Getting to Know Our Guests: Characterizing the Reproductive Microbiome by Next Generation Sequencing

A report by Ram et al. [2011] in this edition of SBiRM offers some insight into the challenge of characterizing the microflora that each of us carries around mostly unnoticed on or within us. In an obstetrical context, many millions of commensal bacteria and other organisms populate the lower genital tract (LGT). Although cervical mucus maintains a sterile environment above, perturbations of the vaginal microflora can permit pathological species to gain a foothold and compromise upper genital tract sterility. Chlamydia is the main etiological factor responsible for the rise in sexually transmitted disease among the young and infection with this and other pathological organisms including Gonococcus, Ureaplasma, Mycoplasma, Candida, Gardnerella, and Trichomonas provoke immune responses that can lead to inflammatory and frequently permanent damage [Larsen and Hwang 2010]. Moreover, poorly understood changes in the LGT flora associated with vaginosis may be connected to recurrent miscarriage [McGregor and French 2000] and preterm labor has long been thought to have a bacteriological component [Srinivasan et al. 2009]. However, our knowledge of the composition and diversity of the healthy LGT's microflora and how it might help protect this special environment from potentially invasive pathogens is still rather poor [Verstraelen et al. 2009].

Hitherto, our limited understanding of human microflora populations has come from conventional culture using complex or defined media. These traditional methods rely on easily established biochemical requirements for growth and well proven indicators of identity (such as the Enterotube™). They are, however, limited to those (mainly or potentially) pathological species that are amenable to external culture. To offset this limitation, culture-free molecular assays have been developed more recently that depend on the amplification and detection of species-specific DNA (or RNA) sequences. PCR and L (ligase) CR tests for Chlamydia, for example are now routinely used for screening and diagnosis. The tests often target sequences encoding the organism's outer membrane protein [Watson et al. 2002]. However, these tests rely on prior knowledge of the specific target sequence, which inevitably depends on having sufficient DNA for sequencing at least once in the pipeline for assay development. This reliance is impractical for species that cannot easily be cultured and is too labor intensive and slow to allow better characterization of complex populations.

Hence, following on from these target-directed sequence amplification approaches, strategies relying on sequence motifs of prokaryotic 16S rRNA genes that are sufficiently conserved to allow PCR amplification of 16S DNA from a wide range of bacterial species are now preferred [Baker et al. 2003; Hyman et al. 2005]. Variable sequences within the 16S amplicons can then be used to distinguish between genera and or species in these heterogenous DNA mixes and even estimate relative abundance by counting [Sontakke et al. 2009]. Coupled with the advent and rapid development of powerful, deep sequencing systems (DSS), 16S DNA offers the best chance yet for investigating the complexity of the bacterial flora of the LGT and just as importantly, doing it cheaply and accurately enough to allow detailed comparisons to be made. Obtaining raw sequence is one thing, but for assigning that sequence correctly, a sufficiently long or contiguous sequence read (300 bp or more) must be obtained that includes genera and preferably species-specific information. Of the three working DSS platforms, only Roche 454 can generate sufficiently long runs (∼400 bases) for unambiguous calls. The Illumina sequencing platform as used in the report by Ram et al. (2011) can normally generate only short (typically 50–150 plus bp) reads with somewhat lower base call accuracy; but with millions of short reads available, these disadvantages are eliminated by a combination of massively parallel sequencing and statistical cleaning.

Normally, the Illumina and ABI SOLiD protocols require the production of libraries from the DNA of interest for sequencing; but as a first step, Ram's group showed that unusually complex primers containing the required anchor (adaptor), index and crucially, 16S-specific bases can directly generate 16S amplicons for sequencing. The adapter sequence anneals to the solid substrate for bridge PCR as usual, but sequencing the single-stranded templates uses multiple 16S-specific primers that anneal to complementary sequences along their length. Paired-end sequencing (sequencing from both ends) using this novel primer multiplex approach should generate >300 bp of 16S sequence with overlapping reads for quality control and with sufficient variable sequence for detailed microbial phylogeny (provided sensitivity and accuracy can be maintained). Moreover, the sample multiplexing potential of the Illumina platform can considerably reduce the cost of sequencing provided the complexity of the sampled microflora does not unduly compromise read accuracy.

There are, however, some challenges that remain. Ram et al. (2011) found the standard Illumina analysis pipeline somewhat wanting after finding that it was having difficulty correctly assigning image colors to bases. In the Illumina system, sequence complexity helps ‘train’ the recognition software in each sequencing cycle and the sequences analyzed here were mostly identical so poor trainers. To provide a solution, a special deconvolution algorithm was used that forced referencing of all 16S runs to the initial phiX control cycles where accurate reads were obtained. This may be a useful fall back for a pilot laboratory test but will not be sufficient in practice. For some unknown reason, Ram was unable to prime from the opposite end of the amplicons although the primers used were obviously able to amplify 16S DNA as expected. The authors suggest that their use of a bespoke exchange of the adaptors to opposite ends of their amplicons, which allowed them to test their forward and reverse primers without resorting to paired-end sequencing, may have been responsible. This was presumably considered as a means to reduce cost. Had this worked, the sequencing limitations of the platform would have been demonstrably overcome. If the reverse priming and software problems can be solved, the Illumina platform could be used with confidence for this type of study and would be much more cost effective than either 454 or SOLiD.

In the wider world of phylogeny, however, proof of principle experiments such as those described by Ram et al. will need considerable refinement before the platform can be adopted more widely. Despite its high cost compared with Illumina IGA sequencing, the sixteen NIH funded teams undertaking the recently launched human microbiome demonstration projects [Peterson et al. 2009] appear to have settled on the Roche 454 platform. Nevertheless, there is always room for new approaches in the study of complex biomasses and the strategy described here promises to make such studies available to all without compromising quality.