Field of Science


If I do an experiment that uses 50 nm magnetic beads, am I using nanotechnology? If the experiment involves a new use for these beads, am I developing nanotechnology? More generally, is all molecular biology nanotechnology?

Anyway, what am I going to do with these beads? I want to find out whether the USS polarizes the direction of DNA uptake. That is, when the USS sequence interacts with the DNA uptake machinery on the cell surface, does only the DNA on one side of the canonical orientation get pulled into the cell, at least until the end has been brought in?

My latest model of the mechanism of uptake initiation predicts that at the initiation step a type 4 pseudopilus pulls in the DNA on the right side of the USS core, while the core is bound tightly to a receptor protein on the surface (this could be the secretin pore-forming protein). If the core remains tightly bound, uptake could continue only on this side until the end of the fragment is reached. This would require that the force exerted by the pseudopilus on the DNA be less that that needed to pull the USS core free from its receptor. Bringing in the end of the DNA would allow transport across the cytoplasmic membrane, which would reverse the direction of uptake. Under this version of the model, this reversal of direction and the stronger pull exerted on the DNA by the cytoplasmic membrane machinery would detach the USS core from the receptor and bring in the DNA on the left side.

This model of uptake is much more detailed than the available evidence supports. I see this as a strength, not a weakness. The model makes many very specific and testable predictions, and at this stage of investigation it's more important for a hypothesis to be testable than to be correct. Maybe that's true at any stage.

I still like the idea of using laser tweezers to investigate the polarity question, but my physicist collaborator points out that this single-molecule method may lack the resolution we need to answer the question. Tweezers are best at measuring forces, not movement of cells.

So I'm back to thinking about blocking the ends with beads. Our previous attempts used what we now realize were giant beads (streptavidin-agarose), far bigger than the cells and much too big and porous for the task. This improved approach will use beads that are only 50 nm across, which should be plenty big enough to prevent uptake as the secretin pore is thought to be only 6-7 nm across at its widest. They can be purchased with streptavidin bound to their surface, making it easy to attach them to DNA just as we did with the agarose beads.

Our original experiments labeled one end of a short USS-containing fragment with 32P and blocked the other end with a (giant) bead. We predicted that the orientation of the the USS with respect to the bead would determine whether the 32P got inside the cell, but found that orientation had little effect. I think we should repeat these experiments using the 50 nm magnetic beads, paying careful attention to the kinetics. The original experiments used 220 bp fragments; because such short fragments might behave differently than long ones, we should also repeat the analysis with fragments that are one or a few kb long (easy - we just use the whole USS plasmid rather than its 220bp insert).

Maybe we can then combine use of the beads with the tweezer analysis, examining forces rather than distance changes as originally planned. If the tweezer analysis of the force acting on the DNA shows that the inner membrane machinery exerts a stronger force than the pseudopilus, then blocking the right end of a long fragment with a bead (left end held in the laser trap) should allow weak-force uptake but prevent the transition to strong-force uptake. Flipping the orientation of the USS should. . . . . no, the transition would still be blocked, this time by the large trapped polystyrene bead at the other end. The flipped USS orientation would affect movement of the cell rather than the kind of force.

How bacterial conjugation works

I just had a lively discussion with the guy in the next office about whether the single-stranded DNA that enters a bacterial cell by conjugation becomes double-stranded before it recombines into the chromosome. The textbook he's using to teach his new bacterial molecular genetics course say yes, but this seems rather improbable to me.

Another faculty member found a paper from about 10 years ago that concludes that the DNA becomes double-stranded before it recombines. But I don't think this double-strandedness was necessitated by the findings of their experiments; rather it seems to be mainly driven by the assumptions of the researchers.

The paper had another surprise. It showed the DNA being transferred into the recipient cell as a loop, with the 'leading' end being held back in the 'donor' cell, and the loop being pushed into the recipient. I've only seen it represented differently, with the leading end moving into the recipient cell as if it was pulling the DNA strand along behind it. But now I think more carefully, the protein attached to the leading end isn't expected to have any power to 'pull' anything once it's in the recipient cell, so the DNA must be being transported just by the machinery in the membrane of the 'donor'.

I'm putting 'donor' in quotes, because that's how most people think of the process of conjugation. They've been taught to think of the cell that has a plasmid kindly donating a copy of it to a cell that lacks the plasmid; if the plasmid is attached to chromosomal DNA the recipient cell is blessed with a copy of part of the donor's chromosome. But plasmids are genetic parasites, and the 'donation' is really an infection. The recipient is passive, and the plasmid DNA is forcibly inserted into it by the 'donor'. To do this the plasmid uses the proteins it codes for, and proteins already produced by its cell.

Why is H. influenzae's CRP so feeble?

One discovery from the grad student's work on Sxy and CRP is that the H. influenzae CRP protein binds CRP sites much less strongly (with much lower affinity) than the E. coli CRP protein does. This is a bit surprising. The two proteins have quite similar sequences, and all the amino acid residues expected to directly contact the DNA are identical.

He's going to contact a lab that has done extensive structural analysis of E. coli CRP, to see how difficult it would be to see how well H. influenzae CRP will superimpose on the E. coli structure. One possibility he suggested is that the dimerization domain of H. influenzae CRP could be weak. This would cause the protein to spend less time assembled into the dimers that most readily bind DNA.

Until now we (at least I) had thought that the affinity of CRP for different genes was determined by how well the gene's CRP site matched the protein's requirements for binding. This would have been optimized for each gene by natural selection acting on mutations in its CRP site. But now I'm wondering whether natural selection has also acted differently on E. coli and H. influenzae CRP proteins to fine tune their affinity for all the sites in their respective genomes.

H. influenzae has only about 40% as many genes as E. coli, and about 40% as many CRP sites regulating them. But I can't think of any way that would favour a 100-fold difference in CRP affinity for the same CRP site, which is what the grad student has found.

The assays were done under exactly the same conditions, but this doesn't ensure that the proteins responded identically to these conditions. I wonder if the binding conditions used for these measurements (optimized for E. coli CRP) might be unsuitable for H. influenzae CRP.

What part of "match" don't we understand?

The overall goal of the work the post-doc and I describe in our Defining the USS manuscript is to find out whether the sequence specificity of the H. influenzae DNA uptake machinery matches the consensus of the uptake signal sequence (USS) repeats in its genome. We've done good jobs of characterizing both the uptake specificity and the USS consensus, but now we're discovering that comparing them is trickier than we had realized.

The figure shows both results, with the genome consensus shown as a SequenceLogo at the bottom, and the effect on uptake of changing each position shown by the bar chart above. Because all the positions in the core (the AAGTGCGGT at the left) have a very strong consensus in the genome, we expected that changing any of these positions would dramatically decrease uptake. But instead we see that changes to different core positions have very different effects. Changing the first position doesn't decrease uptake at all (within our limit of detection) but changing any of the central three knocks it way down.

One confounding factor comes from how the genome consensus is described; the Y-axis of the logo is not a linear scale but a logarithmic scale reflecting 'information content' of the consensus. Another may come from how the uptake effects were measured; cells were deliberately given less DNA than needed to saturate the uptake machinery.

But the biggest complication is that, although we have a simple 'molecular drive' hypothesis describing how biased DNA uptake leads to accumulation of the preferred sequence in the genome, we are only beginning to develop ways to evaluate the different components of this model. This means that we can't predict exactly what sequences will accumulate in response to any specific uptake bias.

The post-doc describes the discrepancy between the two parts of the figure as maybe resulting from 'saturation' of the evolutionary process causing USS accumulation. If a small uptake bias acting over millions of generations is enough to drive the preferred base at a particular position to a very high frequency, then a larger bias may not make much difference to the outcome. For example, say USSs with a C or G or T at position 1 are taken up 95% as well as a USS with the consensus A at that position, but this bias provides sufficient long-term drive to cause 98% of the USS population to have As at that position. If so, increasing the bias might not increase the frequency of As by very much.

So what will we say in our manuscript? We could start by discussing the implications for the role of the USS in DNA uptake. We'll point out that uptake is very sensitive to the central positions of the core. Even if the other positions in the USS (core and flanking AT-rich segments) are all perfectly matched to the genome consensus, changing any one of these central bases drastically reduces uptake. This suggests that these positions make very important contacts with the uptake machinery. Changes at any of five other core positions reduce uptake to 20-60% of the control (perfect USS), so these also probably make important contacts. Changing the initial A, or pairs of bases in the flanking segments, reduces uptake only modestly, suggesting that the functions of these in uptake are dispensable if the rest of the USS is perfect.

We could go on to discuss the implications for accumulation of USSs in the genome, as I do above, and end by pointing out that what's needed is a better understanding of the consequences of molecular drive (and maybe put in a plug for our Perl simulation model).
The post-doc and I got about half-way through our Defining the USS paper today. We fixed up the Methods section and almost all of the Introduction, and walked ourselves through the Results. Maybe because we hadn't looked at them for a couple of weeks, we had some new ideas.

We had been commenting on how the USSs in coding sequences show significantly different USS motifs, depending on which reading frame they're in. But now we look again, we realize that the real story isn't how they are different, but how similar they are. Despite coding for entirely different amino acids, they all have strong matches to the canonical USS core and flanking segments. The differences are so small that they might even be attributable to random effects due to small sample sizes.

This might mean that needing to code for proteins doesn't significantly constrain USS sequences. Rather, USSs accumulate only in places where the existing protein coding constraints meet their need for uptake efficiency.

We also clarified our analysis of the correlations between the bases at different positions in the USS, especially for USSs that are likely to be acting as terminators. Because terminators act by folding into hairpins, we expected to see that the parts of these USSs that come together when folded would show correlations. But they don't. Instead the strong correlations are all between bases that are adjacent in the sequence or separated by only a single base, just as they are in the non-terminator data set. We think this means that the correlations arise from the need for adjacent base pairs to physically cooperate when the USS is kinked during uptake, and that selection for terminator function is much weaker.

Defining the uptake signal sequence

The big goal for today is to make progress on the manuscript where we pin down the specifics of the Haemophilus influenzae 'uptake signal sequence' (USS). This DNA sequence can be defined in two ways, both as the sequence motif preferentially taken up by competent H. influenzae cells, and as a sequence motif that's strikingly overabundant in the H. influenzae genome. The two are of course related; we think that the abundance on the genome is a direct consequence of preferential uptake and random (unbiased) recombination between incoming DNA and the chromosome.

The most glaring hole in the manuscript right now is the Methods section where I need to describe the bioinformatics methods I've used to analyze the USS motif in the genome. The post-doc responsible for the DNA uptake experiments has already written her part of the Methods, and today we're going to sit down together and write my part, as well as work on polishing the Results (I think this is in OK shape) and the Discussion (needs work!).

Polishing the Sxy manuscript

We're fixing the final details on the Sxy manuscript; I'm hoping to have it submitted in the next 24 hours. Below I'll try to summarize what it says, in less technical terms than the Abstract uses.

We already know that the Sxy protein regulates expression of competence genes; here we're examining how Sxy itself is regulated. Our most powerful tools are regulatory mutations that turn Sxy on when it would otherwise be off. The paper starts by describing new mutations that, taken all together, strongly suggest that expression is controlled by changes in RNA folding. We conclude this because all of the mutations change how RNA can fold, but only one of them changes the Sxy protein sequence (and that in a trivial way).

The paper then presents data showing that the RNA folding changes don't affect how much Sxy RNA is made, but how efficiently the RNA is used to make Sxy protein. Then more data shows that normal cells translate their Sxy RNA more efficiently when they're starved, but the mutant cells translate it efficiently even when they're not starved.

The cell uses the genes/proteins that Sxy regulates to take up DNA, so these results help us understand how being starved causes cells to take up DNA. We have argued that cells take up DNA as food (not to get different versions of their genes as others have assumed), so these results strengthen the evidence that DNA uptake is an adaptive response to starvation.

What's UP, Sxy?

One of the grad students has been doing experiments to clarify how the transcriptional activator protein Sxy turns on genes that have its recognition sequence, the "CRP-S" site. He's given me a draft of a paper he's writing about this work.

Sxy works by interacting with another transcriptional activator protein called CRP. CRP's job is to bind to DNA at CRP sites and, by bending the DNA, help RNA polymerase start making RNA. CRP-S sites are much harder to bend than normal "CRP-N" sites, and we have been thinking that Sxy acts by helping CRP to bind (if it can't bend the DNA it lets go).

But several pieces of his data tell us that binding isn't enough. A protein from E. coli (we work mostly in another bacterium, Haemophilus influenzae) binds CRP-S sites fairly strongly in the test tube, but when it's in H. influenzae cells it can't turn the CRP-S genes on without Sxy's help. And changing the CRP-S sequence so H. influenzae's CRP can bind it without Sxy in the test tube doesn't enable CRP to turn the gene on in cells without Sxy's help.

So the grad student is suggesting that Sxy may also affect how RNA polymerase interacts with the DNA. RNA polymerase binds to DNA more effectively if its 'tail' makes contact with an AT-rich sequence called the "UP" sequence. Many genes have an UP sequence, and at normal CRP-N sites CRP is thought to help RNA polymerase's tail bind to it.

He's found that the genes with CRP-S promoters have what look like three precisely-spaced UP sequences beside their CRP-S sites, and in his draft paper he proposes that Sxy acts partly by helping RNA polymerase make contact with these sites. We could test this by changing these sequences to not fit the UP consensus, and seeing if this makes the promoter unable to be activated by Sxy.

This work doesn't address the question of whether Sxy does help CRP bend the DNA. But we won't be able to address this until we have conditions where purified CRP and Sxy both interact. Right now we have the CRP but not the Sxy - it's a very difficult protein to work with.

Welcome, visitors from the Just Science links

Unlike most of the other science blogs you might check out, this is a real 'research blog'. In it I write about the experiments and analyses we do in my lab, and about what I think they mean. I've been doing this for about 6 months, and I find that writing it helps me think more clearly about our work.

I don't expect casual visitors (or any visitors except those from my lab) to follow the complexities of the science, but it might give them a sense of what more-or-less ordinary* scientists do and how we think.

*No, of course I don't really think I'm any more ordinary than any other scientist (we all think we're special), but I'm a Canadian so I feel obliged to pretend.

DNase I: friend and foe

My posting has been a bit infrequent because I've been focused on getting drafts of our two grant proposals ready for pre-submission critiquing by the volunteer reviewers (blessed be their names and their progeny). But that's done, with a lot of help from a grad student and a postdoc. I've committed myself to the Week of Science Blogging, so I'll be posting at least daily for the next week.

We put a fair bit of effort into crafting a tightly focused "Significance" section to end each proposal (after the detailed descriptions of the experiments we propose). It's important to avoid ending with strings of platitudes, so ours is a bit on the edgy side. On the other hand we can't afford to sound too arrogant (this is a Canadian granting agency).

Some science: We had been hoping to be able to use preparations of cell membranes to study how a radioactively-labeled 'test' DNA fragment interacts with the uptake machinery. But we had overlooked the problem of all the cellular DNA that will be released when we break the cells open to get the membranes.

When the cells are first opened the uptake proteins will bind the abundant cellular DNA, and unless we get rid of the DNA they've bound, they will never bind our test DNA. In principle we could get rid of the cellular DNA by adding a lot of DNase I, an enzyme that breaks down DNA. But then we would need to get rid of ALL of the DNase I , as otherwise it will break down our test DNA!

The postdoc (who discovered the problem) is going to do some preliminary testing to see if this analysis can be salvaged, perhaps by repeatedly washing the membranes to get rid of the DNase I.