By adapting existing next-generation sequencing strategies to profile DNA methylation patterns in small amounts of starting material, a team from the UK, Sweden, and Germany hopes to garner longitudinal epigenetic information from newborn bloodspot samples that may eventually turn the samples themselves into a resource for understanding future disease risk.
"If we could use genetics and epigenetics at birth, that could provide an interesting and potentially useful avenue for prediction," Vardhman Rakyan, an epigenetics researcher at the Queen Mary University of London, told In Sequence.
Rakyan was co-senior author on a study appearing online last week in Genome Research in which he and his colleagues extracted DNA from small pieces of filter paper spotted with newborn cord blood samples, known as Guthrie cards, and looked at the quality and consistency of the cytosine methylation data that could be obtained from these samples using a bisulfite conversion-based array or a modified methylated DNA immunoprecipitation sequencing, or MeDIP-seq, method.
Although each Guthrie card typically contains four bloodspots from the same individual, the researchers were intent on findings ways of using DNA from just one of these spots, which typically yield between 50 and 200 nanograms of genetic material. To do that, they tweaked the MeDIP-seq protocol, making it suitable to do methylation profiling with around 200 nanograms of DNA rather than microgram amounts, as is more standard.
"We scaled down almost every step. So we used, for example, less antibody; we scaled all the volumes down, accordingly," Rakyan explained. "During the PCR step we had to do replicates and additional amplification cycles."
Once they had determined that they could get consistent methylation information from Guthrie cards and other samples using array- and sequencing-based approaches, the investigators went on to compare newborn methylation profiles in Guthrie cards from three individuals with methylation patterns found in blood samples from the same individuals three years later.
Based on the patterns at differentially methylated sites in the genomes of these three individuals, the study authors proposed that methylation profiles remain quite stable in each individual over time. If so, they argued, newborn blood samples may provide a chance to peek at epigenetic profiles that may be clinically relevant later in life.
The researchers are now transitioning over to bisulfite sequencing for ongoing Guthrie card studies aimed at exploring such questions. By comparing methylation profiles at birth with those found in samples collected throughout an individual's lifetime, they hope to tease apart some of the methylation profiles at play in various human traits and diseases.
"There's a lot of interest now in performing epigenome-wide association studies, or EWASs, similar to a GWAS," Rakyan said. "One of the biggest things, though, is if you find an epigenetic difference in an individual with any given disease, it's impossible to say whether that epigenetic difference is due to the disease itself or has any role in causing the disease — it's cause versus consequence."
"Having Guthrie cards, in some ways, helps you to address this because you're getting the epigenetic profile at birth and, in most cases, prior to overt disease of any kind."
Arrays vs. MeDIP-seq
For the current study, the researchers opted for the MeDIP-seq method rather than whole-genome bisulfite sequencing in part due to cost concerns.
Moreover, since MeDIP-seq was a method that had been developed in their lab, Rakyan said, the team was already familiar with it and felt confident making subtle modifications and miniaturizations to the protocol to make it more suitable for testing the newborn bloodspots.
In parallel to their validation of the sequencing-based method, the team tested the bloodspot samples using Illumina 450K arrays, which assess roughly 450,000 cytosine methylation marks across the genome. These include sites where cytosine and guanine nucleotides are found together frequently, called CpG islands, as well as gene promoter and enhancer regions.
The Guthrie cards used for the study had been spotted with newborn cord blood more than a decade ago and assessed as part of a Swedish diabetes prediction study published in 2004. To ensure that the methylation profiles they found in the bloodspot samples were as accurate as possible and able to pick up known tissue-specific methylation marks, the team also tested their methods in fresh blood and sperm samples collected from unrelated individuals.
After determining that they could get fairly reliable methylation profiles using the 450K array and genomic DNA from Guthrie cards, the investigators tested their MeDIP-seq method too, comparing results across sample types and methylation testing methods.
Each MeDIP sample was sequenced on one lane of the Illumina GAIIx, providing around 20 million or 30 million reads — enough to get good data for most methylated cytosines, Rakyan explained, though he said that even deeper coverage would reveal regions with smaller methylation differences.
Preliminary experiments indicated that the methylation profiles uncovered by sequencing the Guthrie card samples corresponded well with those detected on the Illumina 450K array, the researchers reported, showing between around 75 percent and 93 percent agreement.
"For regions of medium to high CpG density, the correlation was really, really good," Rakyan said. "When you start getting low CpG density, MeDIP-seq — or MeDIP in general — doesn't seem to work as well."
Overall, though, the consistency they saw between the two methods emboldened the researchers to do more elaborate analyses on samples from a few individuals.
For two girls and one boy sampled in the Swedish study, they looked not only at Illumina 450K-based methylation profiles at birth, based on genomic DNA from Guthrie cards, but also at those found by MeDIP-seq using DNA from whole-blood samples taken when the same children were three years old.
Again, the team used the Illumina GAIIx to do MeDIP-seq on one sample per lane at a cost of around £1,000 ($1,580).
Rakyan estimated that it would now be possible to test around a dozen samples by MeDIP-seq for the same price by multiplexing on the HiSeq 2000, though he cautioned that researchers who are keen to multiplex on the Illumina instrument will need to make their own adaptors, since Illumina's TruSeq adaptors are themselves methylated.
Results from the newborn and three-year-old comparison highlighted a slew of sites in the genome that were differentially methylated between the three individuals.
But in an attempt to focus on the methylation profiles that were likely to be mediated by environmental influences rather than genetic differences, Rakyan explained, the researchers decided to ignore differentially methylated regions found in the neighborhood of the 2.5 million SNPs that are assessed by Illumina's Omni2.5S array.
After its filtering steps, the team was left with between 10 and 13 regions showing consistent differential methylation patterns from one individual to the next, both in the newborn bloodspot samples and in the three-year-olds.
"Those were the differences we highlight in our paper, which might not be due to genetic differences," Rakyan explained. "Those are the ones that we really show are stable."
"Given the stringency of our filters," the study authors noted, "this is almost certainly an under-estimate of the level of inter-individual DNA methylation variation that exists between any two unrelated human individuals."
'An Amazing Resource'
It will likely be some time before sequencing-based methylation profiling at birth becomes standard, if ever, the study authors explained, since so much remains unknown about the ties between the epigenome and disease risk.
Still, since the use of newborn bloodspot samples for testing for genetic conditions such as phenylketonuria, cystic fibrosis, and sickle cell disease has become so common, they argued that Guthrie card samples may make it possible to do some of the very studies needed to tease apart the relationship between epigenetics and disease.
"If they are stored — and, in some places, they are being stored indefinitely — and if proper consent can be obtained … then this is an amazing resource for those types of studies," explained Rakyan.
"There's no collection that needs to be done," he explained, "unlike any prospective study, where you recruit pregnant moms and then you follow their kids, and so on."
Despite the success that he and his colleagues have had so far in using MeDIP-seq to profile cytosine methylation in DNA from Guthrie cards, Rakyan said the most appropriate method for testing these sorts of samples still depends on the number of samples and the budget available.
In situations where there are thousands of samples to test, for instance, he predicted that sequencing-based strategies would still be too expensive for researchers in smaller labs.
"Right now it's just really large centers or really large consortia that are doing thousands of samples by sequencing," Rakyan said. "As of today, if you want to do thousands of samples, array-based approaches like the Illumina 450K are preferable, because they're cheaper and easier to perform."
On the other hand, he argued that it is far more feasible to consider looking at methylation profiles in 100 or so samples using sequencing-based methods. Again, cost may dictate whether that sequencing analysis takes the form of an enrichment-reliant method such as MeDIP-seq or wholesale bisulfite sequencing of the genome.
In the long term, though, given the limited amount of DNA available in each sample, Rakyan said he would be in favor of doing the most comprehensive analysis possible if newborn methylation profiling were to become more widespread.
"These samples are very interesting, because you can never go back and get more," he said. "These Guthrie cards typically have four spots on them. If you use one of them, that's it, it's gone forever."
"So one thing we would definitely discourage people from doing is looking at one gene or two genes or three genes," Rakyan said. "Forget about that. We need to go whole genome."
From that perspective, even a genome-wide method such as MeDIP-seq may fall short when thinking about doing standardized, methylation profiling at birth, Rakyan said.
In contrast, by using an approach such as whole-genome bisulfite sequencing that looks at every base in the genome, he said, "you're getting the definitive methylation profile. You're not going back five years from now and doing the same experiment using a different method."
Ideally, he speculated that the introduction of nanopore sequencing technologies, which are expected to be cheaper and more powerful than current methods, could make it much more realistic to think about doing whole-genome bisulfite sequencing on many, many samples.
In the meantime, Rakyan said that he and his colleagues have already started to phase out MeDIP-seq in favor of bisulfite sequencing for their research purposes. That move has them again looking for methods that make it possible to profile methylation with lower starting levels of DNA than is standard.
"Sequencing costs have gone down so much that it makes sense," he said. "And for most projects in the lab, we're not looking at thousands and thousands of samples, so for the samples we're looking at we can afford to do it."
Rakyan said his team has already gotten bisulfite sequencing to work with 100 to 200 nanograms of DNA. He also pointed to the possibility of trying their hand at the tagmentation approach that University of Washington researchers described for doing bisulfite sequencing with as little as 10 nanograms of DNA (IS 4/10/2012).
For their part, the authors of the current study are already planning to take a crack at using Guthrie card samples for epigenetic disease studies. On that front, collaborators from Skåne University Hospital, led by co-author Åke Lernmark, have already obtained parental consent to use DNA from Guthrie cards collected for tens of thousands of children in Sweden since the late 1990s, along with blood samples collected annually since birth.
Among the children in the cohort are more than 100 who have now developed type 1 diabetes, Rakyan noted, making it possible to compare methylation patterns before and after the disease developed.
"That's just one disease," he added. "There are other children [in the cohort], unfortunately, who are also suffering from other diseases. This means we can try to understand what role the epigenome or DNA methylation plays, if any, in these conditions."