At times the peer-review process can be very frustrating.  It is probably wiser not to publish this but I feel like I have to speak out about this somehow---if for no other reason than to illustrate part of the publication (and also grant review) process.  Our two recent publications were in review at various journals for years.  It is painful to see other publications come out that are similar to what you have been trying to publish and to see so much time going by on the calendar during the process.  The peer review process is necessary, and on the balance helps improve the quality of scientific publications.  I can honestly say that some of my publications have greatly benefited from reviewer feedback.  However, this is not always the case.  I think removing the authors names and institutions from the material the reviewer sees, among other steps, would improve the process.  At any rate, I have saved the following review as an example of what we went through to get our papers published (this one is from a previous submission of the engineering underdominance manuscript).  This reviewer followed our manuscript submission from one journal to another (the same person was picked repeatedly by the editors).  We could tell this because some of the reviews contained identical word-for-word copy and paste sections despite our revisions to try to address and clarify them.  I have added the emphasis but the words are taken directly from the review.  If you read through to the end you can see that the reviewer made up speculative scenarios that were not in our paper as a basis to reject it.

August 21, 2012

Unfortunately, the authors have failed to generate the underdominance-based gene drive system they claim to have created, and thus the paper must be rejected. In fact, their text is written in such away that the fact that they have not in fact created a functioning gene drive system only becomes clear when one reads the methods.

In brief, their strategy is to use RNAi, with the dsRNA generated targeting a ribosomal protein, and being expressed under the control of a gal4 uas element. Flies carry this piece of DNA and in addition a rescuing version of the ribosomal protein that is resistant to the dsRNA-dependent RNAi. Flies carrying this chromosome (the third chromosome) are argued as showing gene drive (below). But of course this is nonsense. This chromosome does nothing at all in a wildtype population. Absolutely nothing at all. It does not drive. It does nothing but float in a wildtype population.

The only context in which the transgene-bearing third chromosome does show drive involves a trick, hidden in the methods. The authors take their rnai-rescue line and introduce into this genetic background a transgene on the second chromosome bearing an actin promoter driven version of GAL4, which drives transcription from UAS elements. Thus, in this new strain, acting gal4/Cyo balancer; uas rnai-rescue, we now get effects. The actin-GAL4-driven dsRNA knocks down the expression of the ribosomal protein; in the presence of one copy of the rescue construct (in heterozygotes) they are less fit, and show a lower viability. Very importantly for the authors technique, they are not dead. Otherwise they would never have been able to cross them with each other to generate homozygotes. In homozygotes for the third (The second chromosome balancer prevents GAL4 from ever becoming homozygous), the animals are more healthy than heterozygotes (but of course only heterozygotes that carry a gal4-bearing second chromosome), but less fit than wildtype.

This genetic behavior allows them to show a form of underdominance - in an extremely contrived and artificial situation - that fairly boggles the mind. They take their actin-gal4/cyo; rnai-rescue flies and cross them to their "wildtype" strain, which is balanced on the second chromosome for actin-gal4/cyo but wildtype for the third chromosome. In other words, their wildtype strain always contains actin-gal4, but not in their actual construct. In the text the authors say that the use of gal4 is not essential for their system and is only meant to be illustrative. But this is nonsense. They simply cannot generate drive without the presence of a gal4 that is introduced into their genetic background.

Not surprisingly, in the above, contrived situation the authors can show underdominant behavior. When the frequency of the rnai-rescue chromosome reaches some frequency that depends on the relative fitness costs associated with being heterozygous and homozygous for the chromosome (but only in the presence of a actin-gal4, the benefits of being homozygous for the rnai-rescue chromosome outweigh the benefits of carrying a wildtype chromosome (which at this point finds itself in unfit heterozygotes most of the time) and the uas-rnai-rescue chromosome drives to allele fixation. Below this threshold the chromosome disappears from the population.

Again, the paper must be rejected because they have simply not created a driving underdominant chromosome that can drive itself into a wildtype population. It is remarkable that the authors believe that they can get away without generating a simple three gene construct in a model organism such as drosophila, that would show drive.

In thinking about why the authors might not have done such an obvious and essential experiment one can only conclude that they probably tried but were unable to generate the relevant transformant. There is probably a very good reason for this, and it relates to the reason why no underdominant gene drive systems have ever been created using modern genetic engineering techniques: It is very hard (if not impossible) to generate the initial unfit heterozygote, which will ultimately give rise to the more fit homozygote. Unfortunately, without a demonstrated solution to this problem all engineered underdominance systems are so much hot air. They look good on paper, but in fact cannot be generated in actual living organisms.

For the record, our plan from the beginning was to try different expression patterns of the RNAi "poison" using the GAL4/UAS system to see what expression pattern resulted in the best underdominance properties.  The first expression driver we tested, act5c-GAL4, worked so well that we immediately wrote up and submitted our results for publication.  To get a single locus system the act5c-GAL4; UAS-RNAi could be substituted with act5c-RNAi to remove the GAL4-UAS middleman.  However, this seemed so trivial we didn't consider it important in reporting our first results.  The hemizygous (single insert) individuals are viable and fertile and in fact we used them in crosses to generate homozygous individuals with an alternatively marked GFP system.  Anyway, tested objective scientific facts seem to fall secondary to dogmatic subjective opinions (see above), even in the scientific literature.  How much, from other studies, is modern research missing out on?

The reviewers seem to have wanted us to repeat the experiments in Drosophila with a single insert.  However, (a fact of modern science) we have moved on to new job positions in different labs and Drosophila was only used as a convenience to arrive at, in a timely manner, some definitive results.  We want to move quickly to our real target, mosquitoes and other species.  In the longer view it is a waste of time to back up and try to make the reviewer happy (which is obviously a lost cause from the start).  But, unfortunately, reviewers like this are all too common (I am restraining myself from providing additional examples) and delay or prevent getting published and getting necessary funding to continue the research.