Retrograde signaling from the brain to the retina modulates the termination of the light response in Drosophila

Abstract

A critical factor in visual function is the speed with which photoreceptors (PRs) return to the resting state when light intensity dims. Several elements subserve this process, many of which promote the termination of the phototransduction cascade. Although the known elements are intrinsic to PRs, we have found that prompt restoration to the resting state of the Drosophila electroretinogram can require effective communication between the retina and the underlying brain. The requirement is seen more dramatically with long than with short light pulses, distinguishing the phenomenon from gross disruption of the termination machinery. The speed of recovery is affected by mutations (in the Hdc and ort genes) that prevent PRs from transmitting visual information to the brain. It is also affected by manipulation (using either drugs like neostigmine or genetic tools to inactivate neurotransmitter release) of cholinergic signals that arise in the brain. Intracellular recordings support the hypothesis that PRs are the target of this communication. We infer that signaling from the retina to the optic lobe prompts a feedback signal to retinal PRs. Although the mechanism of this retrograde signaling remains to be discerned, the phenomenon establishes a previously unappreciated mode of control of the temporal responsiveness of a primary sensory neuron.

For an organism to extract useful information about the photic environment, its visual system must to able to respond quickly to alterations in light intensity. Much of this speed is built into the phototransduction apparatus. For example, in the visual system of Drosophila, the cascade of events that follow absorption of a photon by rhodopsin produces within milliseconds a change in current across the photoreceptor (PR) membrane (1, 2). This cascade starts with the conversion by photoactivated rhodopsin of a G protein to a form that stimulates phospholipase C, whose products (by a pathway that is not fully understood) promote the opening of TRP family ion channels. No less important to an organism's fitness is the speed with which the photoresponse terminates when light levels drop. Rather than rely on the spontaneous decay of the light response, visual systems use dedicated mechanisms to speed the process (3). In Drosophila, as in other organisms, a prominent deactivation pathway features complexation of rhodopsin with arrestins. The intricacy of the termination response is evidenced by its dependence on genes (e.g., inaC) that seem to function separately from the arrestin pathway. To date, however, all of the elements identified as being involved in restoring the system to its resting state have been limited to PR components. Here we show that, when darkness follows a few seconds of exposure to light, prompt restoration of the resting state depends on synaptic output from PRs to second-order neurons. Our work implies a previously unsuspected level of feedback control of the termination of the photoresponse, one that operates on phasic properties of primary sensory neurons.

Materials and Methods

Fly Stocks. The control stocks used in this study were as follows: Canton-S, obtained from J. Steven de Belle (University of Nevada, Las Vegas, NV); w¹¹¹⁸ made congenic to Canton-S by repeated backcrossing (4); the parental w¹¹¹⁸ strain from which the Exelixis set of inserts (5) was derived; and an isogenic derivative of this strain containing a homozygous viable insert (e04146) in a randomly selected gene (CG11674) that we find has no effect on visual physiology. The last two strains and an isogenic derivative containing a homozygous viable insert (e02083) in the ort gene were obtained from the Bloomington Stock Center (Bloomington, IN). A yw;;arr2⁵ stock was provided by Hong-Sheng Li (University of Massachusetts, Worcester, MA). From William Pak (Purdue University, West Lafayette, IN) we obtained stocks of w;Hdc^JK910 and w;Hdc^P211 (6). The w mutation was removed from these stocks by outcrossing to Canton-S. The Pak laboratory also provided w;inaC^P209 and w;;ort^US2515 stocks. The Cha-Gal4-19B (7) and UAS-shi^ts1 strains (8) were obtained from Benjamin White (National Institute of Mental Health, Bethesda). These were crossed to generate Cha-Gal4/+;UAS-shi^ts1/+ offspring, which were raised at 19°C and tested after being shifted to 30^°C while mounted on a temperature-controlled stage. Handling and maintenance of the fly stocks were routine except that dark-rearing was used for light-sensitive stocks and their controls.

Pharmacological Treatments. Solutions of imidacloprid, neostigmine bromide, and edrophonium chloride (Sigma-Aldrich) were made in Drosophila Ringer's solution (130 mM NaCl/4.7 mM KCl/1.8 mM CaCl₂/1.7 mM KH₂PO4/0.35 mM Na₂HPO₄, pH 7.0). Flies were tethered for electroretinogram (ERG) recordings as described below and a drop of drug (or control) solution large enough to cover the entire surface was gently applied to the fly eye (9). The procedure for delivery of volatile anesthetics has been reported (4), as has the procedure for recording an ERG under anesthesia (10).

Electrophysiological Recordings. ERGs were typically taken from 2- to 7-day-old females that had been raised at 25°C, collected under a brief exposure to CO₂, and granted an overnight recovery period. The flies were tethered without recourse to anesthesia, by using either vacuum (10) or confinement to a truncated micropipette tip, a simplified version of the technique developed by Carlson and colleagues (11). Flies were dark-adapted for 10 min and stimulated by a train of five pulses of orange light (either 3 sec or 200 msec in duration), delivered at 15-sec intervals; the traces were averaged and analyzed as described below. For intracellular recordings, we followed procedures used by the Pak and Montell laboratories (12, 13). In brief, 2- to 7-day-old females that had been raised at 19°C were momentarily exposed to CO₂ and placed sidewise onto double-stick foam tape; the legs and wings were pressed onto the tape, and the head was fixed to the tape with a drop of Loctite Super Glue. Then, by using a pair of 25-gauge needles as scalpels, a small hole (covering <20% of the surface of the eye) was made in the dorsal posterior edge of the cornea. After covering the hole with vacuum grease (Dow-Corning), a glass electrode filled with 2 M KCl (50-75 MΩ resistance) was lowered into the retina. Penetration of a PR was marked by the appearance of a potential difference vs. a low-resistance glass reference electrode that was filled with Ringer's solution and placed in a drop of electrode cream applied to the thorax. The best estimate of the resting membrane potential (RMP) was evaluated by comparing the potential at the end of an experiment with the value obtained when the recording electrode was removed from the eye and placed in a drop of Ringer's solution in contact with the reference electrode. After penetration, flies were dark-adapted for 5 min and stimulated with single pulses of orange light (18 mW/cm² at the surface of the eye). Acceptable recordings were judged as those in which the RMP was less than -20 mV and the light-induced potential (LIP) (defined below) at the end of a 3-sec light pulse was >10 mV.

To quantify ERG and intracellular recordings, baseline voltages were measured just before lights-on (V₁) and just before lights-off (V₂) as before (10); the difference defined the LIP. The kinetics of photoresponse termination were routinely quantified as the time after lights-off when the measured voltage fell to LIP/2.

Results

A convenient way to monitor the performance of the visual system of Drosophila is provided by the ERG, a record of extracellular potentials measured at the surface of the eye in response to a pulse of light (14). In a typical ERG, the potential that is maintained for the duration of the light pulse and decays thereafter is thought to reflect currents arising in the PRs, whereas the rapid transients at the beginning and the end of the light pulse represent currents arising in the optic lobe (OL). Most ERG studies use a standard light pulse with a 2- to 3-sec duration. However, we recently showed that the sensitivity of the off-transient to various treatments varies with the length of the light pulse (10), implying that the operational circuitry for visual processing changes during a standard pulse. To explore the range of this variability, we compared the effects of several genetic and pharmacological manipulations on short-pulse and long-pulse ERGs. What emerged was a differential dependence of photoresponse termination on events in the underlying OL.

Effects on the ERG of Genetic Disruption of Synaptic Communication from the Retina to the Brain. The axons of insect photoreceptors communicate with second-order neurons of the OL by releasing histamine from their terminals (15). The importance of such signaling is evidenced in ERGs by the absence of on and off transients in recordings made from mutants that fail to synthesize histamine as a result of a defective histidine decarboxylase (Hdc) gene (6). The same is true for mutants that fail to respond to the neurotransmitter as a result of a defective ora transientless (ort) gene, which encodes a subunit of the histamine-gated chloride channel (16). We have confirmed these observations, but, in addition to the previously reported phenotype, we find that these mutants have a problem with the termination process. Specifically, after a 3-sec light pulse, the return to baseline tends to be slower in the ERG of mutant flies than in that of control flies. Fig. 1A shows a representative example from a classical mutant, ort^US2515, and a standard wild-type strain, Canton-S. Immediately after lights-off, both the wild-type and mutant ERGs start a sharp recovery, but soon the mutant lags behind the wild type. This effect can be quantified by measuring the time taken for the ERG to decay toward the baseline value (17). Such measurements made on a collection of ERGs from individual flies reveal that, although there is some overlap in the values, the wild-type and mutant records are clearly distinct (Fig. 1B). To confirm that the phenotype is a consequence of a defective ort gene and not the result of an adventitious difference between the strains, we examined a recently isolated mutant (ort^e02083) and its isogenic control. Again, both the representative examples and the quantitative measures from a collection of ERGs show that mutant flies have slower termination kinetics (Fig. 1). ERGs from two Hdc mutants also have an initial fast recovery phase followed by a return to baseline that is slower than that of control flies (Fig. 3 and data not shown).

Fig. 1.

Comparison of ERGs from wild type and ort mutants. (A) Representative ERG traces made by using a 3-sec light pulse. (Left) The ort^US2515 mutant is compared with a Canton-S control [ort⁺(CS)]; both strains are white-eyed because they carry the w¹¹¹⁸ mutation. (Right) The ort^e02083 mutant is compared with an ort⁺ isogenic control [ort⁺(iso)]; both strains are orange-eyed because they carry a miniwhite transgene in addition to the w¹¹¹⁸ mutation. To assist comparison of the ERG decay phase, the voltage scale has been normalized so that the height of each trace is the same at the moment of lights-off. (B) Time required for the decay of the ERG from the value at the moment of lights-off to a value that is halfway to the baseline. Decay times for the indicated strains after a 3-sec light pulse (Left) and comparable values after a 0.2-sec light pulse (Right) are shown. The decay time values are presented as a box plot in which the box encompasses the data points that are ±25% around the median value (shown by a horizontal line). Vertical error bars encompass data points that lie 1.5 box heights above and below the box limits, and outliers are shown individually as open circles. Number of flies tested: ort^US2515, 6; ort⁺ Canton-S, 6; ort^e02083, 7; ort⁺ isogenic control, 7. SP, short pulse; LP, long pulse.

Fig. 3.

The effect on the ERG of pharmacological and genetic manipulation. (A-D) Wild-type Canton-S flies treated with various drugs. (A) Ringer's solution control. (B) Neostigmine, 162 μM. (C) Imidacloprid, 1 nM. (D) Enflurane, 0.43%. ERGs were recorded either 10 min after Ringer's solution or a cholinergic drug was applied to the surface of the eye or after a 1-hour exposure of the whole animal to humidified air containing the volatile anesthetic. (E-H) Genetic manipulation. Flies of genotype Cha-Gal4/+;UAS-shi^ts1/+ were raised at 19°C and either tested at that temperature (E) or shifted to 30°C for 10 min before recording (F). For comparison are shown ERGs recorded at 25°C from flies carrying the Hdc^JK910 mutation (G) or the arr2⁵ mutation (H). All ERG traces have been normalized as in Fig. 1.

Given the consistent effect on termination kinetics of the ERG phenotype seen with 3-sec light pulses, it is noteworthy that this phenotype is much diminished after a 0.2-sec pulse of light. This can be seen in the plots of decay times, in which the mutant values for short pulses tend to be both smaller than those from long pulses and closer to values for control flies (Fig. 1B). More impressive is the comparison of values from individual flies. For the mutants, in 12 of 13 cases the decay time after a short pulse is smaller than that recorded from the same fly after a long pulse. In contrast, this is true for only 3 of 13 wild-type flies. Comparison of short- and long-pulse ERGs from Hdc mutants reveals the same pattern (Fig. 3 and data not shown). The possibility that the reduced phenotype implies that a different neurotransmitter is effective after short pulses is ruled out by the fact that both ort and Hdc mutants lack transients under these conditions. To define the time at which synaptic communication is important for the termination phase, we used light pulses of intermediate duration. We found marginal differences between the ort^e02083 mutant and its isogenic control with light pulses of 0.5 sec but clear slowing of the mutant relative to the wild type with 1-sec pulses (data not shown); the same was true for the Hdc^JK910 mutant. Thus, our experiments show that a change in visual processing occurs after a few hundred milliseconds of exposure to light, a change that makes photoreceptor signaling to the OL important for termination of the photoresponse.

Intracellular Recording from Photoreceptors of Wild-Type and Mutant Flies. What is the source of the currents that persist after the end of a long light pulse in the synaptic mutants? On the one hand, the simplest explanation is that the currents come from PRs, which are unable to promptly terminate the photoresponse in the mutant lines. On the other hand, because the ERG is recorded extracellularly, it is also able to pick up currents generated in pigment cells, neurons from the OL, etc. One must thus entertain the hypothesis that currents from these other cell types are responsible for the altered ERG and the PR response is normal. If this were the case, one would predict that intracellular recordings from PRs would be unaffected by mutations that interfered with synaptic signaling. We have tested this possibility and ruled it out.

Specifically, we used microelectrodes to record the light response from PRs of two ort mutants and two control strains. As seen by the termination metric from a collection of recordings from individual flies (Fig. 2A), after a 3-sec light pulse about half of the mutant records have termination kinetics that are clearly outside the range obtained from wild-type flies. Representative examples of wild-type records and mutant records with slow kinetics are shown in Fig. 2B. We also note that, after short light pulses, the termination kinetics of the mutant collection differ from those of controls, but the phenotype is less prominent than after long pulses. Indeed, with the exception of a single outlier, the mutant flies that display slow termination kinetics after a long pulse have a faster repolarization after a short pulse. Thus, although the mutant phenotype displayed in intracellular recordings is of limited penetrance (an observation we discuss below), it recapitulates the essential features seen with ERG traces. Most importantly, as judged by the resting membrane potential (RMP) and by the magnitude of the change in this potential induced by a 3-sec pulse of light (LIP), wild-type PRs and both fast and slow subpopulations of mutant PRs have similar fundamental physiological parameters. Specifically, for these three groups the mean value ± SEM of the RMP is, respectively, -43.9 ± 4.2, -38.4 ± 6.6, and -49.7 ± 8.5, and the corresponding LIP values are 15.2 ± 1.2, 13.7 ± 1.0, and 12.5 ± 0.7. Thus, the mutant phenotype is not merely a reflection of a globally damaged PR but is a specific change in the response profile. Although we had difficulty in obtaining recordings from Hdc mutants that were comparable in basic physiological parameters with their controls, a similar effect on termination kinetics was seen (data not shown). We conclude that at least part of the ERG defect of ort and Hdc mutants reflects altered termination of the photoresponse in PRs. Because these mutations disrupt synaptic communication between PRs and the second-order neurons, we infer that feedback from the OL to the PRs influences the termination process.

Fig. 2.

Intracellular recordings from PRs of wild-type and mutant flies. (A) Time required for the decay of the intracellular potential from the value at the moment of lights-off to a value that is halfway to the baseline. Decay times for the indicated strains after a 3-sec light pulse (Left) and comparable values after a 0.2-sec light pulse (Right) are shown. Each circle presents the value calculated from a trace of an individual fly; a second trace from the same fly (data not shown) gave a comparable value. The strains used were the same as those in Fig. 1 except that the isogenic control for the ort^e02083 strain did not carry a miniwhite transgene and was thus white-eyed. Number of flies tested: ort^US2515, 7; ort⁺ Canton-S, 6; ort^e02083, 8; ort⁺ isogenic control, 7. LP, long pulse; SP, short pulse. (B) Representative intracellular traces made by using a 3-sec light pulse. (Upper) A record from each of the two ort⁺ strains. (Lower) A record from each of the two mutant lines. The examples shown are from mutant flies with a decay time typical of the slow population. Roughly half of the records from the mutant lines were indistinguishable from the wild-type records and are not shown. For each trace shown, the vertical scale has been normalized as in Fig. 1. For the two control flies shown, the unnormalized value of the LIP was 16.6 and 13.9 mV; for the two ort mutants, it was 11.0 and 12.5 mV.

Testing the Feedback Hypothesis. If activity in the fly brain influences events in the retina, alteration of brain neurotransmitters should have a corresponding effect on the ERG. To test this idea, we manipulated the cholinergic system, whose neurotransmitter is synthesized in the brain but not in the retina (18). Interference with cholinergic transmission indeed produced effects on the shape of the ERG at the end of a 3-sec light pulse. In one series of experiments, we applied cholinergic drugs to the surface of the eye, relying on fluid movement down the bristle shaft for prompt delivery to the synaptic region (9). Although application of diluent was without effect (Fig. 3A), when the acetylcholinesterase inhibitor neostigmine (Fig. 3B) or edrophonium (data not shown) was applied to the surface of the eye, within minutes the long-pulse ERG developed a slowed return to baseline. Application of the cholinergic agonist imidacloprid had a similar effect (Fig. 3C). The traces shown in Fig. 3 are representative of a larger set whose half-decay times (measured as in Fig. 1B) were as follows (mean ± SEM): control flies treated with Ringer's solution, 0.36 ± 0.06 sec (n = 7); neostigmine-treated flies, 0.82 ± 0.11 sec (n = 16); and imidacloprid-treated flies, 1.52 ± 0.23 sec (n = 7). Although all these drugs nominally increase cholinergic transmission, they also may exhaust or desensitize the cholinergic system (19). That such depression is responsible for the drug effects on the ERG is suggested by the fact that similar effects are produced when release of acetylcholine from presynaptic terminals is acutely depressed. To inhibit release, we used the system developed by Kitamoto (8), in which a shift to a nonpermissive temperature instantaneously disables cholinergic exocytosis. Compared with the ERGs of controls maintained at the permissive temperature (Fig. 3E), the cholinergically disabled flies had a substantially slowed return to baseline after a 3-sec light pulse (Fig. 3F). It should be noted that, as illustrated in Fig. 3, all of the cholinergic manipulations had much smaller effects on ERGs that were elicited with short pulses of light. This pattern is, of course, reminiscent of that obtained with ort mutants (Fig. 1) and Hdc mutants (Fig. 3G) but is distinct from that seen with mutants in genes known to directly affect the PR termination process, like arr2 or inaC. In these cases, termination of the ERG is comparably slow after both long and short pulses (Fig. 3H and data not shown). This difference not only eliminates the possibility that the different termination kinetics seen in the synaptic mutants represent an instrumental artifact but, as discussed below, also sets limits on the possible mechanism(s) by which synaptic crosstalk between the retina and the brain might affect termination of the photoresponse.

It must be noted that, under the conditions we used, interference with cholinergic signaling has little or no effect on the potential that is maintained for the duration of the light pulse, indicating that the phototransduction process itself is intact. Furthermore, the on-transient is largely unaffected by these manipulations, showing that photoreceptors are able to communicate with neurons in the underlying lamina. However, the off-transient is reduced in amplitude, especially after a long light pulse (Fig. 3). It is hard to decide whether the diminution of the off-transient is merely a consequence of inefficient repolarization of PRs (10) or whether cholinergic signaling plays a more active role in the generation of this feature. Nevertheless, the consistent effects of cholinergic interference on the ERG return to baseline after long light pulses support the hypothesis that retrograde signaling from the brain influences the termination of the photoresponse.

In higher organisms, general anesthetics are well known to be potent inhibitors of synaptic transmission (20). We previously showed that these agents do not affect the on-transient, the most immediate reflection of histaminergic signaling from PRs to large monopolar cells (10). However, anesthetics might produce effects at other places or at other times in the feedback circuit. Indeed, ERGs from flies exposed to the volatile agent enflurane closely resemble those from flies with defects in retrograde signaling (Fig. 3D). All anesthetics tested produced similar effects (data not shown), but their potencies, measured relative to those in locomotor assays (21), varied. This divergence suggests that events at the end of a long light pulse could provide a useful tool for exploring distinct contributions to the anesthetic spectrum in Drosophila.

Discussion

Because of its importance to the functioning of the visual system, photoresponse termination has been extensively studied (3, 22). Whereas the focus of past work has been on processes that are intrinsic to the PRs, our work demonstrates that higher-order neurons also influence the speed with which Drosophila PRs return to the resting state. This influence requires transmission of visual information from the retina to the OL and thus appears to represent retrograde signaling from the brain to PRs.

The basic tool for the bulk of our observations was electroretinography. The consistent result was that manipulations that interfered either with cholinergic signaling generated in the brain or histaminergic signaling generated in the eye yielded ERG traces with a decidedly slow return toward the baseline. Although the ERG is a simple and reliable method, it cannot by itself distinguish the cellular source of each component of the waveform. To test whether the signal that lingers after lights-off comes at least in part from PRs, we turned to the much more demanding technique of in situ intracellular recording from these cells. Although the ort gene, which is needed for proper reception of the histamine signal, is not expressed in PRs, we found that intracellular recordings from ort mutants were not normal. Specifically, about half of the recordings showed dramatically slowed termination kinetics. We do not know why this phenotype was not seen in every impaled cell. One possibility is that the requirement for feedback is stochastic, with some PRs endowed with a termination apparatus that is sufficiently robust to endure the lack of a retrograde signal. An alternative scenario claims that feedback is normally a universal requirement, but a subpopulation of individual PRs has been able to compensate for its lack. Also under consideration is the possibility that, although all impaled cells have comparable basic physiological parameters, subtle differences in the impalement have revealed in some cells a weakness that is common to all of the mutant PRs. Despite our uncertainty about the basis of its partial penetrance, the mutant phenotype seen with intracellular recordings does show that communication from PRs to the brain influences the process by which PRs recover from a light pulse. Because such recovery can be closely coupled to the process by which PRs adapt to ambient light (23), it is possible that retrograde signaling will also be found to influence the latter.

Our work prompts the traditional questions of who, where, when, and how about the retrograde signal. Its identity remains to be determined because, even though the cholinergic system is required for feedback, we cannot distinguish whether acetylcholine itself is the retrograde messenger or whether cholinergic transmission is required in order for another transmitter/modulator to be released from higher-order neurons. Additionally, we can only speculate about the location to which the retrograde signal is delivered. Feedback synapses exist between second-order interneurons and PR axons (24), and our observations might provide a physiological rationale for this anatomic feature. However, nonsynaptic delivery of a small molecule from the OL to the PR axons or even to the rhabdomeres cannot be ruled out. Another unresolved issue is whether the signal is delivered only in response to a light flash or whether it is released tonically from the OL. The latter hypothesis is limited by our observation (S.R., unpublished work) that the shape of the ERG of wild-type flies is unaffected by long periods of dark adaptation. If tonic retrograde signaling occurs, it must therefore rely on a spontaneous background level of histaminergic transmission. Finally, there is the issue of how the retrograde signal affects the photoresponse. Here, a key observation is that the importance of such signaling depends on the duration of the light pulse, having a much smaller effect on termination with short (≈0.2 sec) than with longer pulses (1-3 sec). An obvious distinction between these two regimens is the amount of calcium that enters the PR, suggesting that signaling serves to counteract the effects of prolonged light-induced calcium entry.

Despite the fact that much remains to be learned, our observations place limits on some candidate mechanisms. For example, altered ERGs can be seen a few minutes after manipulation of the cholinergic system. This argues against effects of feedback on much slower processes that can influence termination, such as the translocation of arrestin from rhabdomeres to cell bodies (17, 25). Nor is feedback unconditionally needed for the activation of arrestin. If it were, loss of feedback would mimic a loss-of-function mutation in the arrestin gene. However, the restoration of the resting state after a short pulse of light is much more severely affected by an arrestin mutation than by interference with communication between the retina and the OL. If feedback acts on the termination mechanism itself, it must be through a more subtle change than preventing complete inactivation of arrestin. Moreover, scenarios in which loss of feedback produces a partial reduction of arrestin function are severely restricted by the fact that arr2/+ heterozygotes have normal termination kinetics (R.L.S., unpublished data). Of course, feedback might act on processes in the PRs other than termination of the phototransduction cascade. One attractive possibility involves the regulation of the ion channels that help to repolarize PRs after light-induced currents have ceased (23). Indeed, it has been shown that serotonin, which may be released from nonsynaptic varicosities just below the retina (26), modulates currents carried by Shaker channels (27). However, it seems unlikely that modulation of Shaker channels is the principal feedback mechanism because, as judged by the shape of the ERG, the termination kinetics after a long light pulse are unaffected by genetic and pharmacological inactivation of Shaker (10). If feedback works through a channel mechanism, affecting currents arising in the PR cell body and/or in the PR axons that invade the OL, it would have to be through activation of leak channels or other members of the voltage-gated potassium channel superfamily.

Regardless of uncertainties about its cellular basis, the phenomenon we have described adds significantly to the repertory of known feedback mechanisms onto primary sensory elements. In the visual system of vertebrates and invertebrates, feedback is used to sharpen the visual field (28, 29). Another example of the brain influencing sensory function is through circadian output to PRs (30). In these examples, as well as others involving feedback onto mechanosensory neurons (31), the fundamental effect of retrograde signaling is to control sensitivity of the primary afferent to its stimulus or to control tonic output from these neurons. In contrast, the feedback we have described seems to be focused on the temporal domain, changing the speed with which sensory cells recover.