fixing faulty circuits + drift-diffusion model

Question 1

briefly, what is the general idea behind fixing faulty circuits?

Answer 1

Forcing neurons that arent working to activate/remain inactive when they are supposed to, mainly using optogenetics atm

Question 2

what is one method of fixing faulty circuits?

(hint - channelrh___)

Answer 2

Express channelrhodopsin to depolarise target neurons in blue light

and/or halorhodopsin to hyperpolarise in yellow light

Question 2

where is halorhodpsin potentially useful?

is it reasonable to believe ChR would work?

Answer 2

Halorhodpsin can be useful in neurological disorders involving overactivity, like epilepsy

ChR has been expressed in the motor cortex of mice, blue light stimulation caused the mouse to run around/move a lot. So it works in vivo, theoretically could work in humans

Question 3

explain how small organic compounds can be used to fix faulty circuits

Answer 3

Small organic compounds that change between cis and trans isomers via light

Delivered by injection
The fixed key part of the molecule is the azobenzene hinge (– benzene ring – N=N – benzene ring –)
with the ‘active’ parts of the molecules on either side, which can be changed

the change in configuration could cause blocking of a channel for instance…

Question 4

explain how small organic compounds can allow for photocontrol of the GABA receptor

Answer 4

Basically you put a structure that will bind away from active site of the GABA receptor (maleimide) on one end of the hinge (for the cis attachment), then a GABA ligand on the other end to activate the receptor when in the trans configuration. Use light to change the configuration, going between on (hyperpolarised as it’s a Cl- channel) and off

Question 5

what are some important details about these small organic compounds?

hint - size, and specificity/precision

Answer 5

the molecules are small enough to fit through the pores of some receptors, like TRPV1 and P2X, so can enter the neuron via these, and block other channels from the inside

Note - this can be super specific; the diffraction limit of light is 240-250 nm, much less than size of neuron so you can activate a small section of the membrane

Note - possible for Ach receptors, K+ channels, glutamate receptors

Question 6

what is retinitis pigmentosa?

what might you do to treat it?

Answer 6

Blindness - gradual reduction of visual field due to degeneration of photoreceptors

Want to use optogenetics to stimulate the retina
OR may need to stimulate the visual cortex (e.g. if the optic nerve is damaged/did not develop, stimulating retina would be pointless)

Question 7

what are three problems when trying to use optogenetics for treating retinitis pigmentosa?

Answer 7

Going straight to the visual cortex means you lose the complex computations performed by the retina

Different ganglion cells in retina have different functions and project to different brain areas, e.g. saccadic eye movements input to superior colliculus, which is deeper in the brain and harder to reach to stimulate

Even if stimulating the retina, Stimulation of RGC with simple stimuli is useless (e.g. a motion sensitive RGC is active only when there is motion). The different ganglion cell types respond to different stimuli

Question 8

how could electric stimulation be used in the treatment of retinitis pigmentosa?

Answer 8

Basically use a microelectrode array in the ganglion cell layer (more accessible?), ideally @ the fovea as this processes high spatial resolution details
You stimulate the electrodes based on whats infront of the patient

Problems so far -
This has been done, but so far patients have only been able to see light, not make out objects/read etc…

Question 9

what are some issues/shortcomings when using electric stimulation to treat retinitis pigmentosa?

Answer 9

This problem may be due to stimulating RGCs but not photoreceptors or bipolar cells

If you stimulated a direction specific ganglion cell, you could cause perception of motion that isn’t actually there in a patient

***You need to consider that there are different types of GC cells, and need to stimulate them differently

Question 10

how has a mouse study got around some of the problems encountered when trying to treat retinitis pigmentosa with optogenetics?

what did the study find?

Answer 10

stimulate deeper layers like bipolar cells or even the remaining photoreceptors

Done in a mouse study -
Expressed halorhodopsin in remaining photoreceptors

GCs were still responding, meaning everything downstream of photoreceptors was still functional

Saw that - direction selectivity, centre-surround organisation, ON vs OFF cells still in tact

Treatment of the mice restored vision enough to navigate a water maze
Currently - moving tot rail with patients, using channel and halorhodopsin to mimic on and off cells, and targeting BP cell dendrites

Question 11

epilepsy -
what is the problem and consequence?

what are the possible causes?

Answer 11

Problem - too much excitatory (glutamate) input vs too little inhibitory input (GABA)
ALSO could be due to astrocytes - e.g. not clearing synaptic cleft of excitatory NTs
Result = seizures

Could be due to mutations increasing Na+ channel activity increasing excitability, or decreasing K+ channel activity reducing hyperpolarisation

Question 12

What are two obvious options for treating epilepsy with optogenetics?

Answer 12

Channelrhodopsin in GABA neurons to increase inhibition
OR
Halorhodopsin in glutaminergic neurons to decrease stimulation

Question 13

what are four problems/choices that need to be made when treating epilepsy with optogenetics?

Answer 13

The brain is big - you need to know which part you are targeting specifically, which changes from patient to patient (and possibly seizure to seizure?)

When to use the light - you must be able to detect the seizure just before it starts, in order to know when to start the light

Do you reduce excitation, increase inhibition or even target astrocytes? again likely to differ between patients

Chronically implanted electrodes cause death of surrounding neurons over time, which may also happen due to light - high in energy etc…

Question 14

mouse models for epilepsy have shown?

Answer 14

Experiments have used both halorhodopsin (decrease excitation) and channelrhodopsin (increase inhibition)

Have been shown to reduce excitation/stop seizures in temporal lobe in one experiment and the thalamus in another

Question 15

what are David Marr’s three levels of analysis, and what are they in the context of this lecture?

Answer 15

1. Computational problem. What problem are you trying to solve?
   How do we integrate noisy sensory input and decide the correct action?

2.Algorithm. What abstract method(s) of processing information could you use to solve this problem?
Drift-diffusion model of evidence accumulation

3.Physical implementation. How can you implement the algorithm to solve the problem, on actual physical hardware, like a biological brain?
Ramping up neural activity to a threshold

Question 16

explain what the drift-diffusion model is

Answer 16

Neutral starting point (tho in some cases could this be biased towards one outcome?)

Evidence is accumulated - as in sensory input. This rate of evidence accumulation is random and so trajectory fluctuates as you take in evidence for either option. The clearer the evidence/the less noise there is, the quicker you reach a decision

Boundaries - these are predefined thresholds that trigger a decision when reached - two boundaries (UB and LB), one for each alternative. Evidence accumulates towards a boundary, crossing it results in a choice being made for the corresponding option

Question 17

what typical test is used when investigating decision making?

Answer 17

the subject looks at a view of moving dots and must look right if they decide the dots are moving right etc…

Reaction time = time between beginning of stimulus (dots moving) and eye saccade (eye moving one side or the other)

Question 18

explain what’s going on with the distribution of reaction times when looking at a histogram

Answer 18

Gaussian distribution is expected for random variables, yet the distribution for reaction times from the moving dots experiment was non-gaussian (not normal/bell shaped) and instead was skewed right (longer tail to the right)

1/RT gave a normal distribution

Explained by the idea that it is the rate of evidence accumulation that is random, this has the gaussian distribution

Question 19

what is the relationship between reaction time and rate of evidence accumulation?

Answer 19

Reaction time is inversely proportional to the rate of evidence accumulation

Question 20

what experiment (seen before basically) was used on drosophila and what was seen?

Answer 20

What -
Trained to associate an odour with a shock, walk into a chamber where they can choose between a side with the bad odour, vs a much lower concentration of bad odour
In the ‘easy’ task it’s odour vs. 0.1 concentration
In the ‘hard’ task it’s odour vs. 0.9 concentration

Found -
Saw an increase in RT or hesitation the more similar the odour strength was, struggles to make decision and tries to collect more evidence (smelling both sides)

RT histogram also skewed right, 1/RT normal distribution also

Question 21

what was the experiment done on mice for decision making?

what was observed?

Answer 21

Mice take in a certain smell, a sound signals their time is up, they then must decide which side of the chamber to go to for a treat (so i’m assuming they would have been conditioned for each odour)

The longer the mice had to make the decision, the more accurate they were

The harder the task - (e.g. weak/mix of similar odours) the longer the mice needed to make a decision. More noise = needs longer to reach bounds

Question 22

speed-accuracy trade off -

what is it in terms of favouring accuracy, or favouring speed?

Answer 22

Accuracy - in order to favour being accurate, you would have higher decision bounds, so lots of evidence must be collected before threshold is reached

Speed - to favour speed, you would lower the decision bounds, so it’s quicker to reach the threshold, but less evidence will have been accumulated, and the decision is more likely to be inaccurate, especially if there’s more noise = more fluctuation in trajectory etc…

Question 23

in decision making (the DDM), how can the bou9ndaries be flexible?

Answer 23

possible that as time goes on during a decision making process, the boundaries gradually get lower, in order to ‘just make a decision’ if its taking too long for evidence accumulation to reach on threshold or the other

Question 24

how did scientists investigate what was happening in neurons during this decision making?

Answer 24

Scientists looked at part of the brain (LIP) just before the motor command to move eyes left or right

did the moving dot experiment and recorded from a neuron that increased in activity when the monkey was about to move its eyes to the right

Question 25

what did the scientists find when investigating decision making at the level of the neuron?

Answer 25

There is an increase/ramping up of activity in this neuron during sensory perception (looking at the dots when they are moving, to the right) - evidence gathering before the decision

The stronger the sensory input the faster the increase in activity/ramping up is seen. E.g. a stronger motion to the right = more evidence gathered all saying the same thing / less noise

The neuron was actually less active/inhibited the stronger the evidence for the opposite outcome - i.e. the more the dots moved to the left

when they aligned the trials by when the decision was made -
Found the same peak was reached for each ‘motion strength’ so regardless of how strong the evidence was. This represents the boundaries, which act as a threshold (got to peak of reach ‘motion strength 70 ms before the decision)

Question 26

what was seen in drosophila with the FoxP mutation?

why was this the case?

Answer 26

Fly experiment from earlier, but found that flies with a mutation in FoxP took much longer to decide (tho accuracy wasn’t affected)

FoxP mutant Kenyon cells had excess K+ channel expression, making them ‘leaky’

This slowed down their rate of evidence accumulation as its harder to depolarise the cell up to spike threshold