Neuroethology: Electric Fish MCB105, 23 March 2006 Walter Heiligenberg Weakly electric fish Completely understood circuits Algorithms that generalize to other systems

Lecture 07 - main goals: Understand the neural processes underlying the jamming avoidance response in the weakly electric fish Jamming Avoidance Respond – reflex behavior

Delay lines, A before B, direction selectivity Lecture 07 – Circuit Motifs Delay lines, A before B, direction selectivity General Principle: timing difference Delay lines – Reichardt detector

The Organism: Electroreception All organisms that show electroreception Sense and generate electric fields Means of communication, navigation, killing(hunting) Neuromuscular junction much larger than in other animal kingdom Huge synaptic current to cause electric field (neurotransmitter: Acetylcholine release)

The Organism: Weakly-Electric Fish Eigenmannia Eigenmannia (weakly electric fish) is what we will study. Independent, convergent evolution. Electric eel is different – can actually kill prey with shock

The Organism: Weakly-Electric Fish, Eigenmannia They can generate an electric field. (Frequency 200 Hz). Electric Organ Discharge from the tail. Lateral line is covered with electric field detectors Similar to sonar (sound), but feed back is instantaneous (speed of light)

The Organism: Weakly-Electric Fish, Eigenmannia, Electric Organ Discharge (EOD) Electric organ –consists of myocytes (muscles). Motor neuron innervates this cell, resulting behavior is discharge of electric fields. Electric Organ

The Organism: Weakly-Electric Fish, Eigenmannia, Electrosensation Fish lives in murky waters – so light is not effective, and they use electric fields to sense their environment (as well as to communicate with each other). Fish can measure deviations in this field using receptors on the body of the fish. Objects need to have lower conductance than water to be detected. Bats/Dolphins do similar things (echolocation), but use sound and delays. Electric field has no delay – just looking at perturbations.

The Problem: Fish-Fish Interference 2 fish hunting/exploring – if they get close to each other, two signals interfere. ~250Hz sine wave is usually the field they generate (220-270Hz). 2 frequencies close to each other – you get a beat (interference) pattern. When microphone shrieks – this is an example of such interference. Avoid jamming, move your own emitter frequency away from interference frequency Simplest strategy: drop or raise frequency sample environment, if it gets worse, change direction

The Problem: Interference Self-generated (S1) vs alien-generated (S2) When two waves have slightly different frequencies, they interfere constructively (positive) at some points and destructively (negative) at other points forming a beat pattern. Electric lateral line measures the sum of the two frequencies. If two frequencies are close to each other, you get beat pattern (S1 + S2). Only get beat pattern if they are slightly different frequencies. If same frequency, whether they are in phase or out of phase, you don’t get a beat.

The Behaviour: Jamming Avoidance Response (JAR) What is the behavior we are studying? Take a fish that generates an electric field (e.g. 250 Hz), then jam him with another frequency (either using another fish, or 2 electrodes in the water artificially). You can hear his own discharge by putting an antenna in the water and listening from a loudspeaker. Fish’s frequency moves away from the interfering frequency. Always move instantly in the correct direction! This ultimately achieves the goal of reducing beat frequency. How does he know that the interfering frequency is lower or higher than his own? While he can measure his own frequency, he doesn’t actually use this method to determine whether his own or alien frequency is higher. What does he use instead?

The Behaviour: Jamming Avoidance Response (JAR) What gets measured? How does the network arrive at the decision? Florian shows demonstration using Labview. Different features of the beat interference pattern: Not only that the envelope is going up or down, but also the zero crossings are changing (pay attention to which comes first, sum/alien frequency or own, and when).

The Circuit: Electroreceptors Stimulus Behaviour What can be measured? Primary receptors are the tuberous receptors. P and T units P – primary sensory receptors that are amplitude-modulated. Firing rate modulates with amplitude of electric field. T -respond to zero-crossings. Biophysical properties of receptors are still unclear, but you know what they respond to and how they respond.

The Circuit: Electro-Lateral Line Stimulus Behaviour ON OFF Downstream of the receptors , in the electro-lateral line (ELL), there are two types of cells. Spherical cell that receive input from T-receptors (phase coders). We will discuss the spherical cells later. Pyramidal cells that receive input from P-receptors (amplitude coders). There are two types of pyramidal cells , E unit (“on-cell”) and I unit (“off-cell”). Inhibitory interneurons in between ensure that when one fires, the other is silent. E-unit will fire when P receptors fires, I unit will fire when P receptors are silent.

The Circuit: Electro-Lateral Line Stimulus Behaviour General principle: Derivative On- vs off-cells What do these pyramidal (E and I cells) code for? E-cell fires when amplitude is rising. I-cell fires when amplitude is falling (note that these cells are coding for the derivative of firing– changes in amplitude versus absolute amplitude). Desensitization (makes sure cell will reset)

See notes below for explanation about processing of differential phase and why it is important for the JAR. Next slide will tell you how it is calculated/implemented in the Torus Semicircularis circuitry. What signals do the fish measure to compute whether it’s electric field frequency is higher or lower than the alien fish’s electric field? There are two: Phase (zero crossings) and amplitude of the beat pattern, which is the sum of it’s own frequency and the alien fish’s frequency. (From now on, I will call this the sum frequency) Phase (zero crossings) and amplitude of his own frequency. Note that he cannot measure anything about alien frequency itself– although this is what he wants to detect. More specifically, the fish can get two pieces of information: Is the envelope of the sum frequency (i.e. beat pattern) going down or up? The amplitude modulators (P receptor cells) can detect that. Whether the sum frequency is generally rising or falling. If E cells that P afferents project to are active -- amplitude of beat is going up. If I cells are active – amplitude of beat is going down. 2) Do the zero crossings of the sum frequency come before or after it’s own frequency? Zero crossings: T-receptor cells will fire every time the wave crosses the zero line. There are T- cells all along the body of the fish. Some of these T-cells have “privileged” positions relative to the EOD such that they can measure the fish’s own zero crossings. Others just measure the sum frequency detected in environment (own + alien). This basically means the fish can measure phase (zero-crossings) of it’s own discharge frequency, and of the summed frequency. Below I will describe in excruciating detail all the possible combinations of phase and amplitude and what they mean! Recap: When two waves have slightly different frequencies, they interfere constructively (positive) at some points and destructively (negative) at other points forming a beat pattern. At the peaks or troughs of the beat pattern, they will be complete in or out of phase respectively. When the amplitude of the beat pattern envelope is declining (I cells active), the waves are not completely in phase. The wave with the higher frequency will cross zero first. Thus, if the alien frequency is higher than the fish’s own frequency (Df>0*), the sum** frequency will cross zero BEFORE the fish’s own frequency (I cells fire AND phase advance*) Alternatively, if the alien frequency is lower than the fish’s own frequency (Df<0), the sum frequency will cross zero AFTER the fish’s own frequency (I cells fire AND phase delay) When the amplitude of the beat pattern is increasing (E cells active), the waves are not completely in phase. The wave with the lower frequency will cross zero first. Thus, if the alien frequency is higher than the fish’s own frequency (Df>0), the sum frequency will cross zero AFTER the fish’s own frequency. (E cells fire AND phase delay) Alternatively, if the alien frequency is lower than the fish’s own frequency (Df<0), the sum frequency will cross zero BEFORE the fish’s own frequency (E cells fire AND phase advance). *Nomenclature: Df and delay/advance are named with respect to alien frequency (alien minus own frequency). **Note that since the sum frequency is own + alien frequency, if alien frequency’s zero-crossings come first, so would the sum, and vice versa. So these two are related. In summary: If alien frequency is higher than own frequency (Df>0), E AND delay (E cells will fire AND phase delay will be encoded) OR I AND advance (I cells will fire AND phase advance will be encoded) If alien frequency is lower than own frequency (Df>0), E AND advance (E cells will fire AND phase advance will be encoded) I AND delay (I cells will fire AND phase delay will be encoded)

The Circuit: Torus Semicircularis (L6) Stimulus Behaviour Now we get to how the phase advance or delays described in the previous slide are actually computed by the circuit. This computation occurs in Torus Semicircularis (TS Layer 6) If you remember, spherical cells in the ELL receive input from T cells that fire when there are zero crossings. These spherical cells then project to the TS Layer 6. In this example, let’s assume that spherical cell A is measuring the fish’s OWN zero crossings, and B, C, and D are along the fish’s body and measuring the sum frequency (alien + own electric fields). In this example, this particular circuit detects when sum zero-crossings arrive BEFORE it’s own zero-crossings (i.e. phase advances). Note that there is also a circuit which will operate in the opposite way, detecting phase delays, that is not shown here! The circuit makes the decision very quickly, detecting differences on the order of a millisecond. There are giant cells, and small cells. Why does size of cell matter. RC time constant – the larger the cell, larger capacitance and resistance – takes longer time to depolarize. Thus, giant cells can act as delay lines. Converse for small cells. -------- Let’s look at the Ha-Hb small cell. Since the cell is small, it has a short time constant, which means it is a precise coincidence detector --- the two signals from spherical cell A and B need to arrive at the same time for for small cell Ha-Hb to fire. In this example, small cell Ha-Hb will fire if sum frequency detected by spherical cell B crosses zero BEFORE it’s own frequency detected by spherical cell A. This is because it small cell receives information about it’s OWN zero-crossings from spherical cell A , as well as delayed information (of about 1ms) about SUM zero-crossings from spherical cell B. Delay occurs because B goes through a giant cell before going to Ha-Hb small cell. Again, giant cells have longer time constants and take a longer time to depolarize, thus they are delay lines. Therefore If B fires before A (sum zero-crossings before own zero-crossings, or a phase advance), then the two inputs will coincide. This would cause small cell Ha-Hb to fire. ----- Again, this is only one example. A different circuit exists for when OWN zero-crossings comes BEFORE sum zero-crossings. In this case, the input from the spherical cells detecting OWN zero-crossings will have to go through a giant cell delay line, and the input from the spherical cells detecting SUM zero-crossings would not be delayed.

The Circuit: Torus Semicircularis (L6) Stimulus Behaviour General Principle: timing difference Delay lines – Reichardt detector Therefore, timing of zero crossings (sum relative to own) is computed in the Torus Semicircularis, and is very essential for the fish to decide whether alien frequency is higher or lower. Another example of the use of delay lines: Direction selective cell (Reichardt detector) For a right to left moving signal, the right neuron gets activated first BUT delayed, and subsequently the left on gets activated. So they meet downstream at the same time, and that neuron knows that the stimulus is moving from right to left. Also in barn owls for ITDs!

The Circuit: Torus Semicircularis (L8) Stimulus Behaviour Phase Advance AMP “OFF” (I cells) AMP “ON” (E cells) Next, still in the Torus Semicircularis (but Layer 8) Remember that the fish still needs to combine zero crossing input with amplitude information. This is because within one cycle of the beat pattern, zero crossings will switch depending on whether beat amplitude is rising or falling. So you need another detector.. This is called the sign-selective neuron. Combines E-cell with phase advance, or I-cell with phase delay, or E-cell with phase delay, or I-cell with phase advance.

The Circuit: Torus Semicircularis (L8) Stimulus Behaviour Phase Advance AMP “OFF” (I cells) AMP “ON” (E cells) So there will be sign-selective neurons for all the different combinations of E vs I and phase advance vs delay. E + Advanced Cells (Df < 0) I + Advanced Cells (Df > 0) Cells still have body-location specific RFs.

The Circuit: Nucleus Electrosensorius Stimulus Behaviour Remember that there are two combinations each (of E vs I, and phase advance vs delay) that signal Df > 0 or Df <0. This information is combined in higher order sign-selective neurons of the Nucleus Electrosensorious. Recap: E + Delay or I + Advance – Alien frequency higher (Df>0) E + Advance or I + Delay – Alien frequency lower (Df<0).

The Circuit: Pre-Pacemaker Nucleus Stimulus Behaviour Pre-pacemaker nucleus (PPn) Neurons in the PPn are excited by Df<0 and inhibited by Df>0 cells that project from the Nucleus Electrosensorius. If Pre-pacemaker nucleus is excited (i.e. Df<0), then the downstream Pacemaker Nucleus will be EXCITED (i.e. increase it’s firing rate).

The Circuit: Pacemaker Nucleus Stimulus Behaviour Finally, the pacemaker nucleus (Pn) When PPn increases it’s firing rate, Pn will also increase it’s firing rate. This is because Pn discharges at the firing rate of PPn. So, when Df<0 (fish’s own frequency is more than the alien’s), PPn and Pn will increase their firing rate. Since Pn paces the electric organ discharge, this means that the electric organ will now discharge at a HIGHER frequency! Conversely, when Df>0 (alien’s frequency is higher than fish’s own frequency), , PPn and Pn will decrease their firing rate. So now the fish lowers it’s own frequency. In both cases, the fish’s own frequency diverges from alien frequency. Mission accomplished!

The Circuit: SUMMARY Stimulus Behaviour The entire beautiful circuit that we understand completely.

The Point: Lessons of Neuroethology (The BIG cell bias) We are biased towards big cells – but small cells are important –in this circuit they are actually the coincidence detectors!

The Point: Lessons of Neuroethology (The BIG brain region bias) Also, a word of caution about ablation experiments. This ablation in the tectum initially looked like it affected electrosensation – but turned out the researchers had ablated the axons from the ELL to Torus Semicircularis. So they implicated the wrong brain region in JAR! But now everything has been clarified.

The Point: Lessons of Neuroethology (The Engineer’s (Physicist) Bias) Finally, the brain doesn’t necessarily do stuff the way an engineer might design it to. For example, there is information in the fish’s circuits (in it’s pacemaker neurons) that tells the fish what it’s own frequency is. Turns out the fish doesn’t use that for the Jamming Avoidance Response!

Paper Heiligenberg Chapter 5 General Principles in the Neuronal Organization of the Jamming Avoidance Response. Kawasaki (1988) Anatomical and Functional Organization of the Prepacemaker Nucleus in Gymnotiform Electric Fish: The Accommodation of Two Behaviors in One Nucleus

Matlab Module: Reichardt Detector