LGN to Ctx anatomy
Notes from the literature:
Coleman et al. 2010 (Bear Lab, mouse):
During the critical period, density of thalamocortical (TC) synapses identified using EM drops by ~30% after 3 days of MD. Terminal area also decreases after 3 days MD (also EM). At 7 days MD, there is no significant difference in TC synapse density or size. Conclusion drawn from this is that first contralaterally-driven TC input decreases during the 3 days of MD and that ipsilaterally-driven TC input increases after 7 days MD. However, no direct evidence is presented to support this, all synapses were from a heterogeneous population of contra and ipsi LGN cells in the binocular region of the LGN.
Erisir and Dreusicke 2005 (ferret):
Density of TC synapses drops some 30% from P35 to P90, but the average size of the TC synapses increases over the same time
Prevailing interpretation is that connections are "inappropriate"; but Arani finds very compact RFs in his intracellular recordings; another possibility is that connections are NOT inappropriate, but initially an individual LGN cells makes many synaptic contacts onto a cortical neuron, that later become fewer, larger synaptic contacts. (Is specificity of initial connectivity our big question?)
In addition, the proportion of TC synapses onto GABAergic neurons in layer 4 goes down over the same time period
Troyer and Miller (http://www.jneurosci.org/content/18/15/5908) calculated that a given cortical cell probably receives between 50-150 LGN inputs by
assuming that there are 1000 X-type retinal ganglion cells per mm^2
assuming that each X-type retinal ganglion cell makes a single contact with 4 postsynaptic LGN neurons
assuming that each cortical cell samples space according to a probability function with a gabor shape (that matched the cortical cell's receptive field)
Sharpening of Topographical Projections and Maturation of Geniculocortical Axon Arbors in the Hamster (Journal of comparative neurology 1988 277(4).)
Naegele, Janice R., Jhaveri, Sonal, and Schneider, Gerald E.
Fine grain survey of retrograde labeled thalamocortical axons in hamster from immediately after birth to adulthood. There are two very interesting findings. 1) the pool of LGN cell projecting to cortex shrinks both in number and in spatial extend within the LGN during development. Consolidation of cells projecting to a given injection site appears adult-like a few days after eye opening, but does appear to perhaps undergo some experience-dependent refinement. 2) Axonal arbors in V1 begin very simple and small and continue to become more complex after eye opening. However, visual experience seems necessary for spatial restriction of the arbor and development of type-specific (two types in adults) morphology, both through retraction of off-target branches and outgrowth of new branches.
Suppose we sparsely labeled LGN neurons with CRE and sparsely labeled cortical neurons with CRE. Sometimes, the labeled cells will be connected. (In a variation, we could examine cells with antibodies, image the results to get the cell type, and then not expand the marker proteins.) By also labeling VGLUT2 and PSD-95, we could observe:
The number of LGN synapses of any type onto the cortical neuron
The number of LGN synapses from THAT cell onto the cortical neuron (is it a single site, 5 sites, 15 sites, ?)
By dividing the number of LGN synapses of any type onto the cortical neuron by the number of LGN synapses formed by a specific LGN neuron we could estimate the number of LGN neurons that contact a given cortical cell
The spatial extent of the LGN axon in the cortex**
The same measurements could be done in animals at different stages of development (maybe 3 stages??) to observe how the number of synaptic contacts from individual neurons onto single neurons change, as well as the extent of the LGN axons.
The same measurements could be done with monocular deprivation. Different conclusions in the literature: Bear/Nahmani say there's a decrease in synapse density, but Silver and Stryker (in cat) say no decrease in total synapse density until 4-7 days of monocular deprivation.
Could be done in ferret or mouse (or both).
By doing these studies at earlier ages in ferret, could examine the number of contacts of individual LGN neurons onto subplate neurons (maybe this is where many initial synaptic connections are refined)
** done by Kaas's lab, "The postnatal development of geniculocortical axon arbors in owl monkeys"; found that axon arbors did not decrease or increase for P cells and for M cells, axon arbors increased rather than decreased
** maybe also done by Antinini and Stryker, but with different conclusions
Who cares how many LGN cells converge onto 1 cortical cell? We know it is probably between 50-150 in cat/ferret and maybe less in the mouse. Suppose we find it is 20, or 150, what does that tell us about how the brain works?
It tells us a lot. Are cortical cells "relaying" or are they performing some computation? It could also greatly inform our planned electrophysiological experiments to study convergence in ferret. It would also greatly inform our models of development (do these counts change over typical development?)
Is there a massive re-organization with normal experience, or a tiny re-organization with normal experience?
If we could address re-organization of different inputs (such as from the two eyes) then that would tell us a lot about how experience could, in principle, modify these connections.
Papers still to read:
Those papers on reorganization of axons after monocular deprivation in Stryker's lab
Resolution (in cortex and behavior) continues to be refined over some 4-8 weeks of experience. I have always naively thought this was due to elimination of "inappropriate" connections in the cortex, but maybe it's all in the LGN input?
During early development (first arrival of LGN axons into layer 4) do we know anything about the number of synapses and branches? Is there elimination there? (There is eventually elimination there but it is unclear if number of contacts of LGN neurons change.)
What about S1?
Activity-dependent cre expression vector show to be capable of labeling eye-specific inputs to V1.
See especially Fig. 2 and Fig.3