A Massachusetts General Hospital and Harvard Medical School-led team has developed a kit-free method for preparing barcoded DNA libraries for Illumina sequencing from picogram-scale amounts of DNA, such as those present after some chromatin immunoprecipitation experiments.
The group described its approach in BMC Genomics earlier this month, illustrating that it was possible to prepare and sequence Illumina libraries using 100 or so picograms of Drosophila embryo DNA that remained following chromatin immunoprecipitation experiments to snag sequences associated with a particular histone mark — ChIP-seq experiments similar to those that the researchers had in mind when they began developing the method.
"I work with really small samples of primary tissue," first author Sarah Bowman, a molecular biology and genetics researcher affiliated with Massachusetts General Hospital and Harvard Medical School, told In Sequence. "And I knew that when I was able to harvest those cells and do experiments on them, I wasn't going to be able to get a lot of DNA out because I'm not starting with very many cells."
"I needed to have a method that worked better for creating sequencing libraries from these really small amounts of DNA," Bowman said. "I wanted to have something simple that would work with small amounts of DNA and also barcode it for cost efficiency."
Indeed, when they sequenced fruit fly libraries prepared from miniscule DNA amounts, she and her colleagues got ChIP-seq profiles that were on par with those previously reported by members of the modENCODE consortium, who used more standard DNA inputs.
That, in turn, has bolstered the researchers' confidence in applying the approach more routinely. For her part, Bowman is continuing to apply the sample prep protocol to small amounts of fruit fly material, often harvested using cell-sorting techniques. Other researchers working in the same lab have started using roughly the same sample prep strategy for a range of sequencing experiments on mouse and mammalian cells.
Several situations or applications are prone to lower-than-optimal levels of available DNA for sequencing, the study's authors noted.
Given the research questions she's interested in, for instance, Bowman often finds herself looking at subsets of cells within Drosophila embryos that yield relatively little DNA — even before sub-dividing that pool of genetic material further with steps such as chromatin immunoprecipitation.
Several groups have published schemes for preparing sequencing libraries with small quantities of DNA, including protocols specifically designed to be compatible with ChIP-seq experiments (IS 10/11/2011). And just last spring, Rubicon Genomics launched its ThruPlex sample prep kits for dealing with DNA inputs as small as 50 picograms.
When Bowman first began looking at sample prep protocols for dealing with small DNA inputs a few years ago, though, she found that some were incompatible with ChIP-seq (owing to a lack of DNA fragmentation during library prep), while others were complex or time-consuming.
Some relied on an extra, pre-amplification step running through the night, for example, and others used in vitro transcription to bolster DNA levels she noted, and options for doing multiplexed sequencing on such miniscule samples were limited.
"When we began the project, the only protocols out there for working with small amounts of DNA did not have barcoding yet," Bowman said. "And they also seemed really complicated. Since making a sequencing library is basically just making lots of copies of DNA fragments, they seemed almost needlessly complicated to me."
In the interest of finding a relatively simple, straight-forward, and time-efficient strategy for doing barcoded Illumina sequencing on very small samples, the researchers decided to try their hand at developing a kit-independent method based roughly on Illumina's TruSeq sample prep method, incorporating their own tweaks and modifications into the general protocol.
The group began considering modifications to the method in around 2009, looking not only at existing Illumina sample preparation protocols but also at published strategies for streamlining or improving parts of the sample prep workflow.
The resulting sample prep scheme, developed over several years, involves what Bowman called "really modest tweaks" to the Illumina sample preparation protocols. "It just [involved] taking each step and trying to think a lot about maintaining yield at every step and perhaps using improved enzymes to help things out," she said.
For instance, the end polishing and A-tailing steps done after DNA fragmentation remain roughly the same. To nip in the amount of DNA needed for library prep, though, the team uses universal rather than indexed adapters so that multiple barcodes can be tacked on during the amplification step, if desired.
"Ligation of universal adapters to DNA fragments creates products that are extended by PCR to produce barcoded samples containing the identical sequences used for Illumina TruSeq multiplexing," the study authors explained. "These oligos produce libraries that are compatible with conventional data analysis pipelines."
The group's modified method also foregoes the usual selection step and uses quantitative PCR to monitor the amplification step to cut it off at the reaction plateau.
On the size selection side, the researchers reasoned that they could probably get away with eliminating gel-based size selection since smaller chunks of DNA are preferentially copied during the amplification step, leading to a built-in form of size selection.
"As long as your starting DNA is well-fragmented and pretty well in the range of what you want to end up with, it doesn't seem like that's necessary," Bowman said. "And size selection is where you can really lose a lot of yield."
By actively monitoring the amplification stage of the sample preparation process, meanwhile, the group is able to keep tabs on the reaction and ensure efficiency by stopping the reaction when it reaches a plateau while guarding against over-cycling, which can introduce artifacts.
With typical sample preparation strategies, there are very few points in the protocol where it's possible to peek at the sample and see if things are working as anticipated, Bowman noted, especially when the sample is very small from the get-go.
By monitoring the amplification step, though, "you should be able to see a pretty predictable upswing in the amount of DNA in the tube," she explained. "If you follow the amplification reaction in real time, you can see that happening. And that's been really helpful for us to know at that point: whether things seem to be working well or not."
To look more carefully at the performance of their library prep protocol, the team took a crack at using it on samples coming out of Drosophila embryo ChIP-seq experiments targeting histone 3 trimethylated lysine 27, or H3K27me3, a repressive histone mark known for turning up in specific regions of the fruit fly genome.
Those sample prep validation experiments also dovetailed with the researchers' ongoing efforts to isolate specific Drosophila embryo segments and discern their histone methylation profiles.
"This was both a test to see whether we had the right segments and a test of the whole protocol — from cell-sorting out of the Drosophila embryo, to preparing the chromatin from a tiny number of sorted cells and doing ChIP-seq on a tiny number of cells," Bowman said.
The team used its protocol to prepare between 50- and 150 picograms of DNA coming out of two such ChIP-seq experiments before sequencing the libraries by single-end sequencing with the Illumina HiSeq 2000.
When they compared the resulting H3K27me3 profiles with data published by modENCODE researchers, who did comparable experiments using around 5- to 50 nanograms of Drosophila DNA, authors of the new analysis found histone enrichment in similar stretches of the fruit fly genome.
"While the biological samples and chromatin fragmentation were not identical," they noted, "we found enrichment of similar genomic regions at multiple levels of scale."
"Furthermore, a genome-wide assessment demonstrates strong overlap between enriched regions in the picogram- and nanogram-scale experiments, and good reproducibility of the results from the picogram samples."
The picogram library preparation method is designed to be compatible with either single- or paired-end Illumina sequencing. In its current form, the protocol takes around the same amount of time as sample prep using Illumina kits, Bowman said.
Those involved believe the library prep approach will find favor with other researchers interested in doing ChIP-seq on small samples or specific sub-populations of cells. With a few tweaks here and there, they noted that the protocol should also be useful for those pursuing a range of other sequencing applications that hinge on smaller-than-usual DNA inputs.
"If you have a protocol where you're starting with a small amount of material and you yield even less — or you have a protocol where you can start with a large amount of material but the region you're interested in, or the technique that you use, only yields a tiny amount of DNA — that would be an appropriate place to use this protocol," Bowman said.
Representatives from Illumina were not available to comment on whether the company might consider incorporating any of the DNA library prep tweaks outlined in the new paper — or in other studies by independent researchers pursuing specific Illumina sequencing applications — into its own sample prep kits down the road.