Field of Science

Tweezers-prep progress

1.  I tested cell adhesion to the coverslips the RA had cleaned for me - it wasn't dramatically better or more consistent than either uncleaned or acid-washed coverslips (after all had been coated with poly-L-lysine using his spreading method.  Overall plenty of cells are adhering, given that I'll only need a small number of cells adhered for each chamber.  I think part of the trick was inverting the slides for 10 minutes before washing away the nonadherent cells, so the cells settle on the coverslip.  I'm also giving the slides a few brisk taps during the washing, so that loose cells detach and wash away.

2.  Incubating the chambers with concentrated BSA protein didn't significantly reduce the number of H. influenzae or B. subtilis cells that stick to the poly-L-lysine coated coverslips.  Nor did it prevent the polystyrene beads from sticking to the coverslips.  A Google search for treatments that might block/inhibit the poly-L-lysine surface revealed only that people are coupling poly-L-lysine and polyethylene glycol or other materials into 'block' copolymers, and testing the ability of poly-L-lysine to 'inhibit' cell growth and various enzymatic reactions.  I didn't test BHI directly, but cells growing in BHI did adhere to the coverslips so I don't think it's much of an inhibitor either.  I don't know how big a problem the unwanted sticking is going to be, but I'd welcome suggestions for other possible blocking agents to test.  Just for fun I'm going to test milk.  Not reconstituted powdered skim milk, but whatever's in the food fridge (2% milk, I think).

3.  I made my big batch of competent cells and froze 34 tubes (0.5 ml), as well as 14 tubes of log-phase cells.  But I haven't yet tested how competent they are.

4.  Several tests showed that the H. influenzae cells that are stuck to the coverslips are otherwise healthy.  In the first test I filled the chambers with sBHI plus low-melt agarose and incubated them overnight.  I didn't seal the ends of the chambers but put the slides in a humidified box, but the chambers dried out completely overnight.  But I could see that there had been lots of cell doublings before the medium dried up.  So next I tested sealing the chamber ends with nail polish.  The tweezers lab people use wax from a candle, because they worry that the nail polish might be toxic, but I didn't have a candle.  The nail polish is indeed toxic (probably because acetone is quite water-soluble); the cells close to the ends of the chamber didn't divide at all.  But the cells in the middle part of the chamber grew well, producing nice tight microcolonies by the end of the day.  I brought in a candle, and today I'll see if I can improve my skill at applying melted wax to chamber ends.

5.  I think today I'll also test whether the cells on the coverslips can be transformed, by washing in MAP7 DNA, and then DNaseI and then sBHI agarose with added novobiocin.  Provided I have thousands of cells stuck to the coverslip, I should be able to find rare transformant colonies.  I'll try this with B. subtilis too, selecting for Trp+.

Uptake sequence manuscript done, back to the tweezers experiments

Yesterday I finished revising all the components of our manuscript on uptake sequence evolution, and sent them to the co-author who's handling the resubmission.  it was already a good manuscript, but now it's very very good, so if the editor and reviewers don't like it we'll just sent it somewhere else.

I've been working on the components of the optical tweezers experiments, with modest progress.  The tweezers apparatus still had problems when I was there the other day (it wasn't aligned properly?), but I have lots of other parts to work on.  Here's a list:
  • The RA of the tweezers lab showed me a way to pre-clean coverslips, and I need to test the ones he prepared to see if they give more consistent results than the ones I've been using. (Coat his and mine with poly-L-lysine, make into chambers, test binding of competent cells.)
  • I need to test whether washing treated coverslips with culture medium (BHI) or protein (BSA) blocks cell and bead binding, because once I have cells bound to the surface I don't want beads to stick to it.
  • I need to make a big batch of competent cells and freeze them in 0.5 ml aliquots, because the cells I made before were contaminated.  (The contaminants were flagellated cells that otherwise looked just like H. influenzae - they were the spinning wiggling cells I described in my last post.)
  • I need to test whether cells that have stuck to a coverslip are still able to grow.  To do this, my plan is to replace the medium in the chamber with medium containing low-melt agarose, and once this has set incubate the chamber at 37 °C for a few hours and then examine the cells under the microscope, looking for clusters of cells (microcolonies).  As a control I'll first try this with cell-free chambers and agarose medium that already contains growing cells.
  • For today I'm just going to work with H. influenzae, but I should also test these factors with B. subtilis cells.

Haemophilus doesn't have flagellae!

I'm still testing treatments to get cells to stick to coverslips (with more success):

I used more poly-L-lysine this time, but the important thing I realized was that I need to tap the slides briskly to loosen the cells that are just sitting on the surface but not attached,a nd to rinse a lot of medium through the chamber.  This let me see that there were usually substantially more B. subtilis cells attached to the coverslip than to the glass slide.

So then I tested whether my competent H. influenzae cells would also bind to the coverslips.  They did, better than the B. subtilis cells.  I could see that many of the cells were only attached at one end, with the other end moving in the medium.  But what was surprising was how vigorously the free parts of the cells were moving around.  Not just bouncing back and forth (can Brownian motion be that vigorous?), but sometimes spinning around almost like cells tethered by their flagella (see this movie, which I think is Salmonella but might be E. coli).

But H. influenzae definitely doesn't have flagella.  So I'm plating the cells I have onto LB and BHI plates with and without hemin and NAD, to check that some flagellated imposter hasn't snuck in to my culture.  (It would have to make colonies that look just like H. influenzae colonies, and be sensitive to novobiocin...)

Sticking cells to coverslips (one step forward, two steps back?)

I coated some coverslips with poly-L-lysine, using the rub-until-dry method I was shown.  I then assembled the coverslips into 'chambers' like those described here, and tested whether competent B. subtilis cells would stick to them.

I wanted to use chambers for these tests because that's what I'll be using with the optical tweezers.  Another benefit turned out to be that I could compare how well cells stuck to the coated coverslip surfaces to how well they stuck to the untreated surfaces of the glass slides that form the bottoms of the chambers.

I let the cells sit in the chamber for 5-10 minutes at room temperature, and then washed them out by flowing about 10 volumes of medium through the chamber, introducing it in tiny drops from a pipette tip at one end and absorbing the flow-through with a square of blotting paper at the other end.  Then I looked at the chamber under the microscope, comparing the cells on its upper and lower surfaces.

In most tests I saw little or no difference between the treated and untreated surfaces; quite a few cells were stuck on both.  The largest volume of poly-L-lysine solution I used did give some patches where many cell stuck, but these were at the edges of the coverslip, where the rub-until-dry treatment hadn't reached well.  Cell density (on both surfaces) was also generally higher at the edges of the chambers - I think this may just v=be because the washing is less effective at the edges.

I also tested my DNA-coated beads.  They didn't stick any better than the cells did, which is good.  But there were lots of clumps of beads - perhaps I didn't vortex them well.


In a previous experiment I had better results with coverslips that had been presoaked in acid alcohol, so I'll try this again.

I'll try higher concentrations of poly-L-lysine.

I'll try using more dilute cells.

I'll try H. influenzae cells as well as B. subtilis cells, because their surfaces have very different chemistries.

I'll try incubating the chambers+cells upside down before washing the cells out, so the cells will settle on the coverslip rather than the slide.  Though, for the tweezers experiments, it doesn't matter which surface the cells are stuck on.

I'll try marking a reference spot on the coverslip, so I can track whether cells present after the first wash are removed by the second one.

I'll reread the papers that did tweezers studies of competent cells, and email their authors for advice.

Transformation by DNA on beads

Yesterday I incubated some of my competent H. influenzae cells with DNA carrying a novobiocin-resistance gene.  One tube got the normal prep of chromosomal NovR DNA.  Two others got DNA that had been cut with one of two restriction enzymes (either EcoRI, average fragment size 6 kb or XhoI, average fragment size 12 kb) and then had biotin incorporated at the ends of the fragments, and three others got streptavidin-coated polystyrene beads with the biotinylated DNA bound to them.

I had calculated that the first batch of DNA+beads had about 250 ng of DNA per ml, so I used the same volumes of beads for the others, and diluted the non-bead DNAs to 250 ng/ml before using the same volumes.  I incubated the DNA plus cells for 15 minutes and then plated the cells on Nov agar, and (diluted) on plain agar.  Today I counted the colonies and calculated the transformation frequencies.

Point 1.  Cutting the DNA with XhoI reduces its transforming ability by 4-fold, but cutting with EcoRI reduces it by about 100-fold.  So I checked where these two enzymes cut relative to the NovR (gyrB) gene.  EcoRI cuts inside the gene, but XhoI only cuts at sites 1 kb and 3 kb on either side of it. 

Point 2.  DNA attached to beads transforms!  This wasn't a sure thing, for lots of reasons.

Point 3.  DNA attached to beads transforms 10-30-fold worse than the same amount of free DNA.  This is not surprising because most of the DNA on the beads will be inaccessible because it's tangled up with or behind other DNA fragments.

Now I'll celebrate by spreading some poly-L-lysine on some cover slips, using the new method I was taught yesterday, so tomorrow I can test whether cells bind to these and whether they stay alive after binding.

Tweezers progress

Yesterday I took my new batches of frozen competent cells (B. subtilis and H. influenzae) across town to the biophysics lab, so I'll have cells there to test without needing to make fresh ones.  While I was there I learned a better way to coat cover slips with poly-L-lysine, which I hope will give more reproducible attachment of competent cells.  And I spent more time making microscope-slide chambers and learning about the tweezers apparatus (mostly standing by while the expert grad student made adjustments to the optics and electronics).

I also attended a seminar about DNA bending.  Short fragments of double-stranded DNA (~100 bp) have been reported to circularize much more efficiently than predicted by their expected persistence length (paper by Cloutier and Widom), and one proposed explanation is the formation of tiny 'bubbles' in the DNA structure – places where several base pairs have separated although the DNA backbones remain intact.  Even though such short single-stranded regions are expected to be very transient they can have a big impact on the probability that the ends of the DNA will meet, allowing DNA ligase to join them and circularize the DNA.

Blocking with BSA makes a big difference

The streptavidin-coated magnetic beads that hadn't been blocked with BSA aggregated into big clumps when mixed with cells, but the treated beads remained separate.  The clumping was independent of whether the beads had been incubated with biotinylated DNA, so I think the clumping was because most cells stuck to more than one bead and most beads stuck to more than one cell. 

So then I checked the streptavidin-coated polystyrene beads I'll use for the tweezers experiments.  Luckily adding cells didn't make them clump.

Now I'm making fresh batches of competent H. influenzae and B. subtilis to freeze, so I'll have consistent cells to work with.

Do cells bind to DNA stuck on beads?

I need to plan the experiment(s) where I test whether competent cells bind to DNA on magnetic beads.

I have 1 ml each of two types of 1µ and 2.8µ magnetic 'dynabeads'.  The beads have been coated with streptavidin, and one batch of each size has then been blocked with BSA, which apparently reduces the surface charge and makes them better for binding protein but not so good for nucleic acids and I think more prone to clumping in high-salt buffers.  I don't know how the different surfaces will affect non-specific cell binding, something I want to avoid (Invitrogen recommends  including 0.01-0.1% Tween 20 to reduce non-specific binding, where that won't interfere with the assay).  And I don't know how the surface properties of these  beads compare to the properties of the polystyrene beads I'll be using for the tweezers experiments.  (I need to keep reminding myself that I'm only using the magnetic beads to check whether cells will bind to DNA on beads, so I shouldn't waste a lot of time optimizing the assay.)

I can separate the beads from 150 µl of liquid by simply drawing the mixture up in a pipette tip, holding the tip in the nanobead magnet rack for 10 seconds, and slowly expelling the liquid; almost all the beads remain behind on the side of the tip.  I can then resuspend the beads by drawing clean buffer up into the tip, away from the magnet.

So:  First mix one aliquot of beads with DNA (Invitrogen says to do this in TE + 1 M NaCl).  To start I'll just use one size of beads, arbitrarily the 1 µ ones.  I know that the stocks contain 10 mg beads per ml, but I don't know how many beads this is.  I was going to find that out, by diluting some beads and looking at them under the microscope (we have a hemocytometer).  But here's a rough calculation:  If the density of the beads is a bit higher than that of water, then a single bead 1 µ in diameter will have the same mass as a 1 µ cube of water, which is 10^-9 mg.  So 10 mg of beads/ml is ~10^10 beads/ml.  As I did with the polystyrene beads, I'll wash the beads several times to remove unbound DNA.  I don't think I need to check that DNA has bound by using the sensitive fluorescence assay - I'll leave that to do if I don't see a difference in cells associated with beads with and without DNA.

Then I'll thaw some frozen competent H. influenzae cells, wash away the glycerol and resuspend them in BHI.  Then I'll  mix them with DNA-treated beads and control (no DNA) beads, at a concentration of, say,  ~ 5x10^8 cells and beads per ml.  I'll incubate cells plus beads briefly (1 min at 37°C?) and then wash the beads three times, saving the eluate/supernatant/whatever it should be called.  Then I'll plate the beads (several dilutions) and the eluates.  I'll also look at the beads under the microscope (maybe do that first).
  1. Make washing buffer, wash beads, resuspend in BHI, in MIV and in PBS.  
  2. Check under microscope to see if they're clumping.
  3. Incubate beads with DNA, wash well.Thaw and wash cells.  Resuspend in BHI or MIV.
  4. Mix cells with beads, incubate briefly and wash.
  5. Assess binding by plating and/or microscopy.

Revisions almost done

Today I finally finished what I very much hope is the last major revision of the uptake-sequence variation manuscript.  I've rewritten half of the Introduction and all of the Discussion, using my new non-adversarial, everyone-wins framework.  I've pulled together all the data for the supplementary table that serves only to show how quantitative our results are.  I've redrawn one figure with new data (same results as the old data).  I've gone back through the Responses to Reviewers, changing the responses to match what we've now improved.  I've started to draft a cover letter to the Editor.

Then I emailed everything to both my coauthors and to the three people in the lab (postdoc, RA and visiting researcher), asking for only essential changes and polishing of the writing.  With luck it will be resubmitted sometime next week, and with even more luck the critical reviewer will find our changes acceptable.

Now I've run out of excuses for not cleaning up my office.

Framing the uptake sequence problem (Intro and Discussion)

I think the last post may have been a bit incoherent, but it led me to a new perspective on the problems posed by uptake sequences, one that I think gives a much better frame for the manuscript.

INTRODUCTION (new frame):

(After introducing uptake sequences and uptake biases...)

Why bacteria take up DNA is controversial, and presence in two bacterial groups of DNA uptake sequences and their associated uptake biases pose problems for both major hypotheses.

It's generally assumed that bacteria take up DNA to get benefits from homologous genetic recombination, and that uptake sequences plus biases are a mate-choice adaptation to maximize these benefits by excluding DNAs that are not from close relatives.  Although this is intuitively appealing, it is evolutionarily problematic, both because it requires simultaneous evolution of bias in the uptake machinery and genomic sequences matching this bias, and because the genomic sequences can only be 'selected' after the cell carrying them is dead.  (There's also the bigger problem that the presumed benefits of recombination are expected to be, on average, very small or nonexistent.)

The alternative hypothesis is that bacteria take up DNA as a source of nutrients (initially nucleotides but also carbon, nitrogen and phosphate), for which the very existence of uptake sequences plus bias is counterintuitive.  If DNA in the environment is valued only as nucleotides on a string, all DNAs should be equally useful.  Although the sequence bias might play a mechanistic role in DNA uptake (such biases are typical of proteins that bind DNA, even ones whose functions are sequence independent), the high density of the preferred sequences in the genome is perplexing.

The phenomenon of molecular drive may resolve the worst of these problems for both hypotheses, by providing a hypothesis-neutral explanation for uptake sequence abundance.

(Explain molecular drive here.)

If molecular drive is indeed an inevitable consequence of biased DNA uptake and homologous recombination, its action may remove the biggest obstacles for both hypotheses.  Below we use a computer simulation of genome evolution to test its requirements and consequences.



DISCUSSION (new frame):

Summarize the findings.  They are robust.

What's been gained:  Proponents of the mate-choice hypothesis now need only explain how natural selection for the benefits of recombination would favour uptake specificity in the genes encoding the uptake machinery - the corresponding uptake sequences will inevitably accumulate in the genome as the specificity strengthens.  Proponents of the DNA=food hypothesis need only explain how sequence bias would evolve for mechanistic benefits in DNA uptake; uptake sequences in the genome can be ignored.

The above paragraph is really too adversarial a perspective.  It's now much clearer what information is needed to explain uptake specificity.  First we need a much more detailed characterization of the real uptake biases (Neisserial and Pasteurellacean).  Second we need to know what role uptake bias plays in the process of uptake in each organism.  Does a dedicated cell-surface protein pre-screen DNA fragments for uptake sequences before uptake is initiated?  Do uptake sequences provide structural flexibility for DNA bending or kinking during initiation of uptake?  Do uptake sequences play any role after initiation?  Do they affect DNA synapsis or other stages of recombination?

We also need new explicit models of the evolutionary forces that would act on uptake genes and preferred sequences.  Can selection for genetic benefits of recombination be strong enough to cause evolution of uptake bias?  Or, vice versa, can exclusion of unrelated DNAs reduce the costs of DNA uptake?  Analysis of protein sequences in genomes with and without uptake sequences suggests that their evolutionary costs are small, but a theoretical framework for this is lacking.  Because the model presented in this paper tracks only a single focal genome, it is not suitable for investigating effects on organismal fitness (whether due to the costs of uptake sequences or to the genetic benefits of recombination).

Discussion draft

(Here I'm just trying to lay out the philosophical framework of the Discussion.)

Uptake sequences might seem to be a minor trivial problem (only two relatively unimportant bacterial groups), but they  have important implications for the evolution of sex.  The key question is, are uptake sequences (= sequence-biased DNA uptake plus abundant preferred sequences in the genome) evidence of selection for benefits of homologous recombination?

Start with recap of what needs to be explained.  It's difficult to disentangle the hypotheses:  Are bias and accumulation the two interdependent components of an adaptation to promote uptake of fragments that can contribute to beneficial recombination, as often assumed?  The big problems with this are the weakness of the hypothesized benefits of recombination in current models of the evolution of sex, and the difficulty of selecting for sequences that only act after the owner is dead.  Alternatively, each component (uptake bias and uptake sequences in the genome) could have a separate explanation; for example, uptake sequences might have a cellular function, and sequence bias might play a mechanistic role in uptake.

But an acceptable model for bacterial uptake sequences must explain the very strong correspondence between the sequences that are preferred by the uptake machinery and those that are overrepresented in the genome.  Given an abundant sequence with a cellular function, might bias favouring it evolve because of genetic benefits of homologous recombination (mate choice), or because of mechanistic benefits of evolving high-affinity DNA-binding proteins specializing in a commonly available sequence?  Conversely, given an intrinsically biased uptake machinery due to mechanistic constraints or the need for high-affinity DNA binding, might the preferred sequences accumulate in the genome regardless of recombination benefits?

This last is the simplest hypothesis to test.  Our results show accumulation of uptake sequences like those in real genomes, provided only that DNA uptake is strongly biased and homologous recombination sometimes occurs.  The results don't require that the homologous recombination provide any genetic benefits.  However, the effects we have found will act regardless of whether the benefit to the cell comes from the DNA's nucleotides or its genetic information.  They also don't need selection for homologous recombination, as that occurs anyway in repair-capable cells

So now we have shown that, provided DNA is taken up and bias exists, uptake sequences will accumulate.  If DNA uptake is selected because it provides food, then uptake sequences have not evolved as markers of sequence homology, and we expect to find that uptake is biased for non-genetic reasons.  If DNA is instead taken up for its genetic information, we need to study how the accumulation of uptake sequences affects the genetic benefits, and how the benefits can affect the evolution of the gene causing the bias..

Our finding that uptake sequences accumulate without selection removes one of the problems with the uptake-bias for sex model (that of needing to simultaneously select for bias and uptake sequences.  Instead now just need to test whether uptake bias will evolve (gradually) because of benefits of homologous recombination, with uptake sequence accumulation following passively along.  Our model of uptake sequence evolution is not designed to evaluate this because it is not population-based (it tracks only a single focal genome).  (Can I mention here my new modeling of the effect of selection at one position?)

An alternative approach is to investigate the role bias plays in the uptake mechanism.  Finding that bias is created by a mechanism-independent protein that pre-screens sequences for uptake sequences would support the hypothesis that bias exists to promote homologous recombination.  Conversely, finding that bias makes a mechanistic contribution to the process of uptake would be consistent with non-genetic functions of DNA uptake.  Our present focus is on properly characterizing the true uptake bias of the H. influenzae uptake system,  identifying the gene or genes responsible for the bias, and finding out the role of sequence biases in the mechanism of uptake

Still (STILL!) working on the uptake sequence variation manuscript

(If the damned thing takes much longer we're going to have to ask the Editor for an extension!)

It's my fault - I'm stalled at fixing up the Discussion.

The problematic reviewer felt that the model didn't make testable predictions, so I wanted to include a brief discussion of how it could be used to evaluate more complex hypotheses about uptake sequence evolution.  Unfortunately, a proper test that includes selection will require the model to follow a population of cells or genomes, rather than a single focal genome.  Such a model would necessarily be  more complicated than ours, and if it was set up like ours the run times might be prohibitively long.  (Of course a clever programmer might find ways to streamline it without losing scientific relevance.)

But I thought of a simple test of whether uptake sequences accumulate near to positions that are under fitness selection in the diverging sibs of the focal genome.  Modifying the program to do this took only about 12 lines of code, but getting this code to work properly took me most of a day (spent chasing curly brackets and finding out the correct way to use the Perl 'substr' function).

This new version of the program includes selection at position 5000 of the evolving genome.  As before, each DNA fragment in the environment is first scored for sequences matching the uptake motif, and the resulting score determines its probability of recombining with the focal genome.  But now, if the fragment overlaps position 5000, it is also checked for its base at that position.  Fragments with an A at position 5000 keep their original uptake-sequence score for the recombination step, but fragments with the other bases have their scores reduced by a factor of 0.7, 0.4 or 0.1 (for G, C and T respectively).  This is intended to simulate poor survival of cells with these bases.  If the focal genome has an A it will have a near-normal recombination around position 5000 (except for the 1/100 fragments that carry mutations to less favoured bases), but if it has one of the other bases it will have reduced recombination except for the higher recombination of fragments carrying mutations there).  Recurrent mutation at position 5000 (in the focal genome and in the divergent fragments) may create a recurrent benefit of recombination, and if uptake sequences promote beneficial recombination, might select for uptake sequences close to position 5000.  On the other hand, if recombination more often brings in harmful mutations, uptake sequences close to the selected position might be selected against.

So I examined the final locations of uptake sequences in genomes from a bunch of runs that started either with 10 kb random-sequence genomes or with 20 kb genomes pre-seeded with uptake sequences (one very close to position 5000).  In the random-sequence genomes there were just as many uptake sequences around position 5000 as anywhere else, and in the pre-seeded genomes the uptake sequence at position 4982 was no more and no less stable than any other uptake sequence.

This isn't a very good test, in lots of ways (in fact it's quite awful), but I think it will show the Editor that  the model is indeed testable, and that we have made a reasonable effort to satisfy the reviewer.  In the manuscript's Discussion I'll describe it in less detail than I have above, and I won't present any data ("Redfield, unpublished").  And I'll explain that a proper test that incorporates selection for beneficial alleles will require a population-based version of the model.

[I also still have to assemble some new data into a replacement for one of the figures, and to go back over the latest changes one more time before sending them to my coauthors on last time.]