This article was originally published March 24.
In collaboration with researchers from several academic and clinical research centers, a team from the Massachusetts-based enzyme development company Kapa Biosystems has been delving into potential problems and bottlenecks in preparing next-generation sequencing libraries — particularly those prepared from formalin-fixed, paraffin-embedded (FFPE) samples.
The effort is intended to improve library prep kits and their applications, Maryke Appel, technical director at Kapa, told In Sequence.
The firm is preparing to release a new library prep kit known as Hyper Prep in the next two weeks. While that kit is not specific to FFPE samples, Appel noted that results so far suggest that it shows promise for dealing with the types of low-quality or low-input DNA that often characterize such samples.
The approach also appears to address some of the sequencing adaptor ligation issues that she and her colleagues identified when scrutinizing libraries being constructed from FFPE DNA using the company's existing library prep kit.
In a poster presented at the Advances in Genome Biology and Technology meeting in Marco Island, Fla. in February, researchers from Kapa, Washington University, the Memorial Sloan-Kettering Cancer Center, Case Western Reserve University, and the University of California, Davis described findings from that study.
For the analysis, the group used the Kapa HTP library prep kit to produce sequencing libraries with DNA from multiple sets of FFPE samples and standard tumor cell line controls.
With that library preparation approach, which uses Kapa's high-fidelity DNA polymerase enzyme, somewhere on the order of 20 to 25 percent of input DNA ends up in the library as adaptor-ligated inserts when dealing with DNA from a typical sample, Appel noted. But that usually drops to just 1 to 10 percent when the DNA comes from a sample that's been formalin-fixed and paraffin-embedded.
In an effort to understand such disparity, the researchers used Kapa's quantitative PCR-based Human Genomic Quantification and QC assay to assess DNA from FFPE and cell line samples along the way during the library prep process, particularly after steps to add Illumina sequencing adaptors to input DNA molecules and amplify those adaptor-ligated molecules.
"Most people look at the final product — they look at yield and then they go into sequencing," Appel said. "We started placing a lot of emphasis on looking at what happens after ligation and how much of the input DNA you really can convert to adaptor-ligated library."
In so doing, the researchers determined that the ligation step actually has a more pronounced impact on the quality and yield of DNA in sequencing libraries produced from FFPE DNA than do amplification efficiency.
That finding is somewhat surprising, Appel noted, since many investigators suspected that damaged, degraded, and deaminated bases in DNA from FFPE samples would be especially prone to amplification problems.
On the amplification side, the team did see some inefficiency in the FFPE samples. But that efficiency was not dramatically different than that observed in libraries prepared from a typical DNA inputs.
A more pressing concern, according to the new analysis, is the lack of conversion to adaptor-ligated DNA for many of the input molecules from a given FFPE sample.
"The conversion of input DNA to adapter-ligated library proved to be significantly lower for libraries prepared from FFPE [versus] control DNA," authors of the poster wrote.
"This results in lower library diversity, higher duplication rates, and lower coverage," they continued. "For optimal results with FFPE sample, input and library construction parameters should be optimized to achieve the highest yields of both adapter-ligated and amplified library."
That suggests that a significant proportion of the starting DNA from FFPE may be lost to sequencing, Appel said. But it also highlights the futility of trying to do extra rounds of amplification during later stages of sequencing library prep for such samples.
"Whatever molecules you don't put adaptors on are lost forever, basically. You can't sequence them," she explained. "You can amplify the DNA to make more, but when you amplify you're just creating duplicates."
So while researchers may gravitate toward doing more amplification cycles when input DNA quality is low, results from the study suggest that this may compound rather than solve the problem.
When faced with low-quality samples, Appel argued that the best approach is likely to try to ramp up sample input during library preparation, if possible. Because input DNA is often limited when dealing with FFPE samples, she noted that another option may be to aim for lower DNA outputs, making it possible to dial down amplification cycles so that rarer molecules are not swamped out of the library.
The study did not address the source of the problem at that conversion step, though the team suspects it is related to damage at the ends of DNA molecules that interferes with successful end-repair processes and adaptor ligation.
Appel noted that Kapa has reformulated its end-repair approach and related enzymatic reactions and chemistry for the upcoming Hyper Prep kit, which appears to greatly diminish the ligation bottleneck during low-quality DNA library prep.
In their AGBT poster, she and her colleagues also outlined a quality control step for analyzing input DNA to help predict the success or failure of library preparation. There, they found that input DNA that scored below a particular threshold with Kapa's hgDNA Quantification and QC kit were far more apt to show library prep problems.
The assay uses sequences from a single human DNA locus, present at known copy numbers, to create standard curves for interpreting the concentrations of unknown samples, even at dilute solutions, Appel noted.
But because it uses multiple primer sets to generate amplicons of different lengths, the assay can provide information about DNA quality, she explained. That's because longer amplicons are less likely to be produced as the number of damaged input molecules increases.
"We do the quantification twice. We get an absolute quantification and then we divide the concentration that you get with the long amplicon by the one you get with the short amplicon," Appel said.
"If the DNA is not really damaged then the copy number is the same, so you can amplify the [long amplicon] as efficiently as you can amplify the [short] fragment," she explained. On the other hand, when DNA is damaged, the concentration of the long fragment is lower, leading to a Q-ratio or quality score that's lower than 1.
In particular, findings from the study indicated that when input DNA from FFPE samples has a Q-ratio greater than 0.4, it is generally able to be successfully sequenced in target capture workflows, whereas DNA with Q-ratios below 0.4 may not be worth pursuing.
"There's kind of a cut-off point," Appel said. "If the Q-ratio is less than 0.4, you can expect the sample to fail in some way."
She conceded that an upfront QC step adds time and expense in the research setting and may not be something researchers want to invest in for every sample. Even so, Appel noted that there are instances when that increased knowledge about samples on hand can help in evaluating sequencing output or in pooling samples with similar quality scores.
In CLIA labs, on the other hand, investigators are often more apt to invest in upfront QC before deciding whether to process a given sample. "Our QC assay provides that initial entry point to say, 'If the Q-ratio is less than 0.4, you're taking a chance,'" Appel said.
Authors of the AGBT poster did not look at whether conversion is a bottleneck when preparing sequencing libraries with FFPE DNA using kits from other companies.
In advance of the Hyper Prep launch, though, researchers from Kapa did compare adaptor-ligated DNA conversion and amplification rates for the new kit with the Kapa HTP kit and with library prep kits from other firms, according to Appel, who said both Kapa kits outperformed alternative library prep approaches on those metrics.
She explained that FFPE libraries are generally produced using the same library prep protocols as those for DNA from other sources with slight tweaks to the protocol, in some cases, to try to maximize adaptor ligation and the like.
Similarly, the anticipated Hyper Prep kit is designed to fit into Illumina sequencing pipelines in place for a wide range of applications, Appel said, though it does appear to have advantages when dealing with low-quality or low-input samples such as DNA from FFPE samples and cell-free circulating DNA.
Although Kapa has standard library preparation kits designed to be compatible with Ion Torrent sequencing chemistry, it has not yet developed a version of the Hyper-prep kit for Ion Torrent.
Kapa is continuing to collaborate with researchers from various research and clinical settings, including undisclosed CLIA labs, to continue determining the best balance between quality checks and usable data in various settings.
With respect to FFPE samples specifically, Appel noted that there is more work needed to look at whether there are ways to diminish DNA damage and fragmentation.
She said the FFPE field may also benefit from efforts to improve aspects of the initial tissue retrieval, handling, sample fixation, and DNA isolation processes, though those are not steps that Kapa currently intends to investigate.
"The challenge with FFPE is not just going to be solved on the sample prep level [alone]," she said. "There are a lot of people who want [improvements] right from the beginning of the workflow, from the way tissues are being fixed and treated right from the beginning."