Flagellar synchronization through direct hydrodynamic interactions
Abstract
Flows generated by ensembles of flagella are crucial to development, motility and sensing, but the mechanisms behind this striking coordination remain unclear. We present novel experiments in which two micropipetteheld somatic cells of Volvox carteri, with distinct intrinsic beating frequencies, are studied by highspeed imaging as a function of their separation and orientation. Analysis of time series shows that the interflagellar coupling, constrained by lack of connections between cells to be hydrodynamical, exhibits a spatial dependence consistent with theory. At close spacings it produces robust synchrony for thousands of beats, while at increasing separations synchrony is degraded by stochastic processes. Manipulation of the relative flagellar orientation reveals inphase and antiphase states, consistent with dynamical theories. Flagellar tracking with exquisite precision reveals waveform changes that result from hydrodynamic coupling. This study proves unequivocally that flagella coupled solely through a fluid can achieve robust synchrony despite differences in their intrinsic properties.
https://doi.org/10.7554/eLife.02750.001eLife digest
Sperm cells, as well as many bacteria and algae, propel themselves using whiplike appendages called flagella. Similar, shorter structures called cilia are also found on the surface of many cells, where they perform roles such as moving liquids over the cell.
Each cilium or flagellum beats at its own characteristic rhythm, but there are many situations where cilia or flagella must synchronize their beating with other nearby cells. For example, an egg cell is swept along the Fallopian tube by the coordinated beating of the cilia lining the tube. Bull sperm cells are also known to synchronize the beating of their flagella when swimming close to each other.
It has been suggested that the movement of the fluid surrounding the beating flagella could be the source of this synchronization. Experiments have produced results that match up with mathematical models describing this fluid movement. However, these experiments have often been designed in ways that didn’t fully exclude other possible sources of synchronization, such as chemical signalling, or—for flagella located on the same cell—a physical connection between the flagella.
To overcome this shortcoming, Brumley et al. used highspeed imaging to watch the flagella of cells of Volvox carteri—a species of green alga—that were separated so that they could only communicate through the movement of the fluid around them. The flagella were still able to synchronize their beating, even when the two flagella naturally beat at substantially different rates.
The distance between the flagella affects how well the beating synchronizes. When close together, the flagella can lock into the same rhythm for thousands of beats. However, as they move further apart, random biochemical fluctuations within the cells reduce the extent to which the flagella can synchronize.
The flagella can also synchronize so that they move in the same direction at the same time, or in opposite directions, depending on how they are oriented relative to each other. Moreover, the results confirm that the fluid flow produced by a beating flagellum is sufficient to synchronize the beating of other nearby flagella.
https://doi.org/10.7554/eLife.02750.002Introduction
Despite the elegance and apparent simplicity of the eukaryotic flagellum and its shorter ciliary version, the collective motions exhibited by groups of these organelles and the resultant fluid flows are far from trivial. For example, the unicellular biflagellate alga Chlamydomonas reinhardtii executes diffusive ‘runandturn’ locomotion (Goldstein et al., 2009; Polin et al., 2009) through stochastic switching between synchronized and unsynchronized swimming gaits—a process which could enhance searching efficiency and assist in the avoidance of predators (Stocker and Durham, 2009). Ensembles of cilia and flagella exhibit stunning temporal coordination, generating flows that transport mucus and expel pathogens (Button et al., 2012), establish the leftright asymmetry in developing mammalian embryos (Nonaka et al., 2002), and transport ova in human fallopian tubes (Lyons et al., 2006).
The origin of flagellar synchronization has been the subject of intense theoretical investigation for many decades. One of the earliest experimental results was Rothschild's qualitative observation (Rothschild, 1949) that the flagella of bull spermatozoa tend to synchronize when they swim close to one another, coupled only through the fluid surrounding them. Much more recent observations of selforganised vortex arrays of swimming sea urchin spermatazoa near surfaces (Riedel et al., 2005) provide further evidence for synchrony mediated purely by hydrodynamic coupling. Motivated by Rothschild's observation, Taylor (Taylor, 1951) developed a mathematical model in which two laterally infinite, inextensible sheets with prescribed sinusoidal travelling waves of transverse deformation interact with each other through a viscous fluid. He found that the rate of viscous dissipation is minimised when the two sheets are in phase. While minimisation of dissipation often holds in real physical systems, it is not in general a fundamental principle from which to deduce dynamical processes. Rather, an explanation for synchronization should capture the forces and torques associated with the underlying molecular motors that drive flagella, their elasticity, as well as the viscosity of the surrounding fluid.
Since Taylor's work a myriad of increasingly complex models of flagellar synchronization have been proposed. Hydrodynamically coupled filaments or chains with various internal driving forces exhibit a general tendency towards synchrony (Machin, 1963; Gueron et al., 1997; Guirao and Joanny, 2007; Yang et al., 2008; Elgeti and Gompper, 2013). At the same time, minimal models of coupled oscillators in viscous fluids (Vilfan and Jülicher, 2006; Niedermayer et al., 2008; Uchida and Golestanian, 2011, 2012; Brumley et al., 2012) offer great insight into the emergence of metachronal coordination. Such models have been investigated experimentally with light driven microrotors (Di Leonardo et al., 2012), rotating paddles (Qian et al., 2009) and colloids in optical tweezers (Kotar et al., 2010), and have also given rise to interpretations of the synchrony and coupling interactions between pairs of flagella of the model alga Chlamydomonas (Goldstein et al., 2009).
Although experimentallyderived coupling strengths between micropipetteheld Chlamydomonas flagella are consistent with predictions based on direct hydrodynamic coupling (Goldstein et al., 2011), it has been proposed (Friedrich and Jülicher, 2012; Geyer et al., 2013) instead that this coupling is too weak to overcome noise, and that residual motion of elasticallyclamped cells could play a role in synchronization. The recent observation (Leptos et al., 2013) of antiphase synchronization in a nonphototactic mutant of Chlamydomonas points as well to the possible role of internal mechanical coupling between flagella. Clearly, examining the synchronization between flagella on a single cell it is difficult to establish with certainty the origins of the coupling mechanism due to the likely presence of biochemical and elastic couplings of as yet unquantified strength between flagella.
In order to disentangle the hydrodynamic from the intracellular contributions to flagellar synchronization we conducted a series of experiments in which two physically separated flagellated cells, which exhibit distinct intrinsic beating frequencies in isolation, are coupled solely and directly through the surrounding fluid. These experiments can be viewed as natural generalisations of earlier work in which vibrating microneedles (Okuno and Hiramoto, 1976) or micropipettes (Eshel and Gibbons, 1989) are used to modulate and entrain the beating of a single sperm flagellum. Owing to the natural distribution of beating frequencies of the flagella of its surface somatic cells, the colonial alga Volvox carteri is ideally suited to this purpose. Each somatic cell possesses two flagella which beat in perfect synchrony, facilitating their treatment as a single entity, henceforth referred to as the flagellum. Somatic cells were isolated from adult Volvox colonies and held with micropipettes at a controllable separation d (Figure 1A,B). The spatial and orientational degrees of freedom associated with this configuration enabled comprehensive analysis over a wide range of hydrodynamic coupling strengths. We found that closelyseparated pairs of cells can exhibit robust phaselocking for thousands of beats at a time, despite a discrepancy in their intrinsic frequencies of as much as 10%. Both inphase and antiphase configurations were observed, depending on the alignment of the directions of flagellar propulsion. Furthermore, with increasing interflagellar spacing we observed for each flagellum a marked change in the beating waveform, a key finding that lends support to models of synchronization that rely on waveform compliance to achieve phaselocking.
Results
One cell
We begin by characterising the flow generated by a single beating flagellum. Despite the fact that a flagellum is a spatiallyextended object with considerable internal dynamics, it has become clear in recent years that the flow fields generated by its beating may be described, on suitable length scales, in terms of geometrically simpler force distributions. In the simplest case, often used in models of synchronization (Vilfan and Jülicher, 2006; Niedermayer et al., 2008; Uchida and Golestanian, 2011), that would be just a single sphere tracing out a closed orbit in space under the action of internal driving forces. The timeaveraged flow around the flagellum would then be approximated by the flow from a point force at a suitable average location. A point force F exerted on a viscous fluid at a location x_{0} produces a velocity field, known as a Stokeslet, of the form (Blake and Chwang, 1973) ${u}_{i}={F}_{j}/8\pi \mu r\left({\delta}_{ij}+{r}_{i}{r}_{j}/{r}^{2}\right)$, where the vector r = x − x_{0}, $r=\left\mathit{r}\right$ and ${\delta}_{ij}$ is the Kronecker delta. A recent study (Drescher et al., 2010) of freely swimming Chlamydomonas cells has shown that the timeaveraged flow field is consistent in its magnitude and topology with a threeStokeslet model (one for each flagellum and one for the cell body).
In our experiments, all flow fields were obtained using particle image velocimetry (PIV) (Raffel et al., 2007). Figure 2A shows the instantaneous flow field at four different times near a single cell, and it is clear that the magnitude (colour) and direction (vector field) of the flow vary during the cycle, as expected from the distinct power and recovery strokes. Examining the timeaveraged velocity field (Figure 2B, obtained by averaging data from four cells), we see that for distances larger than 20 μm from the flagellar tip, both upstream (red) and downstream (blue) components of the flow obey a Stokeslet decay (u ∼ 1/r) (Figure 2C). This trend is maintained over a range consistent with the distances sampled for our twocell experiments (below).
Let us now examine more closely the timedependent flows. Figure 3A and Figure 3—figure supplement 1 show a fit of the instantaneous flow fields of each frame to a Stokeslet form, using the position x_{0} and the magnitude and direction F as fitting parameters. The results of this procedure are illustrated in Figure 3A as the average trajectory ${\overline{\mathit{x}}}_{0}\left(t\right)$ (the closed white curve) executed by the Stokeslet over approximately 10^{3} beats. Figure 3B shows ${\overline{\mathit{x}}}_{0}\left(t\right)$ (solid red line) together with a scatter plot of x_{0}(t) from individual frames (red dots). The black arrows along the average cycle illustrate the position, orientation and relative magnitude of the Stokeslet at evenlyspaced times along the average cycle. Importantly, the orientation of the point force does not coincide with its direction of motion, a feature to be expected given the anisotropic drag on the flagellum. Figure 3C shows the magnitude of the fitted Stokeslet over all beats. The amplitude of this force exhibits very strong periodic variations, and is approximated by $F\left(t\right)/8\pi \mu ={A}_{0}\left(1+{A}_{1}\text{sin}\left(2\pi t/T\right)\right)$ with A_{0} ≃ 1076 μm^{2}/s and A_{1} ≃ 0.56.
This determination of the magnitude of the effective Stokeslet describing the flow field around a cell can be compared with an estimate based on the observed motion of the flagellum itself. Figure 4A shows snapshots of a typical flagellum captured over a full beat cycle, superimposed at 2 ms intervals, together with measured instantaneous velocities along the filament. With resistive force theory (RFT), the results of the tracking procedure are used to derive estimates for the forces produced by the flagellum. First proposed by Gray and Hancock (1955), RFT considers the anisotropic drag experienced by a long rodlike flagellum moving through a viscous fluid, and assumes that each unit segment of the flagellum experiences a local drag that is proportional to its local instantaneous velocity. The force density f along the flagellum is approximated by
which is readily computable from experimental data, where the constants of proportionality, ${C}_{\perp}$ and ${C}_{\Vert}$, are the normal and tangential resistance coefficients respectively. We chose ${C}_{\perp}$ and ${C}_{\Vert}$ according to the classic model of Lighthill (1975), with ${C}_{\perp}=4\pi \mu /\left(\text{ln}\left(0.18\lambda /a\right)+0.5\right)$ and ${C}_{\Vert}=2\pi \mu /\left(\text{ln}\left(0.18\lambda /a\right)0.5\right)$, with an aspect ratio λ/a = 80.
The total instantaneous force F(t) produced by the flagellum is given by ${{\displaystyle \int}}_{0}^{l}\mathit{f}\left(s,t\right)ds$, where l is the total length of the flagellum and s its arclength parameterisation. In Figure 4B we plot the normal (blue) and tangential (red) components of f, for characteristic power and recovery stroke waveforms (solid and dotted lines respectively). To construct a limit cycle representation of the cyclic force variation, we define an effective centreofmass for the flagellum, $\mathit{x}\left(t\right)={{\displaystyle \sum}}_{i}^{N}\left(\left{\mathit{f}}_{i}\right{\mathit{x}}_{i}\left(t\right)\right)/{{\displaystyle \sum}}_{i}^{N}\left(\left{\mathit{f}}_{i}\right\right)$, averaging over all N discretised force vectors f_{i} applied at points x_{i} along the flagellum. Figure 4C depicts the trajectories of integrated force F in this coordinate representation (red arrows). An average limit cycle representation (black arrows) is obtained from measurements taken from ∼100 beats: resultant force directions are seen to vary continuously along the cycle. The RFT result overestimates the force production during the recovery stroke, where the assumption of locality breaks down. It is encouraging to see that the amplitude of this force is similar to the value calculated earlier using Stokeslet fitting, though it should be noted that these results correspond to two different cells.
Two cells
To investigate the effect of hydrodynamic coupling on pairs of flagella, we captured pairs of cells and aligned them so that their flagellar beating planes coincided (Figure 1A). Videos of hydrodynamically interacting flagella were first processed by subtracting a 30 frame running average. Median filtering was undertaken using 3 × 3 pixels^{2} regions. At each cell–cell separation d, we recorded flagellar dynamics over ∼100 s, and extracted flagellar phases ${\varphi}_{\mathrm{1,2}}$ from Poincaré sectioning of the dynamics (Goldstein et al., 2009; Polin et al., 2009) by monitoring the signal in respective interrogation regions (Figure 1B, Figure 1—figure supplement 1), so that the respective flagella passed through precisely once per beat. Recording the passage times between beats allowed reconstruction of the flagellar phase ${\varphi}_{\mathrm{1,2}}$. The timedependent interflagellar phase difference $\text{\Delta}\left(t\right)=\left({\varphi}_{1}{\varphi}_{2}\right)/2\pi $ was used to characterise the synchronization properties of the two cells.
The measured phase difference Δ(t) is shown in Figure 1C for one pair of cells at four different spacings (see Video 1). We measured beat frequencies ${\omega}_{1}$ and ${\omega}_{2}$ for the two flagella in isolation, and define $\delta \omega ={\omega}_{1}{\omega}_{2}$ to be their intrinsic frequency difference. Calling L = d/l the cell–cell separation normalised by the average flagellar length l of each pair, Figure 1C shows that for $L\gtrsim 2$ hydrodynamic coupling is negligible and Δ(t) drifts approximately linearly with time depending on $\delta \omega /\omega $ (8.2% here). For intermediate values of L, the flagella exhibit short periods of synchrony interrupted by brief phase slips. However, when the same cells are brought closer to each other, they phaselock for the entire duration of the experiment. This conclusively demonstrates that robust and extended flagellar synchronization can arise in physically separated cells purely through the action of hydrodynamics. For different pairs of cells (n = 11), a similar behaviour is observed.
Next we examine in detail the experimental time series Δ(t). Consider first the synchronous periods within the full time series of Figure 1C. Fluctuations about the phaselocked states Δ_{0} (Figure 1D) are Gaussian with a variance proportional to L, as seen by rescaling as $\left(\text{\Delta}{\text{\Delta}}_{0}\right)/{L}^{1/2}$ (Figure 1E). Gaussian fluctuations suggest a description of the dynamics of Δ(t) based on a Langevin equation with an effective potential V(Δ) having a quadratic minimum at Δ_{0}. We then write
where $\nu \left(\text{\Delta}\right)=\delta \nu \text{\Delta}+U\left(\text{\Delta}\right)$. The quantity $\delta \nu $ is the intrinsic frequency difference for the two phase oscillators, U an effective potential which has period one in Δ, and $\xi \left(t\right)$ is a Gaussian white noise term satisfying $\langle \xi \left(t\right)\rangle =0$ and $\langle \xi \left(t\right)\xi \left(t\prime \right)\rangle =2{T}_{\text{eff}}\delta \left(tt\prime \right)$, where T_{eff} is an effective ‘temperature’. To leading order $U=\mathit{\u03f5}\text{cos}\left(2\pi \text{\Delta}\right)$, where ε is the interflagellar coupling strength. The observed dependence on L of the distribution of Δ fluctuations is a natural consequence of Equation 2 if $\mathit{\u03f5}\propto 1/L$. We test this scaling below. Intraflagellar biochemical noise leads to stochastic transitions between adjacent minima of the tilted washboard potential ν(Δ) (Goldstein et al., 2009; Polin et al., 2009). For each video, the autocorrelation of Δ is used to extract the model parameters $\left(\mathit{\u03f5},\delta \nu ,{T}_{\text{eff}}\right)$ as described previously (Goldstein et al., 2009; Polin et al., 2009).
Cells aligned so that their power strokes point in the same direction (as in many ciliates) exhibit inphase (IP) synchrony (Δ_{0} ≃ 0), indicating a coupling strength $\mathit{\u03f5}>0$. Rotation of pipette P_{1} (Figure 1B) by 180^{∘} so that the power strokes are opposed (as in the Chlamydomonas breaststroke) changes the sign of the coupling strength and gives rise to antiphase (AP) synchronization (Δ_{0} ≃ 1/2), in agreement with theory (Leptos et al., 2013). Figure 5A depicts the nondimensionalised coupling strength $\kappa =\mathit{\u03f5}/\overline{\omega}$ for all experiments, where $\overline{\omega}$ is the average beat frequency across all experiments for a given pair of cells. The dependence on the interflagellar spacing $\left\kappa \right\propto {L}^{1}$ is consistent with the intrinsic flagellar flow field presented in Figure 2. For both the inphase and antiphase configurations, we fit $\left\kappa \right=k\times {L}^{1}$ finding k_{IP} = 0.016 and k_{AP} = 0.014 respectively. At a given L, IP pairs exhibit on average a marginally stronger coupling than AP ones, possibly due to the fact that flagella in IP are on average closer together than in AP. The average values of the other model parameters are $\langle {T}_{\text{eff}}/\overline{\omega}\rangle =0.005\pm 0.003$ and $\langle \delta \nu /\overline{\omega}\rangle =0.058\pm 0.033$, with $\langle \overline{\omega}\rangle =33.0$ Hz. As a crosscheck, we can estimate directly the effective internal noise from the distribution of beating periods of separated cells, and find $\langle {T}_{\text{eff}}/\overline{\omega}\rangle =0.002$, consistent with the value above.
The average measured flagellar frequency ω for the two cells in each experiment is shown in Figure 5B, nondimensionalised by the average value for each cell ${\overline{\omega}}_{\text{cell}}$ across different spacings. Figure 5C illustrates the measured frequency difference as a function of L. The data exhibit an apparent bifurcation near L = 1, beyond which phase drifting occurs over time. Integration of Equation 2 in the absence of noise yields a predicted value for the observed frequency difference in terms of the model parameters: $\delta \omega /\delta {\omega}_{\text{far}}=\sqrt{1{\left(2\pi \mathit{\u03f5}/\delta \nu \right)}^{2}}$, for $\mathit{\u03f5}\left(L\right)<\delta \nu /2\pi $ and $\delta \omega /\delta {\omega}_{\text{far}}=0$ otherwise. The orange curve in Figure 5C illustrates this prediction, calculated using the average extracted model parameters. In the presence of noise this sharp bifurcation becomes rounded and shifted (Risken, 1989), as shown in green in Figure 5C. It is evident that noise plays an important role in determining the observed location of the bifurcation point.
Waveform characteristics
Although coupling is established purely through hydrodynamic interactions, the process of synchronization hinges on the ability of the flagella to respond differentially to varying external flows. For sufficiently strong coupling, different cells can adopt a common phaselocked frequency through perturbing one another from their intrinsic limit cycles. Indeed, models of coupled flagella involving hydrodynamically coupled semiflexible filaments (Gueron et al., 1997; Guirao and Joanny, 2007; Elgeti and Gompper, 2013; Yang et al., 2008) show a tendency towards metachronal coordination, though the precise role that flexibility plays in facilitating synchrony is unknown. Minimal models in which spheres are driven along flexible trajectories (Brumley et al., 2012; Niedermayer et al., 2008) reveal that deformationinduced changes in the phase speed can facilitate synchrony. However, functional variations in the intrinsic flagellar driving forces could lead to synchrony even for fixed beating patterns (Uchida and Golestanian, 2011, 2012).
Through dynamic tracking (Wan et al., 2014), we followed the evolution of the flagellar waveforms for several thousand consecutive beats. One example is shown in Figure 6A where the extracted waveform is shown at various stages through the beating cycle, overlaid onto logarithmicallyscaled residence time plots. The same pair of flagella is compared at close and far cell–cell separations (7.3 μm and 72.2 μm respectively). In order to characterise flagellar waveform changes as the cells are brought closer together, we define three angles x_{a}, x_{b}, x_{c} (radians) with respect to the cell body axis (Figure 6B). Figure 6C shows the temporal evolution of these angles for the right flagellum, corresponding to the close (red), intermediate (green) and wide (blue) separations. In particular, the most significant difference is observed in the x_{c} component (distal part of the flagellum). Similar results are found for the other cell, indicating that the interaction is mutual. Figure 6—figure supplement 1 and Figure 6—figure supplement 2 illustrate the robustness of these results for multiple cells and different configurations. Taken together, the results in Figure 6 demonstrate that accompanying the robust hydrodynamic phaselocking is a change in the flagellar waveform. For the first time, we have shown by systematically varying the cell–cell spacing that each flagellum can directly alter the beating profile of its neighbour simply through hydrodynamic interactions.
Discussion
Understanding the mechanisms giving rise to robust phaselocking of flagella can be broken down into two distinct components, namely (i) identification of physical or chemical coupling between the flagella and (ii) characterisation of the response of each flagellum subject to these external stimuli. Theoretical studies and experiments suggest that cell body rocking of freely swimming Chlamydomonas can induce synchrony (Friedrich and Jülicher, 2012; Geyer et al., 2013), and experimental investigations of such cells hint that hydrodynamic interactions between the flagella and cell body could be important for locomotion (Kurtuldu et al., 2013). At the same time, the synchronization properties of immobilised Chlamydomonas cells are generally consistent with models in which the flagella interact purely hydrodynamically (Goldstein et al., 2009, 2011; Polin et al., 2009), although the observed antiphase synchronization of the ptx1 mutant has been suggested (Leptos et al., 2013) to implicate intracellular coupling such as through elastic filaments at the basal bodies. In addition, intracellular calcium fluctuations in Chlamydomonas are known to affect the flagellar dynamics (Yoshimura et al., 2003; Leptos et al., 2013), so these studies cannot exclude the possibility that flagellar synchrony is regulated primarily by chemical or other nonmechanical means. Indeed, Machemer showed that membrane voltage affects the metachronal wave direction in Paramecium (Machemer, 1972).
In the present experiment, cells are held with micropipettes in order to preclude all chemical coupling of the two flagella other than the possible advection of molecules by the flow. The fact that the coupling strength for IP and AP pairs is almost identical despite the pronounced difference between the associated flows rules out coupling via chemical means. In these experiments hydrodynamics alone is responsible for the interflagellar coupling. Yet, this may be direct interaction between the flagella or indirect, through some residual motion of the cell bodies. If such motion does play a role in the observed synchronization, it is useful to estimate the effective spring constant associated with angular displacements of the cell. This can be done by estimating the hydrodynamical torques at one cell due to the flagellar beating of a nearby cell and using an experimental bound on the observed angular displacement. The fluid speed u at a distance r from the origin due to a second cell whose flagellum exerts a (point) force of magnitude F varies as u ∼ F/8πμr, and when acting on the flagellum of the first cell will produce a torque scaling as ${C}_{\perp}u{l}^{2}/2$, where l is the length of the flagellum. If we set this equal to the torque kα of a rotational spring, where k is the spring constant and α the angular displacement, we obtain an estimate for k from an upper bound α^{*} for the rotation. Using l = 20 μm, r = 5 μm, and F/8πμ ∼ 2 × 10^{3} μm^{2}/s, we find k > (3 × 10^{2}/α^{*}) pN·μm. A similar argument can be constructed by considering the vorticity at the central cell due to the flow from another, and yields the same result. We have performed experiments of exactly this type, in which we measured the rocking motion of a deflagellated cell placed in the flow field of a cell with a beating flagellum (Video 2), and found α^{*} ≤ 0.01 (about 0.5°). This strongly suggests k > 3 × 10^{4} pN·μm, which, by the calculations of Geyer et al. (2013), is sufficiently large to suggest that rocking does not play a significant role in the synchronization observed in the present experiments.
Two spheres of radius a in an unbounded fluid of viscosity μ, driven along circular trajectories of variable radius (stiffness λ) are able to synchronize their motions through flowinduced changes in their respective phase speeds (Niedermayer et al., 2008). Calculating the spring stiffness λ = R/l^{3} in terms of the flagellar bending rigidity R and length l yields the scaled coupling strength for this dynamical system, ${\kappa}_{\text{spheres}}=\left(27\mu \pi {a}^{2}{l}^{2}\overline{\omega}/2R\right)\times {L}^{1}$. Estimating (Niedermayer et al., 2008) a = 0.1 μm and R = 4 × 10^{−22} Nm^{2}, and using measured values of the other parameters from the present experiment ($\langle \overline{\omega}\rangle =33.0$ Hz and $\langle l\rangle =19.9$ μm), we obtain ${\kappa}_{\text{spheres}}=0.014\times {L}^{1}$. This minimal model, in which synchronization is facilitated through hydrodynamic interactions, compares favourably with the measured flagellar coupling strengths presented in Figure 5.
Conclusions
The experimental study presented in this article reveals unambiguously the importance of hydrodynamics in achieving flagellar synchronization. Physical separation of the cells precludes any form of chemical or direct mechanical coupling, leaving hydrodynamic interactions as the only mechanism through which synchronization can occur. The process of phaselocking is extremely robust, with cells sufficiently close to one another exhibiting uninterrupted synchrony for thousands of consecutive beats. Accompanying this synchrony is a characteristic shift in the flagellar waveform. The extracted interflagellar coupling strength is consistent with hydrodynamic predictions and the measured flow fields generated by individual flagella. Additional experiments were undertaken using a uniflagellar mutant of the unicellular alga Chlamydomonas. Although its flagellum is shorter and its waveform is different to that of Volvox, we also observed hydrodynamic phaselocking in these experiments. Owing to the ubiquity and uniformity in the structure and function of flagella in various eukaryotic species, the results of the present study are expected to generalise to other systems, and may be of significant value for a wide range of theoretical models.
Materials and methods
Cell growth and imaging
Volvox carteri f. nagariensis (strain EVE) were grown axenically in Standard Volvox Medium (SVM) (Kirk and Kirk, 1983) with sterile air bubbling, in a growth chamber (Binder, Germany) set to a cycle of 16 hr light (100 μEm^{−2}s^{−1}, Fluora, OSRAM) at 28^{∘}C and 8 hr dark at 26^{∘}C. Individual biflagellate cells were extracted from Volvox colonies using a cell homogeniser, isolated by centrifugation with Percoll (Fisher, UK), and inserted into a 25 × 25 × 5 mm glass observation chamber filled with fresh SVM. Cells were captured using micropipettes and oriented so that their flagellar beating planes coincided with the focal plane of a Nikon TE2000U inverted microscope. Motorised micromanipulators (Patchstar, Scientifica, UK) and custommade stages facilitated accurate rotation and translation of the cells. The flow field characterisation and pairwise synchronization analyses were imaged using a 40× Plan Fluor objective lens (NA 0.6). A higher magnification 63× Zeiss W PlanApochromat objective lens (NA 1.0) was used to conduct separate experiments for the waveform analysis. For each experiment, we recorded videos with a highspeed video camera (Fastcam SA3, Photron, USA) at 1000 fps under bright field illumination.
One cell
Request a detailed protocolSpatiotemporal analysis of the flow field associated with individual isolated cells was achieved through seeding the fluid with 0.5 μm polystyrene microspheres (Invitrogen, USA) at a volume fraction of 2 × 10^{−4}. We recorded ∼30 s long videos, each one corresponding to approximately 1000 flagellar beats. The timedependent velocity field was reconstructed using an open source particle image velocimetry (Raffel et al., 2007) toolbox for MATLAB (MatPIV).
Two cells
Request a detailed protocolFor each pair of Volvox somatic cells, we investigated the synchronization properties as a function of interflagellar spacing. A number of videos were taken at various separations (varied nonmonotonically). In many cases we also rotated the micropipettes (see Figure 1B) so that the flagella were beating in the same plane but opposite directions. There are two such ‘antiphase’ configurations possible, in which the flagella beat towards and away from one another respectively. Both of these states are referred to as antiphase in the extraction of parameters in Figure 5 and Figure 7.
Additional model parameters
Request a detailed protocolThe stochastic Adler equation was used to model the dynamics of Δ(t) as described in Goldstein et al. (2011). Figure 7A,B show the amplitude C_{0} of the autocorrelation function of Δ and the values of the average synchronous period ${\tau}_{\text{sync}}$. Fluctuations of the phase difference Δ about the synchronized states are well described by Gaussian distributions, with variances C_{0} proportional to the interflagellar spacing L. The coupling strength ε exhibits excellent agreement with the hydrodynamic predictions. Figure 7C,D show the dependence of the effective temperature ${T}_{\text{eff}}/\overline{\omega}$ and intrinsic frequency difference $\delta \nu /\overline{\omega}$ as a function of L = d/l for every pair of flagella measured.
Proximity to pipettes
Request a detailed protocolIn order to study the dynamics of hydrodynamically coupled flagella, the two cells were held using orthogonallypositioned glass pipettes. This geometry allowed us to investigate both inphase and antiphase configurations for the same pair of cells, through the simple rotation of one pipette. At the same time, however, this meant that the two cells were held from different directions with respect to their flagella, and that one of the two pipettes was oriented along the direction of the flagellar power stroke, which is the main flow direction. This can cause two problems. Firstly, the flow field of a cell held by the side could be significantly different from that presented in Figure 2. Secondly, the holding pipettes could distort the scaling of the flagellar flow with cell–cell separation from the ∼1/r scaling presented in Figure 2C. We investigated these problems with the series of experiments shown in Figure 8. One cell was held at its posterior pole by a pipette (Figure 8A) and the flow field measured. A second micropipette was then moved progressively closer, eventually to the point of contact with the cell (Figure 8D). It is clear that the second pipette affects the flow, but mostly in the region between the two pipettes.
Let us consider the region upstream of the cell (above the cell in Figure 8). For a cell held from the side, this is the region where the other cell will be. Here the flow is only minimally affected, with an average relative change between Figure 8D,A below 8%. A large contribution is represented simply by a ∼7% decrease in flow speed. Taking this decrease into account, the average relative change is about 5%. As a result, these experiments allow us to consider the flow generated by a cell held from the side as identical to that generated by a cell held from the back, at least in the region of interest to our experiments. By comparing the flows for different positions of the second pipette, we can also quantify its effect on the flow field that would be experienced by the second cell. For each configuration of pipettes, this can be estimated as the relative difference between the unperturbed and the perturbed flows in the region where the flagella of the second cell would be, here considered to be a 10 × 10 μm^{2} region 20 μm to the left of the tip of the incoming pipette. The difference ranges from ∼5% to ∼10% and ∼13% for Figure 8B–D respectively (in the last case we choose a position approximately 10 μm below and 20 μm to the left of the pipette tip). These represent the typical error contributions from neglecting, as we have done in the text, the influence of the pipettes on the flows generated by the cells.
Minimal model with variable forcing
Request a detailed protocolWe used the Stokeslet approximation to the flow field of an isolated cell in Figure 3, to test the effect of force modulation on synchronization within the class of minimal models which abstract the beating flagellum as a sphere driven along a closed orbit (Niedermayer et al., 2008; Uchida and Golestanian, 2011, 2012). We simulated two spheres of radius a = 0.75 μm in an unbounded fluid of viscosity μ = 10^{−3} Pa·s, driven along coplanar circular orbits of radius r_{0} = 8 μm by a force $F\left(\varphi \right)/8\pi \mu ={A}_{0}\left(1+{A}_{1}\text{sin}\left(\nu \varphi +{\varphi}_{0}\right)\right)$ tangential to the orbit, with A_{0} = 1076 μm^{2}/s and A_{1} = 0.56. Notice that this corresponds to assuming that the point forces in Figure 3B are tangential to the cycle. The value a = 0.75 μm was chosen to ensure that the orbital frequency matched the mean value observed experimentally. The orbits were separated by d = 20 μm and had a radial stiffness with spring constant λ. The limit λ→∞ corresponds to rigid prescribed trajectories (holonomic constraint). For each value of $\lambda \in \left\{1\text{\hspace{0.17em}pN}/\mu \text{m},5\text{\hspace{0.17em}pN}/\mu \text{m},\infty \right\}$ we ran five sets of simulations, corresponding to $\nu \in \left\{\mathrm{0,1,2}\right\}$ and ${\varphi}_{0}\in \left\{0,\pi /2\right\}$. Choosing ν = 2 is equivalent to modulating the driving force with the experimental amplitude but at a frequency double the experimental one. Although this is not what we observed, it is still interesting to consider, since in this configuration it is the frequency that contributes most to synchronization through force modulation.
As a consequence of the phasedependent driving force, the geometric phase ${\varphi}_{i}$ of an individual isolated oscillator does not evolve at a constant rate in time. We thus chose to rescale the phase $\text{\Phi}=\text{\Phi}\left(\varphi \right)$ so that in the absence of hydrodynamic interactions, $\dot{\text{\Phi}}=2\pi /T=\text{constant}$. Both the geometric phase difference $\delta ={\varphi}_{1}{\varphi}_{2}$ (thin curves) and its rescaled value ${\delta}_{\text{rescaled}}={\text{\Phi}}_{1}{\text{\Phi}}_{2}$ (thick curves) are shown for each simulation in Figure 9. These results show clearly that within the boundary of the model we are considering, the two oscillators synchronize through a coupling between hydrodynamic stresses and orbit compliance (Niedermayer et al., 2008) with no noticeable effect from force modulation.
Repeating the simulations with a stiffness derived from the flagellar bending rigidity as in the main text, λ = 0.05 pN/μm, radius a = 0.1 μm, and reducing the force amplitude to A_{0} = 143 μm^{2}/s to keep the revolution frequency at the experimental value, yields the results in Figure 10. Again, the synchronization is achieved only through interaction between hydrodynamic stresses and orbit compliance.
References

Fundamental singularities of viscous flowJournal of Engineering Mathematics 8:23–29.https://doi.org/10.1007/BF02353701

Hydrodynamic synchronization of light driven microrotorsPhysical Review Letters 109:34104.https://doi.org/10.1103/PhysRevLett.109.034104

Direct measurement of the flow field around swimming microorganismsPhysical Review Letters 105:168101.https://doi.org/10.1103/PhysRevLett.105.168101

Emergence of metachronal waves in cilia arraysProceedings of the National Academy of Sciences of the United States of America 110:4470–4475.https://doi.org/10.1073/pnas.1218869110

External mechanical control of the timing of bend initiation in sea urchin sperm flagellaCell motility and the cytoskeleton 14:416–423.https://doi.org/10.1002/cm.970140311

Flagellar synchronization independent of hydrodynamic interactionsPhysical Review Letters 109:138102.https://doi.org/10.1103/PhysRevLett.109.138102

Cellbody rocking is a dominant mechanism for flagellar synchronization in a swimming algaProceedings of the National Academy of Sciences of the United States of America 110:18058–18063.https://doi.org/10.1073/pnas.1300895110

Noise and synchronization in pairs of beating eukaryotic flagellaPhysical Review Letters 103:168103.https://doi.org/10.1103/PhysRevLett.103.168103

Emergence of synchronized beating during the regrowth of eukaryotic flagellaPhysical Review Letters 107:148103.https://doi.org/10.1103/PhysRevLett.107.148103

Cilia internal mechanism and metachronal coordination as the result of hydrodynamical couplingProceedings of the National Academy of Sciences of the United States of America 94:6001–6006.https://doi.org/10.1073/pnas.94.12.6001

Spontaneous creation of macroscopic flow and metachronal waves in an array of ciliaBiophysical Journal 92:1900–1917.https://doi.org/10.1529/biophysj.106.084897

Hydrodynamic synchronization of colloidal oscillatorsProceedings of the National Academy of Sciences of the United States of America 107:7669–7673.https://doi.org/10.1073/pnas.0912455107

Flagellar waveform dynamics of freely swimming algal cellsPhysical Review E 88:13015.https://doi.org/10.1103/PhysRevE.88.013015

Antiphase synchronization in a flagellardominance mutant of ChlamydomonasPhysical Review Letters 111:158101.https://doi.org/10.1103/PhysRevLett.111.158101

The reproductive significance of human fallopian tube ciliaHuman Reproduction Update 12:363–372.https://doi.org/10.1093/humupd/dml012

Ciliary activity and the origin of metachrony in Paramecium: effects of increased viscosityThe Journal of Experimental Biology 57:239–259.

The control and synchronization of flagellar movementProceedings of the Royal Society B: Biological Sciences 158:88–104.https://doi.org/10.1098/rspb.1963.0036

Mechanical stimulation of starfish sperm flagellaJournal of Experimental Biology 65:401–413.

The FokkerPlanck equation: methods of solution and applicationsThe FokkerPlanck equation: methods of solution and applications, Springer, 2nd edition, chapter 11.

Analysis of the swimming of microscopic organismsProceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences 209:447–461.https://doi.org/10.1098/rspa.1951.0218

Generic conditions for hydrodynamic synchronizationPhysical Review Letters 106:58104.https://doi.org/10.1103/PhysRevLett.106.058104

Hydrodynamic synchronization between objects with cyclic rigid trajectoriesThe European Physical Journal E 35:1–14.https://doi.org/10.1140/epje/i2012121355

Hydrodynamic flow patterns and synchronization of beating ciliaPhysical Review Letters 96:58102.https://doi.org/10.1103/PhysRevLett.96.058102

Lag, lock, synch, slip: the many ‘phases’ of coupled flagellaJournal of the Royal Society Interface 11:20131160.https://doi.org/10.1098/rsif.2013.1160

Gravitaxis in Chlamydomonas reinhardtii studied with novel mutantsPlant & Cell Physiology 44:1112–1118.https://doi.org/10.1093/pcp/pcg134
Decision letter

W James NelsonReviewing Editor; Stanford University, United States
eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.
Thank you for sending your work entitled “Flagellar Synchronization Through Direct Hydrodynamic Interactions” for consideration at eLife. Your article has been favorably evaluated by Randy Schekman (Senior editor), a Reviewing editor, and 3 reviewers.
The Reviewing editor and the reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission.
The manuscript by Brumley et al. reveals mechanisms of beat synchronization in pairs of flagella. This is biologically relevant as pairs of flagella serve as a minimal model system for synchronization in collections of flagella such as ciliary carpets in mammalian airways, oviduct epithelium, etc. whose coordinated beating in metachronal waves is important for efficient fluid transport. Secondly, this minimal system allows to link biology and physics in a quantitative way to generate mechanistic insight.
The work reports a key experiment; in fact this experiment should have been conducted long ago. By use of two spatially separated cells, any means of chemical coupling or direct mechanical coupling is effectively ruled out. This sets this particular experiment apart from previous studies.
The 3 reviewers and the Reviewing editor agreed that the manuscript is wellcrafted and contributes to a timely topic. However, the authors should address the following significant points:
1) The analysis of waveform is elegant, but there is a lack of statistics or a way to compare results from multiple cell pairs in Figure 4, which just shows one example of a pair of cells. At the very least, several other examples need to be shown, perhaps as figure supplements, to show that the effect of distance on waveform isn't just some idiosyncrasy of this particular cell pair. What would be more ideal is if there is some way to parameterize the limit cycles so that their shape could be described by a single number or some small number of parameters. The cycles look teardrop shaped so maybe one could use a Pyriform (“pear shaped quartic”), or the Lemniscate of Bernoulli. Regardless of how it is accomplished, having some ability to compare results from different cells statistically is very important.
2) It is unclear whether there is a true bifurcation in the stochastic case. If delta_omega is defined as the expectation value of dphi1/dtdphi2/dt, then delta_omega likely varies smoothly with epsilon, although it may indeed display a steep change beyond a certain value. A more stringent terminology is required.
3) Were different stages used for the two micropipettes to rule out any direct mechanical coupling between them? Otherwise, there is a concern that potential oscillations of the micropipettes caused by the flagellar oscillations could by relayed by such a coupling. One cilium is certainly too weak to generate such a strong vibration to excite the manipulator and then back to the other cilium. There is of course the potential that the stage itself is vibrating and then causes both of them to oscillate. Sperm cells have been held on vibrating micro needles, and vibrating micro needles were brought close to flagella (Eshel 1989 and Okuno 1976); from there the authors could potentially estimate how strongly the manipulator would have to vibrate.
4) Referencing previous published work: the authors should do a better job in giving credit to previous relevant work (which goes far back and another detailed literature search is advised). Giving corresponding credit to this previous work does not diminish the relevance of the present paper given its experimental execution and analysis.
a) A clear statement in the paper should be made that (pure) hydrodynamic synchronization between cilia/flagella has been demonstrated previously. The authors cite Taylor 1951 in a manner that suggests that Taylor made pure theoretical arguments (which is actually true), but Taylor himself cites earlier experiments that describe the synchronization of sperm flagella.
b) It would be worth commenting on the dominating synchronization mechanism of the flagellar pair in single clamped Chlamydomonas One group (Goldstein et al. PRL 2009) proposed direct hydrodynamic interactions between the two flagella of Chlamydomonas as main synchronization mechanism, while others have argued that this synchronization mechanism might be too weak to overcome flagellar noise and that instead a residual motion of elastically clamped cells could accelerate flagellar synchronization: VF Geyer, F Jülicher, J Howard, BM Friedrich, PNAS 110, 18058–18063 (2013).
c) Riedel 2005 showed synchronization of sperm flagella, with a similar conclusion on the hydrodynamic origin of the coordination.
d) There have been also multiple papers where sperm cells have been held on vibrating micro needles, or vibrating micro needles were brought close to flagella, see for example Eshel 1989 and Okuno 1976 – and there are likely more.
e) The cited reference Geyer et al 2013 combines theory and experiments, which should be mentioned. Also, the discussion of elastic contributions to flagellar synchronization should make reference to this cited reference, which first introduced alternative mechanisms of flagellar synchronization independent of direct hydrodynamic interactions.
https://doi.org/10.7554/eLife.02750.019Author response
1) The analysis of waveform is elegant, but there is a lack of statistics or a way to compare results from multiple cell pairs in Figure 4, which just shows one example of a pair of cells. At the very least, several other examples need to be shown, perhaps as figure supplements, to show that the effect of distance on waveform isn't just some idiosyncrasy of this particular cell pair. What would be more ideal is if there is some way to parameterize the limit cycles so that their shape could be described by a single number or some small number of parameters. The cycles look teardrop shaped so maybe one could use a Pyriform (“pear shaped quartic“), or the Lemniscate of Bernoulli. Regardless of how it is accomplished, having some ability to compare results from different cells statistically is very important.
We have now expanded Figure 6 (previously Figure 4) to include several figure supplements that illustrate the robustness of the results on waveform deformations.
We agree with the referees that the issue of limit cycle parametrisation is interesting, and indeed we had already thought about it. Ideally, we would like to project out limit cycles on some basis set of curves, which could then be used to quantify the limit cycles’ deformations. We believe that it is the extent of these deformations that should be compared between cells, rather than the limit cycles themselves. Unfortunately, we did not find any convincing way to put this idea into practice. In particular it was not obvious to us what would constitute a good metric in this space of curves. We decided, therefore, to present only the raw deformations. It is our hope that with time these results will inspire the development of a technique to quantify the deformations.
2) It is unclear whether there is a true bifurcation in the stochastic case. If delta_omega is defined as the expectation value of dphi1/dtdphi2/dt, then delta_omega likely varies smoothly with epsilon, although it may indeed display a steep change beyond a certain value. A more stringent terminology is required.
In the presence of noise the bifurcation is indeed rounded, although for the level of noise in our experiments the predicted rounding is small. We have now made this clear in the text.
3) Were different stages used for the two micropipettes to rule out any direct mechanical coupling between them? Otherwise, there is a concern that potential oscillations of the micropipettes caused by the flagellar oscillations could by relayed by such a coupling. One cilium is certainly too weak to generate such a strong vibration to excite the manipulator and then back to the other cilium. There is of course the potential that the stage itself is vibrating and then causes both of them to oscillate. Sperm cells have been held on vibrating micro needles, and vibrating micro needles were brought close to flagella (Eshel 1989 and Okuno 1976); from there the authors could potentially estimate how strongly the manipulator would have to vibrate.
The microscope used in these studies sits on a very high quality floating vibration isolation table in the quietest room of our microscopy suite, with the room air supply silenced and delivered through tiny perforations in the ceiling tiles, preventing any vibrationinducing jets of air. The two micropipettes were held on physically separated supports in either of two ways. In one, each micropipette is mounted on a separate multiaxis translation stage supported on a vibrationallydamped rod attached to the optical table but with having no physical contact with the microscope. In the second approach, one of those micropipettes is instead supported on a miniature translation stage mounted directly on the microscope body, while the other is still mounted on a separate manipulator. With all these precautions we are very confident the observed flagellar dynamics is indeed intrinsic to the cells. We thank the referee for highlighting the earlier works on sperm cells and vibrating micro needles, which we have now referenced in the introduction.
4) Referencing previous published work: The authors should do a better job in giving credit to previous relevant work (which goes far back and another detailed literature search is advised). Giving corresponding credit to this previous work does not diminish the relevance of the present paper given its experimental execution and analysis.
a) A clear statement in the paper should be made that (pure) hydrodynamic synchronization between cilia/flagella has been demonstrated previously. The authors cite Taylor 1951 in a manner that suggests that Taylor made pure theoretical arguments (which is actually true), but Taylor himself cites earlier experiments that describe the synchronization of sperm flagella.
Thank you for this comment. We are aware that Taylor’s work was indeed motivated by an observation of Rothschild that the flagella of closelyseparated bull spermatozoa tend to be synchronized, and have now included this reference in the Introduction. We have further taken this opportunity to expand the Discussion in the Introduction about hydrodynamic mechanisms. It is important to note that neither Rothschild’s observations nor those of the more recent work we cite provided unequivocal proof of purely hydrodynamic synchronization. For example, a mechanism akin to cell body rocking may have played a role there as well.
b) It would be worth commenting on the dominating synchronization mechanism of the flagellar pair in single clamped Chlamydomonas One group (Goldstein et al. PRL 2009) proposed direct hydrodynamic interactions between the two flagella of Chlamydomonas as main synchronization mechanism, while others have argued that this synchronization mechanism might be too weak to overcome flagellar noise and that instead a residual motion of elastically clamped cells could accelerate flagellar synchronization: VF Geyer, F Jülicher, J Howard, BM Friedrich, PNAS 110, 18058–18063 (2013).
We have modified the Discussion as requested.
c) Riedel 2005 showed synchronization of sperm flagella, with a similar conclusion on the hydrodynamic origin of the coordination.
We have now included this reference in the Introduction.
d) There have been also multiple papers where sperm cells have been held on vibrating micro needles, or vibrating micro needles were brought close to flagella, see for example Eshel 1989 and Okuno 1976 – and there are likely more.
These references have also been included in the revised Introduction.
e) The cited reference Geyer et al 2013 combines theory and experiments, which should be mentioned. Also, the discussion of elastic contributions to flagellar synchronization should make reference to this cited reference, which first introduced alternative mechanisms of flagellar synchronization independent of direct hydrodynamic interactions.
Done.
https://doi.org/10.7554/eLife.02750.020Article and author information
Author details
Funding
European Research Council (Advanced Investigator Grant 247333)
 Douglas R Brumley
 Kirsty Y Wan
 Marco Polin
 Raymond E Goldstein
Wellcome Trust (Senior Investigator Award)
 Douglas R Brumley
 Kirsty Y Wan
 Raymond E Goldstein
Engineering and Physical Sciences Research Council
 Kirsty Y Wan
 Marco Polin
 Raymond E Goldstein
Human Frontier Science Program
 Douglas R Brumley
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Reviewing Editor
 W James Nelson, Stanford University, United States
Version history
 Received: March 11, 2014
 Accepted: June 24, 2014
 Version of Record published: July 29, 2014 (version 1)
Copyright
© 2014, Brumley 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.
Metrics

 4,539
 Page views

 744
 Downloads

 161
 Citations
Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading

 Physics of Living Systems
Biphasic responses are encountered at all levels in biological systems. At the cellular level, biphasic doseresponses are widely encountered in cell signalling and posttranslational modification systems and represent safeguards against overactivation or overexpression of species. In this paper we provide a unified theoretical synthesis of biphasic responses in cell signalling systems, by assessing signalling systems ranging from basic biochemical building blocks to canonical network structures to wellcharacterized exemplars on one hand, and examining different types of doses on the other. By using analytical and computational approaches applied to a range of systems across levels (described by broadly employed models) we reveal (i) design principles enabling the presence of biphasic responses, including in almost all instances, an explicit characterization of the parameter space (ii) structural factors which preclude the possibility of biphasic responses (iii) different combinations of the presence or absence of enzymebiphasic and substratebiphasic responses, representing safeguards against overactivation and overexpression respectively (iv) the possibility of broadly robust biphasic responses (v) the complete alteration of signalling behaviour in a network due to biphasic interactions between species (biphasic regulation) (vi) the propensity of different coexisting biphasic responses in the Erk signalling network. These results both individually and in totality have a number of important consequences for systems and synthetic biology.

 Computational and Systems Biology
 Physics of Living Systems
Diabetes is caused by the inability of electrically coupled, functionally heterogeneous cells within the pancreatic islet to provide adequate insulin secretion. Functional networks have been used to represent synchronized oscillatory [Ca^{2+}] dynamics and to study cell subpopulations, which play an important role in driving islet function. The mechanism by which highly synchronized cell subpopulations drive islet function is unclear. We used experimental and computational techniques to investigate the relationship between functional networks, structural (gapjunction) networks, and intrinsic cell dynamics in slow and fast oscillating islets. Highly synchronized subpopulations in the functional network were differentiated by intrinsic dynamics, including metabolic activity and K_{ATP} channel conductance, more than structural coupling. Consistent with this, intrinsic dynamics were more predictive of high synchronization in the islet functional network as compared to high levels of structural coupling. Finally, dysfunction of gap junctions, which can occur in diabetes, caused decreases in the efficiency and clustering of the functional network. These results indicate that intrinsic dynamics rather than structure drive connections in the functional network and highly synchronized subpopulations, but gap junctions are still essential for overall network efficiency. These findings deepen our interpretation of functional networks and the formation of functional subpopulations in dynamic tissues such as the islet.