Here's a short summary of what turned into a very long comment. I thought there was a simpler explanation that could also explain the evidence presented, and I have since done some experiments and believe that this alternative is more likely to be true.

In this work, the authors present examples of plant-derived DNA sequences in several independent studies of cell-free DNA. The problem here is that all of these examples suffer from the same shortcoming: namely, that contamination is hard to rule out when sensitive instruments are used to detect parts-per-million signal levels in samples that are already in the low ng/ml range (that is, very dilute).

The authors suggest that contamination can be avoided by good, but standard, laboratory practice and ruled out by independent replication. However, as is well established in forensics and paleogenetics, where dilute DNA samples such as these are much more common, DNA at low concentrations can often be found on 'clean' new equipment, plasticware, and reagents [1-6]. These trace contaminants are typically negligible, but when the intended sample is dilute, they can make up a detectable fraction of the total DNA ultimately used in the experiment. Laboratories that regularly work with dilute samples often take precautions far in excess of those taken here (e.g. UV-treatment of new reagents and positive-pressure rooms [7]), but, even so, allow that trace contaminants are difficult, maybe impossible, to eliminate [8,9].

One way to control for trace contaminants is to simply sequence a sample of purified water. For obvious time and money reasons, people don't typically do this, but in the one case that I could find [10], they recovered at least some apparently plant-derived DNA (see table S15 of that reference). So just how common are these plant DNA trace contaminants, and can they really explain the findings of this paper?

I was concerned about the potentially inflammatory conclusions drawn from these data when such a mundane explanation might serve just as well, so I decided to actually do the experiment myself-- using similar analysis methods to those described here, I looked for plant DNAs in other samples having no plausible connection with food, but sharing with this study a low concentration of DNA. It turned out that I was able to find plant DNAs in these samples as well, appearing at frequencies comparable to, even exceeding, the frequencies described in this paper.

Not only were these DNAs present, but the distribution of DNAs between samples in a single experiment also matched the distributions described here. In one such study, Navin et al [11] separately sequenced the genomes of 100 individual cells from a single tissue, each washed free of the others before preparation. Despite being drawn from the same source, the species of plant DNAs recovered from these samples were heterogeneous (for instance, some showed wheat, whereas others showed lettuce). The analysis I ran also, at least at first glance, appeared to duplicate the statistical properties observed here (fig. 3). Although these results are not immediately intuitive, they would not be surprising with the low concentrations we're dealing with here-- each contaminant might be represented by as little as one DNA molecule per sample, so we shouldn't expect all samples to share the same contaminants.

Finally, they show that while there are many plant-derived reads in maternal plasma, no such reads are detectable in cord blood. However, the proper comparison is between maternal blood and cord blood (full blood has a much higher concentration of DNA than plasma, eliminating the dilute sample issues discussed above). They find only one read in the maternal blood sample, so there is no significant difference between the two samples.

We're currently finishing up a manuscript describing these results in full, with an emphasis on how, especially when working with dilute samples, we need to take into account trace contaminants when analyzing high throughput sequencing data.

Rich Lusk
University of Michigan

1. Elimination of contaminating DNA within polymerase chain reaction reagents: implications for a general approach to detection of uncultured pathogens.
Meier A, Persing DH, Finken M, Böttger EC.
J Clin Microbiol. 1993 Mar;31(3):646-52.

2. Evidence of contamination in PCR laboratory disposables.
Schmidt T, Hummel S, Herrmann B.
Naturwissenschaften. 1995 Sep;82(9):423-31.

3. Residual DNA in thermostable DNA polymerases - a cause of irritation in diagnostic PCR and microarray assays.
Ehricht R, Hotzel H, Sachse K, Slickers P.
Biologicals. 2007 Apr;35(2):145-7.

4. Animal DNA in PCR reagents plagues ancient DNA research.
Leonard JA, Shanks O, Hofreiter M, Kreuz E, Hodges L, Ream W, Wayne RK, Fleuscher RC.
J Archaeol Sci. 2007 Sep;34(9):1361-6.

5. Detection of bacterial DNA in blood samples from febrile patients: underestimated infection or emerging contamination?
Peters RP, Mohammadi T, Vandenbroucke-Grauls CM, Danner SA, van Agtmael MA, Savelkoul PH.
FEMS Immunol Med Microbiol. 2004 Oct 1;42(2):249-53.

6. Contamination of Qiagen DNA extraction kits with Legionella DNA.
Evans GE, Murdoch DR, Anderson TP, Potter HC, George PM, Chambers ST.
J Clin Microbiol. 2003 Jul;41(7):3452-3.

7. George A. Kowalchuk, Jeremy J. Austin, Paul S. Gooding, John R. Stephen, Chapter 12 Valid Recovery of Nucleic Acid Sequence Information from High Contamination Risk Samples – Ancient DNA and Environmental DNA, In: Keith R. Mitchelson, Editor(s), Perspectives in Bioanalysis, Elsevier, 2007, Volume 2, Pages 357-371, ISSN 1871-0069, ISBN 9780444522238, 10.1016/S1871-0069(06)02012-X.

8. An investigation of the rigor of interpretation rules for STRs derived from less than 100 pg of DNA.
Gill P, Whitaker J, Flaxman C, Brown N, Buckleton J.
Forensic Sci Int. 2000 Jul 24;112(1):17-40.

9. An efficient multistrategy DNA decontamination procedure of PCR reagents for hypersensitive PCR applications.
Champlot S, Berthelot C, Pruvost M, Bennett EA, Grange T, Geigl EM.
PLoS One. 2010 Sep 28;5(9).

10. Subglacial lake vostok (antarctica) accretion ice contains a diverse set of sequences from aquatic, marine and sediment-inhabiting bacteria and eukarya.
Shtarkman YM, Koçer ZA, Edgar R, Veerapaneni RS, D'Elia T, Morris PF, Rogers SO.
PLoS One. 2013 Jul 3;8(7):e67221.

11. Tumour evolution inferred by single-cell sequencing.
Navin N, Kendall J, Troge J, Andrews P, Rodgers L, McIndoo J, Cook K, Stepansky A, Levy D, Esposito D, Muthuswamy L, Krasnitz A, McCombie WR, Hicks J, Wigler M.
Nature. 2011 Apr 7;472(7341):90-4.