Six months ago a paper appeared on the Science Express pre-publication site of the prestigious journal Science. It came from a group of NASA-funded researchers, accompanied by the full NASA publicity hoopla, but it was harshly criticized by other researchers, with almost all agreeing that it was so seriously flawed that it should never have been published. The paper has finally been formally published in the print edition of Science, along with eight highly critical Technical Comments and a response to these from the authors (all available here). So this post is an attempt to pull all the scientific issues together. The accompanying post by Marie-Claire Shanahan explores how these disagreements have played out on blogs and in the larger media and what lessons there might be for science education.


Finding life elsewhere in the universe is a mainstay of science fiction, and searching for it is probably the most exciting aspect of the US space program. We have yet to visit another planet, but NASA’s Astrobiology Program is getting ready by examining the constraints that affect life on Earth. Thus they generously fund research into the origin of life, the strategies of organisms that live in extreme environments, and, most relevant to this work, the search for a 'shadow biosphere'.

The shadow biosphere hypothesis posits that life on Earth includes organisms unrelated to the ones we know about. All of these known organisms (plants, animals, protists, bacteria and archaea) have so many features in common that we're sure they descended from one common ancestor, as shown in the colored part of the figure left.

We think that the first ancestor of this known biosphere arose quite soon after conditions on the early Earth became cool enough to support life. This speedy origin of known life tells us that the origin of life is not stupendously improbable, and thus it not only may have happened elsewhere in the universe, but may have happened more than once on Earth.

The conventional view is that any independent lineages of life went extinct without leaving fossil structures or molecules that we would recognize as different from the known biosphere. But some visionaries have proposed that members of one or more unrelated lineages might survive as a ‘shadow biosphere’ (the grey lineages in the figure), unrecognized perhaps because they superficially resemble members of the known biosphere, or because they live in environments we haven’t yet examined, or because they are so different from us that we haven’t even recognized that they are alive.

Finding evidence of a shadow biosphere on Earth would be almost as exciting as finding life on another planet, but what to look for?

Most researchers now think that the ancestor of the known biosphere was a RNA-like molecule, but in principle life could originate with anything that is subject to natural selection – any molecule or molecular complex that replicates imperfectly, producing heritable variation that changes its reproductive success. The known biosphere's metabolites and informational molecules are built from carbon, hydrogen, oxygen, nitrogen, and phosphorus, but alternatives are certainly possible, especially given the spectacular power of natural selection.

Fans of science-fiction will be familiar with the idea that an alternate life form might be built on silicon rather than carbon, since these atoms have similar versatile bonding properties. Arsenic and phosphorus also share bonding properties, and this led a bold young postdoctoral fellow, Dr. Felisa Wolfe-Simon, to hypothesize that shadow-biosphere organisms might use arsenic in place of phosphorus.

This idea had already been considered and discarded by chemists. Arsenic certainly will form many of the same bonds that phosphorus does, and its known toxicity is because some of our enzymes mistakenly incorporate arsenic into biological molecules in place of phosphorus. However many of these arsenylated molecules have been found to be much less stable than their phosphorylated counterparts, and the bonds found in DNA and RNA were thought to have half-lives much shorter than a second.

However Wolfe-Simon didn't think these objections mattered very much in the face of the immense adaptability of living systems, and, after winning a prestigious post-doctoral fellowship from NASA’s Astrobiology Program, she joined the highly regarded arsenic research group of Dr. Ron Oremland to test her ideas.


Wolfe-Simon looked for arsenic-based life in California’s Mono Lake. Lots of organisms live in the lake and its sediments, but the content of arsenic and other minerals is high (200 µM arsenic). To find out if the sediment contained anything that could survive and grow using arsenic rather than phosphorus, she mixed lake sediment with a simple culture medium. The medium had the same basic chemical properties as the lake water but also provided glucose (for energy) and all but one of the basic elements that bacteria need to synthesize new cells. The missing element was phosphorus; instead the medium contained arsenic.

Something grew! The medium eventually became cloudy with millions of bacteria-sized cells. And the cells continued to grow and divide after they were repeatedly diluted into fresh medium, even when the arsenic concentration was 5 mM (25-fold higher than that of Mono Lake). A bit of the culture was spread onto culture plates containing the same medium solidified with agar. Now individual cells grew into clonal colonies, each from a single cell. The fastest growing of these clones was given a strain name (GFAJ-1) and its properties are reported in the Science paper.

When the arsenic concentration was gradually increased still further, GFAJ-1 was found to grow best with 40 mM arsenic; this level is very high (normal E. coli cells can't grow in 1 mM), but not higher than levels tolerated by some known bacteria, including arsenic-resistant strains of E. coli. The cells grew to high densities on positive-control medium with phosphate but no added arsenate, and only slightly on negative-control medium with neither nutrient added.

The big question was whether GFAJ-1 is a member of the shadow biosphere, so the researchers tested whether the cells had DNA. A DNA-extraction procedure produced a substance that not only behaved like DNA but that could be amplified by PCR and sequenced using primers for normal-biosphere ribosomal RNA genes. Phylogenetic analysis of the sequence then showed that GFAJ-1 belongs to a particular genus of known bacteria, Halomonas (indicated by the tiny dark blue wedge in the figure above). So, big disappointment, GFAJ-1 is part of the normal biosphere, not a new life form.

But why could GFAJ-1 grow on medium lacking phosphorus? Its Halomonas ancestors must have had normal chemistry – had it evolved the ability to use arsenic in place of phosphorus?

This would be almost as exciting a result as finding a member of the shadow biosphere, so Dr. Wolfe-Simon and her colleagues did a series of tests.

1. Analysis of the culture media and other solutions by mass spectrometry showed some phosphorus, usually about 3 µM, thought to come from contaminants in the reagents. However the authors felt that this was not nearly enough phosphorus to permit the observed cell growth in the arsenate medium, especially since the cells had barely doubled when neither arsenate nor phosphate was added.

2. Analysis of washed arsenate-grown cells by a variant of mass spectrometry called Nano-SIMS showed that they contained more arsenic and much less phosphorus per g dry weight and per carbon atom than phosphate-grown cells, and about tenfold more arsenic than phosphorus.

3. Crude fractionation of arsenate-grown cells by phenol-chloroform extraction showed that arsenic was present in all fractions, as expected if it had been incorporated into a wide range of biological molecules that would normally contain phosphorus.

4. Synchrotron X-ray studies of whole arsenate-grown cells found that the arsenic atoms had chemical bonds consistent with being part of biological molecules.

5. Analysis of partially purified DNA from these cells found more arsenic and less phosphorus than in DNA from phosphate-grown cells.

The authors concluded not only that the cells had actively accumulated arsenic, but that they had ben able to grow in the arsenate medium because they had substituted arsenic for some of the phosphorus in their DNA and other cellular constituents.


Why do so many other researchers disagree with these conclusions? The high-tech analyses appear to have been competently performed; they were mainly carried out by collaborators at other laboratories, and few concerns have been raised about them. The problems are largely with the preparation of culture media and samples for analysis, and with the interpretation of the results.

Data quality: The raw data in the Supplementary Materials (online-only) are riddled with inconsistencies that the researchers should have recognized as red flags for experimental flaws. Here’s just one example: Although 118 ppb of arsenic were measured in the 'DNA/RNA' fraction from the control phosphate-grown cells, no arsenic was detectable in the same fraction from the arsenate-grown cells (<20 ppb). Yet when the DNA and RNA in these fractions were analyzed after gel electrophoresis, the arsenate-grown preparation contained twice as much arsenic as the control. The authors don’t mention this discrepancy, so we don’t know whether it’s due to errors in sample preparation, in the assay method, or something else.

Cell growth analysis: One of the most glaring inconsistencies was in the analysis of cell growth (the paper’s Figure 1 below, with my annotations in red). Growth was measured two ways, by the turbidity of the culture (Fig. 1A) and by direct cell counts (Fig 1B), but the authors failed to notice a discrepancy between the two arsenate-medium growth curves (the middle lines with square symbols) that largely invalidates the subsequent analyses of cell contents. In Fig. 1A the cells appear to continue growing for 360 hours, but the cell counts for the same cultures (Fig. 1B) show that cell numbers leveled out at about 150 hours. The cause of this discrepancy is revealed by the electron micrographs (Fig 1C, D and E). GFAJ-1 cells looked like typical rod-shaped bacteria when they were grown on phosphate medium (Fig. 1 D), but when they were grown in the arsenate medium (Fig. 1C) they gradually plumped up, and cross-sections showed that they were full of large white bodies. Microbiologists agree that these are not vesicles, as suggested by the authors, but granules of the waxy energy-storage hydrocarbon polyhydroxybutyrate (PHB). PHB is the bacterial equivalent of fat, and cells make it when they have an abundant energy supply (in this case the sugar in the medium) but must stop dividing because they have run out of phosphorus or nitrogen. Why does this matter? Because PHB contains lots of carbon but no phosphorus, its presence seriously skews measurements of the proportions of different atoms. The arsenate-grown cells, being full of PHB, would have much less phosphorus per g dry weight or per carbon than the skinny phosphate-grown cells, even if all the cells contained the same amounts of phosphorus.

Phosphorus requirements: Many bacteria are known to flourish in environments with less than 3 µM phosphorus, and simple calculations (not done by the authors) show that the contaminating phosphate in the arsenate medium could readily support the observed numbers of cells. The arsenate-grown cells may indeed have contained much less phosphorus than the phosphate-grown cells, but this is likely to be because they were being economical with this scarce resource – the DNA-analysis gel photograph showed that they had dramatically reduced their content of ribosomal RNA.

Cell fractionation: The cell fractions were produced by a chemical procedure that partitions cell contents between organic, aqueous and insoluble fractions. The authors focused on the key biological molecules in these fractions (e.g. DNA and RNA in the aqueous fraction), but neglected the very large number of other molecules present. Finding arsenic in the aqueous phase of a phenol extraction is a long way from finding arsenic in DNA and RNA

DNA purification: The most shocking error was the omission of standard steps from the DNA purification. The alcohol-insoluble pellets from the aqueous phase fractions were run in a gel and the DNA-containing sections sliced out for analysis. Normally the DNA is then purified away from the gel slice and washed using a simple ‘spin-column (a ten minute procedure), but for some reason this step was omitted and the atomic analysis was done on the whole gel slice, bringing along all its contaminants and greatly diluting the sensitivity of the assay. In the absence of the final purification steps it’s impossible to know whether the DNA really contained arsenic.


Before considering what other researchers should or shouldn’t do to clarify things, we need to do some ‘Bayesian’ thinking. By this I mean explicitly considering what Bayesian statisticians call the ‘prior probability’ of the authors’ conclusions, and then considering how the new evidence in their paper changes this probability.

Here’s a simple example of such thinking: Say I come to you claiming that space aliens have given me the power to control coin tosses. As evidence I say that I tossed a coin six times and it came up heads each time.You won’t bother to investigate my powers, because you already know that similar claims have been disproven by many previous investigations (the prior probability of my being right is very small), and that the same outcome will occur by chance in 1 of 32 trials (the new supporting evidence is very weak).

For the hypothesis that GFAJ-1 bacteria use arsenic in place of phosphorus, the prior probability is vanishingly small because of a combination of molecular and evolutionary problems:

- The chemists were right. The arsenic bonds needed in DNA and RNA are spectacularly unstable, with half-lives of less than 0.1 second.

- To be incorporated into DNA arsenic would need to first be incorporated into and stable in ATP, and in all the metabolic precursors of RNA and DNA.

- Many enzymes are needed to process these precursors, and they would all have to be able to handle arsenic.

- The complex cellular machinery responsible for replication, transcription and translation would not be able to use as templates DNA and RNA whose backbones had a mixture of phosphorus and arsenic atoms.

- Even if DNA polymerase could use such a template, the irregularity of the backbone would cause a devastating increase in its error rate, driving the cell into mutational meltdown.

· Selection for arsenic use in phosphorus-limited environments will only be strong if cells can completely eliminate their need for phosphorus.

- The many mutations required to modify proteins and enable arsenic use would all have to occur before any use of arsenic would be beneficial.

- Mono Lake provides abundant phosphorus for growth, so the required selection would not occur there.

Any one of these problems is big enough to send the prior probability of arsenic use into the basement. In any case, unless the fundamental principles of chemistry are wrong, bond instability is a death-knell to the authors' conclusions.


What should the authors do? If they want other researchers to take their claims seriously, they need to replicate the cell preparations and fractionations with much stronger reproducibility and control of contamination at all steps.

What should other researchers do? Should they try to replicate any or all of these results? Only if they don’t have anything better to do. Because the probability that the authors are right is so low, these experiments carry a very high ‘opportunity cost’ – researchers would have to take time away from their ongoing research projects, which are much more likely to produce scientific advances than is replicating this work.

What I'm going to do: Given the massive publicity the original claims received, many people think it’s important that somebody show that the claims are wrong. Nobody else seems to be stepping up to the plate, so I’ve sent for the GFAJ-1 strain and I’m going to do the bare-minimum experiments, testing whether arsenate promotes its growth when phosphate is limiting, and whether arsenic is present in its DNA from after growth in limiting phosphate plus abundant arsenate. (I expect to find that it doesn’t and it isn’t.) I'm not going to try to grow cells in the complete absence of phosphate, both because eliminating the contamination would be difficult and because the negative result I expect (the cells don't grow) wouldn't be very compelling. I'll be openly blogging about this work as I do it - you can follow along at my RRResearch blog.


Much information is in blogs and other material that comes from experts but hasn't been formally peer-reviewed. Links to just about all of this are available at Bora's arsenic link-dump. The best peer-reviewed information is in the paper itself, in the eight Technical Comments, and in the authors' response to the comments, all of which are available here.

About the Author: Rosie Redfield is an evolutionary microbiologist and Professor of Zoology at the University of British Columbia; the cocktail-party description of her research asks "Do bacteria have sex?" Her hair isn't always blue (today it's pink). Find her on her blog and Twitter.

The views expressed are those of the author and are not necessarily those of Scientific American.


Related at Scientific American:

Arsenic bacteria have changed education that is

An arsenic-laced bad-news letter: Who is the audience for online post-publication peer review?