Lessons from the past - designing switches with the MS2 hairpin

Talking with Eli, we realized that there are probably a lot of current lab participants who have had little to no experience with designing with the MS2 hairpin as the output signal.  If memory serves me right, it has been almost two years since we’ve gotten new lab data on this type of design, whereas for about two years before that, the MS2 hairpin was the standard for practically all labs.  So we decided to try to dredge up the most important lessons the old timers learned back in their day.


As I was scrolling through submitted designs, one thing in particular jumped out, and that is what it takes to firmly turn off the MS2 binding in the OFF state.

Here is an extreme case of what  won’t make a good switch.

The game will accept this as having satisfied the constraint of not having the MS2 structure, but that is completely misleading.  The game can only show the folding that it calculates to have the single minimum folding energy (MFE).  But in nature, each RNA strand is in constant motion, changing between all possible foldings that are energetically “available”.  The act of binding the MS2 hairpin to the MS2 protein actually lowers the energy,  so in this case, the MS2 protein will be bound such a large percentage of the time that there will be little difference in binding between the OFF state and the ON state. 

But this is only the extreme case.  In practice, we’ve observed that even a predicted folding like this one will keep the switch from being very good:

Only the end of the hairpin is predicted to fold, and even then, it has the C’s slipped out of position.  But the end of the MS2 hairpin is the part that actually binds most strongly to the MS2 protein, and this is close enough that it won’t take a very high concentration of the MS2 protein to force a binding.

To sum things up, for a good scoring switch, you’ll probably want your design to at least break up the end of the MS2 loop in the OFF state.  Or as Eli put it, “Grab it by the hairpin loop”.


Patterns for MS2 position in relation to the aptamer

Exclusion puzzles: Put MS2 next to aptamer in OFF switches

Same State puzzles: Place MS2 distanced to aptamer in ON switches

How you want to distance the MS2 to the aptamer is up to you. There are several ways to do this. Experiment! :slight_smile:

The CRISPR puzzles are bigger than our earlier Riboswitch puzzles, so please do experiment with MS2 position, as you might come up with something different that also work. This is fine as long as you are aware that you are experimenting.

From my experience, the trends shown above are what have the biggest chance of working.

Don’t forget the winning switch mechanism

I see a lot of you forgetting the turnoff sequence to MS2 in exclusion labs.

From past winning design score 94%


This turnoff sequence is a shared sequence between aptamer and MS2 and it takes turns turning either the MS2 or aptamer off.

The value of transfering past winners - Get winning designs for free :wink:
I took the above design when we had a new lab round. I transfered an exact copy of it - which also scored 94%. This was my control design to compare my planned experiment against.

I then did an experiment to it, where I made a targeted destruction of the turnoff sequence.

The point was to demonstrate that the mechanism of the switch would not work without this sequence.

With the turnoff sequence mostly replaced with A’s - new score 64%

The turnoff sequence typically needs to bind in the aptamer sequence.

Different turnoff sequences for different aptamers

Not all aptamers take the same turnoff sequence. Even designs with the same aptamer may take variations in the turnoff sequence. Your job is to figure ways how to do it. A trick is to look for similar or identical stretches of sequence in both the aptamer and the MS2 and then make a middle man sequence.

Sometimes it can be hard. Then one can make half of the turnoff sequence target a bit in the MS2 and the other half target in the aptamer. Then the aptamer and MS2 will still be competing about binding with it.

For more explanation about the turnoff sequence, check out this CRISPR training tutorial.

CRISPR training #1: Shared sequence - Bachelor’s dilemma

MS2 is very strong. On turning it off properly, I said “Grab it by the hairpin loop”. 

What I meant is to get hold of it properly, not just in a superficial way that takes hold of just the opening base pair of the MS2 stem or just a few of them. 

However pairing literally with the sequence in the hairpin loop isn’t the most effective either. 

What I see being particularly effective is to grab it by its magnet segments. And what are magnet segments?

Some RNA bases make stronger bind than others. C’s and G’s in particular. When there are several of these bases, I call them magnet segments. Here are the magnet portions of the MS2. 

Targeting those are very effective in switch elements in general.


In Exclusion FMN labs typically one set of the magnets are grabbed. In FMN labs it is almost always the G’s. 

In FMN Same State labs typically the first 4-5 bases of the bottom of the hairpin stem is grabbed. 


My explanaitions tend go rather long. As means to counter that I have decided to put up a switch guide I have been working on for a while. It is kind of a sum up on the difference between Switch RNA and static RNA.

I will try throw in abbreviated advice on lab. Should help me get more advice up as well. Beware of the mess as it is not quite finished. 

Lock one end of the aptamer with a static stem.

I have seen a lot of solves where the aptamer don’t have a static end in the smaller TEP and FMN puzzles.

A good partial moving switch mostly tend to have one end of the aptamer locked.

For more about why see the section No static stem at the aptamer end opposite the MS2. This post goes into things to avoid when making Riboswitch designs. Hint: There are more hints there. :slight_smile:

Where are free moving aptamers a good idea?

Where unlocked aptamers typically are of most use are in small designs that goes full moving and where the switch elements are not placed optimal in relation to each other.

In some of our current smallest labs where there is not much space for movement, taking that small area behind the aptamer in use for making the switch can make sense.

I don’t want to discuring you guys from experimenting. I am so myself. Just sharing what I think most likely will to lead to successful designs.

Coaxial Stacking

One more thing is really important in relation to success for switch labs in general.

Coaxial stacking

Basically Coaxial stacking has to do with when two stems are adjacent to each other, they get a nice energy bonus for lining up on top of each other. It stabilizes the structure.

Omei realized the potential of coaxial stacking in relation to Eterna RNA designing. This is something he is investigating. For details see this background post.

Example of coaxial stacking in Exclusion NG 2 lab

I think coaxial stacking is why the turnoff sequence likes to sit right next to the MS2. And also why the static stem appears to turn up at very specific spots.

Example of coaxial stacking in Same State NG 2 lab

This aptamer design is of the type that did best in this lab - of one aptamer designs . It holds coaxial stacking opportunities in both the ON and OFF state.

In some cases it is even possible getting the static stem next to both MS2 and the aptamer gate. Or as here two coaxial stems next to the MS2. Assuming that coaxial stacking only makes sense with even amount of stems. They should pair up two and two.

Position of the coaxial stem

In Exclusion:

  • Next to the aptamer gate (Most prevalent for now)

  • Next to the turnoff sequence

In Same State:

  • Next to the aptamer gate - which are the turnoff sequences for Same State puzzles.

  • Next to MS2 (Most prevalent for now)

Which is best?

Coaxial stacking opportunities

Here a Same state NG 2 winner. With coaxial stacking opportunities drawn in.  


Coaxial stacking can happen both with static stems or switching stems.

Get creative with coaxial stems. :slight_smile:

Static lab designing - Important highlights

It’s a real long time since we have done static labs. A good bunch of you haven’t had the slightest hands on experience with static RNA. 

As I took a look at some of the CRISPR control lab designs, I realized I need to highlight some crucial lessons from the past. 

A few of you uses way to much or even too little GU. Same goes for GC and AU. There are certain designs that are doomed to fail as static lab designs. 

The first and best place to send you is Dimension9’s forum post on the topic. 

RNA Lab: “Christmas Trees,” “Cub Scout Projects,” and “Optical Illusions” 

He was the one who realized that the energy models cheated on us, by allowing things that lab will not allow or only allow in rare cases under special circumstances. 

Sum up: 

  • Designs with mainly GC won’t work  (Except in really short stemmed designs)

  • Designs with mainly AU won’t work  (Except in really long stemmed designs)

  • Designs with mainly GU won’t work

Designs with a mixture of mainly GU and AU won’t work either. Similar while seen much rarer, GC with a lot of GU likely won’t work either.

Long Stems crave GU

Static stems start to crave GU when they get something like 8 or more base pairs long.
Lab designs that crave GU

Still don’t overdo. Just throw in a GU - preferably somewhere well burried in the stem.

While long stemmed designs generally are the most tolerant of weird base distributions when it comes to solving pattern, I wish to highlight what will be less likely to work anyway. 

I encouraging experimenting. Just have this in mind so you don’t use all your slots on something that will most likely fail.

Experiments to do

However if you did an experiment like Rhiju with going from Zero GC to Max GC and starting from AU, that is actually a real fine experiment.

I would like to see this done for all other two basepair combos. That would actually be a way to get data on Dimension9’s insights.