*For correspondence: R.E. Goldstein@damtp.cam.ac.uk Competing interests: The authors declare that no competing interests exist. Funding: See page 13 Received: 08 June 2016 Accepted: 17 October 2016 Published: 24 November 2016 Reviewing editor: Richard M Berry, University of Oxford, United Kingdom Copyright Kirkegaard et al. This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited. Aerotaxis in the closest relatives of animals Julius B Kirkegaard, Ambre Bouillant, Alan O Marron, Kyriacos C Leptos, Raymond E Goldstein* Department of Applied Mathematics and Theoretical Physics, Centre for Mathematical Sciences, University of Cambridge, Cambridge, United Kingdom Abstract As the closest unicellular relatives of animals, choanoflagellates serve as useful model organisms for understanding the evolution of animal multicellularity. An important factor in animal evolution was the increasing ocean oxygen levels in the Precambrian, which are thought to have influenced the emergence of complex multicellular life. As a first step in addressing these conditions, we study here the response of the colony-forming choanoflagellate Salpingoeca rosetta to oxygen gradients. Using a microfluidic device that allows spatio-temporal variations in oxygen concentrations, we report the discovery that S. rosetta displays positive aerotaxis. Analysis of the spatial population distributions provides evidence for logarithmic sensing of oxygen, which enhances sensing in low oxygen neighborhoods. Analysis of search strategy models on the experimental colony trajectories finds that choanoflagellate aerotaxis is consistent with stochastic navigation, the statistics of which are captured using an effective continuous version based on classical run-and-tumble chemotaxis. DOI: 10.7554/eLife.18109.001 Introduction Taxis, the physical migration towards preferred or away from undesired conditions, is a feature shared by virtually all motile organisms. Taxis comes in many forms, and in common is an underlying field of attractant (or repellent) and an ability to react and navigate along gradients of this field. Bac- teria do chemotaxis towards nutrients (Adler, 1969; Berg, 1993) and away from toxins (Tso and Adler, 1974). Algae do phototaxis towards light (Yoshimura and Kamiya, 2001; Drescher et al., 2010) and gyrotaxis along gravitational potentials (Kessler, 1985). Chemotaxis provides a mecha- nism for the recognition and attraction of gametes (Vogel et al., 1982) and for complex behavioural patterns such as in the slime mould Dictyostelium discoideum, where cAMP-driven chemotaxis is a critical part of the formation of the multicellular stage of the life cycle (Bonner, 1947). Aerotaxis, defined as oxygen-dependent migration, is well-characterized in bacteria (Taylor et al., 1999), but is poorly studied in more complex organisms. This is despite the essentiality of oxygen for all aerobic life, and the important role that Precambrian oxygen levels played in the emergence and evolution of multicellular animal life (Nursall, 1959). One group of aquatic heterotrophic protists, the choanoflagellates, are of particular interest for the study of how multicellularity evolved. Choanoflagellates are a class of unicellular microorganisms that are the closest relatives of the animals (Lang et al., 2002). This relationship was first proposed by James-Clark in 1866 (James-Clark, 1866), on the basis of the resemblance between choanofla- gellates and the choanocytes of sponges. The sister relationship between choanoflagellates and ani- mals was further confirmed in the genomic era by molecular evidence (King et al., 2008). All choanoflagellates share the same basic unicell structure: a prolate cell body with a single beating fla- gellum that is surrounded by a collar of microvilli. The beating of the flagellum creates a current in the surrounding fluid that guides suspended prey such as bacteria through the collar (Pettitt and Kirkegaard et al. eLife 2016;5:e18109. DOI: 10.7554/eLife.18109 1 of 16 RESEARCH ARTICLE Orme, 2002) where they can be caught and ultimately phagocytosed. The flagellar current also has the effect of causing the choanoflagellate cell to swim. The choanoflagellate Salpingoeca rosetta can form colonies through incomplete cytokinesis (Fairclough et al., 2010). In the presence of certain bacteria (Dayel et al., 2011; Levin et al., 2014), these colonies have an eponymous rosette-like shape as shown in Figure 1. The colony morphology is variable, and the constituent flagella beat independently of one another (Kirkegaard et al., 2016). The random and independent flagellar motion argues against there being any coordination between cells in a colony, and as yet no evidence of any form of taxis for choanoflagellate colonies has been reported. The geometry, flagella independence and lack of taxis observed in S. rosetta colonies contrast with other lineages, such as the Volvocales, a group of green algae (Goldstein, 2015). Phototaxis is clearly observable in both unicellular (Chlamydomonas) (Yoshimura and Kamiya, 2001) and colonial (Volvox) (Drescher et al., 2010) species, in order to maintain optimum light levels for photosynthe- sis. Volvocalean phototaxis is deterministic, requiring precise tuning between the internal biochemi- cal timescales and the rotation period of the organism as a whole. Although S. rosetta colonies also rotate around an internal axis, due to the variable colony morphology and the independent beating of the individual flagella, this rotation rate will itself be random (Kirkegaard et al., 2016), rendering a strategy similar to that of the Volvocales unlikely in S. rosetta. An alternative strategy is stochastic taxis, sometimes referred to as kinesis. The classic example of stochastic taxis is the run-and-tumble chemotaxis of certain peritrichous bacteria (Berg, 1993). By eLife digest Most animals are made up of millions of cells, yet all animals evolved from ancestors that spent their whole lives as single cells. Today the closest single-celled relatives of animals are a group of aquatic organisms called choanoflagellates. Certain species of choanoflagellates can also form swimming colonies. This kind of multicellularity might resemble that seen in the earliest of animals. As such, studies into modern-day choanoflagellates can give insights into how the first animals to evolve might have behaved. Many organisms can find their way towards favorable areas using different strategies. For instance, bacteria can bias their tumbling to gradually swim towards food, and algae can turn and move directly towards light. While choanoflagellates require oxygen, it was not known if they could also actively navigate towards it, or any other resource. Now, Kirkegaard et al. find that the choanoflagellate Salpingoeca rosetta can indeed navigate towards oxygen – an ability called aerotaxis. This was true for both individual cells and for colonies made up of many cells. This discovery suggests that the transition from living as a single cell to living as a simple multicellular organism could still have allowed the earliest animals to seek out and move towards resource-rich areas. Aerotaxis requires cells to both sense oxygen and react appropriately to changes in its concentration. Kirkegaard et al. watched choanoflagellate colonies swimming under controlled conditions and varied the oxygen concentration in the water over time. These experiments revealed that the colonies navigate based on the logarithm of the oxygen concentration, so that at low oxygen levels the cells were even more sensitive to small changes in oxygen concentration. This type of ‘logarithmic sensing’ is similar to how our ears sense sounds and our eyes sense light. Kirkegaard et al. went on to conclude that the colonies were not actively steering in the correct direction directly. Instead, the colonies appeared to choose directions at random and later decide whether such a turn was correct. It remains unclear whether the common ancestor of animals and choanoflagellates could also perform aerotaxis, and if so what mechanisms this involved. Further studies to compare aerotaxis and aerotaxis-related genes in simple animals and other single-celled relatives of animals would be needed to illuminate this. Future studies could also explore the maximum and minimum oxygen concentrations that choanoflagellates can detect, and how well they navigate at these upper and lower limits. DOI: 10.7554/eLife.18109.002 Kirkegaard et al. eLife 2016;5:e18109. DOI: 10.7554/eLife.18109 2 of 16 Research article Biophysics and Structural Biology spinning their left-handed helical flagella in different directions, such bacteria can alternate between swimming in straight lines (running) and randomly reorienting themselves (tumbling). Through bias- ing tumbles to be less frequent when going up the gradient, the bacteria exhibit biased motion towards a chemoattractant without directly steering towards it (Berg, 1993). Here, we study S. rosetta and show that it exhibits aerotaxis, i.e. navigation along gradients of oxy- gen. We further examine and statistically analyse aerotaxis of S. rosetta colonies under spatio-tempo- ral variations of oxygen at the level of total colony populations and at the level of the trajectories of individual colonies. From these experiments we establish two key features of the aerotactic response of choanoflagellates: they employ a stochastic reorientation search strategy and the sensing of oxygen concentration gradients is logarithmic. Finally, we render these results quantitative through the use of mathematical analysis of a modified Keller-Segel model (Keller and Segel, 1971). Results Experimental set-up The study of aerotaxis in bacteria has led to numerous methods for creating spatial oxygen gradients (Shioi et al., 1987; Wong et al., 1995; Zhulin et al., 1996; Taylor et al., 1999), one of which is the exploitation of soft lithography techniques (Adler et al., 2012; Rusconi et al., 2014). Since PDMS, the most commonly used material for microfluidic chambers, is permeable to gases, gas channels can be introduced in the devices to allow gaseous species to diffuse into the fluid. For example, an oxygen gradient can be created using a source channel flowing with normal air and a sink channel flowing with pure nitrogen. Our device, shown schematically in Figure 2, is a modified version of that used by Adler, et al. (Adler et al., 2012). Viewed from above, the sample channel (yellow) consists of a wide observation chamber with thin inlet and outlet channels. The outlet leads to a serpentine channel that hinders bulk fluid flows. On each side of the sample channel are gas channels, the inlets of which are con- nected to a valve system allowing for the flow of air (20% oxygen) and nitrogen. The flow of air and nitrogen can be conducted in any combination and configuration, e.g. oxygen in one channel and nitrogen in the other, and can be easily swapped over. The PDMS chamber is plasma etched to a glass slide, and an extra glass slide is etched on top of the device, preventing air from diffusing in from the surrounding environment. Figure 1. Micrograph of S. rosetta colonies (left) with schematic illustration (right, collars in blue). Scale bar: 50 mm. Cell body diameters are ~5 mm. DOI: 10.7554/eLife.18109.003 Kirkegaard et al. eLife 2016;5:e18109. DOI: 10.7554/eLife.18109 3 of 16 Research article Biophysics and Structural Biology Experiments were carried out immediately after plasma etching as the permeability of gases slowly decreases thereafter. Cultures of S. rosetta were introduced at the inlet of the device, and both the inlet and outlet were then closed to pre- vent evaporative flows. A gradient of oxygen was set up by having air flowing in one of the side channels and nitrogen in the other. Aerotaxis in choanoflagellates Our main experimental result, shown in Figure 3a,b, is the observation that S. rosetta colonies accumulate at the oxygen-rich side and away from the oxygen-poor side, i.e. that they are aerotactic. We also found aerotaxis in the uni- cellular fast swimmer form (Dayel et al., 2011) of S. rosetta, showing that this is not an exclusive phenomenon to colonies. With the present microfluidic device we can explore more details of choanoflagellate aero- taxis by dynamically changing the oxygen bound- ary conditions, for instance by flipping the gradient direction or by removing all oxygen influx after a uniform distribution has been reached. Figure 3c shows the result of such a dynamic experiment over the course of ~3.5 hr. The density is normalized for each frame and the noise present is partly due to colonies missing in the tracking in some frames. Many repetitions of the experiment show that the behaviour in Figure 3c is highly repeatable and robust to changes in the details of the cycling protocol (See Figure 3—figure supplement 1). For consistency the figures in the main text are based on this specific protocol. Whenever one gas channel contains oxygen and the other nitrogen, the colonies swim towards the oxygen-rich side as further shown in Video 1 . In the time after a gas channel swap, the slope of the maximum density reveals the ensemble drift velocity vdrift. When there is oxygen in both gas channels, we observe that the density reaches an approximately uniform distribution within the time frame of the experiment. For periods in which nitrogen flows in both channels, this is not the case. Under these experimental conditions, the colonies accumulate in the middle of the chamber, where there is still some residual oxygen, as further shown in Video 2. The fact that in this nitrogen-only configuration the colonies accumulate mid-chamber shows that accumulation does not depend on the presence of a nearby surface. With only nitrogen flowing, eventually there will be no oxygen gra- dient. Nonetheless, we observed the colonies to stay in the middle of chamber even after 90 min (see Figure 3—figure supplement 1). At that time, the highest oxygen levels are estimated by the diffusion equation to be less than ~0.2%. This contrasting behaviour between the oxygen-only and nitrogen-only configurations suggests an asymmetry or non-linearity in the aerotactic response. If the response to oxygen concentration had been linear, the observation of the density band in the nitrogen-only section would imply similar density bands at the chamber edges in the oxygen-only section, which is not observed. Instead one might hypthesize that the colonies navigate along relative (rc=c) instead of absolute (rc) gradients, i.e. reacting to gradients that are comparable in magnitude to the background concentration. This is also known as logarithmic sensing, and we will confirm in the modelling section that this hypothesis can quantitatively explain the experiments. Navigation strategy Strategies of taxis can be categorized into two main classes: deterministic and stochastic. In both strat- egies the swimming organism measures the attractant gradient (for small organisms by some temporal filter [Block et al., 1982; Celani and Vergassola, 2010]). A deterministic strategy, then, is one in which 10 mm Figure 2. Microfluidic device. (a) Top view of the device. The sample channel (yellow) is loaded with culture and observed in the middle chamber. The side channels (red, blue) are gas channels in which oxygen and nitrogen may be flown. Scale bar: 10 mm. (b) Side view of the device. PDMS is plasma etched to a glass slide, and a cover slip is plasma etched on top, centered on the imaging chamber, also shown in (a). Thickness of the channels are » 115 mm. DOI: 10.7554/eLife.18109.004 Kirkegaard et al. eLife 2016;5:e18109. DOI: 10.7554/eLife.18109 4 of 16 Research article Biophysics and Structural Biology the organism directly steers towards the attractant, such as seen in sperm cells that modulate their fla- gellar beat to adjust directly the curvature and torsion of its swimming path in the gradient direction (Friedrich and Ju¨licher, 2007; Jikeli et al., 2015). Contrasting is a stochastic strategy such as bacterial run-and-tumble locomotion (Berg, 1993), where modulation of the frequency of random reorienta- tions biases the motion in the gradient direction without directly steering towards it. One simple method of taxis results from an organism swimming faster when it is moving up the gra- dient, creating an overall bias towards the attractant. With the detailed colony-tracking in the present study it is possible to test whether this mechanism is in operation with S. rosetta. Figure 4—figure supplement 1 shows examples of tracks during periods of uniform swimming (t = 70 min) and after a gas channel swap (t = 142 min). Figure 4 shows the evolution of the mean colony swimming speed v (green) as well as the component velocities vx (yellow) and vy (purple), averaged over ~150 colonies in each frame. For most times, the component velocities average to zero, but after a gas channel swap the y-component peaks. The ensemble average swimming speed in these sections, however, does not show an increase, suggesting that a velocity modulation is not the method of taxis. To quantify this fur- ther, the inset of Figure 4 shows the swimming speed in these sections plotted against the alignment to the gradient c^ vy=v where c^ ¼ 1 signifies the direction of the gradient. The plot shows a very small (~3%) change in swimming speed going up the gradient. Velocity-biased taxis can be described by vðtÞ ¼ vðp^Þ p^, where e.g. vðp^Þ ¼ v ð1þ g p^  rc=jrcjÞ, g being the velocity-modulation taxis parameter. p^, the direction of swimming, is unbiased by the attractant field c and evolves by rotational diffusion. To obtain a drift velocity ~1/3 of the swimming velocity, as we find for S. rosetta in the following sec- tion, the velocity modulation would have to be g= 2/3 for a two-dimensional swimmer and g = 1 in three dimensions, much larger than the ~3% observed. We conclude that the primary mechanism of aerotaxis in S. rosetta is therefore not a modulation of swimming speed. S. rosetta colonies swim along noisy helical paths, and each colony displays distinct helix parame- ters (Kirkegaard et al., 2016). To perform any kind of statistical angle analysis, we consider ensem- ble average quantities: the speed v and rotational diffusion dr, and average helix rotations out. (b)(a) 500 0 500 y ( m ) (c) 0 50 100 150 200 t (min) 500 0 500 y ( m ) (d) 0 1 2 3 4 ×10−3Air Nitrogen Figure 3. Aerotaxis of S. rosetta colonies. (a–b) Micrographs near an oxygen-rich wall at twice the resolution of that used in the density experiments. Scale bar: 50 mm (a) Colonies approach a wall where the oxygen-concentration is high. (b) Colonies staying near this wall. (c) Density evolution of S. rosetta during experiment. At each time step the distribution is normalized to a probability distribution [colorbar units in mm1]. Colors on the side indicate what gas is flowing in that side channel, red for oxygen and blue for nitrogen. Ncolonies~150, concentration ~5  106 mL1. (d) Keller-Segel model with log-concentration input given by Equation 4, D = 865 mm2/s, a = 1850 mm, vdrift = 5.2 mm/s. DOI: 10.7554/eLife.18109.005 The following figure supplement is available for figure 3: Figure supplement 1. Seperate aerotaxis experiment. DOI: 10.7554/eLife.18109.006 Kirkegaard et al. eLife 2016;5:e18109. DOI: 10.7554/eLife.18109 5 of 16 Research article Biophysics and Structural Biology Figure 5a shows the angular distribution data during the swaps, where, for the purposes of dis- playing all the data in a single graph, we have let  !  for times when the oxygen gradient were pointing down. This distribution favors the up- direction  ¼ p=2. More interesting is the distri- bution of reorientations. For this we define the angle turned by a colony in a time Dt as Df ¼ jðt þ DtÞ p=2j jðtÞ p=2j such that it is posi- tive if the turn is in the direction of the gradient and negative otherwise, and choose Dt low enough that p