This is an idea that has been floating around in the back of my mind for quite awhile, but today, as I was looking at a blank CRISP/Cas9 - FMN 1b-ng - Exclusion puzzle, I realized it provided an ideal test case.
Consider the following.
By simply pasting in the MS2 hairpin and mutating three other bases, I’ve created a switch that satisfies all the constraints except for the ones limiting A’s. There are all kinds of interesting ways all those unmutated A’s can be filled out, but the focus here is highlighting what makes such a simple sequence work as a switch.
It works because the 5-base segment marked in red is an essential component of both the input structure (here, the stem that closes the FMN aptamer) and the output structure (here, the MS2 hairpin). It’s a straightforward application of the Bachelor’s Dilemma pattern – the segment has to make a choice between forming one structure or the other – it can’t do both.
Of course Nature, like bachelors, can be fickle and try to play it both ways at the same time. So for a great switch, the design creator needs to try to discourage that.
What especially excites me is that switches constructed in this way should work (perhaps with minor tweaking) using almost any small molecule as the input and MS2 as the output. The only requirements for the puzzle seem to be that 1) the input aptamer requires a closing stem, and 2) there are enough available bases next to the stem to place the MS2 aptamer.
As I was writing the above paragraph, I decide to test its premise on another current exclusion puzzle using a different input molecule. I selected the TEP 3 Exclusion puzzle, and sure enough, it worked like a charm. In this case, the aptamer stem has a locked base that precluded putting the MS2 sequence directly against the aptamer loop, but that didn’t pose any problem. I just placed the hairpin as close as I could and then started filling out complementary bases. [Full disclosure: If you look closely, you’ll see that for this puzzle I did have to fill in some other bases to enforce all the constraints, but these are for holding static structures, not involved in the switching. You might also notice the length 2 hairpin that shows up in state 2. It occurs in this screenshot and not the first simply because the first screenshot NUPACK and the second with Vienna2. They differ on whether that hairpin is strong enough to be part of the MFE folding, and a design that also balances the MFEs to NUPACK’s liking requires a little more tweaking.]
I have a lot of designs of the same type this round, although I didn’t have a name for them. Many of the #alt MS2 designs I submitted follow this same pattern.
Thank you for replying, Andrew. If there’s an easy way to point out which ones, I would love to look them over to compare how you filled out the designs. It sounds like you selected MS2 sequences where possible, to minimize the distance to the aptamer. That certainly seems like a good idea.
I’m sure I’m not the first to create designs like this. All I’ve done here is to call it a pattern and give it a name. If I weren’t focusing so much on OpenCRISPR submissions right now, I would have gone back through past results looking for examples and seeing how they performed.
Definitely. I wouldn’t expect any of these designs would score well as shown here. In fact, I suspect they might behave like weak Same State switches. i.e. with a fold change of less than 1. To hope for a good Exclusion switch, one will need to bind the MS2 hairpin more strongly in state 2, with this being balanced by a comparable amount of energy reduction in the folding for state 1.
Hi Omei,
I looked through the designs, and the ones with #alt MS2 in the title in these sublabs are the ones I was talking about. I did add nts to most of these designs to further disrupt the MS2.
FMN 1a - Exclusion
FMN 1b - Exclusion
FMN 1a-ng - Exclusion
FMN 2 - Exclusion
FMN 3 - Exclusion
FMN 4 - Exclusion
TEP 1b - Exclusion
TEP 2 - Exclusion
TEP 3 - Exclusion
Tetracycline - Exclusion
Tetracycline inv - Exclusion
A few of the sublabs I tried the MS2 on both sides, but most are on the high numbered side.
Thinking more about this, I see two separate goals here, one being our normal goal of designing switches with high fold changes, and the other being something we have never tried to do before (as far as I know) of designing aptamer-agnostic switches. I would expect the best fold changes for the latter won’t match the former, but a collection of switch sequences that could be joined with (almost?) any RNA aptamer seems like a powerful laboratory tool.
The first requirement for an aptamer-agnostic switch would be that none of the switch bases pairs directly with bases of the aptamer. This is not the way I would normally design a switch, but after some initial failures, I designed a sequence that works using either end of the FMN aptamer, e.g. the FMN 1a and 1b puzzles.
(The Christmas tree stem is paired that way just to call out that it is a static stem that can be lengthened or shortened as needed to fit the constraints of our puzzles.)
This sequence doesn’t work with the other four aptamer/orientations yet, but I realized that I built in kind of an assumption at the aptamer neck that I don’t think is critical to success. And because the tetracycline aptamer uses so many bases, I need to make the switching area a little more compact to fit into our puzzle length. But neither of these seem like serious roadblocks.
Anyway, I’m posting this now to challenge anyone else who finds this an interesting problem. Getting half a dozen essentially different sequences that work in-silico gives us a much chance of having at least one work well in the lab than if we get only one.
So for a great switch, the design creator needs to try to discourage that.
