Variability in Bleach Kinetics and Amount of Photopigment between Individual Foveal Cones

Abstract

Purpose.: To study the bleaching dynamics of individual foveal cone photoreceptors using an adaptive optics ophthalmoscope.

Methods.: After dark adaptation, cones were progressively bleached and imaged by a series of flashes of 545-nm to 570-nm light at 12 Hz. Intensity measurements were made within the foveal avascular zone (FAZ) to avoid confounding signals from the inner retinal blood supply. Over 1300 cones in this region were identified and tracked through the imaging sequences. A single subject was used who demonstrated the necessary steady fixation, wide FAZ, and resolvability of cones close to the foveal center.

Results.: The mean intensity of all cones was well-described by first-order kinetics. Individual cones showed marked differences from the mean, both in rate of bleach and amount of photopigment; there was an inverse correlation between these two parameters. A subset of the cones showed large oscillations in intensity consistent with interference from light scattered within the cone outer segment. These cones also bleached more quickly, implying that rapid bleaching induces greater amounts of scatter.

Conclusions.: Neighboring cones in the fovea display high variability in their optical properties.

Introduction

Do neighboring photoreceptors respond equivalently to equivalent stimuli? This is an important question in the formation of accurate models to describe retinal function. Single cell recordings have established that when a single photon is captured, the change in circulating current shows little variability between receptors. ¹ This would translate into equivalent responses to equivalent stimuli, only if the probability for photon capture is relatively constant between receptors. One way to measure the probability of photon capture is to study the rate at which photopigment is bleached following exposure to light. Photopigment kinetics have previously been well characterized using retinal densitometry, in which measurements are made of the total amount of light reflected from the eye. ^2–4 Alternately, subjective measurement of pigment kinetics can be made using color-matching techniques. ⁵ Both methods give reproducible measures of the optical density and rate of bleach for the average population of cones over an extended area, but do not speak to the properties of individual cones. 

With the recent advent of adaptive optics imaging, it has become possible to visualize individual cone photoreceptors in the living eye. ⁶ Such images universally show a marked variation of the intensity of light returned from the photoreceptors, both across space and across time. ^7–12 This raises the possibility that individual cones may exhibit a wide spectrum of optical and, hence, physiological properties. At first glance, this seems like a simple idea to test using adaptive optics by the use of appropriate stimulating and imaging light on dark-adapted retina. 

However, measurements of cone intensity in adaptive optics images are confounded by interference effects. The outer boundary of the cone outer segment and the ellipsoid of the cone inner segment reflect a relatively significant amount of light. ^13,14 When the temporal coherence length of the imaging light is sufficiently large compared with the double-pass distance between these structures, interference modulates cone intensity. ^8,11 This produces significant variation in intensity between the cones, distributed in a seemingly random fashion according to wavelength scale differences in outer segment length or refractive index. 

Individual cones vary in intensity over a time course of minutes, even when imaged with light of sufficiently broad band so as to have coherence length much shorter than the cone outer segment length. ⁷ This leaves open the possibility that the apparent variability of the cone mosaic may, in part, arise from real differences in the amount of light transmitted and absorbed through cone outer segments. 

Here, this idea was investigated by measuring the amount of photopigment and the rate of pigment bleach in adaptive optics images of the initially dark-adapted cone mosaic. Low-coherence light is used to minimize interference-based fluctuations in cone intensity. 

Methods

This project was carried out in accordance with the tenets of the Declaration of Helsinki, and approved by the University of Melbourne Human Ethics Committee. Informed consent was obtained from all subjects before testing. 

System

Adaptive optics retinal imaging has been covered in detail elsewhere. ⁶ The system used here is a flood-based design. The wavefront beacon is an 835-nm superluminescent diode (Hamamatsu, Hamamatsu City, Japan ). The wavefront sensor is a Hartmann-Shack with lenslets of 0.4-mm pitch and 24-mm focal length (Adaptive Optics Associates, Cambridge, MA), coupled to a 1 MP camera charge-coupled device (CCD) (JAI Pulnix, San Jose, CA) . The deformable mirror is a Mirao-52d (Imagine Eyes, Orsay, France ) with 52 actuators and 15 mm diameter, corresponding to 6 mm in the pupil plane. Adaptive optics correction operates in closed loop at 12 Hz using custom Matlab software (Mathworks, Natick, MA). When the measured root mean square (RMS) wavefront error becomes sufficiently low (typically <0.06 μm over a 5.6-mm pupil, using normalized Zernike coefficients), imaging light is triggered on each iteration of the adaptive optics loop until 100 frames are acquired. Images are captured by two MegaPlus 4020C 4 MP CCDs (Princeton Instruments , Trenton, NJ). Each pixel corresponds to approximately 0.4 μm on the retina assuming a 17-mm equivalent focal length for the eye.  

The imaging source is a 6W supercontinuum laser (Fianium Ltd., Southampton, UK ) that produces broadband laser light pulsed at 80 MHz. Dual eight-channel, acousto-optic tunable filters (AOTFs) (Crystal Technology, Palo Alto, CA ) isolate small wavebands of interest for imaging (4–10-nm full width at half maximum [FWHM] for a single channel, dependant on center wavelength). Each AOTF passes light of a given polarization direction; the output beams are recombined with a polarizing beam splitter to maximize output power. This light is passed through 32 m of 0.37 NA, 200 μm core diameter, step-index optical fiber (Thorlabs, Newton, NJ) to reduce spatial coherence and associated image speckle. ¹⁵ The fiber tip is imaged into the pupil in Maxwellian view, and beam diameter at the cornea is approximately 2.9 mm.  

Two main wavebands were imaged with two retinal cameras, using an appropriate dichroic beam splitter. The first band was created by recruiting six AOTF channels (three per AOTF), with peak output for each channel ranging from 545 nm to 570 nm. This was designed to both bleach cone photopigment, and measure the resulting change in amount of cone photopigment. The extended bandwidth provides short temporal coherence to minimize interference-based fluctuations. ^8,11 The second band was created using two AOTF channels (one per AOTF), with peak output at 690 nm. This band was designed to provide a reference that would minimally interact with photopigment, while achieving a longer coherence length to facilitate gathering of interference-based information. 

Total corneal power was 0.49 mW for each imaging band. Total field of view was 1.3° in diameter; two smaller subfields were used for analysis (described below). Exposure duration was set at 3 ms, controlled by digital input to the supercontinuum's seed laser. Each band was individually 1.44 log units below the maximum permissible exposure (MPE) defined according to current American National Standards Institute guidelines for 10 seconds of imaging under the above conditions. ¹⁶ The corresponding retinal illuminance is 8.92 and 6.83 log td for the 545-nm to 570-nm and 690-nm bands, which should bleach approximately 39% to 65% and less than 1% of the total cone photopigment in the first frame, respectively. The results will demonstrate more rapid bleaching than this on the average, which may indicate more efficient coupling of light via the Stiles-Crawford effect (SCE) due to our Maxwellian view illumination geometry. Fortunately, many individual cones still showed far slower bleaching than the average, allowing accurate measurements to be made of the bleach dynamics despite the limited frame rate.  

The temporal coherence for each imaging waveband was measured using a Michelson interferometer with one mirror on a motorized stage. Fringe visibility was quantified using the average Michelson contrast perpendicular to the fringe orientation. FWHM temporal coherence length in the cone outer segments (n = 1.43 refractive index was assumed¹⁷) was 7 μm and 18 μm for the 545-nm to 570-nm and 690- nm light, respectively. Outer segment length of approximately 0.7° eccentricity is, at the very least, 25 μm to 30 μm. ¹⁸ This gives a round-trip path length difference of greater than 50 μm to 60 μm between light backscattered from the ellipsoid of the inner segment and from the posterior bounds of the outer segment, which is much longer than the coherence length of our 545-nm to 570-nm light; therefore, only reflections from within the cone outer segment can produce significant interference effects with this channel. Much previous cone densitometry work has been made using a broadband krypton flash attenuated by a 550 ± 25-nm filter. ^19–22 For comparison, this is calculated to produce a coherence length of approximately 5 μm by application of the Wiener–Khinchin theorem to an idealized tophat spectral transmittance profile. It is expected, therefore, that this source will mitigate interference effects to a comparable degree as achieved in this previous work. 

Subject Selection

From pilot data initially obtained on eight healthy subjects aged 22 to 28 years, several factors were identified that were critical to the success of the experiment: 

1.  
   Inner retinal vasculature greatly confounds the signal; imaging should occur within the bounds of the foveal avascular zone (FAZ). The subject should, therefore, have a wide FAZ to allow a large number of cones to be studied;
2.  
   Good adaptive optics image quality close to the center of the fovea; adaptive optics systems have difficulty resolving central foveal cones in many subjects ²³ ; and
3.  
   Highly stable microfixation in the presence of the extremely bright bleaching/imaging light, so that all cones of interest remain within the field for the duration of the bleach.

Only one of the subjects met all of these criteria and was analyzed further. The subject, who is also the first author, is a healthy 28-year-old male with emmetropic refraction and normal ocular health. 

Imaging

Data were collected from the left eye. The pupil was dilated with one drop of 0.5% tropicamide 15 minutes before data collection. Five imaging runs were conducted, with dark adaptation for 10 minutes before each run. Fixation was directed 0.7° temporal with a dimly lit, calibrated fixation target. This position was sufficiently distant from the foveal center, such that all cones could still be easily resolved, while remaining well within the bounds of the FAZ, as depicted in Figure 1. The FAZ was demarcated using 593-nm light at an inner retinal focus, with simultaneous images of the cone mosaic acquired to allow registration with cone data. Despite the extremely bright 545-nm to 570-nm imaging light, fixation in this subject was kept sufficiently steady, such that the full bleach response was captured in a large majority of the cones. 

Figure 1. 

 

Left: master cone image approximately 0.5° to 1.0° from foveal center, with white points showing cones tracked in each movie. Right: the edge of the FAZ, registered with the cone image but at an inner retinal focus.

Images were flat-fielded to correct for nonuniformities in retinal illumination. The flat-field image was generated by averaging several hundred nonregistered frames that were acquired while the subject altered their fixation in a random fashion. Data in the 1.3° field that fell outside the central 1.0° was further excluded. Normalization via flat-fielding suffices for photometric measurements such as the optical density of a cone, or the plateau intensity reached after bleach. However, measurements of bleach rate should ensure that each cone in the field receives a uniform light diet; given the somewhat Gaussian profile of our illumination, flat-fielding alone may not be appropriate for the estimation of bleach rate. The central 0.5° diameter of the field was far more uniform (<5% maximum deviation), and so, data are also provided on bleach rate in which information outside this region was ignored. 

Images obtained at 545 nm to 570 nm were reduced in size to account for transverse chromatic aberration; the necessary change was empirically determined to be 0.7%. 

Cone Tracking

To reliably locate the same cones in each data sequence, a master image was generated from the average of several hundred frames of non–dark-adapted data, acquired at 690 nm. Over 1300 cones were manually labeled in this image in an area centered about the 0.7° temporal position (Fig. 1). After manual labeling, cone positions were fine tuned by finding the brightest pixel in a three-by-three area surrounding the labeled position, and then taking the center of mass about this pixel in a five-by-five area. 

Dark-adapted data sequences were self registered to generate an average image; this average was then registered to the master image, allowing the location of previously identified cones to be accurately predicted in each data sequence. Cone positions were allowed to drift by up to three pixels from the master position to account for image distortions. Drifts farther than this were classed as errors, and 2% to 5% of cones in each sequence were discarded this way. Figure 2, left, shows a subset of the cones in the 690-nm master image, and Figure 2, right, shows the corresponding cones labeled in one 545-nm to 570-nm bleaching run approximately 1 hour later. Visual inspection of Figure 2 confirms robust identification of labeled cones in this data sequence. 

Figure 2. 

 

Unique cone identification between data sequences. Left: master image at 690 nm after manual labeling of cones. Right: cone mosaic and automatically predicted cone locations approximately 1 hour later with 545-nm to 570-nm light.

Cone intensity was represented by calculating the average intensity of the central three-by-three pixels (∼1.2 × 1.2 μm) of each cone, in each frame, of each data sequence. Figure 3 shows a series of consecutive single frames depicting the areas measured for a small region of interest in one of the data sequences. The top left pane is the average of greater than 80 frames. 

Figure 3. 

 

Cone intensity measurement. Top left: average of grater than 80 frames for one movie. Remaining panels: frames one to seven of the same movie. Average pixel value within each box was used to represent cone intensity. Each image was flat-fielded and stretched to fill its color map.

Kinetics Model

The rate of cone pigment bleach is proportional to the amount of photopigment, to the number of photons navigating the outer segment, the affinity of the photopigment to those photons, and the probability of an absorbed photon resulting in bleaching. ²⁴ This first order relationship can be expressed as:  

 

where p is the density of photopigment (mol/μm²); I_inc is retinal illuminance (td, i.e., proportional to quanta/μm²/s); γ represents the probability of photoisomerization resulting from photon absorption (∼2/3); α is a cross-sectional constant (μm²/mol) indicating photosensitivity of the receptor-pigment complex. It is shown below that it is useful to decompose this further into an optical contribution (efficiency of coupling to the receptor waveguide; efficiency of reflection from tissue posterior to the visual pigment) and a contribution representing the photosensitivity of the pigment: $\alpha ={\alpha }_{optical}.{\alpha }_{pigment}$.  

Ignoring pigment regeneration, solution of equation 1 for the amount of pigment at time t gives:  

 

Immediately after traversing a cone in double passage, the illuminance (I_refl) depends on the amount of light that traversed the outer segment (I_OS) and the amount that was absorbed by the visual pigment:  

 

 

 

 

 

 

 

The light directly leaving the receptor is assumed to be related to the recorded camera signal (C) by a system- and eye-dependent constant that is the same for all cones in a given eye (ρ). The (α.γ) term is expressed as 1/Q_e, where Q_e is the amount of energy (td-s ) required to bleach all but e⁻¹ of the available pigment; this provides a form for the receptor photosensitivity that is more directly comparable to previous work:  

 

There are three key terms that can be readily extracted by fitting camera intensity data with equation 4:  

 

 

 

 

 

 

The plateau intensity is the most reliably measured term, because it is the limiting intensity as t approaches ∞, for which we have ample and stable data points. Variations in this parameter from cone to cone, in the absence of interference effects, most likely reveal differences in receptor waveguiding or in the reflectivity of tissue directly behind each cone. Since the ρ parameter can only be estimated in our setup, we leave the units for this constant as arbitrary. 

After determining the plateau intensity, total absorptance may be calculated by setting t to zero in equation 4:  

 

 

 

 

Where C₀ is the pixel intensity in the first frame. In other words, this term gives the fraction of light absorbed by the cone's available photopigment. It is related directly to the double-pass optical density, and again is reliably fit by the model. 

The final term, representing overall receptor sensitivity (including both optical and pigment factors) was less reliably measured. This is because it describes the shape of the bleaching curve, which is steep since measurements had to be made with a very bright light in order to see the cones. However, if individual cones that are bleaching slowly enough are found, a narrow confidence interval (CI) for this parameter will be achieved, allowing reliable estimates of this value for those cones. 

Important Note concerning Measured Pigment Density

While the term representing density is fit reliably with the model, the model does not take into account the effects of light that reaches the camera without passing through the visual pigment in the outer segment (e.g., cornea, lens, internal limiting membrane; in traditional densitometry the interstitial spaces between the cones are another example). This stray light will reduce the size of the ratio between initial and final recorded intensities, and so underestimate the density of the pigment. ³ This is not a trivial point; recent estimates place the stray light contribution as high as 50% ²⁵ unless efforts are made to explicitly reduce the contribution, primarily by manipulating the size and position of the entrance and exit pupils. ²⁶ Such manipulations are not practical for an adaptive optics system, since large and full pupils are required, both for resolution and sufficient signal to noise ratio to make accurate measurements on individual cones. 

It is possible to circumvent the limitations of stray light by two techniques, both requiring an additional set of measurements to be taken at an additional signal wavelength (e.g., 600 nm). The first technique relies on self screening; a phenomenon where the shape of the absorption spectrum becomes flatter the more heavily concentrated the pigment. The additional measurements allow estimation of the spectrum shape, allowing the pigment concentration to be inferred, ^27,28 although the degree of flattening is low at the modest concentrations of the visual pigment, somewhat limiting accuracy. ³ The other method is to calculate the stray light contribution directly from the additional measurements, and to incorporate this into the traditional calculation of pigment density. ²⁹ Neither approach has been adopted here, as the primary interest is relative differences between cones, which can be estimated quite reliably. The measured density values will be shown to agree well with Rushton's ² initial work that did not exploit self screening or control particularly well for stray light. 

Results

The top five plots in Figure 4 show mean cone intensity for each dark-adapted data sequence at 545 nm to 570 nm. The data have been displaced vertically for clarity. A plateau is reached well within the first seven frames, followed by a second phase of typically greater intensity. This second phase is unlikely to be related to bleaching of photopigment; L and M cone pigment should be bleached greater than 99.9% after seven exposures of 3-ms light at 8.92 log td, while S cones and rods have minimal numbers in this area. The second phase is also not a result of eye movements because the simultaneous 690-nm data (bottom plots) appear flat. The effect may be a result of increased choroidal blood–oxygen saturation following the vigorous retinal stimulation caused by the intense imaging light. Regardless of the cause, the fitting of bleach parameters with equation 1 was limited to the first seven frames of each sequence, plotted as solid lines in Figure 4. The r ² of the fit to the mean data was greater than 0.98 for each sequence at 545 nm to 570 nm. 

Figure 4. 

 

Mean intensity for all cones during bleach. Top five plots show intensity at 545 nm to 570 nm (displaced for clarity), with unbroken lines showing the fit of equation 1 to the first seven data points. Simultaneously collected data at 690 nm is shown in the lowest group of plots (overlapped for clarity). The middle group of plots shows intensity at 545 nm to 570 nm, shortly after a near total bleach.

Although not apparent in Figure 4, the mean first frame intensity, plateau intensity, and photosensitivity varied significantly between data sequences. The topmost data sequence especially showed reduced photosensitivity (slower rate of bleach). Such effects could result from variations in alignment relative to the peak of the SCE. Despite such variations, the absorptance was highly reproducible between sequences: mean ± SD = 0.278 ± 0.01, giving double pass optical density = 0.142 ± 0.006. As discussed in the Methods section, this figure is a marked underestimate since stray light necessarily corrupts the measurements; however, the figure does, very closely, match Rushton's ² data in the presence of stray light, and acts as a suitable metric to compare relative amounts of pigment between cones. 

The middle plots of Figure 4 show data at 545 nm to 570 nm collected in the same retinal location following near total bleach. The bleach was achieved 30 to 50 seconds earlier via another imaging run of 100 frames. The intensity versus time relationship after the bleach is flat as expected. Taken together with the lack of a dark-adapted effect at 690 nm, the correctness of the measured optical density, and the correctness of the derived photosensitivity constant for the receptors (see Discussion), it is highly likely that photopigment bleach is indeed being measured. The only other explanation is some spectrally active optical effect downstream to the phototransduction process. It is noted that any such effect would also affect efforts at conventional retinal densitometry.  

Figure 5 shows the intensity at 545 nm to 570 nm for selected cones in one sequence. Cones were selected to represent two groups that were evident in the data: (1) cones that obeyed the predicted exponential model (bottom plots), and (2) those that clearly did not (top plots). It is assumed that all cones do, in fact, obey an underlying exponential time course, but that in some cones these data are obfuscated by constructive and destructive interference. An interference-driven explanation for these cones is supported by the large amplitude of the fluctuations, by the absence of fluctuations when averaged over a large number of cones (Fig. 4), and by previous work in the literature. ^8,11 There are, therefore, two data sets that may be examined using the exponential kinetics model: (1) the mean of the whole cone population, where interference effects will be averaged out, and (2) the subset of cones that obeyed the exponential model well, and, hence, do not display interference effects. 

Figure 5. 

 

Intensity of selected cones during bleach. Data separated into two groups: top (no symbol) showing interference response, bottom (crosses) showing exponential response.

From the cones in Figure 5 that did obey the model, it seems that there can be marked variation in the ratio between the first and last frames (absorptance), the time to reach plateau (photosensitivity), and the plateau intensity reached. To explore this further, the analysis was restricted to cones that achieved very high fit quality (r ² > 0.95) when fit with equation 1. This constituted 21% to 36% of cones depending on the particular movie. Figure 6 shows the results for one representative movie. Cones were ordered ascendingly by their measured absorptance. Vertical lines indicate 95% CI on the fit parameters for each cone in this subset, while horizontal lines indicate the fit to the mean intensity of all cones. Red CIs indicate cones in this subset that differed significantly from the fit to the mean data at the 95% CI. As discussed above, the absorptance values obtained are underestimates since no correction has been made for the effects of stray light. 

Figure 6. 

 

Bleach parameters for the subset of cones that were well fit by the exponential model (r ² > 0.95), for one representative movie. Vertical lines show 95% CI for each cone. For comparison, horizontal lines show the fit parameters to the average intensity of the entire cone population. Red lines indicate parameters that differed from the average for the entire cone population at the 95% CI, revealing generally greater absorptance and reduced photosensitivity (slower rate of bleach) in this group of cones.

It can be seen that, even in the subset of cones that achieved a high-quality exponential fit, there is marked variability in each of the fit parameters. Further, these cones clearly tended toward greater absorptance and reduced photosensitivity compared with the mean of all cones. There was also a mild tendency toward greater plateau intensity. The same trend was repeated in each of the data sequences. On average across all movies for these cones, at the 95% CI: (1) 55% (4.5%) of cones had greater (less) absorptance than the mean; (2) 44.5% (0.3%) of cones had slower (faster) rate of bleach (i.e., photosensitivity) than the mean; and (3) 43% (28.5%) of cones had greater (less) plateau intensity than the mean. 

It is known that optical density in the central fovea can change rapidly with eccentricity in some subjects due to changes in cone length with eccentricity. Others show a relatively flat response. ³⁰ In this study's subject, over the small (∼0.5°) area examined, there was no correlation found between retinal position and optical density. In other words, density changes resulting from differences in cone length were minimal for this subject in the area analyzed. 

The fitted photosensitivity constants are the least reliable parameter in these data for several reasons. The bright intensity of the imaging light ensures most of the photopigment is bleached in the first frame, making it difficult to accurately characterize the time course of the bleach. Additionally, as discussed in the Methods section, each cone should receive an equivalent light diet in order to make accurate measurements of the bleach constant; therefore, to confirm the observations regarding the rate constant, analysis was further limited only to those cones that remained within the central 0.5° of the field (where illumination varied by no more than 5%), as well as achieving r ² greater than 0.95 fit to equation 1. This corresponded to approximately 85 cones for each movie. The fitted data for each of these cones in one sequence is plotted in Figure 7, after normalization to a common start and end point to facilitate comparison of rate constants. It can be seen that most of these cones had slower rate constants than the fit to the mean data (bold line). 

Figure 7. 

 

Fitted cone intensity for all cones within the central 0.5° diameter (uniformly lit) that achieved good exponential fit. Data normalized to common start and end points. The bold line shows the fit to the mean of all cones. Most cones had slower time course than the mean.

Discussion

The time course of the bleach response has been tracked in a large number of individual cones within the FAZ. The average estimate of double pass optical density (0.14) precisely matches Rushton's ² classical data, and the measurements of photosensitivity (8.2 × 10⁻⁷ td⁻¹-s⁻¹) also compare favorably with Rushton ^4,31 (2.9–6.3 × 10⁻⁷ td⁻¹-s⁻¹) . The figure for pigment density is an underestimate since light that returns to the detector without passing through the visual pigment, ²⁷ was not accounted for; however, the figure was highly consistent between runs, facilitating relative comparisons of individual cones within a given subject.  

Individual cones broadly displayed two types of behavior: some showed random fluctuations that are likely due to interference effects, while others showed a classical first-order kinetics response. The cones in the latter group displayed wide variability in total amount of photopigment and in rate of bleach, and almost all had reduced rates of bleach and greater amounts of photopigment compared with the mean. These parameters could not be assessed in the former group, but by deduction it seems that the cones displaying interference effects must have had greater rates of bleach, and lesser amounts of photopigment. 

Since the coherence length of our source was much shorter than the cone outer segment, interference can only arise due to scatter within the outer segment. ¹² This implies an association between increased scatter within the outer segment and cones that bleach quickly. Single cell recording work has previously shown quantized, step-like recovery of dark current following a bleach. ³² This was interpreted to correspond to occasional, localized, abrupt blockages in the dark current along a length of approximately 1 to 2 μm within the outer segment. If these blockages establish a scattering boundary, those arising close to the bounds of the outer segment will produce interference effects even for short coherence light. Conversely, blockages arising closer to the middle of the outer segment would not produce interference effects with short coherence light. Assuming that the induced scatter is a direct consequence of the current flow, this idea could be verified by measuring the proportion of cones displaying interference-modulated intensity after a pause of approximately 100 ms, which is sufficient to allow near total recovery of the cone circulating current, even after a substantial bleach. ³³ Rods are known to recover circulating current far more slowly after a bleach, ^33,34 and may, therefore, exhibit scatter for a far longer period of time. 

The blockages described above should occur more frequently when the cones are more vigorously stimulated. Differences in cone stimulating protocol between different research groups may, therefore, explain discrepancies in the fraction of interference-driven cones reported. ^8,10,12 In addition, the results show that the amount of light channeled through the outer segment can be highly variable between adjacent cones. Therefore, it might be expected that the amount of scatter will be greater in those cones that show the greatest light throughput. With light of sufficiently short coherence, this would predominantly manifest as a brightness increase that is larger in these cones. Grieve and Roorda ⁹ observed just such a result using 840 ± 50-nm light to image cones after stimulation with visible light; the cones that were initially brightest (greatest light throughput) showed the most increase in scatter after stimulation. 

Changes in near-infrared scatter of the outer segment in response to light have previously been measured in suspensions of rod outer segment fragments. ^35,36 These changes have been attributed to redistribution of products of the phototransduction cascade between membrane-bound and soluble forms. This is another potential source of the scattering within the cone outer segment, though it is a less satisfactory explanation due to the apparently slower time course, and the lack of any known tendency toward the formation of localized boundaries.  

The results presented here show an inverse correlation between rate of bleach (linked to optical coupling) and amount of photopigment. Based on this, and the following lines of evidence, it is proposed that the presence of unbleached photopigment may be associated with reduced optical coupling: 

1.  
   Image quality, as opposed to just brightness, seems to increase as the bleach progresses. In preliminary results from a separate experiment on the same subject, the retina was imaged with 540-nm to 570-nm light after 5 minutes of dark adaptation and also after a 25 × 10⁶ td-s bleach. This procedure was repeated seven times, and the first frames from each run averaged together. Figure 8 shows the resulting dark-adapted (left) and bleached (right) images, showing more well-defined and higher contrast cones after bleaching. After normalizing by the total energy in each image and masking the area filled by the blood vessel, variance was approximately 60% greater in the bleached image. There was also a redistribution of energy from the dimmer pixels (predominantly intercone space) to the brighter pixels (the cones), for example, the energy fraction carried in the dimmest 15% of pixels was approximately 20% less in the bleached image than in the dark-adapted image. This redistribution of energy supports the notion of a change in waveguiding properties following bleach;
2.  
   Similarly, it has been shown that ex vivo two-photon fluorescence retinal image quality is drastically reduced in the dark-adapted retina ³⁷ (Fig. 4 in the cited paper). This occurred despite the fact that the stimulating (730 nm) and excitation (730/2 = 365 nm) lights were negligibly absorbed by photopigment; and
3.  
   The SCE is tightly linked to optical waveguide theory. ¹⁷ It has been shown that the directionality of the SCE increases following a bleach compared with the directionality after several minutes of dark adaptation. ³⁸ Interestingly, there is reportedly a further increase over the bleach case after dark adaptation for 30 minutes, ³⁹ indicating a complex relationship between waveguiding and the adaptation state of the retina. The increased directionality after bleach has been noted to recover over the course of approximately 1 minute, which is faster than the approximately 5 minutes required for the cone pigment itself to recover from a full bleach ³⁸ ; therefore, the effect may be linked to some metabolites of the phototransduction cascade as opposed to pigment.

Figure 8. 

 

Image quality of the retina with 545-nm to 570-nm light when dark-adapted (left) versus bleached (right). The first frames of seven separate imaging runs were averaged to make each image. Cones appear more localized and distinct from the background in the bleached retina. The location shown is approximately 1.25° temporal to the fovea.

The results presented here also indicate intercone variability in the probability for photon catch. It is known that at intensities sufficient to bleach more than a few percent of cone pigment, the cone response is dictated almost entirely by the amount of unbleached pigment. ^34,40 It follows that the probability of photon catch should determine performance under this regime. Such variability may, therefore, impose a large source of noise for visual processing. 

This work was originally spurred by a desire to develop an efficient way to classify cone subtype with adaptive optics. Previous approaches have required the averaging of results from many dark-adapted runs, which is time consuming. ^19–22 Given the broad range in individual cone pigmentation and bleach kinetics that was found here, coupled with the close spectral peaks of the L- and M-cones, the established method of averaging a large number of local cone densitometry measurements does indeed seem to be the only valid approach. Given the success of others with such measurements, it is suspected that the variations in optical density and optical coupling seen here are probably not fixed for a given cone, and instead drift over time based on natural cycling in cone physiology. 

Acknowledgments

The authors are grateful for the extensive and insightful comments made on the manuscript by our colleague Alfredo Dubra and by the anonymous reviewers. These comments aided in improving the rigor of the bleaching model, the visualization of results, and the exploration of all of the relevant literature. 

Footnotes

 Supported by grants from the Australian Research Council Discovery Project (DP0984649), the Australian Research Council Discovery Early Career Researcher Award (DE120101931), and the University of Melbourne Interdisciplinary Seed Fund Scheme.

Footnotes

 Disclosure: P. Bedggood, None; A. Metha, None