1. Brain Rhythms in Health and Disease
Nancy Kopell
Boston University

2. What Are Brain Rhythms, and How Are They Measured?
Electrical activity can be measured in vitro and in vivo:
Single cells
Local groups of cells (LFP)
Non invasively (EEG, MEG)
Peaks in spectral power
delta (1-4 Hz)
theta (4-12 Hz)
alpha (9-11) Hz
beta (12-30 Hz)
gamma (30-90)
Bragin, et al.
Whittington et al.
Jensen et al

3. Why Care About Rhythms?
Strong associations between behavior and rhythms
Repeatable and subtle changes in dynamics correlated to task
Changes happen on time scale of behavior
Brain rhythms associated with active sensing, attention, decisions, motor activity – all forms of cognition
All (?) forms of neurological disease are associated with changes in rhythms
Hypothesis: the changes is rhythms – and associated changes in coordination – are final pathway for disease related changes in cognition.
Hypothesis: physiology underlying rhythms is central to their effects.

4. Working Memory: Different connections for Different rhythms
Alpha (9-11 Hz)
Beta (12-30 Hz)
Gamma (30-90 Hz)
Palva JM et al. PNAS
©2010 by National Academy of Sciences

5. General Mathematical Framework:
Hodgkin-Huxley Equations
C dv/dt = -  I ion + D  2 v -  I synapse
I ion = (g mj h) (v – ER)
Conductance x Electromotive force
m and h satisfy
dx/dt = (x(v) – x)/ x
Activation curves
Synapses: excitatory or inhibitory

6. Pyramidal Interneuron Network Gamma (PING)
Whittington et al., J. Physiol. 1997
Induced by tetanic stimulation (hippocampus in vitro)
Associated with cell assemblies
PING is coherent with heterogeneity, sparse coupling (Borgers, NK)
E-cells are synchronized by I cells, I cells are synch’d by E-cells
Period depends on decay time of inhibition = longest time scale

7. Persistent Gamma Rhythm
Traub, Whittington, Borgers, Epstein, NK
Can be induced in slices with acetylcholine (ACh) agonist and/or kainate
ACh is associated with attention
Lasts a long time
E cells each fire infrequently
Requires noise
Population displays gamma rhythm
More complex biophysical model involves axo-axonal gap junctions; Uses FRB cells (Traub, Whittington…).
Persistent gamma is associated with background attentional state.

8. Gamma and Cell Assemblies
Cell assemblies: Synchronous collections of neurons.
The gamma rhythm (30-90 Hz) is associated with
Sensory processing, attention, awareness
“Binding” (life without a homunculus: Singer, Gray …)
(Possible) Uses:
Potentiating signals
Plasticity (Hebb: “Cells that fire together wire together”)

9. What Makes Gamma Good for Cell Assemblies?
Olufsen, Whittington, Camperi and NK 2003
Suppression is consequence of timing of excitation and inhibition
Gamma formed from simple currents; no memory from cycle to cycle
Other rhythms use currents that last longer than gamma cycle, create memory.
Physiology matters

10. Gamma Filters for Coherent Input
Borgers and Kopell, 2008
Reynolds, Desimone…: Attention comes from biased competition
Attention is associated with more coherent input (Fries et al. Bichot et al.)
More coherent inputs lock out the less coherent ones. E I
Target network
Coherence “protects” chosen stream of information.

11. Bias: Effects of Inhibition, Coherence and Salience
Bias comes from effects of internal inhibition in E/I network.
Larger distractor can affect outcome
Distractor can be shut out by increasing I -E conductance
Protection against distractors requires gamma frequency input.

12. Two Gamma Rhythms in Rodent Au1
Ainsworth, S. Lee, Cunningham, Roopun, Traub, NK, Whittington, 2011
Rhythms induced by bath-applied kainate
Layer 4 rhythm:
Is PING rhythm with E-E connections depending on NMDA
Can be eliminated with NR2C/D blocker
Layer 2/3 rhythm:
Is noise-dependent gamma
Can be eliminated with gap junction blockers

13. Layer 5 Output Switches With Higher Kainate
Low gamma only
Low and high gamma
Layer 2/3
Layer 4
Layer 5
rmp
ipsp A B C
Layer 5 Output Switches With Higher Kainate
E-cells
Ainsworth et al.
Model (S. Lee)
L2/3 L4 L5
100
200
300
400
500 ms
Spikes
LFP
L2/3 L4 L5
100
200
300
400
500 ms
Spikes
LFP
400 nM
800 nM Es Is
L2/3 (s) Eg
L4 (g) Ig Ei
L5 (i) Ii
400 nM
800 nM
No feedback from layer 5
Output follows : L2/3 at low kainate
L4 at high kainate

14. In Vivo Implications (?) of Laminar Dynamics
Strong input to Au1 may go directly from L4 to output.
L4 activity actively impedes influence of signals from L2/3 to L5.
Weak input is gated by superficial rhythm.
Medium input is enhanced by locking of L4 with L2/3 Es Is
L2/3 (s) Eg
L4 (g) Ig Ei
L5 (i) Ii
Later: effects of top-down gamma

15. Cholinergic Modulation Changes Rhythms in Au1
Roopun, Traub, Whittington, Kramer, Barakat, J. Lee, Moore-Kochlac, NK
Slice version of attention:
Produces a new beta2 rhythm (25 Hz) in L5
Produced in Au1 using carbachol
Cell types most implicated: L5 IB cells and LTS cells
IB cells fire a few spikes per “burst”, cut off by LTS inhibition
Frequency determined by:
drive to LTS and IB cells,
decay time of LTS-mediated inhibition.
J. Lee

16. Beta Rhythms and Selective Attention
Top-down attention is associated with beta ( Buschman and Miller 2007)
Fries et al. (2001, 2008.)
Model (J. Lee, M. Whittington, NK 2013)
Top-down signals (at beta) go to deep layers of “attended” column only
Input from thalamus goes to L4, both columns
Excitation to inhibitory cells (deep to superficial) goes within and across cols.

17. Top-Down Beta 2 to Deep Layer Increases Firing Rates
Top-Down Beta 2 to Deep Layer Increases Firing Rates and Gamma Power in Superficial Layers
Control Attended Unattended
L2/3
Within a column
More RS-cell spiking in L4
More gamma
Across columns
Weaker RS activity in L 2/3
Stronger LTS activity in L2/3, more beta L4 L5

18. Rhythms and Cholinergic Modulation
Question: how can cholinergic modulation be specific?
Nicotinic activation depolarizes deep layer LTS cells; muscarinic
activation hyperpolarizes deep FS cells. Both increase LTS activity.
Increased activity of LTS cells potentiates effects of top-down signals, increasing attentional modulation of gamma and firing power in superficial layers.
ACh makes networks more sensitive to top-down signals.
Selective attention and diseases (ADHD, SZ)
If there is pathology in cholinergic modulation, there is
Reduction in LTS activity
Deficits in selective attention.

19. Anesthesia: Dose-dependent Responses to Propofol
Propofol increases the amplitude and decay time of GABAa mediated inhibition
Brown, Purdon

20. Propofol and “Beta Buzz”
Low doses of propofol produce potentiation of beta rhythm (12-30 hz) .
Rampil IJ (Anesthesiology 1998)
Not known: why does enhanced beta cause a “buzz”?
Project: why does propofol enhance beta?

21. Potentiated Inhibition and M-current Interact to Produce a Beta Rhythm
Propofol enables beta rhythm via antisynchrony of LTS cells
McCarthy et al 2008
Model:
e-cells with spiking currents, A-current
LTS cells (i) with spiking currents and M-current
Baseline: interaction of e and LTS cells.
Potentiated inhibition: LTS cells fire by rebound, e cells suppressed

22. M-Current Interaction With Inhibition
GABAa potentiation decreases M-conductance
M-current reduction continues after GABAa has decayed back to baseline
LTS model interneurons
are quiescent if unperturbed
may spike in response to either pyramidal cell or interneuron input
*Low* dose propofol allows rebound spike during repolarization. .
Time to rebound spike .

23. Large Network Dynamics
Propofol enables the formation of antisynchronous clusters of beta frequency spiking interneurons
Spectral power rises in the beta bands with low-dose propofol
Power decreases in the lower frequency bands with low-dose prop.
LTS FS e

24. Loss of Consciousness With Higher Dose of Propofol
Propofol administered in 5 discrete levels
Subjects are asked to respond to tones, words and names
Ching et al. 2010
Cimenser et al. 2011
Feshchenko et al. 2004
Consolidation of power into alpha band as responses begin to cease
Well defined alpha power at deepest levels of anesthesia
Distinct in location from ‘occipital’ alpha oscillation
Simultaneous strong slow oscillations (< 1Hz)
Global coherence is mainly in alpha band.

25. Alpha and Thalamocortical Loop: Network
Ching et al, PNAS 2010
Cortical populations: Pyramidal cells, FS/LTS interneurons
Thalamic population: Thalamocortical relay cells, Reticular cells
RE, TC cells: h-current, T-type calcium (rebound bursting)
Propofol increases the decay time of inhibition throughout the full network.
Questions:
What produces the alpha rhythm?
What produces the coherence (frontal)?
Prior work:
Alpha spindles: Destexhe & Sejnowski, “Thalamocortical Assemblies,” 2001

26. Investigate With Smaller Models
Increasing GABA slows rhythmicity, sustains thalamic participation.
Thalamic locking is specific to alpha frequency input from cortex.
Locking requires hyperpolarization of TC cells
RE cells can coordinate cortical modules

27. Simulation Results
Model produces weak gamma baseline- dictated by timescale of inhibition
Slowing of cortical rhythmicity as GABAa increases (conductance, decay time)
Converges to alpha rhythm
No coherence at baseline and small dose levels
At 3x baseline, coherence is exhibited between the cortical populations c c

28. Implications/Remarks
Many theories of mechanisms in general anesthesia point to a thalamocortical ‘off-switch’ (Alkire et. al., 2008)
Our modeling suggests that the thalamus may be engaged in a pathological rhythm
Pathological synchrony may impede the efficacy of thalamic excitation
Thus: Loss of consciousness may come from too much coupling, not from decoupling.

29. Loss of Consciousness and Changes in Rhythms
At loss of consciousness, there is large increase in alpha and delta power.
relationship between alpha and delta signals depth of anesthesia
Purdon et al ; Soplata et al

30. Beta in the Cortico-Basal Ganglia-Network
Beta pathology in Parkinson’s disease:
Increased beta oscillations in the cortex and basal ganglia correlate with motor symptoms of bradykinesia and rigidity.
Predominating views on the source of beta in Parkinson’s disease:
1. STN-GPe network (STN-GPe pacemaker)
-- (Plenz et al., 1999) .(Terman et al., 2002)
-- Model beta rhythm depends on plastic changes in synaptic strength due to chronic loss of dopamine.
2. Cortex
-- In rat M1 brain slices, high beta arises from the co-application of carbachol and kainate. (Yamawaki et al., 2008)
-- Not clear why beta should be related to loss of dopamine.

31. A New Hypothesis: PD Beta Originates (partly) in the Striatum
McCarthy et al PNAS, 2011
Note: striatum is (one of) first places to feel loss of dopamine from SNpc.
Relationship between Dopamine and Acetylcholine in Striatum
Dopamine tonically inhibits ACh release in the striatum.
 dopamine   ACh
The MSN M-current is reduced by ACh (M1 muscarinic receptors).
Connection to propofol beta: depression of M-current (excitatory) acts like changing the time constant of inhibition to give rebound

32. Model of Parkinsonian Striatum
Model of parkinsonian network:
Only cells are “medium spiny neurons”
 M-current maximal conductance
Biophysical model of MSN network (100 neurons). Top are raster plots of spiking in normal and parkinsonian networks. The raster plot shows obvious patterning of MSNs into a network beta rhythm in the parkinsonian state. Bottom figures are the respective spectrograms showing a waxing and waning beta in the LFP of the normal model striatum and a persistent (and higher power) beta in the model parkinsonian striatum. The beta peak in the power density plot becomes more prominent and at a higher beta frequency in the parkinsonian state.
The striatum has many neuromodulators. Here we focus on dopamine and ach. On dopamine because that is what is lost in PD.
On ACh because there has been a lot of work showing interactions between dopamine and ach and it’s the balance of ach and dopamine that is critical in regulating MSN activity. In fact, up until the1960’s the only pharmacologic therapy for Parkinson’s disease was anticholinergics.
Model predictions:
Increasing cholinergic (muscarinic) tone in striatum will elicit beta oscillations.
Model predicts that patterned spiking of MSNs into a beta rhythm (on a population level) is responsible for this increase in beta power. This is important if the beta rhythm is to be propagated outside of striatum since MSNs are the only output neuron of the striatum.
3. It is a prediction of our model that it is not just ACh that matters, it is muscarinic receptor activation.
Model needs sufficient excitatory input from cortex to produce beta (not necessarily beta input). Lack of sufficient input from cortex will decrease beta in striatum. Details about this are in the paper.
Note (in case someone asks): The model results are largely invariant to network connectivity, network size, and heterogeneity (by giving different MSNs different max GABAat conductance). 32

33. Experimental Confirmation:
Carbachol increases beta oscillations in mouse striatum
Striatal LFP prior to carbachol infusion
McCarthy et al, PNAS, 2011
Striatal LFP after carbachol ( mM) infusion

34. Behavioral Correlates of Increased Cholinergic Tone.
Stimulation of striatal cholinergic interneurons
decreases movement in normal mice
Kondabolu 2016
bradykinesia
akinesia

35. Deep Brain Stimulation (DBS) to the Subthalamic Nucleus
Deep Brain Stimulation (DBS) to the Subthalamic Nucleus (STN) and Parkinsonian Motor Symptoms
DBS mechanism of action is unknown.
What we know:
Fast therapeutic effect
Frequency dependent
Decreased beta (measured after stop of DBS) correlates with improvement of motor symptoms
=> some of the efficacy of high frequency DBS may lie in its ability to reduce elevated beta oscillations.
DBS

36. Modeling the effect of DBS through the indirect pathway
Model of DBS Normalizes Striatum
normal
Modeling the effect of DBS through the indirect pathway
parkinsonian
parkinsonian,
With 130 Hz DBS
Normalization is result of inhibition from GPE.

37. Cellular, Molecular and Circuit Level Changes
Cellular, Molecular and Circuit Level Changes Alter Rhythms in Schizophrenia
Kinds of pathologies in SZ:
genetic changes (risk genes)
GABA alterations
NMDA hypofunction
dopaminergic dysregulation

38. Brain Rhythms and Normal Function
Rhythms aid in local processing
Gamma helps create and protect cell assemblies
Two frequencies of gamma in A1 helps produce routing of signals.
Global coordination builds on local dynamics
Top down attention (beta) leads to increased gamma and activity in superficial layers
Dysregulation of nicotine changes ability of target network to react to attentional signals.

39. **Big Picture** Working hypotheses:
Cognition (and its dysfunction) emerge from dynamic coordination (and its dysfunction)
Alteration in rhythms are common pathways for etiology of disease
Big question:
Under what conditions can alterations in oscillations restore diseased state to normal? .