In the 23 December 2004 issue of Nature, Jingdong Tian et al. describe a new method for "Accurate multiplex gene synthesis from programmable DNA microchips." The name most frequently associated with the paper is that of George Church, a professor at Harvard Medical School.
The authors combine microfluidics, biochemistry, and molecular biology to produce a widget capable of rapid synthesis of long oligonucleotides (oligos). The paper reports an integration of 1) a new way to elute completed oligos from arrays; 2) on chip amplification of oligos; 3) error correction using via "strict hybridization" conditions to remove mistakes; and 4) microfluidic multiplexing, to produce 14.5 kilobase (KB) long fragments of DNA. Slipped in at the end of the paper is the claim that they have already used this technology to successfully fabricate 95-382 KB oligos, assembling them into megabase (MB) length sequences. Although it may receive less press, when the paper comes out describing the latter advance it will mark a significant milestone in the human ability to manipulate biological systems. Organismal length sequences will be well within reach.
Now for the press coverage of the paper. Mr. Wade, in the 12 January 2005 edition of The New York Times, describes it thus;
Researchers have made an unexpectedly sudden advance in synthesizing long molecules of DNA, bringing them closer to the goal of redesigning genes and programming cells to make pharmaceuticals.
But the success also puts within reach the manufacture of small genomes, such as those of viruses and perhaps certain bacteria. Some biologists fear that the technique might be used to make the genome of the smallpox virus, one of the few pathogens that cannot easily be collected from the wild.
With all respect to George Church and his colleagues, and without reducing the significance of their technical achievement, I have to say this actually isn't so much of a surprise. It is true that I have been following this, and that I saw the chip on Erdogan Gulari's desk last winter. In other words, I have had time to get over it. But this sort of thing has been in the air for a while, and Drew Endy and I talked about something similar many years ago at tMSI. I am certain we were not the first to do so.
More interesting is the reduction in cost per base of the synthesis, which Professor Gulari puts at about a penny a base for the long oligos. This is news, and the cost falls completely off the curves I published in 2002. The impact of the paper will only be felt when the technology becomes widely available, which is at least a couple of years out. Unless I misunderstand the market and the state of the technology, the only people with access to synthesis at this scale and cost are the authors of the paper and their pals in academia.
With respect to suggestions that oligo synthesizers should be regulated, my views are well known at this point. In the NY Times piece, Professor Church suggests registration of instruments could go a long way towards increasing security. More information is, of course, better. But we have too much experience forcing people "underground" when the things they want to pursue are restricted or made illegal. I suspect we will be much better off encouraging an open community of people unafraid to talk about what they are up to in their garage. Finally, even if instrument makers are willing to going along with registration, there will be a big hole in the registry due to the aftermarket, and I don't know how to enforce registration of homemade DNA synthesizers. There are arguments that no one will want to build a synthesizer, or to play with what it enables, but I think the history of tinkering is a fairly decisive counterexample. So the real question is, how do you stop people from playing? I don't think you can.
As an advance in the technology, far more interesting to me is a paper by Peter Carr et al, from the Jacobson group at MIT, "Protein-mediated error correction for de novo DNA synthesis". They use the DNA mismatch-binding protein MutS to identify mistakes, which are then removed from the synthesis pool. One round of this procedure improves the error rate to ~1 in 4000 bases, which is a factor of three better than the Tian et al work discussed above. A second round of error correction reduces the error rate to ~1 in 10 KB. This rate is so low that a single round of synthesis and cloning should be sufficient to produce multi-gene cassettes suitable for use in complicated genetic circuits. The combination of the protein-mediated correction and the Tian et al work would be impressive indeed. Since George Church is thanked in the acknowledgments of the Carr paper, no doubt all the right people are considering the possibilities.