In the retina, the visual signal begins as graded changes in membrane potential in photoreceptors, but the output is then delivered by ganglion cells in the form of spikes. Where does the conversion from analogue to digital signaling begin? To reach the inner retina, the visual signal travels through bipolar cells, and it has long been assumed that these are purely graded neurons¹. Although calcium spikes have been recorded in one class of depolarizing bipolar cell in slices of goldfish retina, it has not been clear whether this is an artifact of the preparation^{2, 3}. To investigate whether retinal bipolar cells can encode light with spikes, we made an in vivo investigation by imaging presynaptic calcium in zebrafish expressing the reporter protein SyGCaMP2^{4, 5}.

SyGCaMP2 detected fast calcium transients in bipolar cell terminals, both spontaneously in the dark (Fig. 1) and modulated by light (Fig. 2). The amplitudes and kinetics of these calcium transients were strikingly constant within many terminals (Figs. 1c–d and Fig. 2c–h). The time-course of calcium influx was judged from the time-derivative of the calcium transient and had a width at half-maximum of 50–80 ms (Fig. 1e). This pulse of calcium influx was immediately followed by an exponential decay with τ = 1.2 ± 0.6 s (24 terminals). A similar time-constant of calcium decay has been measured in the synaptic terminal of bipolar cells using synthetic calcium dyes responding to brief (20 ms) depolarizations^{6, 7}. It therefore seems likely that fast presynaptic calcium transients observed in vivo are generated by stereotyped and transient depolarizations. These voltage signals are likely to be calcium spikes: in bipolar cells, these can be generated by voltage-dependent calcium channels in the synaptic terminal which also control neurotransmitter release^{2, 3, 8-11}.

Fast calcium transients were detected by thresholding the derivative of the SyGCaMP2 signal (the choice of this algorithm is discussed in Supplementary Information). In a sample of 1,008 terminals in 9 different fish, 65% generated spikes over an observation period of 60 s. To test whether presynaptic spikes in bipolar cells might be used to encode visual information, we recorded SyGCaMP2 signals in response to a full-field stimulus consisting of a light step followed by modulation at 2.5 Hz. In the 35% of terminals in which we could not detect spikes, slow sustained changes in calcium were observed with variable kinetics and polarities, reflecting ON and OFF cells responding to the stimulus in different ways (Fig. 2a and b). In the 65% of terminals generating fast spike-like calcium transients, these were modulated in a variety of ways. We observed both transient and sustained ON terminals (Fig. 2c), sustained OFF (Fig. 2d) and transient ON terminals responding especially strongly to temporal contrast at 2.5 Hz (Fig. 2e). In other terminals, spikes were not clearly modulated by this stimulus, although we cannot rule out the possibility that other stimuli might have been effective (Fig. 2f). Thus ON and OFF bipolar cells with different kinetics have the ability to generate spikes modulated by light.

It was often possible to detect small and slow changes in baseline calcium coincident with changes in spike frequency (e.g. Fig. 2c and lower traces in Fig. 2d). Further examples of the relation between graded and spike-like calcium signals are shown in Fig. 2g and h. In some ON cells, spikes could occur in darkness, when resting calcium was low reflecting a hyperpolarized state, and in light, when mean calcium levels were higher reflecting depolarization (Fig. 2g, upper). In some OFF cells, spiking occurred continuously in the dark but switched off completely in the light, coincident with a fall in baseline calcium (Fig. 2g, lower). Strikingly, the same terminal could sometimes switch between generating sustained signals and generating spikes (Fig. 2h), reminiscent of the bistable membrane potential observed in isolated bipolar cells^{8, 10}. These results indicate that bipolar cells in the retina of zebrafish respond to visual stimulation with a combination of graded and spiking signals.

These results demonstrate that fast presynaptic calcium transients, likely generated by calcium spikes, are a major feature of signal transmission from a number of functionally distinct types of bipolar cell in the retina of live zebrafish. Thus bipolar cells in the retina of zebrafish respond to visual stimulation with a combination of graded and spiking signals. The functional design of the retina is strongly conserved across vertebrates, so it is worth investigating whether bipolar cells in other species are also capable of encoding a visual stimulus with spikes as well as graded voltage signals.