Limits...
Accurate measurement of 5-methylcytosine and 5-hydroxymethylcytosine in human cerebellum DNA by oxidative bisulfite on an array (OxBS-array).

Field SF, Beraldi D, Bachman M, Stewart SK, Beck S, Balasubramanian S - PLoS ONE (2015)

Bottom Line: We performed four replicates of the same sample of cerebellum and found a high level of reproducibility (average r for BS = 98.3, and average r for oxBS = 96.8).We validated our results using qPCR in conjunction with glucosylation of 5hmC sites followed by MspI digestion and we found good concordance with the array estimates (r = 0.94).We also provide a novel method for validating the presence of 5hmC at low levels, and highlight some of the pitfalls associated with measuring 5hmC and 5mC.

View Article: PubMed Central - PubMed

Affiliation: Cancer Research UK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge, CB2 0RE, United Kingdom.

ABSTRACT
The Infinium 450K Methylation array is an established tool for measuring methylation. However, the bisulfite (BS) reaction commonly used with the 450K array cannot distinguish between 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC). The oxidative-bisulfite assay disambiguates 5mC and 5hmC. We describe the use of oxBS in conjunction with the 450K array (oxBS-array) to analyse 5hmC/5mC in cerebellum DNA. The "methylation" level derived by the BS reaction is the combined level of 5mC and 5hmC at a given base, while the oxBS reaction gives the level of 5mC alone. The level of 5hmC is derived by subtracting the oxBS level from the BS level. Here we present an analysis method that distinguishes genuine positive levels of 5hmC at levels as low as 3%. We performed four replicates of the same sample of cerebellum and found a high level of reproducibility (average r for BS = 98.3, and average r for oxBS = 96.8). In total, 114,734 probes showed a significant positive measurement for 5hmC. The range at which we were able to distinguish 5hmC occupancy was between 3% and 42%. In order to investigate the effects of multiple replicates on 5hmC detection we also simulated fewer replicates and found that decreasing the number of replicates to two reduced the number of positive probes identified by > 50%. We validated our results using qPCR in conjunction with glucosylation of 5hmC sites followed by MspI digestion and we found good concordance with the array estimates (r = 0.94). This experiment provides a map of 5hmC in the cerebellum and a robust dataset for use as a standard in future 5hmC analyses. We also provide a novel method for validating the presence of 5hmC at low levels, and highlight some of the pitfalls associated with measuring 5hmC and 5mC.

Show MeSH
Comparison between oxBS arrays and qPCR.Right: Correlation between 5hmC detected by qPCR vs array probes. Left: Discrepancy between qPCR and arrays quantified as percentage difference relative to qPCR (See also S1 Document). Data points having the same coordinates appear in darker red.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4338296&req=5

pone.0118202.g008: Comparison between oxBS arrays and qPCR.Right: Correlation between 5hmC detected by qPCR vs array probes. Left: Discrepancy between qPCR and arrays quantified as percentage difference relative to qPCR (See also S1 Document). Data points having the same coordinates appear in darker red.

Mentions: We validated these results using an orthogonal assay that employs glucosylation of 5hmC sites followed by MspI digestion, which is specific for CCGG sites, and qPCR. 5hmC glucosyltransferase can be used to modify 5hmC, yielding glucosyl-5hmC, which is not susceptible to digestion by MspI[23]. Therefore by performing qPCR on glucosylated, unglucosylated and undigested DNA it is possible to ascertain the levels of 5hmC at a given site. We chose 27 sites that had a range of 5hmC values according to the oxBS array data from 0% to 40% and including apparently “negative” values. Sites were selected first by 5hmC level then by the availability of a CCGG motif at that site (see S3 Table). The correlation coefficient between the 450K results and the qPCR results was 0.94. Sites with negative values by the oxBS array clustered around zero on the qPCR (See Fig. 8 and S2 Table) demonstrating that these “negative” values are indeed an artefact resulting from the random distribution around 0. The correlation between the oxBS array and the qPCR validation was maintained across the dynamic range of the array experiment from 3% to 40% (see Fig. 8). Two recently published studies have also measured 5hmC in brain tissue using arrays. Stewart et al. [24] used the oxBS array assay while Chopra et al. [22] used TAB array. Our data shows consistency with the dataset from Stewart et al. (see S10 Fig.) whereas the consistency was less good with the TAB array data, suggesting the possibility of method-based differences (see S9 Fig.). However, we must be cautious in drawing conclusions about the comparisons, as the samples were different. The independent validation of our method by a distinct assay (glucosylation and MspI digestion) across the full dynamic range of our data, provides us with confidence in the accuracy of our measurements.


Accurate measurement of 5-methylcytosine and 5-hydroxymethylcytosine in human cerebellum DNA by oxidative bisulfite on an array (OxBS-array).

Field SF, Beraldi D, Bachman M, Stewart SK, Beck S, Balasubramanian S - PLoS ONE (2015)

Comparison between oxBS arrays and qPCR.Right: Correlation between 5hmC detected by qPCR vs array probes. Left: Discrepancy between qPCR and arrays quantified as percentage difference relative to qPCR (See also S1 Document). Data points having the same coordinates appear in darker red.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4338296&req=5

pone.0118202.g008: Comparison between oxBS arrays and qPCR.Right: Correlation between 5hmC detected by qPCR vs array probes. Left: Discrepancy between qPCR and arrays quantified as percentage difference relative to qPCR (See also S1 Document). Data points having the same coordinates appear in darker red.
Mentions: We validated these results using an orthogonal assay that employs glucosylation of 5hmC sites followed by MspI digestion, which is specific for CCGG sites, and qPCR. 5hmC glucosyltransferase can be used to modify 5hmC, yielding glucosyl-5hmC, which is not susceptible to digestion by MspI[23]. Therefore by performing qPCR on glucosylated, unglucosylated and undigested DNA it is possible to ascertain the levels of 5hmC at a given site. We chose 27 sites that had a range of 5hmC values according to the oxBS array data from 0% to 40% and including apparently “negative” values. Sites were selected first by 5hmC level then by the availability of a CCGG motif at that site (see S3 Table). The correlation coefficient between the 450K results and the qPCR results was 0.94. Sites with negative values by the oxBS array clustered around zero on the qPCR (See Fig. 8 and S2 Table) demonstrating that these “negative” values are indeed an artefact resulting from the random distribution around 0. The correlation between the oxBS array and the qPCR validation was maintained across the dynamic range of the array experiment from 3% to 40% (see Fig. 8). Two recently published studies have also measured 5hmC in brain tissue using arrays. Stewart et al. [24] used the oxBS array assay while Chopra et al. [22] used TAB array. Our data shows consistency with the dataset from Stewart et al. (see S10 Fig.) whereas the consistency was less good with the TAB array data, suggesting the possibility of method-based differences (see S9 Fig.). However, we must be cautious in drawing conclusions about the comparisons, as the samples were different. The independent validation of our method by a distinct assay (glucosylation and MspI digestion) across the full dynamic range of our data, provides us with confidence in the accuracy of our measurements.

Bottom Line: We performed four replicates of the same sample of cerebellum and found a high level of reproducibility (average r for BS = 98.3, and average r for oxBS = 96.8).We validated our results using qPCR in conjunction with glucosylation of 5hmC sites followed by MspI digestion and we found good concordance with the array estimates (r = 0.94).We also provide a novel method for validating the presence of 5hmC at low levels, and highlight some of the pitfalls associated with measuring 5hmC and 5mC.

View Article: PubMed Central - PubMed

Affiliation: Cancer Research UK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge, CB2 0RE, United Kingdom.

ABSTRACT
The Infinium 450K Methylation array is an established tool for measuring methylation. However, the bisulfite (BS) reaction commonly used with the 450K array cannot distinguish between 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC). The oxidative-bisulfite assay disambiguates 5mC and 5hmC. We describe the use of oxBS in conjunction with the 450K array (oxBS-array) to analyse 5hmC/5mC in cerebellum DNA. The "methylation" level derived by the BS reaction is the combined level of 5mC and 5hmC at a given base, while the oxBS reaction gives the level of 5mC alone. The level of 5hmC is derived by subtracting the oxBS level from the BS level. Here we present an analysis method that distinguishes genuine positive levels of 5hmC at levels as low as 3%. We performed four replicates of the same sample of cerebellum and found a high level of reproducibility (average r for BS = 98.3, and average r for oxBS = 96.8). In total, 114,734 probes showed a significant positive measurement for 5hmC. The range at which we were able to distinguish 5hmC occupancy was between 3% and 42%. In order to investigate the effects of multiple replicates on 5hmC detection we also simulated fewer replicates and found that decreasing the number of replicates to two reduced the number of positive probes identified by > 50%. We validated our results using qPCR in conjunction with glucosylation of 5hmC sites followed by MspI digestion and we found good concordance with the array estimates (r = 0.94). This experiment provides a map of 5hmC in the cerebellum and a robust dataset for use as a standard in future 5hmC analyses. We also provide a novel method for validating the presence of 5hmC at low levels, and highlight some of the pitfalls associated with measuring 5hmC and 5mC.

Show MeSH