Partial retractions — as opposed corrections or the full monty — are unusual events in scientific publishing. But they appear to come in twos.
The journal Infection and Immunity, the work of whose editor, Ferric Fang, is much admired by this blog, has a fascinating example of the breed in its February issue.
The article in question, by a group from the University of Kentucky in Lexington led by Susan Straley, appeared online in 2007. It was titled “yadBC of Yersinia pestis, a New Virulence Determinant for Bubonic Plague,” and, as the words suggest, involved a gene marker for the virulence of plague. Or so it initially seemed.
But according to the partial retraction, the researchers are walking back one of their main claims. Consider:
Volume 76, no. 2, p. 578–587, 2008. Page 578, Abstract, lines 7 and 8: We retract the last sentence of the Abstract.
That sentence read:
We found that loss of yadBC caused a modest loss of invasiveness for epithelioid cells and a large decrease in virulence for bubonic plague but not for pneumonic plague in mice.
The notice continues:
Page 578, column 2, lines 13 to 18: We retract the last sentence of the Introduction.
That sentence read:
We evaluated the transcriptional and translational expression of these proteins, their possible function as adhesins and invasins for epithelioid cells, and their importance for lethality in bubonic plague and pneumonic plague, and we showed that they are new Oca family members that are essential for lethality of bubonic plague.
The authors also are pulling
Pages 585, Results, column 1, lines 1 to 8 from the bottom: We retract the last three sentences of the paragraph entitled “Tests for importance of yadBC in bubonic plague and pneumonic plague in mice.”
Which reads:
Likewise, an intranasal dose of 5,400 F1 yadC strain Y. pestis CO92.S18 cells was 100% lethal (four mice were tested). The reconstituted yadBC strain CO92.S16 killed six of seven mice given an intranasal dose of 1,800 bacteria, which again was consistent with full virulence. These findings indicated that YadB and YadC are not required for virulence in pneumonic plague, whether F1 is present or not.
And, finally, the authors have removed:
Page 585, Discussion, column 2, lines 19 and 20 from the bottom: We retract the phrase “severely compromised.”
Which stated:
Interestingly, our LD50 test showed that deletion of yadBC had no significant effect on virulence when the intranasal route of infection was used, whereas the virulence of the yadBC mutant was severely compromised when the skin route was used for challenge. These findings add YadB and YadC to the growing list of properties, such as the capsule F1, that play a role in bubonic plague but are not required for lethality in pneumonic plague (5, 10), and they suggest that the initial niches inhabited by the bacteria in the two diseases are very different and determine the subsequent course of infection.
What follows is a lengthy explanation by the researchers about the reason for the partial retraction. We leave it to our readers whether that explanation is adequate:
In recent studies, Tanya Myers-Morales, Annette Uittenbogaard, Amanda A. Gorman, and Susan C. Straley have discovered that the yadBC mutant used for measuring virulence by the intradermal (i.d.) route became corrupted at the last step of its creation (transformation of the Lcr virulence plasmid), with the revised conclusion that yadBC is not a major virulence determinant for the skin infection. The original result, described in the second paragraph under “Tests for importance of yadBC in bubonic plague and pneumonic plague in mice” at the end of the Results section, is therefore retracted. Statements repeating the conclusion drawn from those results therefore also are incorrect: the last sentences of the Abstract and Introduction and the phrase “severely compromised” in the Discussion. The results and conclusions in the rest of the paper are not affected, because the strains used were Lcr− or were fully virulent.
The original ΔyadBC mutant had been made independently in two backgrounds: Pgm+ Lcr− (avirulent) and Pgm− Lcr+ (conditionally virulent from an intravenous route). Then, working in biosafety level 3 (BSL3) containment with select-agent security, we had reconstituted potential virulence of the Pgm+ Lcr− ΔyadBC mutant by restoring the Lcr plasmid. During this last step, that ΔyadBC strain evidently acquired a virulence-abolishing mutation that did not show up in our in-vitro phenotypic characterization assays. That strain was avirulent by the i.d. route, as described in our published article. Meanwhile, a complemented ΔyadBC/BC+ strain (with restoration of yadBC in the native location) did not acquire a virulence-abolishing mutation when we restored the Lcr plasmid, and it was virulent by the i.d. route. Hence, we reported that yadBC is required for lethality of bubonic plague. However, when we recently found that the Pgm− Lcr+ΔyadBC strain that we had made in parallel at the outset was fully virulent in the systemic plague model, we became concerned that something may have been wrong with the ΔyadBC strain that we had reconstituted to virulence in BSL3. It was important to know if yadBC is actually required for virulence in bubonic plague and perhaps not in systemic plague or whether it is not required for virulence in either case. Therefore, we derived a Δpgm derivative of the Pgm+ Lcr+ΔyadBC strain that had been avirulent by the i.d. route, tested its virulence in systemic plague, and found that it also was avirulent in that model. We now knew that that strain was corrupted. So we restored the Lcr plasmid again to the Pgm+Lcr− ΔyadBC strain. This time it did not acquire a virulence-abolishing mutation and was only slightly attenuated by the i.d. route (two of six mice given a dose of 71 CFU died and two of two mice given a dose of 524 CFU died). This result shows that the effect of yadBC in bubonic plague is subtle or redundant. We feel that the further lethality studies required to determine a true 50% lethal dose value are not justified, because many mice would be required to quantify a relatively small attenuation.
We tried to reach Straley, whose group holds a patent for the “compositions and methods for treating a Yersinia pestis (Y. pestis) infection” but haven’t heard back. But Fang tells us that he felt a partial retraction in this case was the best way to handle the article, which has been cited 22 times, according to Thomson Scientific’s Web of Knowledge:
I think that partial retractions are only appropriate under exceptional circumstances, but this was one of those circumstances. The decision was in keeping with ASM Journals policies.
The partially-retracted publication concerns two genes (yadB and yadC) of the plague bacillus that were reported to be required for virulence. I was contacted by the senior author after she determined that the mutant strain used in the virulence experiment was corrupted, and follow-up work failed to confirm a robust virulence phenotype. This invalidated the central conclusion of the paper, but did not affect all of the other data contained in the paper, which did not involve the problematic mutant strain. I was reluctant to simply issue an Authors’ Correction, as ASM Journals policy states that these should be limited to “errors of a scientific nature that do not alter the overall basic results or conclusions of a published article.” However a complete Retraction would have called the entire article into question, whereas in this case it was quite straightforward to specify which data were in error and which were still believed to be valid. Therefore the author and I agreed that a Partial Retraction was most appropriate. Retractions are a rather crude tool, and I believe it is an important responsibility of an Editor to try to use them in a way that best serves both the authors and scientific community. I hope we achieved that in this case.
Sounds like the data should of been split between two papers, if one particular subset of the data is so independent from the rest that it really can be filleted out without affecting the rest of the data and conclusions.
What’s the opposite of salami-slicing…too many results spoiling the broth..overstuffing the minestrone?
I’ve seen other instances like this pop up involving the use of transgenic animals. In that case, what is called a ‘germline deletion’ had occurred – where during the ongoing breeding of animals a mutation is introduced which is inherited by all subsequent offspring. The mutation just happened to delete the leptin gene, producing a massively obese phenotype which the authors obviously (mistakenly) put down to the transgene (unaware of the germline deletion). It was only through a massive effort on the part of another research unit which published a paper questioning the findings that the germline deletion was discovered.
These cases highlight the problematic aspects of working with transgenic animals – the research unit where I work uses them a lot, and it’s commonly said that to properly characterize a transgenic animal, i.e. just to examine whether the genetic manipulations are working as planned, is about a year’s work. This is hardly ever done in the rush to publish.
Furthermore, I’ve even been in a seminar where a research unit had gone through this painful process of characterizing an animal, only to discover the transgene was not performing as planned, and then to have an audience member say that they used to work with the people who created that transgenic line, and they themselves had done this work and concluded years ago that it was a poorly performing transgenic line. In other words, the people giving the seminar had to completely re-do all that work at their own expense simply because none of that information was publicly available! What we really need is a ‘transgenic animal’ equivalent of http://www.antibodypedia.com or something similar.