Evaluating the latest high-throughput molecular techniques for the exploration of microbial gut communities

Abstract

The human gut microbiota has become the subject of an increasing amount of attention, due to an emerging understanding of its role in maintaining health throughout our lives. Since only a small proportion of the gut bacteria can be quantified using traditional plate culturing methods, culture-independent approaches are required for determining the structure of complex microbial communities. To avoid cloning bias and low phylotype coverage that affects amplicon cloning and sequencing strategies, high-throughput methods such as phylogenetic arrays and massively parallel sequencing are now being used to find more than just the most abundant taxa, at significantly lower costs and higher speeds. The target for these methods is the 16S ribosomal RNA gene that is present in all prokaryotes. Since the gene is too long to be sequenced using high-throughput methods, regions of high variability (from V1–V9) are selected for amplification and either direct sequencing, or hybridization against phylogenetic microarrays. In our recent study,¹ we compared sequencing of amplified V4 and V6 regions using 454 FLX Pyrosequencing² with the HITChip, an oligonucleotide microarray for taxonomic profiling of human intestinal tract communities based on concatenations of known V1 and V6 regions.³ We found good correlations between the phylogenetic classifications stemming from the two technologies, especially at lower-order ranks (phylum, class, order, and to a lesser extent, family), which indicates high robustness of both approaches. However, the V6 regions proved to be much less suitable for taxonomic classification than the V4 region, probably due to this region simply being too variable. Although this study was, to our knowledge, the deepest sequencing of single gastrointestinal samples reported to date, the microbial richness levels had still not leveled out, with up to 1,800 unique phylotypes detected in one community. Encouragingly for studies with lower sequencing coverage per sample, we also noticed that a fifth of the sequencing depth (40,000 as opposed to 200,000 reads) was sufficient for capturing a majority of microbial diversity within a sample.

Different variable 16S rRNA regions have been used for amplicon pyrosequencing in many compositional studies during the last few years. Furthermore, recent advances in sequencing technologies has yielded nearly a doubling of pyrosequencing read-lengths up to 500 bp (454 Titanium), and a 2.5-fold increase in read number per pico-litre plate. The Illumina sequencing platform today offers paired-end 100 bp reads and more than a 20-fold increase in throughput per plate, compared to a pyrosequencing plate. Leveraged by this technology, we are currently extending the published study to compare six variable tandem regions sequenced from one single gut, using the improved 454 and Illumina technologies. For both the longer Titanium reads, and for the shorter but more numerous Illumina reads, a preliminary analysis of the data indicates an increased level of resolution in terms of numbers of genera detected, with about 80% more genera detected using Illumina reads. Preliminary taxonomic analysis also suggests significant variation in overall classification patterns achieved depending on the choice of variable regions, or sequencing technology; we saw differences in phylum distributions of up to 25 percentage points both when comparing regions using the same technologies, as well as when comparing technologies using the same region. These differences could be due to either natural variations in classification robustness afforded by the different regions, or due to primer biases. In either case, this inherent variation may be a factor for comparing published studies based on different rRNA gene regions. We are currently investigating the cause of these variations.

In silico simulations, conducted both by us and by other groups,^4–6 showed that the V4 region was one of the more robust variable 16S rRNA regions, in terms of giving reliable taxonomic classifications. Together with our finding that 40,000 reads are sufficient for capturing a large part of the natural diversity in a gut community,¹ it gave us confidence to perform a large-scale compositional study of hundreds of Irish elderly subjects (http://eldermet.ucc.ie), at three different time-points. Interestingly, we see a strong trend of intra-individual variance being much smaller than the inter-individual variance, and we are now correlating the variations in microbiota between subjects with an array of clinical data for each subject.

Addendum to: Claesson MJ, O'Sullivan O, Wang Q, Nikkilä J, Marchesi JR, Smidt H, de Vos WM, Ross RP, O'Toole PW. Comparative analysis of pyrosequencing and a phylogenetic microarray for exploring microbial community structures in the human distal intestine. PLoS One. 2009 Aug 20;4(8):e6669. doi: 10.1371/journal.pone.0006669.