After preparation of the library (and careful quantitation) and preparation of the amplified template comes the main event: the sequencing itself. While there are several methods available, the methods can be divided into three broad divisions.
The three divisions are (firstly) Pyrophosphate Release (named for the original patent by Mostafa Ronaghi and others in 1998 when he was a graduate student at the Royal Institute of Technology in Stockholm); this is the method that uses individual nucleotide flow across all templates, and then detects the signal. (Pyrosequencing – now owned by QIAGEN – detects pyrophosphate, but is not a ‘next-generation’ sequencer as it is not massively parallel; however Roche / 454 FLX used essentially the same method.) Jonathan Rothberg, who parallelized the FLX pyrosequencing method at Curagen, simply changed the detection method with Ion Torrent.
The second division is reversible terminator chemistry, developed by Solexa (now owned by Illumina), and the basic approach was also used by the single-molecule company Helicos Biosciences. (Helicos, while still a publicly traded company, is for all intents and purposes defunct as a vendor now.)
The third division is sequencing by ligation, invented by George Church and commercialized by Life Technologies as SOLiD and now the 5500xl, and also appears in modified forms as the Dover Polonator and the Max-Seq by AZCO Biotech.
The second and third approaches will be covered with separate posts.
Pyrophosphate release was a new approach to sequence detection. Sanger sequencing (my prior post here) uses fluorescence-labeled dideoxy-terminators and detects the terminal base with a direct detection of the fluor. Pyrosequencing detects the by-product of the incorporation of a particular base (G, A, T or G) by an enzymatic cascade that detects the presence of PPi (also called diphosphate), by combining the free PPi with adenosine 5′ phosphosulfate in the presence of an enzyme ATP sulfyrase, producing an ATP molecule. (This is the same ATP molecule that is the ‘energy currency’ within living cells.)
At this point a second reaction occurs, thus the process is called an enzyme cascade: the ATP molecule serves as fuel for a luciferase enzyme to convert luciferin to oxyluciferin, and in the process of this conversion visible light is produced. (Luciferase is the source of bioluminescence, of which the most famous example is firefly luciferase. Being something of a newcomer to the East Coast, I can testify to the fact that fireflies are amazing creatures, and makes being a kid in summer a lot of fun.)
The visible light produced from the enzyme cascade is then detected via a camera, and the residual ATP molecules and nucleotides are degraded by an apyrase enzyme.
So the process after you get templated beads with 10′s of thousands or 100′s of thousands of identical library molecules on it is as follows: apply a universal sequencing primer, load the instrument with millions to 100′s of millions of the templated bead preparation, and add a polymerase (which will bind tightly to the free 3′ hydroxy group of the sequencing primer).
Add a single NTP (say GTP) across all the wells of beads, and detect the signal from the individual beads that have a corresponding C base. (About 25 percent of all the wells should ‘fire’, if the four bases GATC are randomly represented.)
If there are two ‘C’ bases present (called a homopolymer, in this case a two-base homopolymer), there will be twice the signal; if there are three bases, three times the signal etc.
Thus the discriminatory power of the system will degrade the longer the homopolymer stretch, as it is straightforward to detect a single signal versus no signal at all, but much more difficult to detect say the difference between a stretch of 12 and a stretch of 13 in the detection system.
The Ion Torrent technology’s advantage over the Roche / 454 FLX system is its inherent scalability (Roche has maximized the real estate of the bead / camera system, and cannot scale it further other than increasing its read-length, which it recently has to about 700 bases long). However the Ion Torrent technology inherits the basic fact that the majority of its errors are homopolymer in nature. Not that the error is terribly worse that its competitors (there’s quite a marketing war over accuracy right now, so I’m careful in choosing my words), but that the type of error is different due to the difference in approach.
The Ion Torrent technology differs from what has been described above in that it does away with the enzymatic cascade altogether, detecting another by-product of the base incorporation, the production of a hydronium ion (H3O+). So instead of detecting pyrophosphate (PPi) it detects the release of the hydronium ion, shifting the pH of an unbuffered system so that microelectronics can detect it via an ion-sensing field effect transistor, in a massively parallel way.
Yet the basic approach remains – first invented in the mid-1990′s – and this approach is known by its pyrophosphate-releasing roots. Here’s a YouTube video animating the Pyrosequencing process.
Stay tuned for the second and third approaches.