Go against the flow
How fish detect and keep their position in water flowing with constant velocity
Understanding brain function generally requires a deep engagement with the matter at least at two separate levels. These are the level of behavioral algorithms on the one hand and the level of neural implementation on the other. These questions can be rephrased as “What exactly is the brain doing?” and “How does it do whatever it is doing?”. Central to this division of labor in the brain sciences is the idea, inspired by David Marr, that the brain implements algorithms, whether for sensory processing, decision making, or motor control, and that these algorithms can be inferred from careful observation of ethologically relevant behaviors. Once a particular algorithm is understood and delineated, we can interrogate its neural implementation by measuring from and manipulating the underlying neural circuits in the context of behavior. Uncovering behavioral algorithms usually does not require modern and cutting edge technologies like optogenetics, wholebrain imaging and genetic editing, it merely requires precise observation, thorough experimental design and, most importantly, rigorous and deep thinking. Such basic scientific qualities are somewhat threatened, and often considered quaint, in the modern era of big data, high throughput and cutting edge technologies. In light of these concerns it comes as a truly pleasant surprise that a study based exclusively on such somewhat antiquated techniques can still find a place in a high impact journal and generate plenty of enthusiasm in the scientific community. The study published recently in Nature by the group of Florian Engert at Harvard in collaboration with the group of Ruben Portugues at the Max Planck Institute of Neurobiology relied entirely on the old-fashioned technique of careful behavioral observations and could have been accomplished in the same form half a century ago. As such it might as well considered “timeless”.
In short, they wanted to know how a larval zebrafish, when placed into a flowing body of water, can detect the presence of the current and then effectively swim against it. More precisely, the group, spearheaded by the team of Pablo Oteiza and Iris Odstrcil, identified the lateral line as one of the main sensory modalities the fish can use to detect that it is drifting with respect to the shore – and more importantly - they could uncover the precise algorithms that translate this sensory input into the appropriate motor commands. Specifically, they found that local flow gradients across the fish’s body, i.e. minute differences in water-flow over the animal’s right side when compared to its left side, turned out to be the essential cue for appropriate navigation. Up to now the primary source of confusion in the field of rheotaxis was based on a widespread and faulty intuition about moving frames of reference, an issue that was first tackled rigorously by Galileo Galilei, whose works can be found amongst the references cited in the paper: since the days of Galileo, we know that the laws of physics should be the same if the reference frame is moving at a constant velocity. In a train moving at constant velocity, we would find it equally difficult to walk towards the front or the rear of the wagon. In fact, if we closed our eyes, we would find it impossible to know that we were moving at all. A fish in a stream moving at constant speed is like one of us in a train – and if it is swimming in the dark, it should not know which way the flow was moving, and would get swept down the river.
However, this is not the case. In fact, most fish tend to swim against the flow, a behavior that ensures that they stay in the same position with respect to the riverbed. This behavior is called rheotaxis and is independent of any visual cues: fish will do it in complete darkness. But how?
The elegance of the story uncovered by the groups of Florian Engert and Ruben Portugues has its root in the relative simplicity of the identified algorithm on the one hand, and the deep bewilderment that pervaded the community about the general nature of the fish’s performance on the other. This behavior, up to now, was truly very hard to explain.
The first piece of the puzzle was in the identity of the responsible sensory receptors. Pablo Oteiza, one of the two leading authors of the study, explains: “Fish have a sensory organ called the lateral line. It is a dense collection of hairs that is distributed around their body and senses local perturbation in water flow. It mediates what is known as the sense of distant touch because anything moving in water will create a flow disturbance. Larval zebrafish swim in events called bouts that last 250 milliseconds and then rest 1 second in between. We show that larval zebrafish, while they are in between bouts, use this sensory modality to collect a snapshot of the small circular flow around their body, which is induced by linear local flow velocity gradients. If we perturb the lateral line organs along the tail fish cannot sense the presence of a current and consequently fail to perform rheotaxis.”
Iris Odstrcil, the other leading author, continues: “By measuring this flow gradient before and after a bout, the fish can effectively compute how it has changed as a consequence of its last bout. If the gradient decreases the fish will continue swimming normally. However, if the gradient increases, the larva will change its behavior and will be more likely to turn.” In the study, the authors proceed to use computer simulations to show that this simple algorithm does indeed lead to the rheotaxis behavior they observed experimentally. “Frames of reference are sometimes a little bit tricky to get our head around. The water in a swimming pool on the back of a truck moving at constant velocity will not move relative to the occupant of the pool,” says Florian Engert. “A rubber duck in the center of the swimming pool would stay in the center. The same should be true for a fish in a river. Now we know why it is not”.
The lateral line is an often overlooked sensory modality because mammals such as ourselves do not have it, but it is extremely interesting. Ruben Portugues explains that “at each time-step, the fish must integrate the information from all the different hair-cells around its body and use this to compute an estimate not of the flow, but the gradient of the flow. In essence it is performing a vector integral, memorizing this result, performing it again and comparing it.”
“That is quite something no?” concludes Pablo. “We now need to figure out how their little brain does that!”
Press release by Harvard University