Our results demonstrate that cannabinoids modulate the transmission of visual-evoked responses by ON-S RGCs. Moreover, our data shows that in the retina there is a functional cannabinergic system that modulates the strength of the transmission of visual signals by ON-S RGCs as well as their spatial tuning. Thus, endocannabinoid signaling in the retina seems to play a similar role in the modulation of the strength of synaptic transmission as that described in other areas of the CNS (Chevaleyre et al., 2006).

The reduction in spontaneous firing rate induced by the cannabinoid receptor agonist WIN55212-2 is consistent with previous reports showing that WIN55212-2 reduced the frequency of excitatory and inhibitory spontaneous synaptic currents in mouse (Middleton and Protti, 2011) and rat (Wang et al., 2016) RGCs. The observed reduction in firing rate suggests that WIN55212-2 produced a shift in the balance of excitation and inhibition, more strongly affecting the excitatory input onto ON-RGCs. A recent report found that WIN55212-2 did not have any effect on spontaneous firing rate in rat RGCs (Jiang et al., 2013). Their recording conditions, however, differed from our study in that recordings were obtained indistinctly from different types of RGCs in retinal slices in light-adapted conditions. Interestingly, we found that the CB1R inverse agonist AM251 showed the opposite effect leading to an increase in firing rate. This is, as far as we know, the first physiological evidence of basal cannabinergic activity in the mammalian retina that modulates RGC output to the rest of the brain. Assuming that the retinal cannabinergic system functions in a similar way to the endocannabinoid system in other brain areas, it can be postulated that an endocannabinoid “tone” is constitutively active in proportion to the activity of ON-S RGCs and that endocannabinoids travel retrogradely to activate cannabinoid receptors on amacrine and bipolar cells and lower their release probability. Alternatively, the effects of AM251 can be explained by constitutive activity of cannabinoid receptors independently of their activation by endocannabinoids. Although this is a possible scenario, a currently accepted interpretation of the effects of inverse agonists is that they act by blocking endogenously produced endocannabinoids that provide autocrine or paracrine stimulation of CB1 receptors, giving the appearance of constitutive activity (Howlett et al., 2011). Experiments blocking the synthetic enzymes of endocannabinoids (Farkas et al., 2010) or chelating postsynaptic calcium (Hentges et al., 2005) could further distinguish whether the inverse agonist acts by blocking a constitutively active receptor or by displacing an endogenous cannabinoid that provides a basal tone.

Although the presence of an endogenous cannabinoid tone in the mammalian retina suggests a role in retinal processing, its physiological effects on the retinal circuit had not been described thus far other than at the cellular level using strong stimulation protocols, which are not regarded as physiological (Diana and Marty, 2004). Wang et al. (2016) demonstrated that depolarization-induced suppression of inhibition, a well described short-term plasticity phenomenon in cerebellum and hippocampus, also takes place in RGCs. Application of depolarizing pulses to RGCs led to a suppression of their mIPSCs and this effect was eliminated by a CB1R antagonist, suggesting that eCBs are indeed released from RGCs in an activity-dependent manner. We probed the effects of cannabinoid receptor modulation on the response strength of ON-S RGCs and showed that the reduction in peak spike response induced by WIN55212-2 was the result of a reduction in the amplitude of the peak light-evoked postsynaptic potential. This effect may be due to: (1) a decrease in excitation, (2) an increase in inhibition, (3) a decrease in excitation and an increase in inhibition or (4) a decrease in excitation and a decrease in inhibition. Conductance analysis revealed that it was in fact owing to a greater decrease in the strength of the light-evoked excitatory synaptic inputs in relation to a reduction in inhibitory inputs, thus leading to an overall decrease in net excitation. Interestingly, a recent study on amphibian RGCs also found that cannabinoid receptor activation produced a reduction in visually evoked excitatory and inhibitory synaptic inputs in amphibian RGCs (Miraucourt et al., 2016). However, in contrast to what we found, this study, also reported that activation of CB1Rs hyperpolarized resting potential by inhibiting the Na⁺-K⁺-2Cl^- co-transporter 1 (NKCC1); hyperpolarization would lead to removal of inactivation of voltage-gated sodium channels and therefore enhance RGC excitability. Although we cannot rule out a direct effect of cannabinoids on the Na⁺-K⁺-2Cl^- co-transporter 1 in RGCs, if WIN55212-2 were to induce a shift of -3.5 mV in the reversal potential of chloride (as reported by Miraucourt et al., 2016), modeling of such a change predicts that it would result in an increase of only 2.3% in G_exc and a reduction of 5% in G_inh. Such changes in G_exc and Ginh would not only be in opposite directions, contrary to our findings, but their magnitude would be significantly smaller compared to the blocking effects of WIN55212-2 on synaptic conductances that we found (G_exc = 39 ± 15% and G_inh = 23 ± 17%). Moreover, no significant labeling for NKCC1 has been detected in the IPL and the ganglion cell layer in the adult mouse retina (Li et al., 2008). Thus, this discrepancy in the effect of cannabinoid receptor activation on membrane potential and cell excitability may be due to different mechanisms of action operating in these two species. The WIN55212-2-induced reduction of the inhibitory conductance can be explained by a direct effect on amacrine cells as well as indirectly due to a reduction of the excitatory inputs that drive feedforward inhibition from amacrine cells onto ON-RGCs (Murphy and Rieke, 2006). The mechanism underlying the spatial modulation of the receptive field profile by cannabinoids seems to involve fine tuning of the strength of these two neurotransmitters of opposing effects in different synapses. WIN55212-2 decreased the degree of surround inhibition of excitatory but not of inhibitory inputs. This decrease in surround inhibition of the excitatory conductance can be explained if the excitatory inputs that drive the amacrine cells that provide presynaptic inhibition to ON bipolar cells are stronger than those involved in feedforward inhibition or if cannabinoid receptors are differentially distributed in these subsets of amacrine cells. Signal transmission through such networks, however, is the most parsimonious explanation as RGCs integrate inputs that are the result of signal processing in the inner and outer plexiform layers and most synapses involved in retinal processing can potentially be modulated by cannabinoids (Bouskila et al., 2016). Further research using genetic labeling, cell ablation and/or optogenetic techniques such as those employed in recent studies of the retinal circuit to target specific bipolar and amacrine cell types (Lee et al., 2011; Jacoby et al., 2015; Tang et al., 2015; Tien et al., 2016) may help dissect more precisely the target/s of (endo) cannabinoids in the signaling pathways to ON-S RGCs.

The CB1R inverse agonist AM251 caused a significant increase in the peak spike response and in surround inhibition demonstrating that, in dark-adapted conditions, ON-S RGCs are modulated by endocannabinoids. Given the non-linear relationship between amplitude of postsynaptic potentials and spike response, the increase in amplitude of the light-evoked postsynaptic potential, although it was found to be non-significant, is likely to be responsible for the significant increase in peak spike response. Conductance analysis showed that the magnitude of the peak excitatory conductance was increased to a greater extent than that of the peak inhibitory conductance. Whilst the increase in excitation indicates a stronger bipolar cell output, the increase in inhibition may originate from a direct effect on presynaptic cannabinoid receptors in amacrine cells and/or a stronger excitatory drive to amacrine cells that provide feedforward inhibition. In the presence of AM251 the inhibitory conductance was less inhibited by surround stimulation, leading to stronger inhibitory inputs from the surround which in turn would cause a sharper tuning of the spike response. This increase in the strength of inhibitory inputs under surround illumination is likely to arise from an increase in the strength of the excitatory inputs that drive inhibitory amacrine cells that provide direct inhibition to ON-S RGCs. Surprisingly, the spatial tuning of the excitatory conductance was not modified by AM251 as would be expected from its enhancing effect on the magnitude of inhibition in response to surround stimulation. This, as previously discussed, suggests that amacrine cells providing presynaptic inhibition to the bipolar cells that drive ON-S RGCs are likely to be different from those involved in feedforward inhibition, therefore a reduction in excitation will impact these forms of inhibition in different ways.

The receptive field of RGCs at high luminance levels are sharply tuned and have a strong inhibitory surround whilst at low luminance levels their spatial tuning is broader as a consequence of weaker inhibitory surround mechanisms (Derrington and Lennie, 1982; Shapley and Enroth-Cugell, 1984; Farrow et al., 2013). Studies on amphibian and fish retina suggested that the cannabinoid system is involved in modulating retinal sensitivity under different luminance conditions, in dark and/or light adaptation and in contrast adaptation (Fan and Yazulla, 2004, 2005; Yazulla, 2008; Miraucourt et al., 2016). Moreover, cannabinoids were shown to have an antagonistic effect to dopamine, a transmitter involved in the switch from rod to cone mediated vision, in cones and bipolar cells of the goldfish retina (Fan and Yazulla, 2005) and in the mammalian retina cannabinoids were shown to suppress dopamine release (Schlicker et al., 1996; Weber and Schlicker, 2001). Our findings, that activation of cannabinoid receptors reduces response strength and surround inhibition and that reduction of endocannabinoid activity leads to stronger responses and sharper spatial tuning in ON-S RGCs, are consistent with the luminance-dependent changes in gain control and receptive field properties of RGCs and with the postulated role of cannabinoids as a “dark signal” (Yazulla, 2008; Farrow et al., 2013). It remains to be elucidated, however, under what luminance and physiological conditions the cannabinoid system is active in the retina and whether or not the eCB system is involved in plasticity phenomena such as contrast adaptation or dark-adaptation in the retina.

Morphological and functional expression of the endocannabinoid system has also been demonstrated in other areas of the visual pathway, such as the dorsal lateral geniculate nucleus (dLGN) and visual cortices (V1 and V2) in rodents and primates (Eggan and Lewis, 2006; Dasilva et al., 2012; Yoneda et al., 2013; Abbas Farishta et al., 2015; Javadi et al., 2015). Visual-evoked responses from rat dLGN neurons show two populations of cells that respond differentially to cannabinoids: the majority (72%) of dLGN neurons are inhibited by cannabinoid agonists, an effect that is prevented by AM251, whilst the remaining 28% are stimulated (Dasilva et al., 2012). At the level of visual cortex, the role of the endocannabinoid system in the development of GABAergic neurotransmission (Jiang et al., 2010) and ocular dominance plasticity (Liu et al., 2008) in rodents is very well documented. Cannabinoid agonists have also been reported to modulate visual responses in the primate primary and secondary visual cortices by decreasing EEG power, LFP power and coherence whereas single cell responses show modulation of their temporal dynamics (Ohiorhenuan et al., 2014). This suggests that the cannabinoid system exerts its effects at different levels of the visual system. Although a more systematic study of the effect of cannabinoids in different types of RGCs and other neurons in the different areas of the visual pathway is necessary, the overall effects at different stages of visual processing seem to be a consistent cannabinoid-induced reduction of visual responses.

Taken together, our results demonstrate the functional expression of the retinal cannabinoid system and show how activation of cannabinoid receptors modifies the response strength and the spatial tuning properties of ON-S RGCs to acquire properties characteristic of low light level conditions. This suggests that alterations, either pathological or induced by exocannabinoids, in the cannabinergic system, might have profound effects on the transmission of light signals and consequently in vision.