Cystoid Macular Edema in Retinitis Pigmentosa: Pathophysiologic Insight Using Swept-Source OCT and Angiography

Abstract

Objective

Cystoid macular edema (CME) is relatively common in patients with retinitis pigmentosa (RP), but its pathophysiology is poorly understood. This study aims to provide insight into pathophysiologic mechanisms of CME formation using swept-source (SS) OCT angiography (OCTA).

Design

Retrospective cross-sectional study at a tertiary referral center.

Participants

Seventy consecutive patients with molecularly confirmed RP (associated with RHO, USH2A, or RPGR) with and without CME.

Methods

All patients underwent complete ophthalmic examination with multimodal imaging, including SS-OCT with SS-OCTA (PLEX Elite 9000). Images were then analyzed in a semiautomated fashion using ImageJ.

Main Outcome Measures

Structural and vascular abnormalities seen on SS-OCT and SS-OCTA in patients with RP, and spatial overlap between CME areas, vascular flow loss in the deep capillary plexus (DCP), and preserved external limiting membrane (ELM) and ellipsoid zone (EZ).

Results

One hundred twenty-nine eyes from 70 patients (mean age 37 years, range 6 to 78) with molecularly confirmed RP were included. The mean best-corrected visual acuity measured 0.35 logarithm of the minimum angle of resolution (range –0.12 to 2.7). Overall, 31% of patients had CME (10% in RPGR, 36% in USH2A, 48% in RHO). There was a strong correlation between regions of CME and ELM preservation (97% spatial overlap overall) as well as DCP flow loss (83% spatial overlap). There was moderate correlation between regions of CME and EZ preservation (66% spatial overlap).

Conclusions

Cystoid macular edema is relatively common in RP, with different prevalence depending on genotype. There is strong spatial overlap between areas of DCP loss and ELM preservation regardless of genotype, suggesting a shared mechanism of impaired fluid clearance at the level of the DCP.

Financial Disclosure(s)

The author has no/the authors have no proprietary or commercial interest in any materials discussed in this article.

Retinitis pigmentosa (RP) is the most common inherited retinal dystrophy and represents a heterogeneous group of diseases that share features of retinal degeneration, primarily of the rod photoreceptors, leading to symptoms such as nyctalopia and poor peripheral vision.¹ A common contributor to central vision loss in RP is cystoid macular edema (CME), with reported prevalence estimates ranging from 15% to 70%.2, 3, 4, 5 However, the pathophysiology of CME in RP is poorly understood. Unlike CME in more common conditions such as diabetic retinopathy, CME in RP has minimal leakage seen on fluorescein angiography,^2,6 suggesting a lack of exudation related to the CME and implying that potential impaired fluid clearance mechanisms may contribute to its pathogenesis.

Recent advances in retinal imaging including OCT angiography (OCTA) allow for detailed structural and vascular analysis, providing insights into pathogenic processes. Prior studies using OCTA in RP have shown reduced retinal capillary flow, with greater flow reduction in the deep capillary plexus (DCP) than in the superficial capillary plexus (SCP).7, 8, 9 However, the clinical impact of such retinal vascular flow impairment in RP is not known, and there have been few prior studies investigating potential relationships between vascular flow loss and CME.

Spaide recently proposed a vascular theory of CME formation, highlighting 3 pathways of intraretinal fluid clearance primarily mediated by Müller cells: (1) through the internal limiting membrane (ILM) into the vitreous, (2) through Müller cells via aquaporin channels in the DCP, and (3) through the external limiting membrane (ELM) to the retinal pigment epithelium (RPE) and choroid.¹⁰ Because the ILM is generally intact in RP, we hypothesize that in patients with RP, the primary mechanism may be impairment of fluid clearance at the level of Müller cell nuclei and the DCP, particularly in regions of intact ELM where fluid cannot clear through the RPE. If so, CME in RP would be expected to correspond to regions of DCP loss and ELM integrity. In contrast, CME might be seen more rarely in regions where retinal degeneration has resulted in ELM loss.

In this study, we evaluate the retinal structural and vascular changes on OCTA in RP. Specifically, we correlate vascular changes in the retinal DCP with the distribution of the CME and in relation to areas of preserved ELM and ellipsoid zone (EZ). We relate these imaging findings to the expression of aquaporin in the human retina both by single-cell RNA sequence database query and immunohistochemical analysis of human donor eyes. Based on the study findings, we propose a theory of impaired intraretinal fluid clearance at the level of the Müller cells at the DCP as a potential shared pathophysiologic mechanism of CME in RP, regardless of genotype.

Methods

This was a retrospective observational case series evaluating clinical and multimodal imaging findings for consecutive patients with molecularly confirmed RP seen at the University of Iowa by the Inherited Retinal Disease Service from September 17, 2017 to August 30, 2023. This study adhered to the tenets of the Declaration of Helsinki and was conducted in accord with regulation set forth by the Health Insurance Portability and Accountability Act. Institutional review board approval was obtained from the Human Subjects Committee of the University of Iowa prior to conduct of the study.

Molecular confirmation of RP was performed through tiered molecular testing informed by the clinical phenotype, with assessment of pathologic variants through the John and Marcia Carver Nonprofit Genetic Testing Laboratory (University of Iowa, Iowa City, Iowa), as previously described.¹ To explore the OCTA findings across various forms of RP but limit the heterogeneity introduced by rare causes of RP, we identified patients with confirmed pathologic variants in either USHA, RPGR, or RHO, the most common causes of autosomal recessive, X-linked, and autosomal dominant RP, respectively. For each patient, the age, gender, and genotype were recorded. The best-corrected visual acuity (BCVA) for each eye, manifest refraction, and clinical findings on slit lamp and dilated fundus examination were extracted from the chart and recorded.

All patients included in the study underwent swept-source (SS) OCT and SS-OCTA imaging (PLEX Elite 9000; Carl-Zeiss Meditec Inc). The SS-OCT instrument uses a tunable laser with center wavelength between 1040 and 1060 nm, with an optical axial resolution of about 6 μm in tissue. The images utilized in this study were acquired using a version of the instrument with scanning speed of up to 100 000 A-scans per second. Scan protocols included 16-mm high-resolution horizontal line scans through the fovea and optic nerve, 12 × 12-mm SS-OCT macular volume scans, and 6 × 6-mm and 12 × 12-mm SS-OCTA scans centered on the fovea. For patients with multiple visits in the study period, the images from the initial visit were used. In our case series, no patients had CME outside of the central 6 × 6-mm scan area. As such, the 6 × 6-mm SS-OCTA scans were used for all subsequent analyses.

Swept-source OCTA scans were reviewed for quality using the manufacturer’s software (PLEX Elite, Version 2.0, Carl-Zeiss Meditec Inc). Eyes were excluded from image analysis if the OCTA was poor quality prohibiting identification of retinal boundaries for subsequent segmentation (e.g. poor signal due to media opacity), if there was concomitant or confounding macular pathology (e.g., epiretinal membrane), or alternate potential causes of CME (e.g., uveitis or diabetic retinopathy). Any segmentation errors were manually corrected utilizing the instrument’s software to ensure proper identification of the vitreous border at the ILM and outer retinal boundary at the RPE. The ILM was assessed for integrity and lack of vitreoretinal interface abnormality in all patients. For analysis of the vascular plexuses, we used the instrument’s algorithm after segmentation correction, which identifies the SCP from the ILM to the boundary of the inner plexiform layer and inner nuclear layer. The DCP is marked from the inner plexiform layer to inner nuclear layer boundary to the outer plexiform layer to outer nuclear layer. To evaluate the integrity of the EZ band and ELM, a custom avascular slab was designed with 30 to 60 micron offset anterior to the RPE, depending on the severity of the outer retinal degeneration in each individual case.

To colocalize areas of CME and DCP loss, the structural and vascular analysis panels from the DCP slab were extracted and imported into ImageJ (National Institutes of Health). In the structural DCP slab, automated thresholding was used to identify CME after manually excluding the peripheral macula to account for fringe or vignetting artifacts outside areas of central pathology of interest. In the DCP angiography slab, the DCP was skeletonized and blurred, followed by binarization and smoothing, to identify the central area of decreased DCP vessel density, using techniques as previously described.¹¹

To evaluate the EZ and ELM area, patients with a completely preserved EZ or completely absent EZ within the 6 × 6-mm scan area were noted. Those with partially preserved EZ and ELM underwent further processing using the modified avascular slab. Local contrast was first enhanced and the image was then blurred and binarized, with the peripheral macula manually excluded as before. The edges were then smoothed and isolated to identify the borders of EZ and ELM.^12,13

The identified en face borders of the EZ and ELM were compared with cross-sectional line scans to ensure proper segmentation. The analyzed slabs were then analyzed to measure area of CME, DCP loss, EZ preservation, and ELM preservation. The percentage overlap was used to colocalize and compare areas of CME and DCP loss with EZ and ELM preservation. To assess reliability of the metrics, 16 eyes (about 10% of the sample; 8 eyes with CME, and 8 without CME) were selected at random, and repeatability was assessed for each metric by intraclass correlation coefficient (ICC). A 1-way analysis of variance was then used to compare these metrics across patients with different genotypes (RHO, RPGR, and USH2A), and a 2-tailed 2-sample t-test was used to compare areas of DCP loss and EZ or ELM preservation between the CME and no CME groups. To explore structure and function relationships, for all patients, the correlation between BCVA and area of EZ and ELM preservation, DCP loss, and CME area was assessed with the Pearson correlation coefficient. A chi-square test was used to compare the binary incidence of CME between the specific gene groups as well, with P value <0.05 considered significant for all statistical analyses performed. Statistical analysis was performed in Excel (Microsoft).

Human donor eyes were obtained from the Iowa Lions Eye Bank following full consent of the donors’ next of kin and in compliance with the Declaration of Helsinki. Sucrose cryopreserved fixed cryosections (n = 3 control, n = 3 RP, n = 3 diabetic macular edema) were collected and immunohistochemistry was performed as described previously.¹⁴ Sections were incubated with antibodies directed against aquaporin-4 (AQP4) (antibody 5A4W4, Novus Biologicals) at a concentration of 1.8 to 18 μg/mL and visualized with Alexa-546-conjugated donkey antirabbit secondary antibody (Invitrogen; 5 μg/mL). For some experiments, sections were triple labeled with anticollagen IV (M3F7, Developmental Studies Hybridoma Bank) at a concentration of 0.24 μg/mL and visualized with Alexa-647 donkey antimouse secondary antibody (Invitrogen; 10 μg/mL) and with fluorescein-conjugated Ulex europaeus agglutinin-I lectin (Vector Laboratories, diluted 1:50 from the manufacturer). In some cases, sections were dual labeled with anti-AQP4 and cellular retinaldehyde binding protein, a marker of Müller cells and RPE (Abcam, ab15051, 1 μg/mL). Photomicrographs were collected on a fluorescence microscope (Olympus BX41). Specificity of the anti-AQP4 antibody was confirmed by blocking with an excess of recombinant AQP4 protein (Abnova), as described previously.¹⁵

To evaluate expression of aquaporins across the retina, the Spectacle website database was interrogated (https://singlecell-eye.org) using data from a study on regional gene expression in neural retina.^16,17

Results

A total of 129 eyes from 70 patients (28 male, 42 female) were included for analysis. Average age was 37 years (median 36, range 6-78). The mean BCVA at final visit was 0.35 logarithm of the minimum angle of resolution (median 0.24; range –0.12 to 2.7). Out of the included eyes, 29 were from patients with RHO-associated RP, 39 from RPGR-associated RP (including 9 manifesting carriers), and 61 from USH2A-associated RP (41 syndromic and 20 nonsyndromic cases). Demographic and baseline characteristics are summarized in Table 1.

Table 1.

Baseline Characteristics of Included Patients Overall and by Affected Gene

	Overall	RHO	RPGR	USH2A
Number of eyes	129	29	39	61
Number of patients	70	15	21	34
Mean age (standard deviation)	37 (17)	38 (18)	33 (17)	40 (18)
Number of female patients (%)	42 (60)	10 (67)	8 (38)	24 (71)
Number of right eyes (%)	66 (51)	16 (55)	20 (51)	30 (49)
Mean BCVA (logMAR)	0.35	0.42	0.41	0.28

BCVA = best-corrected visual acuity; logMAR = logarithm of the minimum angle of resolution.

Cystoid macular edema was present in 31% of eyes overall (40/129 eyes), with prevalence varying by genotype, including 48% of eyes (14/29) with RHO-associated RP, 10% of eyes (4/39) with RPGR-associated RP, and 36% of eyes (22/61) with USH2A-associated RP (Table 2). The difference in prevalence of CME among the different genotypic groups was statistically significant (P = 0.002). Of note, 24 eyes assessed had no remaining EZ or ELM identifiable, and none of these patients had CME.

Table 2.

Summary of Areas (mm²) of ELM and EZ Preservation, DCP Loss, and Prevalence of CME by Genotype

	Overall	RHO	RPGR	USH2A	P
ELM area (mm2)	9.68	9.48	10.01	9.57	0.97
EZ area (mm2)	5.25	4.07	7.92	4.10	0.15
DCP area (mm2)	1.82	1.57	2.50	1.51	0.002∗
Number of eyes with CME (%)	40 (31)	14 (48)	4 (10)	22 (36)	0.002∗

CME = cystoid macular edema; DCP = deep capillary plexus; ELM = external limiting membrane; EZ = ellipsoid zone.

^∗Statistical significance (P < 0.05).

Example images from a patient with USH2A-associated RP are shown in Figure 1 to illustrate the slabs used for identifying CME as well as areas of DCP loss and ELM or EZ band preservation. Figure 2 shows example images of each slab with en face segmentation as well as a composite image overlaying CME with areas of DCP loss and ELM or EZ band preservation. Repeatability of these metrics was excellent, with an ICC >0.90 for all metrics, including ELM area (ICC = 0.93), EZ area (ICC = 0.96), DCP loss (ICC = 0.96), and CME area (ICC = 0.92). For all included eyes, the average area of ELM preservation was 9.68 mm², the average area of EZ preservation was 5.25 mm², and the average area of DCP loss was 1.82 mm². When stratifying by affected gene, there was no statistically significant difference in areas of ELM or EZ preservation; however, eyes from patients with RPGR-associated RP had a larger area of DCP loss compared to those from USH2A- and RPGR-associated RP. The areas of ELM and EZ preservation, and DCP loss are reported for all patients and by genotype in Table 2. In all eyes, there was a weak correlation between BCVA and areas of ELM (r = 0.37) or EZ preservation (r = 0.26), and no correlation between BCVA and DCP loss (r = 0.05).

Figure 1.

Example images from a patient with USH2A-associated retinitis pigmentosa and CME. Fundus photograph (A) with overlay of the 6 × 6-mm scan area (dotted box) imaged by OCT and angiography. Corresponding line scan (B; yellow line in A) shows CME with central preservation of the ELM (white arrowhead) and EZ band (yellow arrowhead). Cystoid macular edema (arrowhead) is seen en face(C) on the structural DCP slab, with corresponding areas of DCP loss (D; arrowhead). A custom outer retina slab (E) results in en face visualization of areas of ELM (white arrowhead) and EZ preservation (yellow arrowhead). CME = cystoid macular edema; DCP = deep capillary plexus; ELM = external limiting membrane; EZ = ellipsoid zone.

Figure 2.

Example segmentation and overlay for a patient with RHO-associated retinitis pigmentosa and CME. Panels on the left show en face segmentation of areas of CME (A), DCP loss (B), and EZ band (white arrowhead) as well as ELM (yellow arrowhead) preservation (C). Segmented areas shown in A–C are colored for ease of comparison (A’–C’). Overlay composite image of A’–C’ is shown in panel D, with each segmented area (CME, DCP, EZ, ELM) shown as labeled. CME = cystoid macular edema; DCP = deep capillary plexus; ELM = external limiting membrane; EZ = ellipsoid zone.

When comparing eyes with or without CME overall, there was no statistically significant difference in BCVA, or area of ELM or EZ preservation. However, when evaluating only eyes with identifiable EZ and ELM (i.e., excluding patients with complete EZ or ELM loss), eyes with CME had a larger average area of DCP loss (P= 0.03), but still no statistically significant difference in area of ELM or EZ preservation (Table 3).

Table 3.

Comparison of BCVA, Areas of ELM Preservation, EZ Preservation, and DCP Loss (mm²) between All Eyes with and without CME

	CME	No CME	P
Number of eyes	40	89
BCVA (logMAR)	0.29	0.38	0.33
DCP loss area (mm2)	1.73	1.44	0.03∗
ELM preservation area (mm2)	7.54	9.68	0.31
EZ preservation area (mm2)	1.99	3.10	0.07

BCVA = best-corrected visual acuity; CME = cystoid macular edema; DCP = deep capillary plexus; ELM = external limiting membrane; EZ = ellipsoid zone; logMAR = logarithm of the minimum angle of resolution.

^∗Statistical significance (P < 0.05).

In patients with CME, there was high average spatial correlation between areas of CME and ELM preservation (97%) as well as with CME and areas of DCP loss (83%). As expected, the degree of spatial overlap between areas of CME and EZ preservation was lower (66%). There was no statistically significant difference when comparing the degree of spatial overlap across different causative genes. The metrics of area overlap are summarized in Table 4. In eyes with CME, there was moderate negative correlation between BCVA and the area of CME (r = –0.43).

Table 4.

Summary of Percentage of Area Overlap between CME and ELM, EZ, or DCP, Respectively

	Overall	RHO	RPGR	USH2A	P
CME and ELM	97%	99%	100%	95%	0.19
CME and EZ	66%	77%	70%	58%	0.16
CME and DCP	83%	74%	85%	88%	0.06

CME = cystoid macular edema; DCP = deep capillary plexus; ELM = external limiting membrane; EZ = ellipsoid zone.

When evaluating the expression of aquaporins using publicly available human retinal single cell RNA sequencing data sets, AQP4 was found to be the major aquaporin of the neural retina, with highest expression in Müller and amacrine cells (Fig S1, available at www.ophthalmologyscience.org).^16,17 To evaluate the expression of AQP4 relative to retinal structures including Müller cells and the retinal vasculature, immunohistochemistry was performed on a series of human donor eyes. The pattern of labeling was consistent with the distribution of Müller glia and astrocytes with labeling throughout the neural retina ending abruptly at the ELM as well as throughout the ganglion cell layer (Fig 3). While AQP4 was present throughout the retina, labeling was especially robust around the microvasculature (identified by anticollagen IV and Ulex europaeus agglutinin-I lectin) in each of the vascular plexuses. The AQP4 signal was consistently external to the anticollagen IV signal, also consistent with a Müller cell (rather than endothelial or pericyte) source. No signal was detected in the RPE or choroid. In this sample set, no consistent differences in AQP4 expression were observed between normal retinas and those with macular edema.

Figure 3.

Immunofluorescence of AQP4 in human retina in relation to capillaries. Left: AQP4 localization. Note the space filling, Müller-like distribution and abrupt termination of labeling at the external limiting membrane, as well as enriched signal surrounding the retinal vasculature (arrows). Middle: visualization of the retinal vasculature labeled with UEA-I. Right: merged image with AQP4 (red), UEA-I (green), and 4′,6-diamidino-222-phenylindole (DAPI) counterstain (blue). Lower panel: higher magnification view of the relationship between the endothelium (green), its basal lamina (blue) and AQP4 distribution. Nuclei are pseudocolored white. Scale bar = 25 μm. AQP4 = aquaporin-4; GCL = ganglion cell layer; ILM = internal limiting membrane; INL = inner nuclear layer; IS = inner segments; ONL = outer nuclear layer; OS = outer segments; UEA-I = Ulex europaeus agglutinin-I.

Discussion

Cystoid macular edema is a well-known contributor to central vision loss in patients with RP, but the underlying pathophysiologic mechanisms of CME formation in the context of RP remain elusive. Using SS-OCT and OCTA in a consecutive series of patients with the most common forms of autosomal recessive (USH2A), autosomal dominant (RHO), and X-linked (RPGR) RP, we show that CME is common overall (31%), with the prevalence of CME varying by genotype (most common in RHO-associated RP at 48%).

Due to the relative lack of vascular leakage in RP, including on fluorescein angiography, we hypothesize that CME in RP may be related to impaired fluid clearance through Müller cells, particularly at the level of the DCP, which is not well-visualized by fluorescein angiography.¹⁸ Supportive of this hypothesis, eyes with CME in our study had high spatial correlation with areas of DCP flow loss on OCTA, as well as ELM preservation. Using single-cell RNA sequencing data from human retina, we further showed that AQP4 is the most highly expressed aquaporin in the retina and is expressed predominantly in Müller and amacrine cells. Immunohistochemical staining of human donor eyes confirmed colocalization of AQP4 with Müller cells and retinal vasculature.

Cystoid macular edema occurs due to an imbalance of fluid flow within the retina. As there is no lymphatic system with the eye, intraretinal fluid is primarily regulated by the retinal vasculature and Müller cells. Spaide previously described 3 main routes for fluid clearance in the retina as moderated by Müller cells: up through the ILM into the vitreous, down through the ELM into the RPE and choroid, and within the retina through the DCP, as partially constrained by the plexiform layers (inner and outer), which straddle the inner nuclear layer.^10,19, 20, 21, 22 The DCP has a tight anatomic relationship with the Müller cell bodies, and the interface between these cell types colocalizes with AQP4. Exudative conditions such as a diabetes and retinal vein occlusion have been shown to have flow loss in the DCP in addition to expected changes in the SCP.^9,23, 24, 25 Based on the published literature, a schematic of fluid flow mediated by the Müller-DCP complex is depicted in Figure 4.

Figure 4.

Schematic of intraretinal fluid clearance as mediated by Müller cells. The inner and outer retinal fluid boundaries are determined by the ILM and ELM, respectively. Intraretinal fluid outflow is mediated by aquaporin channels (inset, bottom left) expressed at the junction of the DCP and Müller cell bodies (inset, top left). DCP = deep capillary plexus; ELM = external limiting membrane; EZ = ellipsoid zone band; GCL = ganglion cell layer; ILM = internal limiting membrane; INL = inner nuclear layer; IPL = inner plexiform layer; ONL = outer nuclear layer; OPL = outer plexiform layer; RNFL = retinal nerve fiber layer.

Photoreceptors are highly metabolic cells, producing water as a byproduct of multiple pathways, including glycolysis.^19,26 In RP, the primary loss of photoreceptors in RP coupled with a relatively preserved choroid leads to hyperoxygenation of the outer retina, which then causes reactive vasoconstriction of the retinal vasculature. Accordingly, previous studies of RP using OCTA have shown a relative decrease in vessel density in OCTA in both the superficial and DCPs.27, 28, 29, 30 Evidence suggests greater susceptibility of the DCP (compared to SCP) to these vasoconstrictive changes, including lower vessel density and greater flow loss on OCTA in the DCP in eyes from patients with RP.7, 8, 9^,31 Similar findings have also been demonstrated longitudinally in a mouse model of RP.³² In this study, we show colocalization of CME in RP with regions of DCP loss, implying that in these areas, decoupling of the Müller–DCP complex may result in impaired fluid clearance via aquaporin channels, resulting in nonexudative CME. A summary of this proposed theory is shown in Figure 5.

Figure 5.

Proposed pathophysiologic mechanism for the development of CME in retinitis pigmentosa. Left to right: photoreceptor loss in retinitis pigmentosa results in decreased oxygen consumption and increased intraretinal oxygen tension. This leads to retinal vascular constriction preferentially impacting the DCP, which is exposed to higher oxygen tension compared to the SCP. Deep capillary plexus loss results in decoupling of AQP4 channels at the Müller cell-–DCP complex. This leads to impaired fluid clearance in regions of DCP loss and the formation of nonexudative CME. AQP4 = aquaporin-4; CME = cystoid macular edema; DCP = deep capillary plexus; SCP = superficial capillary plexus.

In RP, the inner retina, including the ILM, is largely preserved, thus maintaining the normal fluid barrier to the vitreous cavity, highlighting the potential importance of the AQP4-mediated mechanism of intraretinal fluid clearance. The ELM likely also plays an important role in fluid clearance. Changes in the ELM and EZ have been shown to correlate with visual acuity in retinal vascular disease including diabetic macular edema.³³ In RP, EZ and ELM preservation have been correlated with disease severity and other features of disease including SCP loss and response to therapy.^34,35 In our study, ELM preservation showed the highest colocalization to CME, suggesting that CME occurs in regions of preserved ELM. This could explain the relatively lower incidence of CME in advanced RP, as no patient with complete EZ and ELM loss in our cohort had CME.

Limitations in our study include limited sample size and retrospective cross-sectional design. Retinitis pigmentosa is a heterogeneous disease with many causative genes, and we limited our analysis to only the most common genes in RP for autosomal recessive, X-linked recessive, and autosomal dominant inheritance. However, these genes differ in mechanism of retinal degeneration, which could explain the difference in prevalence in CME. A larger sample size could improve the power of detecting differences in OCTA parameters, including within smaller subgroups such as specific genotypes. The cross-sectional study design here limits assessment of the temporal relationship DCP loss and the development of CME. A longitudinal study which follows patients where CME develops or resolves could provide further evidence to support the role of DCP loss in CME formation. Cystoid macular edema tended to occur in patients with intermediate severity; no patients with complete ELM/EZ loss, and only 1 patient with complete ELM/EZ preservation, had CME in our cohort.

Sound pathophysiologic theory is the basis for further scientific discovery. In RP, improved understanding of fluid clearance could lead to better therapies to prevent and treat vision loss from CME. Moreover, insight into mechanisms of impaired retinal fluid flow may have relevance to a wide range of other conditions with CME, including diabetes, retinal vein occlusion, macular degeneration, and Irvine–Gass syndrome. For example, Müller cell markers like AQP4 and glial fibrillary acidic protein have been shown to be altered in the aqueous humor of patients with diabetes, and the degree of altered expression has been correlated with the level of retinal disorganization.³⁶ Alterations in Müller cell morphology and AQP4 expression have also been demonstrated in outer retinal atrophy secondary to age related macular degeneration.³¹

Although these pieces of evidence suggest that AQP4 channels play a critical role in intraretinal fluid management, the process of fluid clearance is dynamic, and other mechanisms may also be involved. For example, osmotic gradients within the retina tissue are also important, and some studies have suggested that AQP4 expression may be downregulated in the context of hyperosmotic fluid, including in macular edema.³⁷ Other water and solute transport mechanisms, including the inward rectifying potassium channel Kir4.1, may also play a role.^38,39 More studies are needed to understand the role of aquaporins and the dynamics of intraretinal fluid through the Müller–DCP complex in healthy and disease states.

Manuscript no. XOPS-D-25-00997.