SLTAB2 is the paramutated SULFUREA locus in tomato

Highlight The chlorotic phenotype of the paramutated sulfurea tomato plants is due to the epigenetic silencing of the SLTAB2 gene, involved in the translation of photosystem I.


Introduction
Paramutation involves the transfer of epigenetic marks from a (paramutagenic) silent gene to the active (paramutable) allele so that it becomes heritably silent and paramutagenic. Several plant species exhibit paramutation and the best characterized examples, the b1 and pl1 loci in maize, have been linked to the process of RNA-directed DNA methylation (RdDM) (Chandler and Stam, 2004;Hollick, 2012) in which paramutagenic small interfering (si)RNAs mediate silencing of the paramutable allele. This simple model does not, however, explain why most siRNA loci are not paramutagenic: there must be other factors.
To shed light on the mechanism of paramutation we are analysing the tomato SULFUREA (SULF) locus. The silent sulf allele has a chlorotic phenotype (Hagemann, 1958) that is associated with reduced auxin (Ehlert et al., 2008). A sulf homozygote is seedling lethal but a viable heterozygous sulf/+ plant has large chlorotic sectors that are due to paramutation of the active allele to a silenced state in early development. This system is like classic maize paramutation because the paramutated state is heritable and paramutagenic (Hagemann, 1969). SULF maps to the pericentromeric heterochromatin of chromosome 2, at approximately 29 cM from the S locus (Solyc02g077390, compound inflorescence) (Hagemann and Snoad, 1971) but the affected gene could not be mapped precisely due to low recombination frequency in this region (The Tomato Genome Consortium, 2012).
A gene orthologue of the Arabidopsis ATAB2 is strongly down-regulated in chlorotic sectors of sulf/+ tomato (Ehlert et al., 2008) but it was previously excluded as the SULF gene because it is still expressed at detectable levels. However, from analysis of transcriptome, methylome, and small RNA populations of wild-type and paramutated tomato leaves we show here that paramutation of SLTAB2 is responsible for the sulf chlorosis and the decrease in auxin levels. SLTAB2 silencing was associated with changes in DNA methylation and siRNA levels at its promoter, a signature of RdDM.
Additional evidence supporting the identification of SLTAB2 as SULF is from virus-induced gene silencing (VIGS) of the SLTAB2 promoter resulting in methylation of the target DNA sequence, silencing of its expression, and a phenocopy of the sulf chlorosis. Together, these results support a causal role of siRNAs and RdDM in paramutation but, unlike the maize examples, the SLTAB2/SULF locus lacked repeated sequences. Mapping of SULF to SLTAB2 and further comparison with maize will help build a general model of paramutation in plants.

Plant material and growth conditions
Atab2 T-DNA knockouts (GABI-KAT line 354B01) and wildtype Col-0 seedlings were sown on 1/2 strength Murashige-Skoog medium, 1× Nitsch&Nitsch vitamins, 0.8% agar, 1.5% sucrose, pH 6; stratified for 72 h at 4 °C in the dark and transferred to shortday conditions (8 h light at 23 °C and 50 µmol photons m -2 s −1 , 16 h dark at 21 °C). Whole seedlings were collected after 7 d of growth. Tomato plants were raised from seeds in compost (Levington M3) and maintained in a growth room at 23 °C with 16/8 h light/dark periods with 60% relative humidity, at a light intensity of 150 µmol photons m -2 s −1 . Young leaves were collected from 1-month-old plants. Sulf and sulf/+ tissue was collected from sulf/+ plants that had both fully yellow (sulf) and fully green (sulf/+) sectors.

Transcriptome analysis
Total RNA samples were prepared from 100 mg of leaf tissue using TRIzol (LifeTechnologies). For qPCR, 5 µg of total RNA was first DNase treated using Turbo DNase (Ambion), following the manufacturer's guidelines. cDNA was then synthesized using random hexamers and Oligo(dT) and SuperScript III (LifeTechnologies), according to the protocol. qPCR was performed on a Roche LC480 with SYBR in technical triplicates. mRNA abundance was normalized by the geometric mean of two housekeeping genes TIP41 and EXPRESSED (Coker and Davies, 2003). Genotyping of amplified cDNA was performed by digesting 100 ng purified SLTAB2 amplicon with BaeI (NEB) for 12 h at 25 °C in 1× NEB2.1, 100 µg ml −1 BSA, and 20 µM SAM as per the manufacturer's instructions, and electrophoresis on a 1.5% agarose gel. Strand-specific RNA-Seq libraries for two wild-types and three pairs of sulf and sulf/+ were made and indexed with the ScriptSeq v2 kit (Epicentre) according to the protocol after RiboZero treatment (Plant leaf, Epicentre), and sequenced as a pool on one lane of HiSeq 2000 100PE. Sequences were trimmed and filtered with Trim Galore! with default parameters and 11-29 million reads per library were concordantly aligned on Heinz genome SL2.50 and ITAG2.4 gene models using TopHat2 v2.0.13 (Kim et al., 2013) (with parameters -r 200 --mate-std-dev 100 -N 3 --read-edit-dist 3 --library-type fr-firststrand --solexa1.3quals, and version 2.2.4.0 of Bowtie2; alignment rate 66-73%; see Supplementary Table S1 at JXB online). Differential expression analysis was performed on raw counts on annotated mRNAs (ITAG2.4) with DESeq2 v1.8.1 (Love et al., 2014). Genes were considered differentially expressed when the adjusted P-value was <0.05. Hierarchical correlation clustering of the genes differentially expressed between wt and sulf was performed in SeqMonk (v0.32.0). Gene Ontology analysis was performed with the goseq package (v1.20.0, Young et al., 2010) using previously published gene ontology annotation (Koenig et al., 2013), normalizing with mRNA length and running with the following parameters: method=Wallenius, repcnt=2 000, use_genes_without_cat=F. Categories were considered to be over-represented if the associated P-value was <0.05 after Benjamini-Hochberg correction.
sRNA-Seq sRNAs were cloned from 10 µg total RNA (from the same tissue as used for RNA-Seq, for two wild-types and two pairs of sulf and sulf/+) using the Illumina TruSeq Small RNA cloning kit and libraries were indexed during the PCR step (12 cycles) according to the manufacturer's protocol. Gel size-selected, pooled libraries were sequenced on a HiSeq 2000 50SE. Sequences were trimmed and filtered with Trim Galore! (with the adapter parameter -a TGGAATTCTCGGGTGCCAAGG) and 14-20 million reads per library were mapped without mismatches and clustered on Heinz genome SL2.50 using the ShortStack software v2.1.0 (Axtell, 2013;Supplementary Table S2). sRNA counts on the defined loci were analysed with DESeq2 v1.8.1. Uniquely mapping reads on DMR1 and DMR2 were normalized with edgeR's implementation of TMM size factors, on all sRNAs present in all libraries and with at least 10 total counts (Robinson et al., 2009), and a Poisson regression was applied to the normalized counts (generalized linear model in R, with the genotype variable taking values wt, sulf/+, and sulf).
Methylome analysis DNA was extracted from 100 mg of leaf tissue (from the same sampling as for RNA-Seq and sRNA-Seq, for two wild-types and two sulf) using the Puregene kit (QIAGEN). Bisulfite library preparation was performed with a custom protocol similar to Urich et al. (2015). 1.2 µg DNA was sonicated on a Covaris E220 to a target size of 400 bp and purified on XP beads (Ampure, ratio 1.8). DNA was end-repaired and A-tailed using T4 DNA polymerase and Klenow Fragment (NEB) and purified again using XP beads (ratio 1.8×). Methylated Illumina Y-shaped adapters for paired-end sequencing were ligated using Quick-Stick Ligase (Bioline). 450 ng of purified (ratio 1.8×), adapter-ligated DNA was bisulfite-converted using the EZ DNA Methylation-Gold Kit (Zymo Research) according to the manufacturer's instructions. DNA was barcoded using 12 cycles of PCR amplification with KAPA HiFi HotStart Uracil+Ready Mix (Kapa Biosystems) with PE1.0 and custom index primers (courtesy of the Sanger Institute). Pooled libraries were sequenced to a depth of about 5× on a HiSeq 2500 125PE. Sequences were trimmed and filtered with Trim Galore! (default parameters), then mapped on Heinz genome SL2.50 using Bismark v0.14.3 (Krueger and Andrews, 2011) (first in paired-end mode with options --scoremin L,0,-0.2 -p 4 --reorder --ignore-quals --no-mixed --no-discordant -X 1500 --unmapped --ambiguous, then unmapped read1 was mapped in single-end mode with the same quality parameter -N 1, Supplementary Table S3). Reads were deduplicated with bismark_deduplicate and methylation calls were extracted using Bismark methylation_extractor (with options -r2 2 for paired end reads). Methylated and unmethylated counts for cytosines of both strands were pooled into contiguous 200 bp bins and separated by context (CG, CHG, and CHH) with a custom python script. Bins with fewer than 10 counts were excluded from the analysis. Bins are considered differentially methylated if the maximum P-value of the two chi-square tests (wt1 versus sulf 1, wt2 versus sulf 2) is <0.01. Analysis of methylation by McrBC was performed as previously described by Bond and Baulcombe (2015). For Sanger bisulfite sequencing, 450 ng of DNA was bisulfite-converted, amplified with primers specific to the region of interest, A-tailed, and cloned into pGEM-T easy (Promega) following protocols similar to the library preparation. Sequences aligned with MUSCLE were then analysed with CyMATE (Hetzl et al., 2007).

VIGS
DMR1a (606 bp) and DMR1b (562 bp) genomic inserts were cloned into the binary TRV RNA2 vector using the KpnI and XhoI restriction sites of the multiple cloning site as described previously (Liu et al., 2002;Bond and Baulcombe, 2015). Cotyledons of tomato seedlings were agro-infiltrated 10 d after sowing with a 1:1 mixture of Agrobacterium tumefaciens (strain GV3101:pMP90+pSOUP) carrying TRV RNA1 and RNA2 at OD 600 =1.5. Symptoms of SLTAB2 silencing were visible from 2 weeks post-infection.

Auxin quantification
Endogenous levels of free IAA were detected by LC-MS/MS method as described in Novák et al. (2012). Briefly, 10-20 mg fresh tissue of the control and mutant lines were collected, extracted in icecold 50 mM sodium phosphate buffer (pH 7) and purified by SPE on hydrophilic-lipophilic balance reversed-phase sorbent columns (Oasis HLB, 1 cc/30 mg, Waters). To each extract, 5 pmol of 13 C 6 -IAA were added as internal standards to validate the quantification. Purified samples were analysed by the LC-MS/MS system consisting of an ACQUITY UPLC System (Waters, Milford, MA, USA) and Xevo TQ-S (Waters) triple quadrupole mass spectrometer. Quantification was obtained using a multiple reaction monitoring (MRM) mode of selected precursor ions and the appropriate product ion.

Oligonucleotides
Please refer to Supplementary Table S4.

Accession codes
All sequencing data have been deposited in the Sequence Read Archive under the BioProject SRP066362.

Pericentromeric SLTAB2 is strongly down-regulated in sulfurea
The tomato lines in this study had either unsilenced SULF loci (wild type) or they were the progeny of a cross between the wild type and a plant with sectors with silent sulf. Some of the F1 plants were wild type and fully green or, like the sectored parent, they had green and chlorotic sectors consistent with a heterozygous sulf/+ epigenotype with paramutation. The chlorotic sectors would have had the homozygous sulf/sulf epigenotype (referred to as sulf) and the non-paramutated green sectors would be sulf/+ (Fig. 1A).
Based on the understanding of paramutation in maize, the SULF locus would be suppressed in sulf (paramutated yellow sectors), partially silent in sulf/+ (non-paramutated green sectors) and fully expressed in the wild type (wt). It would also be located in the pericentromeric heterochromatin of chromosome 2 upstream of the euchromatic S locus (Hagemann and Snoad, 1971). To find loci with these characteristics, we analysed transcripts of wild-type, non-paramutated sulf/+, and paramutated sulf leaves using mRNA-seq. We identified 2 237 differentially expressed genes between sulf and wt (P <0.05) that clustered into four main categories with distinct Gene Ontology enrichments ( Fig. 1B; Table 1). Consistent with a decrease in photosystem I and the quantity of pigment in sulf leaves (Ehlert et al., 2008), many photosynthesis-related genes were down-regulated (Table 1, class II). The down-regulation of photosystem I was also detectable in the non-paramutated heterozygous sulf/+ leaves (Table 1, class I).
Genes associated with various stress responses were up-regulated in sulf (Table 1, class IV) and were probably a secondary consequence of the sulf phenotype.
Of these differentially expressed genes, 36 were both down-regulated in sulf and located upstream of the S locus on chromosome 2. Among these candidates for SULF, Solyc02g005200 particularly stood out as being the most repressed in sulf (15-fold reduction) and at the predicted map position of SULF (29 cM from S locus, when the centromere-S distance is 30 cM). The other candidates mapped to the euchromatin or the transition zone between heterochromatin and euchromatin ( Fig. 1C; Supplementary Table S5). Further qPCR analysis confirmed the strong down-regulation of Solyc02g005200 in sulf (26-fold) and revealed variable levels in sulf/+, compatible with mono-allelic expression (Fig. 1D). Solyc02g005200 is the orthologue of Arabidopsis thaliana ATAB2 that is likely involved in the translation of mRNAs for both photosystems (Barneche et al., 2006) and we refer to it as SLTAB2.
To confirm the heritability of the SLTAB2 silent epiallele, we analysed SLTAB2 in the F1 progeny of chlorotic sulf/+ (S. lycopersicum cv. Lukullus) crossed with S. pimpinellifolium. Out of 22 F1 plants, 8 displayed a paramutated phenotype with yellowing of parts of the leaves (Fig. 2A). The expression of SLTAB2 in these chlorotic plants was reduced by about half in their green sectors compared to wild-type plants, and 9-fold in their yellow sectors (Fig. 2B). Furthermore, in a PCR test that differentiated the polymorphic alleles from the two parents, we only detected expression from S. pimpinellifolium (Fig. 2C). These data are consistent with SLTAB2 being SULFUREA: in the green tissue the S. pimpinellifolium allele would have been expressed (but not the silent allele from the sulf/+ S. lycopersicum parent), and in the chlorotic tissue, it would have been paramutated. In addition, by confirming the heritable silencing of the S. lycopersicum allele in the chlorotic plants, these data confirm that SLTAB2 silencing is not merely a consequence of the sulf phenotype.
Furthermore, consistent with equivalence of SULF and SLTAB2, the Arabidopsis T-DNA knockout of ATAB2 is seedling-lethal in heterotrophic conditions and deficient in green pigment (Barneche et al., 2006;Fig. 3A). This mutant also has less auxin than the wild type (Fig. 3B). These three phenotypes all resemble sulf.

Paramutation is associated with changes in SLTAB2 promoter DNA methylation
We predicted, based on the analysis of the maize b1 gene (Stam et al., 2002), that the DNA of the paramutagenic sulf would be hypermethylated. After a genome-wide analysis of differentially methylated regions (DMRs) between wt and sulf, we looked for candidate loci in the appropriate region of chromosome 2 and adjacent to genes that were differentially expressed. Genome-wide there were thousands of such DMRs in the CHH context and hundreds in the CG and CHG contexts, with the CHH DMRs being predominantly hypermethylated in sulf whereas CG and CHG DMRs were evenly split between hyper-and hypo-DMRs (Fig. 4A). On chromosome 2 there were several differentially expressed genes with adjacent CHH DMRs but only the SLTAB2 locus had strong DMRs in all cytosine contexts (Fig. 4B).
Closer inspection revealed that there are two adjacent DMRs in the immediate promoter of SLTAB2, DMR1 Values are log 2transformed library-normalized/median-normalized counts. (C) Down-regulated genes in sulf on chromosome 2. Log 2 fold-change wt/sulf of significantly down-regulated genes (adjusted P <0.05). Percentage CHG methylation and gene density (genes/Mb) are plotted along chromosome 2 to show the distribution of heterochromatin. The S locus (red dotted line) marks the right-most border for the possible location of SULFUREA (Hagemann and Snoad, 1971). The end of the chromosome is marked by a solid red line. (D) SLTAB2 expression in wt, sulf/+ (green) and sulf (yellow) leaves. Horizontal bar shows the mean of five biological replicates. 26-fold reduction in sulf compared with wt (P-value=0.0002324, two-tailed t test), while the difference between wt and sulf/+ is not significant (P-value=0.1771). and DMR2 ( Fig. 4C; Table 2). DMR2 overlaps annotated repeats directly upstream of SLTAB2 and was hypomethylated in sulf in the CHG and CHH contexts (78% and 4% methylation, respectively, compared with 93% and 27% in the wild type). DMR1, by contrast, overlaps the transcriptional start site of SLTAB2 from 300 bp upstream, encompassing the first two exons and introns, and it was hypermethylated in sulf in all contexts (73% mCG, 62% mCHG, and 4% mCHH compared with 24%, 4%, and 0.7%, respectively, in wt). This hypermethylation of the transcriptional start site of SLTAB2 is consistent with a decrease in transcription in sulf and it further strengthens the case that SLTAB2 is SULF.
Additional evidence for the relevance of DMR1 methylation for paramutation is our finding that the silencing of the S. pimpinellifolium allele in the chlorotic F1 (sulf/+×S. pimpinellifolium) coincided with a hypermethylation of this region (Fig. 4D). The silent S. lycopersicum allele was methylated in all parts of the paramutated plants, whereas the S. pimpinellifolium allele was unmethylated in green sectors and methylated in yellow sectors.
We also predicted, based on the maize paramutation examples, that sulf would correlate with 24-nt siRNAs. At a genome-wide scale, there was a distinct increase in 23-24nt siRNAs in sulf compared with the wild type (Fig. 5A), in line with the pattern of CHH hypermethylation. At DMR1 of SLTAB2, the 23-24-nt siRNAs were more abundant in paramutated sulf (Fig. 5B) than in the wild type whereas at DMR2 the 23-24-nt siRNAs were less abundant than in the wild type (Fig. 5C).

VIGS of SLTAB2 phenocopies sulf
To test the involvement of SLTAB2 in sulf paramutation further, we used virus-induced gene silencing (VIGS). VIGS, when targeted to transcribed sequences, leads to knock-down of mRNA levels by post-transcriptional gene silencing but, when targeted at DNA sequences [e.g. the FWA promoter in A. thaliana (Bond and Baulcombe, 2015)], it can initiate heritable DNA methylation and transcriptional gene silencing. We cloned two segments (a and b, Fig. 4C) of DMR1 into tobacco rattle virus (TRV)  (Fig. 6A). The similarity of this VIGS phenotype to leaves of sulf is further evidence that SLTAB2 and SULF are equivalent.
DNA methylation analysis of chlorotic sectors by McrBC suggested that there is an epigenetic component to the silencing of SLTAB2 by TRV-DMR1b: the targeted DNA was as strongly methylated as in sulf samples (Fig. 6C). The involvement of epigenetics is further supported by the lasting VIGS phenotype several months post-inoculation (Fig. 6B) during which time the level of the virus vector decreased. From these data we conclude that the DMR1b region of SLTAB2 has the predicted characteristics of the paramutagenic component of sulf because it is susceptible to epigenetic modification.

Discussion
In this paper we present several lines of evidence that SLTAB2 is SULF: it maps closer to SULF than any other genes with the predicted pattern of mRNA accumulation in sulf/+ and sulf tissue (Fig. 1B); it encodes a protein required for photosystem I production that explains the chlorotic phenotype; the orthologous atab2 mutation has the same chlorosis and auxin-deficient phenotype as sulf (Fig. 3); and there is a definite DNA methylation mark at the silent SLTAB2 allele in sulf that is inherited from sulf/+ both in selfed and outcrossed progeny (Fig. 2). This epigenetic mark  is transferred to the previously active allele in F1 progeny (Fig. 4D) and is associated with 24-nt siRNAs (Fig. 5). The final evidence for the equivalence of SLTAB2 and SULF is from the finding that VIGS targeted to the SLTAB2 DMR can recapitulate both the physiological and epigenetic features of sulf (Fig. 6). Negative values indicate hypermethylation in sulfurea, positive values hypomethylation. DMRs whose downstream gene is differentially expressed (DE) between wild-type and sulfurea are coloured in blue, while DMRs whose downstream gene is not differentially expressed (non-DE) are coloured red. (C) Methylation over SLTAB2 promoter, plotted in 200 bp sliding windows (step of 100 bp). Repeats are plotted as blue arrows. (D) Wild-type sulf×S. pimpinellifolium F1s bear no methylation on either the S. lycopersicum or S. pimpinellifolium alleles at DMR1 (region 7 267 172 to 7 267 898 encompassing the S. pimpinellifolium SNPs, 11 clones per allele). In the green sectors of paramutated F1s only the S. lycopersicum allele is methylated (14 clones, 7 clones for S. pimpinellifolium allele), whereas in the yellow sectors methylation is found on both alleles (11 and 9 clones).
We envision that paramutation occurs when methylation of DMR1 DNA by paramutagenic siRNAs starts a positive feedback loop in which Pol IV is recruited to the silent locus by SHH1 (Law et al., 2013). The recruited Pol IV (Blevins et al., 2015;Zhai et al., 2015) would transcribe siRNAs that would mediate maintenance of the silent state of SLTAB2 and its transfer to paramutable alleles.
SLTAB2 had been ruled out previously as SULF because it is expressed at detectable levels in sulf and it was thought that, unlike tomato, the mutation of the Arabidopsis orthologue could be rescued on sucrose (Ehlert et al., 2008). An alternative candidate for SULF was implicated in the tryptophan-independent pathway of auxin biosynthesis. With our new data, however, we show that the previous exclusion of SLTAB2 was not valid because targeted suppression by VIGS or mutation of this gene produces an accurate phenocopy of the sulf phenotype including, in the atab2 mutant, an auxin defect and seedling lethality.
A likely explanation for the sulf phenotype based on silencing of SLTAB2 invokes the failure to translate the psaB mRNA as described for the orthologous mutations ATAB2 in Arabidopsis and TAB2 in Chlamydomonas (Dauvillée et al., 2003;Barneche al., 2006). The PsaB protein is the reaction centre protein of photosystem I and, in its absence, the thylakoid membranes would fail to form, pigments would not accumulate at the normal levels, and the leaves would be chlorotic. An auxin defect of sulf is a likely consequence of the PsaB defect, as observed in the atab2 mutant (Fig. 3). In addition, the genome-wide hypermethylation in the CHH context in sulf is reminiscent of transient hypermethylation in response to stress, already described in phosphate-starved rice (Secco et al., 2015) and virus-infected Arabidopsis (Bond and Baulcombe, 2015). Several cases of paramutation have been tied to the hypermethylation of regulatory tandem repeats (Stam et al., 2002;Chandler and Stam, 2004). By contrast, the sulfurea paramutation is correlated with increased DNA methylation at the transcriptional start site (DMR1) where there is no repeated DNA (Fig. 4). There are annotated LTR fragments at DMR2 that is directly adjacent to DMR1 but, unlike the classic systems, this region of the paramutagenic sulf allele showed a reduction in sRNAs and hypomethylation. Hypomethylation of DMR2 may be a consequence of a different chromatin state of the silenced allele and, although its repeats may not be directly involved in the silencing of SLTAB2, they and the largely heterochromatic region in which SLTAB2 is embedded may contribute to its paramutability, poising it for silencing.
The opportunity to study paramutation via SLTAB/SULF in tomato has several advantages over the various maize systems that have been most informative until now. First we have a VIGS system so that establishment of the epigenetic mark can be tracked directly in tomato mutants that are defective for components of the RNA silencing pathways. We will also be able to use VIGS on sulf/+ plants to test the role of various tomato genes in the establishment and maintenance of paramutagenicity and paramutability.
A second benefit of the sulf system is the possibility of studying the transfer of the epigenetic mark in vegetative tissue. With the well-studied maize paramutation systems this transfer is likely to occur early in embryo development and is not readily accessible to molecular analysis, whereas, in tomato, it will be taking place in or close to vegetative meristems. It will still not be easy to access the cells in which the allelic transfer is taking place but we may be able to use the DMR1-specific siRNAs as markers of the transfer process. These RNAs are rare in total plant extracts ( Fig. 5) but they may be more abundant at the primary sites of paramutation. Having identified the SULF gene, we should also be able to complement the physiological consequences by providing a transgene without the target DNA of paramutation so that we can grow plants with a sulf/sulf epigenotype.
These various experimental tools will allow us to explore the differences of the tomato and maize paramutation systems. For example, the apparent target of paramutation in sulf has no tandem repeats and is close to the transcriptional start whereas, in maize, at the B locus, they are essential and separated from the transcribed region by 100 kb. Answers to these and other questions will allow us to explore the frequency of paramutation-like events in plant breeding and evolution. Several recent findings indicate that such events are not restricted to the few well-characterized examples of paramutation in maize and other species (Regulski et al., 2013;Greaves et al., 2014). They may be frequent and have an effect on transgressive and heterotic phenotypes.

Supplementary data
Supplementary data are available at JXB online. Table S1. Summary of RNA-Seq alignments. Table S2. Summary of sRNA-Seq alignments. Table S3. Summary of whole-genome bisulfite sequencing alignments. Table S4. List of primers. Table S5. Summary of down-regulated genes in the genetically defined region of SULFUREA.