Formation of connectome: Formation of connectome: nature versus - PowerPoint PPT Presentation

Formation of connectome: Formation of connectome: nature versus nurture nature versus nurture Alexei Koulakov, Cold Spring Harbor Lab Alexei Koulakov, Cold Spring Harbor Lab 1

Cold Spring Harbor Laboratory, New York 2

How the brain works Dendrites Synapse Action potential (spike) Axon Cell body (soma) ~10 10 neurons ~10 14 synapses 3

How much information can be stored in connections? H   log 10 neurons in cortex N ~10 4 s ~10 synapses per neuron  Ns ~ N  H Ns log N ~ 400 terabytes ~ 45 years of HD video Wei, Tsigankov, Koulakov, Annals of NYAS (2013)

Genomic bottleneck  9 3 10 bp ~ 1 GB Human genome contains of information How can 1GB of information set up 400 TB of connections? Obviously, each synapse cannot be specified in the genome individually Some simplifying rules are necessary Genome development 1GB rules Cortical networks Sperry, PNAS (1963) 400TB Wei, Tsigankov, evolution Koulakov, Annals of NYAS (2013) Zador, Nature Commun. (2019) 5

Genomic bottleneck Genome development 1GB rules Cortical networks evolution 400TB   y x Neurodevelopmental rules contained in the genome (1GB) carry information about the capacity of humans for intelligent behavior 6

Q1: What are the rules for genes to specify connections? Q2: How can genetic information be combined with experience- dependent network plasticity? 7

Brain regions 8

Cortical maps: topography Kalatsky and Stryker, 2003 Many visual areas represent world topographycally 9

Maps = connections Visual cortex Thalamus T ha la mus Superior colliculus Supe rio r Co llic ulus axons Retina 10

Topographic map Eye Superior colliculus axons How does a neuron know its position and know where to go? 11

Chemoaffinity hypothesis Roger Sperry Nobel Prize in Medicine 1981 I t seems a necessary conclusion … t hat t he cells and f ibers of t he brain and cord must carry some kind of individual ident if icat ion t ags, by which t hey are dist inguished one f rom anot her … 12 —Roger Sperry, 1963

Molecules define neuron’s position Brown et al Cell (2000) 13

Horizontal direction is imprinted in the eye and in the target by molecules Superior colliculus Retina ephrins-A EphA axons EphA/ephrin-A chemorepulsion 14

Electrostatic model EphA level = � � � � ephrin-A level =     A A H q ( r ) i i i Total number of bound EphA receptors H is minimized - repulsion 15

Vertical axis Retina Superior colliculus EphB ephrins-B Y X EphB/ephrin-B chemoattraction 16

Attraction in Y direction: another type of charge � � EphB level = � � ephrin-B level =         A A B B H q ( ) r q ( ) r i i i i i i The same axon carries two charges of different colors that interact with two potentials. 17

Monte Carlo simulation ephrin-A e   E T / probability EphA or � � ephrin-B EphB or � � 18

More general set of rules:         A A B B H q ( ) r q ( ) r i i i i i i    1 0           H M q r M    i i   0 1 i   , A B ,  W     M W q    ˆ H W ( )  ij jr ij i j i i j ,   , M  Matrix was optimized by evolution to yield the capacity for general intelligence 19

A1: Genetic information is converted into connectivity patterns by virtue of molecular tags (Sperry) Q2: How can genetic information be combined with experience dependent network plasticity? 20

Activity of neurons during topography formation: retinal waves Data courtesy David Feldheim (UCSC) 21

What happens if there are no retinal waves Disrupted retinal waves Normal retinal waves ( 2 -/- mice) Superior colliculus Superior colliculus Retina Retina McLaughlin et al, (2003) Conclusion: axons with correlated activity are attracted to each other in the target 22

Attraction between connections in the target Correlated activity: Strong attraction i j Uncorrelated activity: Weak attraction   r r i j          H C U r r activity ij i j Correlation in 2 ij activity due to waves in retina 23 Distance in retina

Brain versus physics G 1     H m m   grav i j  2 r r  i j i j   k 1  H q q   elect r i j  2 r r  i j i j        H C U r ( r ) activity ij i j 2 i j       2      2 U r ( r ) exp r r / 2 i j i j charges do not separate, cannot introduce potential 24

Correlated activity leads to sharper projections SC, γ = 0 γ = ¼ γ = 1 Retina … through attraction between axons neighboring in retina 25

Unified Hamiltonian: molecules experience defined by genes learning nature nurture               H M q r C U r ( r )  i i ij i j 2 i ij   , A B , Sperry Hebb         H M q W C W W U  i j ij ij ik im km 2 ij ijkm 26 

Test of the model: maps is ephrin knockout mice 27

Maps in ephrin knockouts Normal mouse Superior colliculus Tracer injection in the eye Mutant ephrinA -/- mouse r c Feldheim, Kim, Bergemann, Frise, Barbacid, and Flanagan, 2000 28

We have plenty of intuition about systems with attraction - gravity Collapse induced by gravity Collapse induced by attraction between axons 29

Maps in ephrin knockouts � � � � Superior colliculus � � � � retina SC SC r c retina Feldheim, Kim, Bergemann, Frise, Barbacid, Tsigankov and Koulakov (2006) and Flanagan, 2000 30

Topographic maps in ephrin triple-knockout (ephrin-A2/3/5 -/-) mice retina Superior colliculus 200 A 150 100 50 Position in the eye 0 A -50 -100 B -150 0 20 40 60 80 100 120 200 B 100 0 -100 Data from -200 0 20 40 60 80 100 120 David Feldheim (UCSC) and Position in the target Jianhua Cang (Northwestern) (2007) 31

Ocular dominance patterns 32

Ocular Dominance Pattern Left eye Right eye Horton and Hocking, 1996 33

3-eyed frog Normal frog 3-eyed frog 3 1 2 1 2 Each eye completely crosses over to the other side 34 Cline, Debski, Constantine-Paton (1987)

Two eyes projecting to the same target E1 E3             1 0     E1 col col   ˆ H q r C U r r  C i i ij i j 2   0 1 E3 i ij col Eye 1 Eye 1 & 3 Eye 1 Eye 3 Eye 3 q q � 35

Conclusions: “Nothing in biology makes sense except in the light of evolution” - Theodosius Dobzhansky Nothing in intelligent behavior makes sense except in the light of biology Genomic bottleneck principle #1: information about brain’s capacity for general intelligence is compressed into < 1GB of mammalian genome - Wei, Tsigankov, Koulakov, Annals of NYAS (2013) Genomic bottleneck principle #2: The need to compress information about brain architecture into a small volume (<1GB) endowed mammalian brain with general intelligence. - Tony Zador, Nature Communications (2019) 36

Neurotheory Lab is hiring! koulakov@cshl.edu 37 darkstar.cshl.edu