Direct squencing from the minimal number of DNA molecules needed to fill a 454 picotiterplate

Abstract

The large amount of DNA needed to prepare a library in next generation sequencing protocols hinders direct sequencing of small DNA samples. This limitation is usually overcome by the enrichment of such samples with whole genome amplification (WGA), mostly by multiple displacement amplification (MDA) based on φ29 polymerase. However, this technique can be biased by the GC content of the sample and is prone to the development of chimeras as well as contamination during enrichment, which contributes to undesired noise during sequence data analysis, and also hampers the proper functional and/or taxonomic assignments. An alternative to MDA is direct DNA sequencing (DS), which represents the theoretical gold standard in genome sequencing. In this work, we explore the possibility of sequencing the genome of Escherichia coli fs 24 from the minimum number of DNA molecules required for pyrosequencing, according to the notion of one-bead-one-molecule. Using an optimized protocol for DS, we constructed a shotgun library containing the minimum number of DNA molecules needed to fill a selected region of a picotiterplate. We gathered most of the reference genome extension with uniform coverage. We compared the DS method with MDA applied to the same amount of starting DNA. As expected, MDA yielded a sparse and biased read distribution, with a very high amount of unassigned and unspecific DNA amplifications. The optimized DS protocol allows unbiased sequencing to be performed from samples with a very small amount of DNA.

Figure 1. Flowchart of the minimal library preparation protocol.

Panel A: The experimental work started with cell sorting, where 20,000 cells were separated in two replicates to confirm the whole experiment. The DNA from 20,000 cells was extracted and split into halves, where one half was amplified with GenomiPhi (MDA) and a second half was processed without whole genome amplification (DS). The shotgun libraries were prepared with the same alternative protocol for the both MDA and DSsamples. Library quality control points were the test PCR with emPCR primers to prove the removal of self-ligated adaptors and the library concentration checking with qPCR. The MDAsample and DSsample with different MIDs in two repetitions were combined into two sequencing runs as sho! wn in the scheme. Panel B: DNA amount requirements in the standard Rapid Library Preparation Method Manual GS FLX+ Series – XL+ (May 2011) compared with the amounts actually needed for sequencing on a selected PTP region. The minimal amount of prepared library required for proceeding to emPCR step in the standard 454 protocol may lose 99% of input DNA during the library preparation step. Then, this amount is diluted to a working stock of 10–7 molecules, defined as the best starting point to perform the emPCR titration step. However, if the exact number of molecules is quantified with qPCR, the emPCR titration step can be omitted, so actually only 0.13 pg of prepared library are needed for sequencing on 1/8 region of PTP (equivalent to 340,000 ssDNA molecules). This allows to use an alternative shotgun protocol where the DNA losses are reduced.

Figure 2. Results of E. coli genome mapping and blast to NCBI database.

Proportions (in %) of Mbp mapped by SSAHA2 to E. coli genome are shown for MDA and DS sequences, separately for each sequencing run. It can be observed that the percentage of mapped DS reads were significantly higher than the MDA reads. The reads that were not mapped to E. coli were analyzed by blast in “nr” database. However, most reads remained unidentified, especially in the case of MDA.

Figure 3. Distribution of coverage throughout the E. coli genome.

The comparison of the genome coverage obtained by MDA and DS methods. The genome coverage of MDA reads was characterized by unequal distribution with many gaps and several areas with extremely high coverage (up to 121 x), while the highest coverage obtained by DS was only 15 x and it was better distributed throughout the whole genome.

Figure 4. Clustering analysis of the k-mer abundance distribution.

Comparison of the relative abundances of 6-mer in the different datasets using hierarchical clustering. As observed, the most likely conformation shows aggregation of E. coli with DS methodology, while B. subtilis is associated with MDA.