Serum Thioredoxin-80 is associated with age, ApoE4, and neuropathological biomarkers in Alzheimer’s disease: a potential early sign of AD

This study included 99 patients equally distributed between the clinical diagnostic groups subjective cognitive impairment (SCI) mild cognitive impairment (MCI) and AD from the Karolinska University Hospital memory clinic in Huddinge, Sweden. The demographical characteristics of the GEDOC memory clinic population are described in Table 1. The clinical and demographical data relevant for the study (e.g., ApoE genotype) was acquired from the central GEDOC database (Karolinska University Hospital, Stockholm, Sweden).

Diagnosis for MCI was done using the consensus criteria for MCI which require the presence of both subjective and objective cognitive impairment including one or several cognitive domains, but no dementia or impairment of daily living activities [20]. Dementia diagnoses were carried out following the criteria of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV), as previously described [21].

Routine neurological and physical examinations were carried out at the memory clinic as described previously [21] and included Mini-Mental State Examination (MMSE), blood tests and CSF sampling. CSF samples were collected as previously described [21]. Fresh samples were used to measure soluble Aβ42, t-tau, and p-tau concentrations in CSF with commercially available sandwich enzyme-linked immunosorbent assays (ELISA) (Innogenetics, Belgium) according to standardized protocols in the memory clinic.

This exploratory sub-study included baseline data from 47 FINGER trial participants (21 women and 26 men, mean age 71 ± 5.1 years) who underwent neuroimaging in Turku (Finland) [22] with serum samples available for Trx80 measurements (For CONSORT flowchart, see Fig. 1A). They were selected at the time when MRI resources became available, and if there were no contraindications. The demographical and clinical characteristics of these participants have been previously described [22, 23], and they were not different from the rest of the FINGER participants.

The FINGER population characteristics [24], trial protocol details [25], main relevant findings [26], and neuroimaging sub-study have been previously published in detail [22, 27]. The demographical, clinical, and neuroimaging data relevant for the study was acquired from the central FINGER database (Finnish Institute for Health and Welfare, Helsinki, Finland). In brief, the data collection was ongoing between September 7, 2009, and November 24, 2011, where 2654 individuals were screened, and 1260 participants of ages between 60 and 77 years from the general population were recruited based on cognitive performance at the mean level or slightly lower than expected for their age according to Finnish population norms for the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) [28] and 6 points or higher in the Cardiovascular Risk Factors, Aging and Dementia (CAIDE) risk score [29]. Diagnosis of dementia or substantial cognitive impairment was used as reason for exclusion as well as any condition affecting safe participation.

The study was approved by the Regional Ethical Review Board in Stockholm, and written informed consent was obtained from all participants.

Mini-Mental State Examination (MMSE) was conducted as standard cognitive testing for all GEDOC and FINGER participants. Higher scores indicated better performance.

All participants underwent a brain 3T MRI (3D Turbo Field Echo sequence, voxel size 1.0 × 1.0 × 1.0 mm, total slices 160, field of view 240 × 240 mm, repetition time 8.1 ms, echo time 3.7 ms, Philips Ingenuity TF PET/MR, Amsterdam, the Netherlands). The specific protocol for MRI imaging has been previously described [22]. T1WI and FLAIR images were quality checked and visually inspected for any abnormalities by an experienced neuroradiologist. Images were excluded if there were brain lesions potentially affecting volumetry and/or scanning issues. Regular phantom scans were performed, and quantitative measures of signal-to-noise ratio, uniformity, and geometric distortion were carried out.

Freesurfer image analysis (version 5.3.0) was used to measure cortical thickness and regional brain volumes. Manual editing was performed in cases where automated white matter (WM) segmentation presented geometric inaccuracies in boundaries between CSF, gray, and white matter. Brain volumes were adjusted for head size to account for between-patient variations in head size [30]. Cortical thickness measurement in AD signature regions was calculated as the mean average of cortical thickness in the entorhinal, inferior temporal, middle temporal, and fusiform regions [31].

White matter lesions (WML) volume was measured through the segmentation of WM hyperintensities on T1 and FLAIR images according to a previously described method [32]. The segmentation was done in 3 steps based on the expectation-maximization algorithm: [1] from the T1 images, WM was segmented into two classes, normal-bright WM regions, and hypointense WM regions [2]; FLAIR images were segmented to three classes: CSF, normal brain tissue and hyperintense voxels [3]; the WM and subcortical regions were segmented into two classes from the FLAIR images. The segmentation of WM hyperintensities was regarded as the class with higher intensities [32, 33].

Determination of Trx80 levels in serum was performed with a sandwich ELISA as previously described [34] but with a few modifications. In brief, standard samples of recombinant human Trx80 (#11522), coating anti-Trx80 monoclonal mouse antibody (clone 7D11 #11543), and detection antibody (biotinylated goat polyclonal anti-Trx1 #11541) were purchased from Cayman chemicals (USA). Standard dilutions of Trx80 (0.2–125 ng/ml) and serum samples were prepared in blocking buffer (0.5% BSA, 0.05% Tween-20); 50 μl of standards or samples were added in duplicates and incubated overnight (O/N). Biotinylated goat anti-Trx1 antibody was prepared at a concentration of 2 μg/ml in blocking buffer (50 μl/well) and incubated 2 h at room temperature. Streptavidin-HRP (#OR03L, Merck, USA) was diluted to 10 ng/ml in PBS and incubated 45 min at room temperature. TMB substrates (#34021, Thermo Fisher Scientific, USA) were mixed 1:1, added 100 μl/well, and incubated 25 min in the dark; 50 μl/well of stop solution (2N sulfuric acid, DY994, R&D systems, USA) was added and absorbance was immediately measured at 450 nm in the spectrophotometer (Tecan Safire 2 Multi-Detection Plate Reader, Switzerland).

In the FINGER study, venous blood samples were taken at baseline in fasting status and using EDTA tubes. Plasma aliquots were stored at – 80 °C until analysis. A panel of cytokines, chemokines, and growth factors (IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12(p40), IL-12(p70), IL-13, IL-15, IL-17, IFN-α2, IFN-γ, MIP-1α, TNF-β) were analyzed with a multiplex suspension array system using Bioplex Luminex 200 instrument (Bio-Rad Laboratories, Hercules, CA, USA) and the MILLIPLEX® MAP Human Cytokine/Chemokine panel (Merck Millipore, Darmstadt, Germany).

The assays were performed in one batch, and samples preparation and setting of the system running protocol were done following the manufacturer’s instructions. All samples and standards were run in duplicate and were measured as pg/ml. Quality controls were performed according to the manufacturer guidelines to ensure accuracy of measurements. After the plate reading, the results files were generated using Bio-Plex Manager software 4 (Bio-Rad Laboratories, Hercules, CA, USA).

Human brain tissue samples were obtained from CIEN foundation tissue bank for neurological research (BT-CIIN). Information regarding age, sex, ApoE4 genotype, postmortem interval, and dementia diagnosis from the brain tissue is displayed in Supplementary Table 1.

Immunoblotting was performed as described [35]. Briefly, human brain tissue was homogenized in lysis buffer (50 mM Tris-HCL, 150 mM NaCl, 1% Triton-X) containing phosphatase and protease inhibitors (Sigma-Aldrich, USA). Homogenates were then centrifuged at 15000×g for 15 min 4 °C, and samples were mixed with equal volumes of loading buffer (160 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 0.01% bromophenol blue, 100 mM DTT). Processed samples were then run in 12.5% polyacrylamide gels (Bio-Rad, USA) and transferred to BioTraceTM nitrocellulose membranes (GE healthcare, USA) for 2 h room temperature at 50 V. Membranes were then blocked for 1 h in 5% skim milk in TBS-Tween-20 (TBS-T) prior to overnight incubation with diluted primary antibody at 4 °C. Primary antibodies: 1:1000 anti-Trx80 monoclonal mouse antibody (clone 7D11 #11543, Cayman chemicals, USA), 1:1000 anti-Trx1(human) Polyclonal Goat Antibody (#11538 Cayman chemicals, USA), and 1:10000 monoclonal Anti-β-Actin mouse antibody, clone AC-15 (Sigma-Aldrich, USA). Fluorescent secondary antibodies (1:10000, LI-COR Biosciences, USA) were used for 2 h at room temperature, and bands were visualized using ODYSSEY Infrared Imaging System (LI-COR Biosciences). Band intensity signal was quantified by ImageJ software. Each band signal value was normalized against loading control (actin signal value).

Demographic data were compared across clinical diagnostic groups with ANOVA for continuous and χ² for categorical variables. Simple regression analysis was applied when exploring the relationship between age, sex and ApoE4 carriership and Trx80 levels. Further associations were analyzed by multiple linear regression models adjusted for age and clinical diagnosis when appropriate. For variables that were not normally distributed (CSF AD biomarkers (Ab42, t-tau, p-tau and p-tau/t-tau ratio), inflammatory cytokines and Trx80), zero-skewness log transformation was applied. ANCOVA tests adjusting for age were used to compare CSF biomarker and Trx80 levels between groups and Trx80 levels between ApoE4 carriers and non-carriers. Student t-test was used to compare hippocampal volumes between groups. Correction for multiple testing has not been applied, and all the analyses performed are post hoc. Level of significance was set to p < 0.05 in all analyses. Analysis was performed using the Stata software, version 14 (StataCorp), and GraphPad Prism, version 9 (GraphPad Software, CA).

Demographical and clinical data as well as CSF and serum measurements from a subset of the Karolinska University Hospital memory clinic cohort GEDOC are listed in Table 1. The MCI and AD groups were significantly older than the SCI group (69 ± 7 and 73 ± 10 versus 59 ± 8 years, p < 0.0001), had lower level of education (12 ± 4 , 10 ± 3 versus 14 ± 3 years, p < 0.001), and had lower MMSE scores compared to SCI (27.30 ± 1.84, 24.52 ± 3.63 versus 29.24 ± 1.16 points, p < 0.0001). As expected from previous reported measurements [2], both MCI and AD groups had significantly lower Aβ42 levels compared to SCI (881.87 ± 350.37 ng/l and 658.32 ± 302.59 versus 935.32 ± 229.86 ng/l, p < 0.001), higher t-tau levels than the SCI group (324.58 ± 162.98, 499.71 ± 228.49, versus 224.97 ± 91.82 ng/l, p < 0.0001), and higher p-tau levels in CSF than SCI group (59.13 ± 25.95, 81.03 ± 28.17 versus 46.46 ± 15.32 ng/l, p < 0.0001).

The baseline characteristics of the FINGER sub-study (n = 47) are listed in Table 2. In addition to serum Trx80 levels, the herein analyzed data contains demographics, cognitive measurements, distribution of ApoE genotype, neuroimaging, and inflammatory marker data. Since only baseline data were included in this exploratory study, both control and intervention groups are analyzed as one “at risk of dementia” category.

As reported in Table 1, the SCI group in GEDOC was significantly younger than MCI and AD groups. According to what previously has been shown [17], age (β = 0.33, p = 0.00) but not sex (men, 13.33 ± 3.40 ng/ml; women, 19.41 ± 3.44 ng/ml; p = 0.26) had a significant influence on Trx80 in the GEDOC cohort. In contrast to the GEDOC cohort, where there was a significant difference in age among participants, in the FINGER cohort, there was no association between serum Trx80 levels and age (β = − 0.04, p = 0.77) nor with sex (men, 21.26 ± 8.95 ng/ml; women, 18.24 ± 5.41 ng/ml; p = 0.26). Based on this, we decided to include FINGER sub-study participants in the comparison since they represent non-demented subjects at risk for AD, whose average age (70 ± 5 years) is comparable to MCI and AD groups. When serum Trx80 levels were analyzed between these 4 groups (Fig. 1B), we observed that the SCI group had significantly lower Trx80 levels (3.56 ± 4.87 ng/ml) than the FINGER sub-study (19.46 ± 38.37 ng/ml; p < 0.01), MCI group (28.13 ± 30.39 ng/ml; p < 0.001), and AD group (25.50 ± 28.58 ng/ml; p < 0.001). When comparing the age matched FINGER sub-study participants to AD, significantly higher Trx80 levels were seen in AD (p < 0.02), and a similar trend could be observed for MCI group (p < 0.10). Based on this result, we combined the data from both sub-studies (GEDOC and FINGER) to perform further analysis on serum Trx80 associations (n = 145).

In agreement with the above results, Trx80 levels significantly correlated with age (β = 0.23, p = 0.01) but not with sex (β = 0.14, p = 0.12) in the merged group (Table 3). When adjusting for age, there was a significant association between Trx80 levels and ApoE4 carriers (β = 0.34, p = 0.00).

Data on clinical CSF AD biomarkers were available from the GEDOC memory clinic cohort. To explore how Trx80 varies with known AD pathology, we investigated the association between serum Trx80 levels and CSF Aβ42, t-tau, p-tau levels, and p-tau/t-tau ratio in the GEDOC cohort (Table 4). No associations between Trx80 and Aβ42 or Trx80 and tau were found when adjusting for age. As shown in Table 1, serum Trx80 levels are significantly higher in MCI and AD groups in comparison to SCI. Since serum Trx80 is increased with dementia and taking into account that Aβ42, t-tau, and p-tau levels are determinant in the development of AD, we next analyzed whether these markers correlated with serum Trx80 levels at different disease stages when adjusted for age (SCI, MCI and AD, Table 4). Higher Tx80 levels were significantly associated with lower p-tau levels (β = − 0.48, p = 0.02) in the SCI group, and a similar trend could be observed for t-tau (β = − 0.36, p = 0.08). In the MCI group, higher p-tau/t-tau ratio was associated with higher Trx80 levels (β = 0.41, p = 0.05). There were no further significant associations for Aβ42, t-tau, and p-tau or the ratio in MCI or AD groups.

We next sought to determine if Trx80 is related to brain volume measures. We found that there was a negative association between hippocampal volume and serum Trx80 levels (β = − 0.32, p = 0.01, Table 5). No other association was found significant between gray matter volume, cortical thickness, or white matter lesions and serum Trx80 levels.

No clinical cutoff for Trx80 has been reported; however, we wanted to explore if there were differences in neuroimaging data among FINGER participants when they are divided into two groups according to their serum Trx80 levels. Participants with the highest serum Trx80 levels had significantly lower hippocampal volume than the participants with the lowest serum Trx80 levels (7.07 ± 1.05 ml; 7.82 ± 0.78 ml, respectively, p < 0.01; Fig. 1C). No differences were found when comparing gray matter volume, cortical thickness or white matter lesions between participants with the highest and lowest Trx80 levels.

Since previous studies reported that Trx80 is implicated in the inflammatory response present in atherosclerotic lesions [17, 19], and it induces a pro-inflammatory response in monocytes and macrophages [15, 36, 37], we investigated possible associations between serum Trx80 levels and serum inflammatory cytokine levels of the participants. Higher Trx80 levels were associated with higher levels of IL-8 (β = 0.33, p = 0.05), IL-13 (β = 0.33, p = 0.05), Mip-1a (β = 0.39, p = 0.03), and bradykinin (β = 0.32, p = 0.04), and a similar trend, however not significant, was observed for IL-4 (β = 0.32, p = 0.06), IL-5 (β = 0.31, p = 0.07), IL-6 (β = 0.30, p = 0.07), interferon-α (β = 0.29, p = 0.08), interferon-γ (β = 0.28, p = 0.10), and TNF-β (β = 0.30, p = 0.08).

As shown in Fig. 1D, ApoE4 carriers from both GEDOC and FINGER cohorts had significantly higher (approximately twofold) Trx80 levels than non-carriers (30.09 ± 41.00 ng/ml; 13.39 ± 18.88 ng/ml, respectively, p < 0.01). Due to the significant role of ApoE4 in AD onset and progression, we investigated the associations between Trx80 and CSF AD biomarkers in ApoE4 carriers and non-carriers (Table 6). Higher serum Trx80 levels were associated with lower Aβ42 in CSF (β = − 0.46, p = 0.04) in non-ApoE4 carriers when adjusted for age. A similar trend could be observed when adjusting for disease diagnosis (β = − 0.36, p = 0.07). This association could not be observed in ApoE4 carriers. Total tau and p-tau did not show any significant associations with Trx80 in these groups. The ratio p-tau/t-tau was significantly associated to serum Trx80 levels in ApoE4 non-carriers (β = 0.49, p = 0.00).

We also investigated whether the associations between ApoE4 and MMSE scores with Trx80 were affected over the disease spectrum (Table 7). To this end, we analyzed these associations within each cohort (FINGER participants and memory clinic patients (SCI, MCI and AD)). A positive association was found between serum Trx80 levels and ApoE4 in FINGER participants (β = 0.30, p = 0.05) and in the MCI group from GEDOC cohort (β = 0.59, p = 0.02). This association was not found in the SCI neither in the AD groups. Regarding cognitive score data, there was a significant association between high Trx80 levels and low MMSE scores only the MCI group (β = − 0.41, p = 0.05).

Finally, we measured Trx80 protein levels in post-mortem human brain homogenates of AD and non-AD patients. Demographical and clinical data from the brain samples has been collected in Supplementary Table 1. AD samples were grouped by ApoE genotype in ApoE3/ApoE3 (E3/E3) or E4/E4 (Fig. 1E). Despite the low number sample size (n = 3 per group), only E4/E4 AD patients had significantly lower Trx80 levels than E3/E3 non-demented controls (p < 0.01). As previously reported by our group [14], Trx80 levels in E3/E3 AD patients were also lower than the non-AD controls, although this difference did not reach statistical significance (p = 0.08) in this short sample number study. There were no significant differences in the precursor Trx1 protein levels between groups.

The main limitation of this study is the relatively small sample size of our cohorts that affects the statistical power. This is especially a problem when comparing ApoE4 carrier data between groups. Because no correction for multiple testing has been applied and all the analysis are post hoc, the findings presented above should be considered exploratory and will need to be further verified in studies with larger sample sizes.