Optogenetic system and method

Abstract

An optogenetic system and method for preventing or halting seizures. The system and method use a sensor for monitoring the activity of a neural network containing a group of target neurons, and generating an input signal indicative of said activity, the target neurons being excitatory neurons. Excitatory stimulation is delivered in the form of an optical signal by an optical stimulator to the target neurons, the optical signal being determined based on the input signal, to reduce the overall activity of the target neurons.

Claims

1. An optogenetic system for preventing or halting seizures by stabilizing brain regions, the system including: a sensor for monitoring an activity of a neural network containing a group of target neurons, and generating an input signal indicative of said activity, the group of target neurons being exclusively excitatory neurons which express optogenetic actuators that cause the cells on which they are expressed to execute excitatory behavior in response to an optical stimulus; an optical stimulator for delivering an excitatory stimulus in the form of an optical signal to the group of target neurons, the optical signal determined based on the input signal, to reduce the overall activity of the group of target neurons; and a processor configured to generate an output signal based on the input signal, wherein the output signal is used to generate the optical signal which forms the excitatory stimulus, wherein: the input signal is oscillatory in nature, and the output signal has the same frequency as the input signal, and wherein the output signal is phase shifted relative to the input signal to provide a phase-shifted optical signal, and the processor is configured to determine or calculate the phase shift, wherein the phase shift is predetermined and fixed, and wherein the processor is configured to calculate the phase shift of the output signal based on the input signal, wherein the processor is configured to identify a range of phases during the oscillation of the input signal at which the delivery of the excitatory stimulus causes a brain region to approach a stable state, and to calculate the phase shift of the output signal based on this identification.

2. The optogenetic system of claim 1, wherein the sensor is located on an implantable device.

3. The optogenetic system of claim 2, wherein the sensor includes an electrode.

4. The optogenetic system of claim 3, wherein the sensor includes an array of electrodes.

5. The optogenetic system of claim 3, wherein the electrode is configured to monitor a local field potential of the neural network containing the target group of neurons.

6. The optogenetic system of claim 1, wherein the optical stimulator is located on an implantable device.

7. The optogenetic system of claim 6, wherein the optical stimulator and the sensor are located on the same implantable device.

8. The optogenetic system of claim 1, wherein the optical stimulator includes a light-emitting element configured to provide light having a light intensity of no less than 0.5 mW/mm.sup.2, and wherein the light emitting element comprises a light emitting diode or a laser.

9. The optogenetic system of claim 8, wherein the optical stimulator includes an array of light-emitting elements.

10. The optogenetic system of claim 1, wherein the input signal is an electrical signal.

11. The optogenetic system of claim 1, wherein the output signal is an electrical signal.

12. The optogenetic system of claim 1, wherein, either: (a) the processor is configured to detect, within the input signal, an event signal which is indicative of an onset or presence of a seizure or seizure-like event, or (b) the sensor is configured to detect an event signal which is indicative of the onset or presence of a seizure or seizure-like event, and to send the input signal, and information relating to the event signal to the processor.

13. The optogenetic system of claim 1, wherein the optical stimulator is configured only to deliver the excitatory stimulus in response to a detection of an event signal.

14. The optogenetic system of claim 13, wherein the processor is configured to generate the output signal only upon the detection of the event signal.

15. The optogenetic system of claim 1, wherein the phase shift is no less than 60° and no more than 165°.

16. The optogenetic system of claim 1, wherein the phase shift is either: (a) no less than −30° and no more than 30°, or (b) no less than 150° and no more than 210°.

17. The optogenetic system of claim 1, wherein the processor is configured to perform band-pass filtering on the input signal, the band-pass filtering having a predetermined filter frequency.

18. The optogenetic system of claim 17, wherein the filter frequency is no less than 1 Hz and no more than 20 Hz.

19. The optogenetic system of claim 1, wherein the processor is configured to perform thresholding and rectification on the input signal, such that an amplitude of the output signal is proportional to an extent to which the input signal exceeds a predetermined threshold.

20. The optogenetic system of claim 1, wherein the output signal is used to control an intensity of the light emitted by the optical stimulator to generate the optical signal which forms the excitatory stimulus.

21. The optogenetic system of claim 1, wherein the output signal is used to modulate a width or frequency of pulses of light, having a constant intensity, to generate the optical signal which forms the excitatory stimulus.

22. A method of halting or preventing seizures by stabilizing brain regions, the method involving the steps of: monitoring an activity of a neural network containing a group of target neurons, the group of target neurons being exclusively excitatory neurons which express optogenetic actuators that cause the cells on which they are expressed to execute excitatory behavior in response to an optical stimulus; generating an oscillatory input signal indicative of the activity; generating an output signal using a processor based on the input signal, wherein the output signal has the same frequency as the input signal, wherein the output signal is phase shifted relative to the input signal to provide a phase-shifted optical signal, and wherein the processor is configured to determine or calculate the phase shift; determining, based on the input-signal, an optical signal for stimulating the group of target neurons, such that when delivered to the group of target neurons, the overall activity of the group of target neurons is reduced; and delivering an excitatory stimulus to the group of target neurons in the form of the optical signal, wherein the phase shift is predetermined and fixed, and wherein the processor is configured to calculate the phase shift of the output signal based on the input signal, wherein the processor is configured to identify a range of phases during the oscillation of the input signal at which the delivery of the excitatory stimulus causes a brain region to approach a stable state, and to calculate the phase shift of the output signal based on this identification.

23. The optogenetic system of claim 1, wherein the group of target neurons are genetically modified to express the optogenetic actuators using a promoter, the promoter being present in excitatory neurons and not present in inhibitory neurons.

24. The optogenetic system of claim 23, wherein the promoter is an EMX promoter.

25. The optogenetic system of claim 24, wherein the promoter is an EMX1 promoter.

26. The optogenetic system of claim 24, wherein the modified gene product is EMX1.

27. The optogenetic system of claim 1, wherein the optogenetic actuators are opsins.

28. The optogenetic system of claim 27, wherein the opsins are Channelrhodopsin 2 (ChR2), Chronos, Chrimson, or ReaChR.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Embodiments of the invention will now be described by way of example with reference to the accompanying drawings in which:

(2) FIG. 1 shows a schematic diagram of an optogenetic system as provided by the present invention.

(3) FIG. 2 shows a flowchart of the steps performed by the controller shown in FIG. 1.

(4) FIGS. 3A and 3B show state space representations of the brain.

(5) FIGS. 4A and 4B show plots of the stimulation applied to the target neurons, and the activity monitored by the sensor, varying with time.

(6) FIGS. 5A, 5B and 5C show an embodiment of the optogenetic system of the present invention.

(7) FIG. 6 shows a series of plots of the activity of target neurons, when stimulated at different phase shifts by the optical stimulator.

(8) FIG. 7 shows a plot of power modulation for different frequencies and phase shifts of optical stimulation.

(9) FIG. 8 shows a plot of the time after seizure onset against the burst duration, with the points colour-coded according to the phase shift of the applied optical stimulation.

(10) FIGS. 9A and 9B show plots of burst durations against the phase shift, for (A) the first burst, and (B) the subsequent bursts.

(11) FIG. 10 is a schematic of closed-loop optogenetic experiment set-up.

(12) FIG. 11 shows, for an exemplary closed loop algorithm: A Example filter kernels used for filtering the LFP. Here the band-pass centre frequency is 10 Hz. B Gain response of the filter. C Phase response across the pass-band.

(13) FIG. 12 shows plots relating to the generation of rhythmical LFP oscillations by closed-loop optogenetic stimulation in neocortical slices from EMX-ChR2 mice.

(14) FIG. 13 shows a plot for optogenetic modulation of SLE durations by closed-loop optogenetic stimulation in neocortical slices of EMX-ChR2 mice.

(15) FIG. 14 shows an expanded plot for optogenetic modulation of SLE durations by closed-loop optogenetic stimulation in neocortical slices of EMX-ChR2 mice (Colour code for different phase-shifts is inset).

(16) FIG. 15 is a plot showing that closed-loop stimulation of excitatory network reduces high frequency power of the shortened SLEs during 90-180° phase-shifts.

(17) FIG. 16 shows plots of Epileptiform interictal spikes in a 4AP in vitro model and in a computational model.

(18) FIG. 17 shows plots of Epileptiform ictal discharge in the 4AP in vitro model and in the computational model.

(19) FIG. 18 shows plots from an example closed-loop stimulation on interictal spikes in vitro—(a) Several cycles of closed-loop stimulation with 8 phase conditions (0:45:315 degrees) shown in yellow. Grey is the recorded LFP time series, red shows the LED output signal. Each phase condition lasted 20 seconds. The LED was only switched on halfway through each phase condition. (b) Spectrogram of the recorded LFP time series in (a). (c) Zoom-in of a 90 degrees phase condition from (a).

(20) FIG. 19 shows plots from an example closed-loop stimulation on interictal spikes in silico—(a) Several cycles of simulated closed-loop stimulation with 8 phase conditions (0:45:315 degrees) shown in yellow. Blue is the simulated LFP time series, red shows the simulated LED output signal. Every setting is as in FIG. 20. (b) Spectrogram of the simulated LFP time series in (a). (c) Zoom-in of a 90 degrees phase condition from (a).

(21) FIG. 20 shows plots from an example closed-loop stimulation on ictal discharges in vitro and in silico. (a, b) Overview of the experiment. 5 complete cycles of 8 phase conditions (shown in yellow) were tested. Blue shows the raw LFP recording, and orange shows the LED light output. (c, d) Example ictal discharges from (a/b) colour coded by phase condition.

(22) FIG. 21 shows discharge duration in different phases of closed-loop stimulation compared to the unstimulated condition. Mean and standard deviation of discharge duration is shown for each phase condition (blue). The unstimulated condition is shown in grey. (a) In vitro durations. (b) In silico durations.

DETAILED DESCRIPTION

(23) FIG. 1 shows a schematic diagram of an optogenetic system according to the present invention. The system illustrated, and which is described below, was used in experiments to test the effectiveness of systems and methods of the present invention in mice. In the system shown, experiments were carried out on a cortical slice from an EMX-ChR2 mouse, having a thickness of 400 μm, shown schematically at the bottom of FIG. 1. ChR2 is selectively expressed in excitatory pyramidal neurons using an EMX promoter. A recording electrode is implanted into the cortical slice, in order to monitor the activity (in this case, the local field potential) of cells within the slice, which includes contributions from excitatory neurons (e.g. glutamatergic cells) and inhibitory neurons. In the experimental setup, the recording electrode is in the form of a multi-electrode array (MEA). Signals recorded by the MEA are then transferred to an amplifier, from which they are transferred to a controller in the form of an input signal. A closed-loop algorithm as described in the previous section is then applied to the input signal in the controller to generate an output signal, the closed-loop algorithm controlled by the PC connected to the controller. The output signal is then sent to the light source whereupon it is converted into an optical signal. In this embodiment, the optical signal (i.e. the excitatory stimulus) is delivered via a 200 μm patch cord, to a 473 nm LED cube, the intensity modulation of the light controlled by the controller. In order to induce seizure-like activity in the cortical slice, 4-aminopyridine (4AP) is applied.

(24) FIG. 2 shows a flow chart of the method performed by the controller (or processor) shown in FIG. 1. The input signal such as an LFP is received from the sensor (i.e. the recording electrode), at the controller, whereupon it is band-pass filtered, as described earlier in the application. The seizure frequency is generally 1-20 Hz, but this may vary from patient-to-patient. After band-pass filtering, the signal is phase shifted by Δϕ as discussed below. Then, thresholding and rectification is performed to give an output signal, which is then used to modulate the intensity or pulse width/frequency of light delivered by the stimulating optrode as shown in FIG. 1, to give a stimulation having a form similar to that shown in FIG. 4A or 4B.

(25) FIG. 3A shows a schematic state space representation of brain activity in a recorded area. Specifically, we take a neural population view/model, where a population of excitatory neurons, and a population of inhibitory neurons interact with each other. The activities of these populations are not fixed: rather, they are dynamic, and evolve continuously with time, and the instantaneous brain state is determined by the instantaneous activities of the excitatory population and the inhibitory population. This is shown in FIG. 3A, in which the activity of the inhibitory population (I) is plotted on the x-axis, against the activity of the excitatory population (E) on the y-axis. A given location on the graph represents an individual state, which in this simplified model, is determined by only the values of E and I. The arrows represent the evolution of the system, i.e. as it evolves from one state to another state. The central shaded region represents the set of stable states. In these regions, there is little to no fluctuation in the activities of the inhibitory and excitatory populations. This principle can be applied to background and seizure states.

(26) The state space representation of the brain is also useful for understanding what happens during a seizure. This is shown by the non-shaded region with the larger arrows, in FIG. 3A. During a seizure or seizure-like event, the brain state rapidly changes, moving elliptically as shown in FIG. 3A. This represents a relatively high-frequency, (usually around 1 to 20 Hz) fluctuation in the activities of the inhibitory and excitatory populations. Large excitatory activity excites the inhibitory population, which then in turn inhibits the excitatory population. The decrease in the activity of the excitatory population then reduces the excitation of the inhibitory population, causing the activity of the excitatory population to rise again. This process repeats itself to give the elliptical cycles shown in FIG. 3A.

(27) In the present invention, unlike known techniques, the seizures or SLEs are halted or prevented by the application of an excitatory optical stimulus to excitatory neurons. This may seem counter-intuitive, but the mechanism behind such a technique may be best understood with reference to FIG. 3B. In this plot, the ellipse represents the evolution of the brain state, and demonstrates its cyclical nature. By applying an excitatory stimulus, in this case optogenetically, the population of excitatory neurons in the neural network of interest (i.e. that which is monitored by the sensor) is increased. The effect of this increase in population is for the state to move to a vertically higher position on the plot shown in FIG. 3B. This is shown by the arrows on the bottom half of the ellipse. As can be seen, when the excitatory stimulus is applied at certain times in the cycle, the state is urged towards the stable region. By carrying out this process repeatedly, the brain state is stabilized, and the seizure is brought to an end, or prevented entirely. What is demonstrated by FIG. 3B is that it is not only the amplitude, but the phase of the stimulus which is instrumental in producing the desired response.

(28) This phase difference of the excitatory stimulus is shown more clearly in FIG. 4A, which compares a plot of the activity of the excitatory population and the applied stimulus, and how they change with time. For clarity, the excitatory population is shown here as sinusoidally-varying, but clearly this need not be the case. In the example shown in FIG. 4A, the excitatory stimulus is applied with a 90° phase shift, as compared with the excitatory population. The benefit of this may be seen from FIG. 4B. As well as being phase-shifted, the input signal (shown in the top plot) has also undergone thresholding/rectification. Hence, no stimulation is delivered to the target neurons during certain phases of the oscillation. This may be understood from the state space representation shown in FIG. 3B. The excitatory stimulus (of FIG. 3B) is applied only during the regions when the activity of the excitatory population is falling, i.e. on the lower half of the ellipse. It is during this stage of the oscillation that excitatory stimulation causes the activity of the excitatory population to increase, and thus move towards the target region. By applying the excitatory stimulus at during this phase of each oscillation, the state of the brain spirals in to the stable region of the plot. No stimulation is delivered during the stages where the activity of the excitatory population is increasing, i.e. at the upper half of the ellipse, since this would further destabilize the state of the brain and prolong the seizure.

(29) FIG. 4B shows a similar plot to FIG. 4A. In reality, it may be complex to measure the target activity of only the excitatory population. Instead, the LFP may be measured, which represents a combination of the activities of both the excitatory and inhibitory population. Similarly, the excitatory stimulation may excite neurons in both the inhibitory and excitatory population in some fixed proportion (i.e. the stimulation will cause a given state to move diagonally up and to the right, in the representation of FIG. 3B). Due to the contribution to the LFP of the inhibitory neurons, the LFP may be phase-shifted relative to just the activity of the excitatory population, and accordingly 90° may not be the ideal phase-shift. Rather, some phase shift Δϕ may be more effective. This parameter, may be empirically determined, e.g. on a patient-by-patient basis, and may depend on the relative proportion of inhibitory and excitatory neurons in a given neural network. The results of experiments at different values of Δϕ are discussed later in this application.

(30) FIG. 5A shows a diagram of an embodiment of the optogenetic system of the present invention in place in a patient or user. Here, the implantable device is in place in the user's brain tissue, with the optrodes (which are shafts with light-emitting elements and recording electrodes mounted thereon) extending into the brain tissue at the top of the head. The optrodes pass through a base plate, discussed in more detail with reference to FIG. 5C. Cables, e.g. ribbon cables, connect the end of the optrode (which is not implanted inside the brain tissue) to a base station located on the outside of the skull, which itself is connected to a chest unit via connecting leads. In use, the recording electrodes send the input signal to the chest unit via connecting leads to the chest unit. The chest unit contains the processor, which is used to process the input signal, and accordingly generate an output signal, which is then sent back to the optrodes via the connecting leads, in order to generate the desired signal in the form of an excitatory stimulus.

(31) FIG. 5B shows a more detailed view of an optrode. The optrode is made up of two main components: the CMOS head and the shaft. The shaft is an elongate body, having a thickness of approximately 200 μm, in the present embodiment. Three recording electrodes are located on the shaft, along with eight μLEDs. The three electrodes are evenly-spaced, with the μLEDs located in the spaces between the recording electrodes, also evenly-spaced. Each of the recording electrodes and μLEDs are connected to the CMOS head, shown at the rear end of the shaft. The connections are preferably inside the shaft itself. The CMOS head includes electronics for processing the signals received, and for sending the signals to the chest unit. The CMOS unit is also configured to send the output signal to the μLEDs, in order to generate the optical signals forming the excitatory stimulus. There may be recording electrodes and μLEDs present on both sides of the shaft, but only one side is shown in FIG. 5B.

(32) FIG. 5C shows the implantable device in more detail, with an array of fifteen optrodes in place. Specifically, each of the optrodes extends through an aperture in the base plate, with the shaft extending from one side, and the CMOS head at the other. A region of the base plate does not have any apertures, and acts as a connector area for connecting to a base station which, in use, is located on the outside of the skull. The base station may be connected to the CMOS heads of each of the optrodes, and also to the chest unit. In use, the shafts of the optrodes is located in the user's brain tissue, with the base plate, CMOS heads and base station located on the user's scalp.

(33) FIGS. 6 to 9 show results of the experiment set up as described above.

(34) FIG. 6 shows several plots, in which SLEs are aligned to the time of onset. In the drawing, different traces are colour-coded to represent the different phase-shifts of the signals applied to the target neurons, relative to the input signal. Black traces indicate no stimulation. A key feature of these plots is the length of the oscillatory component of the seizures. In particular, it can be seen that relative to no simulation, some phase shifts significantly decrease the length of the seizure, while other phase shifts lengthen it. Similar effects are evident from FIGS. 7 to 9. Specifically, FIG. 7 shows the power modulation at different frequencies during the SLEs produced by closed-loop stimulation with different phase shifts. Phase is unwrapped and plotted twice up the vertical axis. Red indicates increased power relative to no stimulation, and blue indicates reduced power. Note the band of blue across high frequencies at a phase shift of 90-180°. High frequency stimulation delivered to excitatory neurons can reduce the overall activity within the slice, if delivered at an appropriate phase relative to spontaneous activity.

(35) FIG. 8 shows a plot of the duration of bursts during SLEs. The dots are colour coded according to the phase shift of the applied stimulation, with the same colour code as for FIG. 1. Similarly, FIGS. 9A and 9B show plots of phase difference, and the burst modulation (i.e. the percentage during for stimulation on, relative to stimulation off, 100% means no effect, whereas 50% and 200% mean respectively that the duration is half and twice what it is without stimulation), for both the first burst of SLE activity, and subsequent bursts. It is shown that an stimulation tends to increase the length of the first burst, but for phase differences between 90 and 180°, this increase is smallest. Then, for all subsequent bursts, the duration is shortened (relative to no stimulation) for delivery of excitatory stimuli which differ in phase from the activity of the target neurons by 90 to 180°.

(36) Below are described two exemplary sets of experiments, one in vitro and the other in silico that have been conducted by the inventors to demonstrate the phase-dependent modulation of epileptic activity using closed-loop optogenetic stimulation.

Example 1—Phase-Dependent Modulation of In Vitro Epileptic Activity Using Closed-Loop Optogenetic Stimulation

(37) This example considers phase-dependent modulation of epileptic activity using closed-loop optogenetics in rodent brain slices selectively expressing Channelrhodopsins-2 (ChR2) either in excitatory pyramidal neurons using an Emx1 promoter, or in a subset of inhibitory cells using the parvalbumin (PV) promoter.

Experimental Details

(38) Brain Slice Preparation

(39) Coronal neocortical brain slices (400 μm) were prepared from Emx1-ChR2 and PV-ChR2 mice, which provides selective neuronal expression of channelrhodopsin-2 in glutamatergic cells (Gorski et al., 2002.sup.1). The mice were perfused using the same ice-cold oxygenated (95% O2/5% CO2) sucrose-containing artificial cerebrospinal fluid (sACSF) used for cutting the brain slices; (sACSF in mm: 252 Sucrose, 24 NaHCO3, 2 MgSO4, 2 CaCl2, 10 glucose, 3.5 KCl, 1.25 NaH2PO4). Rodent brain slices were cut using a 5100 mz vibratome (Camden Instruments). The slices were later transferred and incubated at room temperature in a brain tissue interface holding chamber until later electrophysiological recordings. During recordings, the slices were perfused with the oxygenated normal ACSF (in mm: 126 NaCl, 24 NaHCO3, 1.2 MgSO4, 1.2 CaCl2), 10 glucose, 3 KCl, 1.25 and NaH2PO4) held at 33-34° C. .sup.1Gorski J A, Talley T, Qiu M, Puelles L, Rubenstein J L, Jones K R (2002) Cortical excitatory neurons and glia, but not GABAergic neurons, are produced in the Emx1-expressing lineage. The Journal of neuroscience: the official journal of the Society for Neuroscience 22:6309-6314.

(40) Electrophysiological Recording

(41) To test the proposed closed-loop algorithm in vitro, local field potentials (LFP) from rodent brain slices were recorded, using these to control optical stimulation in a closed-loop manner. All the electrophysiological recordings were performed using an interface recording chamber and 16-channel linear multi-electrode array probe (NeuroNexus Technologies: A16×1-2 mm-100-177 probes—shanks are 100 μm apart; recording site area on each shank, 177). The recording sites on all shanks were located 50 μm from the tip of the electrode. Using this MEA probe, 1500 μm of brain tissue could be sampled at any time either in the same cortical layer or across cortical layers. The impedance of the MEA electrodes used ranged between 0.4-2 MΩ.

(42) LFP signals were amplified using a MP8I headstage and PGA amplifier (Multichannel Systems) with a combined gain of ×1000, and sampled with a Micro1401-3 data acquisition box (CED, UK) at approx. 10 KHz and visualised using Spike2 software running on Windows (Win 7) computer.

(43) Epileptiform Activity Patterns

(44) The convulsant compound 4-aminopyridine (4-AP; 200 μM) was bath applied to induce epileptiform activity in rodent brain slices. Two patterns of spontaneous LFP activity were observed after 4AP application which we have termed ‘interictal activity’ and ‘ictal burst activity’. ‘Interictal activity’ was characterised in the absence of optogenetic stimulation by generally flat LFP with only occasional LFP transients (‘spikes’) which did not develop into oscillations. Note that despite the absence of seizure-like events (SLEs), we call this ‘interictal’ due to the similarity between this pattern and interictal discharges observed in clinical recordings. ‘Ictal burst activity’ was characterised by intermittent LFP transients that developed into SLEs comprising multiple bursts of oscillatory activity. The difference between these patterns likely reflects inherent variability in the animals and/or brain slice preparations.

(45) Closed-Loop Optogenetic Stimulation

(46) Spike 2 software was used to select one of the acquired LFP channels as input to the closed-loop controller (FIG. 10). As seen in FIG. 10, the local field potential (LFP) recorded by one electrode is processed and fed-back via a real-time closed-loop algorithm to modulation an LED light source. An experimental session comprises periods of closed-loop stimulation using different phase-shifts, as well as interspersed periods with no stimulation.

(47) More specifically, the output of the controller was a 0-5V voltage signal sent to an external LED driver (DC4104; Thorlabs) using the appropriate cable (CAB-LEDD1; Thorlabs). The LED driver was configured such that the 0-5V voltage range was converted to a constant current range of 0-1000 mA. This current drove a blue LED light source (473 nm, M470F1; Thorlabs) coupled to a 200 μm diameter optical fibre (M89L01-200; Thorlabs). The LED driver DC4014 was connected to the LED light source M470F1 via the DC4100-HUB connector hub from Thorlabs. For control experiments, we used a different wavelength of light source (590 nm, 590F2; Thorlabs), which falls outside the activation spectrum of ChR2 opsins.

(48) Closed-Loop Algorithm

(49) The closed-loop algorithm was implemented in custom-designed hardware based around a PIC dsPIC30F4013 microcontroller running at 30 MHz. The microcontroller sampled the LFP signal from one electrode at 500 Hz, applied a phase-shifting finite impulse response (FIR) filter, thresholded (above the background noise level) and half-wave rectified this signal to generate an output which controlled the LED intensity. The FIR filter convolved the input signal with a kernel given by:
e.sup.−kt.Math.cos(2πft+φ)
where k determines the filter band-width and was equal to 1.25; φ determines the extent to which the output is phase-advanced from the input and cycled from 0 to 315° in 45° steps; f determines the center frequency of the pass-band and we used different values in different experiments. In general we chose f to reflect the dominant frequency of ‘ictal burst activity’ (10-20 Hz), and for ‘inter-ictal’ recordings, we chose frequencies between 2-20 Hz. The total kernel length was 512 samples. FIG. 11 shows typical filter kernals used, as well as the resultant gain profile and phase-shifts. The overall gain of the closed-loop algorithm was chosen such that typical seizure activity resulted in a full-scale activation of the LED light source, with minimal saturation.

(50) Experiments comprised periods of closed-loop stimulation (LED On) and periods of no stimulation (LED Off). For ‘ictal burst recordings’, the duration of these periods was adjusted to ensure that generally at least one spontaneous seizure occurred during each duration.

(51) Analysis

(52) Data were analysed using custom scripts written in Matlab (Mathworks, USA). For ‘interictal activity’, frequency-domain analysis was performed on the entire period of each stimulation condition. For ‘Ictal burst activity’ we analysed only time-periods containing SLEs. The onset of each SLR was identified by an initial threshold crossing below −0.3 mV. Analysis was performed only for a time window of 18 s, chosen to capture the duration of SLEs. In each case, average power spectra for the LFP signal were compiled over the relevant time periods for each condition. Power modulation was calculated as a ratio (in decibels) relative to power during LED Off periods.

(53) In addition, the duration of seizure bursts within SLEs was calculated. The LFP was first high-pass filtered (at 8 Hz) and rectified the LFP. A burst was defined as any time-period for which this signal was not less than 0.1 mV for more than 0.2 s. Bursts with a duration less than 10 ms were excluded. Because the duration of the first burst within the SLE was typically longer than subsequent bursts, these were analysed separately.

Results and Discussion

(54) Excitation of Glutamatergic Pyramidal Neurons During ‘Interictal Activity’

(55) In neocortical slices from Emx1-ChRs mice which did not exhibit spontaneous SLEs (‘interictal activity’), closed-loop optogenetic stimulation could reliably elicit rhythmical LFP oscillations (FIG. 12). In FIG. 12, A shows example traces with no stimulation and during closed-loop stimulation with different phase-shifts; B shows LFP power spectra for different stimulation conditions; and C shows power modulation (relative to LED-Off) as a function of frequency and phase-shift. Note that phase is unwrapped and plotted over two cycles. In this case, closed-loop stimulation generated an oscillation with a fundamental frequency that increased with increasing phase-shift.

(56) In general, these oscillations resulted from a spontaneous large amplitude spike which was fed back by the closed-loop algorithm. We found that the fundamental frequency of the oscillations (and the higher harmonics) increased with increasing phase-shift. Note that by convention we denote positive phase-shifts as a phase-advance of the output relative to the input. Therefore we can alternatively consider that the frequency of oscillation decreased with increasing phase-delay of the (output) optical stimulation relative to the (input) LFP. This makes intuitive sense since the stimulation resulting from a brief LFP spike will resemble the impulse response of the filter. Thus, the peaks of this impulse response (and hence the timing of optical stimulation) will occur later with increasing phase-delay, leading to a slower oscillation. Note also that the amplitude of the induced oscillation varied with stimulation condition, and was greatest when the phase-shift generated a frequency of oscillation that matched the pass-band of the filter, while at other frequencies the closed-loop stimulation did not generate sustained oscillatory activity. However, since the ‘interictal activity’ was characterised by an absence of oscillatory activity in the LED off condition, a general increase in LFP power was observed in all stimulation conditions.

(57) Excitation of Glutamatergic Pyramidal Neurons During ‘Ictal Burst Activity’

(58) Also examined was the effect of closed-loop optical stimulation of pyramidal neurons in the case that spontaneous SLEs were induced by 4-AP. FIG. 13 shows all SLEs observed in a single experimental session, aligned to the event onset and colour-coded according to stimulation condition (see FIG. 14). In this recording, SLEs consisted of an initial negative deflection followed by a burst of oscillation at around 15-20 Hz lasting a few seconds. After a period of quiescence, subsequent bursts were seen. Each entire SLE lasted about 18 s and occurred with an inter-event interval of approximately 1 minute. Interestingly, closer inspection of the SLEs (FIG. 14) suggests that the duration of the oscillatory bursts varied with closed-loop stimulation condition.

(59) FIG. 15A shows power-spectra for SLEs divided according to stimulus condition, and FIG. 15B shows the modulation of power at each frequency relative to the LED Off condition as a function of the phase-shift applied by the closed-loop algorithm. Again it is clear that some phase-shifts induce a pronounced oscillation with a fundamental frequency (and higher harmonics) that increased with increasing phase-advance (or decreasing phase-delay). Note however that since there was spontaneous activity in the LED Off condition, closed-loop stimulation could either increase or decrease power at different frequencies depending on the phase-shift used. Phase-shifts between 90-180° resulted in reduced power at the dominant frequency of the spontaneous SLEs (15-20 Hz). Interestingly these phase conditions also led to a reduction in power at higher frequencies up to 1000 Hz in the LFP. These higher frequencies are thought to reflect the spiking activity of local neurons, including the pattern of recruitment of neurons to seizures recorded in human patients (Schevon et al., 2012; Weiss et al., 2013.sup.2). Therefore it appears that excitatory stimulation, delivered selectively to excitatory cells, could nevertheless result in a net reduction of activity during SLEs when timed appropriately to the ongoing seizure oscillation. .sup.2Schevon C A, Weiss S A, McKhann G Jr, Goodman R R, Yuste R, Emerson R G, Trevelyan A J (2012) Evidence of an inhibitory restraint of seizure activity in humans. Nature Communications 3:1060

(60) The duration of oscillatory bursts within each SLE was also examined. The duration of the first burst within the SLE was either unchanged or extended depending on closed-loop phase-shift (15C). However, the duration of subsequent bursts was suppressed for phase-shifts between 90-180° (15D) (the same conditions associated with a broad reduction in LFP power). In FIGS. 15C and 15D, the red dashed line shows average duration in LED Off condition. Shading indicates s.e.m. Note that phase-shifts associated with higher/lower LFP power are also associated with longer/shorter oscillatory bursts.

Example 2—Phase-Dependent Modulation of in Silico Epileptic Activity Using Closed-Loop Stimulation

(61) This example considers computational modelling work that parallels the in vitro closed-loop optogenetic stimulation experiments discussed above.

(62) Methods

(63) Modelling Epileptiform Activity

(64) The model used here is a variant of the classic Wilson-Cowan neural population model [Wilson and Cowan, 1972.sup.3], which is described in detail in previous publications [Wang et al., 2012, Wang et al., 2014.sup.4]. The two-variable version of it is used, which models the neural tissue as a single excitatory population and a single inhibitory population. This model is able to capture epileptiform spikes and epileptiform discharges [Wang et al., 2012.sup.5], which are the two key activity types from the experimental data. .sup.3Wilson, H. and Cowan, J. (1972). Excitatory and inhibitory interactions in localized populations of model neurons. Biophysical Journal, 12(1):1-24.sup.4Wang, Y., Goodfellow, M., Taylor, P. N., and Baier, G. (2014). Dynamic Mechanisms of Neocortical Focal Seizure Onset. PLoS Comput Biol, 10(8):e1003787.sup.5Wang, Y., Goodfellow, M., Taylor, P., and Baier, G. (2012). Phase space approach for modeling of epileptic dynamics. Phys Rev E, 85(6):061918

(65) Briefly, the differential equation system used is:

(66) $\begin{array}{cc}\frac{\mathrm{dE}}{\mathrm{dt}}=\left(-E+\mathrm{Sigm}\left(a*E-b*I+P\right)\right)/{\tau }_e\frac{\mathrm{dI}}{\mathrm{dt}}=\left(-I+\mathrm{Sigm}\left(c*E-d*I+Q\right)\right)/{\tau }_i& \left(1\right)\end{array}$
where E (I) is the activity of the excitatory (inhibitory) neural population. The parameters a; b; c; d determine how strongly each population influences themselves and the other population. The parameters τ.sub.e and τ.sub.i are time constants, dictating how quickly a population reacts to incoming input.

(67) Finally the sigmoid function is:

(68) $\mathrm{Sigm}\left(x\right)=\frac{1}{1+\exp \left(-\left(x-4\right)\right)}.$

(69) This system will be simulated with noise of amplitude na (reflecting synaptic noise and non-specific input from the surrounding tissue). In other words, the Euler-Maruyama solver can be used to simulate the system as a stochastic differential equation system.

(70) Depending on the parameter settings, the above described system is able to simulate epileptiform interictal spikes, similar to those observed in the in vitro experiments. Essentially, two parameter settings were used, one for the interictal spikes, and one for the ictal discharges. Table 1 lists these in detail for each configuration (interictal vs. ictal) in this example.

(71) TABLE-US-00001 TABLE 1 Interictal Ictal a 17 17 b 15 10 c 40 40 d 0 0 τ.sub.e 0.06 0.0264 τ.sub.i 0.06 0.012 P 0 −0.3 Q −7 −15

(72) FIG. 16 (a) shows an example of an epileptiform interictal spike from the in vitro experiments. FIG. 16 (b) in comparison, shows a simulated interictal spike from the model. It is important to note here that the spiking activity arises spontaneously in the model as a result of the noise input. FIG. 16 (c) shows a longer simulation with many spontaneous spikes. In dynamical systems terms, the system is in a stable node, with a homoclinic bifurcation (of the saddle and a limit cycle) nearby in parameter space (bifurcation parameter P). Hence the spike can be found as an excitable transient in phase space, and the added noise can occasionally perturb the system onto this transient. This mechanism has been used before in several studies to model interictal spikes, e.g. [Wendling et al., 2002.sup.6, Wang et al., 2012].sup.6Wendling, F., Bartolomei, F., Bellanger, J., and Chau-vel, P. (2002). Epileptic fast activity can be explained by a model of impaired GABAergic dendritic inhibition. European Journal of Neuroscience, 15(9):1499-1508

(73) The model outputs (E(t) and I(t)) is suggested to reflect the multi unit activity (MUA) in the in vitro recordings, whereas the local field potential (LFP) is perhaps best modelled as hf.sub.0.1(E+I). The hf( ) function is a high pass filter, in this case with a cut-off frequency of 0.1 Hz. This is to simulate the high-pass filtering properties of the in vitro recording equipment. Such a filter introduced slow waves after the spike, which resembled those observed in the in vitro recording (FIG. 16). In the strict sense, the modelling of the LFP should be an explicit part of the differential equations, as for example in the Jansen-Rit model [Jansen and Rit, 1995.sup.7]. In the Wilson-Cowan approach, this has been neglected, hence a simple E+I is assumed to represent the LFP. .sup.7Jansen, B. and Rit, V. (1995). Electroencephalogram and visual evoked potential generation in a mathematical model of coupled cortical columns. Biological Cybernetics, 73(4):357-366

(74) Analogous to the modelling of spikes, the system described by equation 1 is capable of producing transient epileptiform ictal discharges, similar to those observed in the in vitro experiments. FIG. 17 (a) shows an example of an epileptiform ictal discharge from the in vitro experiments. FIG. 17 (b) in comparison, shows a simulated ictal discharge from the computational model. It is important to note here that the entire ictal discharge arises spontaneously in the model as a result of the noise input. FIG. 17 (c) shows a longer simulation with many spontaneous spikes. In dynamical systems terms, the system is in a stable node, with a coexisting bistable limit cycle (separated by the separatrix of a saddle in phase space). The noise input can then drive this system to and from the limit cycle, leading to the appearance of transient ictal discharges in the simulated time series. This bistability mechanism has been a popular one to model seizures [Lopes Da Silva et al., 2003.sup.8, Breakspear et al., 2006.sup.9, Wang et al., 2014]. .sup.8Lopes Da Silva, F., Blanes, W., Kalitzin, S., Parra, J., Su czynski, P., and Velis, D. (2003). Epilepsies as dynamical diseases of brain systems: basic models of the transition between normal and epileptic activity. Epilepsia, 44:72-83.sup.9Breakspear, M., Roberts, J., Terry, J., Rodrigues, S., Mahant, N., and Robinson, P. (2006). A unifying explanation of primary gen-eralized seizures through nonlinear brain modeling and bifurcation analysis. Cereb Cortex, 16(9):1296-1313

(75) Modelling Closed-Loop Stimulation

(76) To model optogenetic input to the system, a simple approach was chosen. The light input is simply added as an additional input term modified by a weight I:

(77) $\begin{array}{cc}\frac{\mathrm{dE}}{\mathrm{dt}}=\left(-E+\mathrm{Sigm}\left(a*E-b*I+P+l*\mathrm{LED}\left(t\right)\right)\right)/{\tau }_e\frac{\mathrm{dI}}{\mathrm{dt}}=\left(-I+\mathrm{Sigm}\left(c*E-d*I+Q\right)\right)/{\tau }_i& \left(2\right)\end{array}$

(78) The detailed dynamics of the Channelrhodopsin [Grossman et al., 2011.sup.10] and the subsequent conversion of this signal to a corresponding postsynaptic potential are simplified here. It is assumed that the LED input is proportional to its equivalent population input. .sup.10Grossman, N., Nikolic, K., Toumazou, C., and Dege-naar, P. (2011). Modeling Study of the Light Stimulation of a Neuron Cell With Channelrhodopsin-2 Mutants. IEEE Transactions on Biomedical Engineering, 58(6):1742-1751

(79) In the experimental closed-loop system, the output of LED(t) is determined by the recorded ongoing activity of the slice. As described earlier, this is currently essentially a phase shifted, rectified version of the filtered recording. In the model, the same algorithm was followed as in the in vitro experiments. In the simulations, the term LED(t) is evaluated at each time point of the Euler-Maruyama algorithm that solves the SDE equations. As the recorded signal (i.e. the simulated LFP), a linear combination of E(t) and I(t) was used, as the real LFP will be a linear combination of population EPSPs and IPSPs, depending on the recording location.

(80) Results

(81) Interictal Spike Activity

(82) The closed-loop stimulation, in vitro, is able to stabilise a single interictal spike into a continuous low frequency oscillations, depending on the phase setting of the stimulation. FIG. 18 (taken from an example test) shows an example of this. In several cycles of running through 8 phase conditions (0, 45, 90, 135, 180, 225, 270, and 315 degrees), the behaviour is seen to be qualitatively very robust. In phases, where a stable oscillation is found, the period of the oscillation is also determined by the phase (seen in the spectrogram in FIG. 18 (b)). In some phases, no stable oscillations can be found.

(83) It was possible to capture most of the qualitative observations in the in silico model using the same closed-loop stimulation. FIG. 19 shows very similar observations in the simulations as in the experimental results of FIG. 18. First, depending on the phase, an oscillation is stabilised, with a few phases not being able to sustain such oscillations (FIG. 19 (a)). This behaviour is also stable across several cycles of running through the 8 phase conditions. In phases, where stable oscillations are found, the frequency is also dependent on the phase (FIG. 19 (b)). Note that the phases in which stable oscillations are found in the model depends essentially on how the simulated LFP signal is defined (in what linear combination of E(t) and I(t) it is calculated as).

(84) The interpretation for these results is fairly simple from a dynamical systems perspective. The fact that such a simple model can capture most experimental observations indicates that the interictal spikes can be understood as an excitable system (near a simple homo-clinic bifurcation). When excited, the system completes one cycle of oscillation.

(85) When using closed-loop stimulation, this oscillation can be stabilised. In dynamical systems terms, we can understand the stimulation as a third coupled variable, and the different phase condition can be understood as the delay of this third variable. In such a framework, it is easy to see that some phases will lead to the stabilisation of the oscillation (i.e. pushing the system beyond the bifurcation point), whereas other phases stabilise the fixed point.

(86) Ictal Burst Activity

(87) The in vitro model displays ictal burst activity, which consists of consecutive periods of beta-range high amplitude oscillatory activity, separated by background activity. During closed-loop stimulation of ictal burst activity in vitro, the essential finding was that the duration of the periods of oscillatory activity is modulated by the phase condition (FIG. 20 (c)). Some phase conditions (90 and 135 degrees) appear to shorten the duration, whereas other phase conditions appear to prolong the burst duration. However, the overall discharge duration was not changed through the stimulation. The shortening of such periods of oscillation is remarkable, given that we used EMX-ChR, i.e. probably activating principal neurons.

(88) The in silico model at the moment is only capturing the periods of beta-range oscillations as separate events. I.e. the periodic bursting is not captured. However, it is possible to test in the simple model the effect of closed-loop stimulation. In the model the phase condition also shows an effect on the duration of the oscillatory events. In certain phase conditions, the event is prolonged by the stimulation. A slight shortening of the oscillations can also be observed at phase 0 (FIG. 20 (d)).

(89) To quantify this, the mean ictal burst duration is shown for each phase condition in FIG. 21. Significant changes (prolongation and shortening of bursts) are observed compared to the control condition, although the effect in the model is less prominent than the in vitro experiments. This may be due to the parameter setting of the model, which has not yet been optimized to reproduce the experiment quantitatively. Note again that the exact phase conditions for prolonging or shortening the burst duration in the model depend on the choice of how the LFP is modelled (what linear combinations of E(t) and I(t)), and the feedback strength. In this case, these parameters have not changed from the previous interictal model.

(90) A possible interpretation of these results is that a simple bistable model of seizure activity is able to capture the prolongation, and shortening of oscillatory activity. In a similar analogy to the interictal case, the prolongation of the oscillation is fairly intuitive.

(91) While the invention has been described in conjunction with the exemplary embodiments described above, many equivalent modifications and variations will be apparent to those skilled in the art when given this disclosure. Accordingly, the exemplary embodiments of the invention set forth above are considered to be illustrative and not limiting. Various changes to the described embodiments may be made without departing from the spirit and scope of the invention.

(92) All references referred to above are hereby incorporated by reference.