After reading about the meat and potatoes of Chen and friends’ paper, you may or may not agree that Layer 2/3 “dynamic zone” interneurons seem to control ocular dominance plasticity. Below, I have dissected the paper in more depth to try to get at this issue, but to get into the nitty gritty details I need to start by giving some more background.
Layer 2/3 interneurons have recently been shown to undergo a strong ocular dominance shift following monocular deprivation. This is paralleled in general in visual cortex neurons, where following monocular deprivation, responses of binocular neurons to stimulation in the deprived eye get weaker, while responses to stimulation in the non-deprived eye strengthen. It was recently found that this effect is actually much stronger in Layer 2/3 interneurons due to a larger desensitization to deprived-eye input to these cells. Although they don’t mention it very clearly, Chen and friends predict that this effect is due to the Layer 2/3 dendritic branch plasticity that they identified and characterized in the 2 previous papers I mentioned in the meat and potatoes. Specifically, the rational is that following monocular deprivation, the deprived eye will have less physiological input into binocular V1, resulting in loss of synapses from carrying input from the deprived eye, particularly from Layer 2/3 interneurons (judging by the fact that their electrical properties are affected more strongly by monocular deprivation.) The idea is that the dendritic branch tip retraction seen by this group has something to do with physical loss of synapses (which they actually show later on). However, which comes first is something they don’t guess at – does the monocular deprivation cause retraction of axons resulting in dendritic tip retraction, or does the interneuron sense a decrease in input activity and retract its branch tips? This is the question of who is sensing activity levels, and will be an interesting question to answer.
So, the overarching idea is that these Layer 2/3 interneurons are the central force driving adult visual cortex plasticity. Again, as I mention earlier, this notion began with the Nedivi lab’s identification of a superficial “dynamic zone” of layer 2/3 that contains these interneurons with highly dynamic dendritic trees. The idea is further bolstered by the older findings that cells in cortical layers 1-3 (extragranular layers) have a greater potential for adult plasticity than cells in other layers. So it seems like these Layer 2/3 interneurons, particularly those in the “dynamic zone”, are master controllers of plasticity in the visual cortex.
The first step in testing this theory was to see what happened to binocular V1 interneuron branch tip dynamics following monocular deprivation. (They identified binocular vs monocular V1 using Optical Intrinsic Signal Imaging.) Prior to monocular deprivation, around 3% of branch tips either elongated or shortened. In contrast, following monocular deprivation (always in the contralateral eye) branch tip dynamics increased significantly to 9% at 4 days post-deprivation, and 8.5% 7 days post-deprivation. However, overall branch length was conserved, indicating that the proportions of tips that elongated and retracted were the same in the normal and deprived conditions. When Chen and friends assessed branch tip dynamics at 14 days post deprivation, they were below baseline levels. Amazingly, this was exactly what Chen and friends had predicted, since it was already known from physiological work that it takes about 7 days for ocular dominance shifts to occur following sensory deprivation. To verify that branch tip dynamics actually involve changes in synapse number, Chen and friends verified by electron microscopy that a newly elongated branch tip contained synapses, and found this to be the case. They only did this for one tip though, so all this tells us is that branch tips can support synapses. After seeing the huge electron microscopic reconstructions in these papers that I reviewed, I am very disappointed that Chen and friends was convinced by a single branch tip. Aside from providing real evidence for their claim, assessing synapse size and density on these branch tips could be very interesting. Are these new synapses stronger or more numerous to allow their effects to be felt better?
Chen and friends also saw that monocular deprivation led to decreased monocular V1 tip dynamics at 4 days post-deprivation, and a return to baseline at 7 days post. They brush this off as merely falling in line with the idea that the plasticity is occurring in conjunction with ocular dominance shift in the binocular cortex, but it occurs to me that this decrease in dynamics reflects a stabilization of branch dynamics. Why might this stabilization occur?
Chen and friends then got more specific. They note that evidence compiled from the primate visual cortex over the past number of years shows that Layer 2/3 is a location where bottom-up, feed forward input from layer IV neurons (which receive sensory input from the thalamus) converges with top-down feedback input that is thought to modulate attention and stimulus salience. Thus, it turns out that the L ayer 2/3 interneurons are perfectly positioned to receive feed forward input from below, and feedback input from above. So Chen and friends tested whether there is any difference in plasticity of interneurons that are positioned higher in the cortex versus those located deeper. What they found fits really well with the idea of monocular deprivation inducing bottom up remodeling of Layer 2/3 interneurons and shifting of ocular dominance. Under control conditions, dendrites located deeper in the Layer 2/ 3 account for ~37% of remodeling (both elongation and retraction) while dendrites that project into Layer 1 account for ~63% of remodelling. However, following 4 days of monocular deprivation, Chen and friends find that dendrites located deeper in the cortex account for around 90% of the retractions, while the distribution of elongated dendrites is similar to control. Then, between 4 and 7 days post-deprivation, dendrites in the lower Layer 2/ 3 account for about 63% of elongations (increasing from 37% in control and 0-4 days post deprivation) and the distribution of retractions return to control levels. I find this truly amazing. Basically what Chen and friends have shown here is that directly following deprivation, you get a net retraction of dendrites on these interneurons, presumably resulting in a loss of synapses. THEN, a few days later, you get a regrowth of dendritic branches, presumably resulting in an increase in synapses. Furthermore, this remodeling is occurring in cortical lamina where sensory input is arriving. When you throw this together, and put some faith in the correlational nature of this study, it looks like we are seeing the functional rewiring of the visual circuit that accounts for an optical dominance shift seen following monocular deprivation; in other words, we have presumably witnessed an initial loss of contralateral input, followed by formation of ipsilateral input (remember we are looking at binocular V1).
Of course, the big question after seeing all these dendritic dynamics is what is actually going on with the connectivity; what’s happening to the synapses? We saw earlier that Chen and friends made a lame duck attempt at proving their newly elongated branch tips contain synapses, and we have to assume their sample size of one is accurate here, but to change the circuitry you might need rewiring at higher levels than just the Layer 2/3 interneurons. The group started by looking at boutons on Layer 2/3 interneuron axons. They see that under normal conditions, about 2.5% of boutons are eliminated and a similar proportion of new ones form. However, under monocular deprivation, bouton elimination drops to 0.7% and formation increases to a whopping 7.5%. This is a solid finding, and probably reflects the real synapse dynamics, but boutons aren’t the best readout of synapses, since they are identified by eye, and telling whether or not a bouton is a synapse or just a bump isn’t exactly a science. The best way to look at synanpses is to look for actual synaptic machinery immunohistochemically. Thankfully, Chen and friends also stained for VGAT, the transporter that loads GABAergic vesicles with GABA, and which thus labels inhibitory synapses. Using this approach they assessed inhibitory synapse number onto Layer 2/3 and Layer 5 pyramidal neurons, both of which receive input from Layer 2/3 interneurons. While they found no change in inhibitory synapses onto Layer 2/3 pyramidals, they saw a 40% increase in inhibtory synapse density on Layer 5 pyramidal neuron apical dendrites. Whether or not these new synapses are strictly from Layer 2/3 interneurons isn’t clear from this data, but given the increase in axon bouton density, it seems likely that this is the case. One piece of information that the group failed to provide was where the axonal boutons were located – if there was an increase in their numbers in Layer 5, I would be more convinced by this section of the paper. However, it looks likely that synapse number is indeed increasing.
Now for the sensationalist part. Chen and friends hit on a previous paper indicating that fluoxetine (read Prozac) “restores a juvenile level of ocular dominance plasticity in the adult […] rodent,” due to its inhibition of restrictive, intrahemispheric GABAergic transmission. In agreement with this finding, Chen and friends saw that fluoxetine almost doubles Layer 2/3 interneuron branch dynamics when compared to control. This on its own is pretty cool – a structural correlate of antidepressant treatment. But they took it a step further. You will recall from earlier that monocular deprivation results in an initial decreased number of elongations, followed by an increase. When Chen and friends combined fluoxetine treatment with monocular deprivation they did not encounter this early decrease in elongations! Without looking at synapse turnover rates, it is hard to tell what this means, because it could either be that the same old synapses are staying around, and there is increased elongation OR that the synapses are disappearing as with normal monocular deprivation but the plasticity that will eventually restore synapses is initiated earlier, potentiall leading to a quicker ocular dominance shift.
Being optimistic and slightly irresponsible, it is encouraging to think that the latter is the case because this could mean that combining fluoxetine treatment with some kind of instructive stimulus – such as cognitive behavioural therapy or talk therapy – might lead to quicker and potentially more effective rewiring of brain circuitry, improving recovery from depression, or perhaps even making it possible in intractable cases. I’m sure this is already going on to some extent, or at least I hope so, but it is comforting to see that there is at least a far fetched and whimsical cellular argument for its success.
With that, we are left with a major cliff hanger. Hopefully you have as many follow up experiments flitting around your head as I do. To begin with, Chen and friends finish their paper with the evidence that fluoxetine increases Layer 2/3 interneuron dendrite elongation immediately following ( ie 4 days post-) monocular deprivation. But in the rest of the paper they also looked at branch tip dynamics at 7 days post-deprivation – what happens after 4 days when normal monocular deprivation starts to cause increased elongation? Are there new synapses formed already? Do the elongation dynamics increase even further? Then, does fluoxetine paired with deprivation cause more inhibitory synapses to form than just deprivation on its own? Does it increase the extent of the ocular dominance shift? Is there a detrimental effect? If you open the eye again following short deprivation does the fluoxetine condition allow for better recovery? There are many questions to ask, but I will leave these few with you and open it up for discussion below.