A positive association between phosphorylated Tau217 (pTau217) and neural correlations is prevented by human leukocyte antigen allele DRB1*13:01

Abstract

Subtle deviations in zero-lag cross correlations (synchronous neural interactions, SNI) differentiate healthy from pathological brain states. Here, we assessed blood biomarkers of dementia and neurodegeneration in relation to SNI in 175 women participating in a longitudinal study for a total of 348 acquisitions. Of seven biomarkers tested (Aβ_1–40, Aβ_1–42, Aβ_1–42/Aβ_1–40 ratio, neurofilament light chain [NFL], total Tau, phosphorylated Tau181 [pTau181], phosphorylated Tau217 [pTau217]), only pTau217 showed a significant positive association with SNI (P = 0.0027) which was negatively associated with cognitive performance. Apolipoprotein E (ApoE) did not influence the pTau217-SNI association, whereas human leukocyte antigen (HLA) DRB1*13:01 reduced the association. The pTau217-SNI association was pronounced in participants seropositive for human herpes virus 1 (HHV1) and human endogenous retrovirus K (HERVK). This effect was abolished in carriers of HLA DRB1*13:01, consistent with the strong predicted binding affinity of this allele with proteins of these viruses. These findings highlight the association of pTau217 with brain function and point to a central role of HLA-mediated immune responses in preserving brain health.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-44894-7.

Introduction

Magnetoencephalography (MEG) and cognitive function

Dynamic communication across the massive, interconnected network of neurons lies at the heart of brain function. Magnetoencephalography (MEG) measures the magnetic fields generated by motion of ions across cell membranes associated with excitatory synapses along the dendrites of pyramidal cells in the cerebral cortex, providing a direct, non-invasive measure of neuronal activity with millisecond temporal resolution^1,2. MEG has been instrumental in distinguishing healthy brain function from anomalies associated with conditions affecting the brain including cognitive decline^3–11. Previous research has shown that subtle decrements in cognitive performance are significantly associated with increased synchronous neural interactions (SNI), even among cognitively healthy individuals, whereas high performance on cognitive screening measures is associated with neural network decorrelation¹¹. Given the relevance of neural decorrelation for information processing^12,13, subtle increases in SNI may portend cognitive impairment associated with restriction in dynamic communication across neural ensembles that impact information processing¹¹, the progression of which may be associated with objective cognitive decline and neurodegeneration.

Blood biomarkers for dementia

Blood biomarkers have emerged as a promising tool for detecting proteins that are associated with neurodegeneration and dementia^14,15. With the advent of ultrasensitive techniques, biomarkers obtained from the blood have proven to be highly accurate, approaching the accuracy of cerebrospinal fluid (CSF) biomarkers¹⁶. The most commonly studied biomarkers include those evaluating amyloid-beta (Aβ_1–40 [Aβ40], Aβ_1–42 [Aβ42], Aβ42/Aβ40 ratio) and phosphorylated tau (pTau181, pTau217) proteins, which are commonly associated with Alzheimer’s dementia, as well as markers of neurodegeneration including total tau and neurofilament light chain (NFL)^17–20. Remarkably, an increase in biomarker levels can precede dementia diagnosis by up to 20 years^21,22, highlighting dementia-related brain changes that occur well in advance of overt disease. Here, in a sample of non-demented women, we evaluated the association of blood biomarkers of dementia and neurodegeneration with SNI. In addition, we assessed potentially moderating effects on that association conferred by genes and viruses that have been implicated in dementia risk or protection.

Apolipoprotein E (ApoE)

With regard to genetic influences, it is well-established that individual variation in ApoE influences risk of Alzheimer’s dementia (AD). Specifically, the presence of ApoE2 is associated with decreased risk of AD whereas ApoE4 increases AD risk^23–26. Furthermore, ApoE4 has been linked to poorer cognitive performance in cognitively unimpaired individuals²⁷. Finally, ApoE4 has been associated with altered brain function not only in AD²⁸ but also in otherwise healthy adults²⁹.

Human leukocyte antigen (HLA)

In contrast to detrimental effects of ApoE4 on brain function, certain HLA alleles have been associated with protective effects. Specifically, DRB1*13 alleles (DRB1*13:01, DRB1*13:02) have been associated with protection against age-related brain atrophy and neural network variability (i.e., “noise”) even in the presence of ApoE4^30–32. Various HLA alleles have been identified as a risk or protective factors for several human diseases³³ including dementia^34–38. HLA genes code for cell-surface proteins that are involved in the adaptive immune system response to foreign antigens such as viruses³⁹; thus, the beneficial effects of certain HLA alleles are presumed to be related to an effective immune system response aimed at neutralizing virus (or other foreign) antigens that could otherwise contribute to disease^40,41.

Human viruses

Several viruses have been implicated in conditions affecting the brain including dementia⁴². In particular, human herpesviruses (HHV), several of which are neurotropic, have been implicated in AD^43–49 and in impaired cognitive performance in healthy adults⁵⁰. HHV infections are very common in the general population^51,52 with seroprevalence estimates often > 90% for several HHVs⁵³. We recently documented age-related increases in blood biomarkers of dementia in cognitively unimpaired women who were seropositive for HHVs compared to women who were seronegative, independently of ApoE4 presence⁵⁴; that effect was strongest for HHV1 (herpes simplex virus 1, HSV1), HHV4 (Epstein-Barr Virus), and HHV6 (Roseola virus). In addition to HHVs, other common viruses, including human endogenous retroviruses (HERVs) and human papilloma virus (HPV) are associated with neurodegenerative disorders in observational and mechanistic studies^55–61.

This study

In this work, we integrated measures of brain function, blood biomarkers, genetics, and virus serology. Using a cohort of women who had not been diagnosed with dementia, we examined associations between blood biomarkers of dementia/neurodegeneration and neural network interactions, and evaluated the possible moderating effects of ApoE genotype, HLA-DRB1*13 alleles, and virus seropositivity.

Results

General

Descriptive statistics of age (Fig. 1A), Montreal Cognitive Assessment (MoCA) scores (Fig. 1B), seven biomarker measurements, and synchronous neural interactions (SNI) are given in Table 1. The ApoE distribution is presented in Table 2. With respect to HLA, we determined the presence/absence of HLA DRB1*13:01 and DRB1*13:02 for the 175 participants (Table 3). The seropositivity rates across the 348 acquisitions varied among the viruses tested (Table 4). HHV3 had a 100% seropositivity rate, so no further analyses were conducted for this virus.

Fig. 1.

Frequency distributions of age (A) and MoCA scores (B). N = 348 acquisitions.

Table 1.

Descriptive statistics for the biomarkers used.

	Mean	SEM	Median	N
Age (y)	69.2	0.63	70.7	348
MoCA	27.4	0.11	28.0	348
Aβ40	53.4	2.75	46.3	348
Aβ42	8.33	0.35	7.78	348
Aβ42/Αβ40	0.208	0.016	0.149	313
NFL	52.5	3.10	32.1	347
tTau	3.51	0.18	2.53	318
pTau181	0.441	0.030	0.250	348
pTau217	1.070	0.094	0.545	348
SNI	0.0367	0.000331	0.0354	348

Values for MoCA (includes education point) and Aβ42/Αβ40 are pure numbers. All other biomarker values are pg/ml determined in serum.

Table 2.

ApoE genotypes of participants and ApoE2, ApoE4 groups.

ApoE	Ν	%
22	1	0.6	ApoE2[ε2ε2, ε3ε2]	23
23	22	12.6
24	3	1.7
33	108	62.1
34	34	19.5	ApoE4[ε4ε4, ε4ε3]	40
44	6	3.4
Total	174	100%

Ν = 172 participants for whom ApoE determinations were available. See text for details.

Table 3.

Genotypes of HLA DRB1*13:01 and DRB1*13:02 (N = 175 participants).

	DRB1*13:01		DRB1*13:02
	N	%	N	%
Absent	153	87.4	160	91.4
Present	22	12.6	15	8.6
Total	175	100%	175	100%

Table 4.

Prevalence of seropositivity for the viruses investigated.

Virus strain		Total N	N seropositive	% seropositive
HHV1	HSV1	345	198	42.2
HHV2	HSV2	342	77	22.5
HHV4	EBV	346	333	95.7
HHV5	CMV	347	185	53.2
HHV6	6A, B	331	304	87.4
HERVK		348	238	68.4
HERVW		348	317	91.1
HPV	6/11/16/18/31/33/45/52/58	348	183	52.6

Numbers in the column total indicate the number of tests that gave unambiguous results in the ELISA.

Association of biomarkers with SNI

The possible association of the various biomarkers with the MEG measure SNI was evaluated using linear mixed-effects models [LMEM; see “Methods” section] where Subjects were the participants (coded as nominal variable), Repeated were the visits (coded as consecutive integers starting with 0), Random command included an intercept, SNI was the dependent variable, and age, Aβ40, Aβ42, Aβ42/40, NFL, Total Tau, pTau181, and pTau217 were covariates. The results are shown in Table 5 where it can be seen that only pTau217 had a significant, positive effect on SNI (P = 0.0027). Therefore, subsequent analyses were carried out only for pTau217.

Table 5.

Results of biomarker-SNI associations.

Biomarker	P value	Biomarker-SNI association
Age	0.154 (NS)
Aβ40	0.335 (NS)
Aβ42	0.128 (NS)
Aβ42/40	0.776 (NS)
NFL	0.074 (NS)
Total Tau	0.577 (NS)
pTau181	0.376 (NS)
pTau217	0.0027	Positive

LMEM analysis. P values refer to the testing the statistical significance of biomarker-SNI association (controlling for age).

Effect of ApoE genotype on pTau217-SNI association

This effect was evaluated by performing LMEM as above (Subjects, Repeated, Random intercept), where SNI was the dependent variable, ApoE2_4 was a fixed factor for ApoE2[ε2ε2, ε3ε2] and ApoE4[ε4ε4, ε4ε3] groups, coded as a binary fixed factor), age and pTau217 were covariates, and ApoE2_4 × pTau217 was an interaction term. No statistically significant effect was found for the ApoE2_4 × pTau217 term (P = 0.295) indicating that the null hypothesis that the SNI vs. pTau217 slopes did not differ between the two ApoE2_4 groups could not be rejected.

Effect of HLA DRB1*13:01 and DRB1*13:02 alleles on pTau217-SNI association

This effect was evaluated by performing a LMEM analysis where SNI was the dependent variable, DRB1*13:01 was a fixed factor (coded as a binary fixed factor), age and pTau217 were covariates, and DRB1*13:01 × pTau217 was an interaction term, The results revealed a statistically significant effect of DRB1*13:01 × pTau217 interaction term (P = 0.002) indicating that the SNI vs. pTau217 parameters (slopes) were not parallel between the carriers and non-carriers of the DRB1*13:01 allele. These slopes were compared directly by performing two LMEM analyses, one for the DRB1*13:01(−) group and another for the DRB1*13:01(+) group. In these analyses, SNI was the dependent variable, and age and pTau217 were covariates. The slope in the DRB1*13:01(+) group was negative and significantly smaller than that in the DRB1*13:01(−) group (positive) (Z = 4.646, P < 0.00001, Paternoster test). These results indicate that the presence of DRB1*13:01 allele lowers the positive dependence of SNI on pTau217. In contrast, the presence of DRB1*13:02 did not have a statistically significant effect.

Effect of viruses on pTau217-SNI association

The effect of seroprevalence (presence/absence of seropositivity) on pTau217-SNI association of each virus was evaluated in the same way described above for other factors. The results are shown in Table 6. It can be seen that only seropositivity for HHV1 and HERVK had a statistically significant effect, both increasing the dependence of SNI on pTau217. Both HHV1 and HERVK effects were absent in carriers of DRB1*13:01 (Table 7). DRB1:13:02 did not have statistically significant effect. Notably, DRB1*13:01 showed strong predicted bindings (IC₅₀ < 50 nM) to HHV1 and HERVK (best predicted binding affinity 22.3 and 10.0 nM, respectively) (Table 8).

Table 6.

Effect of virus seropositivity on pTau217-SNI association.

	Interaction in LMEM	Comparison of slopes (Paternoster test)
HHV1	P = 0.008	P = 0.0027 (Virus present > Virus absent)
HHV2	NS
HHV4	NS
HHV5	NS
HHV6/7	NS
HERVK	P = 0.006	P = 0.0071 (Virus present > Virus absent)
HERVW	NS
HPV	NS

Table 7.

Effect of HHV1 and HERVK seropositivity on pTau217-SNI association in the absence/presence of DRB1*13:01.

		Interaction in LMEM
		DRB1*13:01
		Absent	Present
Virus	HHV1	P = 0.037(+)	NS
	HERVK	P < 0.001(+)	NS

Table 8.

Viruses tested for predicted binding affinity.

Virus	Protein description	Uniprot ID	AA	N epitopes
HHV1	Envelope glycoprotein D	Q69091	394	380
HERVK	10 Pol protein	P10266	1014	1000

MoCA

MoCA is an estimate of overall cognitive function. Here we evaluated the effects on MoCA of age, SNI, all 7 biomarkers, and DRB1*13:01 and DRB1*13:02 using LMEM analysis, where the MoCA score was the dependent variable, as described below.

Effect of age

In this LMEM analysis age was the sole covariate. We found a highly statistically significant negative effect of age on MoCA (P < 0.001), i.e. MoCA score decreased with age. In addition, in the same model, we evaluated the effect of individual biomarkers on MoCA, controlling for age. Of the 7 biomarkers, only pTau217 had a highly significant negative effect on MoCA (P < 0.001).

Effect of pTau217-SNI

In this LMEM analysis age, SNI and pTau217 were covariates and pTau217-SNI was an interaction term. We found a highly statistically significant negative effect of the pTau217-SNI interaction on MoCA (P < 0.001).

Effects of DRB1*13:01 and DRB1*13:02

The analysis above was performed for the absence/presence of DRB1*13:01 and, similarly, for DRB1*13:02. In the absence of either of these two HLA alleles, the pTau217-SNI association was associated with a highly significant negative effect on MoCA (P < 0.001 for DRB1*13:01 and P = 0.004 for DRB1*13:02). In contrast, no such significant association was found in carriers of either allele (P = 0.232 for DRB1*13:01 and P = 0.073 for DRB1*13:02).

Discussion

pTau217 and brain function

Despite individual differences in brain anatomy, synchronous neural interactions are remarkably consistent across individuals³ such that deviations are informative about brain health⁴. Here we show that brain network function is strongly associated with serum pTau217. We have previously shown that decrements in cognitive functioning are associated with increased neural network correlations which are thought to constrain information processing by reducing neural network flexibility¹¹. Here, in an extension of those findings, we documented that cognitive performance declined as the pTau217-SNI association increased. The strong association of pTau217 with SNI was particularly robust in the presence of HHV1 or HERVK seropositivity. Remarkably, the pTau217-SNI association and its effect on MoCA were eliminated in the presence of HLA-DRB1*13:01, which is characterized by high-affinity binding to the viruses examined. These findings underscore the association of serum pTau217 on brain function and provide additional support for robust protective effects of DRB1*13:01, highlighting the central role of HLA-mediated adaptive immune responses in maintaining brain health.

Of the seven biomarkers evaluated in the present study, only pTau217 was significantly associated with brain functioning. Several prior studies have documented the sensitivity of pTau217. For example, pTau217 outperforms other biomarkers in diagnosis of AD, even at early stages of the disease, and in distinguishing AD from other neurodegenerative disorders^62–65. Furthermore, blood pTau, and particularly pTau217, has been associated with longitudinal cognitive decline in a sample of non-demented individuals⁶⁶, and has been shown to predict AD diagnosis when combined with ApoE and cognitive performance⁶⁷. The mechanisms underlying the role of pTau217 on neurodegeneration are still under investigation; however, evidence suggests that toxic gain-of-function resulting from trans-synaptic spread of misfolded tau proteins as well as reduced clearance of pathogenic tau, neuroinflammation, and interaction with other proteins (e.g., amyloid beta) likely contribute to pathology, including neuronal dysfunction and synaptic degeneration^68,69. Indeed, tau has been linked to numerous synaptic functions in healthy brains including regulation of synaptic mitochondria and neurotransmitter receptors as well as directly interacting with proteins and post-synaptic signaling complexes, any of which may be impaired by tau alterations including phosphorylation⁷⁰. Notably, since MEG measures synaptic activity^1,2 it may be optimally equipped to capture pTau217-associated synaptic degeneration. It is worth noting that the effects of pTau217 on brain function documented here were independent of ApoE. While some evidence suggests that pTau is more prone to drive trans-synaptic progression of tau pathology and tau-mediated synaptic disruption in ApoE4 carriers⁷¹, ApoE4-associated AD risk is variable⁷² and moderated by multiple factors that influence brain health and cognitive function^{32,34,73–75}.

Effect of virus seropositivity on pTau217 association with SNI

Causal connections between microbial infection and neurodegeneration, particularly AD, have been investigated for decades⁷⁶, with several HHVs and other infectious agents implicated in cognitive decline and AD^50,77,78. We recently reported age-related increase in blood biomarkers of AD in HHV seropositive individuals⁵⁴. Here we found that history of infection with HHV1 or HERVK, both of which have been implicated in AD and neurodegenerative disease^{43,44,56,57,79}, increase pTau217-SNI associations, controlling for age. These findings add to the extensive body of research linking viruses to cognitive decline and highlight the influence of virus infection on blood biomarkers of AD and brain function. Many viruses are neurotropic and/or capable of entering the brain via peripheral nerves or the bloodstream^80,81, resulting in neuroinflammation⁷⁷. Moreover, both tau and HHVs induce activation of HERVs⁸¹ which are increasingly recognized for their involvement in the pathogenesis of many brain disorders⁵⁵; however, morbidity and mortality associated with HERVK is partially moderated by HLA^41,46,82.

HLA DRB1*13:01 reduces pTau217 association with brain function

Previous studies have documented protective effects of DRB1*13 alleles on brain atrophy, neuronal dysfunction, and various conditions affecting the brain^30–37. In the present study, we found that the significant association between pTau217 and brain function was eliminated in DRB1*13:01 carriers; notably, this protective effect persisted even in the presence of virus seropositivity. Since HLA is instrumental in facilitating the human immune system response to viral antigens, the protective effects observed here are presumably due to enhanced immune system response conferred by DRB1*13:01. Indeed, DRB1*13:01 was found to bind with strong affinity to both HHV1 and HERVK, suggesting ability to eliminate and/or neutralize these viruses that might otherwise contribute to disease. The protective effect of DRB1*13:01 on pTau217-SNI association is in line with a proposed role of HLA in dementia prevention⁴¹. We proposed that HLA alleles capable of effectively binding viral or microbial antigens are likely to promote health. In contrast, insufficient HLA-antigen binding results in antigen persistence and neuronal damage⁴⁰. One role of ApoE is neuronal repair, particularly in the case of ApoE2 and ApoE3; however, ApoE4 expression is associated with neurotoxic fragments⁸³. Thus, synthesis of ApoE4 resulting from antigen persistence is likely to contribute to neurodegeneration. From this perspective, HLA exerts a primary influence on viral contributions to neurodegeneration.

Summary, considerations, and limitations

Taken together, the present findings suggest that serum tau phosphorylation is significantly associated with neural dysfunction in non-demented women, particularly in those with history of certain viral infections who are lacking protective HLA alleles. Most participants in this sample were cognitively normal based on MoCA, a well-validated cognitive screening tool. Even so, blood biomarkers of dementia were detectable, and, in the case of pTau217, significantly associated with neuronal functioning. Given the negative association of pTau217-SNI with cognitive performance, we anticipate that these effects may be even more pronounced in samples of cognitively impaired individuals, although this remains to be investigated. Importantly, the protective effects documented here for DRB1*13:01 do not exclude contributions from other HLA alleles. HLA is highly polymorphic, and several variants have been implicated in brain function³² and AD risk^34–38. In addition, the present study was limited to a subset of viruses; however, numerous other viruses and bacteria have been implicated in neurodegeneration and dementia. Finally, the study exclusively focused on women; in light of sex differences in dementia and immune responses^84,85, it remains to be seen whether the current findings extend to men. Future studies evaluating sex differences regarding the influence of viruses and genetics on blood biomarkers and neuronal function are warranted.

Materials and methods

Participants

A total of 175 women participated in the study as paid volunteers. The data were obtained as part of an ongoing longitudinal study involving annual data acquisition; consequently, the number of annual visits varied for participants, for a total of 348 visits. Women were excluded from the study if they, at the time of consent, had been diagnosed at any point in their lifetime with a neurological disorder, any autoimmune disorder associated with neurocognitive dysfunction (e.g., systemic lupus erythematous, rheumatoid arthritis), any major medical condition affecting brain function (e.g., brain cancer, head injury with cognitive sequelae), serious psychiatric diagnoses (e.g., bipolar disorder, schizophrenia, any history of psychiatric hospitalization), or any recent/current medication or treatment known to affect brain function (e.g., radiation, chemotherapy). Finally, with respect to education, most participants were well educated: 7% had completed high school or equivalent, 17% had completed 3 years of college or less, and 76% of the had completed 4 years of college or more. Written informed consent was obtained from study participants. The institutional review board and relevant committees of the Minneapolis VA Health Care System approved the study protocol which was performed in accordance with the Declaration of Helsinki.

Cognitive assessment

MoCA⁸⁶ was administered to screen participants for cognitive impairment. MoCA assesses several domains of cognitive function including executive function, memory, language, and abstract reasoning, among others. Scores for each domain were added to reflect a total score ranging from 0 to 30 (without the education point).

ApoE genotyping

Determination of ApoE genotype was performed as follows for 174/175 participants (ApoE could not be determined in one participant due to technical reasons). DNA samples were genotyped using PCR amplification followed by restriction enzyme digestion⁸⁷. Each amplification reaction contained PCR buffer with 15 mmol/L MgCl 2 nanograms of of genomic DNA, 20 pmol ApoE forward (5N TAA GCT TGG CAC GGC TGT CCA AGG A 3N) and reverse (5N ATA AAT ATA AAA TAT AAA TAA CAG AAT TCG CCC CGG CCT GGT ACA C 3N) primers, 1.25 mmol/L of each deoxynucleotide triphosphate, 10% dimethylsulfoxide, and 0.25 μL AmpliTaq Gold DNA polymerase (Thermo Fisher Scientific, Waltham MA, USA). Reaction conditions in a thermocycler included an initial denaturing period of 3 min at 95 C, 1 min at 60 C, and 2 min at 72 C; followed by 32 cycles of 1 min at 95 C, 1 min at 60 C, and 2 min at 72 C; and a final extension of 1 min at 95 C, 1 min at 60 C, and 3 min at 72 C. PCR products were digested with HhaI and separated on a 4% Agarose gel which was stained with Ethidium Bromide. Known ApoE isoform standards were included in the analysis.

Determination of blood biomarkers in serum

We determined the serum levels of the following dementia-related biomarkers: Aβ40, Aβ42, Neurofilament light (NFL), total Tau (tTau), phosphorylated Tau 181 (pTau181), and phosphorylated Tau217 (PtAU217), as follows. Αβ40 and Αβ42 were determined using the FUJIFILM Wako ELISA kits (Catalog # 296-64401 and 298-64601). NFL and tTau were determined on the Ella Automated Immunoassay System (ProteinSimple, part of Bio-Techne) using the Simple Plex Human Total Tau (Catalog # SPCKC-PS-009562) and NFL (Catalog # SPCKB-PS-002448) cartridges. pTau181 and pTau217 were determined on a MESO SECTOR S 600MM using the S-PLEX Human Tau kits (Meso Scale Discovery, Catalog # K151AGMS and K151APFS). Assays were performed and values calculated following the manufacturers’ instructions. Serum samples were diluted 1:2 by the recommended diluent.

HLA genotyping

DNA isolation was carried out from whole blood or saliva samples using commercially available kits (blood: ArchivePure cat# 2300730 from 5Prime distributed by Fisher Scientific or VWR; saliva: Oragene-Discover cat.OGR-500 coupled with prepIT purifier reagent cat.PT-L2P/DNA Genotek Inc. Ottawa, ON, Canada). The purified DNA samples were sent to HistoGenetics (http://www.histogenetics.com/) for high-resolution HLA Sequence-based Typing (SBT; details are given in https://bioinformatics.bethematchclinical.org/HLA-Resources/HLA-Typing/High-Resolution-Typing-Procedures/ and https://bioinformatics.bethematchclinical.org/WorkArea/DownloadAsset.aspx?id=6482). Their sequencing DNA templates are produced by locus- and group-specific amplifications that include exon 2 and 3 for Class I (A, B, C) and exon 2 for Class II (DRB1, DRB3/4/5, DQB1, and DPB1) and reported as Antigen Recognition Site (ARS) alleles as per ASHI recommendation⁸⁸.

Virus seropositivity assays

We determined the seroprevalence of IgG antibodies against Human Herpes Virus 1(HHV1), HHV2, HHV3, HHV4, HHV5 and HHV6 using commercially available enzyme-linked immunosorbent assay (ELISA) kits; we could not obtain reliable results from kits for HHV7 and HHV8. The ELISA tests were performed as per the manufacturer’s instructions and recommendations. Serum samples were processed using a mini-automated 5-in-1 workstation (Crocodile cat. 84024-01; Berthold Technologies, Oak Ridge, TN, USA). The workstation includes the ELISA microtiter plate reader which was read at dual wavelengths for absorbance at 450 nm and 620 nm as reference wavelengths. Details of the virus-specific ELISA kits are as follows: (a) HHV1: Human Anti-Herpes simplex virus Type 1 IgG ELISA Kit (HSV1), Abcam Inc., Boston, MA USA, cat. ab 108737; (b) HHV2: Human Anti-Herpes simplex virus Type 2 IgG ELISA Kit (HSV2) Abcam cat. ab 108739; (c) HHV3: Human Anti-Varicella-Zoster Virus IgG ELISA Kit (VZV) Abcam cat. ab 108782; (d) HHV4: Human Anti-Epstein Barr virus IgG ELISA Kit (EBV-VCA) Abcam cat. ab 108730; (e) Human Anti-Cytomegalovirus IgG ELISA Kit (CMV) Abcam cat. ab 108724; (f) Human Herpesvirus 6 IgG ELISA Kit cat. KA1457, Abnova, Taiwan. Ambiguous readings were not used in the analyses. (g) ERVK6 ELISA Aviva Systems Biology kit OKEH08186 at 1:4. (h) Syncytin-1/ERVW-1 Abbexa kit abx522100 at 1:5. (i) HPVL1 ELISA: Anti-HPV L1 IgG (types 6/11/16/18/31/33/45/52/58) Alpha Diagnostic International kit 550-500-PHG at 1:200, All assays were performed according to the respective manufacturers’ instructions.

In silico determination of predicted best binding affinities of DRB1*13:01

We estimated in silico the predicted best binding affinity of DRB1*13:01 to the viruses that were associated with biomarker-brain associations (HHV1 and HERVK) using the Immune Epitope Database (IEDB; http://tools.iedb.org/mhci/) NetMHCpan (ver. 4.1 BA) tool⁸⁹. We used the sliding window approach^90,91 to test exhaustively all possible linear 15-mer epitopes of the virus proteins analyzed (Table S1, Supplementary Material); the epitope length of 15 amino acids is optimal for HLA-II molecule binding^92–94. The method is illustrated in Fig. 2 for HERVK. For each pair of peptide-HLA molecule (pHLA-II) tested, this tool gave as an output the IC₅₀ of the predicted binding affinity; the smaller the IC₅₀, the stronger the binding affinity. Given a protein of N amino acid length and an epitope length of 15 AA, there were N-15 + 1 IC₅₀ values. The predicted best binding affinity (PBBA) for each pHLA-II pair was the minimum IC₅₀ value of all epitopes tested for the pair. An IC₅₀ value of IC₅₀ < 50 nm is regarded strong⁹⁵ (“hit”). The number of epitopes tested for each virus is given in Table S1.

Fig. 2.

Schematic diagram to illustrate the sliding window approach for in silico testing epitope binding affinities to DRB1*13:01. See text for details.

Magnetoencephalography (MEG)

All participants underwent a MEG scan. As described previously⁴, participants lay supine within the electromagnetically shielded chamber and fixated their eyes on a spot 65 cm in front of them, for 60 s. MEG data were acquired using a 248-channel axial gradiometer system (Magnes 3600WH, 4-D Neuroimaging, San Diego, CA), band-filtered between 0.1 and 400 Hz, and sampled at 1017.25 Hz, corresponding to a sampling interval of 0.983 ms. Data with artifacts (e.g. from excessive subject motion) were eliminated from further analysis. MEG records were visually screened and rejected if artifacts were present (e.g. from excessive subject motion, eye movements, blinking, or environmental noise).

Processing of the raw MEG series was performed using programs in Python⁹⁶. Single trial MEG time series from all sensors underwent ‘prewhitening’⁹⁷ using a (50, 1, 3) Auto Regressive Integrated Moving Average (ARIMA) model to obtain innovations (i.e. residuals)⁹⁶. All possible pairwise zero-lag crosscorrelations, (Synchronous Neural Interactions, SNI; N =  = 30,628 sensor pairs) were computed between the prewhitened MEG time series of each MEG scan. Crosscorrelations were transformed to using Fisher’s⁹⁸ z-transformation to normalize their distribution:

1

Finally, since we focused on the strength of neural interactions, irrespective of their sign, we used the absolute value of for further analyses:

2

Statistical analyses

Statistical analyses were performed using the IBM-SPSS statistical package (version 30) using standard statistical methods. Since data at different visits were acquired from the same participants in several cases, a linear mixed-effects models (LMEM) analysis was performed using the mixed models linear procedure in SPSS, where Subjects were the participants and Visits were the Repeated variable (expressed as consecutive integers with the first one set to zero), Random command with intercept; SNI was the dependent variable, various binary groups (ApoE, HLA) were fixed factors, and age and biomarkers were covariates. Additionally, the effect of fixed factors (ApoE, DRB1 alleles) on the pTau217-SNI association was assessed by including a Fixed Factor main effect and a Fixed Factor × pTau217 interaction term to test the null hypothesis that the SNI vs. pTau217 parameters (slopes) are not parallel. If this interaction term was statistically significant, rejecting the null hypothesis above, the two slopes were compared directly using the Paternoster test⁹⁹:

3

where and are LMEM parameters estimates (slopes), SE are their standard errors, and Z is the normal deviate. All P values reported are 2-sided, .

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

G.S. performed the laboratory experiments. A.C.L. acquired and pre-processed MEG data. A.P.G. performed data analysis. L.M.J. and A.P.G. wrote the paper. All authors edited and approved the paper.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Partial funding for this study was provided by the University of Minnesota (the Anita Kunin Chair in Women’s Healthy Brain Aging, the Brain and Genomics Fund, the McKnight Presidential Chair of Cognitive Neuroscience, and the American Legion Brain Sciences Chair) and the U.S. Department of Veterans Affairs. The sponsors had no role in the current study design, analysis or interpretation, or in the writing of this paper. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

Written informed consent was obtained from all study participants. All experiments were approved by the Minneapolis VAHCS.