Contamination Redux

Contamination has as many heads as hydra. Unfortunately, all the recent publications on this subject w.r.t. XMRV haven't done much more than reiterate doubts expressed less clearly in earlier papers. Let me systematically consider each of the multiple meanings of contamination, so far as I have been able to assign fixed meanings.

Meaning one:

Huber found her laboratory was contaminated by mouse cells from previous work. This was said to show results in Lombardi et al. were the result of such contamination. The catch here was that the rooms in question had (deliberately) never been used for work on mouse samples, and tested clean by the same test for mouse mitochondrial DNA used to show Huber's contamination. Mitochondrial DNA is many times as common as nuclear DNA in cells. In any case, it would take a foolhardy scientist to tell Sandra Ruscetti she was naive about contamination by mouse cells.

Meaning two:

Dr. Stoye has just shown that you can find contamination by fragments of mouse DNA in some laboratory reagents if you analyze the DNA content of water samples without calibrating your assay against positive controls. (This is like turning the gain on a radio all the way up when it is not tuned to a station, and announcing the discovery of noise.) The catch here is that the Lombardi/Mikovits group did test the batches of reagents for contamination before using them, rather than attempt to untangle a mess afterward, ala Huber. (They even told other researchers which brands of reagents were safest, not that anyone paid attention.) They also had positive controls which limited the required sensitivity.

Meaning three:

We now have the shocking discovery by Sfanos et al. that many prostate cell lines can be contaminated. (Dr. Miller has shown that 22Rv1 is contaminated by a virus resembling XMRV, though identity of the contaminant to XRMV may not be completely established. This line was not used in the published research.) Cell lines are created to allow investigation of pathological processes. If a virus causes prostate cancer, it should be expected to show up in these cultures. If the cultures cannot support the virus, it is all too likely the cell line will be of little use for research on cancers caused by virus. The implication here is that shoddy laboratory work allowed indicator cell lines to become contaminated -- and nobody noticed.

The problem, once again, is that the specific cell lines used by Lombardi et al. were tested for contamination repeatedly. They have been tested since publication. Even if this does not rule out contamination, the use of negative controls appears adequate to distinguish positive results from contamination by cell lines even if those cell lines are contaminated.

Meaning four? recombination origin:

Crucial parts of denier arguments on recombination depend on failure to detect virus at critical points. At other times the authors are convinced their assays are 100% effective. This is an unsubstantiated claim. Calibration with V62 plasmids is a long way from detecting latent provirus integrated into nuclear DNA. There is a straightforward way around this, even without assuming any patients are infected, which no one has bothered to attempt. (I'll come back to this later.)

We have not heard a peep about the effect of hypermutation by APOBEC3G on inserted provirus in any study by deniers since an original mention by Paprotka. Some provirus does appear to have the characteristic G -> A substitutions. We do not know which assays are able to handle this. This is not a case where you can get by with sloppy reasoning rescued by the sloppy nature of biology. There are good reasons to believe the vast majority of copies are latent, not active.

What is becoming clear is that a single retrovirus is showing two distinct behaviors. As an RNA virus in an active infection, it is unstable and easily hypermutated. Most copies seen in plasma will not be identical to reference strains. Once provirus is inserted in the nucleus it benefits from cellular mechanisms to protect and repair the cell's own DNA; the rate of mutation drops dramatically. This does not mean it is free from changes which took place prior to insertion. (Nor does this mean it is immune to damage by reactive oxygen species (ROS). Most diseases do not affect mitochondria directly, do not produce ROS, and do not activate DNA damage repair pathways. There is some evidence this disease does all three.)

The resilience of this virus to hypermutation means that it is possible to get (some) replication-competent virions from (some) hypermutated provirus. This will not be identical to reference strains. However, if you protect it, and allow it to continue its carefree existence as an active infection for a period of time, sooner or later it will produce sequences closely resembling reference strains. During the period before this takes place, you will have no detectable virus. When it does reappear, you can exclaim "contamination!" This can take place in the best of laboratories, even Dr. Rein's.

The way to convince me you really can detect integrated provirus despite hypermutation, etc. is to separate a blood sample into parts, infect one, allow active infection to proceed for a time, force active viral infection into latency, transfer a few cells with latent provirus to the uninfected part, and detect them. (You can split spiked samples a second time to allow culturing which will show that spiking was successful.) I'm betting the problem will turn out to be much harder than previously imagined.

This isn't the way virologists have been operating for that last 50 years. During that time there have been a few regrettable oversights, HIV-1 and HTLV-1 among them. The Lombardi/Mikovits paper did not spell out all these points because they were actually doing original research, not simply going through the motions. Some discoveries have been made since that time. This is the nature of research.

If you only allow those with complete explanations to perform research and publish, you can pretty nearly eliminate fundamentally-new discoveries. In the case of human retroviruses, virologists have almost painted themselves into a corner where no undiscovered viruses will ever be discovered unless they are introduced to the population de novo. Someone needs to reconsider the tacit assumptions implicit in current practice.

There is a lot above to digest. In a previous iteration of this post, I stopped at about this point. What follows is a change in perspective. The sociology of science becomes significant.

Various forms of contamination in biological research are one of those indelicate facts of life everyone knows, but would rather not mention in public discourse. Most battles with contamination take place in private and are never published.

Establishing the source of a particular instance is generally done by similarity to known contaminants. All too often the route of contamination is never identified. This has unpleasant implications researchers don't even like to think about.

If contamination is as ubiquitous and sinister a problem as sometimes claimed, and most instances are not published, it is safe to assume the vast majority of published work might be weakened or invalidated by undetected contamination. This is offset by a belief that certain laboratories, or certain researchers, are immune to the problem. Inside information indicates this is not entirely true.

If you are trying to avoid contamination, it is important to study good laboratory practice. The fact that it still happens in the best of laboratories reveals a second aspect. Social networks play an important role in deciding who gets smeared with accusations in public.

Professional reactions to public disclosure of the recombination origin hypothesis included the revealing comment that this should have been taken directly to Dr. Silverman, so he could participate in a joint paper. This is basically saying you don't do this to respectable researchers, and it was still possible to separate Silverman from the controversy over ME/CFS and XMRV, which is clearly not respectable.

The proposed recombination origin also obeys the important principle of conservation of preconceptions. As said above, current opinions very nearly rule out the discovery of any human retrovirus unless it is entirely novel. Showing a laboratory origin, which seems reasonable if you avoid some implications, relieves everyone of the need to rethink opinions formed a decade or more back.

The transition from peaceful bliss to desperate struggle with contamination does not always follow any simple pattern. I am going to propose one. Contamination problems are commonly detected shortly after a laboratory starts working with samples derived from actual patients. There are many other factors which have changed over the last 30 years, but that one is constant. It is part of the reason this specific field has a reputation as a graveyard of careers.


I agree, excellent post anciendaze!

Someone needs to reconsider the tacit assumptions implicit in current practice.
Given what we've seen transpire over the past nearly 2 years, including clear manipulation of the scientific process as well as attempts to isolate and viciously "disprove" the WPI's work, at what point do innocently flawed assumptions begin to unmistakeably resemble deliberate attempts to discredit and bury under the guise of a scientific form of plausible deniability?

Blog entry information

Last update

More entries in User Blogs

More entries from anciendaze