Field of Science

Now no growth in arsenate

Results from growing phosphate-limited GFAJ-1 with and without 40 mM arsenate for 40 hr, in new screw-capped plastic tubes:

This is good in that growth in the absence of arsenate is cleanly limited by the amount of phosphate the medium contains.  A parallel culture with 1500 µM phosphate reached an OD of 0.63.  But the cells in medium with 40 mM arsenate don't appear to have grown at all.

I'm away for the rest of the week (Science Online London!), so I put the tubes back into the incubator, to see if the arsenate cultures just grow  v-e-r-y  s-l-o-w-l-y.

Two weeks later:  No, the arsenate cultures didn't grow at all.

Screening for loss of transformability

In preparation for some mutant-creation work we want to put into our grant proposal, I've been developing an efficient way to screen hundreds of cells for loss of transformability.  We want to be able to use this method to identify cells carrying the point mutation we are engineering, after the mutation has been recombined into the chromosomes of some cells in a competent culture.  Of course we can't know for sure that the mutation will eliminate transformation (finding this out is the point of making the mutant), but this phenotypic screen will complement our planned PCR screen for the mutation itself.

The mutant DNA isn't ready to recombine into cells yet (it needs to be put into a longer DNA fragment), so I'm developing my method with DNA carrying a point mutation that inactivates the recA homolog rec-1; this completely prevents recombination and thus eliminates transformation. 

My initial test system is transformation with rec-1 mutant chromosomal DNA; if the desired mutant DNA isn't ready the tests can be repeated with PCR-amplified rec-1 mutant DNA.  After some fumbling around I now have a screen that works.

The basic procedure is to mix competent cells with rec-1 mutant DNA, plate the cells on plain sBHI agar, and let them grow overnight into colonies, some of which will now carry the rec-1 mutation.  To find these, large numbers of cells from individual colonies are then picked up with a pipette tip, mixed with DNA carrying a novobiocin-resistance gene, and transferred to sBHI agar containing that antibiotic.  Colonies of normal cells will then give some novobiocin-resistant colonies, but colonies of nontransformable mutants won't.

The screen takes advantage of several things we've discovered over the years.  One is the fact that many cells in an H. influenzae colony spontaneously develop competence.  When the cells in a colony are lifted from the agar and mixed with DNA carrying a selectable mutation, some of the cells take up the DNA and acquire the mutation. 

For wildtype cells the frequency of this 'colony transformation' is quite low if the mutant DNA is the chromosomal DNA of a mutant cell, because most of the DNA the cells take up comes from other parts of the chromosome. But the transformation frequency is much higher is cells are instead given a pure DNA fragment carrying the mutation, either a cloned gene or a PCR product.  We have a cloned novobiocin-resistance allele that transforms very well.

We also can greatly increase the frequency of this transformation by doing the initial transformation of our candidate point mutation into hypercompetent cells rather than normal cells.  Because 100-times as many of the cells in the colony are competent, we don't need to plate a large number of the cells from each colony in order to distinguish colonies that give transformants from colonies that don't.  Instead we can just put a small spot of cells from each colony onto the antibiotic plate.  Because we know that the hypercompetence mutation only affects regulation, using these cells won't compromise our ability to detect effects of our engineered mutations on the uptake machinery.

One other minor but very useful thing we know is that some antibiotic-resistance mutations don't require 'expression time'.  With most resistance mutations, cells that acquire the mutation by transformation take an hour or more of culture without antibiotic to express the proteins that will now make them resistant to the antibiotic.  But this expression time isn't needed for novobiocin and kanamycin - instead cells that have taken up novR or kanR alleles can form colonies even if they are immediately placed on antibiotic agar.

In my optimized procedure, each colony is picked up with a pipette tip, and the tip is touched to a sBHI plate to preserve the colony in case it turns out to be the mutant we're looking for.  The rest of the cells are then quickly suspended in 50 µl of sBHI containing 1 ng of the novR DNA fragment (in a well of a 96-well microtiter plate), and 5  µl of that is immediatelyspotted onto an sBHI+novobiocin plate (16 spots per 60-mm plate).  After overnight incubation almost all the spots contain hundreds of colonies - the rare ones that don't contain colonies are the desired mutants.

This screen is fast and efficient enough for our needs.  In a few hours I can test several hundred colonies - since we expect our engineered-mutant DNA fragments to transform at better than 1%, this should be enough to find each mutant.  In the first test I found one rec-1 mutant colony among the 64 I screened.

Aug. 29:  In a replicate experiment I've screened another 144 colonies and found 3 mutants.

Plasmid transformation results

I repeated the plasmid transformation tests, using the same two plasmids but this time using plasmids prepared from both E. coli and H. influenzae, transforming each prep into both RbCl-competent E. coli and MIV-competent H. influenzae.

First consider plasmid pSU20, a small low copy-number plasmid with no H. influenzae uptake sequences.  When grown in E. coli, this gave about ????? transformants in E. coli but none in H. influenzae.  The same plasmid grown in H. influenzae gave about 330 transformants in both E. coli and H. influenzae.  (The two plasmid preps were used a different concentrations, but for each the same concentration was used for both E. coli and H. influenzae.)  This result tells us that H. influenzae cells do take up plasmids that have no uptake sequences, but that the plasmids from E. coli don't survive, probably because they lack methylation to protect them against the H. influenzae cells' restriction enzymes.

The situation is more complicated with the other plasmid, a pSU20 derivative containing a 9.3 kb insert of H. influenzae DNA.  The insert has several uptake sequences, but it also carries the gyrB gene and appears to be quite toxic.  The E. coli-grown plasmid gave only about 400 transformants in E. coli but more than 2000 in H. influenzae.  This was a high-concentration plasmid prep, so I think the high concentration must have allowed some plasmids to escape restriction, and the low transformation of E. coli must be because the insert is toxic.  The H. influenzae-grown prep gave fewer transformants, about 160 in H. influenzae and fewer than 10 in E. coli (no colonies from 100 µL of transformed cells).

Conclusion:  Yes, plasmid transformation definitely does work in H. influenzae, but it doesn't work especially well.

How do I not understand these results? Let me count the ways.

Because I've had inconsistent results growing GFAJ-1 with arsenate and limiting phosphate, I set up a big series of cultures in screw-top glass tubes (the previous cultures had been in small flasks).  The results with the culture tubes are completely different and also internally inconsistent.  Foor convenience I've mainly assessed growth using 'optical density' (OD) to measure the turbidity of each culture, rather than taking the trouble to dilute and plate the cultures and count colonies.

The first surprising result is that the presence of 40 mM arsenate had little or no effect on growth, regardless of the level of phosphate provided (compare the blue, red and green bars for each phosphate level).  Previously I've only sometimes seen very slow growth in 40 mM arsenate cultures with high phosphate (1.5 mM) or with no added phosphate, and other times seen no growth at all.

Second, some cultures with little or no added phosphate grew to much higher densities than previously or than parallel cultures with similar amounts of phosphate (e.g. 4 µM and 5 µM phosphate with no arsenate, and 3 µM and 7.5 µM phosphate with 10 mM arsenate).  Based on colony counts of previous experiments, I expect the 3 µM phosphate cultures to have about 100-fold fewer cells than cultures with 1500 µM phosphate

Third, the culture densities differed dramatically from tube to tube, with little correlation with the amount of phosphate I added. Part of the explanation for this might be that the culture tubes were heavily and variably contaminated with phosphate (they had been previously used, many years ago, and I had rinsed them and their caps well but not acid-cleaned them).  But the contamination would have to have been consistently lower in the tubes that I added no or little phosphate to (because other cultures that differed by only 1 µM added phosphate differed by OD's of 0.2, equivalent to about 5 x 10^8 cells/ml).

(I looked at some of these cultures under the microscope and saw no evidence of contamination with other bacteria, nor did I see any unusual colonies when I plated some of them.)

Fourth, I plated the cells from a few cultures, and one's cfu counts were way lower than expected from its OD.

Fifth, by mistake I initially set up these cultures in base medium with 2 mM phosphate rather than in base medium with no added phosphate.  I didn't throw this set of cultures out but incubated them with the correct cultures.  These grew to higher densities than the low-phosphate cultures, but most didn't grow as dense as similar cultures had previously in flasks.

But again there was substantial tube-to-tube variation in density, and no effect of arsenate at either 10 mM or 40 mM.  Phosphate contamination of the culture tubes can't explain this variation, as all the cultures had ample phosphate.

To rule out effects of contamination with phosphate (or whatever), I'm now going to set up some similar cultures in new plastic screw-top tubes.  I won't bother with 10 mM arsenate and I'll reduce the number of different phosphate concentrations.  I'll also test the effect of filling the tubes to different levels, in case reduced aeration is a factor.

Yes, I have no idea what's going on!

Good astrobiology papers

A few months ago I asked for examples of good astrobiology papers, but the response was disappointing.  This prompted me to suggest a discussion at Science Foo Camp titled "Astrobiology: Buzzword or Science". 

Sarah Kendrew, one of the participants in this discussion, has now posted a thoughtful follow-up on her One Small Step blog: Astrobiology-where's-the-bacon?.  She points to good work on identifying planets that could support Earth-like life, and discusses the need for more collaboration between astronomers and biologists.

Enceladus photo from NASA
The astrobiology discussion at SciFoo was fruitfully combined with a discussion on the ethics of planetary exploration, posed by Jill Tarter, titled "What if Mars has Martians?".  This was something I hadn't given much thought to.  The big issue at present is the possibilities of contamination of other systems with organisms from Earth, and of Earth with hypothetical organisms from other systems.  The consensus was that 'certainty' is a better word than 'possibility', at least where contamination originating on Earth is concerned.  This applies not only to manned exploration but to unmanned visits too, but nobody was quite ready to say that these should all be stopped.

In addition to the above discussion, Jill and other participants were able to provide some history of NASA's Astrobiology program.  I and others had wondered if this was initiated mainly as a public relations exercise, but it was a real grass-roots initiative by NASA scientists.

Mutation probably isn't the explanation

After reading my latest post about the so-far-mysterious non-reproducibility of GFAJ-1 growth in arsenic medium, a reader sent the following suggestion by email:
I've nothing very definite in mind, but such inconsistency, if not due to genetic heterogeneity in the inoculum, suggests the possibility that some heterogeneity is arising during the PO4/As-limited growth period, some cultures stochastically acquiring the ability to pass the barriers by some standard or non-standard (mutational) process. Perhaps you need a way to cell-sort that could separate fast-growing from slow-growing cells in a culture that is in the process of escaping the growth limitation.
I'm pretty sure the discordant results aren't due to mutation, for two reasons.

First, the #4 cultures that grew quickly appeared to do so with no lag and no period of slow growth during which mutations could arise.  The #4 cultures became turbid just as fast as the control #3 cultures without arsenic, so I think that all of the cells in the inoculum contributed to the growth.

Second, one of the experiments where #4 didn't grow (expt. 2) was begun with cells taken from the fast-growing #4 culture of the previous experiment (expt. 1), rather than from the frozen stock of phosphorus-depleted cells used for the other three experiments.  If the growth in expt. 1 was due to a mutant clone, the #4 culture in expt. 2 should also have grown fast, but it didn't.

Inconsistent effects of 40 mM arsenic

OK, I've now done the same experiment four times and gotten two different results, twice each.  Each time I set up six 10 ml cultures of GFAJ-1, each inoculated with about 10^4 cells from one frozen stock (three times) or from a previous culture (one time).

Most of the results are consistent.  Each time, the cultures with no arsenate grew as they had previously, to about 2x10^9 cfu/ml (#1), 2x10^7 cfu/ml (#3), and 0.5-1x10^7 (#5).  Each time, the cultures with 40 mM arsenate and 1.5 mM phosophate or no added phosphate grew little or not at all in the first two days.  YThe two times I've done a longer follow-up they both grew very slowly, to final densities a bit lower than their no-arsenate controls.

But the results with 40 mM arsenate and 3 µM phosphate (#4) are very inconsistent.  Twice they behaved like the other arsenate cultures, growing very little at first and, in the one follow-up, slowly growing to a density slightly lower than the control.  But twice they grew as fast as the control and to about double the density.  I have no idea why.

I think I should consider the fast growth of two #4 cultures to be the anomaly, as the failure of the #4 cultures to grow the other times is entirely consistent with the behaviours of the other arsenate cultures (#2 and #6).  But how to investigate why?  Rather than just repeating the cultures again and again, I should try varying the conditions a bit.  Screw-cap tubes instead of foil-capped flasks?  Slightly different amounts of phosphate?  Less arsenate?

If I had more time I'd test all these variables.  But my big new genetics course starts in two weeks, I'm going to London next week for a week (part brief visit to family, part Science Online London), the CIHR grant proposal needs, by Sept 15, both revision and new supporting data (some of which I'm trying to generate), and the postdoc and I still need to finish his uptake-specificity paper by then (he still needs to send me his new analyses).

Setting up the cultures isn't a big deal, but monitoring them takes time every day, because I'm diluting and plating them to get colony counts.  Maybe if I just monitor growth visually (by turbidity) and not by plating.  Wolfe-Simon et al. grew all their cultures in screw-capped glass tubes, so maybe I'll just set up a lot of these, with more diverse concentrations of phosphate and arsenate.

OK, plasmid transformation does work in H. influenzae...

At least it worked for me, for one of the DNAs I tested.  But the results weren't very clean or very well controlled so I need to do the experiment again.

This time I'll use two different preps of the same shuttle vector pSU20 - one prep of plasmid grown in E. coli and one of plasmid grown in H. influenzae.  (I would have done this the first time but couldn't find a H. influenzae-grown prep.  Today I streaked out the right cells from our freezer stocks, and tomorrow I'll make a plasmid prep for this experiment.)

I expect the E. coli-grown plasmid to transform quite poorly, because the DNA will lack methylation at the HindII and HincII restriction sites.  In my first experiment the difference was dramatic: I got 10,000 transformants into E. coli but none into H. influenzae.  The Rd strain of H. influenzae carries both of these restriction systems (Ham Smith got a Nobel Prize for discovery of these), and its cytoplasm is chock full of the HindIII and HincII restriction enzymes.  Although its own DNA is appropriately methylated and thus immune to cleavage, incoming DNA from other species is cut.  The restriction enzymes only cut double-stranded DNA so they don't affect chromosomal transformation (watching the DNA uptake movie might help this make sense), but plasmids introduced by electroporation or plasmid transformation remain double-stranded, so they're vulnerable.

I'll also use E. coli-grown and H. influenzae-grown preps of the noviobiocin-resistance plasmid pRRnov1.  In my quick-and-dirty experiment I used a very old tube of this plasmid, grown in H. influenzae.  This gave about 100 transformants into H. influenzae but none into E. coli.

Thinking about the impact of endogenous restriction enzymes prompted the postdoc to wonder if this might be a limiting factor in the RAs' recombineering work - it might help explain why her attempts to make 'unmarked' deletion mutants sometimes succeeds and sometimes fails.

First, my understanding of the process she uses:

She starts with mutant cells she's created - in these cells a specific competence gene has been deleted and replaced with a cassette giving resistance to spectinomycin.  Because this cassette can disrupt expression of downstream genes she wants to excise it from the chromosome, leaving an 'unmarked' mutation.

To excise the SpcR insert, she first electroporates a plasmid with the flp recombinase (pRSM2947?) into the H. influenzae cells carrying the SpcR-marked mutation.  The SpcR cassette is flanked by DNA sequences that that are cut and rejoined by the Flp enzyme, excising the cassette.  The plasmid is also temperature-sensitive (Ts) so cells must be plated at 30°C.  If this electroporation works (sometimes it doesn't, perhaps due to restriction)) she gets H. influenzae cells that carry the excision plasmid.  

She then induces the excision genes (by adding tetracycline), plates the cells, and screens for colonies that are now SpcS.  Plating at 37°C ensures that the cells also lose the Ts plasmid.  Sometimes the induction works well (more than 1% of the colonies are SpcS) and sometimes it doesn't work at all.

So the other day the postdoc pointed out that restriction and subsequent repair of an E. coli-grown plasmid transformed into H. influenzae might sometimes produce defective versions of the plasmid, versions that are able to replicate and that confer the expected antibiotic resistance but that have defects in other plasmid-borne functions.  

When the RA electroporates pRSM2947 into the marked mutants, I think she uses a plasmid stock that was grown in E. coli.  Cleavage of this plasmid by the endogenous restriction enzymes might sometimes give plasmids with defects in the flp recombinase gene.  These cells would then fail to produce any unmarked derivatives on induction with tetracycline.  Even if she routinely checks the transformants for presence of the pRSM2947 plasmid, she might not detect that the plasmid now has a defect.  So maybe one way to improve the efficiency of the excision process is to start with a stock of pRSM2947 that has been grown in H. influenzae

How is this work going to be useful for our upcoming grant proposal, you might ask?  I'm asking myself that too.  More on this later.

Update:  The RA tells me that she switched to using H. influenzae-grown plasmid a while back, but that this didn't solve the excision problems.

Does plasmid transformation work in H. influenzae?

The word around the lab is that the standard technique for transforming H. influenzae cells with plasmid DNA hasn't worked for anyone in years, so I'm testing it myself.  It always worked OK for me, but I haven't done it in a long time.  The technique is simple: cells are made naturally competent in the usual way (starvation in MIV medium) and incubated with plasmid DNA.  The plasmid DNA is efficiently taken up into the periplasm, but it can't get across the inner membrane because that translocation needs a DNA end and the plasmid is circular,  The cells are then given an osmotic shock (30% glycerol for 10 minutes), which somehow drives the plasmid into the cytoplasm.

Why do this now?  Lord knows I have lots of other stuff to deal with, but plasmid transformation is potentially something we'd want to include in our upcoming CIHR grant proposal, preferably with some preliminary data.  The only major concern raised by the reviewers of our previous submission was that we didn't propose to evaluate the effects of point mutation changes in competence genes, just of deletions of the whole genes.  The RA has developed a very nice way to make point mutations in genes cloned in E. coli, but we don't have an efficient way to transfer them into the H. influenzae chromosome.  Well, we can transfer them into the chromosome by transforming competent H. influenzae cells with the mutated DNA, but this is only efficient if the DNA fragment is quite long (see this post), and finding the desired transformants requires laborious screening because there's no associated selectable marker.  We're in the process of making a suitably long fragment of an interesting mutation she's generated, and I hope to be able to test this in the next week or two.

But the post-doc pointed out that we should also consider ways of testing point-mutation changes borne on plasmids, because these can easily be selected for.  The strategy would be to test cells containing the mutant plasmid and a deletion that removes the chromosomal copy of the gene.  One complication is that the plasmid-borne competence gene must have a promoter to drive its expression - ideally this will be its own competence-regulated CRP-S promoter.  I'll have to check whether the short cloned fragment containing the RA's interesting mutation includes its promoter - if not, the longer versions will.

In most cases we'd be working with a gene that is essential for natural competence, so there are two ways to construct these strains.  The first is to start with the strain carrying the chromosomal deletion and introduce the plasmid by electroporation.  The RA has found that electroporation sometimes works very well and sometimes doesn't work at all - I don't know if she's ruled out the effects of the restriction enzymes in the H. influenzae cytoplasm, which are expected to cut up DNAs containing unmodified HindIII and HincII sites.  The second way is to start with competent wildtype cells, introduce the plasmid by plasmid transformation with selection for the plasmid vector's antibiotic resistance, and then introduce the deletion by normal chromosomal transformation with selection for the inserted spectinomycin-resistance cassette.

I think this work needs a lot more thought, and soon, because the grant proposal deadline is Sept. 15.  But last night I did a test of the basic plasmid-transformation method, using several DNA preps of plasmids based on the chloramphenicol-resistant shuttle vector pSU20 (two very old, and one new but of plasmid grown in E. coli).  I transformed them into both competent H. influenzae and competent E. coli (as a control).  I'll get the E. coli results today, but will have to wait till tomorrow for the H. influenzae results because the colonies grow very slowly on chloramphenicol plates.

We now resume our #arseniclife blogging...

When we paused last week, I had just noted the unexpected result that, in two of three tests (#1184 and #1189), GFAJ-1 cells in 3 µM PO4 + 40 mM AsO4 grew to a higher density  than the cells in 3 µM PO4 (no AsO4), but the cells with more and with less phosphate didn't grow much at all.  In the other test (#1188) the cells didn't grow at all in the arsenic cultures; I won't say anything more about these cultures unless I get more results like this.  All cultures without arsenic grew as they had in previous experiments.

I left the #1189 arsenic cultures in the incubator while I was away, and over this ~225 hr incubation the 1.5 mM-phosphate' and no-added-phosphate cultures grew to densities close to those of their no-arsenate controls.  I don't know why they grew so much slower than the 3 µM phosphate+arsenate culture.

I'm going to repeat these cultures once more.  If I see the same results (fast high growth of cells in 3 µM phosphate + arsenate), I'll dilute the 3 µM phosphate cultures into 400 ml of the same medium ± arsenate, to get lots of cells for my first arsenate-grown DNA prep.


What's SciFoo?  Science Foo Camp, of course!  (Here are links to Wikipedia and Twitter.)  You've probably never heard of this 'unconference' - it's small, you need an invitation, and nobody talks about their data.

It happens at the Googleplex (Google's headquarters in Mountain View California.  About 200 people, mostly scientists but also journalists, artists and teachers, gathered for two days of high-quality conversations.  Some of the conversations happen in specific blocks of time, and the rest over meals and drinks and demos and just hanging around.  Much more like a series of wonderful dinner parties than like a typical scientific meeting.

The bike is part of Google's fleet (photo from aemkei's Flickr photostream).  I was going to take one for a spin, but they're 'fixies' (fixed-wheel, no brakes) and I didn't think my cycling skills were flexible enough to save me from crashing when I wanted to stop.

Polishing my SciFoo talk - last slide

My 5-minute 'Ignite' talk at SciFoo is tonight, and I still need to work out the point I'll make with my last slide.  (It's too late change the slide - our Powerpoint files had to be submitted two days ago.)

I'll have just said that big mistakes were made by everyone involved (authors, reviewers, editors, publicists, journalists).  Then I want to say something about how scientists in particular need to be able to admit their errors - we're working not only at the frontiers of knowledge but at the frontiers of our abilities.  Failure to admit we've been wrong is a betrayal of the scientific process.

Can I say that in 15 seconds, without rushing my words?  Maybe, but it would be better to cut a few words, or at least a few syllables.

How about this?

Previous slide:  Everyone involved made big mistakes (authors, reviewers, editors, publicists, journalists).  But the big betrayal wasn't the errors but the failure to admit them.

Last slide:  Scientists work not only at the frontiers of knowledge but at the frontiers of our abilities, and learning to be wrong should be part of our training.

A good paper, but is it 'astrobiology'?

A while back I asked for examples of good astrobiology papers.  After a bit of back-and-forth with some commenters, I clarified my search:  I was looking for papers that reported competent experimental research and self-identify as 'astrobiology'.  And I was looking for personal recommendations to papers that had been read by the person recommending them, not just pointers to journals or reports where I could read astrobiology papers and decide for myself whether they were any good.

The results were disappointing.  The problem wasn't that the commenters recommended papers that turned out to be lousy, but that they didn't recommended any recent papers at all, except for a reanalysis of Miller's original spark-discharge material from his 1953 experiment.  The problem wasn't that the commenters were ignorant of the field - one was the Director of the NASA Astrobiology Institute, who directed me to a recent report on the Institute's success but didn't give personal recommendations for any of the publications it lists.

I'm thinking about this issue now because of a paper that just appeared in PNAS, titled Carbonaceous meteorites contain a wide range of extraterrestrial nucleobases (authors Callahan et al.).  CNN's Lightyears blog headlines this as DNA discovered in meteorites, but most reports were more sensible.  (The image below is from the Lightyears blog.)

To my non-expert eye this looks like good science.  The results are not really surprising (purines have been found in meteorites before), but they're very solid.  What makes this work important is that the authors surveyed a wide range of meteorites, carefully eliminated sources of external contamination with terrestrial purines, and showed that the distribution of purines found in the meteorites matched that produced by laboratory reactions simulating space chemistry but not that of terrestrial contaminants (e.g. they found not only the common terrestrial contaminant adenine but also 2,6-diaminopurine).

So is this good paper a good astrobiology paper?  I don't think it qualifies.  Although most of the authors are supported by NASA, they nowhere mention astrobiology or consider whether their work has implications for the origins of life anywhere but on Earth.  

(I'm officially a zoologist, so now I'm off to the San Francisco Zoo, planning to be back in time for the SciFoo meet and greet this afternoon.)

Preparing my Ignite talk for Science Foo Camp

I'm allowed 5 minutes and 20 slides (15 seconds each) to tell the #arseniclife story to about 250 scientists at Science Foo Camp on Saturday.

I'm hand-drawing the slides, so as to have simple appealing illustrations rather than dense data.  This is taking ages as I lack both the artistic skills and the technical skills to do it efficiently, but I think it will be worth the effort.  Fitting everything smoothly into 5 minutes also takes a lot of practicing - luckily each practice takes only 5 minutes.

Control DNA preps for my mass-spec colleagues to test

I've got lots of GFAJ-1 cells in the fridge, so I think I'll make some DNA preps from them to send to my colleagues at Princeton for the initial mass-spec measurement of baseline levels of arsenic.  They've done the initial controls, determining that their setup's detection limit is about 1-^-7 - 10^-8 M arsenate, equivalent to replacement of about 0.01% of the phosphates in the DNA backbone with arsenate.  This is about 300-fold lower than the 4% replacement claimed by Wolfe-Simon et al.

The cells were grown in AML60 medium with either limiting or ample phosphate - I don't think it matters which.  I'll do one DNA prep with the limiting-phosphate cells.  As a control for carry-over arsenate contamination, I'll add sodium arsenate to the other culture (40 mM) and let them steep in it for a little while, then wash away the arsenate and do another DNA prep.  I'll also take some purified DNA, soak it in 40 mM arsenate for a little while, and then repurify it.  If my DNA purification methods are good, none of the DNAs should have detectable arsenic.

Growth of GFAJ-1 in 40 mM arsenate (not)

My initial attempts to grow GFAJ-1 in media containing 40 mM sodium arsenate have given unexpected results.  The first time I thought it might be a screw-up on my part, but the second time suggests a problem with the cells.

The experimental design is simple.  I set up the six flasks shown below, and inoculated each with about 2 x 10^4 GFAJ-1 cells/ml.  If we ignore the presence of glutamate in the medium, flask 1 corresponds to the +P/-As condition of Wolfe-Simon et al.'s Fig. 1B (shown below the flasks), and flasks 3 and 4 correspond to  the -P/+As and -P/-As conditions.

Because I've already found that the cells grow in limiting phosphate without any arsenate,  I expected dense growth (~ 10^9 cells/ml) in flasks 1 & 2 (lots of phosphate), thinner growth in flasks 3 & 4 (limiting phosphate) and even less growth in flasks 5 & 6 (very limiting phosphate).  The graph below shows the results (blue line is initial cell denbsity).  Cultures 1, 3 and 5 grew as expected, but I was surprised to see very little growth in flask 2 and quite thick growth in flask 4 (both with arsenate).  

The most likely explanation was that I'd made a mistake in setting up the flasks - perhaps putting the 1.5 mM PO4 into flask 4 instead of flask 2.

So I set up the cultures again, this time starting them from cells taken from flask 4, because these appeared to have grown well in the presence of arsenate.  (The other cells had been growing without arsenate for about 2 months, and I was worried that they might need time to adjust to this stressor.)  These cultures started with about 10^3 cells/ml.  The graph below shows interim results after only 14 hr incubation; again the blue line is the initial cell density.

All of the cultures without arsenate have grown well; none have run out of phosphate yet.  But there hasn't been any growth in the cultures with 40 mM arsenate; instead most of the cells have died.  These numbers are the density of the cultures yesterday morning.  This morning cultures 1 & 3 were turbid, as expected, but all the arsenate cultures looked clear.

Because these arsenate  cultures looked unexpectedly thin yesterday, I set up another replicate set of 6, using phosphate-depleted cells from the freezer.  I'll have results from these tomorrow, along with the final results from the above cultures.*

So maybe I'll need to gradually re-accustom GFAJ-1 to growing in high arsenate, by first growing them in a lower concentration.  This will take some time, and will have to wait until I get back from Science Foo Camp next Sunday.  I'll also leave the present cultures for another week, to see if cells eventually grow.

Quick check that I made up the sodium arsenate correctly:  The molecular weight of sodium arsenate is 312 (this is the dibasic, heptahydrate version = Na2HAsO4 7H2O), so I put 31.2 g into 100 ml final volume, to give a 1.0 M solution.  I then added 400 µl of this to each 10 ml culture to give 40 mM.

*Tantalizing note added Aug. 8:  Well, my third set of cultures are behaving like the first set - the cells in 3 µM PO4 + 40 mM AsO4 have grown up slightly more dense than the cells in 3 µM PO4 (no AsO4), but the cells with more and with less phosphate haven't grown much at all.  So I probably didn't make a mistake with the first set after all.  I don't know what this could mean, but it is sort-of consistent with the -P/+As and -P/-As results in the Wolfe-Simon et al. figure above...  Unfortunately we'll all have to wait until next week for me to do any more investigating.

A Haemophilus transformation experiment

In addition to my GFAJ-1 arsenic experiments, I'm also doing work on our real research program, natural transformation in Haemophilus influenzae.  We need to (yet again) revise our DNA uptake grant proposal for the Sept. 15 deadline, and I'm doing some work that will let us strengthen that proposal.

The previous version of the grant included a substantial section on making marked and unmarked knock-out mutations of all the competence genes, and one of the reviewers sensibly asked if we could also make point-mutation changes to these genes.  This would indeed be very valuable, so we're trying to generate evidence that we can do this.

The biggest obstacle is not making the mutant versions of the genes (the RA can do that easily in E. coli), but identifying the rare H. influenzae cells that have integrated them into their chromosomes.  We'd like to simply transform the mutated DNA fragments into competent cells, but this works very inefficiently when the DNA fragments are short.  Unfortunately, short fragments work best for the mutagenesis steps.

My latest experiment tested the effect of fragment length on transformation efficiency  I used a cloned DNA fragment containing the novR (novobiocin resistance) allele of the gyrB gene.  This is a point mutation, in the middle of a 9.3 kb fragment.  I did three test transformations, each using the same amount of the plasmid.  Test 1 used plasmid DNA cut to release the 9.3 kb fragment intact.  This gave a transformation frequency of 7 x 10^-3, about 10-fold higher than the control transformation using chromosomal DNA from a novR strain.  Cutting the plasmid with restriction enzymes that gave a 4.8 kb novR fragment reduced the transformation frequency only slightly, to 4 x 10^-3, but cutting with enzymes that gave a 2.6 kb fragment reduced it 10,000-fold, to 4 x 10^-7.

Several factors complicate this experiment.  I don't know the plasmid DNA concentration; I think it was about 100 ng/ml, which would be less than saturating.  The plasmid insert contains several uptake sequences, but I don't know their distribution (I could dig this info out).  The digest that gave the 4.8 kb fragment may not have been complete, in which case the true transformation frequency might be lower.  The control transformation frequency was lower than I expected

We can conclude that our mutagenesis experiments should use fragments that are at least 5 kb long.  If this gives a transformation frequency of only 4 x 10^-3 we'll have to do lots of screening (possible but tiresome), but the transformation frequency can probably be increased ten-fold by using more-competent cells and more DNA.

First evidence refuting Wolfe-Simon et al.'s results

I belatedly realized that the data I posted yesterday are quite important, in that they contradict the growth results in Figure 1 of Wolfe-Simon et al.'s paper.

Here's their Fig. 1B (growth curves based on cell counts), annotated with a table of the levels of phosphate and arsenate they found in their media.

Wolfe-Simon et al. used this graph as evidence that the GFAJ-1 bacteria could use arsenic for growth, in place of phosphorus.  They claimed that growth in the 'added AsO4' medium must be due to the cells using arsenic because (1) this -P/+As medium had no added phosphate, and the small amount of phosphate contaminating most batches was insufficient to support this growth and (2) the cells did not grow in the plain medium (-P/-As), to which neither phosphate nor arsenate had been added.

In my blog and in my published Technical Comment I argued that the ~3 µM phosphate contaminating the medium was indeed sufficient to support the observed growth of ~ 2 x 10^7 cells/ml.  The calculation underlying this argument is shown below.  Given the uncertainty in genome size and in the % of cellular P needed for DNA, my new results nicely support this argument - adding 3 µM phosphate to the -P medium supported growth to 1.7 x 10^7 colony-forming units/ml (equal to at least 1.7 x 10^7 cells/ml).
My new results also contradict Wolfe-Simon et al.'s observation that cells could not grow under limiting phosphate if arsenic was not provided.  No arsenic was added to my low-phosphate media, but the cells grew just fine.  I think that GFAJ-1 might have failed to grow in Wolfe-Simon et al.'s batch of -P/-As medium because this batch contained much less phosphate than the other media.  The levels of phosphate contamination they detected (red boxed area above) were very unpredictable (3.7 µM in one batch, <0.3 µM in another, and 7.4 µM in the wash solution), and the Methods don't say that all cultures in this figure were made with the same batch.

I've now set up some new cultures, testing whether the presence of 40 mM arsenate affects growth in media with no added phosphate or with 3 µM or 1.5 mM phosphate.  I hope the sudden introduction of 40 mM isn't too big a shock to the cells; they've been growing without any arsenate for the past two months.
In other progress, the Research Associate has obtained the 16S rRNA sequence of the strain I'm working with.  It exactly matches the sequence of GFAJ-1 in GenBank, so now we're sure these are the right cells.  This confirmation is especially important since the growth properties I'm finding for GFAJ-1 don't match those reported by Wolfe-Simon et al.

Growth of GFAJ-1 under phosphate limitation

Before we can test whether the bacterium GFAJ-1 can really incorporate arsenic into its DNA when phosphorus is scarce but arsenic is abundant, I need to find out how it behaves when phosphorus is scarce and no arsenic is available.  Here are the results.

The first graph shows how dense cultures got when no phosphate was added to the basic growth medium.  I described this experiment in two posts a few days ago (planning, results), but I didn't show the data.  I started the cultures with cells that were either phosphate-replete or phosphate-depleted, and at densities ranging from 10^4 to 10^7 cells per ml.  All of the cultures except the most dense phosphate-replete one ended up at densities of 1-3 x 10^6 cells/ml.  This tells me that, even though I didn't add any phosphate to the basic medium, each ml contained enough phosphate to produce about 2 x 10^6 cells.  This is about ten-fold lower than the growth observed by Wolfe-Simon et al, and so my experiments should use medium with enough phosphate added to produce about 2 x 10^7 cells/ml.

The second graph shows how small amounts of added phosphate increase the cell density.  These cultures all had the same initial cell density (10^4 cells/ml), but they had different amounts of added phosphate.  As before, medium with no added phosphate yielded about 2 x 10^6 cells/ml, and medium with 1500 µM phosphate yielded about 3 x 10^9 cells/ml.  Media with 3 µM phosphate (the concentration of contaminating phosphorus in most of the media used by Wolfe-Simon) gave a cell density just below 2 x 10^7, and 6 µM phosphate gave about 4 x 10^7.  So I think I'll use 4 µM phosphate as my final phosphate-limiting medium.

Yesterday I also (very carefully) made up my 1.0 M stock of sodium arsenate.  I put on all the safety gear I usually eschew (lab coat, gloves, dust mask, full face shield) and moved our top-loading balance into the fume hood so any arsenic dust would be sucked away rather than entering the lab air supply.  (I even put the balance into a plastic bag so no arsenic dust could get into its innards.)  The sodium arsenate turned out to be in the form of granular crystals (like coarse salt or sugar), not fine powder, so dust wasn't a problem.  But when I carefully unscrewed the cap a lot of grains fell out of it (maybe 0.5 g!), so I was really glad to be working in the hood where cleanup was easy.

Now the arsenate solution is sitting on my lab bench, conspicuously labeled POISON.  Beside it sits a big bottle to discard the waste culture medium into  - this will eventually be sent to our chemical waste facility for disposal.  I'll set up another bottle for waste tips and other solids contaminated with significant levels of arsenic.