1. Peter Andras School of Computing and Mathematics Keele University
Voltage – sensitive dye imaging and computational modelling of the crab stomatogastric ganglion
Peter Andras
School of Computing and Mathematics
Keele University

2. Overview Crab stomatogastric ganglion (STG)
Voltage-sensitive dye (VSD) imaging
VSD imaging of the crab STG
Computational analysis of the imaging data
Computational modelling of STG neurons and networks
Summary

3. Why study the STG ? Stomatogastric ganglion (STG) of crabs
26 neurons arranged in a relatively flat sheet
Relatively isolated (one input nerve from higher ganglia)
Complex behaviour – central pattern generators (CPG) – pyloric and gastric mill rhythms
Ideal model system for studying neural activity patterns, neuromodulation, motor pattern generation, restorative re-wiring

4. Voltage – sensitive dye (VSD) imaging
The dye molecules attach to the cell membrane, pick up photons (green), go through internal reorganisation (charge shift) and then return to the baseline state and release a photon (red fluorescence) – this is what we measure: the change in the intensity of the emitted fluorescent light; very small up to 10 % intensity change per 100mV change in the membrane potential
The membrane potential difference represents an electric field across the membrane and this field may facilitate or may reduce the chance of return of the excited molecules to their baseline state, influencing the intensity of the generated fluorescent light
The dye may be injected into selected neurons or the whole ganglion may be bathed in a dyed saline solution
Red photon
MiCAM 02 SciMedia
Green photon
Olympus BX 51 WI

5. VSD impact on the STG Do the neurons take up the dye ?
Is the dye toxic or not ?
Does it have a significant impact on what the neurons do or not ?
Stein, W, Andras, P (2010). Journal of Neuroscience Methods, 188:

6. VSD impact on the STG
There is an impact on STG neurons with strong light on the neuropil or neuron cell bodies, but no impact if light is directed onto axons
The impact is reversible following the stopping of illumination
Low light level minimizes the light induced impact

7. Imaging of dye filled neurons
Neurons are filled through micro-electrodes with di-8-aneppq dye using 1s duration steps of positive current (+10nA)
Alternative: dye filling with pico-spritzer
Current based dye filling takes around min per neuron, much quicker with pico-spritzer
Stein, W, Städele, C, Andras, P (2011). Journal of Neuroscience Methods, 194:
Stein, W, Städele, C, Andras, P (2011). Journal of Visualized Experiments, doi: /2567.

8. Imaging of dye filled neurons
Can we see the temporal difference in the recorded signal that is due to the propagation of the signal within the neuron ?

9. Imaging of dye filled neurons
Simultaneous imaging of a pair of interconnected neurons (PD and LP)
Can we record the spikes reliably using averaging ?

10. Imaging of dye filled neurons
Can we record the spikes reliably using single-sweep recording without averaging ?
Can we record small membrane potential change events (e.g. the impact of inhibitory input) ?

11. Imaging of dye filled neurons
Importance:
The recording of interconnected neurons in their physiological environment is critical for the understanding how neurons interact to deliver the emergent functionality of the neural system formed by them
Most other systems where simultaneous recording of neurons is possible and demonstrated it is either not know whether the recorded neurons are connected or not (it is also not known way how they are connected – e.g. mammalian cortex, vertebrate retina, large invertebrate ganglia) or the functionality of the system formed by the neurons is not known (e.g. neurons in cell culture)

12. Optical recording of many neurons
Bath application of the dye to the whole ganglion using dyed saline with di-4-ANEPPS dye
Takes around 30 minutes to bath the ganglion that is followed by 30 minutes of washing away of the excess dye
Städele, C, Andras, P, Stein, W (2012), Journal of Neuroscience Methods, 203: 78-88

13. Optical recording of many neurons
Typically we have can have around 20 neurons in the field of view of the camera (compare with max 4 neurons that can be recorded using intracellular electrodes)
Averaged recording over many cycles

14. Optical recording of many neurons
Can we record reliably individual spikes, general spiking activity and the impact of inhibitory event following the bath loading of the VSD ?

15. Optical recording of many neurons
Can we identify and record spike bursts and individual spikes within a burst generated by neurons following bath loading of the dye ?

16. Optical recording of many neurons
Recording of lobster STG neurons – 3D arrangement
Can we record slow and fast activity of neurons in details ?

17. Optical recording of many neurons
Can we record neural activity in axons within nerves ?
Can we track the propagation of activity within the axon ?

18. Optical recording of many neurons
Importance:
Bath loading of the dye reduces the time required to load many neurons
We can record the simultaneous activity of many STG neurons (much more than what is possible using microelectrodes)
The method allows the investigation of the emergence of joint activity of neurons in neural systems in their natural physiological context
We can track the activity travelling in single axons within a bundle of axons in a nerve

19. Computational data analysis
Recording of multiple PY neurons without and with exposure to dopamine
Expectation: dopamine de-synchronises the activity of PY neurons

20. Computational data analysis
Depolarized activity plateu
Feature points:
minimal slope
maximal slope
beginning of top zero slope
end of top zero slope
Trace features:
length of depolarized activity plateu
length of hyperpolarised inhibition period
Joint activity features:
length of temporal distance between matching feature points
Hyperpolarised inhibition period
Delay between matching feature points
Steyn, J, Andras, P, (in preparation)

21. Computational data analysis
The dopamine has differential effect on different PY neurons, shifting their feature points differently through the modulation of their activity  De-synchronisation of PY neurons

22. Computational data analysis
11 pairs of PY neurons compared using the measurements of 4 different feature points
F-test used to compare variances of temporal differences for the with and without dopamine cases
22 out of 44 statistical test indicate that dopamine de-synchronises the PY neurons, only one indicates the opposite, while 21 tests show no significant change

23. Computational data analysis
Importance:
Novel method for reliable analysis of imaging data for the assessment of the dynamics of temporal relationships between neural activities
First time to show in a physiologically valid context that the expected de-synchronisation of PY neurons following dopamine exposure actually does happen

24. Computational modelling
Hodgkin – Huxley models of STG neurons
Include equations for Calcium concentration
Two compartments: axon and soma/dendrite
Typically about 20 differential equations per neuron

25. Computational modelling
Others have shown that many conductance parameter settings lead to very similar behaviours of model neurons
The conductances obey correlation rules – these are maintained by exposure to neuromodulators and get relaxed in the absence of this
Often published models are very finely tuned and small changes stop the proper working of these models
Often joining model neurons that work well independently leads to non-functional network models

26. Computational modelling
We built a large working simulation of the pyloric rhythm network of the STG – 1 AB, 2 PD, 1 LP, 5 PY neurons – about 180 differential equations
Novelty: neurons belonging to the same class are simulated by multiple model neuron – others usually model these by a single model neuron
Aim: model the de-synchronsation impact of dopamine on PY neurons

27. Computational modelling
Measurements:
Inter-Spike Interval (ISI) the distance between the temporally closest spikes considering pairs of PY neurons
Spike Distance (SD) uses ISI to calculate a more precise measure of the synchronisation of two spike trains
Both are implemented in the Spiky spike sequence data analysis tool developed by Thomas Kreuz

28. Computational modelling
We simulated slightly different PY neurons by randomly setting their conductance parameters while respecting the correlational rules for conductances of appropriate ionic currents
100 simulations were run and we measured the ISI and SD spike synchronisation measures considering pairs of simulated PY neurons
Dopamine exposure was simulated by reducing the strength of gap junction connections between simulated PY neurons – this effect is known from experimental data

29. Computational modelling
Sufficient reduction in gap junction strength causes statistically significant change in the ISI and SD measures of PY synchronisation
Steyn, J, Alderson, T, Andras, P, (in preparation)

30. Computational modelling
Computational analysis of the impact of the variability of conductance parameters on model PD neurons
Biological PD neurons are never fully synchronous and the delay between spikes changes and the joint spike patterns may change under the impact of neuromodulators

31. Computational modelling
One PD has fixed conductance parameters the other has variable parameters
The models reproduce the observed biological behaviour of PD neurons
Dos Santos, Andras, P, (in preparation)

32. Computational modelling
Importance:
Computational modelling can lead to testable hypotheses about the explanations of observed biological phenomena
Separate modelling of multiple neurons of the same kind can help in understanding and modelling of activity and modulation dependent changes in the roles of neurons

33. Other related work
New voltage sensitive dyes based on the Bodipy molecule – in collaboration with Prof Andrew Benniston (Newcastle), current Leverhulme project
Combined use of MEA stimulation and VSD recording to restore activity in damaged STG – recent project in collaboration with Prof Alex Yakovlev and Dr Patrick Degenaar (Newcastle)

34. Conclusions
VSD recording of the STG provides a radically new neural system in which deciphering of how neural interactions lead to emergent system level behaviour becomes possible – the closest alternative systems (in terms of accessibility and functionality) are the enteric plexus of guinea pigs and snail & nudibranch ganglia, but none of them is known in comparable details as the STG
The ability to record individual axons in nerves provides an opportunity for unprecedented insight into how complex control of the motor patterns is realised by higher neural centres in the context of the STG
The reliable computational analysis of noisy optical imaging data is complicated and we developed novel methods that support this – we showed statistically convincingly the expected de-synchronisation of PY neurons in the STG
The computational modelling of separate copies of neurons of the same kind allows the computational analysis of impacts of neuromodulation of neural activity synchronisation and also on other kinds of correlated neural activities in principle, potentially leading to new testable hypotheses about dynamic changes in the roles of neurons

35. Acknowledgements Filipa Dos Santos (Keele) Jannetta Steyn (Newcastle)
Thomas Alderson (Newcastle / Ulster)
David Fourie (Newcastle / Singapore)
Wolfgang Stein, Carola Staedele (Illinois State)