Methodological Approaches Used to Assess Fish Gastrointestinal Communities - Aquaculture Nutrition: Gut Health, Probiotics and Prebiotics - page 112

(e.g. TOPO TA cloning kit, Invitrogen; pGEM-T, Promega). The benefit of such an approach is
the culture-independent nature, and libraries of near full-length 16S rDNA sequences improve
the accuracy of species identification. Ward et al. (2009) demonstrated with the use of clone
libraries that the gut microbiota of the omnivorous yellowbelly rockcod/bullhead notothen
( Notothenia coriiceps ) exhibited greater diversity than that of the carnivorous blackfin/Scotia
Arc icefish ( Chaenocephalusaceratus ). However, many researchers have suggested that meth-
ods based on 16S rDNA gene sequences using universal primers may not accurately reflect the
true underlying diversity of a given environment (Suzuki and Giovannoni 1996; Marchesi etal .
1998) and a large number of clones are required to provide reliable quantitative information.
In addition, technical challenges such as PCR bias, varying ribosomal DNA copy numbers,
and the efficiency of DNA extraction procedures all have the potential to significantly skew
abundance estimates; therefore, assumption of a direct relationship between the number of
sequences of a particular type in a clone library and the number of organisms in the environ-
ment may be inaccurate (Marchesi et al. 1998; Suzuki and Giovannoni 1996). In addition, the
number of clones required (e.g. hundreds to thousands per sample) to obtain good coverage of
gut microbial communities makes clone library preparations impractical.
5.2.5 Next-generation sequencing (NGS)
As described within this chapter, there are significant methodological challenges in survey-
ing the tremendous and dynamic diversity of bacterial taxa within the fish intestine. By one
estimate of human intestinal bacterial community richness, the number of unique bacterial
phylotypes exceeds 1200 (Rajilic-Stojanovic et al . 2007). Bar-coding methods (e.g. DGGE,
TTGE and RISA) have utility in simultaneous comparisons of multiple samples, yet the num-
ber of unique bands typically observed from a DGGE gel is orders of magnitude less than
the extant intestinal diversity. While culture-independent methods avoid the most significant
bias of laboratory cultivation, the biases intrinsic to PCR and bar-coding methods neglect less
abundant bacterial taxa that may play critical roles in gut nutrition and disease.
Advances in DNA sequencing technology now allow massively parallel sequencing of PCR
amplicons or genomic samples, providing sufficient coverage to include less abundant gut
microorganisms. Each sequencing technology has its own set of biases and limitations. For
amplicon sequencing the PCR biases (e.g. preferential amplification of abundant templates)
are also present. In addition, some critical parameters to consider for NGS applications are:
(1) the length of the sequencing read, (2) the numbers of sequencing reads, (3) sequencing
errors, (4) the ability to include multiple samples, and (5) the cost per sample. The specific
sequencing technology adopted will influence each of these parameters. For example, sequence
reads with an average length of ∼450 base pairs (bp) can be generated with a GS FLX sys-
tem (454 Life Sciences, Branford, CT), whereas sequence reads with an average length of
∼98 bp can be generated with a Genome Analyzer IIx (Illumina, Inc., San Diego, CA). As
NGS technology is evolving rapidly and more companies are in this field (e.g. ABI SOLiD,
Pacific Biosciences, Ion Torrent), clearly the number of reads, read lengths, and read accu-
racy will change over time. Sequence read length impacts phylogenetic resolution depending
on the targeted domain of the 16S rRNA molecule (Youssef et al. 2009) and gene annotation
accuracy (Wommack et al. 2008). Clearly longer sequence read lengths are preferable for 16S
rRNA gene surveys, although by targeting specific phylogenetically informative domains (i.e.
the V3 region of 16S rRNA) a shorter read length can still be used to achieve comparable
results to nearly full-length sequences (Kim et al. 2011). The number of sequence reads is