Vitreous STMN2 levels reflect TDP-43-associated neurodegeneration in postmortem eyes and brains

Abstract

Stathmin-2 (STMN2) levels decline in brains with transactive response DNA binding protein-43 (TDP-43) inclusions. TDP-43-related changes could extend to ocular structures, although vitreous STMN2 levels remain uncharacterized. This exploratory study analyzed 72 post-mortem brains and eyes depending on the presence or absence of TDP-43 inclusions in the brain and across neuropathological diagnostic groups (Alzheimer's disease [AD], chronic traumatic encephalopathy [CTE], AD and CTE, or neither). Results showed decreased vitreous STMN2 levels in TDP-43-positive cases but no association with diagnostic groups. Vitreous STMN2 was correlated with vitreous neurofilament light chain. Diminished vitreous STMN2 levels might indicate TDP-43-associated neurodegeneration.

Introduction

Transactive response DNA binding protein (TDP-43) is a nuclear regulator of RNA transcription, splicing, and transport.^1–3 Aberrant post-translational modifications of TDP-43—hyperphosphorylation, ubiquitination, or cleavage—can disrupt these crucial functions, leading to mislocalization, cytoplasmic accumulation, and aggregation into pathological inclusions.^4–6 These inclusions are most characteristic of amyotrophic lateral sclerosis (ALS)^4,6 and a subset of frontotemporal lobar degeneration (FTLD)^4,6,7 as well as limbic predominant age-related TDP-43 encephalopathy (LATE), a condition associated with advanced age. ⁸ That said, TDP-43 inclusions have also been reported in Alzheimer's disease (AD)^9,10 and chronic traumatic encephalopathy (CTE),^11–13 suggesting a broader neurodegenerative role for this protein. Loss of TDP-43 function in all these proteinopathies results in nonsense-mediated decay or premature polyadenylation of other essential transcripts such as stathmin-2 (STMN2),^14–18 an axonal maintenance factor, and unc-13 homologue A (UNC13A),^19,20 a regulator of synaptic vesicle release. Diminished levels of STMN2, in particular, impair the regenerative capacity of neurons, ¹⁴ ultimately triggering axonal degeneration and gradual decline in function.

As a result of the intricate embryological and vascular relationship between the brain and eyes,^21,22 retinal TDP-43 inclusions have also been detected in ALS, ²³ FTLD,^24–26 AD, ²⁵ and CTE. ²⁷ Other neuropathological hallmarks such as amyloid-β (Aβ) and phosphorylated tau (p-tau) found in the retina^24–27 have also been measured across the retinal-vitreous interface in ocular fluids.^28–30 Such extensions of neurodegeneration into the eye could contribute to visual symptoms in ALS, ³¹ AD,^32–34 and CTE, ¹³ though causation remains unclear. In terms of cognitive dysfunction, levels of Aβ₄₀, Aβ₄₂, and total tau (tTau) in the vitreous humor were found to correlate with scores on the Mini-Mental State Exam. ³⁵ These findings underscore the potential of the vitreous humor in particular, given its proximity to retinal tissue, to detect early neurodegenerative changes^36,37; the protein composition of the vitreous humor is distinct from cerebrospinal fluid (CSF) or plasma^29,30,35 and could provide greater insights. Moreover, emerging techniques for sampling vitreous fluid could be less invasive than lumbar puncture, while offering comparable biomarker performance.^38,39

Given that STMN2 in neural tissue can serve as a sensitive indicator of TDP-43-related neurodegeneration,^14–16 it could function similarly as an ocular fluid biomarker. The purpose of the present study was to better characterize the relationship between STMN2 levels in the vitreous humor and: (a) TDP-43 inclusions, neurofibrillary tangles, or neuritic plaques in corresponding brains, (b) neuropathological diagnostic groups (AD, CTE, both, or neither), (c) repetitive head impact exposure (RHI), and (d) other cytoskeletal biomarkers in the cortex, vitreous humor or CSF.

Methods

Brain and eye donors

All donors were part of the Boston University Alzheimer's Disease Research Center (ADRC) and United Neurological Injury and Traumatic Encephalopathy (UNITE) brain banks; inclusion criteria have been described elsewhere. ⁴⁰ All brain donors who also donated eyes were included in the study. Donations with post-mortem intervals greater than 72 h were excluded. Consent for use of tissues was obtained from the legal healthcare proxy. All procedures were approved by the Boston University Medical Center Institutional Review Board (IRB H-37370). Post-mortem eyes and brains underwent comprehensive neuropathological examination. For each case, vitreous humor was collected during autopsy, with eye laterality chosen at random for analysis. Samples were immediately placed at −80°C and stored until assay to prevent protein degradation. The mean post-mortem interval was 36.4 h.

Neuropathological analysis

Neuropathological processing involved a comprehensive evaluation for neurodegenerative conditions, following standardized procedures established by the UNITE Brain Bank. ⁴⁰ All cases were assessed using paraffin-embedded tissue sections obtained from predefined brain regions, which were subjected to histochemical and immunohistochemical staining protocols. Neuropathological analysis focused on the extent and distribution of beta amyloid, neurofibrillary tangles, alpha-synuclein inclusions, and phosphorylated TDP-43 inclusions. Immunohistochemistry for p-TDP-43 (Cosmo Bio, Tokyo, Japan; clone 11-9) was performed on sections including the amygdala with the entorhinal cortex (Brodmann area [BA] 28), hippocampus at the level of the lateral geniculate nucleus (LGN), and dorsolateral frontal (BA 8,9) at a titration of 1:3000 and processed on a Leica Bond RX automated stainer (Leica Biosystems, Wetzlar, Germany) according to the manufacturer's instructions. Positive controls were verified with each batch of stains. Cases were considered TDP-43-positive if inclusions were detected in any of these regions.

Cases meeting criteria for intermediate or high AD neuropathological change, as defined by the National Institute on Aging and Reagan Institute (NIA-RI), ⁴¹ were classified as AD. AD neuropathologic change was assessed using the NIA-RI criteria, rather than the more recent NIA-AA guidelines, because most cases in this cohort had previously been evaluated with NIA-RI. Prior studies also suggest that the CERAD neuritic plaque score and Braak neurofibrillary tangle stage components of NIA-RI criteria provide strong predictive value for cognitive impairment in AD. ³¹ CTE was diagnosed using the NIH/NINDS consensus criteria,^42,43 which require abnormal perivascular accumulations of hyperphosphorylated tau in neurons, with or without astrocytes, predominantly at sulcal depths. CTE severity was staged based on the regional distribution of p-tau pathology. ⁴⁴ Control cases were defined as those with no significant neurodegenerative pathology, i.e., no evidence of AD, CTE, or other neurodegenerative lesions except for age-related changes, such as Braak stage I, upon neuropathological examination.

Immunoassay measurement for STMN2, amyloid, tau, and NfL proteins

Protein biomarker levels were quantified in the vitreous humor from post-mortem eyes and cortical brain tissue using Meso Scale Discovery (MSD, Rockville, MD, USA) quantitative immunoassays. The biomarkers analyzed included STMN2, tTau, and neurofilament light chain (NfL). tTau and NfL were selected based on prior reports by our group demonstrating their presence in antemortem and post-mortem vitreous humor samples.^29,30,35 All assays were performed with the investigators blinded to the TDP-43 status of each case.

Vitreous humor samples were centrifuged at 12,000 rpm for 15 min to remove cellular debris, aliquoted into 100 µL portions, and stored at −80°C. For immunoassay preparation, vitreous fluid was diluted 1:1 with 1% Blocker A (MSD #R3BA 4) in wash buffer to achieve a total volume of 200 µL. The diluted samples were further centrifuged at 17,000×g for 15 min at 4°C, and the supernatant was used for the assay.

Brain tissue was processed by dilution in ice-cold 5 M guanidine hydrochloride/50 mM Tris-HCl buffer (pH 8.0) supplemented with protease inhibitors (Thermo Scientific, #78439) and phosphatase inhibitors (Sigma, #P5726 and #P0044). Tissue homogenization was performed using a TissueLyser LT (Qiagen, #85600) at 50 Hz for 5 min. Homogenates were incubated overnight at room temperature with shaking, diluted 1:300 with ice-cold 1% Blocker A in wash buffer, and centrifuged at 17,000×g for 15 min at 4°C. The supernatant was aliquoted and stored at −80°C until analysis.

Levels of vitreous STMN2 were assayed without dilution using human Stathmin 2 (STMN2) enzyme-linked immunosorbent assay (Abbexa Ltd, abx 585079) following the manufacturer's instructions. Other immunoassays were conducted on the MSD platform following the manufacturer's instructions with all samples measured in duplicate. Samples were analyzed for NfL using the R-Plex human neurofilament L antibody set (F217X-3, MSD) per the manufacturer's instructions. Tau-specific antibodies BT2 and HT7 (Thermo Fisher Scientific, #MN1010 and #MN1000B) were employed for the detection of tTau. Sulfo-tag conjugated secondary antibody was used for signal detection, and the multi-detection SECTOR 2400 Imager (MSD) was utilized for biomarker quantification.

Statistical analyses

All fluid biomarker concentrations, including vitreous STMN2, were log-transformed because residuals were not normally distributed. Differences in vitreous STMN2 concentrations were analyzed using the one-sided Mann-Whitney U test based on the a priori hypothesis that STMN2 levels in vitreous humor are lower when TDP-43 inclusions are present in corresponding brains. Vitreous STMN2 was compared across neuropathological groups (i.e., AD, CTE, AD + CTE, and control) using the Kruskal-Wallis test. Vitreous STMN2 levels were also compared across Braak stages (B0-3) and Consortium to Establish a Registry for Alzheimer's Disease (CERAD) scores (C0–3) using Kruskal–Wallis tests. Spearman's rank correlation was used to test the association between vitreous STMN2 and (1) years of RHI exposure, as well as (2) other vitreous biomarkers (i.e., tTau, NfL). All analyses were performed in R, version 4.3.1, with p ≤ 0.05 defining statistical significance. No adjustments for multiple comparisons were applied in this exploratory analysis.

Results

Sample features

A total of 72 brain and corresponding eye post-mortem specimens were examined. The median age at death was 73 (IQR: 65 to 81.5) years and 69 (95.8%) specimens were from men (Table 1). Of these specimens, 14 (19.4%) were controls, 13 (18.1%) had AD pathology only, 35 (48.6%) had CTE pathology only, and 10 (13.8%) were diagnosed with combined AD and CTE pathology (Table 1).

Table 1.

Sample characteristics of post-mortem study specimens by neuropathological diagnosis.

Study parameter [Median (IQR) or n (%)]	Entire cohort (n = 72)	Controls (n = 14)	AD only (n = 13)	CTE only (n = 35)	AD + CTE (n = 10)
Age, y	73 (65–81.5)	48.5 (46–68)	75 (70–82.5)	76 (66–83)	80 (72–85)
Sex, Male	69 (95.8)	13 (92.8)	12 (92.3)	34 (97.1)	10 (100)
Exposure Type
Football	39 (54.2)	7 (50.0)	5 (38.5)	21 (60.0)	60 (60.0)
Military	7 (9.7)	2 (14.3)	5 (38.5)	0 (0)	0 (0)
Multiple	18 (25.0)	2 (14.3)	2 (15.4)	11 (31.4)	3 (30.0)
Other	8 (11.1)	3 (21.4)	1 (7.7)	3 (8.6)	1 (10.0)
Contact Sport History, y	11 (8–15)	8 (3–11)	8 (1.5–13)	14 (11–18)	11.5 (9–15)
Braak Stagea
0	15 (21.7)	10 (71.4)	0 (0)	5 (14.3)	0 (0)
I-II	22 (31.9)	3 (21.4)	0 (0)	17 (48.6)	0 (0)
III-IV	23 (33.3)	1 (7.1)	8 (61.5)	9 (25.7)	5 (50.0)
V-VI	9 (13.0)	0 (0)	5 (38.5)	1 (2.9)	5 (50.0)
CERAD Score
0	36 (50.0)	12 (85.7)	0 (0)	24 (68.6)	0 (0)
1	14 (19.4)	2 (14.3)	0 (0)	11 (31.4)	1 (10.0)
2	14 (19.4)	0 (0)	8 (61.5)	0 (0)	6 (60.0)
3	8 (11.1)	0 (0)	5 (38.5)	0 (0)	3 (30.0)
TDP-43 Positivity	27 (37.5)	0 (0)	6 (46.2)	14 (40.0)	7 (70.0)
TDP-43 Inclusions
Amygdala	24 (33.3)	0 (0)	5 (38.5)	13 (37.1)	6 (60.0)
EC	26 (36.1)	0 (0)	6 (46.2)	13 (37.1)	7 (70.0)
Hippocampus	23 (31.9)	0 (0)	5 (38.5)	12 (34.3)	6 (60.0)
DFC	11 (15.3)	0 (0)	2 (15.4)	8 (22.9)	1 (10.0)

IQR: interquartile range; AD: Alzheimer's disease; CTE: chronic traumatic encephalopathy; CERAD: Consortium to Establish a Registry for Alzheimer's Disease; EC: entorhinal cortex; DFC: dorsolateral frontal cortex. ^an = 3 with missing data (2 with other tauopathies excluding Braak staging, 1 unknown).

Relationship between vitreous STMN2 levels and TDP-43 inclusions in the brain, neuropathological diagnosis, Braak stage, CERAD score, and repetitive head impact exposure

Log-transformed vitreous STMN2 levels were significantly lower in cases with TDP-43-positive brains (Figure 1, p = 0.040). Across neuropathological groups (mdn_control = −2.07, n_control = 14; mdn_AD = −3.18, n_AD = 13; mdn_CTE = −3.02, n_CTE = 35; mdn_{AD  +   CTE}  = −2.01, n_{AD  +   CTE}  = 10), no significant differences in vitreous STMN2 levels were detected (Supplemental Figure 1, p = 0.126). Across Braak stages (mdn_B0 = −2.36, n_B0 = 15; mdn_B1 = −2.96, n_B1 = 17; mdn_B2 = −2.80, n_B2 = 22; mdn_B3 = −2.83, n_B3 = 10) and CERAD score (mdn_C0 = −2.91, n_C0 = 33; mdn_C1 = −2.41, n_C1 = 14; mdn_C2 = −2.27, n_C2 = 12; mdn_C3 = −3.21, n_C3 = 8), no significant differences in vitreous STMN2 levels were found either (p_Braak = 0.802, p_CERAD = 0.845). Similarly, no significant association was detected between vitreous STMN2 levels and duration of RHI exposure (Spearman's rank correlation coefficient [r_s] = 0.097, p = 0.437).

Figure 1.

Log-transformed vitreous STMN2 concentrations in relation to brain TDP-43 pathology. Cases with TDP-43 inclusions showed reduced vitreous STMN2 compared to TDP-43-negative cases. Error bars represent the first quartile, median, and third quartile.

Relationship between vitreous STMN2 levels and other fluid biomarkers

Concentrations for all fluid biomarkers and findings from fluid biomarker associations are presented in Table 2. A significant relationship was found between vitreous STMN2 and vitreous NfL (Supplemental Figure 2, r_s = 0.274, p = 0.026) but not CSF NfL (r_s = 0.258, p = 0.204). Associations with cortical and vitreous tTau were not significant (Table 2).

Table 2.

Biomarker levels and Spearman rank correlations between vitreous STMN2 levels and other fluid biomarkers.

Type	Biomarker	Mean ± SD Raw Biomarker Level	Mean ± SD Log-Transformed Biomarker Level	rs	p
Vitreous	STMN2, ng/mL	0.186 ± 0.214	−2.742 ± 2.052	–	–
Cortical	tTau, μg/g	549 ± 818	5.728 ± 1.001	- 0.103	0.419
Vitreous	tTau, pg/mL	(3.32 ± 7.39) × 106	13.448 ± 1.846	- 0.128	0.311
CSF	NfL, pg/mL	58545 ± 68998	9.577 ± 3.303	0.258	0.204
Vitreous	NfL, pg/mL	45367 ± 263510	4.457 ± 4.535	0.274	0.026*

STMN2: stathmin-2; tTau: total tau; NfL: neurofilament light chain. *p < 0.10.

Discussion

Ocular manifestations of TDP-43-related neurodegeneration have been reported before in retinal tissues in the context of ALS, ²³ FTLD,^24–26 AD, ²⁵ and CTE. ²⁷ Numerous visual disturbances, often nonspecific, have also been associated with these proteinopathies. ^{13 32–34,45} Nevertheless, the role of TDP-43 inclusions in the mechanisms underlying these visual changes or as indicators of broader neurodegenerative processes remains unclear; that is, these symptoms are not specific to retinal changes and causal relationships remain under investigation. ⁴⁶ Investigating the potential consequences of retinal TDP-43 inclusions, such as alterations in vitreous STMN2 levels, could reveal deeper mechanistic insights into both possibilities.

Although past studies have shown that STMN2 levels are reduced in brains with TDP-43 inclusions,^14–18 our findings extend this association to the vitreous humor, demonstrating that vitreous STMN2 levels are significantly reduced in post-mortem eyes corresponding to cases with TDP-43-positive brains. In contrast, vitreous STMN2 levels were not significantly related to Braak stage or CERAD score, indicating that these reductions are relatively specific to TDP-43 pathology rather than the overall burden of AD-related neurofibrillary tangles or amyloid plaques.

At the same time, the absence of significant differences in vitreous STMN2 across neuropathological diagnostic groups of AD and CTE might seem inconsistent with prior research on the relationship between STMN2 and TDP-43 in neurodegenerative proteinopathies.^9–13 This discrepancy likely reflects the heterogenous nature of TDP-43 within AD and CTE, where inclusions occur only in a subset of cases and vary widely in their regional density and distribution.^9–13 In addition, the low sample size of this exploratory study could have limited the power required to detect these group-level differences. Finally, co-existing neuropathological processes such as amyloid-β ⁴⁷ and tau ⁴⁸ deposition could attenuate the association between vitreous STMN2 and diagnostic group, even when TDP-43 pathology is present.

While both NfL and tTau are established markers of axonal degeneration, ⁴⁹ a significant correlation was observed between vitreous STMN2 and vitreous NfL, but not with CSF NfL, cortical tTau, or vitreous tTau. This observation aligns with recent evidence demonstrating that loss of STMN2 function leads to neurofilament-dependent axonal degeneration. ¹⁶ Furthermore, these findings could reflect the distinct biological roles of these biomarkers, ⁵⁰ with vitreous STMN2 better corresponding to changes in axonal integrity, reflected in NfL release, rather than microtubule dysfunction and neurofibrillary tangle formation. These associations could also differ depending on the relative dominance of TDP-43 versus tau pathology, so further exploration in mixed-pathology contexts is needed. Finally, tau biomarkers might reflect a later-stage process of neurodegeneration, whereas STMN2 and NfL are more sensitive to early axonal changes. In brief, these findings are consistent with the underlying biological processes: loss of STMN2 reflects TDP-43–mediated axonal dysfunction, aligning with NfL as a marker of axonal injury, whereas tau accumulation is more distinct from STMN2.

The absence of a significant relationship between vitreous STMN2 and duration of RHI exposure is notable, given the established link between RHI, TDP-43 inclusions, and CTE pathology. ¹² Several factors could explain this discrepancy, including the limitations of retrospective self-reported RHI exposure data, low sample size, and the potential for more complex relationships between exposure and vitreous STMN2 expression.

Future studies should address the limitations of this exploratory analysis, including its sample size and cross-sectional design. Longitudinal studies with larger cohorts are needed to further evaluate vitreous STMN2 as a biomarker and to explore its temporal dynamics relative to the progression of TDP-43-associated neurodegeneration. Agreement validation studies are also needed to evaluate differences in STMN2 levels between CSF and ocular compartments (i.e., vitreous humor, aqueous humor, tears), to ensure ocular sampling reflects central pathology with sufficient fidelity to support clinical use. Plus, our cohort was almost entirely male (95.8%), which could limit generalizability of these findings; recruiting more female subjects is needed to determine if these findings hold across sexes. Premortem clinical information (i.e., cognitive status or ophthalmologic exam results) were also unavailable for these cases, preventing correlational analyses between vitreous STMN2 and functional impairments. Finally, unrecognized ocular conditions (e.g., macular degeneration) could have influenced vitreous protein composition, although none were documented for these cases.

Nevertheless, these findings contribute to the growing recognition that ocular compartments can reflect neurodegenerative processes. Multiple steps are needed, however, before vitreous STMN2 could serve as a screening tool for neurodegenerative processes in clinical practice. To start, more fine-tuned ophthalmological procedures are needed for collecting vitreous humor – vitrectomies solely for routine screening are too invasive. That said, patients already undergoing certain ophthalmic procedures could contribute vitreous samples without additional risk. ⁵¹ In a similar vein, when patients receive intravitreal injections for ophthalmic diseases, a small amount of vitreous fluid is released. This vitreous reflux mirrors the composition of the vitreous humor and could be safely collected using micropipette sampling or absorbent strips.^38,39 TDP-43-related changes in STMN2 transcripts precede overt neurodegeneration^4–6; thus, vitreous STMN2, once minimally invasive vitreous sampling becomes more feasible, could flag incipient disease at a prodromal stage. Equally important, these alterations in STMN2 appear unique to TDP-43 pathology such that drops in vitreous STMN2 could distinguish TDP-43-associated conditions from other pathologies. Additionally, combining vitreous STMN2 with advanced imaging modalities sensitive enough to reflect retinal changes in mild traumatic brain injury or AD,^52,53 such as optical coherence tomography, ⁵² could boost prognostic value and provide a more comprehensive understanding of the interplay between ocular and cortical neurodegeneration.

Overall, this study demonstrates that TDP-43-associated neurodegeneration impacts ocular compartments and underscores the potential of STMN2 as a vitreous biomarker. Although further research is needed to elucidate the diagnostic and prognostic utility of vitreous STMN2, ocular fluid biomarkers are a promising avenue for advancing our understanding and detection of neurodegenerative diseases.