Field of Science

What am I doing? What should I be doing?

Here's an attempt to get better focused and organized:  (Trigger alert: most of this post is not about the real science but about the tiresome details that get in the way.)



What I've been doing:

1. Preparing for a big mutagenesis project:

I want to isolate a lot more hypercompetence mutants.  I have a plan, and I've pre-tested the mutagenesis step, but a key component is a way to efficiently transform rare competent cells in a large population of non-competent cells.  The best way to do this is with a cloned or PCR'd DNA fragment containing a selectable point mutation giving antibiotic resistance.  In the past I've used a cloned novobiocin-resistance fragment.  I had thought that a PCR'd fragment would work just as well, but tests show it to be only twice as good as chromosomal DNA (it should be at least 10-20 times better).

So I've been struggling to get good plasmid preps.  I have two plasmids carrying this 9.4 kb novR fragment.  One is the original clone, in an old natural ampicillin-resistant plasmid called pRSF0885.  I have very old plasmid DNA preps, and some frozen cells from 1990 with the plasmid.  The cells are still viable and resistant to ampicillin but my plasmid minipreps give no plasmid DNA.  The old plasmid DNA efficiently transforms cells to AmpR, using a method that should produce plasmid transformants rather than replacement of the chromosomal allele, but again my plasmid preps give no or almost no plasmid DNA.  I also have the same novR insert cloned into the CmR vector pSU2718 (almost identical to pSU20), but the plasmid yield is miniscule.  So I'm beginning to suspect that something may be wrong with my plasmid prep method - I need to streak out something foolproof and prep it.

2. Tracking down the cause of non-reproducible colony counts:

Counting colonies is our most important research technique (yes, in some ways we're a very low tech lab).  Normally the results are nicely reproducible, but lately I and the PhD student have been seeing bizarre discrepancies between replicate plates and between different volumes or dilutions of the same culture.  After eliminating other variables (plates too dry? plates too wet? new brand of plates (nasty VWR plates with rounded edges?)? dirty spreading beads?) I did a test of 8 different combinations of different sources of BHI medium and of agar.  This suggested that the problem may be with the old BHI agar we've been using (we were given two big buckets of it by a colleague).  The good news is that this BHI agar is almost all gone, so the problem will solve itself.  The bad news is that this BHI agar is almost gone, so we're going to have to start paying for more!  (And I'm not confident that the problem is completely solved - there are still discrepancies, and tracking down the cause of irreproducibility is inevitably an exercise in frustration.)

What I should be doing:

3. Finishing up the honours student's Actinobacillus pleuropneumoniae experiment:

One of last year's honours undergrads left us with an almost-ready-to-submit manuscript lacking only some RNA-seq analysis and a remake and retest of one mutant.  The other honours undergrad (now our summer student) has been working on the RNA-seq analysis, but we're still waiting for the final missing sequencing data.  My job is to remake the mutant and test its competence phenotype.

The problem with her original mutant was that misannotation of a gene overlap had resulted in a mutant that deleted the last five amino acids of a gene that needed to be intact.  Last month I spent ages designing primers to delete the correct segment, but the primers ahve jsut been sitting on my bench.  I need to reconstitute them and (tomorrow) get instructions form the PhD student on where the PCR reagents are and how to use the PCR machine.

4. Learning R:

Our Life Sciences Centre R group is holding lots of workshops and help sessions, but by bad luck they've mostly been at times I can't attend.  The summer student has given me some R homework, but it's buried on my desk.

5. Clean my desk:

To find the R homework.

Completing the toxin/antitoxin project

The honours undergraduate has finished her work on the toxin/antitoxin genes, and has submitted and successfully defended an extraordinarily good thesis.  The thesis is written as a polished scientific paper, which is ready to submit except for some gaps in the data.  My job now is to fill these gaps.

She began by confirming the mutant phenotypes. We've already shown (Sinha et al. 2012) that the toxin knockout prevents DNA uptake as well as transformation. Under competence-inducing conditions the antitoxin knockout prevents transformation, and the toxin knockout and double knockout both have normal transformation.  This is consistent with our hypothesis that the Toxin protein does something that prevents competence development, and the Antitoxin protein prevents this.

Her next step was a phylogenetic analysis of the toxin and antitoxin genes.  This work is complete. She identified three groups with the gene pair, and used synteny analysis to show that the pair entered these groups by independent horizontal transfer events.
Because there is no outgroup she was unable to root the tree, so the order of transfer events is not established.

Actinobacillus and Haemophilus species are all members of the Pasteurellaceae, and A. pleuropneumoniae has a competence system very like that of H. influenzae.  Her next project was to knock out the A. pleuropneumoniae toxin and antitoxin homologs, so see if they affect competence in the same way the H. influenzae homologs do.  She used PCR to amplify and clone the genes with flanking DNA, and then used inverse-PCR to create linearized versions of this plasmid lacking either or both of the genes.  (I'd better diagram this out because I'm going to have to do some similar work.)


She then ligated SpcR cassettes into the linear PCR products, creating circular plasmids with the SpcR cassette replacing each (or both) genes, and transformed these back into the A. pleuropneumoniae chromosome, selecting for the cassette.

Once she'd made her knockout mutants she checked their competence phenotypes, and found that all were normal.  She would have concluded that the A. pleuropneumoniae toxin does not block competence development, but in the interim she had discovered that the GenBank record for the toxin gene was incorrectly annotated, missing the last 5 amino acids because these overlap with the start of the antitoxin gene.  This meant that the antitoxin knockout she had created was also missing the end of the toxin gene, and therefore might have actually been a double knockout (if the terminal toxin amino acids are important for its function).

So one of my jobs is to redo the antitoxin mutagenesis, using a new primer that preserves all of the toxin gene, and check the growth and transformation phenotypes of this new correct mutant.  If I find that it has a transformation defect I'll also check its DNA uptake.

 She then examined the growth properties of the H. influenzae and A. pleuropneumoniae mutants.  The A. pleuropneumoniae mutants all grew normally, but the H. influenzae antitoxin knockout grew significantly slower in log phase than the other H. influenzae strains.  This suggests that the H. influenzae toxin, when unopposed by antitoxin, interferes with cell growth even under conditions where the cells are not competent.  The toxin knockout also grew a bit slower.

She then examined the RNAseq data for the wildtype and knockout strains.  Each culture was sampled and sequenced at four time points: T0 is log phase in rich medium, just before transfer to the competence inducing medium MIV, and T1, T2 and T3 are 10, 30 and 100 minutes after transfer to MIV.  

Here we discovered another problem – two of the ‘antitoxin-knockout’ cultures used for these analyses were incorrect - they were instead knockouts of a different competence gene, comN.  So we have three replicates of the toxin and double knockout strains, but only one correct replicate of the antitoxin knockout strain.  The two missing replicates have been recultured and their RNA preps remade, and they are now being sequenced.  My contribution here will be to help the summer student finish the analysis of this data.  (By 'help' I mean pester him with questions and requests for explanations...)

Analyzing the data she had showed that log phase expression of the toxin mRNA increases dramatically when the antitoxin is missing - this is consistent with characterization of related toxin/antitoxin pairs, where the antitoxin protein represses transcription of the toxin/antitoxin operon.  She found lots of other interesting (confusing/perplexing) effects.  Since most of these are based on the single antitoxin-kockout replicate, we'll examine them more thoroughly once we have all three replicates.



Still planning the mutagenesis project

OK, I'm abandoning the old mutagenized stocks and starting the mutagenesis project from scratch.

I want to take advantage of having selectable alleles linked to each of the genes where mutations produce hypercompetence.  That's 1) the StrR point mutation linked to sxy (about 50 kb; 50% cotransduction), 2) the CatR cassette linked to murE (about 4 kb; 90% cotransduction?), and the CatR cassette tightly linked to rpoD (~100 bp; > 90% cotransduction).

Last time I tried to do this by doing the EMS-mutagenesis in vitro (See blog posts: here and here). Call this strategy C.   I directly mixed EMS with DNA of three strains carrying the above selectable markers and wildtype alleles of the hypercompetence genes, and then transformed the mutagenized DNAs into competent wildtype cells, selecting for each of the marker strR and CatR alleles.  But this failed.  I didn't get any low-level novobiocin resistance mutations that would have indicted that the EMS caused mutations, and I didn't get any hypercompetent mutants.  I suspect that transforming cells with EMS-damaged DNA is a very inefficient way to create mutations.

This time I'm going back to doing the mutagenesis in vivo (call this strategy B), incubating the three marked strains with EMS and allowing them to grow for 1-2 hr to convert the damage into mutations.  Then I'll isolate the DNA from each mutagenized culture, and use this to transform wildtype cells to the three markers.  This will give me pools of cells that have experienced a high mutation rate in the neighbourhood of sxy, murE and rpoD.

But before I do this I need to do the math, to see if this is really any better than just mutagenizing wildtype cells and screening them all for hypercompetence (call this strategy A).  Having the linked markers will certainly be handy later, once I've found hypercompetence mutants and want to find out which gene they're in.  But will using them as described above really let me find more mutations in these specific genes?

OK, I've laid out a situation with realistic numbers, and I've run it by the PhD student and the summer student.  Bottom line:  Strategy B does not enrich for mutations in the desired region.  It's only strength is that it allows use of much higher concentrations of EMS than would be tolerated in Strategy A.

Here's the numbers analysis for strategy A:

  1. Start with 10^9 cells.  (Below I'll consider whether fewer would be OK.)
  2. Treat with EMS (0.08M for 30 min).  Previous work suggests that this creates about 1 mutation per surviving cell.
  3. Grow cells for about 3 hr or more, keeping cell density below OD600= 0.2.  The goal is to allow enough time for the cells to recover from the DNA damage, undergo two rounds of DNA replication to convert some damage into G->A mutations, and express the mutant hypercompetent phenotype.  Assume that the cell numbers increase 10-fiold during this period.
  4. How many mutations will we have?  The genome has about 4 x 10^5 Gs.  Mutations at some of these will be lethal or sub-lethal, so assume about 2 x 10^5 positions where mutants have normal or near-normal growth.  With 10^10 cells, we will have about 5 x 10^4 occurrences of each mutation, on average.
  5. How many hypercompetence mtuations will we have? 10 of these are positions we have already found to be sites of hypercompetence mutations, so in our 10^10 cells we'll have at elast 5 x 10^5 hypercompetent cells.
  6. To select for these hypercompetent cells, transform the OD600=0.2 cell population with the 8 kb NovR DNA fragment from plasmid pRRnov1.  This DNA transforms much better than the equivalent PCR fragment or than chromosomal DNA with the equivalent mutation.
  7. The normal cells will transform at a frequency of about 10^-8 - 10^-7.  (I'm guessing here; with chromosomal DNA it's ~10^-9.) That would give about 100-1000 novR colonies from the 10^10 cells.
  8. The hypercompetent cells will transform at higher frequencies:  rpoD: cells with the known rpoD mutations will probably have a TF of 10^-4 - 10^-3, so the 5 x 10^4 cells of each mutation would give about 5-50 novR colonies.  sxy:  cells with the known sxy mutations will probably have a TF of 10^-3 - 10^-2, so the 5 x 10^4 cells of each mutation would give about 50-500 novR colonies.  murE:  cells with the known murE mutations will probably have a TF of 10^-2 - 10^-1, so the 5 x 10^4 cells of each mutation would give about 500-5000 novR colonies.   
  9. According to this analysis, most of the colonies will be hypercompetent mutantx, and most of the hypercompetence mutants we find will probably have their mutations in murE. Of coulrse this only considers the hypercompetence mutations we already know about.
Things to do first:  
  1. Check that our EMS stock is still good, by mutagenizing wildtype cells and scoring low-level novobiocin resistance.  A resistance frequency of ~5 x 10^-6 is approximately one mutation per viable cell.
  2. Make a big prep of pRRnov1.  The yield of this plasmid is often poor so this may take several attempts and large volumes.  Check the transformation frequencies and efficiencies this DNA gives.

Are the old mutagenized cells worth using?

Last month I wrote a post about an old experiment (What can I recover from an old failed experiment).  I concluded that some of the frozen mutagenized cells might be worth using, but that I would first need to test how effectively they had been mutagenized.  Now it's time to do that.

The cells were incubated with the mutagen ethane methylsulfonate (EMS), which alkylates G bases and causes G->A transition mutations.

I can test the efficiency of this mutagenesis by plating cells on novobiocin agar plates, using a lower-than-normal novobiocin concentration (1 µg/ml rather than 2.5 µg/ml).  I also need to test the overall viavbility of the frozen cells, by plating them on plain agar plates, but I'll need to do this anyway to test the mutagenesis. 

I have two tubes of each of six cultures (three strains at two EMS concentrations).  Each was grown for 90 min after the EMS treatment, to allow DNA replication to convert the DNA damage to mutations:
  • B and C: KW20 (wildtype), 0.05 mM and 0.8 mM EMS
  • D and E: RR514 (StrR, linked to sxy), 0.05 mM and 0.8 mM EMS
  • F and G: (RR805 (CmR, linked to murE), 0.05 mM and 0.8 mM EMS

I'll thaw and test one tube of B and one of C; these should be representative of the others.

Rather than just discarding the remaining cells I thawed, I could at the same time put these cells through the next step, by growing them at low density and transforming them with either novR chromosomal DNA or a novR PCR fragment (better because higher transformation efficiency).  But I won't do this, because first I need to find the novR PCR fragment and test it.

I just checked my old notes.  In expt #1290 I had tested the PCR products (novR and nalR) and found that they gave only slightly higher transformation frequencies than MAP7 chromosomal DNA.  This is surprising, since the effective concentration of the resistance-conferring fragments is much higher in the PCR DNA prep (maybe 10-50-fold higher, depending on the concentration of the PCR DNA, which wasn't measured.  The MAP7 DNA was used at a concentration that's saturating for transformation.

BUT, on more carefully re-reading my old notes (as usual not as limpidly clear as I would desire), I can't be sure that the cells I have saved are the cells I need (they might instead be cells that have already been incubated with the wrong DNA).  I think I'd better abandon this mess and start fresh.

Wildtype strain weirdness

While looking over some of the RNA-seq analyses done by our summer student (former undergrad, someday grad student somewhere), I noticed something unexpected.

We know that cultures in the rich medium sBHI become moderately competent when they reach high density.  Their transformation frequencies reach about 10^-5 - 10^-4, which is 100-fold lower than fully induced cultures but 1000-fold higher than log phase cultures.  Consistent with this, in the microarray analyses we did about 12 years ago we saw modest induction of all the competence genes except ssb in this condition (we didn't publish the data but described it as 4-20-fold induction).

So I expected to see similar induction in in rich medium in the RNA-seq data.   But there's no consistent induction at all; most genes don't show any change at all, and a few go up or down a bit.  You can look at the complete data here, but unfortunately the individual graphs are very low resolution.  Here's a blowup of one pair of graphs, for comE, which encodes the secretin pore:

On the left are cells in the competence-inducing medium MIV.  We see very strong induction of comE in wildtype (KW20) cells (brown dots and line), and no induction when the Sxy regulator is knocked out (blue dots and line).  On the right are cells in rich medium.  Here we see strong induction in the presence of the sxy1 hypercompetence mutation (blue dots and line) but no induction at all in the wildtype KW20 cells.

So the summer student did more analyses.  First he did more work with the RNA-seq data:

1.  The baseline (log phase) levels of expression of each competence gene are the same in cultures grown for the rich-medium experiments and cultures grown for the MIV-induction experiments rich medium.  This suggests that there was nothing very wrong with the rich medium cultures.  (The rich medium experiments took their log phase samples from very dilute cultures (OD600 = 0.02), and the MIV-induction ones took theirs at OD600 = 0.2.  Later we plan to use this to test whether both densities are genuine log phase, by seeing if this density difference changes expression of any genes at all.)

2.  The sxy gene is slightly induced as culture density increases, though we don't really know how much it should be induced.

3. The genes required for induction of the competence regulon are intact in the reads of the rich-medium cultures: sxy, crp, cya.  That means we didn't accidentally use a strain carrying a knockout of one of these genes instead of wildtype cells.

4.  To check whether the supposedly wildtype strain might have a knockout of another gene, he looked for reads derived from antibiotic-resistance cassettes.  He found a few reads of a spectinomycin cassette (like the one we have used for many of our knockouts), but the numbers were so small they're probably just contaminants.

Then he managed to do comparisons between the RNA-seq data and the old microarray data:


The blue-line graph shows the microarray data.  The Y-axis is fold-change in expression of each competence gene, and the X-axis is time relative to when a separate sample was removed for competence induction.  The red-line graph shows the RNA-seq data, squished to make the spans of its axes roughly consistent with those of the other graph.  The Y-axis is again fold-change in expression, but now the X-axis is the density of the culture, measured as OD600.  

In the microarray data, some genes aren't induced at all but others are induced as much as 11-fold.  In the RNA-seq data, only one gene is induced even 2-fold, and many are down-regulated.  So we definitely do have a problem.

So then it became my turn to do some experiments to figure out what's going on.  Fortunately I had saved frozen samples of the cells used for every RNA sample we analyzed by RNA-seq.  

First I thawed out and transformed the three samples of supposedly wildtype cells at OD600 = 1.0. Their transformation frequencies were all a lot lower than they should have been ('FK': 3.6 x 10^-7; 'GK': 2.5 x 10^-7; 'HK': 7.6 x 1-^-7, rather than about 10^-5).  That's consistent with a genuine lack of induction of their competence genes (and not with my alternate hypothesis, that there was just some error in the analysis of this RNA-seq data).

I streaked the cells on spectinomycin plates, to check if they had an unexpected spcR cassette.  None grew, and they all grew on the control plates.  So we didn't use any of our spc-cassette knockout mutants by mistake.

Finally I inoculated two of the supposedly wildtype strains ('GK', from a plain-plate colony of the above transformation test, and 'FK', from its OD600 = 0.02 frozen sample) into rich medium, along with a wildtype control strain, and tested competence development under three conditions.  The left columns show the transformation frequency seen 60 min after adding 1mM cAMP to a log-phase sBHI culture - I did these in case we didn't get normal transformation in the other tests, since this would tell us if the strains were somehow unable to produce cAMP (e.g. if they had a phosphotransferase mutation).  The middle columns are cells at high density in sBHI; both of the suspect strains have near-normal transformation frequencies.  The right columns are cells transferred to the competence-induction medium MIV; one of the suspect strains has normal transformation and the other is down 10-fold.


(I must admit that I don't have high confidence in the numbers from this experiment, for several reasons.  First, the colony sizes and counts were very erratic (inconsistent from one dilution or plating volume to another), perhaps because I used mostly old novobiocin plates left by our now-gone Co-op tech. Second, the 'latelog' samples were allowed to grow longer than I intended, so their competence may have been decreasing, and this problem was slightly worse for the GK and FK cultures.  Third, I was testing a new competence protocol (see p.s. below).)

So what have we learned?  Mainly that there's nothing obviously wrong with the 'wildtype' cells used for the sBHI RNA-seq samples.

So what should we do next?  I don't know.  Maybe I should repeat the competence-induction tests with fresh plates, to get better numbers.

p.s.  The KW20 MIV transformation frequency is slightly lower than I usually see, probably because I was testing a new scaled-down protocol that doesn't use an expensive disposable filter funnel to collect and wash the cells.  Our usual protocol is to collect and wash 10 ml of cells using a Nalgene disposable filter funnel (0.2 µ size, designed for water sampling), and then resuspend them in 10 ml of MIV in a flask shaking in the waterbath for 100 min. But my new attention to economy has revealed that each funnel new costs nearly $7.  So this time I just pelleted 2 ml of cells in a microfuge tube, resuspended the cells in 1 ml MIV, pelleted them again, and resuspended them in 2 ml of MIV in a large glass culture tube, which I incubated on the roller wheel in our air incubator.  The transformation frequencies for KW 20 and GK are plenty high.  The lower transformation of the green 'FK' sample may be because I ran out of MIV and skipped its washing step.