Red stripes have higher anthocyanin accumulation and transcript levels of biosynthetic genes

Red stripes of 'Royal Gala' and 'Honeycrisp' contained approximately eight and four times as much anthocyanin as green stripes (83 vs. 10 and 38 vs. 10 μg/g of anthocyanin monoglycoside equivalent for 'Royal Gala' and 'Honeycrisp', respectively). In all cases, the major anthocyanin detected was cyanidin-3-galactoside (Figure 3).

We subsequently compared the transcript levels of regulatory genes MYB10, MYB17, bHLH3 and bHLH33 and biosynthetic genes CHS, CHI, F3H, DFR1, LDOX, UFGT, in RNA isolated from red and green stripes of 'Royal Gala' and 'Honeycrisp' (Figure 4). MYB10 and MYB17 transcript levels correlated with anthocyanin concentration in both 'Honeycrisp' and 'Royal Gala', with higher mRNA levels in red stripes as compared to green stripes (ratios significantly larger than 1, p ≤ 0.05). Transcript levels of structural genes followed the same pattern as those of MYB10 and MYB17. Levels of the two bHLH transcription factors did not differ in green and red stripes (p ≤ 0.05), and therefore correlated poorly with anthocyanin concentration. These results reveal differential transcript accumulation of MYB10 and MYB17 in differentially pigmented stripes, which in turn results in a corresponding modulation of transcript levels of structural genes. MYB10 is a known activator of the apple anthocyanin pathway [17] and MYB17 has been shown to inhibit steps in the anthocyanin pathway [23] and has high sequence similarity to AtMYB4, a repressor of the phenylpropanoid pathway [44, 45]. We decided to further characterize MYB10 coding and upstream regions in order to determine whether sequence polymorphisms can explain different pigmentation patterns.

Low sequence diversity in the MYB10coding region in 'Honeycrisp', 'Connell Red' and 'Fireside'

To study the possibility that sequence differences are the cause of differential color patterns in the peel, we sequenced a total of 94 cDNA clones of the 'Honeycrisp' MYB10 coding region: 47 from a phenotypically uncharacterized 'Honeycrisp' fruit (harvested in late August when pigment pattern could not yet be conclusively determined), 24 from a mature striped and 23 from a mature blushed fruit. Ninety-two percent of the sequences obtained were 100% identical to MYB1-1, an allele of the MYB10 locus [17]. We found three single nucleotide polymorphisms (SNP) that produce changes in protein sequence, but since each one appeared only once in our dataset, and in phenotypically different apples, they most likely represent amplification or sequencing errors. These results indicate low levels of sequence diversity in the MYB10 coding region in 'Honeycrisp', with no evidence suggesting that the blushed/striped phenomenon is associated with modifications at the primary DNA sequence level within the coding region. MYB10 coding sequences from the striped cultivar Fireside (24 clones) and 'Connell Red' (23 clones), a stably blushed sport of 'Fireside', are identical to that of the most abundant version found in 'Honeycrisp' and the previously published MYB1-1 sequence--supporting our conclusion that differences in primary DNA sequence are not the source of differential patterns of apple peel pigment accumulation.

No size variation in MYB10promoter region among apple cultivars

We amplified three fragments (-2029 to -1229, -1411 to -678, and -677 to 47; nucleotide positions on the Genbank accession EU518249 relative to translation start site) collectively spanning about 2 Kb of the MYB10 promoter. PCR results did not indicate any fragment size differences among blushed and striped 'Honeycrisp', 'Connell Red' and 'Fireside' DNA, suggesting no large insertion or deletions were present. We sequenced the PCR products of each of these fragments from three independent reactions and found no sequence differences between blushed and striped 'Honeycrisp', or between 'Connell Red' and 'Fireside', although there were 14 SNPs between 'Honeycrisp' and the other two cultivars.

Neither presence nor transcription of a TRIM element explains apple peel phenotypic variation

Within the Plant & Food Malus gene database [46] was a DNA sequence identical to Genbank accession EU518249, the promoter of MYB10. Further upstream from this sequence, between positions -3038 and -2420 from the ATG translation start site of MYB10 (EU518249, 'Royal Gala') was a sequence with 85% identity to a Malus TRIM element (AY603367), a terminal-repeat retrotransposon in miniature [34]. We checked for the presence of a TRIM element upstream of the MYB10 locus in 'Honeycrisp' (blushed and striped), 'Connell Red', 'Fireside', '1807' (green selection) and 'Geneva' (ultra red cultivar) via PCR, combining a primer designed from the TRIM element (TRIM forward primer) with one designed from the promoter region of MYB10 (primer -1873). Results confirmed the presence of the TRIM element in each of these cultivars in a position identical to that observed in 'Royal Gala' (Figure 5C). We subsequently cloned and sequenced three PCR products from 'Honeycrisp' (blushed and striped), 'Connell Red' and 'Fireside'. Half of the fragments yielded sequences showing 99% or more identity to the previously published (EU518249) upstream region of 'Royal Gala' MYB10. The other sequences were probably amplifications from insertions of similar TRIM elements located elsewhere in the genome, with percent identities to TRIM ranging from 40 to 56.5%.

We tested for TRIM transcript presence in blushed and striped 'Honeycrisp', 'Connell Red', 'Fireside', 'Geneva' (ultra red cultivar) and 'Honeygold' (green cultivar), and found it to be transcribed in all cases. However, a fragment spanning a portion of the TRIM element and extending 500 bp into the upstream region of MYB10 did not amplify from total RNA, indicating that transcription from the TRIM element did not extend into MYB10 in these cultivars. Overall, results indicated that neither the presence of the TRIM element in the MYB10 promoter region nor its transcription explained the differences in peel pigment accumulation among the cultivars studied.

Increased methylation levels in green stripes

DNA samples from green and red stripes of 'Honeycrisp' (2007 samples) and 'Royal Gala' were treated with the methylation-sensitive endonuclease McrBC to ascertain whether the observed differences in transcript accumulation were associated with methylation differences in the promoter or coding region of MYB10 (Figure 5). McrBC preferentially cuts DNA containing methylcytosine on one or both strands, between two recognition sites [5'...Pu^mC(N_40-3000)Pu^mC...3']. McrBC treated and mock-digested templates were compared using real-time PCR, and percent methylation was calculated. In total, 18 fragments starting at the transposon insertion and spanning 2.5 Kb of the promoter region and three exons of MYB10, were evaluated. Results indicated that a region of the MYB10 promoter, encompassing the fragments between nucleotide positions -1411 and -555 is highly methylated (above 60%) in both cultivars. 'Connell Red' and 'Fireside' had low methylation (20-40%) in the -2254 to -2098 fragment and high methylation (95%) in the -846 to -651 fragment, indicating a similar pattern of MYB10 methylation in these cultivars relative to those observed in 'Royal Gala' and 'Honeycrisp' (Figure 6).

Green stripes of 'Honeycrisp' (2007 samples) showed higher overall methylation levels than red stripes throughout the promoter region (Figure 5A). The -704 to -555 fragment was omitted from this comparison since quantification in the McrBC digested samples was highly variable due to extremely low template levels, indicating that this region was so highly methylated that treatment with McrBC resulted in nearly complete digestion of the template DNA. Sequence analysis indicated that differences in predicted methylation levels between regions were not due to difference in the number of potential McrBC recognition sites (data not presented). Similar results were obtained for 2008 fruits, but overall methylation levels were higher than in 2007 and differences between red and green stripes were even greater (Figure 7). These results indicate that while methylation levels are variable between years, green stripes are consistently associated with higher methylation of MYB10 promoter regions. Similar trends were observed in 'Royal Gala' for some of the fragments, except that the differences between red and green stripes were smaller. In total, higher methylation levels were observed for 'Royal Gala' than 'Honeycrisp' (Figure 5B).

In contrast to 'Honeycrisp' red and green stripes, we hypothesized that color differences between red (exposed to light) and green (unexposed to light) regions of the peel of blushed apples are only due to light effects and not to differences in methylation levels. We therefore compared methylation percentages in red (exposed) and green (unexposed) areas of blushed apples and red and green stripes. Results indicated no significant differences (p ≤ 0.05) between red and green regions of the peel of blushed apples. Interestingly, in two out of the three regions studied (-1411 to -1229 and -846 to -651), red stripes have methylation levels comparable to those in the exposed peel portions of blushed apples, while green stripes have methylation levels higher than those of red stripes or red and green regions of blushed apples (Figure 7).

Bisulfite sequencing offers greater resolution than McrBC-based methods for the detection of methylated cytosines. Building on McrBC results, we next pursued bisulfite sequencing of MYB10 promoter regions from 'Honeycrisp' and 'Royal Gala'. Preliminary bisulfite sequencing experiments indicated that cytosine methylation in the promoter region of MYB10 is found in all three methylation contexts (i.e. CHH, CHG, and CG, where H is A, C or T). To avoid amplification bias, we therefore designed degenerate PCR primers to target two different promoter regions. This severely constrained areas that could be targeted, and amplification upstream of -1007 was ultimately unsuccessful using unbiased primers. A comparison of methylation levels between red and green stripes in the -1007 to -684 and -534 to -184 regions confirmed that green stripes are more highly methylated than red stripes (9.3 and 5.2% difference respectively), with 80% and 65% of cytosines showing higher methylation levels in green than in red stripes in the -1007 to -684 and -534 to -184 regions respectively (Figure 8A). Further analysis of the -1007 to -684 region indicated that clones obtained from green stripes have higher overall methylation levels than those obtained from red stripes (Additional files 1 and 2). Observed methylation differences between red and green stripes are modest, but actual differences may be greater. Although great care was taken, manual isolation of red and green stripes from 'Honeycrisp' peels was imprecise, resulting in tissue samples that were substantially enriched for red or green stripes but not entirely devoid of contaminating tissues. Thus, DNA samples used for McrBC- and bisulfite-based analyses, while predominantly derived from the target tissue (red or green stripes) likely represent a mixture of DNA, partially obscuring methylation differences between sources.

Consistent with our preliminary results, different cytosine contexts did not exhibit distinct methylation patterns; all cytosine contexts showed high methylation levels in highly methylated regions and vice versa (Figure 8B). Overall, CHH and CHG methylation was highest (20.2 and 16.9% respectively) and CG methylation was lowest (1.6%). A sequence alignment for the -1007 to -684 region is presented in Additional file 2.

Anthocyanin genes transcript levels correlate with striped patterns

Anthocyanin and transcript quantifications performed in this study suggested a possible mechanism controlling pigment patterns in apple fruit peels. We have found that green stripes are associated with lower anthocyanin accumulation, which is explained by reduced transcript levels of all the anthocyanin pathway genes evaluated, including the structural genes in the pathway, and MYB10, a transcription factor which regulates them. An additional gene evaluated in this study, MYB17, an apple transcription factor that represses anthocyanin synthesis [23] was transcribed in a similar manner to MYB10. As MYB17 is not elevated in green sectors relative to red, we considered MYB10, the main transcription factor regulating the pathway in apple [12, 15, 17, 21], to be the most likely candidate to be involved in peel variegation. We therefore sought to identify a mechanism responsible for MYB10 transcript level differences.

Variegation in apple peels is associated with MYB10methylation mosaicism

Our results indicate an inverse association between methylation levels in the MYB10 promoter and anthocyanin accumulation in striped apple peels. As previously suggested by Cocciolone and Cone [31] for maize, we hypothesized that early in apple fruit development, differences in MYB10 methylation are present among individual cells. Throughout fruit growth, these differentially methylated cells give rise to sectors of peel varying in their ability to accumulate anthocyanins. Our results indicate that DNA methylation in the promoter of MYB10 is associated with reduced transcript accumulation. Specifically, we propose that differential methylation of MYB10 promoter regions in red vs. green stripes of the 'Honeycrisp' peel results in differential accumulation of the MYB10 transcript, which in turn determines both transcription of anthocyanin structural genes and pigment accumulation. DNA methylation may affect MYB10 transcription through interference with the RNA-polymerase transcription complex or by preventing binding of additional factors required for transcription. In Arabidopsis, genome wide studies of DNA methylation have found that methylation within regulatory regions is rare and probably selected against, as it may interfere with transcription initiation [47]. Our results suggest that high levels of methylation in certain promoter regions of a key transcription factor in the flavonoid biosynthetic pathway in apple may play a regulatory role, but it is not inhibitory for gene activity. It is possible that since apple trees are clonally propagated and stay in production for many years (i.e. fruit peel tissue does not derive from cells that have undergone recent meiosis), mechanisms of epigenetic regulation might not be identical to what has been described in more widely studied species such as Arabidopsis and rice.

Different methylation levels early in apple fruit development could be mitotically maintained from those in the meristematic cells that gave origin to the fruit, or could result from demethylation or de novo methylation. Previous results in 'Honeycrisp' suggest that there is at least some mitotic maintenance of methylation states, given that trees clonally propagated from buds on branches with exclusively blushed fruits, tend to produce a higher percentage of blushed fruit [32]. Nonetheless, results from the same study indicated that additional factors can influence the pattern in the peel, namely position of the fruit on the tree and crop load.

The presence of a TRIM transposable element in an upstream region of the MYB10 promoter might influence the changes in methylation observed between different regions of the peel but neither its presence per se nor its transcription appears to be responsible for determining peel pigmentation. The TRIM element identified in this study is located 2.5 Kb upstream of the predicted translation start site, and is present in 'Honeycrisp', 'Royal Gala', and five other cultivars with peel pigmentation ranging from green to ultra red. Lippman et al. [48] indicated that in Arabidopsis transposable elements can determine epigenetic gene silencing when inserted within or very near (<500 bp) a gene. The effect of a transposable element 2.5 Kb upstream of the coding region is unknown. We did not find any evidence of transposable element sequences within the highly methylated promoter region of MYB10.

Within the most methylated region of the MYB10 promoter in this study (-1411 to -555; Figure 5) are five putative E-box motifs [22] which are bHLH-related cis-acting elements (CACATG) [49, 50]. Methylation was absent at the three E-box motifs evaluated using bisulfite sequencing, but highly methylated areas occurred a few nucleotides upstream or downstream of these motifs. This may suggest a potential regulatory role for one or more of these motifs. An assessment of the presence of other types of epigenetic marks such as histone methylation can shed additional light on the mechanism involved in MYB10 regulation. Our results show that methylation followed the same pattern in all three cytosine methylation contexts, with alternation of high and low methylation regions. The high methylation levels observed for CHH and CHG sites, and low methylation levels for the CG context, indicate a pattern not representative of what is generally observed in flowering plants [51, 52]. Broader survey of methylation patterns throughout the apple genome is warranted.

The unstable nature of pigment patterning in 'Honeycrisp' could be a result of a more variable cell to cell methylation pattern than is present in other cultivars producing fruit with consistent pigment patterns, such as 'Royal Gala', 'Fireside' and 'Connell Red'. We speculate that the occurrence of stripes in 'Honeycrisp' is a consequence of higher than normal methylation levels in the MYB10 promoter region in the green stripes, something that occurs only in some fruit and to varying degrees. In contrast, MYB10 methylation levels and thus peel pigmentation remains more constant in 'Royal Gala'.

Plant material

Leaf samples of 'Honeycrisp', 'Connell Red', 'Fireside', '1807', 'Honeygold' and 'Geneva' apple were collected in early spring of 2005 and fruits of the same cultivars were collected at maturity during the 2005, 2006, 2007 and 2008 growing seasons from trees at the Horticultural Research Center in Chanhassen, Minnesota. In February 2008 ('Royal Gala') and 2010 ('Honeycrisp') fruits were harvested at Plant and Food Research orchards (Nelson, New Zealand). 'Royal Gala' apples grown in Chile were purchased in a Minnesota grocery store in April 2008 to be used for methylation experiments. For the MYB10 characterization experiments, whole fruit peels were used. For anthocyanin quantification, transcript analyses and methylation studies, red and green stripes were carefully separated using a razor blade. Since stripes cannot be absolutely classified as green or red, samples are more accurately described as "red stripe enriched" or "green stripe enriched". Both green and red stripes were obtained from the same region of the peel at each time, preventing the possibility that the "red stripe enriched" sample would also be enriched for fruit peel regions more exposed to light and vice versa. For comparisons between different blushed fruit regions, light-exposed (redder) and -unexposed (greener) peel regions were separated. For both blushed and striped fruit regions, peel tissue from at least two apples was pooled. In all cases, leaves and peels were immediately frozen in liquid nitrogen and placed at -80°C before anthocyanins, DNA or RNA was extracted.

Identification and quantification of anthocyanins

Apple peel samples were finely ground in liquid nitrogen and then resuspended in 1 ml methanol and 0.1% HCl. Samples were sonicated for 4 min, stored at room temperature in the dark for 3 h and then centrifuged at 3,000 × g. Aliquots of 1.0 ml were dried down to completion in a Labconco Centrivap Concentrator (Labconco, Kansas City, MO, USA). Samples were resuspended in 20% methanol (250 μl). Anthocyanins were identified by LC-MS analysis as described previously [53]. Identification was based both on masses (M⁺) of molecular ions and characteristic fragments/neutral losses and comparison of retention times and fragmentation with authentic standards of cyanidin-3-O-glucoside and cyanidin-3-O-galactoside. M⁺ fragments observed were 303 Da (delphinidin), 287 Da (cyanidin) and 271 Da (pelargonidin). Neutral losses (i.e. mass differences between fragments) observed were 162 Da (hexoside sugar, e.g. galactose), 146 Da (deoxyhexoside sugar, e.g. xylose) and 132 Da (pentoside sugar, e.g. arabinose). MS cannot distinguish between sugars of the same mass (e.g. glucose/galactose). Anthocyanins and other phenolic compounds were quantified by HPLC as described previously [53]. Quantification was achieved by reference to standards of anthocyanins and other phenolic compounds, using LC-MS data to confirm identification of peaks.

Real-time transcript analysis

Mature 'Royal Gala' peel RNA was isolated by a method adapted from Chang et al. [56], quantified in a NanoDrop nd-100 spectrophotometer running software version 3.0.1 (NanoDrop Technologies, Wilmington, DE) and treated with DNA-free DNAse (Ambion, Austin, TX). For real time-PCR analysis, total RNA was reverse-transcribed into cDNA using the Super-Script III (Invitrogen Corp.) reverse transcriptase kit. Real-time PCR amplification and analysis were carried out using the Roche 480 LightCycler System (Roche Diagnostics, Mannheim, Germany). All reactions were performed using the LightCycler 480 SYBR Green I Master Mix (Roche Diagnostics) following manufacturer instructions. Reactions were performed in triplicate using 5 μl 5X Master Mix, 1.0 μM each primer and 3 μl diluted cDNA. A negative nuclease-free water (Roche Diagnostics) control was included in each run. Primers used are the same as described above. PCRs used an initial denaturation step at 95°C for 5 min, followed by 50 cycles of denaturation for 10 s at 95°C, annealing for 10 s at 60°C and elongation for 20 s at 72°C. Fluorescence was measured at the end of each annealing step at 72°C. Amplification was followed by a melting curve analysis with continual fluorescence data acquisition during the 65-95°C melt curve. The raw data were analyzed with the LightCycler software, (LightCycler 480, Software 1.5) and transcript levels were normalized to actin to minimize variation in cDNA template levels. For each gene, a standard curve was generated using a cDNA serial dilution, and the resultant PCR efficiency calculations (ranging between 1.81 and 2.0) were imported into relative transcript level analysis. Error bars shown in real-time PCR data are technical replicates, representing the means ± SE of three replicate real-time PCR reactions. ANOVA on real-time PCR data was performed as described above.

MYB10characterization

To study sequence diversity in the MYB10 coding region, fruit peel RNA and cDNA were obtained using the modified Lopez-Gomez and Gomez-Lim extraction method as described above. The MYB10 coding region was amplified using PfuUltra™ Hotstart DNA Polymerase (Stratagene, La Jolla, CA) using MYB10 cDNA primers (Table 1) and DNA template. Reactions were performed in a 50 μl total volume (15 ng template, 100 ng/μl each primer, 25 mM each dNTP, 2.5 units AmpliTaq™ (Applied Biosystems), 10X buffer provided by manufacturer and 25 mM MgCl₂). PCRs used 35 cycles of 94°C 30 s, 50°C 30 s, 72°C 120 s (Gene Amp PCR system 9700, Applied Biosystems). Fragments were then A-tailed by incubating 3 μl PCR product with 1 μl AmpliTaq™ (Applied Biosystems), 1 μl buffer provided by manufacturer, 1 μl 2 mM dATP, and 1 μl sterile water for 24 minutes at 70°C. Fragments were then desalted through a MicroSpin™ S-200 HR column (Amersham Biosciences, Piscataway, NJ) according to manufacturer's recommendations. Desalted fragments were cloned into the pGEM^®-T Easy Vector (Promega Corp.), also according to manufacturer's instructions. Plasmids were purified from 3 ml overnight cultures using the Wizard Plus SV Minipreps DNA Purification system (Promega Corp.). To verify insert size, 3 μl of plasmid DNA were digested in 1X manufacturer supplied buffer by 10 units EcoRI (Invitrogen) in a 10 μl total volume at 37°C for 1 h. The entire reaction was loaded and separated on 1% agarose gels in TBE buffer, stained with ethidium bromide, and photographed under UV light. Inserts were compared to DNA standards of known size. Subsequently, undigested plasmids were sequenced using 3.2 pM of M13 forward or reverse primer. All nucleotide sequences were determined by Applied Biosystems BigDye Terminator version 3.1 cycle sequencing on an Applied Biosystems 3130xl or 3730xl automatic sequencer (Applied Biosystems) at the University of Minnesota DNA Biomedical Genomics Center's sequencing and analysis facility. Sequences were analyzed, assembled into contigs, and compared to known sequences using SeqMan™ II (Windows 32 vs. 5.08; DNASTAR Inc, Madison, WI).

For characterization of the MYB10 region upstream of the translation start site (referred to as "promoter" for simplification), leaf tissues or fruit peels were very finely ground in liquid nitrogen, and DNA was isolated using the Haymes' method [57] or using the DNeasy Plant mini Kit (Qiagen). Three promoter regions were amplified using PfuUltra ™ Hotstart DNA Polymerase (Stratagene) using MYB10 primer pairs -2029/-1229, -1411/-678 and -677/47 (Table 1). Reactions were performed as described above, but without additional MgCl₂. PCR fragments were then desalted through a MicroSpin™ S-300 HR column (Amersham Biosciences) according to manufacturer's recommendations and fragments from three independent replicate reactions per sample were sequenced directly using 3.2 pM of the corresponding forward and reverse primers, as detailed above.

Methylation studies

Peel genomic DNA (less than 1 μg) from red or green stripes, or from red and green areas of blushed apples harvested in 2007 and 2008, was digested with McrBC (New England Biolabs, Beverly, MA) in 100 μl total volume including 1X NEB2 buffer, 0.1 mg/mL bovine serum albumin, 1 mM GTP and 40 U McrBC or 50% glycerol (mock digested reactions). Digestions were incubated overnight at 37°C to ensure complete digestion and then incubated at 65°C for 20 min to halt enzyme activity. Real-time experiments were performed on McrBC and mock digested template as described above for 'Honeycrisp' and primers used are presented in Table 2. For each experiment, real-time PCR runs, including a control (mock digested DNA) and a McrBC digested sample, were performed in triplicate, and two or three independent digestions were used. Percent methylation for individual samples was calculated as a function of the delta CT between control and McrBC treated DNA, using the formula:

Student's t-test was used to establish significant differences in template amounts between biological replicates and subsequently calculate sample size. The estimated sample size was used when determining whether significant differences occurred between red and green peel regions.

Peel genomic DNA from red or green stripes (2009 harvested apples) were subjected to bisulfite conversion using the EZ DNA Methylation-Gold™ Kit (Zymo Research Corp., Irvine, CA) following manufacturer's recommendations. Two MYB10 promoter regions were amplified using MYB10 primer pairs -1007/-684 and -534/-184 (Table 1). PCRs were performed using Zymo Taq™ DNA Polymerase (Zymo Research Corp.) in a 50 μl total volume (1.0 μl bisulfite converted DNA as template, 0.6 μM each primer, 250 μM each dNTP, 2.0 units Zymo Taq, 1X buffer provided by manufacturer). PCRs used an initial denaturation step at 95°C for 10 min, followed by of 35 cycles of 95°C 30 s, 50°C 40 s (-1007 to - 684 fragment) or 55°C (-534 to - 184 fragment), 72°C 60 s, and a final elongation step at 72°C for 7 minutes. A secondary PCR was carried out using the same primers and conditions, and 1.0 μl of the primary PCR product. Fragments were then desalted through a MicroSpin™ S-300 HR column (Amersham Biosciences) according to manufacturer's recommendations. Desalted fragments were cloned into the pGEM^®-T Easy Vector (Promega Corp.), also according to manufacturer's instructions. Bacterial colonies were frozen in 100 μl aliquots of Luria broth (Miller) solution with 10% glycerol in 96-well plates and shipped on dry ice to Beckman-Coulter Genomic Services (Beverly, MA) for Sanger sequencing. Percent methylation was calculated based on the methylation status of each cytosine within the two regions sequenced, using 12 to 24 clones per sample.