OK, I finally have conditions where GFAJ-1 cells grow reproducibly in medium containing 40 mM sodium arsenate: tightly sealed screw-cap glass tubes or bottles, half-full, gently rocking at 37 °C.
It's sadly true that I lack any insight into why the cells wouldn't grow in polypropylene screw-cap tubes, or in flasks, or why sometimes they wouldn't grow in anything. Since Wolfe-Simon et al. grew their GFAJ-1 in screw-capped glass tubes, I think I'm adequately replicating their growth conditions.
So now I've grown big batches of cells in bottles and extracted DNA from them. My collaborators tell me they'd like to have about 50 µg of DNA for their cesium chloride gradient purification and mass spectrometry analysis. This requires starting with about 2 x 10^10 cells, given an estimated genome size of about 3.7 megabase pairs, and allowing for some losses in purification. For cells growing in phosphate-rich medium I figured 50 ml of culture would be enough (one with arsenate and one without), and for phosphate-limited cells I tried 500 ml. Harvesting the cells turned out to be a bit tricky, because when I centrifuged them they formed a very loose pellet - I had to take the pellet with about 10% of the supernatant and centrifuge again.
I did only a crude DNA prep, by my standards, but the DNA is much cleaner than one sample in Fig. 2 of the Wolfe-Simon et al. paper. I lysed the cells with lysozyme and then 1% SDS, extracted them once with phenol and once with phenol::chloroform, and added NaCl to 150 mM and 2 volumes of 95% ethanol, all as Wolfe-Simon et al. did. But instead of centrifuging the now-insoluble DNA and RNA, I spooled the DNA fibers out onto the tip of a sealed glass pipette, rinsed them with 70% ethanol, and air-dried them. (I also added RNase A with the SDS to degrade the RNA.) Spooling can only be done with chromosomal DNA, because it requires long fragments at high concentration, but it's the method of choice because it leaves behind all the non-DNA insoluble material that centrifugation would pellet with the DNA.
I then resuspended the DNA by swirling the in Tris-EDTA and gave the clumps time and pipetting to help the fibers disperse. I checked the concentrations using the wonderful NanoDrop spectrophotometer, and ran about 200 ng of each prep in an agarose gel to check its quality (length and cleanliness). The gel photo below shows the results - clean preps of DNA fragments longer than 30 kb (the top size standard is the 27.5 kb HindIII fragment of Lambda DNA).
I have almost a mg of the DNA in the rightmost lane. This was a separate prep - I hadn't yet discarded the high-phosphate plus arsenate cultures I'd done several days before (see Growth!), so I just pooled them all, collected the cells, and did a parallel DNA prep.
One problem with the cultures was that the phosphate-limited cells without arsenate looked very sickly when I harvested them, with orangish clumps of debris in the medium after two days growth and many misshapen cells seen under the microscope. And I only got 7.6 µg of DNA from this culture . So I inoculated a new culture, this time using 1000 ml divided between three bottles. Again the culture looked bad, but I was more careful in harvesting the cells, and would up with about 132 µg of DNA. So on Monday I'll send 50 µg of each DNA to my collaborators.
The critical test will be assaying for arsenic in the DNA from cells grown with limiting phosphate and 40 mM arsenate, since this is the condition that was claimed to cause arsenic to be incorporated into DNA. I'm not sure how important the other culture conditions will turn out to be - if we detect no arsenate at all in this DNA, will we really need the other conditions to make our case? But if we do detect arsenic in this prep, these will be controls for background arsenic levels.
The other odd think about my cultures was that the cells with 40 mM arsenate grew better than the cells without arsenate. This could just be an effect of ionic strength, since I put an equivalent volume of water in the no-arsenate cultures, so I'm going to do a careful dose-response curve with a wide range of arsenate concentrations, using NaCl to keep the ionic strength constant.