Ih Channels Control Feedback Regulation from Amacrine Cells to Photoreceptors

In both vertebrates and invertebrates, photoreceptors’ output is regulated by feedback signals from interneurons that contribute to several important visual functions. Although synaptic feedback regulation of photoreceptors is known to occur in Drosophila, many questions about the underlying molecular mechanisms and physiological implementation remain unclear. Here, we systematically investigated these questions using a broad range of experimental methods. We isolated two Ih mutant fly lines that exhibit rhythmic photoreceptor depolarization without light stimulation. We discovered that Ih channels regulate glutamate release from amacrine cells by modulating calcium channel activity. Moreover, we showed that the eye-enriched kainate receptor (EKAR) is expressed in photoreceptors and receives the glutamate signal released from amacrine cells. Finally, we presented evidence that amacrine cell feedback regulation helps maintain light sensitivity in ambient light. Our findings suggest plausible molecular underpinnings and physiological effects of feedback regulation from amacrine cells to photoreceptors. These results provide new mechanistic insight into how synaptic feedback regulation can participate in network processing by modulating neural information transfer and circuit excitability.


Introduction
Feedback regulation is common in neural circuit information processing. In both vertebrate and invertebrate visual systems, photoreceptor output is feedback-regulated by interneurons, which is an important mechanism for shaping the transmission of light information [1,2]. In the vertebrate retina, bipolar cells receive synaptic input from rod and cone photoreceptors and transfer information to ganglion cells. Meanwhile, the laterally distributed horizontal cells provide a feedback signal to photoreceptor axon terminals, controlling their output gain [3,4]. The structure, function, and development of the vertebrate and insect visual systems possess many evolutionary parallels [5]. In the Drosophila lamina, 12 neuron classes have been identified, and specific interneurons may serve similar functions [6,7]. Serial electron-micrograph (EM) studies have revealed that outer photoreceptor (R1-R6) axons project their outputs to L1-L3 monopolar cells and amacrine cells (AC) and receive synaptic inputs from L2, L4, AC, Lawf, and C3 cells [6][7][8]. Because connectivity in the Drosophila lamina has been elucidated to the level of individual synapses, this system provides a good model to study how the feedback neural circuit works and facilitates network information processing [6,7,9,10].
I h channels, also called hyperpolarization-activated cyclic nucleotide-gated (HCN) channels, are low-threshold, voltage-gated ion channels that are normally activated at negative potentials below −50 mV [19,20]. As I h channels are permeable to Na + and K + and form an inward current at rest, they may depolarize the neuronal resting membrane potential (RMP) and influence excitatory postsynaptic potential kinetics and integration [21][22][23]. A recent study demonstrated that HCN1 colocalizes with low-threshold voltage-gated T-type Ca 2+ channels (Ca v 3.2) in presynaptic terminals and inhibits glutamate release by suppressing Ca v 3.2 activity [24]. In the present study, we examined two I h mutant fly lines that exhibit rhythmic depolarization in photoreceptors without light stimulation. Our results demonstrate that I h channels are expressed in ACs and suggest that I h channels regulate synaptic glutamate release by modulating the activity of Cacophony (Cac) channels. We further showed that the eye-enriched kainate receptor (EKAR) receives the retrograde glutamate signal in photoreceptor terminals. Finally, we investigated how feedback regulation from ACs affects photoreceptor output and fly optomotor behavior. Our studies elucidate the molecular mechanism and physiological roles of feedback regulation from ACs to photoreceptors in the Drosophila visual system.

I h Mutant Photoreceptors Undergo Rhythmic Depolarization
To identify additional genes involved in fly photoreceptor functions, we performed an electroretinogram (ERG)-based genetic screen in the mutant lines from Exelexis collections [25,26] and gene disruption project (GDP) collections [27]. In this screen, we identified two I h mutant lines, PBac{XP}I h f01485 and PBac{XP}I h f03355 , that exhibited normal light responses and distinctive ERG baseline oscillations ( Fig. 1A-C). Although the amplitudes and frequencies of oscillations were variable within individual flies (S1A Fig.), this ERG phenotype was easily detectable and distinct from that of wild-type flies, which never exhibited ERG baseline oscillations (Fig. 1B,C). Strikingly, ERG baseline oscillations in I h mutant flies were sustained for more than 15 min, although their amplitude and frequency were attenuated (S1B Fig.). Similar results were observed in the recombinants with two deficiency lines Df(2R)Exel7131 and Df(2R) BSC700, in which the entire I h gene (Gene ID: 36589) was deleted (Fig. 1B,C) [26,28]. Splicing of the I h gene creates several transcriptional variants that encode I h channels with long or short N-termini and different lengths of the inter-loop regions between the membrane-spanning domains S3-S4 and S4-S5 ( Fig. 1A) [29]. The PBac{XP}I h f03355 mutant contains a piggyBac inserted into the extron of all transcriptional variants of the I h gene, whereas PBac{XP}I h f01485 mutant has a piggyBac inserted into the intron of most transcriptional variants of the I h gene (Fig. 1A) [25]. Reverse transcription polymerase chain reaction (RT-PCR) analysis revealed that the piggyBac insertion completely abolished the mRNA transcription of I h gene in these two mutant lines (Fig. 1D). Using antibodies against the intracellular C terminal domain that exists in all I h channel variants, we revealed four major I h channel variants (170, 125, 73, and 71 kDa) that were expressed in wild-type flies but absent in I h mutant flies (Fig. 1E). In addition, two low intensity bands (74 and 52 kDa) exist in I h mutants (Fig. 1E), which might be nonspecific bands. To further confirm that piggyBac insertion actually disrupts the I h gene and leads to ERG baseline oscillations in PBac{XP}I h f03355 mutants, we performed piggyBac precise excision from PBac{XP}I h f03355 mutants. ERG recording showed that this completely abolished the ERG baseline oscillations phenotype (Fig. 1B,C). Taken together, these results demonstrate that loss of I h channels results in an abnormal ERG baseline oscillation phenotype. Because ERG is an extracellular recording technique that measures the light-induced mass response of the eye, we next conducted intracellular recordings in photoreceptors to investigate whether the ERG baseline oscillation phenotype was due to a photoreceptor abnormality. Interestingly, I h mutant photoreceptors, but not wild-type photoreceptors, showed rhythmic depolarization without light stimulation ( Fig. 2A,B). In the first several minutes of recording, the pacemaker traces in I h mutant flies showed a uniform amplitude (3.10 ± 0.36 mV) and frequency (0.24 ± 0.05 Hz). Each depolarization had a rise time of 0.82 ± 0.12 s, lasted 1.57 ± 0.13 s, and had a decay time of 0.60 ± 0.15 s (Fig. 2A). The frequency of depolarization in the intracellular recordings was much lower than that in ERG recordings, likely due to the desynchronized depolarization of multiple photoreceptors in the extracellular recordings. In addition, I h mutant photoreceptors showed a reduced amplitude (10.3 ± 0.7 mV versus 11.6 ± 0.8 mV, p < 0.01) as well as a prolonged decay time (0.23 ± 0.03 s versus 0.06 ± 0.01 s, p < 0.001) of light-evoked depolarization (Fig. 2C). These alterations might reflect abnormal photoreceptors and/or cell communication in the visual system.
To explore whether loss of I h channels causes a photoreceptor abnormality, we conducted EM studies and biochemical analyses of photoreceptors. EM images showed normal  (Fig. 3A), and the expression and localization of phototransduction components in I h mutant flies were also normal (Fig. 3B,C), indicating that the rhythmic depolarization noted in I h mutant photoreceptors was not due to abnormalities in rhabdomere structure or phototransduction cascades. We also excluded the possibility that the rhythmic depolarization in I h mutant photoreceptors was dependent on phototransduction cascades activation by genetically blocking phototransduction cascade activation through introducing the mutant of norpA, which encodes the key phototransduction component phospholipase C [30]. ERG recordings showed that norpA mutation did not suppress the rhythmical depolarization in I h mutant photoreceptors (Fig. 3D).

Lack of I h Channels in ACs Leads to Rhythmic Depolarization in Photoreceptors
To elucidate how the loss of I h channels leads to rhythmic depolarization in photoreceptors, we examined the expression pattern of endogenous I h channels. In wild-type flies, I h channel staining was observed in the lamina and medulla, whereas photoreceptor cell bodies were either weakly or not at all labeled (Fig. 4A). Although the axons of outer photoreceptors project into the lamina and form synaptic connections with multiple lamina neurons, I h channels were are absent in I h mutant flies. Primer pair CACGCGACCAATCTCATCC/ TCATGGAGTGTTACCCTCG, which can amplify all transcriptional variants, was used in RT-PCR analysis. The tubulin gene was used as a loading control. (E) Western blotting revealed four major I h channel variants (indicated by arrows) expressed in wild-type flies but absent in I h mutant flies. Note that the low-intensity bands presented in I h mutant flies are nonspecific.  undetectable in photoreceptor axons (Fig. 4B). Strong I h labeling was observed in the somata of L1 and L2 neurons, whose membranes had been labeled with mCD8-GFP markers (Fig. 4C,D). I h channels were also found in the somata and processes of ACs, which was identified by expression of mCD8-GFP markers under the control of AC-specific split GAL4 (Lai-GAL4) (Fig. 4E) [31].
To determine whether I h channels in the lamina neurons or glia might contribute to rhythmic depolarization in photoreceptors, we specifically depleted I h channels in various cell types using an RNAi knockdown approach. The RNAi line THU02084.N recognizes all I h transcriptional variants (Fig. 1A) [32], and western blotting analysis demonstrated that RNAi knockdown of I h channels using pan-neural elav-GAL4 but not glia-specific repo-GAL4 successfully repressed the expression of all I h variants (Fig. 5A). Moreover, RNAi against all I h transcriptional variants using elav-GAL4 but not repo-GAL4 phenocopied the abnormal ERG baseline oscillations observed in I h mutant flies (Fig. 5B,C), indicating that the loss of I h channels in neurons, but not glia, results in rhythmic depolarization in photoreceptors. Consistent with no or low I h channel expression in photoreceptors, the Rh1-GAL4 driver did not recapitulate the ERG deficits observed in I h mutant flies (Fig. 5B,C). These results suggest that rhythmical depolarization in I h mutant photoreceptors is caused by abnormal communication between photoreceptors and other neurons.
Photoreceptor terminals receive synaptic inputs from L2, L4, and Lawf neurons and ACs [6][7][8]. Given that I h channels are expressed in L1 andL2 neurons and ACs, we next asked whether the loss of I h channels in these neurons causes rhythmic depolarization in photoreceptors. Although RNAi knockdown of I h channels in L1 and L2 neurons using L1L2-GAL4 [33,34] significantly reduced the levels of I h channels (Fig. 5A), these RNAi knockdown flies did not exhibit obvious ERG baseline oscillations (Fig. 5B,C), suggesting that the loss of I h channels in L1 and L2 neurons is not sufficient to trigger rhythmic depolarization in photoreceptors. Our immunostaining data from Lai-GAL4;UAS-mCD8-GFP flies have shown that the number of ACs is small (Fig. 4E), and anti-I h antibody staining revealed that I h expression in ACs is much lower than that in L1/L2 neurons (Fig. 4E). Therefore, depletion of I h channel from ACs may not lead to a detectable reduction in total I h channels in the whole fly head (Fig. 5A). To validate that RNAi knockdown in ACs works well, we performed immnuostaining analysis and showed that

Feedback Regulation in Fly Vision
RNAi knockdown of I h channels using Lai-GAL4 successfully did deplete I h channels in ACs (Fig. 5D). Interestingly, RNAi knockdown of I h channels using Lai-GAL4 recapitulated the ERG abnormalities observed in I h mutant flies (Fig. 5B,C), indicating that the loss of I h channels in ACs leads to rhythmic depolarization in photoreceptors. Although knockdown of I h channels in L1 and L2 neurons using L1L2-GAL4 caused occasional ERG baseline oscillations, the recombinant of Lai-GAL4 and L1L2-GAL4 resulted in greater ERG baseline oscillations compared with those caused by Lai-GAL4 alone (Fig. 5B,C). These results suggest that the loss of I h channels in L1 and L2 neurons might also contribute to rhythmic depolarization in photoreceptors.
We next generated p[UAS-I h ] transgenic flies and performed rescue experiments. To choose an appropriate variant for transgene generation, we first determined which variant is expressed in the retina and lamina. The RNAi line P{KK100190}VIE-260B only recognizes the long isoforms of I h gene ( Fig. 1A) [35]. RNAi against long I h transcriptional variants using eye-specific eyeless-GAL4 showed normal I h protein levels and ERG response (S2A,B Fig.). Furthermore, pan-neural elav-GAL4 failed to recapture the ERG baseline oscillations (S2B,C Fig.), although it caused a significant reduction in the long isoforms of I h proteins (170 and 125 kDa, S2A Fig.). The above observations indicated that only short isoforms of I h play essential roles in suppressing the ERG baseline oscillation phenotype. Western blotting also revealed that 71and 73-kDa variants but not 170-and 125-kDa variants were highly expressed in the isolated retina and lamina (Fig. 6A). RT-PCR analysis further showed that the annotated transcript form I h -RK, which encodes a 618aa isoform I h -PK protein, was abundant in the retina and lamina (Fig. 6B). Thus, we amplified the cDNA of I h -RK and generated p[UAS-I h K ] transgenic flies to confirm that this transgene can be successfully expressed in neurons (Fig. 6C).
Expression of I h channels in ACs but not in L1 and L2 neurons restored normal ERG activity in I h mutant flies ( Fig. 7A-C). Conversely, expression of I h channels in photoreceptors (Rh1-GAL4) or glia (repo-GAL4) had no inhibitory effects on ERG baseline oscillations ( Fig. 7A,B). Intracellular recordings further validated that the expression of I h channels in ACs suppressed rhythmic depolarization in photoreceptors (Fig. 7D). These observations provide evidence that AC-derived I h channels are critical to inhibit rhythmic depolarization in photoreceptors.

Rhythmic Depolarization in I h Mutant Photoreceptors Is Due to Uncontrolled Synaptic Glutamate Release from ACs
As photoreceptor terminals receive synaptic inputs directly from ACs, we first investigated whether rhythmic depolarization in I h mutant photoreceptors was due to abnormal cartridge structure or connections between photoreceptors and ACs. However, EM images showed no obvious morphological differences between I h mutant and wild-type cartridges (Fig. 8A). Therefore, we suspected that rhythmic depolarization in photoreceptors might be due to abnormal synaptic output from ACs. To test this hypothesis, we blocked neurotransmitter release from ACs by expressing tetanus toxin light chain (TeTxLC) [36] or silencing ACs by expressing a mutant form of open rectifier potassium channel (dORK ΔC ) [37]. TeTxLC and dORK ΔC expression was suppressed during development by the ubiquitous expression of temperature- sensitive Gal80 ts but selectively induced during adulthood by exposure to 30°C for 4 h [38]. Interestingly, intracellular recordings revealed that TeTxLC expression in ACs abolished rhythmic depolarization in I h mutant photoreceptors and led to a prolonged decay time of lightinduced depolarization ( Fig. 8B and S3A Fig.). By contrast, TeTxLC expression in L1 and L2 neurons failed to suppress rhythmic depolarization in I h mutant photoreceptors ( Fig. 8B and  S3A Fig.). Consistently, ectopic expression of dORK ΔC in ACs but not in L1L2 neurons also suppressed rhythmic depolarization in I h mutant photoreceptors and caused a prolonged decay time of light-induced depolarization (Fig. 8C). These observations indicate that rhythmic depolarization in I h mutant photoreceptors is due to abnormal synaptic output from ACs.
Given that ACs are likely glutamatergic neurons [16], we further tested whether rhythmic depolarization in photoreceptors is caused by uncontrolled glutamate release from ACs. Vesicular glutamate transporter (vGluT) functions in loading glutamate into synaptic vesicles and is therefore critical for synaptic glutamate output [16]. Consistently, knockdown of vGluT expression in ACs but not in L1 and L2 neurons suppressed rhythmic depolarization in I h mutant photoreceptors ( Fig. 8D and S3B Fig.). Taken together, the above observations demonstrate that rhythmic depolarization in I h mutant photoreceptors is due to uncontrolled synaptic glutamate release from ACs.
Cac channels can produce high voltage-activated Ca 2+ currents above −30 mV and low voltage-activated Ca 2+ currents between −70 and −60 mV in vivo [42]. I h channels are normally open at potentials more negative than −50 mV, leading to depolarization of the RMP. A recent study demonstrated that HCN1 colocalizes with Ca v 3.2 in presynaptic terminals and inhibits glutamate release by suppressing Ca v 3.2 activity [24]. Thus, it is possible that the loss of I h channel activity hyperpolarizes the RMP of ACs to a small window (−70mV to −60 mV) in which Cac channels are active. If this is true, then we should be able to suppress Cac activity and subsequent rhythmic depolarization in photoreceptors by changing AC RMP. Previous recordings in larval muscle fibers showed that expression of the sodium channel NaChBac evoked robust voltage-gated inward currents that began to activate at approximately −60 mV and peak at approximately −30 mV [43]. Therefore, we genetically depolarized the RMP of ACs by expressing NaChBac [43]. Interestingly, its expression in ACs but not in L1L2 neurons suppressed rhythmic depolarization in I h mutant photoreceptors (Fig. 9C). However, both expression of dORK ΔC and NaChBac in ACs failed to trigger rhythmic depolarization in wildtype flies (Fig. 8C and Fig. 9C). Taken together, these findings suggest that rhythmic depolarization in I h mutant photoreceptors is due to changes in the RMP that relieve Cac inactivation in ACs.

EKAR Receives Synaptic Glutamate Output from ACs
To identify the glutamate receptor that mediates retrograde glutamate signaling from ACs to photoreceptors, we screened 15 known ionotropic glutamate receptor (iGluR) subunits, including three conserved classes (kainate, AMPA, and NMDA types) of cation iGluRs and one chloride channel (GluClα) ( Table 1) [20]. Interestingly, knockdown of ekar (CG9935, Gene ID: 43806) in photoreceptors by two individual RNAi lines, THU3080 and THU4260, suppressed ERG baseline oscillations in I h mutant flies (Fig. 10). However, knockdown of other glutamate receptors in photoreceptors did not show any inhibition effects (Fig. 10).
We also obtained an ekar mutant allele, Mi{ET1}CG9935 MB00001 [44], that does not produce ekar mRNA (Fig. 11A,B). Intracellular recordings revealed that light-induced depolarization of photoreceptors was significantly reduced in ekar mutant flies compared with wild-type flies (11.6±0.8 mV versus 4.3±0.8 mV, p < 0.001, Fig. 11C). However, EM images revealed normal rhabdomere structures in ekar mutants (S4A Fig.), and western blotting showed that ekar mutants express normal protein levels of phototransduction components (S4B Fig.). These observations suggest that EKAR might contribute to the light-evoked depolarization of photoreceptors. To further validate the role of EKAR in mediating the retrograde glutamate signal, we generated I h ;;ekar double mutant flies. Intracellular recordings revealed that ekar mutation suppressed rhythmic depolarization in I h mutant photoreceptors (Fig. 11D,E). Given  that rhythmic depolarization in I h mutant photoreceptors was independent of phototransduction cascades activation (Fig. 3D), the suppression of rhythmic depolarization in I h ;;ekar double mutant flies indicates that the kainate receptor EKAR receives synaptic glutamate output from ACs.

Feedback Regulation Is Critical for Visual Signal Transmission and Motion Detection in Dim Light Conditions
To explore the potential role of this feedback regulation, we first examined whether feedback regulation is required for photoreceptor survival. EM images revealed normal rhabdomere structure in 14-day-old I h and ekar mutant flies raised under regular light cycles (12 h light/12 h dark) or in constant darkness (S5 Fig.). These observations suggest that feedback regulation from ACs to photoreceptor terminals is not essential for photoreceptor survival.
We next performed intracellular recordings to examine photoreceptors' excitability in response to various light intensity stimulations. The results showed that photoreceptors underwent light-induced depolarization in a light intensity-dependent manner (Fig. 12A,B). 10 Lux light stimulation evoked a 2.91 ± 0.19 mV depolarization in wild-type photoreceptors and a significantly reduced depolarization in I h mutant photoreceptors (2.01 ± 0.11 mV versus 2.91 ± 0.19 mV; p < 0.05, Fig. 12A,B), which was indistinguishable with rhythmic depolarization without light stimulation in I h mutant photoreceptors (Fig. 12A). In contrast, 10 Lux light stimulation triggered a significantly reduced depolarization in ekar mutant photoreceptors (0.63 ± 0.10 mV versus 2.91 ± 0.19 mV; p < 0.001) and in Lai-GAL4/UAS-TeTxLC photoreceptors (1.04 ± 0.10 mV versus 2.91 ± 0.19 mV; p < 0.001) (Fig. 12A,B). These observations indicate that feedback regulation from ACs to photoreceptor terminals facilitates photoreceptor excitability and helps maintain light sensitivity in presence of ambient light.
To investigate the potential role of this feedback regulation in visual behavior, we assessed the flies' optomotor responses under various light conditions. We placed single flies on a circular platform and examined their ability to track moving light patterns (Fig. 12C,D). With highintensity moving light patterns (85 and 800 Lux), I h mutant flies but not ekar mutant or Lai-GAL4/UAS-TeTxLC flies exhibited a reduced ability to track moving patterns (Fig. 12E). However, with low-intensity moving light patterns (1 and 10 Lux), I h mutant flies, ekar mutant, and Lai-GAL4/UAS-TeTxLC flies were less able to track moving patterns (Fig. 12E). These findings demonstrate that feedback regulation from ACs to photoreceptor terminals enhances the flies' optomotor response in dim light conditions, whereas uncontrolled feedback regulation also disturbs motion detection in ambient light conditions (Fig. 12E).
Based on these results, we propose a feedback regulation model from ACs to photoreceptors (Fig. 12F). Photoreceptors synthesize the inhibitory neurotransmitter histamine, which is released upon light stimulation [12,45]. Thus histamine hyperpolarizes ACs by opening HisCl 2 channels [13,46,47]. Hyperpolarization of ACs activates I h channels, which depolarizes AC RMP and limits Cac channel activity. Without I h channels in ACs, Cac channels are activated, resulting in Ca 2+ influx and subsequent glutamate release from ACs. EKAR is expressed in photoreceptor terminals and depolarizes photoreceptors upon receiving the retrograde glutamate released from ACs.

I h Channels Regulate Retrograde Glutamate Release by Modulating Cac Activity
It has been shown I h channels are expressed in several classes of interneurons that exhibit spontaneous firing activity and provide tonic inhibition to principal neurons, thus contributing to the regulation of firing frequency and excitability [48]. In this study, we revealed that loss of I h channels in ACs results in rhythmic depolarization in photoreceptors. This phenotype was suppressed by either blocking neurotransmitter release or impairing synaptic glutamate output from ACs. Our studies provide solid evidence that feedback regulation from ACs to photoreceptors is regulated by I h channels. Although expression of I h channels in L1 and L2 neurons failed to suppress rhythmic depolarization in I h mutant photoreceptors, L1/ L2 neuron-expressed I h channels may also contribute to feedback regulation, as knockdown of I h channels using recombinant Lai-GAL4 and L1L2-GAL4 led to enhanced ERG baseline oscillations compared with Lai-GAL4 alone. These findings are consistent with previous morphological studies showing that outer photoreceptor terminals also directly receive feedback inputs from L2 neurons [6][7][8].
Low-threshold Ca 2+ channels are expressed in a variety of tissues such as the brain, heart, smooth muscle, kidney, and various endocrine glands [49]. These channels play important roles in controlling intracellular Ca 2+ levels, modulating neuronal excitability, and regulating hormone and neurotransmitters secretion [50]. Here, we show that cac mutant flies suppressed glutamate release and subsequent rhythmic depolarization in I h mutant photoreceptors. We further showed that rhythmic depolarization in I h mutant photoreceptors were suppressed by changing the RMP of ACs. Our results suggest that HCN channels depolarize the RMP, thereby restricting Ca 2+ entry via Cac channels and preventing glutamate release. Therefore, glutamate release is enhanced in I h mutants, which causes rhythmic depolarization of photoreceptors. In addition, we showed that the released glutamate may induce long-lasting depolarization by opening of EKAR, which may contribute to the slow repolarization at the end of light stimulation in I h mutant photoreceptors. A recent study reports that HCN1 channels localize in the active zone of mature asymmetric synaptic terminals and inhibit synaptic glutamate release by suppressing the activity of low-threshold voltage-gated T-type Ca 2+ channels [24]. Thus, this form of regulation might be a common mechanism by which these channels modulate neuronal excitability.

EKARs Receive Retrograde Glutamate Signals in Photoreceptor Terminals
Intracellular recordings from wild-type and shibire TS mutant flies reveal that cessation of all synaptic feedback to photoreceptors results in a 10-15 mV hyperpolarization shift upon light Feedback Regulation in Fly Vision stimulation [14], suggesting that feedback regulation depolarizes photoreceptors upon light stimulation. However, the receptors that receive the excitatory neurotransmitter are still unknown. Our findings indicate that the EKAR receptor receives the glutamate signal in photoreceptor terminals: both knockdown of EKAR in photoreceptors and mutation of ekar suppressed rhythmic depolarization in I h mutant photoreceptors. This result is supported by the previous microarray data, which showed that EKAR is highly expressed in the eye [51]. Our intracellular recording result also revealed that light-induced depolarization of photoreceptors was significantly reduced in ekar mutant flies compared with wild-type flies. The recordings further showed that I h mutant photoreceptors undergo rhythmic depolarization with slow rise and decay times, which is consistent with the physiological properties of kainate receptors that mediate postsynaptic depolarization with slow rise and decay time [52]. Taken together, these results indicate that EKAR is expressed in photoreceptor terminals and depolarizes photoreceptors upon light stimulation.

Feedback Regulation from ACs to Photoreceptors Helps Maintain Light Sensitivity in the Presence of Ambient Light
In vertebrates, the synaptic input from horizontal cells to cone cells contributes to many visual functions including the formation of center-surround receptive fields, retinal synchronization, and light adaptation [53][54][55]. Fly ACs are structurally equivalent to horizontal cell in vertebrates. A recent study showed that silencing ACs reduces optomotor responses to regressive rotation stimulation, whereas activation of ACs leads to slightly decreased responses to high and low contrast stimulations [31]. Although ACs project most of their synapses to epithelial glia, they also form direct feedback synapses to photoreceptor axons [6]. In this study, we show that feedback regulation from ACs to photoreceptors improves ambient light-induced visual signal transmission and motion detection under dim light conditions, which might be important for fly activity at dawn and dusk. Conversely, uncontrolled feedback regulation in I h mutant flies impairs visual signal transmission and motion detection in ambient light conditions, Feedback Regulation in Fly Vision suggesting that feedback regulation is strictly modulated. In this study, we showed that I h mutants exhibit a significantly reduced ability in tracking the moving patterns with high light intensity. Since L1 and L2 neuron play essential roles in normal motion vision [31,56], loss of I h channels in L1 and L2 neurons may contribute to this reduced ability in motion detection. In Drosophila photoreceptors, the rapid termination of photoresponse in Drosophila is thought to be achieved by fast deactivation of rhodopsin and calcium-mediated intrinsic feedbacks [57,58]. In addition, a slow termination of photoresponse has been reported in mutants with blocked photoreceptor transmission [59], suggesting a retrograde regulation is likely to contribute to the termination speed. In this study, Our ERG and intracellular recordings showed that blocking neurotransmitter release from ACs resulted in slow repolarization at the end of light stimulation. Similar phenotypes are observed in flies with reduced glutamate signal output from ACs and cac mutant flies [60]. Given that most of mutants with slow termination phenotype did not undergo rhythmic depolarization [57][58][59][61][62][63][64], the slower depolarization phenotype is not sufficient to trigger the rhythmic depolarization. These observations suggest that feedback regulation from ACs is essential for the rapid repolarization of photoreceptors at the end of light stimulation. However, repolarization speed is normal in ekar mutant flies. Because ACs form most of feedback synapses to epithelial glia [6]. Rapid repolarization of photoreceptors might be regulated by glia cells. The mechanism that facilitates rapid repolarization at the end of light stimulation need to be further investigated.
In summary, our studies reveal the molecular mechanism and physiological roles of feedback regulation from ACs to photoreceptors. This might represent a general mechanism by which feedback regulation modulates synaptic transmission and facilitates neural circuit excitability and network information processing.

Fly Stock
Transposon piggyBac insertion flies I h f03355 and I h f01485 [25,26] were obtained from Harvard Medical School. The mutant alleles used for other genes in this work are Cac H18 [65], PBac {WHr}Ca-α1T del [42], and Mi{ET1}CG9935 MB00001 [44,66]. UAS-RNAi lines were ordered from Vienna Drosophila RNAi Center and Tsinghua RNAi Stock Centre. L1-GAL4, L2-GAL4, and L1/L2-GAL4 lines [33,34] were obtained from Dr. Jens Rister, and the split Lai-GAL4 line (R92A10AD attP40; R17D06DBD attp2) [31] was provided by Dr. Aljoscha Nern. Other lines used in this work were obtained from the Bloomington Stock Center. The wild-type flies used in this study were w 1118 . All flies were maintained in standard medium at 25°C, with 60%-80% relative humidity. Less than 3-day-old flies were used for ERG recording, and 3-day-old flies were used in optomotor response assays. In all experiments, an equal number of male and female flies were used. Full genotypes of samples shown in main figure panels are provided in S1 Table. Antibodies Anti-I h antibodies were generated in rabbits against a purified glutathione S-transferase fusion fragment (aa332-618) of I h -PK protein and generated by GenScript (Nanjing, China). An affinity column, generated by coupling a MBP-I h fragment (aa332-618) to Sepharose 4B, was used to purify the antibody. The sources of other antibodies were rabbit anti-TRP [67], rabbit anti-Arr2 [61], rabbit anti-INAD [63], anti-PLC antibodies [68], anti-GFP (Abcam), and mouse anti-24B10 [69] and anti-Rh1 (4C5) (DSHB).

Electrophysiological Recordings
Electroretinogram (ERG) recordings were conducted at 25°C as previously described [70]. Less than 3-day-old flies were collected, immobilized with strips of tape and put in darkness for 5 min for adaptation. Two glass microelectrodes were filled with Ringer's solution and placed on the thorax and compound eye. Flies were stimulated with 5-s light pulses (4000 Lux) every 25 s using a Newport light projector. For each fly, ERG recording lasted for more than 100 s. The signal was amplified and recorded using a Warner IE210 Intracellular Electrometer. In ERG and intracellular recording without light stimulation, more than five continuous depolarizations with amplitude >1 mV was defined as an ERG baseline oscillation phenotype. The fraction of flies that exhibited ERG oscillation was quantified and presented in the figures. Intracellular recordings were performed as described previously [63]. Briefly, flies were fixed with strips of tape, and a small opening was made on surface of the eye using fine tweezers. A low resistance (>30 MO) glass microelectrode filled with 2 M KCl was gradually inserted into the opening until light-induced membrane depolarization was observed. A reference electrode was filled with Ringer's solution and placed inside the eye at the retina layer. The signal was amplified and recorded using a Warner IE210 Intracellular Electrometer. The fractions of photoreceptors exhibiting rhythmic depolarization in each genotype were calculated. To quantify the amplitudes of light responses, 10 photoreceptors from 10 flies were measured for each genotype, and the mean±SEM was calculated and showed in figures.

Retina and Lamina Isolation
To separate the retina and lamina from the brain for western blotting analysis, fly heads were cut and immersed in 100% ethanol for 2 h before the retina and lamina were carefully dissected from the brain. Separated tissues (retina and lamina, the head without retina and lamina) were homogenized in SDS-sample buffer. To isolate the retina and lamina for RT-PCR analysis, they were dissected carefully, and the mRNA was extracted from the separated tissues using TRIzol reagent (Invitrogen).

Immunostaining
Section staining was performed as previously described [61]. Fly heads were fixed with 4% paraformaldehyde. After three washes, the fly heads were dehydrated with acetone and embedded in LR White resin, and 1-μm cross-sections were cut across the top half of the eye. The sections were incubated in diluted primary antiserum (Rh1, 1:200; INAD, 1:400; TRP, 1:400) at room temperature for 1 h. After three times of washing in PBS, sections were incubated with diluted secondary antibodies at room temperature for 1 h. The stained sections were examined under a ZEISS Axio observer A1 microscope.
Whole-head staining was performed to locate endogenous I h channel in wild-type adult flies. Fly heads were dissected in PBS buffer and fixed with 4% paraformaldehyde in PBS buffer. After fixation, the heads were double labeled with diluted primary antiserum (anti-I h antibodie 1:50 and 24B10 1:100 or anti-GFP 1:200). After three washes in PBS buffer, fly heads were incubated with diluted secondary antibodies at room temperature for 1 h [71]. After three additional PBS washes, the stained heads were examined under an LSM 700 confocal microscope.

Electron Micrograph
EM was carried out as previously described [72]. Fly heads were fixed at 4°C for 12 h in 2.5% gluteraldehyde, 0.1 M sodium cacodylate (pH 7.2). After three washes with 0.1 M sodiumcacodylate, the heads were stained with 1% osmium tetroxide at room temperature for 1 h. After a standard ethanol dehydration series, the heads were immersed in propylene oxide for two 10-min washes before they were embedded with standard procedures. Thin sections (100 nm) were cut at the top 2/3 of retina to show ommatidia whereas cut at half the lamina to display cartridges. Sections were collected on Cu support grids and stained with uranyl acetate for 8 min, followed by 5 min with lead citrate. Micrographs were taken at 80 kV on a Hitachi-7650 transmission EM.

Optomotor Responses Assay
Fly optomotor responses were tested as previously described [73]. Briefly, 3-day-old flies were collected and their wings were cut off. After recovering for more than 24 h in a 12-h light/12-h dark cycle, a single fly was placed on a circular platform for the optomotor response test. The platform is surrounded with a water-filled moat to prevent the fly from escaping, and the moat was surrounded with a panoramic LED display that controlled by LED Studio software (Shenzhen Sinorad Medical Electronics). Bright and dark stripes were used to generate a clockwise motion light for 90 s followed by an anti-clockwise motion light for another 90 s (180°/s corresponding to a temporal frequency of 4 Hz). The walking traces of flies were recorded by a camera (WV-BP330, Panasonic System Networks), and the data were analyzed in MATLAB (Mathworks). The flies' optomotor responses were quantified by the performance index of tracking time (PITT). The PITT is defined as (P tracking time -P un-tracking time )/ (P tracking time + P un-tracking time ). The probabilities of fly movement in the platform in accordance with LED rotating direction or not are defined as tracking or un-tracking, respectively. The male and female flies were used alternately.

Statistics
Statistical analysis was done by using MS Excel. The numerical data used in all figures are included in S1 Data. For quantitative data of ERG and intracellular recordings, fraction of flies or photoreceptors that exhibit oscillations and standard error of rate are shown. Fisher's exact probability tests were used to compare genotypes. For statistical analysis of depolarization amplitudes and decay time of depolarization in Figs. 2C, 8B, 8D, 11C, and 12B, data are presented as mean ± SEM. Two-tailed Student's t tests were used to compare genotypes. Significance was classified as follows: Ã , p 0.05; ÃÃ , p < 0.01; ÃÃÃ , p < 0.001; n.s. p > 0.05.