Retinitis pigmentosa (RP) is a group of inherited retinal diseases which can lead to photoreceptor degeneration and concomitant progressive loss of peripheral vision which encroaches on the central retina over time (1, 2). RP is the most common inherited retinal disease, impacting approximately one in 4000 individuals globally (1, 3, 4). Functionally, RP leads to reduced vision in low light conditions, reduced peripheral vision, color vision deficiencies, and eventually reduced visual acuity and legal blindness. Mutations in more than 87 genes lead to RP, and there are three main inheritance patterns: autosomal dominant, autosomal recessive, and X-linked (2). Previous studies have reported that disease progression is mediated at least in part by the genetic subtype of RP (5).

Current techniques for diagnosis and monitoring of RP rely heavily on clinical imaging to assess retinal structure. Specifically, fundus imaging can reveal telltale signs of RP such as bone-spicule hyperpigmentation, narrowing of retinal arterioles, cystoid macular edema, and a yellowed appearance of the optic disk. Optical coherence tomography (OCT) has revealed even more structural disruptions in RP including inner retinal thickening, photoreceptor layer and retinal pigment epithelium thinning, and cystoid macular edema (6–8). While these imaging techniques are useful for diagnosing RP, they are generally used after the patient has already reported significant visual symptoms.

Clinical tools such as electroretinography, microperimetry, Humphrey visual field testing, and visual acuity testing have been used to characterize and monitor the functional impact of RP on the retina (9–18). By their nature, these functional assays evaluate large areas of the retina that span hundreds of photoreceptors. Like their structural counterparts, they lack the sensitivity required for monitoring early disruptions in retinal cells, meaning significant damage can occur before manifesting clinically. Indeed, previous studies investigating the relationship between foveal cone metrics such as density, spacing and visual acuity have consistently shown that visual acuity remained within a normal range until there was a 40% - 60% reduction in foveal cones (19, 20).

With the recent advancements in gene therapy, cell replacement, and small molecule strategies showing promising results for future use in the treatment of inherited retinal diseases (21–24) more sensitive tests of photoreceptor function are needed to better monitor disease progression and test treatment efficacy. Optoretinography (ORG) is a non-invasive method of assessing photoreceptor function in vivo using only the light backscattered from retinal cells (25–28). There are at present two methods of obtaining ORGs: phase-based optoretinography (pORGs), and intensity-based optoretinography (iORGs). pORGs are determined using the argument (or phase) of the complex OCT signal, and are typically calculated by determining the change in phase as referenced to the cone boundary. This is then related to the change in optical path length (29–35). Conversely, iORGs are created from changes in the amplitude of backscattered light from photoreceptors using adaptive optics flood illumination ophthalmoscopes (26, 36) adaptive optics scanning light ophthalmoscopes (AOSLOs) (25, 37, 38), and OCTs (39–43). The spatial and cellular selectivity of optoretinography makes it a particularly attractive technique for assessing photoreceptor disease. Indeed, prior work using pORGs in three individuals with autosomal recessive RP showed decreased function in photoreceptors (44). This predicates a broader question about the utility of the ORG for the clinic, and by extension its relationship to existing clinical tools.

In this work, intensity-based optoretinography is used to evaluate photoreceptor function in individuals with retinitis pigmentosa and controls with no reported retinal pathology. Additionally, this work examines the iORG’s relationship to clinical microperimetry and metrics of photoreceptor mosaic structure across multiple retinal locations. The relationship between each of these measurements is then compared to eccentricity-matched control data.

In this study, iORGs were used to assess photoreceptor function in individuals with retinitis pigmentosa and a control population of individuals with no reported retinal pathology. We compared iORG RMS amplitudes in these individuals to measurements of cone structure such as nearest neighbor distance and cone outer segment length and compared iORG RMS to a gold standard clinical measure of retinal function, microperimetry (MAIA). Structural metrics were consistent with previously published values for controls and individuals with RP (50, 62, 63), and we observed higher NND values for individuals with RP at more eccentric locations, consistent with the progression of RP. Additionally, MAIA microperimetry indicated relatively preserved function near many of the RP individuals’ foveae. Like metrics of structure, these results were variable across individuals, consistent with previous reports of varied phenotypes and disease progression across different individuals with RP (64, 65).

For most RP participants, average population-RMS amplitudes were far lower than our control population, with two notable exceptions: 1.5° in 64010 and 2° and 5° in 89385. 64010 has a less severe RP phenotype, and cones close to her fovea have markedly preserved structure and function, though amplitude still declines with eccentricity. However, 89385’s iORG pattern is not consistent with previous reports of functional losses in RP (10, 65, 66) despite their clinical RP diagnosis; indeed, they showed decreased RMS amplitudes at 1°, a normal RMS amplitude at 5°, and another decrease near 8°. While not detected by the NND metric, the structure of 89385’s photoreceptor mosaic appears grossly abnormal ( Supplementary Figure 3 ), with focal patches of missing rod photoreceptors and preserved cones. Thus, while our data is not consistent with a “classical” definition of RP, we cannot rule out other possibilities such as mosaicism (67), which has been reported in individuals with RHO mutations (68, 69). Further, without genetics we cannot rule out other diseases with pigmentary deposits, such as pigmented paravenous retinochoroidal atrophy (70), or even congenital rubella (71). Unfortunately, we were not able to collect a genetic sample from this individual to confirm either hypothesis.

We also interrogated how many cones in each area had putatively “healthy” (e.g. greater than the 5^th percentile of normal cones) individual-RMS amplitudes. This analysis allowed us to determine if cone function in a region is lost uniformly, or if certain cells retained normal function while others declined. Interestingly, three individuals who had substantially reduced population-RMS amplitudes at 1° (13090, 53885, 89385) had ~80% of their individual-RMS amplitudes within normative ranges, suggesting that their low population-RMS amplitudes were being driven by a small subset of cones in those ROIs. The individual-RMS amplitudes of 89385 again stand out as unique compared to other individuals with RP, at the 1° and 2° temporal location there are less cells responding normally than at the 5° and 8° temporal locations. Overall, the individual-RMS analysis highlights that near the fovea there are more cells with normal responses in most individuals with RP compared to their more temporal imaging locations. Perhaps most importantly, this indicates that even in areas with reduced function there are healthy cells remaining that could benefit from maintenance treatments to preserve/prolong their function.

One of the most interesting aspects of these results is the relationship between structure and function in these individuals. When comparing iORG population-RMS amplitude against the NND distance of the same cells, we observed a largely linear relationship in our normal controls, similar to that previously observed by Jiang et al. (29). However, this relationship changes significantly in individuals with RP, taking on the appearance of exponential decay as a function of NND. By extension, this indicates that very small changes in NND correspond to large changes in iORG amplitude in individuals with RP, plateauing at what appears to be the noise floor around ~5 AU, and it appears that the largest amount of change occurs over a NND range of only 2 microns.

The relationship between cone outer segment length as measured through Cirrus OCT and iORG population-RMS was less clear. Typically, individuals with RP had lower population-RMS amplitudes for a given cone outer segment length than controls, however there were numerous instances where the population-RMS amplitude for a given outer segment length from individuals with RP fell within the range of controls, and the prediction intervals for the two populations overlapped. There were several instances where individuals with RP had outer segment lengths that were the same as controls, but with reduced iORG amplitude and the regression for our participants with RP had a slightly steeper slope but had a substantially decreased y-intercept. This disconnect between outer segment length and amplitude in individuals with disease is consistent with prior observations using pORGs (44) and suggests that outer segment length alone can’t account for the observed reductions in amplitude. However, as our Cirrus OCT that lacks the transverse and axial resolution necessary to interrogate the OS length of individual cones, our OS length measurements were in effect an aggregate measurement and may therefore underestimate the remaining OS.

Finally, we compared iORG population-RMS amplitude to a gold standard for assessing cone function, microperimetry (MAIA). For many imaging locations, both reduced population iORG amplitude and retinal sensitivity were observed in individuals with RP when compared to controls. There was some overlap in retinal sensitivity between controls and some individuals with RP, for example the 1.5° location of 64010, the 1° temporal location 13090, and both the 1° and 2° temporal locations fell within a standard error of the retinal sensitivity of controls. Like NND, we observed an exponential relationship between retinal sensitivity and population iORGs, wherein a very small change in retinal sensitivity (<4dB) corresponded to a very large change in population iORG amplitude (>20AU). However, the dynamic range of the two assays were substantially different, with the iORG reaching its noise floor long before the MAIA (between 15–20dB).

This study had some limitations and potential biases that should be considered when interpreting its results. As mentioned above, the presence of cataracts and/or macular edema precluded successful AOSLO imaging in two individuals with RP. While we excluded individuals with cataracts to ensure our full stimulus was reaching the retina, nearly all participants had microedema and/or regions with poor image quality, and in some cases individual cones were unresolvable or ambiguous within the attempted imaging locations (especially in the peripheral retina). The split detection modality was used to aid in visualizing cone inner segments, but in cases where cones were still unresolvable, even with split-detection, that location was omitted from further analyses (53885, 60151). However, omitting individuals or locations with advanced disease can lead to selection bias at the population level, and future work using stimuli with varied imaging entrance locations (to avoid cataracts), or an ORG modality that is nominally less sensitive to inner retinal edema (e.g. AO-OCT) has the potential to reduce this source of bias. Further, as noted above, RP is a group of inherited retinal diseases that encompass many different genetic mutations (2, 72) and exhibit variable disease progression (10, 65, 73) and phenotypes (62). However, this study includes only seven individuals with RP (only 4 of which had genetic information) and is unlikely to be representative of all RP subtypes. Future genotype-specific work will be necessary to establish an iORG “signature”, if any, from the many RP subtypes.

While all efforts were made to maintain a consistent stimulus dose to the retina, inter-individual variability in ocular biometry means that the photons/µm² delivered to each cone varies slightly between individuals. To assess its contribution to the iORG variability observed in this study, the stimulus photon density was calculated for all participants using the model published by Jiang et al. (29) and was plotted against the population-RMS amplitude of each participant ( Supplementary Figure 2 ; data mean subtracted for visualization). The average stimulus photon density at the retina was 2.33x10⁷ photons/µm² (range = 1.86x10⁷ to 2.77x10⁷ photons/µm²). There was no significant difference in stimulus photon density between controls and individuals with RP (Student’s t-test, p = 0.475). A linear regression model was fit to the data and no significant relationship was found between mean subtracted population-RMS amplitude and stimulus photon density at the retina (y = 10–⁷x – 2.58; r^{2 =} 0.0014; p > 0.05). Of note, this model does use a linear retinal magnification factor and doesn’t factor in individual variations in absorptance of ocular structures (cornea, lens) and may exhibit nonlinearities on the extrema when factoring in more complex optical models of the eye (74, 75).

Overall, we found the differences between the iORG and MAIA to be quite interesting and a microcosm of the differences between psychophysical assays that integrate information across the entire visual system (microperimetry), and objective, cell specific assays such as optoretinography. For instance, we observed iORG RMS rapidly approaching and reaching its noise floor around 15–20dB in retinal sensitivity, implying that cones with sensitivities beneath these values are functioning so poorly that there is no optical path length change in these cells, and thus no iORG. However, assuming the participant is attending to their fixation task during microperimetry, there remains retinal sensitivity at this location, indicating that there remains at least some residual function such that the spot can be seen by the participant. Importantly, these data underscore the limitation of what optoretinography measures in this paradigm: path length changes in the cone outer segment. Thus, while optoretinography appears to be closely related to cone function in the early to moderate stages of disease, when a cell reaches a stage where the OS is disorganized or not present, the assay may no longer provide a sensitive assessment of cone health. In many ways, this sensitivity seems a boon for the iORG- and indeed, significant deficits were detected in regions with ostensibly normal retinal sensitivities. However, it does suggest caution in interpreting results from future optoretinography studies of other retinal diseases, as the absence of an ORG signal does not necessarily indicate that the cell is non-functional- it only means that there isn’t a detectable change in the cone outer segment.

Overall, these data suggest that microperimetry and measures of cone structure, at present, remain the best method for detecting complete cell loss, but optoretinography appears to be most sensitive to the earliest changes in cell function. Detecting these early changes in cell function may be valuable as either a screening tool, or for assessing the efficacy of therapeutic interventions.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Institutional Review Boards at the Medical College of Wisconsin (PRO00038673). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

MG: Conceptualization, Data curation, Methodology, Software, Writing – original draft, Writing – review & editing, Formal analysis, Visualization. TC: Data curation, Resources, Supervision, Writing – review & editing, Investigation. RC: Data curation, Resources, Supervision, Writing – review & editing, Conceptualization, Formal analysis, Funding acquisition, Methodology, Software, Validation, Writing – original draft, Visualization.