The team used the device to capture a single cell of a “candidate division” called OP11 from an anaerobic sulfide and sulfur-rich spring in southwestern Oklahoma called Zodletone Spring. There was nothing particularly special about OP11 a priori, Blainey says, because nothing was known about it; the organism was sequenced to fill in a phylogenetic gap.
Yet the resulting sequence provided more than just As, Cs, Gs, and Ts, according to Blainey; it provided glimpses of how OP11 lives —oxidizing organic molecules for energy, just as humans do, and breaking down complex polymers, which humans cannot. “We are actually quite excited about the data we got back—not in the sense that it encodes genes for turning lead in to gold, but that we got information about enough of the right genes to start outlining the nature of the OP11 cell we isolated.”Single cell SNPs
The other major issue with MDA is amplification bias; researchers using MDA consistently observe uneven amplification across the genome. According to Blainey, genome completeness in single-cell genome projects can run the gamut from 0 to 95%; Woyke says her team typically recovers between 20% and 80%.
Such uneven reads tend to confound genome assemblers, which are designed to anticipate relatively even sequence coverage. “This is absolutely not true at all for single cell data and really fouls up the assemblies,” notes Blainey.
Patrick Chain, team lead for metagenomics and next-gen sequencing applications at the Los Alamos National Laboratory, along with LANL colleague Cliff Han, has developed a process to normalize genome coverage by inducing artificial polyploidy—that is, multiple genome copies — by using cell division inhibitors to produce cells with 2, 4, 8, or more genomes per cell. The result, he says, is “better genome coverage and a more normalized distribution of genome coverage after sequencing.”
Meanwhile a group led by Sunney Xie, professor of chemistry and chemical biology at Harvard University, has developed an alternative approach to genome amplification. Called MALBAC, the technique employs a linear thermal cycling amplification step prior to PCR amplification, rather than MDA.
“That solves the bias problem,” Xie says — so much so, in fact, that his team has amplified the DNA from a single human cell with 93% coverage at 30x sequencing depth. The team could even call a single SNP variant based on their data. “If one base is different, we can call it.”Metagenomics meets single cell genomics
Single-cell genomics and metagenomics represent two sides of the same coin — understanding the biology of organisms that cannot be grown in the lab. Metagenomics techniques can reveal the genetic potential of a community, but not the players. Single-cell approaches can close that gap, for instance by providing scaffolds for assembling metagenomics data or reference genomes for variation studies. As a result, for many researchers the two technologies are complementary.
According to Woyke, JGI received 12 single cell genomics proposals during a recent call for proposals, 10 of which combine metagenomics and single cell approaches and account for more than 400 single-cell genomes in total.
“It's not like you should only do one or the other, they inform each other,” says Philip Hugenholtz, director of the Australian Center for Ecogenomics in Queensland.
In one 2009 study (3), for instance, Stepanauskas and his team (including Woyke) isolated two individual “uncultured, proteorhodopsin-containing marine flavobacteria” from the Gulf of Maine, collecting 1.9 Mb and 1.5 Mb of genomic DNA representing an estimated 91% and 78% genome recovery, respectively.
The team used those genome assemblies as scaffolds to “recruit” individual reads from the Venter's GOS to map where in the world's oceans those organisms reside. “In theory we could do the same with single-cell sequencing alone,” Stepanauskas concedes. “But it would be much more expensive.”Single cell viromics
Of course, to get a really complete picture of a microbial community, researchers must study more than just its microbes; there also are the viruses that prey upon them.
It has been estimated there are 10 viruses for every bacterial species, and as Lasken points out, “we don't even know how many bacterial species there are.”
Recently Lasken, along with former JCVI colleague Shannon Williamson, published a proof-of-principle study addressing “single virus genomics” (4). The team mixed two known bacteriophages (T4 and lambda), sorted them into individual “cooled agarose beads” on a microscope slide, and used those as templates for MDA and sequencing. The team showed they could read a single lambda phage at 437x coverage, capturing all but the first five bases of the virus’ genome.
Now, the Lasken team is trying to adopt the technology to environmental samples. But it won't be easy. Viruses represent an even bigger challenge than microbes at the single particle level, as viruses by definition cannot replicate on their own. Researchers trying to interpret viral genomes must therefore figure out not only what their genomes encode, but also which organisms they infect.
Still, says Lasken, “It is fair to say that single viral particle sequencing would solve a very difficult problem of how to get access to this enormous number of viruses in the environment.”
As with most everything in biology, such sequences will be only the beginning. After all, even well-studied of organisms have their secrets. Says Hugenholtz, “There are still genes in E. coli that they haven't worked out what the function is, despite having the genome for almost 20 years.”