Perifoveal Choriocapillaris Flow Deficits Associated With Cerebrospinal Fluid Aβ₄₂/tau in Presymptomatic Alzheimer's Disease

Abstract

Purpose: This study investigated the association between swept-source OCTA (SS-OCTA) choriocapillaris flow deficit percentage (CC FD%) and cerebrospinal fluid (CSF) biomarkers in presymptomatic Alzheimer's disease (AD).

Methods: Twenty-three cognitively healthy (CH) participants, including those with pathological (CH-PAT) and normal (CH-NAT) Aβ₄₂/tau ratios, underwent lumbar puncture for CSF Aβ₄₂/tau and ptau-181 quantification using electrochemiluminescence assays. OCTA en face images of the choriocapillaris were analyzed for flow deficits within a 6 × 6 mm macular grid and 3- and 6-mm Early Treatment of Diabetic Retinopathy Study (ETDRS) perifoveal rings. The association between CC FD% and CSF Aβ₄₂/tau and CSF ptau-181 levels was evaluated.

Results: Perifoveal microvascular changes in the choriocapillaris outer ring (6-mm ETDRS ring) were significantly correlated with the CSF Aβ₄₂/tau ratio in CH-PATs compared to CH-NATs. However, linear regression analysis across all CH participants (CH-PAT + CH-NAT) revealed that only age was significantly associated with CSF ptau-181 levels.

Conclusions: Our findings suggest that this cohort is in an early preclinical AD stage, without ptau-181 biomarker evidence of established AD. SS-OCTA measures may complement CSF A/T (amyloid/tau) levels, helping to define transitional stages from CH-NAT to CH-PAT during presymptomatic AD.

Translational Relevance: By identifying the transition from normal aging to presymptomatic AD, SS-OCTA metrics combined with biofluid markers could enhance AD oculomics by stratifying disease risk and prioritizing treatment interventions.

Introduction

Vascular changes in the central nervous system contribute to the earliest stages of presymptomatic Alzheimer's disease (AD). Neurodegeneration follows a cascade of blood-brain barrier dysfunction. Pericyte degeneration and detachment lead to blood-brain barrier leakage and neuronal damage from oxidative stress and edema. Both the retinal and cerebral vessels undergo structural and functional changes with aging, including endothelial cell dysfunction, increased vessel stiffness, and reduced capillary density. These changes contribute to age-related vascular diseases and neurodegeneration.¹ 

The retina could serve as a window into the brain, providing insights into cerebral small vessel disease through innovative, non-invasive, high-resolution, and fast in vivo imaging techniques.² This relationship underscores the importance of further research into shared mechanisms to develop therapeutic strategies that target both retinal and cerebral microvasculature.³ The perifoveal microvasculature, especially the choriocapillaris, may be an effective screening biomarker for cerebral hypoperfusion, a precursor to dementia.⁴ Decreased perfusion in the perifoveal choriocapillaris may be one of the earliest pre-AD retinal biomarkers because the highly metabolic retinal pigment epithelial (RPE) cells and photoreceptors would be most sensitive to oxidative stress leading to photoreceptor dysfunction.⁵ 

In a previous publication, we demonstrated that the perifoveal choriocapillaris flow deficit percentage (CC FD %) measured within the Early Treatment of Diabetic Retinopathy Study (ETDRS) macular grid, distinguished presymptomatic AD individuals from normal aging controls.⁶ In the present study, we aimed to find the relationship between the CC FD % and cerebrospinal fluid (CSF) biomarkers, including the Aβ₄₂/tau ratio and CSF ptau-181 level, to help differentiate the CH-NAT and the CH-PAT stage of presymptomatic AD. 

Methods

This prospective, cross-sectional observational study received approval from the Institutional Review Boards of the University of California Los Angeles (no. 15-000083) and the Huntington Medical Research Institute (no. 33797). Informed consent was obtained from all participants. The study adhered to the Declaration of Helsinki and the Health Insurance Portability and Accountability Act. 

Study Participants

Participants aged ≥60 years were recruited at the Huntington Medical Research Institutes in Pasadena, California, USA, for cognitive aging research. All participants underwent neuropsychological testing. Biofluid collection and retinal imaging were performed within one year for 92% of participants and within two years for the remainder. 

Retinal Scan Selection

Eligibility for the retinal imaging study included participants with normal cognitive functioning according to age, sex, and education standards based on comprehensive neuropsychological testing. In addition, CSF assays were used to categorize participants according to an Aβ₄₂/tau ratio of 2.7. This ratio is shown to classify over 85% of probable preclinical AD cases among cognitively healthy individuals as those with cognitively healthy pathological Aβ₄₂/tau (CH-PAT, or “pre-clinical AD”) and cognitively healthy non-pathologic Aβ₄₂/tau (CH-NAT).⁷ Participants could have none or only one vascular risk factor, such as hypertension, with less than 1% having only one. A single vascular risk factor is common and not necessarily linked to increased cerebrovascular disease in older adults with AD.^8,9 

Swept-Source OCTA Imaging

OCTA images of the macula (6 × 6 mm) from CH subjects were captured using the PLEX Elite 9000 (Carl Zeiss, Dublin, CA, USA) and assessed for quality. The device features a swept light source with a central wavelength of 1050 nm, a bandwidth of 100 nm, and an axial resolution in the tissue of ∼5 µm. The device operates at 100,000 A-scans per second and uses the optical microangiography algorithm to generate the OCTA image.¹⁰ The choriocapillaris en face OCTA slab was defined by a 16-µm thick slab segmented at the level of the Bruch's membrane with the inner boundary placed 4 µm below the Bruch's membrane.⁶ The boundaries were inspected, and segmentation errors were manually corrected. After the removal of projection artifacts, the images were screened for quality. Images with a signal strength below 7 or the presence of significant artifacts or noise were excluded. Our study cohort were normal aging subjects with no vitreous degeneration causing decreased vision. Based on our study visit eye examination data, the best-corrected visual acuity was 20/20 OU, or 20/25 in either eye. Freiburg contrast sensitivity testing was normal for their age range and helped confirm no age-related vitreous floaters affecting our swept-source optical coherence tomography angiography (SS-OCTA) scan quality. Based on the SD-OCT scans, we selected the eyes without drusen to compute the choriocapillaris flow deficit. Binarized CC en face OCTA images, using a local threshold method—Phansalkar (radius ∼20 µm) were used to quantify the percentage of CC FD. The flow deficits with a <24 µm diameter were excluded from the analysis, because they are considered to fall within a normal intercapillary distance. The function computes the final FD in terms of area % “Analyze Particles” provided by ImageJ and represents the percentage of flow deficits within the 6 × 6 mm CC en face OCTA slab after masking for superficial and large vessels.^11–16 

The computation of CC FD % was assessed within (1) the entire 6 × 6 mm CC en face OCTA slab, (2) the 6-mm ETDRS ring (outer ring), and (3) the 3-mm ETDRS ring (inner ring). Specifically, the thickness of the outer ring is 1500 um; the thickness of the inner ring is 1,000 um. The inner boundary was placed at 1500 µm and 500 µm from the foveal center, respectively, as described in our previous publication by Corradetti et al.⁶ Moreover, the study excluded individuals with previously diagnosed glaucoma beyond its initial stages, those who had undergone prior retinal treatments in the examined eye, patients with a history of ocular inflammation, and those with pre-existing retinal and choroidal conditions confirmed through structural OCT and SS-OCTA. In cases where both eyes were eligible, only one eye per patient was selected based on the highest-quality image (Note: The same retinal imaging dataset acquired from our previous work⁶ was evaluated in this new study for any correlations with CSF AD biomarkers). 

Cerebrospinal Fluid Biomarker Assays

Cerebrospinal fluid biomarker assays Aβ₄₂, total tau, and ptau-181 were performed by electrochemiluminescence on the MESO Quick Plex SQ 120MM platform (MSD, Rockville, MD, USA) at the Huntington Medical Research Institute Analytical Biochemistry Core Laboratory. On the date of sample collection, each participant's CSF was spun in a centrifuge at 3000 RCF for three minutes at 20°C. CSF 1 mL aliquots were transferred, accessioned in the LIMS (Cloud LIMS, Wilmington, DE, USA), and stored at −80°C until the testing date. All assays were performed at 20°C to 26°C to ensure consistent signals between runs. Cerebrospinal fluid samples were prepared with a 2× dilution and tested with a sample volume of 25 µL/well. CSF Aβ₄₂ was measured using the V-Plex Aβ Peptide Panel 1 (6E10) kit; total Tau was measured using the S-PLEX Human Tau (Total) kit; and ptau181 was assayed using the S-PLEX Human Tau (pTau181) kit (MSD). Samples were analyzed as duplicates with an eight-point calibration curve to assess assay and batch performance. Biomarker data were processed using the Discovery Workbench Software (MSD) and verified for quality control, including a calibration curve signal recovery between 75% to 125% and a sample replicate coefficient of variation <15%. Any nonconforming samples or batches were prepared and analyzed again to confirm the results. 

Statistical Analysis

Statistical analyses were performed using SPSS Statistics 29 for Macintosh (IBM, Chicago, IL, USA). One study eye of each CH participant was used to determine the association between OCTA metrics and CSF Aβ₄₂/tau and CSF ptau181 levels. Means and standard deviations were used to report the quantitative variables. Pearson's correlation coefficient was used to analyze the correlation between the choriocapillaris flow deficit parameters and biofluid levels in the PATs and NATs groups. A multivariate linear regression model, adjusting for age, assessed the associations between CC FD % and CSF biomarkers. P values ≤ 0.05 were considered statistically significant. 

Results

Study Cohort Characteristics

Twenty-three eyes of 23 participants were analyzed. Fifty-six percent of subjects with a CSF Aβ₄₂/tau ratio of < 2.7 were categorized into the CH-PAT group, and 44% of subjects with a ratio > of > 2.7 were assigned to the CH-NAT group.⁶ Females comprised 53.8% of the CH-PAT group and 70% of the CH-NAT group. The mean age was 82.2 ± 6.6 (SD) years for CH-PAT and 74.3 ± 11.7 (SD) years for CH-NAT (p = 0.07). Most subjects had low-risk ApoE genotypes for AD: 46.1% of CH-PATs had ApoE ε3/3 and 23% had ApoE ε3/4. In contrast, 70% of CH-NATs had ApoE ε3/4 and 20% had ApoE ε4/4. None of the CH-NATs had ApoE ε4/4 (Table 1). None of the individuals in this study had a history of or showed signs of ocular, neurological, or other systemic diseases.⁶ All images included in the study had a signal strength of at least 9. 

OCTA Parameters and CSF AD Biomarkers

The CC FD % in the outer 6-mm ring (P = 0.0006) was more significantly associated with the CSF Aβ₄₂/tau levels than the inner ring (P = 0.0017) in the CH-PATs versus CH-NATs group and appeared to be an independent predictor for CSF Aβ₄₂/tau in both the 6- and 3-mm ETDRS rings (Fig. 1). Compared to the CC FD % measured in the entire slab and in the inner ring, the CC FD % in the outer ring flow deficit % (P < 0.001) also had the strongest positive correlation with the CSF Aβ₄₂/tau ratio (β-coefficient = 0.206; confidence interval = 0.110–0.302). (Fig. 2) These findings suggest that the outer ring CC FD % may be the strongest independent predictor for the CSF Aβ₄₂/tau ratio in the CH-PATs group (Table 2). On the other hand, none of the OCTA metrics, except for age, were associated with CSF ptau-181 levels in all CH volunteers. 

Figure 1.

 

The top row shows CC slab images from a subject with an abnormal CSF Aβ₄₂/tau ratio; the bottom row shows CC slabs from a subject with a normal CSF Aβ₄₂/tau ratio. The images in column A, both top and bottom, depict the en-face OCTA CC slabs segmented at the Bruch's membrane (BM) level, with the inner boundary 4 um below the BM and the outer boundary 20 um below the BM. The images in column B, both top and bottom, show the FDs as black pixels in the binarized en-face OCTA CC slab using local Phansalkar thresholding. The images in column C, both top and bottom, display the FDs as black pixels in the binarized CC slab with the superficial mask that excludes the superficial and larger vessels from the CCFD % calculation. In column D, both images illustrate the final detected FDs calculated as CCFD %.

Figure 2.

 

Regression analyses (data from Table 1) for the outer ring choriocapillaris (top), the inner ring choriocapillaris (middle), and total choriocapillaris (bottom) flow deficits associated with CSF Aβ₄₂/tau ratio in the PATs group. CC outer, outer choriocapillaris flow deficits; CC inner, inner choriocapillaris flow deficits; CC total, total choriocapillaris flow deficits; PATs, study subjects with pathological CSF Aβ₄₂/tau ratios, a risk factor for progression to dementia.

Table 1.

 

Clinical Features of Our Study Cohort

 

Table 2.

 

Pearson's Correlation Analysis Showed a Statistically Significant Correlation Between the Aβ₄₂/tau Ratio and the CC Outer FD, CC Inner FD, and Total CC FD

 

Discussion

The CC FD % in the outer 6-mm ring was more significantly associated with the CSF Aβ₄₂/tau level than the inner 3-mm ring in the CH-PATs versus CH-NATs group. Most importantly, the outer ring CC FD % strongly predicted the CSF Aβ₄₂/tau ratio in the CH-PATs group. The higher CC FD % within the 6 mm ring (outer ring) in CH-PATs may suggest that CC FD % in the pre-AD eye and the microvascular changes may follow a centripetal spatial pattern that appears to progress towards the foveal center. These findings align with the spatial distribution of the retinal Aβ₄₂ plaques in post-mortem AD eyes,^17,18 in which these plaques were located predominantly in the superior temporal and Inferior temporal retinal periphery and within the innermost retinal layers supplied by the superficial vascular plexus.¹⁸ This unique spatial distribution of the disease distinguishes AD from age-related macular degeneration which does not show a particular predilection for regions within the macula. The increased perifoveal CC FD % in the CH-PAT subjects appears to parallel the cerebral hypoperfusion in early AD. In the TgF344-AD rat model,¹⁹ retinal Aβ deposits have been associated with an inflammatory response mediated by activated microglia and complement proteins in the ocular choroid. It has been postulated that this chronic inflammation may drive chronic oxidative stress-related photoreceptor dysfunction.²⁰ 

Not only did the CC FD % in the outer ring distinguish presymptomatic AD from normal aging individuals, as shown in our previous study,⁶ but it was also significantly predictive of the CSF Aβ₄₂/tau ratios in the CH-PATs versus CH-NATs group. Our results align with another recent study, Kwapong et al.²¹ that showed decreased choriocapillaris perfusion density by SS-OCTA correlated with increased serum Aβ_42, Aβ_42/40. Our SS-OCTA results also appeared to identify retinal vascular changes in a cohort at an earlier AD stage compared to those in other publications. Unlike the study by Kwapong et al.,²¹ which showed that decreased choriocapillaris density correlated with ptau-181 levels, none of our OCTA metrics were significantly associated with CSF ptau-181 at the cognitively normal stage before MCI on all CH participants. The lack of correlation of our choriocapillaris OCTA metrics with CSF ptau-181 levels represents a stage in which our CH-PATs group has not yet developed a definitive trajectory toward symptomatic AD or is still in the cognitively resilient stage. The decreased choriocapillaris density found by Kwapong et al.²¹ appears to represent vessel loss, a later histopathological stage than ours with the increased perifoveal CC FD %. This same group of authors¹⁶ further confirmed that their decreased choriocapillaris density was associated with longer MCI duration and worsening cognitive impairment. 

The major strengths of our study include its prospective nature and our well-characterized and curated cohort of individuals with virtually no comorbidities or concomitant ocular disorders to confound the analysis. The relatively small sample size was the main limitation of our study. A larger sample size in a longitudinal study over at least five or more years is proposed to confirm our preliminary findings in this small pilot study. The refraction, axial length, and lens status were unavailable for all our study subjects. However, as we measured the CCFD only in the 6 × 6 mm images centered in the macula, the axial length would not have affected our results. This study did not include subjects with refractive errors > 3D during the screening. 

In conclusion, our SS-OCTA results uniquely illustrate the spatial progression of the disease, starting from the retinal periphery towards the macula. The choriocapillaris flow deficit in the outer ring of the ETDRS grid was significantly more predictive of CSF Aβ₄₂/tau compared to the inner ring choriocapillaris flow deficit in the CH-PATs versus CH-NATs group. Our results also highlight how SS-OCTA metrics can be analyzed in tandem with biofluid markers to define the transitional stages from CH-NAT to CH-PAT that might eventually lead to symptomatic AD or other dementia types. Longitudinal studies will be required to confirm the temporal sequence of OCTA findings and biofluid markers. Because our study cohort was recruited from a normal cognitive brain aging study, not from a dementia clinic, PiB labeling and amyloid positron emission tomography (PET) scans were not performed at the time of our recruitment. Cerebrospinal fluid biomarkers start to become abnormal around the time of amyloid PET²² abnormality and earlier than neocortical tau PET changes when MTBR-tau-243 and ptau-205 become elevated.^23–25 Quanterix plasma biomarkers, especially ptau-217 alone, have been reliably shown as a tau PET proxy to discriminate intermediate/high from none/low AD neuropathological changes.²⁶ Our study results imply that retinal imaging might complement biofluid markers to help define the transitional stages of CH-NATs evolving to CH-PATs. Disease staging with a panel of molecular markers and non-invasive, in vivo retinal imaging markers could potentially reduce our reliance on costly PET scans (averaging $5000–$6000) that also have the risk of radiation exposure.²⁷ 

Diagnosing AD at its earliest stage is vital for implementing therapeutic interventions to prevent irreversible cognitive decline. If validated longitudinally in larger diverse populations, the correlations of SS-OCTA metrics with biofluid markers could be part of the ATN-IVS (amyloid/tau/neurodegenerative/inflammatory/immune/vascular) framework²⁸ and pre-AD oculomics for stratifying disease risk and prioritizing treatment. 

Acknowledgments

Supported by Kuen Lau Research Foundation. 

Disclosure: J.W. Chan, None; G. Corradetti, Nidek Inc. (R); X. Wu, None; N. Astraea, None; X. Arikaki, None; A.N. Fonteh, None; A.M. Suchy-Dicey, None; S. Sadda, 4DMT (C), Abbvie (C), Alexion (C), Allergan Inc. (C), Alnylam Pharmaceuticals (C), Amgen Inc. (C), Apellis Pharmaceuticals, Inc. (C), Astellas (C), Bayer Healthcare Pharmaceuticals (C), Biogen MA Inc. (C), Boehringer Ingelheim (C), Carl Zeiss Meditec (C, R, F), Catalyst Pharmaceuticals Inc. (C), Centervue Inc. (C, F), GENENTECH (C), Gyroscope Therapeutics (C), Heidelberg Engineering (C, R, F), Hoffman La Roche, Ltd. (C), Iveric Bio (C), Janssen Pharmaceuticals Inc. (C), Nanoscope (C), Notal Vision Inc. (C), Novartis Pharma AG (C), Optos Inc. (C, F), Oxurion/Thrombogenics (C), Oyster Point Pharma (C), Regeneron Pharmaceuticals Inc. (C), Samsung Bioepis (C), Topcon Medical Systems Inc. (C, R, F), Nidek Incorporated (R, F)