Abstract
Purpose.:
To better understand how photoreceptors and their circuits support luminance-dependent spatial visual behavior.
Methods.:
Grating thresholds for optokinetic tracking were measured under defined luminance conditions in mice with genetic alterations of photoreceptor activity.
Results.:
The luminance conditions that enable cone- and rod-mediated behavior, and the luminance range over which rod and cone functions overlap, were characterized. The AII amacrine pathway was found to support low-resolution and high-contrast function, with the rod–cone pathway supporting high-resolution and low-contrast function. Rods alone were also shown to be capable of driving cone-like spatial visual function, but only when cones were genetically maintained in a physiological dark state.
Conclusions.:
The study defined how luminance signals drive rod- and cone-mediated spatial visual behavior and revealed new and unexpected contributions for rods that depend on an interaction between cone and rod systems.
The mammalian visual system is able to adaptively process spatial visual signals over a wide luminance range.
1 This ability depends in part on the complementary irradiance sensitivities of cone and rod photoreceptors,
2,3 with cones active under bright (photopic) illumination, rods active under dim (i.e., scotopic) illumination, and combined cone and rod activation in intermediate lighting conditions (i.e., mesopic
4,5). Retinal circuitry downstream of the photoreceptors processes these luminance signals into spatially organized receptive fields, which give rise to adaptive, visually guided behaviors. Retinal photoreceptor systems make a tradeoff between light sensitivity and spatial resolution to enable this function, with cone circuits having a lower spatial convergence onto bipolar and ganglion cells than rod circuits
6 and superior spatial visual resolution in photopic conditions.
6 In mesopic conditions, interactions between photoreceptor systems in the retina, such as via rods activating cone circuits,
7,8 also shape the spatial characteristics of retinal output in a luminance-dependent fashion.
Cones drive cone bipolar cells, which in turn directly activate retinal ganglion cells (RGCs
9). Rods, however, use two different circuits to convey luminance signals to ON RGCs.
10 In the direct rod circuit, rod cells synapse with rod bipolar cells, which activate AII amacrine cells that connect via gap junctions to cone ON bipolar cells. This pathway is thought to be functional at scotopic light levels.
7 In the indirect rod pathway, rods make electrical synapses with cones, which then use the cone pathways to provide signals to RGCs. The indirect pathway is thought to function at intermediate light levels.
11 Although these circuits have been studied extensively in the isolated retina, how they ultimately contribute to visual behavior has been only recently investigated for non–image-forming functions, such as circadian photoentrainment.
12
In order to understand how photoreceptor systems in the retina contribute to luminance-dependent, spatial visual function, we designed the present study to define the luminance conditions that characterize photopic, mesopic, and scotopic spatial visual behavior in the mouse, and the grating thresholds that define cone- and rod-mediated function. We utilized the virtual optokinetic system (VOS) for these experiments, which enables the measurement of grating thresholds for optokinetic tracking (OKT)—a reflex that is highly quantifiable, is not dependent on the function of visual cortex,
13 is sensitive to manipulations of retinal photoreceptor function,
14 and allows measurement of spatial frequency and contrast thresholds under controlled illumination. Optokinetic tracking measures were made in genetic mouse lines that carried selective deletions of the phototransduction pathways of rods, cones, and intrinsically photosensitive retinal ganglion cells (ipRGCs). An important feature of these mouse lines is that they show no overt signs of retinal degeneration,
12,15–18 which can lead to confounding structural
19 and functional changes
20 in the retina.
This approach enabled us to define the retinal circuits by which rods and cones on their own support spatial visual function. It also facilitated the determination that rods, which typically drive low-resolution and high-contrast function, are capable of driving the high-resolution and low-contrast visual functions typical of cones, depending on the physiological state of cones.
Adult male and female mice of eight different strains on a C57BL/6 background were used. Mice were bred or generated in the pathogen-free vivarium at Mudd Hall at The Johns Hopkins University in accordance with The Johns Hopkins Animal Care and Use Committee regulations. As young adults, mice were transferred to a quarantine facility at Weill Cornell Medical College in New York City for 4 to 6 weeks, where their health was monitored. Once certified healthy, they were transferred to the animal holding facility at the Burke Medical Research Institute. Both Cornell and Burke facilities are AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International) accredited, and experimental procedures were undertaken in accordance with the local Institutional Animal Care and Use Committees and in adherence with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The facilities were maintained at 22°C on a 12-hour light/12-hour dark cycle (lights on from 0600–1800 hours).
Two strains of mice with cone-selective activity were used: cone only ([CO]: Gnat1
−/−; Opn4
−/−16,18) and rod knockout ([RKO]: Gnat1
−/−). The mutations utilized a Gnat1 mutation
16 of the rod-specific transducin. This enabled rod cells to absorb photons but not to amplify the signal. Since transducin is required reduce cGMP and close a cyclic nucleotide-gated channel, rods lacking functional transducin do not hyperpolarize in response to light and remain in a depolarized (dark-like) state. To create CO mice, the transducin mutants were bred into a line lacking melanopsin (Opn4
18), in which the ipRGCs lack the melanopsin-based phototransduction pathway.
Three strains of rod-only (RO) mice were used: ROhypCones (CngA3
−/−; Opn4
−/−, previously known as RO1
12,15); ROdepCones (Gnat2
−/−; Opn4
−/−, previously known as RO2
12,17), and ROΔCones (h.red DTA; Opn4
−/−, previously known as RO3
12,21). For ROhypCones mice, a cyclic nucleotide A subunit (CngA3) mutation rendered cones unable to transduce light
15 by targeting the requisite cone-specific cyclic nucleotide-gated ion channel and eliminating the Na
+/Ca
2+ “dark” membrane current. This leaves cones hyperpolarized and physiologically mimicking a state of constant illumination. A complementary approach was used to produce ROdepCones mice, by rendering cones unable to signal the presence of light because of a missense mutation of the requisite cone-specific transducin, Gnat2.
17 The loss of Gnat2 results in cones maintained in a depolarized or “dark-like” state, similar to the Gnat1 mutant. For ROΔCones mice, cones were physically removed by crossing animals carrying an attenuated form of diphtheria toxin (DTA) under the control of human red-cone opsin with a strain lacking melanopsin.
12 To create RO mice, the three strains of cone mutants were bred with mice lacking melanopsin (Opn4
12). A strain possessing intact photoreceptors (C57BL/6; wild-type [WT]) and strains with melanopsin-only (MO) function (CngA3
−/−; Gnat2
−/−) or with no photoreceptor function (i.e., triple knockout [TKO]; Gnat1
−/−; CngA3
−/−; Opn4
−/−) were used as controls.
Visual thresholds were measured by evaluating OKT in a VOS as described previously.
12,22,23 Vertical sine wave gratings were projected on monitors as a virtual cylinder, which surrounded an unrestrained mouse standing on a platform (
Fig. 1A). The hub of the cylinder was centered between the animal's eyes in real time to fix the spatial frequency of the grating at the animals' viewing position. Rotation of the cylinder (12°/sec) elicited reflexive tracking that was scored via live video (
Fig. 1B). Following dark adaptation (2–10 hours), spatial frequency and contrast thresholds at six spatial frequencies (0.031, 0.064, 0.092, 0.103, 0.192, 0.272 cyc/deg) were generated through each eye separately.
13 Contrast sensitivity is the inverse of the contrast threshold corrected for the output luminance of the computer screens used to present grating stimuli. Contrast thresholds were converted to Michelson contrast using screen luminance values (maximum − minimum)/(maximum + minimum), which are dimensionless. Grating luminance was set by layering combinations of neutral density (ND) filters (0.3, 0.6, 0.9, 1.2 ND; Lee Filters USA, Andover, Hampshire, UK) on the monitor screens. A light meter (ILT-1700; International Light Technologies, Peabody, MA, USA) was able to measure screen luminances (average of the four screens;
Fig. 1C) with up to 4.5 ND, but not in darker conditions. Optokinetic tracking measures under different luminance conditions were made independent of one another; that is, animals were dark adapted on separate days for measures at each luminance. Assessments in low-light conditions were facilitated by infrared lighting on an infrared-sensitive camera (Sony Handycam DCR-HC28; Sony, Tokyo, Japan).
Wild-type mice (
n = 6) readily tracked at maximal illumination (0 ND = 54 lux), generating spatial frequency (0.3918 cyc/deg ± 0.005) and contrast thresholds (maximum sensitivity at 0.064 cyc/deg = 16.26 ± 1.03) comparable to previously published values (
Fig. 1D
22). The spatial frequency threshold declined little (0.3806 cyc/deg ± 0.0037 to 0.3702 cyc/deg ± 0.0021) as luminance was reduced with up to 4.2 ND. However, between 4.2 and 6.3 ND, it declined from 0.3702 cyc/deg ± 0.0016 to 0.1962 cyc/deg ± 0.0021 (
P = 0.012), where it remained until 6.9 ND. At 7.2 ND, no OKT could be elicited (
Fig. 1D). Similarly, the contrast sensitivity function in WT mice was maintained near normal (
Fig. 1E) as filters were added up to 4.2 ND; it then declined precipitously at all spatial frequencies between 4.2 and 6.3 ND, was largely unchanged at 6.3 and 6.9 ND, and was not recordable at 7.2 ND. Based on these results, we hypothesized that OKT function between 0 ND (54 lux) and 4.2 ND (∼1.5 lux) is exclusively mediated by cones, that function between 6.3 and 6.9 ND is exclusively mediated by rods, and that function in a luminance range from 4.2 to 5.4 ND is mediated by both rods and cones.
As negative controls, we investigated whether mice lacking all photoreceptor function (TKO;
n = 8), or rod and cone function (MO;
n = 9) tracked under any luminance. As we have previously shown,
23 the strains did not track (data not shown), confirming that rods or cones are necessary for OKT. Whereas our recent study showed that melanopsin phototransduction and subsets of ipRGCs can modulate the OKT response in the presence of functional rod/cone phototransduction pathways,
24 ipRGCs on their own through the melanopsin phototransduction pathway are not sufficient to drive OKT.
To isolate cone function from rod and ipRGC function, we measured thresholds in double knockout mice (CO: Gnat1
−/−; Opn4
−/−;
n = 9) that left cones as the only functional photoreceptors (referred to hereafter as CO animals
15,18). We found that between 0 and 4.2 ND (
Fig. 2A), the spatial frequency (left) threshold (0.3810 cyc/deg ± 0.0024 to 0.3705 cyc/deg ± 0.030) closely approximated that of WT mice; however, no responses could be elicited at 4.5 ND or darker conditions. Contrast sensitivity measures were similarly affected by changes in luminance: WT-like responses were present from 0 to 4.2 ND, but no tracking could be elicited in darker conditions (
Fig. 2, right). Thus, cones on their own are unable to drive OKT below 2 lux.
To determine whether melanopsin phototransduction plays a role in cone-mediated function, we measured OKT in rod knockout mice (RKO; Gnat1
−/−b16;
n = 9) where cones and ipRGCs are both functional (
Fig. 2B). The spatial frequency threshold of the mice (left) closely resembled CO responses (
Fig. 2A) and was indistinguishable from that of WT animals (0.38675 cyc/deg ± 0.0016 to 0.3696 cyc/deg ± 0.009). The contrast sensitivity responses were similar in CO and RKO animals.
To investigate the luminance conditions that underlie the ability of rod circuitry to drive spatial visual behavior, we utilized a mouse line with cones ablated by the expression of a DTA in the background of a melanopsin deletion (referred here as ROΔCones, previously referred to as RO3 animals
12;
n = 6). ROΔCones mice only have rod function through the AII amacrine cell circuitry. Wild-type–like spatial frequency performance was maintained from 5.4 to 6.0 (0.19525 cyc/deg ± 0.0036;
Fig. 3A), confirming that the AII amacrine cell pathway is sufficient to drive rod responses at low luminance (
Fig. 3). In the absence of cones, however, the threshold between 4.2 and 5.4 ND was much lower than that in WT animals (
Fig. 3A versus
Fig. 1D). Thus, although cones are not capable of driving responses between 4.2 and 5.4 ND, their presence is required for rods to achieve maximal sensitivity. Thresholds did not vary much as luminance increased until 8 lux (0.9 ND); however, between 8 and 29 lux, they declined until no function was measurable at 0 ND. Measures of contrast sensitivity in ROΔCones mice followed a similar pattern (
Fig. 3B): Function was maintained from 6.9 to 1.8 ND; it gradually declined between 1.8 ND and 0.3 ND; and no tracking was present at 0 ND.
Since the results from ROΔCones mice suggested that rods require cones for normal OKT responses, we postulated that the physiological state of the disabled cones would influence the ability of rods to drive visual responses. We investigated this using rod-isolated mouse lines, in which cones were left hyperpolarized (CngA3
−/−; ROhypCones;
n = 12) and physiologically mimicking the light state, or depolarized and mimicking a dark state (Gnat2
−/−; ROdepCones;
n = 6). Spatial frequency thresholds (
Fig. 4A; left) and contrast sensitivity (
Fig. 4A; right) in ROhypCones mice were very similar to those of ROΔCones mice, indicating that hyperpolarized cones do not alter the function of rods. When we measured spatial frequency thresholds in the ROdepCones strain, however, we found (
Fig. 4B) that rod function alone was sufficient to reproduce WT-like function from 3.6 to 5.4 ND—a luminance range that spanned 4.2 ND, which is the luminance boundary of cone function (i.e.,
Fig. 2). Whereas rod function from 0 to 3.6 ND was not as good as in WT or CO mice, it was substantially better than when cones were removed or kept in a light-like state (
Fig. 3). We are confident that responses in the ROdepCones strain were mediated by rods, since the mice did not track at 0 ND. Contrast sensitivity varied with the cone genotype in a similar manner (
Fig. 4B; right): RO animals with cones maintained in a constant dark-like state possessed function comparable to WT and CO animals.
The availability of several mouse lines with defined mutations enabled the quantitative assessment of deficits in spatial visual behavior, as well as the ability to relate those deficits to specific photoreceptor systems. This would not have been possible without the use of genetic models, since the absorption spectra of individual retinal photoreceptors overlap. Thus it is not possible to selectively stimulate one photoreceptor system without influencing another in mice with intact photoreceptors.
Figure 5 graphically illustrates the spatial frequency results of the study, showing with idealized traces (solid lines) how cone (blue)- and rod (orange and green)-mediated responses would function in WT mice with intact photoreceptors. When plotted together, the top boundary of the traces approximates the function of WT mice (
Fig. 1D), indicating that the aggregate responses of the cone- and rod-selective mutant strains emulate those of mice with intact photoreceptors. Thus, cones alone enable function near 0.4 cyc/deg at 1.5 lux and brighter (blue arrow; the present study did not explore the upper boundary of cone function) conditions (blue shading), and rods alone enable function via the AII amacrine cell pathway near 0.2 cyc/deg at 1 lux and dimmer conditions (yellow-orange shading) either through the electrical synapse with ON cone bipolar cells or through the chemical synapse with OFF cone bipolar cells. We note that for the OKT measures reported here, rods were capable of generating visual responses in brighter conditions than for circadian photoentrainment.
12 We surmise that this may be because the OFF pathway, which does not contribute to circadian functions, participates in OKT responses.
Within the range from 29 to 1.5 lux, however, both cone and rod circuits are active (overlapping shading), and rods drive a different set of responses depending on whether cones are left in a physiological “light-like” or “dark-like” state. Indeed within the range from 2 to 1.5 lux, rod-driven responses can match cone-driven responses (
Fig. 5, green line). Moreover, from 1.5 to 1 lux, a luminance range in which cones are not able to drive function (
Fig. 5, blue line), rods in the presence of cones maintained in a light-like state are unable to support function better than ∼0.2 cyc/deg (
Fig. 5, orange line), whereas rods in the presence of cones left in a dark-like state are able to function as well as in WT mice (
Fig. 5, green line). Although the ROdepCones mice were outliers among the three rod-selective strains used in the study, they shared essential luminance characteristics of rod-only function with ROhypCones and ROΔCones strains: They did not respond in bright conditions (i.e., 54 lux), and they maintained responsiveness for three orders of magnitude of reduced illumination below that at which cone-mediated behavior was present. Thus, the distinct pattern of ROdepCones responses was not due to residual cone phototransduction, but was the result of the specific transducin mutation that left the cones in a depolarized state.
Possibly the most surprising finding in this study is the inability of cones to drive spatial visual responses at mesopic light levels. One explanation for this is that maintaining rods in rod transducin knockouts in a dark state lowers the sensitivity of cones to light. It is important to note, however, that our studies did not measure photoreceptor transduction directly, but spatial behavioral responses that rely on circuit processing downstream of the photoreceptors. Thus, the function of cone circuitry to drive OKT may have a different and lower sensitivity to light than cones on their own, as measured with electrophysiological recordings or with cone-driven electroretinograms. It would have been desirable to have a rod mutation that keeps rods in a hyperpolarized state to test this hypothesis. Unfortunately, such animals do not exist at this time.
There are five conclusions of the study. First, the rod AII amacrine cell pathway supports low spatial frequency and high-contrast function. Second, rods are capable of driving WT-like spatial frequency and contrast responses when cones are maintained in a persistent dark-like state. Third, the sensitivity of the cone phototransduction pathway, even at intermediate light levels at which rods apparently signal through them, is not sufficient to drive visual tracking. Fourth, the inability of rods to support function at high luminance drops gradually and is not an all-or-none phenomenon as cone function is in dimmer conditions. Fifth, the AII amacrine cell pathway enables rods to support function under relatively bright conditions, but fails to do so for circadian photoentrainment.
12 Thus, these features define the contribution that rods and cones make to a spatial visual behavior, and show that rods can drive different circuits depending on the luminance conditions, the physiological state of cones, and the nature of the visual function.
It is well established that the AII amacrine cell pathway signals rod function in dim conditions. Recent work in the circadian photoentrainment field has shown that the AII pathway fails to respond to luminance signals at intermediate or high intensities.
12 Here, however, we show that rod signaling through the AII pathway persists up to the upper boundary for rod sensitivity. We observed these effects both in mutants with cones left in a persistent light-like state, and hence shunted the rod signals through the rod–cone pathway, and in animals that lacked cones. Despite the evidence that rods use cone pathways to signal light at intermediate light intensities, animals that have cones as the only functional photoreceptors fail to support vision under the same luminance conditions. This is the first evidence that the rod–cone pathway is engaged even at intensities that are below the sensitivity of the cone phototransduction pathway. Therefore, cones are able to participate in adaptive visual function even when they are not engaging their phototransduction machinery. In addition, rods signaling through the rod–cone circuit show WT-like function. Thus, the rod–cone circuit enables good spatial visual function even when luminance is below the sensitivity of cones. Moreover, rod function does not saturate abruptly but does so gradually, regardless of the cone mutation that was used to isolate rod function.
It was beyond the scope of the present study to investigate the details of how rod-based function was able to rival cone and WT function at intermediate luminance conditions. In WT mice, the pattern of responses as luminance was decreased (
Figs. 1D,
1E) was consistent with other reports
25–27 in showing that the transition between photopic and scotopic states is smooth (
Fig. 1D), a feature of spatial visual function that has been attributed to cone–rod interactions.
28,29 Whereas the details of these interactions are not fully understood, connectomic studies
30 have shown that multiple amacrine cell-based circuits are present in the retina, which have the potential to enable rods to convey their signals through cone circuits. It is possible that such an interaction could be driven through rod–cone connexin 36 gap junctions when the cone is in a hyperpolarized state.
31 The activity of connexin 36 is known to be regulated by dopamine,
32,33 which when released in a luminance-dependent fashion allows the retina to adapt and function over a range of luminance conditions.
34 Future studies in mice with connexin 36 mutations, behaviorally tested in controlled illumination as in the present study, have the potential to evaluate this possibility.
Whereas we did not include mouse models of retinal degenerative disease (RDD) in our study, our observation that the relationship between cone and rod photoreceptor activity and visual function in normal mice with intact photoreceptors is not simple has ramifications for the characterization and treatment of RDD. Indeed, conventional studies of RDD have not been undertaken with this in mind, but with the expectation that graded changes in the photoreceptor array would be accompanied by graded changes in visual function. Most mutations that lead to RDD differentially affect the function of a single photoreceptor system, at least in the early stages of the disease.
35,36 In addition, there is also growing evidence that photoreceptor degeneration can lead to the reorganization of the inner retina
19,37 and the production of maladaptive “noise” in retinal output.
20,38–40 Thus, the degeneration has the potential to alter interactions between photoreceptor systems and to produce a worse functional outcome than would be expected by the degree of photoreceptor survival. Such maladaptive changes may explain the apparent discord between structural and functional measures in humans
41 and rodent models of RDD.
14 It should now be possible to systematically evaluate the relationship between visual function and retinal anatomy and output in mouse models of RDD and to identify the ultimate cause of visual decline. In addition, it should also be possible to evaluate whether treating maladaptive changes in the retina, which are independent of photoreceptor survival, is a feasible therapy.
Disclosure: N.M. Alam, None; C.M. Altimus, None; R.M. Douglas, CerebralMechanics, Inc. (E); S. Hattar, None; G.T. Prusky, CerebralMechanics, Inc. (E)