Rod signaling in primate retina: range, routing and kinetics

Stimulus or context dependent routing of neural signals through parallel pathways can permit flexible processing of diverse inputs. For example, work in mouse shows that rod photoreceptor signals are routed through several retinal pathways, each specialized for different light levels. This light level-dependent routing of rod signals has been invoked to explain several human perceptual results, but it has not been tested in primate retina. Here we show, surprisingly, that rod signals traverse the primate retina almost exclusively through a single pathway, regardless of light level. Indeed, identical experiments in mouse and primate reveal large differences in how rod signals traverse the retina. These results require reevaluating human perceptual results in terms of flexible computation within this single pathway. This includes a prominent speeding of rod signals with light level – which we show is inherited directly from the rods photoreceptors themselves rather than from different pathways with different kinetics.


Introduction
identified above? The results described below show, surprisingly, that the rod bipolar pathway is the dominate route for rod-derived RGC responses across the full range of 142 rod signaling. ganglion cells in primates, regardless of light level. This conclusion is, however, subject 177 to the caveat that NBQX will impact retinal signaling in multiple ways.

179
To further test the hypothesized dominance of the primary rod pathway, we made 180 current-clamp recordings from H1 horizontal cells. H1 horizontal cells receive direct 181 synaptic input from L-and M-cones but not from rods ( Figure 3A; (Kolb, 1970;Rodieck, 182 1998;Verweij, Dacey, Peterson, & Buck, 1999)); thus, these cells provide a readout of 183 rod signals in the synaptic output of cones without the need for pharmacology. H1 184 responses, like the NBQX-insensitive responses of AII amacrine cells, were weak at low 185 backgrounds but became pronounced at backgrounds ≥300 R*/rod/s ( Figure 3D,E). In a subset of horizontal cell recordings, we also compared responses to short-wavelength supplement 2). At backgrounds below 300 R*/rod/s, responses to short wavelength flashes had substantially slower kinetics than responses to long wavelength flashes, consistent with previous work showing that rod-cone gap junctions can transmit rod very similar kinetics, suggesting that at these light levels both responses originated in the cones. Responses of cones were very similar to those of horizontal cells, with the 196 emergence of a sizeable response only for backgrounds ≥300 R*/rod/s ( Figure 3E,F).

209
The dominance of the primary rod pathway in primate was unexpected, as previous the secondary pathway were half maximal at ~5 R*/rod/s, nearly two orders of 218 magnitude lower than the half-maximal intensity of secondary-pathway rod signals in 219 primate. Thus, the secondary pathway in mouse begins contributing to retinal output at 220 light levels that are more than 100-fold below rod saturation, whereas in primate signals 221 from the secondary pathway are largely absent below rod saturation.

225
We next considered whether continuous stimuli might more effectively elicit signals 226 through the secondary pathway than the brief flashes used thus far. Sinusoidal stimuli 227 are of particular relevance since they are used in many of the human perceptual studies 228 that motivate the light level-dependent rod routing hypothesis. We started by recording 229 excitatory synaptic input to an On parasol RGC in response to short-(rod-preferring) and 230 long-(cone-preferring) wavelength sinusoidal stimuli at a mid-to-high mesopic light level 231 (10 R*/rod/s). The contrasts of the short and long wavelength stimuli were adjusted so 232 that when modulated individually at 2 Hz they produced roughly equal amplitude 233 modulation of the RGC's excitatory inputs. These contrast amplitudes were then held fixed as we explored responses to a range of temporal frequencies ( Figure 5A,B). As measured responses to the same stimuli (amplitude and background) used in the 239 parasol recordings. Across all temporal frequencies probed, the ratio of H1 responses to 240 rod-and cone-preferring stimuli was at least 10 times smaller than that for On parasol 241 responses to the same stimuli ( Figure 5C,D). Hence, rod-derived signals in the 242 secondary pathway are too weak to explain rod-derived RGC responses. Instead, the 243 primary rod pathway dominates responses to both sinusoidal and flashed stimuli.

247
Our results thus far suggest that the primary rod pathway continues to convey rod 248 signals to On parasol ganglion cells even when the rods are approaching saturation.

249
Does the primary pathway also dominate responses in Off retinal circuits? Rod signals 250 could reach Off cone bipolar cells from three known sources ( Figure 1): 1) dendritic input 251 directly from rods (i.e. tertiary pathway), 2) dendritic input from cones (i.e. secondary 252 pathway), and 3) axonal inhibitory input from the AII amacrine cell (i.e. primary pathway).

253
The differences between On and Off circuits suggests that both the relative weighting of 254 rod-and cone-derived signals and the routing of rod signals could differ.

256
We first compared the relative weighting of rod-and cone-derived signals in the 257 responses of On and Off parasol and midget RGCs at a mean light level of 20 R*/rod/s 258 ( Figure 6A,B,C). Like the experiments in Figure 5, we began by adjusting the contrasts 259 of rod-and cone-preferring stimuli so that they produced equal amplitude responses in 260 an On parasol RGC ( Figure 6A,C). After achieving a match, we presented these 261 response-equated stimuli while recording from other RGC types in the same piece of 262 retina. In Off RGCs, responses to rod-preferring stimuli were roughly half as large as 263 responses to cone-preferring stimuli ( Figure 6B,C). This indicates in turn that rod signals 264 are less strongly routed through Off cone bipolar cells than On cone bipolar cells at this 265 light level. If the secondary pathway (rod-cone electrical coupling) dominated, rod-and 266 cone-derived signals would be mixed prior to transmission to On and Off bipolar cells, 267 leading to similar weighting in On and Off circuits. Thus, this On/Off asymmetry supports 268 our conclusion that the secondary rod pathway does not convey strong rod-derived 269 signals in primate.

271
Alternatively, rod-derived responses in Off RGCs could arise from the primary pathway 272 via rod bipolar and AII amacrine cells or through the tertiary pathway via direct rod input 273 to Off cone bipolar cells ( Figure 6D). To distinguish between these possibilities, we used

293
We repeated the LY/APB experiments in mouse retina to provide a direct comparison 294 across species ( Figure 6G,H). At a background of 0.5 R*/rod/s, LY/APB reduced 295 excitatory inputs to Off alpha RGCs by ~90%, indicating that the primary pathway 296 dominates signaling at this light level. At a background of 5 R*/rod/s, LY/APB reduced 297 responses by 35%, and at a background of 50R*/rod/s, responses were little affected.

298
This is consistent with the results from Figure 4 that show that sizable rod signals reach 299 AII amacrine cells through On cone bipolar cells in mouse retina, and correspondingly 300 that the secondary pathway conveys significant rod signals at mesopic light levels.

302
Collectively, the experiments in Figures 3-6 indicate a surprising difference between how 303 rod signals traverse the mouse and primate retinas. Specifically, unlike the situation in 304 mouse, the primary rod pathway provides the dominant route that rod-derived signals 305 take through the primate retina across scotopic and mesopic light levels. This 306 dominance of the primary pathway means that perceptual phenomena previously 307 attributed to a change in routing need to be reinterpreted (e.g. section on kinetics below); 308 it also maximizes opportunities to independently process rod and cone signals since 309 they are not mixed until late in the retinal circuitry.

313
Perceptual experiments show that the kinetics of rod-derived signals speed relative to 314 cone-derived signals as light levels increase (Sharpe et al., 1989). This speeding is often 315 attributed to a luminance-dependent change in the dominant route that rod-derived 316 signals take through the retina (reviewed by (Buck, 2004;Sharpe & Stockman, 1999; that such rerouting does not occur. Instead, as described below, the shift in kinetics of 319 rod-derived signals appears to originate within the rods themselves.

321
We first determined whether responses of RGCs under our experimental conditions 322 exhibited kinetic shifts similar to those observed perceptually. Spikes ( Figure 7A) and

357
The speeding of the rod voltage response coincided with a large steady-state 358 hyperpolarization of the rod membrane potential (9.0± 0.7 mV upon a step from 1 to 60 359 R*/rod/s, mean ± SEM, n=10; Figure 7G). To test whether this hyperpolarization could

367
These experiments indicate that the shifts in kinetics of the rod-derived retinal outputs 368 are largely inherited from the rods themselves rather than a light-dependent shift in the

403
This speeding of rod signals has been attributed to a shift in routing of rod signals from the (presumed slow) primary pathway to the (presumed fast) secondary or cone bipolar 405 pathway. Experiments in non-primate retinas, particularly mice, support this hypothesis 406 (Soucy et al., 1998;Deans et al., 2002;Trexler et al., 2005). The work described here

450
Hamburyan and colleagues found that rods could no longer modulate RGC output at 451 backgrounds at or above 10 4 R*/rod/s. It will be interesting to see if slow adaptation 452 mechanisms found in mouse rods are also present in primates.

454
Linking neural circuits to perception

456
Interactions between rod and cone signals affect many aspects of mesopic vision 457 (reviewed by (Buck, 2014;Stockman & Sharpe, 2006)). An important factor controlling 458 these interactions is the relative timing of rod and cone signals. This relative timing shifts 459 considerably between low and high mesopic light levels as rod signals speed relative to

499
Perforated patch clamp recordings from cone photoreceptors were performed in current 500 clamp using electrodes (9-11 MΩ) containing (in mM): 115 potassium aspartate, 1 added to the internal solution at 30μg/mL. Upon sealing on a cell, access was monitored 504 by tracking the membrane potential as well as the response amplitude to a constant 505 amplitude probe flash. Once access equilibrated (~5 -25 minutes), recordings were 506 started. Throughout perforated patch recordings, the membrane potential and response 507 to a reference flash were monitored to ensure that electrical access to the cell remained 508 stable.

510
For suction recordings, a suspension of finely chopped retina was transferred to a 511 recording chamber, pieces of retina were briefly allowed to settle, then perfused (2-3 512 ml/min, 32 ± 1 °C). Suction electrodes (3-4 MΩ, tip inner diameter of ~1.6μm) were filled 513 HEPES-buffered Ames and were voltage clamped at 0 mV. Individual rod outer 514 segments were drawn into the recording pipette under gentle suction.

522
Cell selection criteria 523 524 Prior to recording data from all cells except photoreceptors, the sensitivity of a given 525 piece of retina was estimated from responses of On parasol RGCs. Two criteria were 526 used to determine whether to continue with data collection (Ala-Laurila and Rieke, 2014; which these criteria were met. In suction electrode recordings, rod photoreceptors were selected for recording and analysis if their outer segment had not been obviously 533 damaged by the suction procedure and if they had maximal light responses exceeding 534 20 pA for primate and 8 pA for mouse. In whole-cell recordings, rods were retained for 535 analysis if they exhibited > 8 mV changes in resting potential for light steps from 1 to 60 536 R*/rod/s. In all cases, data from any recorded cells in which these criteria were met was 537 retained and reported in relevant figures.

539
We collected data generally from 5-10 cells for each experimental manipulation; this 540 target population size was based on checking similarity of effects across cells rather 541 than a pre-experiment estimate of the population size required for statistical significance.

554
Rods were selectively stimulated using the 405 nm LED. To test for cone contribution to 555 these responses, we monitored the ratio of On parasol responses elicited by the 520 and 556 405 nm LEDs. This ratio is predicted to differ by a factor of ~2.4 for responses elicited by 557 rods and M cones based on the rod and M cone spectral sensitivities (On parasol cells 558 do not get significant S cone input (Field et al., 2010)). The ratio increased noticeably at 559 100-300 R*/rod/s, and increased by a factor of 1.8 at 300 R*/rod/s. Together with the 560 theoretical factor of 2.4, this indicates that responses to the 520 nm LED represented a 561 ~60% contribution from cones and ~40% from rods at 300 R*/rod/s. The ratio of rod 562 sensitivity to cone sensitivity to the 405 nm LED is predicted to be 2.4-fold higher than 563 that for the 520 LED, indicating that ~25% of the response to the 405 nm LED at 564 300R*/rod/s originated in cones. Cone contributions were substantially smaller at lower 565 background light levels.

569
For spike recordings from ganglion cells, we detected spike times and compiled them 570 into peristimulus time histograms as previously described (Murphy and Rieke, 2006).

571
Assuming linearity, rod responses to sinusoidal stimuli (Figure 7) were predicted by 572 convolving the measured flash responses with sinusoids of the various frequencies.        In the primary pathway rod signals are routed through dedicated rod bipolar cells to AII amacrine cells. AII amacrine cells in turn transmit 'On' signals to On cone bipolar cells through dendro-axonal gap junctions and 'Off' signals to Off cone bipolar cells through glycinergic synapses (not shown, but see C). Cone bipolar signals are subsequently transmitted to retinal ganglion cells. B) In the secondary pathway, rod transmit signals via gap junctions to cone axons, and hence the associated cone circuitry (Kolb, 1977;Schneeweis & Schnapf, 1995b;Deans et al., 2002;Hornstein et al., 2005). C) In the tertiary pathway, rods transmit signals directly to Off cone bipolar cell dendrites (Soucy et al., 1998;Hack, Peichl, & Brandstatter, 1999;Tsukamoto et al., 2001b).  In the presence of dim (3 R*/rod/s) and moderate (30 R*/rod/s) backgrounds, stimuli that produce sub-saturating responses in the RGCs (scaled) are almost entirely blocked by NBQX in AII amacrines. At a background of 300 R*/rod/s (i.e. above rod saturation) flash responses in AIIs are largely insensitive to NBQX application.    : Rod-signals generated by sine-wave stimuli are also restricted to the RB pathway. A) Excitatory synaptic input recorded from an On parasol RGC in response to sinusoidally-modulated short (rod-preferring; blue traces) and long (cone-preferring; red traces) wavelength stimuli. Contrasts were adjusted to produce equal modulation at 2 Hz, and were then held fixed for all subsequent recordings (e.g. different frequencies, horizontal recordings). B) Population data from On parasol RGC recordings; response modulation versus stimulus frequency. C) Physiological response of an H1 horizontal cell (from the same retinal mount) to short and long wavelength stimuli for the same contrast used for the parasol cell in A. D) Population data from H1 horizontal recordings; response modulation versus stimulus frequency.  Figure 6: Rod signals in the tertiary rod pathway are weak. A) Excitatory synaptic input recorded from an On parasol RGC in response to sinusoidally-modulated short (rodpreferring; blue traces) and long (cone preferring; red traces) wavelength stimuli. Contrasts were adjusted to produce equal modulation at 2 Hz, and were then held fixed for subsequent recordings from other cell types (e.g. Off parasol RGC). B) Excitatory synaptic input recorded from an Off parasol RGC from the same retinal mount as A and in response to the same stimuli. C) Relative weighting of rod and cone signals in excitatory inputs to On and Off RGCs. D) Schematic of the primary and tertiary rod circuits that influence Off cone bipolar signaling and the actions of the mGluR6 agonist/antagonist mixture LY/APB. E) Rod signals (20 R*/rod/s) in excitatory inputs to an Off parasol in control conditions and after suppressing activity in all On bipolar cells with an mGluR6 agonist/antagonist cocktail (LY/APB, see Methods). F) Response ratio (cocktail:control) across cells as a function of mean luminance. Data are plotted as mean±SEM. G) Rod signals (5 R*/rod/s) in excitatory inputs to an Off alpha RGC (mouse) in control conditions and after suppressing activity in all On bipolar cells with an mGluR6 agonist/antagonist cocktail (LY/APB, see Methods). H) Response ratio (cocktail:control) across cells as a function of mean luminance. Rod, but not cone, signals in excitatory inputs to Off parasol RGCs arise from the primary rod pathway under mesopic conditions. Example recording of excitatory synaptic input to an Off parasol RGC in response to 2 Hz sine wave modulation of a short (rod-preferring) or long (cone-preferring) wavelength LED. Responses to short, but not long, wavelength light modulations were largely eliminated by a mixture of LY341495 and APB (see Methods).