A team led by researchers at the Oslo University Hospital has adapted a previously published native chromatin immunoprecipitation sequencing method to work with very low input — down to 100,000 cells per immunoprecipitation, which represents a 200-fold reduction from existing native ChIP, or N-ChIP, sequencing protocols
The group described its adapted protocol in a paper published online in BMC Genomics last month. According to the researchers, the approach extends the applicability of N-ChIP sequencing to isolated primary cell types and rare cell populations, and could also potentially remove the need for cell culture, which can introduce additional epigenetic marks.
The study's first author, Gregor Gilfillan, told In Sequence in an email that the group chose to base its work on N-ChIP — specifically, a method developed by a National Institutes of Health team and published in Cell in 2007 — rather than cross-linked, or X-ChIP, because his team was interested in looking at histone modifications, for which N-ChIP is suited, and there has been some suggestion in previous publications that native ChIP methods retain more DNA during purification and preparation steps.
"As far as I am aware, nobody has definitively proven this in a side-by-side comparison," Gilfillan wrote. "Neither have we here, but at least it makes sense that purifying DNA under native conditions will be more efficient than if you have to de-crosslink, which is probably not 100 percent efficient, and involves high-temperature incubations that damage your DNA."
The group sought to improve the NIH method because it contained a cumbersome dialysis step, which Gilfillan said the group believed it could simplify. "Many older ChIP protocols are more complex than they need be and by shortening and simplifying where you can, you lose less material, and in the process gain sensitivity," he wrote.
In the team's BMC Genomics study, Gilfillan and his colleagues describe their adapted protocol and share data from tests of the method using progressively smaller inputs. The group used the parameters of the original technique with its published input of 2 x 107 cells per immunoprecipitation as a benchmark to compare the performance of the low-input adaptation. In all, the team tested five different inputs, ranging from the original 2 x 107 to 2 x 104.
The researchers prepared chromatin from cultured CD4+ lymphocytes immunoprecipitated with an anti-H3K4me3 antibody using the original method and sample size, and the adapted method for each input sample size, sequencing each ChIP-seq library on a single lane of an Illumina GAIIx. The researchers found that additional sequencing was necessary to call peaks for the lowest cell-number input that they tested, so the group performed a second set of sequencing for this sample using a single Illumina HiSeq 2000 lane.
Using the performance of the original NIH method as a benchmark, the group measured the sensitivity of its adapted approach for each input sample size. According to the researchers, 85 percent of peaks could still be detected down to 1 x 105 cells. At the next, and final, input size, 1 x 104, the sensitivity fell to 70 percent.
The group used the presence of peaks that were not present using the original high-input method to calculate a measure of specificity. According to the authors, this specificity held steady with the scaling down of inputs, with all datasets having greater than 90 percent of their called peaks "on target" compared to the dataset using the original method.
Overall, the group found that as input size decreased, the number of unmapped reads and mapped reads derived from duplicates increased. The researchers compared a sample of the unmapped reads to the GenBank nucleotide database and found that only a small proportion in all the samples represented sequencing errors that failed to map to the human genome. The rest likely represent PCR amplification artifacts, the group wrote.
To demonstrate that the method could be used for different histones, the researchers also measured the H3K27me3 mark along with H3K4me3 using 100,000 cells per immunoprecipitation in two precipitations. According to the group, the mutually exclusive nature of the two trimethylations was clearly visible in their resulting profiles.
The group also applied the method to lymphocytes from three pairs of human monozygotic twins. Gilfillan said that this application was the group's primary goal in adapting the method and, according to the group, a more complete report on the results of this study will appear in a separate publication.
Gilfillan said the team hopes to study epigenetic differences in discordant human monozygotic twins where one has a disease such as psoriasis but the "identical" sibling does not. Other groups have studied methylation differences between such twins, and the Oslo team hopes to see how histone modifications may correlate with that methylation data.
According to Gilfillan, the team's adapted N-ChIP sequencing protocol could be a complement to other reduced-input epigenetic methods — like reduced representation bisulfite sequencing — in a variety of studies. "You use RRBS to examine methylation, and then turn to ChIP-seq to look at histone modifications. The ChIP-seq is, however, more demanding in the lab, and you need to be more careful in sample collection and storage," he wrote.
While the group's method significantly lowered the input requirements of current N-ChIP sequencing methods — by two orders of magnitude — Gilfillan said that the team believes that alternative sequencing library preparation methods will be necessary to reduce inputs further.
One such alternative library prep approach, nano-ChIP-seq, was recently published by a group from the Broad Institute, which showed it could generate chromatin state maps from as few as 10,000 cultured embryonic stem cells and 25,000 hematopoietic stem cells (IS 10/11/2011).
Gilfillan said that the Broad team's method, which used cross-linked ChIP rather than native ChIP, would be the method of choice for researchers not working with histone proteins. According to Gilfillan, N-ChIP strategies are unsuitable for transcription factors, for example, which have "more labile interactions" with DNA than histone proteins.
While the approach allows them to go lower with cell numbers, "the additional amplification steps … presumably come at a cost of introducing more PCR bias and amplification artifacts. So if you have 100,000 cells to use, and are only looking at histones, I would choose our method to minimize the amplification employed," he wrote.
In the paper, the researchers also wrote that the success of their low-input N-ChIP sequencing method even using standard library preparation techniques "may reflect the reported higher efficiency of N-ChIP relative to X-ChIP."
However, Gilfillan wrote, "I think we reached the limit of ChIP-seq using standard Illumina library prep techniques here, at least for primary CD4+ cells. So going lower is going to require alternative sequencing library preparation methods such as that from the [Broad] group."
"In addition to the method itself, I think the main message from the paper is, 'Watch out for PCR junk reads and pay attention to sensitivity as you scale down.' These are points where I expect a lot of other ChIP-seq researchers will be wasting valuable sequencing capacity as they attempt to scale down to smaller samples," he said.