Primary retinal tauopathy: A tauopathy with a distinct molecular pattern

Abstract

BACKGROUND

Phosphorylated tau (p‐tau) accumulation, a hallmark of Alzheimer's disease (AD), can also be found in the retina. However, it is uncertain whether it is linked to AD or another tauopathy.

METHODS

Retinas from 164 individuals, with and without AD, were analyzed for p‐tau accumulation and its relationship with age, dementia, and vision impairment.

RESULTS

Retinal p‐tau pathology showed a consistent pattern with four stages and a molecular composition distinct from that of cerebral tauopathies. The stage of retinal p‐tau pathology correlated with age (r = 0.176, P = 0.024) and was associated with AD (odds ratio [OR] 3.193; P = 0.001), and inflammation (OR = 2.605; P = 0.001). Vision impairment was associated with underlying eye diseases (β = 0.292; P = 0.001) and the stage of retinal p‐tau pathology (β = 0.192; P = 0.030) in a linear regression model.

CONCLUSIONS

The results show the presence of a primary retinal tauopathy that is distinct from cerebral tauopathies.

1. BACKGROUND

The deposition of amyloid β (Aβ)‐containing plaques and the formation of neurofibrillary tangle (NFT) pathology in the brain are neuropathological hallmark lesions of Alzheimer's disease (AD), ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ which have also been reported to occur in the retina. ¹ , ⁸ , ⁹ NFTs consist of aggregates of abnormally phosphorylated tau protein (p‐tau). ¹⁰ Pathologic accumulation of tau also exists in non‐AD tauopathies with different molecular patterns of the tau aggregates, which are well illustrated by the relationship of the 3‐repeat (3R) and 4‐repeat (4R) isoforms of tau. ¹¹ In AD and primary age‐related tauopathy (PART), both 3R and 4R tau coexist, whereas in Pick's disease (PiD), 3R tau predominates in contrast to most other cerebral tauopathies with predominant 4R tau. ¹¹ , ¹² , ¹³ , ¹⁴ Different patterns of retinal tau‐related changes in AD, progressive supranuclear palsy, PiD, and corticobasal degeneration have been reported recently. ¹⁵ However, a detailed anatomical and molecular analysis of the retinal tau‐related changes in demented and non‐demented individuals with different ophthalmopathological conditions is missing.

To address this question, we analyzed the characteristics of p‐tau pathology in a large cohort of retina samples from AD donors and healthy controls, as well as in retinal biopsy tissue from different ophthalmological diseases.

2. METHODS

2.1. Study cohort

Tissue was obtained from UZ Leuven Biobank and the Netherlands Brain Bank (NBB: https://www.brainbank.nl). All donors gave informed consent. This project was performed following Belgian law and was approved by the local ethical committee in Leuven/Belgium (S64492). The Amsterdam University Medical Center medical ethics committee approved the donor program of the NBB (2009/148).

In total, 164 retina samples were examined (Table S1 in supporting information). Thirty‐five of them were collected post mortem. In these cases, the result of the neuropathological examination of the brain was available covering Aβ phases, Braak NFT stages, Braak Lewy body disease (LBD) stages, and CERAD (Consortium to Establish a Registry for Alzheimer's Disease) neuritic plaque scores ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ (Table S2 in supporting information). One hundred twenty‐nine retina samples were derived from biopsies of patients with severe ophthalmopathological conditions, including uveal melanoma, painful inflammatory diseases of the eye, or retinal degeneration. A total of 21 samples were from donors with a clinical diagnosis of dementia (17 with neuropathologically confirmed AD, age mean ± standard deviation [SD]: 77.38 ± 9.09), and 143 from clinically non‐demented individuals (age mean ± SD: 69.01 ± 14.20). Based on the clinical diagnosis, we distinguished among five groups: (1) ocular tumors, (2) ocular inflammation, (3) retinal degeneration, (4) AD, and (5) controls.

Fresh‐frozen samples from 12 retinas and 14 brains were used to perform Western blot analysis (Table S3 in supporting information).

RESEARCH IN CONTEXT

1. Systematic review: A literature review using online databases was conducted. Amyloid β and phosphorylated tau (p‐tau) pathology were reported in human retinas. However, molecular characterization of retinal p‐tau pathology and its comparison to Alzheimer's disease (AD) pathology has not yet been determined in a large cohort of cases, including non‐demented individuals. Relevant literature was appropriately cited.

2. Interpretation: Our study suggests the existence of primary retinal tauopathy (PReT) with a molecular p‐tau pattern, distinct from known cerebral tauopathies. Given its association with AD, PReT may represent a prerequisite for retina involvement in AD.

3. Future directions: Usage of ocular imaging, specifically the detection of retinal p‐tau, has potential to aid in the diagnosis of AD. Although PReT is molecularly different from AD, its correlation with AD could be used as a supplementary biomarker for monitoring disease progression. Future research is needed to determine the impact of retinal imaging as an AD biomarker.

2.2. Pathological analysis

Dissected eyes from deceased individuals and biopsy samples after enucleation or orbit exenteration were fixed in 4% paraformaldehyde solution (PFA) for ≈ 24 hours. Then, eyeballs were lamellated sagittally into 4 mm thick slabs and embedded in paraffin. Evisceration samples were completely embedded in paraffin after ≈ 24 hours of PFA fixation. Sections were cut at 5 μm thickness.

Dissected eyes from autopsy cases of NBB were frozen using iso‐pentane at −100°C and stored at −80°C. For histological analysis, eyes were defrosted at room temperature in 4% PFA for 48 hours before dissection through the horizontal and vertical axis into nasal‐superior, nasal‐inferior, temporal‐superior, and temporal‐inferior quadrants. The quadrants were PFA‐fixed for 3 hours and then embedded in paraffin. Retinal tissue blocks were microtomed at 10 μm. Paraffin sections were deparaffinized and heat‐pretreated using Envision^TM Flex Target Retrieval Solution Low pH (Dako). Treatment with 98% formic acid for 3 minutes at room temperature was performed before anti‐Aβ immunohistochemistry. After peroxidase blocking (Dako REAL^TM), primary antibodies (Table S4 in supporting information) were applied and incubated overnight at room temperature. Horseradish peroxidase‐conjugated secondary anti‐mouse or anti‐rabbit antibodies were used to detect the respective primary antibodies; 3,3′‐diaminobenzidine (DAB) was used as chromogen. Sections were counterstained with hematoxylin and mounted using a Leica Autostainer. To visualize fibrillar tau pathology, that is, NFTs, sections were stained with the Gallyas silver‐impregnation method. ²⁰ , ²¹ To determine fibrillar amyloid Congo red and thioflavin S stains were used.

2.3. Biochemical analysis

To confirm and extend the histopathological findings, we examined 14 brain samples (4 AD, 5 PART, and 5 controls cases) containing frozen temporal neocortex (Brodmann area 20) as well as 12 retina samples (8 AD and 4 non‐AD cases; Table S3). Due to the limited availability of brain and retina samples, it was not feasible to analyze retina and brain samples from the same patients.

Briefly, frozen brain/retina tissue samples (50 mg) were homogenized in 500 μl of filtered phosphate‐buffered saline (PBS) containing protease and phosphatase inhibitor‐cocktail (Complete and PhosSTOP, Roche) using a micropestle and subsequent sonication. The homogenate was centrifuged for 5 minutes at 3000 × g and 4°C. The resulting supernatant was retained for further ultracentrifugation. Ultracentrifugation of the supernatant was performed at 121,656 × g for 30 minutes. The pellet containing insoluble proteins was resuspended in PBS followed by sonication. ²² The total protein content of the pellet was determined using the BCA Protein Assay (Bio‐Rad). Finally, proteins were eluted with 1x LDS sample buffer (Life Technologies) and heated to 95°C.

Using sodium dodecyl sulfate polyacrylamide gel electrophoresis, the insoluble proteins were resolved on a precast NuPAGE 4‐12% Bis–Tris gel system (Life Technologies). Proteins were subsequently transferred onto nitrocellulose membranes. The membranes were blocked in 5% bovine serum albumin diluted in Tris‐buffered saline with Tween (TBS‐T) for 1 hour. For the detection of different p‐tau species via western blotting, specific antibodies targeting various tau species were incubated overnight at 4°C (Table S4). The blots were then incubated with a chemiluminescent ECL detection system (SuperSignal West Dura system, Thermo Fisher Scientific) and imaged using a CCD imager Image Quant LAS 4000 (GE HealthCare). We conducted both short and longer exposure times to visualize lower protein levels in the retina samples. After detection, the membranes were washed in TBS‐T for 5 minutes and incubated for 30 minutes in Restore Western Blot Stripping Buffer (Life Technologies), followed by two 5‐minute washes in TBS‐T. Subsequently, the membranes were blocked and incubated with another primary antibody as described above. To avoid inconclusive results due to the use of the stripping method, we always repeated the detection step after stripping the antibodies to document successful stripping.

2.4. Assessment of retinal tau pathology

The presence of p‐tau pathology and its anatomical distribution within the retinal layers was assessed dichotomously and documented separately for each layer based on anti–p‐tau^S202/T205‐stained sections. This antibody distinguishes tauopathic lesions in the retina from physiological retinal tau. ¹⁵ Both peripheral and central parts of the retina were analyzed. The peripheral area was determined by observing the retina near the ora serrata, while the central area was recognized as the portion close to the optic nerve. Accordingly, the peripheral retina represents the peripheral 50% of the retina lengths measured from the optic nerve to the ora serrata whereas the central part represents the proximal 50% of the respective length. For this analysis, we used 70 enucleation cases to ascertain the correct distinction between the peripheral and central parts of the retina. The most advanced p‐tau distribution was recorded for each cease. Microscopic images for documentation and quantitative assessments were taken with a Leica DCC 290 microscope coupled with a Leica DFC7000 T camera. Likewise, the presence/absence of other tau epitopes was determined with the antibodies listed in Table S4 in representative cases.

2.5. Determination of the visual performance

A retrospective analysis of clinical records was performed to evaluate the pre‐operative visual performance. Clinical records were available from 125 biopsied patients. The reports covered the ophthalmological examination of the resected/eviscerated eyes prior to the surgery, including visual acuity assessments according to a standardized protocol. ²³ Eyes with a visual acuity score between 8/10 and 10/10 were considered normal, eyes with a visual acuity score between 1/10 and 7/10 as visually impaired, and eyes with a score lower than 1/10 as “legally” blind. ²³ By applying these thresholds, we semi‐quantitatively determined vision impairment scores distinguishing among cases with normal (0), impaired vision (1), and blindness (2). Data on pre‐operative eye pressure and the presence/absence of retinal detachment were available for 125 biopsy cases.

2.6. Statistical analysis

Statistical analysis was performed using the IBM‐SPSS Statistics 25 software package. Spearman correlation was used to analyze the association between age (at biopsy/autopsy), clinical features, and retinal tauopathy. Semi‐partial correlation controlled for age and sex was used to investigate the association between retinal tauopathy and 4R‐tau presence. Binary and/or multinominal logistic regression models included clinical parameters (clinical diagnosis, glaucoma, dementia, diabetes, etc.), age, and sex as independent variables used to identify the main clinical correlative(s) of p‐tau pathology in the retina. Linear regression analysis was used to study the relationship between retinal tauopathy and visual acuity, internal eye pressure, or retinal detachment, controlled for the clinical diagnosis, age, and sex. To determine the association between retinal tauopathy and brain pathology conditions, we used semi‐partial correlation controlling for age and sex. The detailed characteristics of all statistical tests used are listed in Table S5 in supporting information.

3. RESULTS

3.1. Retinal tauopathy shows a hierarchical spreading pattern in the layers of the retina

Out of 164 analyzed retinas, 41 showed no p‐tau^S202/T205 (Figure 1A). In 123 cases, p‐tau^S202/T205 lesions were present in the outer plexiform layer (OPL; Figure 1B). In 85 of these 123 retinas, neurons of the inner nuclear layer (INL) exhibited cytoplasmic p‐tau^S202/T205 as well (Figure 1C). Out of these 85 cases, 61 showed additional p‐tau^S202/T205 immunopositivity in the inner plexiform layer (IPL; Figure 1D). Accordingly, four stages of retinal p‐tau distribution can be distinguished: stage 0, no p‐tau^S202/T205 pathology; stage 1, p‐tau^S202/T205 in OPL; stage 2, p‐tau^S202/T205 in OPL and INL; stage 3, p‐tau^S202/T205 in OPL, INL, and IPL (Figure 1E). The stages were always determined by estimating the most advanced site exhibiting p‐tau pathology in samples covering the peripheral and central parts of the retina. The more advanced stages of the p‐tau pathology were usually observed in the periphery of the retina when comparing central and peripheral parts (Figure S1 in supporting information; Wilcoxon signed‐rank test: P = 0.0001).

FIGURE 1.

The sequential pattern of primary retinal tauopathy (PReT). A, No p‐tau pathology in the retina (PReT stage 0). B, Initial accumulation of p‐tau^S202/pT205 in the OPL (PReT stage 1). C, Additional presence in neurons of the INL (PReT stage 2). D, Further accumulation of p‐tau in OPL and INL, as well as in the IPL (PReT stage 3). E, Heatmap of the prevalence of p‐tau^S202/T205 accumulation in the retinal layers in percent of all cases investigated at a given PReT stage. Calibration bars in (A) are valid for (A), (B), (C), (D). GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; ONL, outer nuclear layer; OPL, outer plexiform layer, PReT, primary retinal tauopathy; PRL, photoreceptor layer; p‐tau, phosphorylated tau

3.2. Molecular pattern of retinal p‐tau in relation to the stage of retinal p‐tau distribution

To clarify whether the molecular pattern of tau pathology in the retina fits with that reported by the AD brain, ¹³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ five additional antibodies were used for histopathological analysis: anti‐p‐tau^T231, anti‐MC‐1 indicative for the paperclip folding of tau, anti‐p‐tau^S396/S404, anti‐3R, and anti‐4R tau (Table S4). p‐tau^T231 showed a similar distribution pattern among the retinal layers as p‐tau^S202/T205, but p‐tau^T231 accumulated already in 50% of the retinal p‐tau stage 0 cases as determined with anti‐p‐tau^S202/T205 (Figure 2). In stage 1, p‐tau^T231 was in 8% of the stage 1 cases not only visible in OPL, but also in INL and IPL (Figure S2A‐D in supporting information). Similar to anti‐p‐tau^T231, the MC‐1 antibody showed mild p‐tau pathology in OPL, INL, and IPL in 23% of the stage 1 cases (Figure S2E‐H) and in OPL and INL in 31% of the stage 0 cases (Figure 2). 3R tau (Figure S2M‐P) and p‐tau^S396/S404/p‐tau^S396 (Figure S2Q‐T) were visible in OPL, INL, and IPL, as well as in the ganglion cell layer (GCL; Figure S2R) and the retinal nerve fiber layer of all investigated retinas regardless of stage, age, or diagnosis. Moreover, we noted that some cones in the photoreceptor layer were p‐tau^S396/S404 positive (Figure S2S).

FIGURE 2.

Heatmap characterizing p‐tau species in the retina. p‐tau^T231 and anti‐MC‐1 tau aggregates were already observed in the retinal layers in PReT stage 0 cases pointing to the presence of precursor lesions of PReT. The percentage of 4R tau–positive retinas differed from the presented retinal phosphorylation of p‐tau^S202/T205. 4R tau occurred in the presence and absence of PReT, but its overall prevalence correlated with increasing PReT stage. All investigated cases showed positive p‐tau^S396/S404 and 3R‐tau patterns in the retina. Gallyas staining did not detect neurofibrillary tangles. 3R, 3‐repeat; 4R, 4‐repeat; PReT, primary retinal tauopathy; p‐tau, phosphorylated tau

4R tau was immunohistologically detected in the retina in 15% to 39% of the analyzed cases (Figure 2). Here, the expression of this tau isoform was visible mainly in OPL and INL (Figure S2I‐L). We observed an association between the stage of retinal p‐tau pathology and the occurrence of 4R‐tau in the retina (r = 0.211, P = 0.007, N = 164). In all examined cases, Gallyas silver staining did not detect NFTs in the retina samples (n = 164; Figure 2, Figures S2, S3 in supporting information).

Western blot analysis comparing tau in retinas to temporal cortex samples from AD, PART, and control cases confirmed significant differences in the respective molecular tau patterns. First, the amount of tau in a given volume of temporal neocortex tissue was higher than that in the same volume of retina tissue. Second, p‐tau^T231 was observed in the brains of control, PART, and AD cases as a 28 kDa fragment of 3R tau, whereas only traces of p‐tau^T231 were visible in retinas. Furthermore, a smear of tau species/oligomers with a molecular weight > 55 kDa was observed with anti‐3R and anti‐4R tau, anti‐p‐tau^T231, and anti‐p‐tau^S396/S404 in brain tissue from AD and PART cases, while 3R and 4R tau species with a molecular weight above 55 kDa were absent in all retina samples (Figure 3; Figure S4A,B in supporting information). An interesting finding with relevance for physiological phosphorylation processes is the presence of p‐tau^S396/S404 even in control retinas and brain samples, presumably indicating a physiological role of tau phosphorylation at these residues.

FIGURE 3.

Characteristic biochemical pattern of tau species in human retina and brain. The amount of tau in temporal neocortex tissue (Brodmann area 20) was higher compared to retina tissue. Oligomeric and/or phosphorylated tau larger than ≈55 kDa was found in brain samples from AD, PART, and control groups but was not evident in the retina. The p‐tau^S202/T205 antibody was present in both retinas and brains, exhibiting a similar biochemical pattern. A ≈28 kDa p‐tau^T231 fragment of 3R tau in controls and longer ptau^T231 fragments in PART and AD were observed in brain samples but were nearly negligible in the retina. The specific bands corresponding to the 3R and 4R tau isoforms were identified in the retina. The concentration of these tau isoforms suggests their presence as aggregates and/or phosphorylated proteins larger than ≈55 kDa, with the highest concentration observed in AD cases. Notably, p‐tau^S396/S404 was present in retinas and brain samples from control subjects, suggesting its importance in physiological phosphorylation processes. (Eye: AD case nr. 11; PART case nr. 32; Control case nr. 36; Brain: AD case nr. 51; PART case nr. 43; Control case nr. 48). 3R, 3‐repeat; 4R, 4‐repeat; AD, Alzheimer's disease; PART, primary age‐related tauopathy; PReT, primary retinal tauopathy; p‐tau, phosphorylated tau

3.3. Correlation of stage and molecular pattern of retinal p‐tau with the clinical diagnosis and brain pathology

By comparing cases with clinical dementia (n = 21) and non‐demented cases (n = 142), we found all stages of retinal p‐tau pathology in both groups. Binary logistic regression comparing cases with clinical dementia and non‐demented cases revealed an association of the presence of dementia with increasing stage of retinal p‐tau pathology (odds ratio [OR] = 1.685 [95% confidence interval (CI) 1.066–2.664]; P = 0.026; Figure S5 in supporting information). No differences in the hierarchical distribution pattern of retinal p‐tau pathology were observed between the two groups.

Comparing the diagnosis groups (1) ocular tumors, (2) ocular inflammation, (3) retinal degeneration, (4) AD, and (5) controls, higher stages of retinal p‐tau pathology than in controls were found in cases with ocular inflammation (OR = 2.605 [95% CI 1.493–4.545]; P = 0.001) and AD (OR = 3.193 [95% CI 1.597–6.385]; P = 0.001; Figure 4). Two of the 17 AD cases did not show p‐tau in the retina (12%).

FIGURE 4.

Prevalence of PReT stages among age groups (A), clinical diagnosis (B), and groups indicating the vision impairment score (C). A, The stage of PReT significantly increases with age (Spearman correlation r = 0.176, P = 0.024, n = 164 [≥60 (n = 41); 61–70 (n = 33); 71–80 (n = 40); > 80 (n = 50)]). B, Eyes from individuals without eye pathology showed lower PReT stages compared to patients diagnosed with Alzheimer's disease, or inflammatory conditions, which exhibited advanced PReT in most cases (multinominal logistic regression controlling for age and sex [control cases as a reference group]; inflammation: OR = 2.605 [95% CI 1.493–4.545], P = 0.001; AD: OR = 3.193 [95% CI 1.597–6385], P = 0.001; control n = 18, degeneration = 12, tumor n = 76, inflammation n = 40, Alzheimer's disease n = 18, in total N = 164). C, The linear regression model controlled for the respective diagnosis, age, and sex showed an association between PReT stage and the vision impairment score (β = 0.192 [95% CI 0.012–0.224; n = 125, P = 0.030; normal [n = 16], impaired vision [n = 32], blindness [n = 77]). CI, confidence interval; OR, odds ratio; PReT, primary retinal tauopathy; p‐tau, phosphorylated tau

The distribution of the stages of retinal p‐tau pathology (Stages 0–3) was compared among four age ranges: younger than 60, 60 to 70, 71 to 80, and older than 80 years (Figure 4). The stage of retinal p‐tau pathology mildly correlated with increasing age (r = 0.176, n = 164; P = 0.024).

Correlation analysis of primary retinal tauopathy (PReT) stages with cerebral Aβ phase, Braak NFT stage, Braak LBD stage, and CERAD scores for neuritic plaque pathology revealed only a correlation of retinal PReT stages with Aβ phase (Table S5).

3.4. Association between retinal p‐tau pathology and visual performance

The semi‐quantitative vision impairment scores ranging from normal vision (0) to blindness (2) were obtained in 125 cases. In a linear regression model controlled for the ophthalmological diagnosis group, age, and sex, a mild independent association between the stage of retinal p‐tau pathology and increasing vision impairment scores was seen (β = 0.192 [95% CI 0.012–0.224]; P = 0.030; Figure 4), whereas the ophthalmological diagnosis showed a stronger independent association with the vision impairment score (β = 0.292 [95% CI 0.125–0.480]; P = 0.001).

The stage of retinal p‐tau pathology was not associated with glaucoma, diabetes, eye pressure, and the presence/absence of retinal detachment (Table S5).

3.5. Prevalence of retinal Aβ immunoreactivity

By analyzing Aβ pathology in all 164 examined retinas histopathologically, we found small spots of Aβ immunoreactive material of 3 to 10 μm diameter (Figure S6A‐C in supporting information) in 3 out of 17 pathologically confirmed AD cases (17%) and 1 out of 146 non‐demented cases (1%). All three AD cases with Aβ‐positive material had stage 3 of retinal p‐tau pathology. The Aβ‐positive control case did not exhibit retinal p‐tau pathology. The Congo red and Thioflavin S stains did not detect fibrillar amyloid material in the retina of the cases exhibiting retinal Aβ positivity (Figure S7 in supporting information).

4. DISCUSSION

Here, we show that p‐tau in the retina of non‐demented individuals exhibits a characteristic molecular pattern that is different from other tauopathies, including AD (Figure 5). Accordingly, we can argue that p‐tau in the retina in this cohort represents a primary retinal tauopathy (PReT) most frequently found in eyes with pathological alterations of different types.

FIGURE 5.

Schematic representation of the pathological features of PReT compared to other tauopathies affecting the brain. Given that PReT showed no glial p‐tau, overlap with the age‐related tau astrogliopathy (ARTAG) can also be excluded. ¹⁴ Neuronal tau pathology of any kind is marked in orange, astroglial tau pathology of any type in green, and oligodendroglial tau inclusions are depicted in blue. Another distinction criterium is the molecular pattern in the western blot. PReT tau has a molecular weight of ≤ 55 kDa whereas all other tauopathies also exhibit tau species with a molecular weight > 55 kDa. AD, Alzheimer's disease;⁴³ AGD, argyrophilic grain disease;³⁶ CBD, corticobasal degeneration; CTE, chronic traumatic encephalopathy;^45–47 NFTs, neurofibrillary tangles; PART, primary age‐related tauopathy; ⁴⁴ Pick's disease; ⁴⁰ , ⁴⁸ PReT, primary retinal tauopathy; PSP, progressive supranuclear palsy; ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴²

The hierarchical distribution pattern of PReT, as assessed with anti‐p‐tau^S202/T205 throughout the retinal layers, showed four stages (0–3) that slightly correlated with the presence of ocular inflammation, the neuropathological diagnosis of AD, and age. In contrast to other tauopathies, 4R tau was only occasionally detectable in PReT. In none of the cases studied here, fibril formation indicated by argyrophilic material in the Gallyas staining was observed. Furthermore, our western blot results show significantly lower amounts of p‐tau in a given volume of retina tissue compared to neocortex samples from AD cases and the absence of tau species/oligomers with a molecular weight > 55 kDa in the retina may explain the lack of fibrillar p‐tau aggregates observed in our retina samples. Moreover, the biochemical pattern points to significant differences between the retinal p‐tau pathology and AD/PART brain lesions. In contrast to the brain, we observed a strong immunohistochemical expression of 3R tau and p‐tau^S396/S404/p‐tau^S396 in all retinal layers suggesting that some posttranslational tau modifications are endogenously present in the retina. In the brain, the pronounced accumulation of p‐tau^S396/S404 in axons and the neuropil is considered an initial step in the manifestation of tau pathology as “initial neuropil tau” (IN‐tau). ²⁴ The physiological expression of p‐tau^S396/S404 in non‐affected areas of the brain is low. Another difference to cerebral AD tau pathology is that the paperclip formation of p‐tau was found in the retina as early as tau pathology started. In contrast, it follows the initial pathological phosphorylation of tau in AD pathogenesis. ²⁴ These histopathological differences were supplemented by differences in the tau‐band pattern observed in western blots. Accordingly, the molecular composition of PReT differed from that found in brains from symptomatic and preclinical stages of AD and other tauopathies in the brain. Except for PiD, most other tauopathies are 4R or 3R/4R tauopathies and are, therefore, expected to induce also 4R tau accumulation in the retina. Den Haan et al. reported a different pattern of immunoreactivity for p‐tau lesions in six AD and six control retinas. ⁸ However, they did not find MC‐1 immunoreactive tau. Given the small number of cases in this study, ⁸ cases with retinal presence of MC‐1 tau may not have been included by chance. An alternative explanation could be the different fixation method compared to our biopsy cases and better preservation of the conformational MC‐1 epitope in our cases than in the samples of den Haan et al. ⁸ that underwent freeze–thaw cycles. Taken together, the molecular pattern of tau aggregates in PReT differs from AD brain and other known tauopathies, arguing in favor of a novel tauopathy that develops in the retina, that is, PReT.

The stages of PReT were associated with inflammatory conditions, AD, and dementia. Given the association with AD, one could argue that PReT is part of AD. An argument for this hypothesis is the fact that the molecular pattern of AD covers all p‐tau species found in PReT and that the Aβ phases in the brain correlate with the PReT stages. However, 2 out of 17 AD cases did not show PReT, and a correlation between PReT stages and the Braak NFT stages could not be seen in our post mortem cohort. As explained above, the different molecular patterns, including the differences in the tau‐band pattern observed in western blots, provide another argument against the hypothesis that PReT is solely the retinal manifestation of AD. Instead, it suggests that PReT is a retina‐specific tauopathy with a distinct molecular pattern of tau aggregates. Thus, PReT may more likely represent a prerequisite for developing AD‐related tauopathic lesions in the retina. Whether the immunohistochemical detection of 4R tau indicates the progression toward AD cannot be excluded.

An important finding is the possible association of PReT with an impairment of vision observed in individuals who underwent eyeball resections or eviscerations. Controlling for clinical diagnosis, we found a mild association between PReT stage and vision impairment. This association would have a clinical impact by adding another potential disease that may contribute to visual impairment: PReT. However, no statement can be made about a causative relation. Our finding requires future confirmation by independent groups as well as experimental validation.

The observed PReT pattern fits with earlier reports about retinal tau pathology in glaucoma, AD, and ophthalmologically healthy individuals. ¹ , ⁸ , ¹⁵ , ²⁹ Another argument in favor of PReT being a distinct tauopathy rather than an early stage of retinal AD pathology is the fact that Aβ‐positive material was seen only in 3 out of 123 cases with retinal tauopathy, from which all 3 had symptomatic AD. However, the fourth case with anti‐Aβ‐positive material exhibited no p‐tau pathology. Most AD cases in our sample did not exhibit Aβ‐positive material in the retina. Accordingly, PReT most frequently develops in the absence of retinal Aβ‐positive material. Given its size varying between 3 and 10 μm in diameter in our cases, Aβ‐positive material in the retina is very small compared to amyloid plaques in the brain, usually measuring 50 to 100 μm in diameter, ³⁰ and their relevance remains unclear. Given the small size of the retinal Aβ‐positive material, one cannot exclude that these immunostained structures could represent unspecifically labeled material. However, an increased frequency of Aβ in AD retinas, as reported by other groups, ³¹ , ³² , ³³ , ³⁴ may argue in favor of an AD‐specific feature.

Given the recent advancement in retinal imaging techniques for detecting neurodegenerative lesions in the retina, ⁴ it is important to take into account PReT as a distinct retinal tauopathy that may not necessarily indicate AD. However, PReT, particularly in advanced stages, is correlated with AD. Hence, monitoring the progression of retinal tau pathology in AD patients, who have been diagnosed with other biomarkers, could provide insights into the progression of AD or the effectiveness of drugs targeting tau. Further clinical and experimental studies are needed to assess the potential of retinal tau imaging for monitoring AD and its therapy.

Some limitations apply to this study: (1) Because the biopsies were retrieved from living donors, these cases lack neuropathological characterization. To overcome this limitation, we included 32 brain autopsy cases from the NBB with available retina sections and a post mortem confirmation of the AD diagnosis. (2) By using samples from two different cohorts, the fixation conditions varied. To avoid sensitivity issues for the detection of p‐tau and Aβ, we used those anti‐p‐tau^S202/S205 and anti‐Aβ_17‐24 antibodies as benchmark antibodies which were previously shown to deliver robust and comparable staining results regardless of tissue fixation and embedding conditions. ³⁵ Accordingly, the staging of PReT and the detection of Aβ are comparable among all studied retinas from both cohorts. (3) Retinal imaging data and AD biomarker data were not available. Accordingly, conclusions about the impact and correlation of PReT with such biomarkers could not be analyzed. (4) Functional confirmation in a retinal tauopathy model system has not been carried out in this initial description of the distinct molecular pattern of the primary retinal tauopathy PReT compared to AD. Functional studies on tau seeding in the retina and its interplay with the brain are needed to better understand the role of PReT in neurodegenerative disorders of the brain, such as AD.

5. CONCLUSION

Our study provides evidence for the existence of a distinct primary retinal tauopathy (PReT) in the human retina. To what extent PReT can act as a prerequisite for developing AD‐associated lesions in the retina or contributes to vision impairment requires further research.

CONFLICT OF INTEREST STATEMENT

J.J.M.H. received grants from the Dutch Research Council (ZonMW) and Alzheimer Netherlands; and performed contract research or received grants from Merck, ONO Pharmaceuticals, Janssen Prevention Center, DiscovericBio, AxonNeurosciences, Roche, Genentech, Promis, Denali, FirstBiotherapeutics, and Ensol Biosciences. I.S. received consultancies and research grants from Alcon, Allergan/AbbVie, EyeD, Horus, Mona, Santen, and Laboratoires Théa. D.R.T. received speaker honorary from Biogen (USA); travel reimbursement from UCB (Belgium); and collaborated with Novartis Pharma AG (Switzerland), Probiodrug (Germany), GE HealthCare (UK), and Janssen Pharmaceutical Companies (Belgium). All payments were made to the respective institution. The other authors have nothing to declare. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human donors in the study were fully informed and consented to participate in accordance with ethical guidelines and the Belgian and Dutch law. This study was approved by the UZ Leuven ethical committee (S64492). The Amsterdam UMC medical ethics committee approved the donor program of the NBB (2009/148).

Supporting information

Supporting information

Supporting information

Walkiewicz G, Ronisz A, Van Ginderdeuren R, et al. Primary retinal tauopathy: A tauopathy with a distinct molecular pattern. Alzheimer's Dement. 2024;20:330–340. 10.1002/alz.13424