1. Bill Kristan Section of Neurobiology Division of Biological Sciences UC San Diego La Jolla, CA The dynamics of decision-making by leech neurons Neurobiology of Decision-Making CSH 24 May 2005

2. We make all kinds of choices: Choosing whether Choosing when Choosing how Choosing which (respond or ignore) Choosing what(feed on it, fight it, love it) (direct/indirect, strongly/subtly) (now, later) (regular/diet, large/medium, left/right)

3. We make all kinds of choices: Choosing whether Choosing when Choosing how Choosing which (respond or ignore) Choosing what(feed on it, fight it, love it) (direct/indirect, strongly/subtly) (now, later) (regular/diet, large/medium, left/right)

4. Leeches crawl in shallow water (3 mm)

5. Leeches swim in deep water (33 mm)

6. Anatomy of the leech CNS

7. Leech preparations

8. Leech locomotory behaviors Swimming 1 sec Dorsal Ventral 10 sec CCC CCC EEE G10 G13 Elongation (E) Contraction (C) Crawling

9. Cell 204 DE motor neuron 5 sec 10 mV Weeks & Kristan, 1978 Swim initiation by a command neuron, cell 204

10. Invertebrate model of choice: inhibition of command neurons Command neurons Command neurons

11. Problem #1: some “command neurons” are activated in incompatible behaviors. Shaw & Kristan, 1997

12. Semi-intact preparation Intracellular Extracellular Stimulate cell R3b1: Esch,Mesce & Kristan, 2002 Problem #2: Some “command neurons” are bifunctional Shallow water Deep water

13. using a combinatorial code by multiplexed neurons Decisions appear to be made interactively

14. Leech nervous system can swim or crawl to the same stimulus Use isolated nerve cord Head brain Tail brain G15 same stimulation elicits different responses: S Stimulate (S) one nerve electrically R, record (R) from another Stimulus Swimming Crawling Imaging Briggman, Abarbanel & Kristan, 2005

15. Optical activity in motor neurons during swimming Data of Adam Taylor

16. Membrane potential trajectories of 144 neurons dF/F (%) Swim Time (msec) Cell number Crawl Time (msec) Cell number Briggman, Abarbanel & Kristan, 2005

17. Discrimination by single neurons Non-discriminating (ND) Early Discriminating (ED) Late Discriminating (LD)Transiently Discriminating (TD) Swim Trials Crawl Trials p > 0.000001 p < 0.000001 Nerve DT Single cell DT Single cells with early DTs Briggman, Abarbanel & Kristan, 2005

18. PC1 PC2 PC3 Cell Number Principal component analysis, across neurons

19. Discrimination of single cells vs. neuronal populations { Cells contributing to Linear DiscriminantSingle cells with early DTs LDA DTs Earliest cell DTs Nerve DTs Single cell DTs {

20. Polarizing cell 208 biases behavioral choice Hyperpolarized Trials (-1.5 nA) Depolarized Trials (+1.5 nA) Intracellular Stimulation DP Nerve Stimulus Briggman, Abarbanel & Kristan, 2005

21. Leeches make behavioral choices sequentially R3b1 swimming “Locomote” crawling 204 swimming “Swim!” 28 swimming “Bend up (down)” Cell Active during Command 208 swimming “Do something!” shortening

22. x y z Swim CPG Crawl CPG Rest state Stimulation Decision making and dynamically:

23. Many leech neurons take part in both swimming and crawling. Only a few differ early -- using a combinatorial code. -- among multiplexed neurons Decision-making may depend on dynamic interactions (good candidates for being decision-makers). About half of them have different activity in the two behaviors. CONCLUSIONS

24. COLLABORATORS: DavidKleinfeld RogerTsien Tito Gonzalez Peter Brodfuehrer HenryAbarbanel KRISTAN LAB: Brian Shaw TimCacciatore Adam Taylor TeresaEsch KarenMesce Kevin Briggman Cast of characters (PanVera)(AuroraBiosciences) Dyesprovided by Vertex Pharmaceuticals Gary Cottrell

25. Linear Discriminant Analysis (LDA) Cell A Cell B Cell C LD Direction a b c

26. PC1 PC2 Principal component analysis, by participating neurons: PC1 PC2 A B PC1 PC2 PC3 Cell Number Briggman, Abarbanel & Kristan, 2005

27. What’s ahead? x p(success of A) p(choice A) = response threshold p(choice A) x (benefit of A - risks) = p(any behavior) x sensory processing x (positive - negative) modulation Input to decision makers: source of modulation site of action effects of feeding classical conditioning: bias to swim or crawl

28. KRISTAN LAB: Brian Shaw Tim Cacciatore Adam Taylor TeresaEsch KarenMesce KevinBriggman Cast of characters by Dyes provided (PanVera) (AuroraBiosciences ) Vertex Pharmaceuticals COLLABORATORS: DavidKleinfeld Roger Tsien Tito Gonzalez Peter Brodfuehrer Henry Abarbanel Gary Cottrell

29. Discrimination by single neurons Non-discriminating (ND) Early Discriminating (ED) Late Discriminating (LD)Transiently Discriminating (TD) Swim Trials Crawl Trials p > 0.000001 p < 0.000001 Nerve DT Single cell DT (Data of Kevin Briggman)

30. Discrimination by single neurons Single cells with early DTs

31. Swimming… in 144 neurons Data of Kevin Briggman dF/F (%) Cell number Time (msec)

32. Cell Stimulation type: produces: Active during: Command: A swimming swimming “Do something!” shortening Is leech decision-making hierarchical? B swimming or swimming “Get out of here!” crawlingcrawling C swimming swimming“Swim!” D bending swimming “Move this muscle group”

33. Cell 204 is inhibited during shortening…. Data of Brian Shaw

34. Record optically from 91 neurons simultaneously Swim Shorten (Data of Kevin Briggman)

35. ShortenSwim Swim – Shorten Subtract no-swim from swim traces for each neuron (Data of Kevin Briggman)

36. Reasonable hypothesis: Decision-making neurons have different activity trajectories in different behaviors.

37. Swimming Time Component amplitude PCA1 PCA2 PCA3 Shapes of the first 3 components: Use principal component analysis (PCA) to detect neurons with different activity trajectories in different behaviors. Crawling (Data of Kevin Briggman)

38. Response Variability Tail brain G12 G15 Head brain (Data of Kevin Briggman)

39. Swim Crawl (Data of Kevin Briggman) A few neurons have different trajectories in swimming and crawling:

40. Another reasonable hypothesis: Decision-making neurons have the earliest differences in their trajectories.

41. Use PCA analysis to follow the trajectories of neuronal classes over time. C1 C2 C3 With just 3 neurons, we could plot their activity on separate axes: 1 st Frame Stimulus delivered 1 st Swim Burst PC 1 PC 2 PC 3 For assemblies of neurons, we plot their PCA components:

42. Do we make choices hierarchically? (sequential, linear) …..or interactively? (feedback, resetting) …..or simultaneously? (nonlinearity, overlapping function) “Choice-makers” will be active in sequence “Choice-makers” activity will bounce back and forth “Choice-makers” will all be active at the same time

43. Neuronal circuits for whole-body shortening, swimming Mechanosensory neurons Trigger interneurons Gating interneurons Oscillator interneurons Motor neurons

44. Shortening and swimming circuits overlap

46. Measuring voltage changes with FRET dyes (FRET = fluorescence resonance energy transfer) (Developed by Tito Gonzalez & Roger Tsien)

47. FRET VSD optical signals (Figure provided by Tim Cacciatore)

48. Tr2 terminates swimming

49. Swimming…. in 90 neurons Data of Kevin Briggman

51. FRET-based dyes can detect synaptic potentials

52. Tr2 spikes elicit 1-1 EPSPs

53. Multiple responses to identical stimulation Head brain Tail brain G15 Use isolated nerve cord same stimulation elicits different responses: Stimulate (S) one nerve electrically S, record (R) from another R Swimming Behavioral state ? ? Behavioral choice ? ? Data of Kevin Briggman Crawling

54. Do we make choices hierarchically? ”Let’s go out to dinner” “Some place new” “Italian” “Mario’s!” (implies sequential, linear) …..or interactively? ”I’m not that hungry; let’s do that new Bistro” ”I don’t want to dress up; how about our favorite sushi bar?” “The Lakers are playing tonight; maybe we should order in.” (implies feedback, resetting) …..or simultaneously? {impossible to express verbally} (suggests nonlinearity, overlapping function)