‘Retraction Watch’ on the Lo-Alter retraction, 26 December 2011

Published by ‘Retraction Watch’, 26 December 2011.

Just days after the retraction of a paper in Science that had claimed a link between chronic fatigue syndrome (CFS) and the virus XMRV, the authors of a similar paper in the Proceedings of the National Academy of Sciences (PNAS) have retracted theirs.

The PNAS paper, “Detection of MLV-related virus gene sequences in blood of patients with chronic fatigue syndrome and healthy blood donors,” was published online on August 23, 2010 by Shyh-Ching Lo, Harvey Alter, and colleagues. Here’s the notice, which PNAS says will be available on its site sometime this week:

The authors wish to note the following: “Although our published findings were reproducible in our laboratory and while there has been no evidence of contamination using sensitive mouse mitochondrial DNA or IAP assays or in testing coded panels, we have the following concerns:

1. The original chronic fatigue syndrome (CFS) patient samples were of insufficient volume to distribute to other laboratories for independent confirmation.
2. Only one (1) of many laboratories has found a similar association between polytropic murine leukemia viruses (pMLV) and CFS and a careful study of 100 CFS patients (2), as well as a coded panel recently constructed by the National Heart, Lung, and Blood Institute (NHLBI) (3), have found no evidence for either xenotropic murine leukemia virus-related virus (XMRV) or pMLVs in CFS patient samples.
3. Our attempts, through collaborations, to demonstrate antibody in affected patients, to isolate the virus by culture, or to show integration sites in the human genome have failed to support the initial findings.
4. While recall of eight patients from the original cohort 15 y later showed pMLV gag sequences in seven, the copy number was very low and phylogenetic analysis showed these sequences were not direct descendents of the original dominant strains (4). Still later samples from four of these patients tested negative in the NHLBI panel. While this result could be explained by viral clearance over time, it fails to support a sustained retroviral infection in human cells.

The notice refers to an ongoing study led by Ian Lipkin:

Although a more definitive, National Institute of Allergy and Infectious Diseases (NIAID)–sponsored, coded panel of samples from 150 well-characterized and geographically diverse CFS patients and controls is being assembled for further study, in consideration of the aggregate data from our own laboratory and that of others, it is our current view that the association of murine gamma retroviruses with CFS has not withstood the test of time or of independent verification and that this association is now tenuous. Therefore, we retract the conclusions in our article.”

We asked Columbia University virologist Vincent Racaniello, who has been following the CFS-XMRV story for years as part of his This Week in Virology podcast, to explain what the original paper found, and what the retraction meant:

…in 2009 Lombardi et al published a Science report indicating they had detected the new retrovirus XMRV – first detected a few years earlier in prostate tumors – in the blood of a high proportion of patients with chronic fatigue syndrome. Many other laboratories tried to reproduce this finding, but none found XMRV in CFS patients.

In 2010 Alter and colleagues reported finding retroviral sequences in blood from a substantial number of CFS patients. No viruses were isolated in this study; only viral sequences were obtained by PCR. These sequences were not XMRV, but rather were closely related to endogenous retroviruses of mice called polytropic murine leukemia viruses. (Polytropic means the viruses can infect many species, including mice; xenotropic means the viruses, though originating in mice, only infect non-mouse species).

As to the retraction notice itself:

Curiously, they begin the retraction by writing that they could not detect contaminating mouse DNA in their samples – which was most certainly present and lead to their detection of MLV-like sequences in the first place. This failure remains puzzling and unexplained; but as they report in the next paragraph, they appear to have run out of material to distribute to other laboratories for ‘independent confirmation’. This is just as well: it’s better not to waste more time on what clearly is a case of contamination.

Lo et al provide three additional reasons why they are retracting this paper. They note that no one has been able to reproduce their findings, including the Blood Working Group as discussed above. They have not been able to find (along with collaborators) anti-XMRV antibody, XMRV virus, or viral integration sites in patient samples. It sounds as though they have done quite a number of experiments with others using their precious samples – so it’s not clear why their collaborators could not also look for contaminating mouse DNA (see previous paragraph). Finally, they mention their finding from the PNAS paper that a second set of samples taken 15 years later from the same CFS patients also were positive for MLV-like viruses. They write that phylogenetic analyses revealed that these sequences were clearly not descendants of the original strains. The sequence data used to make this conclusion were available at the time of the PNAS publication, so it is not clear why this evolutionary incompatibility was not noted at the time.

The authors made the link between their paper and the now-retracted Science paper clear in their original paper’s conclusion:

Although we find evidence of a broader group of MLV-related viruses, rather than just XMRV, in patients with CFS and healthy blood donors, our results clearly support the central argument by Lombardi et al. (3) that MLV-related viruses are associated with CFS and are present in some blood donors.

But that wasn’t the case, says Racianello:

The Lo-Alter finding was viewed by many as supporting the findings of Lombardi et al; but in truth they only confused the matter. The viruses pinpointed by Lo-Alter in CFS patients were not XMRV, and it made no sense that CFS would be caused by such a diverse range of viruses. A second report in 2011 reported MLV-like sequences in a CFS cohort but many other studies failed to find any kind of retrovirus in the blood of CFS patients.

Just this year it became clear that XMRV is a laboratory-generated recombinant virus: it arose during the passage of a prostate tumor in nude mice in the early 1990s. This made it highly unlikely that it could be associated with human disease. Lombardi et al retracted a part of the 2009 Science paper that reported nucleic acid sequence; they noted that their samples were contaminated with XMRV plasmids. What remained of the paper were serological and virus culture experiments that were not specific for XMRV. Last week the remainder of this paper was editorially retracted by Science. What that meant is that there is no longer any confidence that XMRV, or any related retrovirus, causes XMRV.

So what’s happened to the XMRV-CFS hypothesis since Lo and Alter published their paper?

The first blow was the finding in several laboratories that reagents used to carry out PCR are often contaminated with mouse DNA. The presence of this adventitious DNA can lead to detection of MLV-like sequences that resemble those found in the Lo-Alter study. The implication was clear: the Lo-Alter findings were wrong, a result of contamination of PCR reagents with mouse DNA.

The next blow was a report of the Blood XMRV Scientific Working Group, which was assembled to determine if XMRV constituted a threat to the blood supply. In this study, sets of coded samples previously shown to be XMRV positive, as well as samples from healthy controls, were blinded and provided to 9 laboratories for analysis by PCR, virus culture, and serology. Two laboratories reported evidence of XMRV in the coded samples. Only WPI identified positive specimens by PCR: two from negative controls, and one from a CFS patient. The Lo laboratory did not detect any positives by PCR, using the same nested assay that they had previously reported in their PNAS paper. The samples tested included 5 specimens that were positive in the Lo-Alter study. In other words, Lo-Alter could not detect retroviruses in the same samples in which they had previously detected them.

And what’s next? Does the retrovirus-CFS story have a future?

With the retraction of the Lombardi et al and Lo-Alter papers, this brings to an end any hope that there might be a retrovirus associated with CFS. Many (including Lipkin) have suggested that a related human gammaretrovirus might be involved in CFS. But I don’t see how a lab contaminant can point you in the direction of a bona fide etiologic agent. Contaminants just cloud our vision, they don’t improve it.

As for the Lipkin study – the more I think about it, the less compelling it seems. Many laboratories have failed to find any retrovirus in CFS patients. The story is over. Why do we think that the results from one laboratory will clear the matter up further? On the contrary – if the Lipkin study is positive, I would not believe it, in the face of all the other negative results we have seen. As I always say, trust science, not scientists.

The paper, which has been cited 76 times, according to Thomson Scientific’s Web of Knowledge, was subject to a minor correction of GenBank accession numbers.

The timing of the two retractions — bookending a holiday weekend — appears to be coincidental. We asked Science last week whether they had tried to coordinate publication with PNAS, and the journal told us they weren’t aware of the PNAS retraction.

The announcements themselves marked a bit of a switch; typically PNAS has not included retraction notices in its press materials, while Science has, as our sister blog Embargo Watch noted last week. The opposite was true this time. We’ve urged journals that use press releases to to press-release retractions whenever they appear.

ME CONNECT HELPLINE

0344 576 5326

Available every day of the week between these times: 10am - 12noon, 2pm - 4pm and 7pm - 9pm.

Calls cost the same as other standard landline numbers (starting 01 or 02). If you have a call package for your landline or mobile phone then calls will normally come out of your inclusive minutes.