Spot 42 RNA doesn't explain our ∆hfq competence phenotype

Gisela Storz from NIH gave a great seminar here yesterday about the many roles that small RNAs and very small proteins play in E. coli gene regulation.  It gave me an idea about the role the small-RNA-accessory protein Hfq might be playing in competence regulation.

One example she discussed is the abundant Spot 42 RNA, so named because it was discovered as an RNA chromatography spot long before any function was discovered.  This small (109 nt) RNA acts in a feed-forward regulatory loop with the cAMP-dependent transcriptional activator CRP.  This regulation fine-tunes the control of the CRP-activated genes that provide the cell with alternative carbon sources when its preferred sugars are depleted.

When the preferred sugars are available, Spot 42 RNA is expressed and limits the translation of genes for using the alternative carbon sources.  Expression of these genes is already low because there is no active CRP to stimulate their transcription, and Spot 42 RNA prevents translation of any transcripts that might inadvertently get made.  The contribution of Hfq (the yellow donut) is to help Spot 42 RNA interact with its target mRNAs and prevent their translation.

But when the preferred sugars run out, active CRP represses transcription of the spf gene, so Spot 42 RNA isn't made. (Yes, I know I said that CRP is a transcriptional activator, but it can also be a repressor when its binding site is on top of or downstream of the promoter.)  Now the transcription that CRP stimulates produces mRNAs that are efficiently translated.

So in E. coli Spot 42 RNA improves the stringency of the regulation that we have been giving CRP full credit for.

Where does H. influenzae competence come in?  Our Hfq knockout lowers competence, and the other honours student showed that this is because it makes competence more dependent on high cAMP than it usually is.  So I got excited by the idea that maybe Hfq's role in competence had to do with its role in mediating Spot 42 activity on CRP-dependent transcripts.  This could either be activity on CRP-dependent transcripts of the competence-specific transcriptional activator Sxy, or on the various CRP-dependent transcripts of the DNA uptake machinery.

But this falls apart in two independent ways.

First, my logic is backwards.  Now that I've created these diagrams it's obvious that Hfq's contribution to competence regulation acts in the opposite direction.  Knocking out hfq would be expected to increase the expression of competence genes under non-inducing conditions, not decrease it under inducing conditions.

Second, it looks like H. influenzae doesn't even have a Spot 42 homolog.  Homologs are ubiquitous in the Enterobacteraceae, and have recently been shown to also be common in the Vibrionaceae, with about 85% sequence identity.  But BLAST searches of the Pasteurellaceae with either type of sequence didn't find a single homolog.  Given the relationships shown in the tree below. I think the ancestral Pasteurellacean must have lost its Spot 42 homolog.

Still no cloning success

Here's what I've done and what I've learned:

At the end of my last post I was about to test the toxicity to E. coli of the 'toxin' gene adjacent to the gene I'm trying to knock out.

I did this by instead trying to create the same plasmid the undergrad had created, one that (unintendedly) deletes the end of the toxin orf as well as the adjacent gene.  So I repeated the inverse-PCR reaction using her old primers instead of my new ones, and used this fragment in my kinase-ligate-transform experiment.  I tested both the ability of this fragment to self-ligate after being treated with kinase, and its ability to ligate to a SpecR PCR fragment that had been treated with kinase.  The former produced only four AmpR colonies, and the latter no AmpR SpecR colonies.  I didn't do plasmid preps on the AmpR colonies to see if they contained the expected plasmid, though maybe I should have.

Although failure of this experiment does not disprove the hypothesis that the toxin gene is toxic to E. coli, it does disprove the hypothesis that the hypothesized toxicity is the reason that my previous experiments failed.  Sorry, that's a nasty sentence - try again: Since this experiment worked for the undergrad last spring but not for me now, toxin-toxicity is unlikely to be why my experiments are failing.

Next I made a list of the various approaches I could try - I'll get back to these below.

The next test I did was to see whether I could ligate the SpecR PCR fragment to a plasmid that didn't need phosphorylation, and whether simple blunt-end ligation was working at all.  If this worked I'd know that the problem is the kinase reaction.  So I cut a simple plasmid (pUC18) with EcoRI (Makes sticky ends) and separately with SmaI (makes blunt ends).  Both these cuts leave ends with 5' phosphates that should be good substrates for ligase.  I heat-inactivated the enzymes and set up three ligation reactions:

1.  EcoRI-cut plasmid with ligase
2. SmaI-cut plasmid with ligase
3. SmaI-cut plasmid with SpecR fragment and ligase

Results: 

1. ~3500 AmpR colonies (all of the ligation reaction)
2. 639 AmpR colonies (all of the ligation reaction)
3. 198 AmpR colonies (half of the ligation reaction) and 1 SpecR AmpR colony (other half of the reaction).  But my plasmid miniprep of cells from this colony didn't give any plasmid

As a positive control, the same amount of uncut plasmid gave about 3000 colonies.

So I concluded that my blunt-end ligation reaction conditions were fine.  Can I then conclude that the problem is the kinase reactions?  Unfortunately this wasn't a very stringent test of the ability of the SpecR fragment to be blunt-end-ligated into a vector, because I didn't pay attention to the relative proportions of the vector and insert.  I should have used a limiting amount of the vector and lots of insert, but I actually used about equal amounts.  Since self-ligation of the vector is a unimolecular reaction it is expected to be much more efficient than insertion of the Spec fragment. 

This experiment used up the last of my control SpecR AmpR plasmid stock so I grew up the cells and did a miniprep.  But this didn't give any plasmid at all either, just some chromosomal DNA!  Can something also be wrong with the plasmid prep solutions, or my procedure?

The grad student is trying to use the same kinase to label his DNA with 32P-ATP.  He's not having much success either so he's going to test my fragments, which are much simpler substrates than the sheared chromosomal DNA he's been using.

Maybe the toxin is toxic!

Recap of the last few posts:  Starting with a plasmid containing the toxin-antitoxin operon of Actinobacillus pleuropneumoniae, I've been trying to create a derivative plasmid whose toxin gene is functional but whose antitoxin gene has been replaced by a specR cassette.  This involves several steps: PCR of the specR cassette and inverse-PCR of the plasmid (to produce a fragment lacking the antitoxin gene), phosphorylation of the specR fragment with T4 kinase, ligation of the two fragments, transformation into E. coli, and selection for SpecR and AmpR cells.  The PCR steps work but the rest fails to produce any resistant colonies

After the first attempt failed I introduced several controls: EcoRI-cut pUC18 as a ligation control, another SpecR AmpR plasmid as a transformation control, and the kinase-treated inverse-PCR fragment as a kinase control.  The ligation and transformation controls worked, so I decided the kinase was at fault.  This hypothesis was supported by finding that I had been using a long-expired stock of kinase rather than the one bought earlier this year.

But two more attempts using the new kinase have also failed.  The first used the supplied kinase buffer and my stock of ATP, and the second used new ligation buffer (recommended by the supplier) which contains its own ATP.  The second time I also preheated and rapid-chilled the substrates, which is recommended to help expose the blunt ends to the kinase.  Both times I got no transformants from either the test or the kinase control, but got lots with my transformation control plasmid. (I didn't bother repeating the ligation control.)

I've been trying to think of what else could be going wrong with the kinase reaction, but it just occurred to me that maybe there's a completely different problem - maybe this toxin is toxic to E. coli.

Although the honours student had mentioned this concern when she handed this project over to me, I had discounted it because the H. influenzae homolog is not toxic in either H. influenzae or E. coli. But both my desired construct and the recircularized inverse-PCR fragment I'm using as a control are expected to express the toxin, possibly at high levels.  So maybe my reactions are all working, but the plasmid they produce is not tolerated by E. coli.

How to test this?  Directly testing for lethality is tricky.  But I can do a different kinase control using a different inverse-PCR fragment, one that won't express the toxin. If the problem is the toxin, this should give AmpR transformants.  I can also use the same fragment with the spec cassette, and I should now get SpecR AmpR transformants.  I have all the honours student's primers and her CR conditions so this should be straightforward.

Think Check Submit - can't we do better than this?

The Scholarly Kitchen blog has a post about a new initiative from a large consortium of scholarly publishing societies and individual publishers, intended to help inexperienced researchers avoid journals from 'predatory publishers'.  This is a very worthwhile goal, but the actual advice provided so far isn't going to exclude most of the bad guys.   

The first step just explains why researchers need to be careful where we publish:

The third step is just reassurance:

The second step is the one that matters; it tells researchers what they should look for:

There's nothing wrong with this advice, but it's certainly not treading on any publishers' toes.

Most importantly, there's no mention of the most valuable resource we have, Beale's List.  This is a frighteningly long list of open-access scholarly publishers whose tactics are potentially, possibly or probably predatory.  It's maintained by Jeffrey Beale, a librarian at the University of Colorado, at his Scholarly Open Access blog.  The last time I checked, a couple of years ago, there were about 300 publishers on this list, but today there are 882!  And this is just the publishers - most of these have multiple journals.

Beale's list isn't just a list.  Beale also provides explicit sets of criteria for evaluating individual publishers and journals.  The absence of Beale's List from the THINK CHECK SUBMIT campaign isn't really surprising, but it reinforces my concern that we can't rely on the publishers to look out for researchers' interests.

New kinase stock found

Along with new stocks of other enzymes.

Somebody apparently thought it was a good idea to stop putting enzyme stocks in their usual place (the 'Special Enzymes' box) in the -20 °C freezer, and instead put them in this new 'coloning' box.

They then put the new box in a bin in the freezer where its label couldn't be seen.

It's the kinase!

Yesterday's experiment worked very well, in that the thorough controls clearly tell me where the problem is.  But the actual experiment produced only three candidate colonies.

Control E: No DNA. No SpecR or AmpR colonies.  GOOD-selective plates kill non-resistant cells

Control F: AmpR SpecR Plasmid.  p∆TA::Spec: ~350 AmpR and ~350 SpecR transformants   GOOD-selective plates select for cells carrying the resistance genes on a plasmid, and the competent cells transformed efficiently.

Controls D and G: EcoRI-cut pUC18 ± ligase.  ~12,500 AmpR colonies after ligation, only about 250 without ligase.  GOOD- ligation worked.

Control C: No-ligation control. Kinased spec PCR product plus not-kinased inverse-PCR product, ligation reaction with no ligase.  No SpecR or AmpR colonies.  GOOD-The fragments do not spontaneously circularize and transform cells, and the fragment mixtures do not contain any unwanted intact plasmid.

Control B: Kinase control.  Ligation of kinased inverse-PCR fragment.  Should have given AmpR colonies, but none.  BAD- Kinase failure.

Experiment: Ligation of kinased spec PCR product plus not-kinased inverse-PCR product.  Gave only 1 AmpR colony and two SpecR colonies.

Next steps:  

I've streaked the three candidate colonies on both Spec and Amp plates.  The desired plasmid should confer resistance to both.

And I looked at the expiration date on the tube of BioLabs T4 polynucleotide kinase I've been using. 03/09!!!!  AAARRRGGGHH!!!!

Have I been using the wrong tube of kinase?  Is this not the kinase that the former undergrad and sabbatical visitor used successfully last year?  Searching the 'Special Enzymes' freezer box turned up another tube of T4 polynucleotide kinase, but this one looks even older.  So I've just emailed the undergrad and sabbatical visitor to ask what they used.

Positive control problem solved

I did the test experiment described in the previous post, and then spent the past few days figuring out why my positive control transformation didn't work any more.

The test experiment was to kinase, ligate and transform into DH5alpha the product of the inverse-PCR reaction.  If the T4 polynucleotide kinase reaction worked, its blunt ends would acquire 5' phosphates that would allow it to be circularized by T4 DNA ligase, and to then transform DH5alpha to ampicillin resistance.  The negative control was no DNA and the positive control was the same p∆TA:spec plasmid that had given thousands of AmpR and SpecR transformants in the previous experiment.

Sounds great, but this time the positive control didn't give any transformants at all!  Background small colonies were frequent, possibly because the plates were a bit old and the ampicillin had lost its potency, so I didn't trust the few larger colonies on the inverse-PCR reactions plates.  I streaked a few of the large colonies to check if they were genuinely AmpR - one was.

I repeated the control transformation and negative control with new Amp plates; the no-DNA control plates were clean but so were the p∆TA:spec plates.

I thought the problem might be the plasmid, but I wasn't sure I have another reliable positive control. So I did a miniprep from the one genuine AmpR colony I had streaked and transformed the cells with that DNA.  I also had the usual no-DNA control, the undergrad's p∆TA:spec, another plasmid made by the undergrad (used successfully by me as the inverse-PCR template, and some pUC18 left by a sabbatical visitor.

Success all around.  The miniprep DNA, the other undergrad plasmid and the pUC18 all gave lots of transformants (the photo shows part of a pUC18 plate), and the undergrad's p∆TA:spec and the no-DNA control gave none.  I don't know why the p∆TA:spec plasmid worked well in my first experiment - maybe I had grabbed a 'wrong' (i.e. good DNA)tube.

Next step, repeating my original experiment (the one in the previous post), this time with better controls.

1. DNA clean-up: I did a new inverse-PCR reaction because the old one got used up in the tests.  I need to start by doing a spin-column cleanup of it.
2. Two kinase reactions: (i) the 'blurry' spec PCR product and (ii) the not-blurry inverse-PCR product.  Heat-inactivate the kinase before step 3 (65°C 20 min).  This time I'll use a newer stock of ATP, and the official kinase buffer.
3. Four ligase reactions: A. The kinased spec fragment plus not-kinased inverse-PCR fragment. B.  (kinase control) The kinased inverse-PCR fragment. C. (negative control) The not-kinased spec fragment plus the not-kinased inverse-PCR fragment . D. (positive control) pUC18 cut with EcoRI and heat-inactivated (65°C 20 min).
4. Six transformations: Ligations A, B, C and D, plus 1 µl pUC18 as positive control and no DNA as negative control.

Preparations:  We have enough kinase, and I've just sent the grad student to buy more ligase.  Luckily I have lots of frozen competent cells for the transformations.  I'll need to digest the pUC18 and check it in a gel, and pour lots of Amp plates and some Spec plates.