Contrast Sensitivity and Visual Acuity in BCM Carriers is Similar to Normal Observers
On psychophysical tests performed without AO, BCM carriers performed in the normal range. This confirms the findings of previous researchers, who showed that BCM carriers usually appear normal on most tests of visual function. This is in line with the large amount of variability in what is considered to be “normal” for the visual system. The contribution of this work is to establish such apparently normal vision given the significant amount of disruption observed at the level of the cone mosaic. Indeed AO-corrected measurements revealed a very different pattern of visual resolution, with BCM carriers having much poorer resolution than controls once optical aberrations are no longer a limiting factor. In fact, contrary to the relationship found in normal eyes, where optical aberrations typically limit visual resolution in the fovea, cone spacing can limit visual resolution at the fovea in some BCM carriers.
The relative role of optical quality and retinal sampling in shaping our visual capacities is illustrated by considering closely the two carriers who underwent visual resolution testing both with spectacle correction only and with AO-correction. Despite some differences in the stimulus used in each experiment (e.g. polychromatic with spectacle correction only in Experiment I (MARSC) and monochromatic with AO-correction in Experiment II (MARAO)), it is interesting to compare the results obtained for these two carriers (JC_1043 and JC_1041). At the PRLF, JC_1041 obtained MARSC of 0.75 arcmin and a nearly identical MARAO of 0.74 arcmin. At 2.5 degrees the results were also nearly identical, where JC_1041 obtained MARSC of 2.12 arcmin and MARAO of 2.09 arcmin. For this observer, it appears that high order ocular aberrations, which are minimized with AO, had little effect on visual performance. The other BCM carrier tested in both experiments, JC_1043, obtained quite different measurements in each experiment. At the PRLF, JC_1043 had a MARSC of 0.86 arcmin and MARAO of 0.72 arcmin. At the 2.5° location she obtained MARSC of 2.19 arcmin and a MARAO of 1.46 arcmin at a slightly greater eccentricity of 2.64°.
The first patient, JC_1041 was thus not ‘optically’ limited under normal viewing conditions (ie. for MARSC
measurements), but rather was limited by retinal sampling in both, while the latter JC_1043 was ‘optically’ limited under normal viewing conditions, as she improved when optical aberrations were minimized with AO (ie. for MARAO
measurements, as normal observers do). However, she failed to reach the performance levels we have observed previously in normal eyes 
. This illustrates how differences in optical quality alone can impact visual resolution, obscuring any differences that may exist downstream in the visual system of different observers. Visual resolution tests with AO can reveal these small differences, as has been shown previously in myopia 
Fixational Eye Movements are Abnormal in Some BCM Carriers
The average SD of fixation for the normal observers when looking at the stimulus was similar to that observed by others 
. The larger SD of fixation when both groups of observers looked at the fixation target relative to the stimulus was not surprising as target size, luminance and color (all of which were different for the fixation target relative to the stimulus) have been shown to influence fixation stability 
. The relatively larger spread of fixation found in the BCM carriers is consistent with the abnormalities in eye movements previously observed in BCM carriers 
. It should be noted that JC_1043 had a pattern of fixation that was within the normal range at all locations; she is the carrier who also had the best resolution, highest cone density (lowest Nc
), and presumably the least amount of cone loss. This is consistent with previous findings that showed that although fixational eye movements may be abnormal in some BCM carriers, they are not abnormal in all BCM carriers 
. An alternative explanation for the fixational instability seen in the BCM carriers might be that they simply were less familiar with the procedure than our normal controls.
It is interesting to note that for test locations at the PRLF, MARAO
was moderately correlated with the SD of fixation for the carriers (R2
0.70) but not for the normal observers (R2
0.31). However, the larger motion probably did not cause the resolution deficit in the carriers, as it has been shown that visual resolution is largely unaffected by retinal image motion 
. In fact, the relative fixational instability seen in two of the three carriers is probably a result of their increased Nc
(and lower MARAO
), consistent with the hypothesis of Steinman and colleagues, that one fixates accurately in order to see clearly, not because one sees clearly 
. It is possible that the fixation control mechanism is relaxed in BCM carriers because the larger cone spacing tolerates a larger degree of image motion without interfering with vision. That is, the carriers are probably less stable than normal because they can tolerate a larger amount of retinal image motion without it interfering with their ability to see clearly.
Does Cone Loss Lead to RGC Loss in the BCM Carrier Retina?
In normal observers, the Nyquist limit of the mosaic of midget retinal ganglion cells (NmRGC
) predicts the reduction in visual resolution seen outside the foveola 
is set by the density of mRGC receptive fields. The transition from Nc
-limited visual resolution in the foveal center to NmRGC
-limited resolution across the visual field is a consequence of eccentricity-dependent changes in retinal circuitry in the cone to midget bipolar cell to mRGC network 
. When the ratio of mRGCs to cones is 2 or greater, as it is expected to be in the center of the normal fovea 
, the expectation is that each cone will have a so-called ‘private line’ connection 
to each of an ON- and OFF- centered mRGC. When this circuitry exists, the Nyquist limit of mRGC receptive fields is identical to Nc
). Under these conditions MARAO
is expected to match Nc
In the normal retina, the density of both mRGC receptive fields and cones decrease with eccentricity 
, as does the mRGC-to-cone ratio 
. When this ratio falls below 2, the centers of mRGC receptive fields begin to receive input from more than one cone. This compromises resolution, resulting in MARAO
no longer matching Nc
but rather matching NmRGC
. If the reduced number of cones in the BCM carrier retina was paired with a normal number of RGCs, the BCM carrier retina would have a higher mRGC-to-cone ratio across the retina than is found in normal eyes. For example, if a BCM carrier fovea had half the cones of a normal retina but a normal complement of mRGCs, the mRGC-to-cone ratio might be 4
1 instead of the 2
1 ratio found in the normal retina. Because the mRGC-to-cone ratio governs the eccentricity at which the transition from Nc
limited resolution to NmRGC
resolution occurs, it is expected that this transition would occur at a more eccentric location in the carrier retina than in the normal retina, extending the ‘private line’ connection to greater eccentricities. This hypothesis is summarized in supplementary figure S1
, which shows the circuitry and acuity predictions made under this scenario. Under this scenario, Nc
might be expected to match MARAO
across the full range of test locations examined herein. However, we found that MARAO
are matched only at the PRLF and that MARAO
falls off at a greater rate than predicted by Nc
just outside the PRLF, with a pattern that is similar to that observed previously for normal observers 
This result supports the hypothesis that there are postreceptoral changes in the organization of the BCM carrier retina. One possibility is that there is coordinated variation between the number of cones and the number of ganglion cells: the number of ganglion cells is reduced in BCM carriers in a systematic way. Support for this hypothesis comes from the fact that we show that the theoretical NmRGC
based upon mRGC receptive field density measurements from normal eyes 
does not fit with the data obtained from BCM carriers, except for the most eccentric test location of JC_1043 (where she had normal cone density). BCM carrier performance at all other test locations was worse than predicted by the model of Drasdo et al. 
, providing further evidence that there is loss of mRGCs. That this loss is local, and is a consequence of the cone loss, is supported by the results from JC_1043 at the most eccentric test location. This BCM carrier was within the normal range on all measures at this test location, and fit well with the model prediction for NmRGC
limited resolution at that test location. This finding is consistent with coupled cone-RGC loss, as a reduction in mRGC density is predicted only where reduced cone density is observed.
The Relationship between AO-corrected Visual Resolution (MARAO) and the Nyquist Limit of the Cone Mosaic (Nc) in BCM Carriers is Similar to that Observed for Normal Eyes Despite Evidence for Coupled Cone-mRGC Loss in the Carrier Retina
Examining in detail the relationship between Nc
found in the BCM carrier retina, it can be seen that the relationship observed is similar to what was observed in both the normal retina and to what is predicted from the model of Drasdo and colleagues 
. As can be seen from , for a given Nc
, BCM carriers actually achieved better MARAO
than the myopic control eye and fell well within the limits measured previously for normal eyes 
. However, the eccentricities at which equivalent MARAO
pairings were found in the BCM carriers were much closer to the PRLF than in a normal retina. For a given MARAO
, the difference between Nc
is similar in both the normal and carrier retina. In fact, corresponding values of Nc
found in the carriers were similar to model predictions at all test locations, the only difference being that they were not found at the model-predicted retinal eccentricities.
re-plots the data shown in , along with the data from the normal observers from Rossi & Roorda 
, and a curve showing the theoretical relationship between Nc
and the NmRGC
predicted from the Drasdo et al. model 
and the cone density data of Curcio et al. 
. For each eccentricity at which an Nc
measurement was calculated, a corresponding value of NmRGC
was computed using the general model of mRGC receptive field density of Drasdo and colleagues 
. It can be seen that both the normal and BCM carrier data points fall near the model prediction curve. Points near the PRLF for the empirical data are slightly shifted upward from model predictions. This is not surprising as Nc
was estimated at the PRLF for four of the five observers from Rossi and Roorda 
; these estimates contain errors and are, on average, higher than Nc
estimates derived from the peak density measurements of Curcio 
. It is also likely that the estimates of Nc
from the average data of Curcio contain some error because the true conversion factor (m
) for those eyes between mm and degrees of visual angle is unknown. The large range of cone densities observed at the foveal center also makes predictions at this location subject to the largest amount of variability.
Relationship between MARAO and Nc in BCM carriers is similar to normal eyes.
The agreement between model predictions and empirical data was better outside the PRLF, where measurements of Nc were made directly. It is interesting to note that the BCM carrier data points of MARAO values beyond the range of those obtained for normal observers still follow the trend predicted by the model. This shows that the carrier retina can be thought of, in effect, as an eccentricity-shifted version of the normal retina. For example, observer JC_1041 had a MARAO of 1.41 arcmin when her Nc was 1.06 arcmin, falling near the model prediction for that MARAO–Nc pairing; however, the model predicts that this combination would be found at an eccentricity of ~2.3° in a normal retina, whereas it was observed at an eccentricity of only 0.86° degrees in JC_1041, a difference of 1.44°.
What does this tell us about the retinal circuitry of the BCM carrier retina? It seems to indicate that for a given cone spacing, downstream neural circuitry is similar in both normal and BCM carrier retinas. This suggests that the main determining factor for the size of mRGC receptive fields appears to be the spacing between cone photoreceptors, as the relationship between Nc
appears to be identical in both groups. As such, this finding indicates that NmRGC
is largely determined by Nc
. This suggests that the number of mRGCs in the retina is directly related to the number of cones and that the cone loss seen in the carriers probably led to a subsequent loss of RGCs. It is hard to imagine how the visual system might deal with a mismatch between the number of cones and mRGCs, particularly in the case where there would be either redundant circuitry at the midget bipolar or midget ganglion cell level, which leads to such strange predictions as spatially redundant mRGC receptive fields in the center of the BCM carrier fovea (see supplementary figure S1
). More plausible is the prediction that the ’private line’ might persist to larger eccentricities than found in the normal retina, well outside the PRLF. However, it is not clear how a cone-mRGC ratio of greater than 2
1 (which is implied if a normal complement of mRGCs is paired with a reduced number of foveal cones) would be implemented at the foveal center or if it would be advantageous. The ON- and OFF- center mRGC sub-mosaics are thought to be spatially redundant in the central fovea, requiring 2 mRGCs per cone; it does not seem plausible for there to be multiply redundant ON and OFF arrays or that there is a mechanism for such circuitry to be implemented. There may be a spatial limit to the size of a single-cone centered mRGC receptive field, as a discord between MARAO
at the PRLF in the carrier eye with the largest cone spacing was observed. However, as a discord between MARAO
at the PRLF was also observed for the control eye, this question cannot be answered here, and it is likely that cortical factors are also involved in limiting MARAO
in the BCM carrier at the foveal center, as is probably the case in myopia 
. The primary limitation of this study is that we did not obtain fMRI and mRGC estimates in all subjects, precluding direct comparison between the cortical representation of the fovea and our predictions of RGC densities in the same subjects. However, the results from the group of BCM carriers that we did measure clearly suggest that the cortical retinotopic organization in BCM carriers is no different from that seen in normal eyes. Further understanding of the organization of RGC receptive fields near the PRLF and the relationship between mRGC density and the cortical representation is required.
What emerges from the results of this set of experiments is perhaps a simple explanation about what governs the foveal overrepresentation in the cortex (ie. cortical magnification). It is well known that in normal eyes, cones greatly oversample the retinal image, but there are other constraints on the information available to the cortex than those specified by the number of cones. Despite reduced cone density, all BCM carriers appeared within the normal range on conventional tests of visual function: this is perhaps the key finding here. This suggests that it is not the number of afferents that drives the foveal overrepresentation, but rather the content of the information that is relayed to the cortex from the retina across the visual field. Under normal viewing conditions, filtering of information limited performance at all locations in the visual field for all observers. Although the filtering mechanisms differed for BCM carriers and normal observers, the resulting information reaching the cortex under normal viewing conditions was similar. At the PRLF, the filtering mechanism was the optics of the eye for the normal observers but cone spacing for most of the BCM carriers. Outside the PRLF, optical filtering probably continued to limit performance in normal eyes very close to the PRLF, but at the more eccentric locations, neural filtering (ie. convergence) imposed the limit in both groups. The filtering mechanism is invisible to the cortex; the information it receives is comparable in each case. In both groups, the central visual field was the most finely sampled area and thus contained the most information about each unit of visual space. The cortical area represented by the fovea and periphery was similar in both groups, suggesting that the information processing requirements were much the same.
This study demonstrates how genetic, behavioral, and imaging techniques can be applied in concert to develop a more comprehensive understanding of the interdependence between different structures within the visual system. This approach is challenging in that it requires technical expertise that few, if any, individual laboratories possess and so requires collaboration amongst several investigators. Despite these challenges, this approach may be well suited to study other conditions that disrupt the organization of the interrelated components of the visual system, such as inherited retinal diseases.