Why did high A+T content create problems for the Plasmodium falciparum genome project?

Question

The main paper for the Plasmodium palciparum genome project (Gardner et al., 2002) repeatedly mentioned that the unusually high A+T content (~80%) of the genome caused problems. For example they imply that it prevented them using a clone-by-clone approach:

Also, high-quality large insert libraries of (A + T)-rich P. falciparum DNA have never been constructed in Escherichia coli, which ruled out a clone-by-clone sequencing strategy.

And that it made gene annotation difficult:

The origin of many candidate organelle-derived genes could not be conclusively determined, in part due to the problems inherent in analysing genes of very high (A + T) content.

Question:
What is the biological significance of high A+T content, and why would it cause problems in genome sequencing?

Ref:
Gardner, M.J., Hall, N., Fung, E., White, O., Berriman, M., Hyman, R.W., Carlton, J.M., Pain, A., Nelson, K.E., Bowman, S., Paulsen, I.T., James, K., Eisen, J.A., Rutherford, K., et al. (2002) Genome sequence of the human malaria parasite Plasmodium falciparum. Nature. 419 (6906), 498–511.

Answer 1

The sequencing technologies that were developed in the last 20 years have a range of optimal use at an average A+T/G+C rate. Both highly AT-rich and GC-rich regions are complicated to process by the different sequencing technologies. Each technology has different ranges of usage, but to name one, Illumina technology prefers sequences in the middle range. If you try to sequence an AT-rich genome with the Illumina standard protocol, you will sequence an incomplete genome, the fragments of which are not a perfect reflection of the original complete genome. Other technologies claim to be completely unbiased to nucleotide content. Pacific Biosciences is one of them, and people seem to agree on that claim, after having analyzed the data that is produced by their machines. Oxford Nanopore Technologies claims that they have almost no biases, but as of today (2012-06-13), there is no confirmation of that by external analyses.

Beyond sequencing problems, the software used to assemble and annotate the sequences may also be prone to errors in AT-rich and GC-rich regions. But many of those problems stem from the incompleteness of the sequencing.

Answer 2

I can't comment on how A+T richness complicates the sequencing process itself, but I can comment on complications that arise when annotating the sequence. Ab initio gene predictors are often based on hidden Markov models that are very sensitive to base composition in the genome (di-nucleotides, tri-nucleotides, etc). These gene finders typically perform very poorly if they are run on a genome that has a much different base composition than the one on which it was trained. This could explain some of the difficulty they has with analyzing genes in the genome.

Answer 3

Often sequencing involves a step of amplification of genomic material. The standard way to perform this is with PCR, but PCR is biased and does not amplify very AT-rich regions well. With multiple rounds of PCR, even low-abundance regions that are not as AT-rich might come to dominate the sample and hide the AT-rich sequences.

This is not only a problem for de novo sequencing, but for many sequencing-based techniques (RNA-seq, ChIP-seq, your-favorite-seq...). Alternative methods have been employed in plasmodium, but they are not as standard (yet?).

See, for example, H2A.Z Demarcates Intergenic Regions of the Plasmodium falciparum Epigenome That Are Dynamically Marked by H3K9ac and H3K4me3 at http://www.plospathogens.org/article/info:doi/10.1371/journal.ppat.1001223

Answer 4

In the past, before massively parallel sequencing, they made a library of cloned sequences and transformed these into E. coli. High AT sequences are difficult to maintain in E. coli (perhaps due to similarity to promoters?).

Answer 5

A lot has already been said in previous answers so I am just gonna add briefly two potential issues with strong AT/CG bias:

1) Potential for polymerase slippage due to homopolymers: this introduces errors in general because you may have unwanted indels in the reads as well as purely incorrect bases being incorporated. This is a problem that can happen even with PCR (although there's a lot of choices now if u want to spend). So in general higher error rates and higher read failure.

2) Difficulty of the machine to separate the signals of individual nucleotides for SANGER (it gets all blurred) or calibration errors with next gen sequencing. So higher read failure (bad quality).

3) Assuming everything is now fine, still lower complexity regions can be VERY hard to map, let alone assemble a complete genome from scratch.

Hope this helps!