Building a Better Mouse

Share on FacebookShare on Google+Email this to someoneTweet about this on Twitter

‘Smart’ mice teach scientists about learning process, brain disorders

Mice genetically engineered to lack a single enzyme in their brains are more adept at learning than their normal cousins, and are quicker to figure out that their environment has changed, a team led by researchers at UT Southwestern Medical Center has found.

The group is also beginning a search for drugs that might create the same effects without genetic manipulation and monitoring the animals’ health and behavior over time.

The key in this study was being able to “knock out” the gene for Cdk5 only in the brain, and only when the mice were adults. This technique, only recently developed and called conditional knockout, allows much more sophisticated experiments than traditional knockout, which entirely eliminates the gene.

Normally, Cdk5 works with another enzyme to break up a molecule called NR2B, which is found in nerve-cell membranes and stimulates the cell to fire when a nerve cell signaling molecule, or neurotransmitter, binds to it. NR2B previously has been implicated in the early stages of learning.

The new research showed that when Cdk5 is removed from the brain, the levels of NR2B significantly increase, and the mice are primed to learn, Dr. Bibb said.

Evolved biological systems aren’t optimal…just good enough. Knowing how the system works should lead to interventions that improve function. In the coming decades there should be nutritional, drug, training, genetic, and cybernetic enhancement of brain function.



  1. of course joe tsien made an adult mouse with extra NR2B several years ago (the doogie mouse).. NR2B is preferentially expressed in young mice and has a longer calcium conductance, purportedly increasing the time window for coincidence detection between pre and post synaptic firing.. in adults NR2B is usually replaced with NR2A and less learning.. presumably this developmental switch happens for a reason..  
    i’m by their reversal learning conclusions in the real paper.. i’m not on completely firm ground but basically i thought fast reversal learning was a sign of poor memory for the initial training.. guess you can interpret how you like..

  2. “presumably this developmental switch happens for a reason” 
    Yes. I’d like to know what the trade offs are. Perhaps too much adaptation to new surroundings hampers long term skill development in adult mice. High level learning might depend on the stability of lower level learning. 
    It would be nice to have a “smart” pill that temporarily increased learning rate.

  3. Any good links or papers on such developmental stage specific expression? How do they make it adult specific?

  4. in this case they used an inducible transgenic system (discussed here at some length). 
    basically, they surrounded part of the cdk5 gene with sites that can be recognized by an enzyme that chops out anything between those two sites.. the enzyme isn’t naturally found in mice. so they crossed altered cdk5 mouse with one that can express the enzyme.. here’s the trick though: the enzyme is setup so that it is only works in the presence of tamoxifen. so you just feed the mice tamoxifen when you are ready to knockout the gene.  
    one problem is that the knockout isn’t reversible, so you can’t tell whether you affected memory acquisition, consolidation, storage, or retrieval.  
    with the doogie mouse i believe they set up their transgene under the control of the promoter for alpha CamKII which happens to be modulated by development. this gives less strict control obviously, but its a very common application now.  
    as far as genes that switch during development, all of the pkc isoforms except zeta and delta increase over the first 4 weeks, nr2b switches out and nr2a switches in, AMPA receptor subunit GluR4 switches out for GluR1, and there was a recent paper showing PSD-95/MAGUK family members doing a dvelopmental swithc too.. i’ll have to find links later..

  5. ‘Presumably’… Yes, you’re correct in that there is likely causation for the switch from NR2B & NR2D to NR2A & NR2C.  
    So, basically, NMDAR’s are heteromeric and composed of assemblies of NR1 + NR2x/NR3; each with distinct expression topologies and, just as importantly, distinct gating properties. NMRAR’s have pretty damn slow gating dynamics generally, having an offset decay time constant of: NR2A (~120ms), NR2B & NR2C (~400ms), NR2D (~5,000ms).  
    It’s a dual edged sword, so you, presumably, want to keep the channel opened longer to accrue a bigger EPSC, inflate the probability of PSP firing and facilitate the ‘coincidence detection’ someone previously mentioned.  
    OTOH, hyperactivating ionotropic glutamate receptors is an excitotoxic mechanism. NMDA channels as mentioned have a really long open time and have been shown to be pivotal in mediating this. I’m not familiar with the exact pathways of glutamate/calcium toxicity, I work with the receptors; Sorry. 
    IMHO, To disagree with fly’s Building a Better Mouse post, he said, “Evolved biological systems aren’t optimal…just good enough.” I disagree with this comment on some levels and the NR2B thing is a good example. After the developmental switch, of the adult isoforms: NR2A is ubiquitous, NR2C is localized in the cerebellum, and of the transient ones: NR2B becomes restricted to the forebrain and NR2D that was in the diencephalon nearly disappears. So there is a neat little distribution of NR2 subunits that appears quite smart. 
    Not only for the neuroprotective role, but if you could look at where the optimal ‘fitness’ is on a high dimensional landscape taking into account the excitotoxicity, energetic costs, information density, etc, I would guess you are pretty close to optimal for the closed system as currently laid out. It’s not a global optimal solution, you could likely build a better receptor that does the same job; or if you could dynamically and locally swap current receptor isoforms at current receptor recycling rates based on some sort of feedback loop, perhaps (overhead costs?!). But the system works pretty darn good. 
    Many people, IMHO, fall into this pit of thinking that biology is this binary system in which you can flip a global switch or ubiquitously swap subunit A for B with no other effects instead of modeling it as a dissociative or whathaveyou network.  
    Marc Hauser has a story how he used to ask his students what single sensory parts would you reversibly take from any animal. So, everyone wants the visual circuits of an eagle, auditory cortex of a bat and olfactory bulb of a dog… but then, he makes the point, who wants to smell a millimole of urine on a hydrant at 200 yards? So, there is something to be said for looking at the whole system in toto; it might not be that bad perceptually for a dog, but I’m another story. 
    That being said, the article is a step in the right direction as it’s a targeted, local approach.

  6. Stupid me, I saw Fly’s comment mid way through and went off on a tangent! 
    On why the developmental switch happens at that specific moment, it’s likely also related to the relative oxygen tolerance of each subunit. NR2B is hypoxia invariant for example. NR2C (xp only in cerebellum) currents are increased by hypoxic challenge, thus it’s latent expression to avoid the excitotoxic effect’s I mentioned before. Mice exposed postnatal ~D14 to hypoxic conditions have exaggerated, huge, motor deficients.  
    Usually in the NR2B discussions people point to Zhuo’s work which shows increased pain/irritation perception from NR2B as the reason. I’d tend to disagree since if the learning advantage was that significant, the ‘pain’ could just be inhibited or renormalized in the peripheral.  
    I’ll stop there since some coworkers are looking at this topic, but this older paper from another lab is decent primer if interested: Oxygen sensitivity of NMDA receptors: relationship to NR2 subunit composition and hypoxia tolerance of neonatal neurons.