This is my latest e mail to Mr Tuller
Absence of evidence in Paprotka et al. (2011).
Papotka et al. (2011) argue that XMRV was created during the creation of the 22Rv1 cell line from the genomic recombination of the genomic RNA of two ERVs they call PreXMRV-1 and PreXMRV-2 co-packaged into the same virion.
Paprotka led by Dr John Coffin used subjective labels to describe these viruses. We will use slightly less subjective labels. Hereafter the ERVs will be called ERV-1 and ERV-2. XMRV will be given the objective label XMRV/VP62.XmU3f and GAGr primers
"To quantify the amount of XMRV DNA in the CWR22 xenografts, we developed a real-time PCR primer-probe set that specifically detected XMRV env and excluded murine endogenous proviruses present in BALB/c and NIH3T3 genomic DNA (Fig. 1C). We used quantitative PCR of 22Rv1 DNA to estimate 20 proviruses/cell and used the 22Rv1 DNA to generate a standard curve. The CWR22 xenografts had significantly fewer copies of XMRV env (<1–3 copies/100 cells) compared to the 22Rv1 cells (2000 copies/100 cells). The CWR-R1 cell line had ~3000 copies/100 cells, and the NU/NU and Hsd nude mice, thought to have been used to passage the CWR22 xenograft, had 58 and 68 copies/100 cells, respectively.”
· The term "XMRV specific" is misleading. The primer probe set (primer 3f-8r) was complimentary to sequences in the env region of XMRV/VP62 and ERV-1. The term XMRV encompasses a wider range of variability than is allowed for here.
· The real-time PCR actually established a figure of 68 copies per 100 cells for ERV-1 and 3 copies of ERV-1 per 100 cells in the CWR22 cell line. There is no history of this primer being able to detect XMRV/VP62 env sequences at a concentration of less than 2000 copies per 100 cells. The serial dilution method used here has come in for a great deal of criticism, and the copy number estimates in mice and the early CWR22 xenografts are highly likely to be unreliable.
· Primer 3f-8r using a single round PCR was unable to detect XMRV/VP62 in the later xenograft.
The following information is erroneous:
“The absence of XMRV in the CWR22 tumor and early passage xenografts. Using qPCR assays, we estimated that the early xenografts contain ~1‐3 XMRV env copies/100 cells (Fig. 1C), which correlated with the amount of mouse DNA in the early xenografts (0.3 ‐1%; Fig. 1D), and the estimated 1 XMRV env copy/cell in the NU/NU and Hsd nude mice (Fig. 1C).”
· This is incorrect from the initial paragraph. The provirus detected was ERV-1 and not XMRV. There is no information on copy number of XMRVenv below 2000 copies per 100 cells. The proviral copy number was at most 68 copies per 100 cells and not 100 copies per 100 cells above. The authors are assuming that the assay could have detected XMRVenv, if present, even though this primer could not detect XMRV in later xenografts when used with the single round PCR.
The following is particularly troubling.
“All six of these PCR primer sets had 100% identity to the published XMRV sequences, and could amplify XMRV from control infected cells as well as PreXMRV‐1.”
The early xenographs were not screened with XmU3f and GAGr.XmU3f GAGr is the primer set that could only amplify XMRV GAG. This primer was not able to amplify XMRV from later xenografts either, using the single round PCR, yet this primer was the one used to determine the analytical sensitivity of the single round PCR assay. Analytical sensitivity for a PCR assay using the same reagents and cycling conditions is primer specific. Vary the primers and the analytical and clinical sensitivities are very likely to change.
The results in Figure 2E for XMRV are not produced using primers that can 'only' amplify XMRV (XmU3f and GAGr), because that primer could not amplify XMRV from later xenografts either.If one looks closely at figure E row 1 it is clear that the primers used were not the ones that could only amplify XMRV gag but primers that could amplify XMRV vp-62 sequences and erv1.Therefore both rows I and 2 should be labeled XMRV and prexmrv1. The later xenografts were not examined for the presence of XMRV using the quantitative PCR with the 3f-8r primers. Thus we do not know whether XMRV could be detected using that assay.Variables governing a PCR assay
There are a number of variables that govern whether a PCR assay can detect a target template in nucleic acid extracted from a biological sample, and the primer sequence is but one of them.
· Absolute and relative concentrations of reagents in the reaction tube.
· Choice of annealing times and temperatures.
· Concentrations of oligonucleotides, primers and magnesium, salt, buffers.
· Concentration of the target template.
· Quality of DNA or RNA in the biological sample.
· Presence of inhibitors in the biological sample or as a result of the nucleic acid extraction process.
All the above are crucial variables
Therefore the assumption that just because a PCR reaction can amplify pristine DNA or copy DNA from serial dilutions in a lab, then it can do so from nucleic acid extracted from a biological sample, is unsafe. Indeed quantitative real-time PCR is very susceptible to the presence of inhibitors, which reduces the amplification efficiency of the PCR reaction.
The calculation of copy numbers using the software used to analyse the results of quantitative real-time PCR assays is based on the assumption that the amplification efficiency of PCR is the same in all reactions considered.
The results from Paprotka (2011) confirm the importance of the concentrations of reagents in the reaction tube because the qPCR reaction using one master-mix using the 3f and 8r primer was able to detect erv1 in NU/NU and Hsd mice, but the same primer with a different master-mix in the single round PCR was quite unable to do so.
We also have the situation where this primer was able to detect ERV-1 in the early xenografts, where the copy number of ERV-1 was much less. This apparent paradox strongly suggests the presence of inhibitors in the DNA extracted from the mice, which affects this very short primer sequence differentially as other longer primers were able to detect Erv 1 sequences in these mice.
There is no information regarding the sensitivity of the quantitative real-time PCR assay using the XMRV only GAG primer used to screen the lab and wild derived mice hence the results have little meaning. These wild derived mice were described as wild in the paper and this is incorrect. ERV-1 was not isolated as a whole provirus from NU/NU or Hsd mice.
Paprotka claim that their assay would have been sensitive enough to detect XMRV in prostate cancer tissue at the concentrations found by Schlaberg et al.(3) (2009) in one patient using IHC, of 1 provirus per 660 cells. The assay constructed in Schalberg et al. used different primers, different cycling conditions and had a theoretical limit of detection of 1 XMRV copy per PCR reaction tube. This is far more sensitive than the assay used in Paprotka et al. (2011), even if the lower limit was 1 copy per 100 cells. Yet this assay only detected XMRV in 25% of the people found positive using the PCR assay.
Analytical sensitivity determined by serial dilutions of pristine DNA in a lab is no measure of a PCR reactions ability to detect a target sequence extracted from a biological sample as the experience of Schlaberg et al. so graphically demonstrates.
Scientific orthodoxy would demand that the xenografts were screened using assays known to be far more sensitive than the single round assay used in Paprotka, such as the nested PCR or nested Reverse transcriptase PCR, or indeed the quantitative real-time PCR devised by Schalberg et al. These have a history of being able to detect XMRV in prostate tissue, while the assays used in Paprotka do not.
Relying on probabilistic arguments and ignoring the issues inherent in detecting target sequences from biological samples is unsafe. History shows that the ability or otherwise of a PCR assay to detect XMRV in prostate tissue depends on choice of primers, cycling conditions and the master mix. These parameters need to be optimized in order to maximize the chances of detecting XMRV if present. Current evidence would also suggest that IHC is a far more sensitive technique for detecting XMRV in prostate tissue than PCR. It is therefore puzzling why it was not used here.Poisson distribution
As copy number falls the number of replicates needed to detect XMRV in a sample isolated from DNA, taken from a biological sample, rises dramatically according to the dictates of the Poisson distribution. The apparent detection of EndoERV-1, but not XMRV, in the early xenographs may well result from an inadequate number of replication PCR runs being undertaken. This would seem reasonable, as the early xenografts directly ancestral to the 22Rv1 cell line were found not to contain erv-2 (Figure 2E, Paprotka).
It is difficult to propose that XMRV was a result of an astronomically unlikely recombination event between two endogenous proviruses if one of the proviruses was not present in the xenografts in the passage immediately before the recombination event allegedly took place (see figure 1A). The experiment certainly needs to be repeated using the statistically appropriate number of replications.
One must now turn to the probability of such a recombination event taking place at all. The authors assume this is fact and base their calculations accordingly. They also assume extramolecular and intramolecular recombination.
"(we) assume that the number of crossovers is distributed according to a Poisson distribution"
"Assuming that crossovers can only occur in the 111 blocks of identity"
"To assign a probability of observing a given pattern we assume that the selection of template for initiation of DNA synthesis is random, the selection for the acceptor template of minus‐strand DNA transfer is random, the identified ≥ 20‐nt identity blocks share the same probability of recombination, and recombination events are independent, i.e. recombination at one block does not affect the probability of a recombination event at any other block. Given these assumptions, the probability of observing a given pattern"
"It is well established that DNA synthesis can initiate from either RNA template, and minus‐strand DNA transfer can occur both inter‐ and intramolecularly. Hence, we hold these two variables constant, assume that the number of crossovers is distributed according to a Poisson distribution, and examine the effect of recombination frequency."
"For example, using the average of 4 crossovers described in the literature, the probability of observing a second independently derived provirus with the same 6 crossovers is 1.3 × 10‐12."
(The above five quotes from reference 2)
The entire argument is based on a series of assumptions, which may or may not hold. They are also assuming that their hypothesis, that XMRV was created by their proposed crossover, is fact. This has not been established, as alternative explanations for their observations are available however unlikely the authors deem them to be. This means that the likelihood of this event happening in the first place is 1.3 × 10-12.
Thus we have two competing explanations. The first is that the PCR assay had insufficient clinical sensitivity to detect XMRV in low copy number for all the reasons discussed above or that XMRV was created by a series of crossovers with a likelihood of occurrence of 1.3 × 10‐12.
The accuracy of these competing hypotheses should be tested experimentally and not determined using probabilistic arguments based on the existing preconceptions of the authors. A number of points presented as factual in the paper are clearly not, and a number of diagrams and assertions are misleading. At the very least they need correction via the standard procedures.ERV-1, ERV-2 and the mice.
The mice tested in Paprotka were NOT from the same supplier as the mice used for xenografting to create the 22Rv1 cell line. The nude mice used to create the cell line were obtained from the Athymic Animal Facility of the Case Western, whereas the mice tested in Paprotka were from,
“…Taconic (NCR nude), Harlan Laboratories (Hsd nude), Charles River Laboratories (BALB/c nude, NIH‐III nude, NIH‐Swiss, and NU/NU nude), and Jackson Laboratory…”
The nude mice Paprotka et al. claim are likely to have been used for in vivo passages of the xenograft (NU/NU and Hsd) are also different to those where ERV-1 and ERV-2 were shown to be integrated in Cingoz (2011). ERV-1 was found integrated in C57L/J and ERV-2 was shown integrated into DBA2J and 129X1/SvJ. All of these mice are hairy and could not have been used to create the 22Rv1 cell line.
Only a small region of ERV-2 was found in Hsd mice from Harlan Sprague Dawley laboratory and no trace of EndoERV-2 was found in the NU/NU mice from the Charles River laboratory (Fig. S3B and S3C, Paprotka).
ERV-1 was also incomplete in NU/NU mice (Fig S6A and S6A, Paprotka).
Hsd is also not a specific strain of mice but the Harlan Sprague Dawley laboratory. Therefore it is not known which strain or strains of mice were tested in Paprotka under the heading Hsd.
· Mice tested in Paprotka from a different supplier as those used to create the 22Rv1 cell line.
· ERV-1 and ERV-2 integrated into none nude strains not used to create 22Rv1 cell line.
· No trace of ERV-2 found in NU/NU mice.
· Incomplete ERV-1 found in NU/NU mice.
· Only a fraction of ERV-2 found in Hsd mice.
There is no evidence that EndoERV-2 (PreXMRV-2) was present in the cells or the mice used for xenografting to create 22Rv1 cells. Thus, XMRV/VP62 cannot be said to have been created during the construction of the 22Rv1 cell line.
1). Paprotka et al: Recombinant Origin of the Retrovirus XMRV: http://www.sciencemag.org/content/early/...ce.1205292
2). Paprotka et al: Supporting Online Material for Recombinant Origin of the Retrovirus XMRV www.sciencemag.org/cgi/content/full/science.1205292/DC1
3). Schlaberg R, Choe DJ, Brown KR, Thaker HM, Singh IR: XMRV is present in malignant prostatic epithelium and is associated with prostate cancer, especially high-grade tumors. Proc Natl Acad Sci U S A 2009, 106:16351-16356.