Developmental Neuroscience · Unit 9: Experience refines the wiring

Opening and Closing a Critical Period

In mouse visual cortex, closing one eye reshapes the cortex only during a few weeks of youth. This lesson asks what sets those weeks: the evidence that a threshold level of inhibition opens the window, the molecular brakes that later close it, and the rodent experiments that reopen it. Afterwards you can read a plasticity experiment as a test of a regulated state rather than an age, and judge what such results do and do not promise for people.

Watch first A Window Held by a Threshold · 2:40

A mouse that lacks one of its two enzymes for making GABA can go through the whole of youth without the critical period that should have reshaped its visual cortex. Give it a dose of a benzodiazepine, even as an adult, and the window opens.

In the lesson on ocular dominance you met the basic result: close one eye of a young cat or mouse, and the visual cortex stops answering it. The same deprivation in an adult does much less. Such stretches of time are windows of opportunity and vulnerability. This lesson takes up the next question, “How critical period is regulated?”, and the knockout mouse in the bold line is where the answer starts.

How critical period is regulated? A threshold of inhibition

Most of the evidence comes from one system: mouse primary visual cortex, with monocular deprivation as the readout. A mouse eyelid is sutured shut for about four days, and responses to each eye are measured. In mice, this shift is largest around postnatal day 28 (P28), and much smaller before about P19 or after about P32 (Gordon and Stryker, 1996). The mouse earns its place because you can change its genes.

One distinction helps here: critical period vs. sensitive period. A critical period begins and ends fairly abruptly; a sensitive period rises and fades gradually. The mouse ocular dominance window is usually called a critical period, though its edges are softer than the name suggests.

The lead came from a knockout. A knockout mouse lacks a working copy of one gene. Mice have two enzymes that make GABA from glutamate, GAD67 and GAD65. GAD65 sits at the synaptic terminal and supplies GABA when neurons fire hard. In 1998 Takao Hensch, Michela Fagiolini, Michael Stryker and colleagues reported that GAD65 knockouts showed no ocular dominance shift after deprivation at the normal age. Their neurons still responded to vision. What failed was the change.

Then the rescue. Diazepam is a benzodiazepine, a positive modulator of the GABA-A receptor: it binds at a site between subunits and makes the receptor’s chloride channel respond more strongly to GABA, without opening it alone (see that lesson, and the lesson on drugs that act on one part of a receptor). Infused into the visual cortex of the knockouts during deprivation, diazepam restored the shift.

Think of a smoke detector that needs a minimum battery charge before it will sound. The knockout’s battery is low. Diazepam tops it up.

Two results from 2000 (Fagiolini and Hensch) turned this into a claim about timing. First, knockouts kept their capacity for a critical period into adulthood, and diazepam triggered the window whenever it was given. Second, in wild-type mice, diazepam given before the normal opening, around the time the eyes open, started the window early. In a later study (Iwai and colleagues, 2003), two days of diazepam just after eye opening closed the window early in wild-type mice, so the normal one did not follow. Boosting inhibition did not itself shift the cortex. It let the cortex respond to deprivation.

Two other manipulations move the window in the direction this model predicts. In mice built by Josh Huang, Susumu Tonegawa, Alfredo Kirkwood, Lamberto Maffei and colleagues (1999), cortical neurons make extra BDNF, a neurotrophin that among other effects promotes inhibitory synapses. Their inhibition matured early, and their critical period ended early. Rearing animals in darkness from birth, which slows the maturation of inhibition, delays the window in rodents and in cats.

Design problem

A GAD65 knockout mouse shows no ocular dominance shift after monocular deprivation at P28. Predict two experiments. First, give diazepam to a knockout at P60 and then close one eye. Second, give diazepam to a wild-type mouse at P15 and then close one eye. What do the two results together tell you about what opens the window?

One way to do it

Start with what the knockout lacks. It has normal eyes, normal thalamic input and responsive cortex, but too little GABA released during strong activity. If the window waits for inhibition to reach a threshold, the knockout’s window has never opened, so it should still be available at P60.

So the first experiment should give a shift. Diazepam raises the effect of the GABA that is released, inhibition crosses the threshold, and the window opens at P60. That is what Fagiolini and Hensch reported in 2000.

The second experiment tests the same threshold from the other side. A wild-type mouse at P15 has not yet built enough inhibition. Diazepam brings it over early, so closing an eye then should cause a shift, and the window should then run its course and end ahead of schedule. Fagiolini and Hensch reported the early opening, and Iwai and colleagues (2003) the early end.

Together the results say that the window is triggered by a level of inhibition, not by the calendar. Age matters only because, in a normal mouse, inhibition reaches that level at a typical age.

Which inhibition: basket cells, alpha1 and Otx2

Cortex has many kinds of inhibitory interneuron. The ones implicated here are parvalbumin-expressing basket cells, fast-firing cells whose synapses wrap the cell bodies of pyramidal neurons. Like most cortical interneurons, they are born in the medial ganglionic eminence of the embryonic forebrain and reach the cortex by tangential migration.

The case for these cells comes partly from receptor subunits. GABA-A receptors are built from five subunits, and the alpha subunit varies. Receptors with alpha1 are concentrated at synapses from parvalbumin basket cells onto cell bodies. In 2004, Fagiolini, Hensch and colleagues used knock-in mice, which carry a chosen point mutation, here one that makes a single alpha subtype insensitive to diazepam. When alpha1 was made insensitive, benzodiazepines no longer opened the window early. When alpha2 was made insensitive, they still did. In mouse visual cortex, then, the drug acts through alpha1 receptors, which points to basket cell synapses.

What makes these cells mature on time? One answer is a surprise. Otx2 is a homeoprotein, a transcription factor with a homeodomain. In 2008, Sayaka Sugiyama, Hensch, Alain Prochiantz and colleagues reported that Otx2 protein builds up in parvalbumin cells of mouse visual cortex although these cells make little Otx2 RNA. Its build-up depended on visual experience, and infusing Otx2 early opened plasticity early. The source was first proposed to be the retina; later work pointed to the choroid plexus, the tissue that makes cerebrospinal fluid. The full answer has been debated, so hold this part loosely.

Closing the window: brakes, not decay

If the window simply faded, nothing could open it again. Several experiments in adult rodents suggest it is held shut.

Perineuronal nets. As the critical period ends, a dense lattice of extracellular matrix, rich in chondroitin sulfate proteoglycans, condenses around many parvalbumin cells. Dark rearing delays it. In 2002, Tommaso Pizzorusso, Maffei and colleagues in Pisa injected chondroitinase ABC, a bacterial enzyme that cuts chondroitin sulfate chains, into the visual cortex of adult rats. Monocular deprivation then shifted ocular dominance, which it did not do in untreated adults.

Myelin signals. Myelin, the insulating wrap on axons (its own lesson), carries proteins such as Nogo that inhibit axon growth through the Nogo receptor. In 2005, Aaron McGee, Stephen Strittmatter and colleagues reported that mice lacking the Nogo receptor kept ocular dominance plasticity after the usual closing age.

Lynx1. Lynx1 is a small membrane protein that binds nicotinic acetylcholine receptors and damps their response. Its level in mouse visual cortex rises in adulthood. In 2010, Hirofumi Morishita, Hensch and colleagues reported that adult Lynx1 knockouts kept strong plasticity, and that blocking nicotinic receptors prevented it.

One caution. Adult mouse visual cortex is not completely fixed. Sawtell and colleagues (2003) found an adult form of ocular dominance plasticity in mice, mostly a strengthening of the open eye’s responses, which differs from the juvenile loss of the closed eye. “Closed” is relative.

Re-opening the critical period?

Why would anyone want to reopen it? The reason is amblyopia, in which one eye sees poorly because its input to the cortex was degraded in childhood, for example by a misaligned or blurred eye. Patching the good eye works best in young children, and much less reliably in adults. A drug that reopened the window could, in principle, give adults another chance.

Several rodent experiments point that way. In 2008, José Fernando Maya Vetencourt, Maffei, Eero Castrén and colleagues reported that weeks of fluoxetine, an antidepressant, restored ocular dominance plasticity in adult rats, lowered intracortical inhibition, and allowed recovery of vision in rats made amblyopic by long deprivation in early life. Elizabeth Quinlan’s laboratory reported that a period of complete darkness restored plasticity in adult rats. And in 2010, Derek Southwell, Sunil Gandhi, Stryker, Arturo Alvarez-Buylla and colleagues transplanted embryonic medial ganglionic eminence cells into young mouse cortex. When the transplanted cells reached about the age at which native interneurons open the window, the host had a second critical period, after its own had closed.

Read these as evidence about mechanism, not as treatments. Each result is in rats or mice, mostly in visual cortex, with deprivations that differ from human amblyopia. Reopening plasticity also reopens vulnerability, since a plastic cortex can be changed for the worse. Small trials of drugs in adults with amblyopia have so far not shown clear benefit, and none of these approaches is an established therapy. Song learning and language have their own windows, taken up in the lesson on sensitive periods.

What the rodent work does show is the lesson’s main point. The critical period is a state of a circuit. Inhibition has to reach a threshold to start it, brakes are added to end it, and in mice and rats both steps can be moved by genes, drugs and experience.

How we measure it

Monocular deprivation with an ocular dominance readout

One eyelid is sutured shut for a few days, then responses of visual cortex to each eye are measured with single-unit recording, visual evoked potentials or optical imaging. A shift toward the open eye shows that the cortex was plastic during the deprivation.

Knockout and knock-in mice

A knockout mouse lacks a working copy of one gene; a knock-in carries a chosen change in it. GAD65 knockouts tested whether GABA synthesis was needed for the window, and knock-ins that make one GABA-A receptor subunit insensitive to diazepam tested which subunit carries the drug's effect.

Local drug and enzyme infusion

A drug or enzyme is delivered into one region of cortex through a minipump or an injection. Diazepam infusion tested whether boosting inhibition is enough to open the window, and chondroitinase ABC injection tested whether the extracellular matrix holds it closed.

Interneuron transplantation

Inhibitory neuron precursors are taken from the embryonic medial ganglionic eminence and injected into the cortex of a young host, where they migrate and mature. Testing the host later asks whether the age of the inhibitory cells, rather than the age of the animal, sets the window.