated the methylation status of MuDR(p1) TIRs in L6 leaves. Using the HinfI restriction sites present in MuDR TIRs and in the flanking p1 sequence, we were able to predict the sizes of hypomethylated as well as hypermethylated restriction fragments. The 497-bp TIR-hybridizing fragment in minimal Mutator line individuals with both MuDR(p1) and Muk demonstrates that the HinfI restriction site in the MuDR(p1) TIR becomes methylated in Muk plants (Fig 6A). SacI restriction sites in the MuDR TIR also become methylated in Muk plants, as seen in Fig 6B. In a line with multiple active MuDR elements present, the SacI sites are not methylated and a 4684-bp band of all of the active MuDR elements is produced. When Muk is present in the same multiple MuDR line, the SacI sites in the MuDR TIR are methylated and do not digest, producing various larger bands with size dependent on MuDR position. Due to the number of MuDR-hybridizing inactive background sequences present in all maize lines, the gradual methylation of the MuDR TIRs could not be assayed.

View larger version (63K): In this window In a new window Download PPT slide   Figure 6. MuDR element TIRs become methylated when Muk is present. (A) MuDR(p1) in the minimal Mutator line has the methylated HinfI restriction product (arrow) only when Muk is present. HinfI sites are present in the the p1 flanking DNA, MuDR TIR, and mudrB portion of MuDR (see Fig 1). The center HinfI restriction site in the TIR is the site that becomes methylated. The methylated band that hybridizes to the TIR probe is 497 bp. (B) MuDR elements from a multiple-MuDR line have methylated TIRs only when Muk is present. SacI sites are found in each of the MuDR TIRs (see Fig 1). When not methylated and probed with an internal region of MuDR, a 4684-bp band (arrow) is produced. When the SacI TIR sites are methylated, larger bands are generated.

MuDR remains inactive multiple generations after silencing by Mu killer:
To determine the stability of the Muk-induced silenced state of MuDR in the absence of Muk, we crossed several Muk/- ; MuDR(p1)/- individuals grown from weakly spotted kernels as female to the a1-mum2 minimal Mutator line tester over several generations (Table 7). The F₁ progeny (generation 1 in Table 7) yielded 9.0% weakly spotted kernels and 91.0% nonspotted kernels on 15 ears. The lack of heavily spotted kernels suggests that MuDR remains relatively inactive even when Muk is segregated away. Molecular analysis showed 46 out of the 46 individuals tested with MuDR(p1) had hypermethylated Mu1 TIRs. Analysis of these 46 silenced individuals also showed no new Mu1 insertions. Further analysis of Muk-silenced MuDR elements over three additional generations of crossing as female to the a1-mum2 minimal Mutator line tester has shown that MuDR(p1) elements do not reactivate when segregated away from Muk (Table 7). Additionally, the decreasing trend in percentage of spotted kernels over the four generations (Table 7) suggests that MuDR may become more deeply silenced through time.

Over the four generations, TIRs from all Mu elements tested remained hypermethylated and no new Mu1 insertions were observed. This analysis included the occasional heavily spotted kernels from Table 7. The heavily spotted kernels in these families were not heritably reactivated. Eleven such kernels were subjected to HinfI digestion, and none of them carried hypomethylated Mu1 elements. Of these, six were also test crossed, and all of them gave rise to mostly nonspotted kernels (8% weakly spotted kernels and no heavily spotted kernels). Thus, we suggest that the occasional heavily spotted kernel represents variation in the efficiency of the maintenance of the silenced state, rather than escape from it. In contrast to lines that carry silenced MuDR(p1) elements, nonsilenced MuDR(p1) elements remain active from generation to generation using both the a1-mum2 reporter (control generation in Table 7) and TIR methylation status (data not shown). These data demonstrate that although the dominant Muk is required to silence MuDR elements, it is not required to maintain MuDR in an epigenetically silenced state. We have also observed that no new Mu1 insertions were generated in progeny of plants carrying silenced MuDR elements, suggesting that Muk also silences germinal Mutator activity.

Mu killer results in decreased mudrA transcript levels:
To test whether the presence of Muk results in the loss of MuDR transcript, total RNA from immature second ears was isolated from F₁ plants derived from the cross Muk/- ; a1-mum2 x MuDR(p1)/- ; a1-mum2. Each plant was genotyped for the presence of MuDR(p1) and Muk, crossed as female to the a1-mum2 minimal Mutator line, and RNA from immature ears was subjected to Northern blot and RT-PCR analysis. mudrA transcript from total RNA was detected only in an active sibling that did not inherit Muk (Fig 7A). Muk/-; MuDR(p1)/-; a1-mum2 individuals grown from weakly spotted kernels exhibited undetectable transcript levels compared to siblings that did not inherit Muk (Fig 7A). No mudrA transcript was detected in plants that lacked MuDR(p1) independent of the presence or absence of Muk. Probes detecting both the 5' (data not shown) and 3' (Fig 7A) ends of the mudrA transcript provided the same results.

View larger version (85K): In this window In a new window Download PPT slide   Figure 7. Expression analysis of the mudrA transcript in a family segregating MuDR(p1) and Muk from the cross Muk/- x MuDR(p1)/-. Both total RNA Northerns (A) and poly(A) RNA Northerns (not shown) provide similar results. In both, the mudrA transcript does not accumulate in individuals without MuDR or with Muk. The mudrA probe used hybridizes to the 3' end of mudrA, as shown in Fig 1. (B) DNase I-treated RNA reverse transcribed with an oligo(dT) primer and amplified using mudrA exon primers for 29 cycles provides similar results as with Northern analysis. The polyadenylated mudrA transcript is not present in individuals with both MuDR(p1) and Muk. RT-PCR controls include a water sample without nucleic acid, a DNA sample that is larger due to the presence of an intron, and a sample of DNase I-treated RNA that was not reverse transcribed (RNA*). mudrA-specific primers used in RT-PCR are shown in Fig 1. In both A and B, + denotes the presence of hemizygous MuDR(p1) or heterozygous Muk, while - denotes the absence of MuDR(p1) or Muk.

Poly(A) transcript was extracted from the same total RNA samples. Northern analysis of the poly(A) RNA provided similar results as obtained using total RNA (data not shown). RT-PCR analysis of total RNA reverse transcribed using an oligo(dT) primer and amplified for 29 cycles provided identical results (Fig 7B). Only an active sibling from the family segregating Muk that did not inherit Muk provided detectable levels of polyadenylated mudrA. Similar results have been obtained with 12 different MuDR(p1)/- active sibling individuals, 22 Muk/- ; MuDR(p1)/- ; a1-mum2 individuals grown from weakly spotted kernels, and 10 individuals without MuDR(p1).

Mu killer results in decreased mudrB poly(A) RNA levels:
The same total RNA samples used in the mudrA expression analysis were used for expression analysis of the mudrB transcript. Total RNA and poly(A) RNA Northern blots were probed with the 5' and 3' regions of the mudrB gene. RT-PCR amplified for 20 cycles was also performed on cDNA primed with either a mudrB-specific primer (B1020r, see Fig 1) or an oligo(dT) primer. RT-PCR products were blotted to nylon and probed with the 5' mudrB probe (see Fig 1).

Surprisingly, mudrB expression was still observed on Northerns using total RNA from Muk/- ; MuDR(p1)/- heterozygotes grown from weakly spotted kernels when hybridized with either the 5' (data not shown) or the 3' mudrB probe (Fig 8A). The mudrA transcript was absent in these same individuals. mudrB transcript of the correct size was present in all 21 Muk/-; MuDR(p1)/-; a1-mum2 F₁ individuals tested. A total of 11 MuDR(p1)/- active siblings with no Muk also had mudrB transcript, while 10 siblings without MuDR were tested. In contrast, Northern analysis did not detect poly(A) mudrB transcript from the 21 Muk/-; MuDR(p1)/-; a1-mum2 samples examined (Fig 8B). To verify these findings, RT-PCR was performed on the same RNA samples (Fig 8C). As with the Northern analysis, RT-PCR resulted in amplification of the mudrB transcript only in Muk/-; MuDR(p1)/-; a1-mum2 samples when the cDNA was primed with a mudrB-specific primer and not when primed with an oligo(dT) primer. These data suggest that in Muk/-; MuDR(p1)/- heterozygotes, mudrB is still transcribed but not correctly processed into mature mRNA.

View larger version (55K): In this window In a new window Download PPT slide   Figure 8. Expression analysis of the mudrB transcript in a family segregating MuDR(p1) and Muk from the cross Muk/- x MuDR(p1)/- (A–C), as well as one generation after initial silencing by Muk (D). Unlike mudrA expression, total RNA Northerns (A) show mudrB still accumulates in Muk/-; MuDR(p1)/- F1 individuals. Poly(A) RNA Northerns (B) do not show mudrB accumulation in the same Muk/-; MuDR(p1)/- samples. The 3' mudrB probe (see Fig 1) was used in A and B, while the 5' mudrB probe produced similar results (data not shown). (C) DNase I-treated RNA was reverse transcribed with either a mudrB-specific primer (B1020r) or an oligo(dT) primer and amplified using mudrB exon primers for 20 cycles, then blotted, and probed with the 5' mudrB probe (see Fig 1). RT-PCR analysis provides similar results as with Northern analysis. The polyadenylated mudrB transcript is not present in individuals with both Muk and MuDR(p1); however, nonpolyadenylated transcript does accumulate. Although mudrB is still transcribed in F1-silencing plants, we assume no functional protein is produced because of the lack of poly(A) mudrB transcript. RT-PCR controls include a water sample without nucleic acid, a DNA sample that is larger due to the presence of an intron, and a sample of DNase I-treated RNA that was not reverse transcribed (RNA*). mudrB-specific primers used in RT-PCR are shown in Fig 1. To determine if mudrB continues to accumulate in the total RNA fraction one generation after initial silencing by Muk, we crossed Muk/-; MuDR(p1)/- individuals from A–C to the a1-mum2 minimal Mutator line and examined progeny by total RNA Northern blot (D). One generation after initial silencing (F2), nonpolyadenylated mudrB transcript does not accumulate. In A–C, + denotes the presence of hemizygous MuDR(p1) or heterozygous Muk, while - denotes the absence of MuDR(p1) or Muk.

To determine if mudrB continues to be expressed in the total RNA fraction in the next generation, a Muk/-; MuDR(p1)/- heterozygote was crossed as female to the a1-mum2 minimal Mutator line tester. Muk in the progeny was scored by crossing the plants as female to an active MuDR(p1)/- ; a1-mum2 individual and assaying for the presence of Muk-induced silencing of the active MuDR(p1) element. Progeny with MuDR(p1) and no Muk one generation after initial silencing by Muk show no mudrB transcript in the total RNA fraction (Fig 8D). A total of 10 individuals with MuDR(p1) and without Muk were tested and mudrB transcript was undetectable in all of them.

Together, these data suggest that although mudrA total RNA transcript levels correlate with Mu TIR methylation and Mutator activity in weakly spotted Muk/-; MuDR(p1)/- F₁ heterozygotes, mudrB is silenced either by an alternative mechanism or at a different time than mudrA.

Small 26-nt RNAs are found in plants with MuDR and Muk:
To test if RNA-based post-transcriptional gene silencing of MuDR was occurring in Muk plants, small RNA Northern blots were used. RNA from the second leaf of seedling plants in a family segregating for Muk and MuDR were examined for the presence of small RNA molecules. These plants were generated from the cross Muk/-; a1-mum2 x MuDR(p1)/-; a1-mum2. A species of small RNA of 26 nt is present in only F₁ individuals with both MuDR(p1) and Muk (Fig 9). This small RNA species hybridizes with both sense- and antisense-transcribed RNA probes complementary to the 5' region of the mudrA transcript. This result has been tested on a total of 14 MuDR(p1)/-; Muk/-; a1-mum2 individuals; 10 MuDR(p1)/-; a1-mum2 siblings without Muk; and 10 each of control MuDR(p1)/-; a1-mum2 individuals, control Muk/- ; a1-mum2 individuals, and control a1-mum2 individuals without MuDR(p1) or Muk. The only individuals that show the small RNA band of 26 nt are the plants with both MuDR(p1) and Muk. Small RNAs homologous to the rest of MuDR were not found (data not shown).

View larger version (69K): In this window In a new window Download PPT slide   Figure 9. Northern blot analysis of total leaf 2 RNA probed with the antisense strand of the 5' region of mudrA. Small 26-nt RNAs are present only in plants with both MuDR(p1) and Muk. The 26-nt RNA band (arrow) hybridizes to both sense (data not shown) and antisense probes. Larger hybridizing bands are products of hMuDR elements and do not contribute to Mutator activity. Two different Northerns with different individuals are shown in A and B. The size range of products shown in A is 18–45 nt. The size of the band specific to Muk/-; MuDR(p1)/- individuals was estimated using known single-stranded RNA molecules of 20 and 25 nt that hybridize to the 3' end of mudrA. The size standard also served as a positive hybridization control. Northerns showing the 26-nt RNAs were stripped of radioactivity and reprobed with the antisense transcript of the 3' region of mudrA to hybridize the size standards. +, presence of hemizygous MuDR(p1) or heterozygous Muk; -, absence of MuDR(p1) or Muk.

Directed attempts to identify Mu killer:
Deletion derivatives of transposons in Drosophila and maize have been implicated in repressing the activity of their cognate full-length elements (CUYPERS et al. 1988 ; LEE et al. 1998 ). Antisense MuDR RNA has been detected in both minimal and complex Mutator lines, which in at least one case is due to read-through of a MuDR deletion derivative (LISCH et al. 1999 ). This antisense RNA could conceivably trigger RNA-mediated Mutator inactivation. In addition to deletion derivatives present in most Mutator lines, all maize lines contain multiple inactive MuDR-homologous sequences (hMuDRs; reviewed in WALBOT and RUDENKO 2002 ). It is possible that one of these inactive background elements expresses an aberrant transcript that can cause Mutator silencing.

To explore these possibilities we attempted to locate a MuDR-homologous sequence cosegregating with Mutator silencing by Southern blot. Ten methylation-insensitive restriction enzymes and five probes that cover the entire MuDR element (including the TIRs) have been used without detecting cosegregation between Muk and any MuDR-related sequence (data not shown). Special care was taken on these Southern blots to ensure that no small hybridizing fragment that cosegregates with Muk was missed. Further, total RNA and poly(A) Northern analysis of >35 individuals also suggests that no detectable aberrant MuDR homologous transcript is associated with Muk. Although it remains a formal possibility that Mu killer is a MuDR-homologous element, to have escaped detection, this element would have to be significantly diverged from functional MuDR elements.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Transposable elements are present in most eukaryotic genomes in multiple copies. Generally, only a subset of the elements present is competent to catalyze their own transposition, and it is likely that even these autonomous elements vary in competence depending on their regional chromatin context. In cases where transposons have been reactivated as a result of mutations in genes responsible for silencing, the transposons examined have been treated as a relatively uniform population (reviewed in OKAMOTO 2001). However, it is likely that only a subset of any given family of transposons is reactivated. In the case of ddm-1 reactivation, for instance, it is clear that only some MULE elements (those located in heterochromatin) were reactivated, suggesting that the position of the elements plays a role in the nature of their silenced state (MIURA et al. 2001 ; SINGER et al. 2001 ). Further, because the autonomous element in these systems has not been identified, it has not been possible to determine which specific autonomous transposon in a given genome has actually been reactivated.

In addition to the heterogeneity of transposon populations, due to the nature of the screens used, only those genes necessary for continued maintenance of transposon silencing have been identified. Little is known about those factors that can initiate de novo silencing. Although many of the mechanisms involved in maintenance of silencing are almost certainly involved in its initiation, additional factors are involved. For instance, although it is clear from work in a variety of systems (particularly those involving transgenes) that double-stranded RNA can trigger silencing, it is likely that the double-stranded RNA trigger is not sufficient to initiate silencing in all cases (TIJSTERMAN et al. 2002 ). This is almost certainly true in Mutator silencing. Since mudrA and mudrB are transcribed convergently from opposite strands, some read-through transcription can and does occur in Mutator-active plants that do not exhibit transposon silencing (HERSHBERGER et al. 1995 ; RUDENKO and WALBOT 2001 ).

Ideally, analysis of initiation of transposon silencing should utilize a single transposon at a known chromosomal position that can be reproducibly and reversibly inactivated. In this respect, the minimal Mutator line presents a unique opportunity to examine the process by which transposons become inactivated. Because this line contains a single active autonomous transposon at a known position, it is possible to examine changes in chromatin configuration, transcription, and transpositional activity simultaneously during the process of silencing.

We have demonstrated that the dominant Muk locus is competent to silence multiple MuDR elements independent of their position. Silencing by Muk is initiated regardless of the gender of the parent from which Muk is inherited. However, the weakly spotted kernel phenotype associated with Muk-induced silencing of the Mutator system is apparent only when Muk is inherited from the maternal parent, presumably due to dosage effects in the triploid endosperm. Whether Muk acts differently in the embryo (and not the endosperm) when inherited from the male or female parent remains to be investigated.

The Muk silencing of active MuDR(p1) elements is not dependent on the presence of a previously silenced MuDR element. The silencing appears to be progressive during plant development and is complete by the production of the mature sixth leaf. As with Mu1 TIRs, the TIRs of MuDR become methylated when silenced by Muk. Importantly, the stable inactive state of a Muk-silenced MuDR(p1) element can be propagated for multiple generations in the absence of Muk, suggesting that Muk is not required for the maintenance of the silenced state. Finally, the decreasing proportions of weakly spotted kernels in subsequent generations suggest that a Muk-silenced MuDR element that has segregated away from Muk may become gradually more inactive over several generations.

MuDR silencing by Muk is associated with the loss of polyadenylated mudrA and mudrB transcript as well as the transient presence of nonpolyadenylated mudrB transcript. The observed differences in the total RNA mudrA and mudrB transcript levels suggest differential regulation of these two genes. The mudrA gene is the putative transposase, and analysis of deletion derivatives has revealed that the loss of mudrA gene product is sufficient to result in Mu element methylation (LISCH et al. 1999 ). Thus, it is tempting to suggest that Muk acts directly on mudrA and that mudrB is then lost because it requires mudrA for continued expression. However, deletions that remove mudrA do not result in the loss of mudrB transcript or protein (LISCH et al. 1999 ), suggesting that the loss of mudrB transcript is not simply due to the loss of mudrA. The loss of polyadenylated mudrB transcript, followed in the next generation by loss of the remaining nonpolyadenylated mudrB transcript, suggests that Muk also affects mudrB, possibly later or by an alternate mechanism than it affects mudrA. Previous studies on the spontaneous inactivation of MuDR have found that the subcellular location of mudrB is altered in Mutator-active vs. -inactive plants (RUDENKO et al. 2003 ). Rudenko and co-workers found a higher proportion of nuclear-retained mudrB transcript in Mutator-silencing plants than in Mutator-active plants. In general, the majority of polyadenylated transcript is present in the cytoplasm, while nonpolyadenylated transcript is located in the nucleus (HUANG and CARMICHAEL 1996 ). Thus, our finding that nearly all of the mudrB transcript in Muk/-; MuDR(p1)/- F₁ plants is nonpolyadenylated is consistent with previous results in Mutator-silencing plants (RUDENKO et al. 2003 ).

By the second generation after silencing by Muk, nonpolyadenylated mudrB is no longer present. Thus, it appears that the continued expression of nonpolyadenylated mudrB is associated with the initiation, but not the maintenance of silencing. It is not clear whether the presence of nonpolyadenylated and potentially nuclear-localized mudrB is a cause or an effect of silencing. Previous workers have suggested a role for increased retention of nuclear-localized transcript in the process of silencing (RUDENKO et al. 2003 ). In those experiments, although the percentage of nuclear mudrB transcript increased, this was largely due to the loss of polyadenylated transcript; the total amount of mudrB transcript in the nucleus remained relatively constant. Similarly, we observe a dramatic change in the proportion of polyadenylated to nonpolyadenylated mudrB transcript. However, that change is due primarily to the loss of polyadenylated mudrB, not an increase in nonpolyadenylated mudrB. Thus, in each of these experiments, increased nuclear retention per se is unlikely to be the cause of silencing, since it was not a variable associated with the process of silencing. One scenario to explain the available data is that silencing triggers the loss of cytoplasmic polyadenylated RNA (first from mudrA and then from mudrB) and those changes in turn lead to transcriptional inactivation of first mudrA and then mudrB. By the next generation, transcriptional repression of both mudrA and mudrB has been achieved. Nuclear run-on experiments will be used to address these issues directly.

Most intriguingly, at least the initial stages of MuDR silencing are associated with the production of small, 26-nt RNA molecules that are homologous to both strands of the 5' end of the mudrA gene. Since the 26-nt RNA is found only when probed with the 5' region of mudrA, this further suggests that Muk affects mudrA before or differently than mudrB. Approximately 26-nt RNAs are part of a larger family of newly identified small RNAs that may have properties different from those of the 21-nt siRNAs known to be associated with RNA degradation (HAMILTON et al. 2002 ). HAMILTON et al. 2002 found that in plants the smaller 21-nt siRNAs are involved in the degradation of the target mRNA, but the presence of longer 26-nt RNA correlates with systemic silencing and methylation of homologous DNA. Since Mu TIR methylation appears to be initiated early in development in tissues where 26-nt RNAs are observed, we suggest that these longer small RNAs may be the trigger for MuDR element methylation.

Despite efforts to locate a MuDR-homologous sequence associated with Muk, none has been found. Also, regardless of its effects on Mu methylation, Muk is not linked to mop1 or either of the two maize DDM1 homologs, CHR101 or CHR106 (data not shown). Further, Muk does not affect the methylation status of the maize centromeric or ribosomal RNA repeats (data not shown), suggesting that Muk is not a global chromatin-remodeling gene such as DDM1.

Future experiments will focus on Mu killer's effects on transcriptional activity of mudrA and mudrB, small RNA production, and DNA methylation at various stages of development. Additionally, we will investigate chromatin compaction, histone methylation, and acetylation of MuDR(p1) when it is being silenced by Muk. We are particularly interested in embryonic tissues in which the process of silencing may be initiating. Because we can separate initiation from maintenance of MuDR silencing, it will also be interesting to examine each of these variables in plants that carry a silenced MuDR element but that lack Muk. The availability of Muk and the single-copy minimal Mutator line should make it possible to carefully dissect a number of aspects of transposon silencing.

	ACKNOWLEDGMENTS

The authors thank Randall Tyers and Nick Kaplinsky for their critical review of this manuscript. The authors also thank Dr. David Braun for supplying the high-copy Mutator tagging line and for his continued feedback. This work was funded by the University of California Berkeley-Syngenta Strategic Alliance.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

BENNETZEN, J. L., 1994  Inactivation and reactivati