Distribution of Big Tau Isoforms in the Human Central and Peripheral Nervous System

Abstract

Objective

Tau is widely studied in neurodegeneration, yet most work has focused on canonical brain tau isoforms. A longer isoform, “big tau,” produced by inclusion of exon 4a, is expressed in the peripheral nervous system (PNS) and central nervous system (CNS) regions. We sought to characterize big tau composition, anatomic distribution, and disease relevance.

Methods

Mass spectrometry (MS) was used to sequence big tau and map its distribution across the human nervous system. Postmortem samples included brain tissue from Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), and controls; spinal cord and peripheral nerves. Big and canonical (“small”) tau isoforms were also quantified in cerebrospinal fluid (CSF) from controls and participants stratified by amyloid status and cognitive impairment.

Results

Human big tau results from insertion of either 355 or 251 amino acids encoded by exon 4a‐long and exon 4a‐short, respectively. Alternative splicing of exons 2, 3, and 10 generates multiple big tau isoforms. Total tau levels were approximately 1,000‐fold higher in the brain than in the PNS; however, the relative abundance of big tau increased from the CNS to the PNS, comprising 50% of the total tau in the periphery and approximately 1% in the brain, primarily localized to the cerebellum. In CSF, big tau levels were unchanged by amyloid abnormalities or cognitive impairment, whereas canonical tau increased with AD pathology.

Interpretation

Big tau represents a distinct tau population enriched in the PNS and uncoupled from disease‐associated changes in brain‐derived tau, suggesting that distinguishing big tau from canonical tau may improve interpretation of tau biomarkers and help differentiate CNS neurodegeneration from peripheral nerve pathology. ANN NEUROL 2026;100:709–736

Neurofibrillary tangles (NFTs) composed of microtubule associated protein tau (MAPT) are one of the key pathological hallmarks of Alzheimer's disease (AD) along with extracellular amyloid‐β plaques. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ Tau pathology within neocortical brain regions is closely linked to cognitive impairment in AD. ⁹ , ¹⁰ , ¹¹ , ¹² Central nervous system (CNS)‐specific tau phosphorylation (p‐tau) have been validated as reliable indicators of CNS tau changes in response to AD pathology. ¹ The presence of AD neuropathological changes in the brain often coincides with an increase in tau and p‐tau species in the cerebrospinal fluid (CSF), making CSF tau biomarkers good candidates for monitoring amyloidosis, NFT formation, and cognitive decline in AD. ¹³ , ¹⁴ , ¹⁵ Recent development of blood‐based tau biomarkers has highlighted that tau measured in plasma reflects at least 2 sources: brain‐derived tau released from the CNS and tau originating from peripheral tissues. ¹⁶ , ¹⁷ , ¹⁸ Peripheral tau species therefore comprise a substantial fraction of the total plasma tau pool, complicating the direct interpretation of blood tau measurements as AD‐specific biomarkers. Discriminating CNS‐derived tau isoforms from those originating in the peripheral nervous system (PNS) may improve the specificity of tau‐based biomarkers for AD and other CNS neurodegenerative disorders. ¹⁹ , ²⁰ Conversely, PNS‐enriched tau isoforms could provide biomarkers for neurological diseases affecting the PNS. ²¹ , ²² , ²³ However, our understanding of PNS‐derived tau in humans remains limited. Characterizing the distribution of PNS‐specific tau isoforms alongside the canonical CNS tau isoforms (referred to as “small tau” henceforth) will be critical for understanding tau biology and its contribution to the pathophysiology of neurodegenerative disorders in which tau is implicated.

Tau is encoded by the MAPT gene located on chromosome 17 in humans, ²⁴ contains 16 exons, ²⁵ and displays remarkable length heterogeneity in its isoforms due to alternative splicing of exons 2 and 3 at the N terminus and exon 10 in the microtubule binding region (MTBR). ²⁶ This leads to the generation of 6 major isoforms in the human CNS; 0N3R, 1N3R, 2N3R, 0N4R, 1N4R, and 2N4R. Expression of these isoforms is developmentally regulated, ²⁷ , ²⁸ with 0N3R being mainly expressed at the fetal stage; then, approximately equivalent expression of 0N3R, 1N3R, 0N4R, and 1N4R is maintained in the adult brain, where 2 N is less abundant. ²⁹ The functional roles of individual tau isoforms remain an active area of investigation, in part because disruption of the normal 3R:4R tau isoform ratio in the brain is a hallmark of several tauopathies. ³⁰ Additionally, multiple post‐translational modifications (PTMs) and truncations lead to a multitude of tau proteoforms known to regulate tau physiological functions and contribute to pathological conformations. ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ Recent advances in Cryo‐EM, mass spectrometry (MS), and genome sequencing techniques have demonstrated the heterogeneity of tau isoforms that are involved in various tauopathies, suggesting that splicing is of key importance in the neuropathological process. ³² , ³³ , ³⁴ , ³⁸ Alternative splicing of the MAPT gene produces 3 transcripts of 2, 6, and 9 kb that are differentially expressed in the nervous system. ³⁹ , ⁴⁰ Tau is also expressed in peripheral tissue (eg, peripheral nerves, heart, skeletal muscle, and the kidneys) outside the CNS where it plays a vital role in metabolism, microtubule formation, and microtubule stabilization. ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ Although the 2 kb transcript produces tau for the cell nucleus, 6 kb transcript expression leads to the common 6 tau isoforms of 45 to 60 kDa, ² whereas the expression of the 9 kb mRNA results in a higher molecular weight isoform (110 kDa) that was originally cloned and sequenced from rat complementary DNA (cDNA) and named big tau. ³⁹ , ⁴⁷ , ⁴⁸

Big tau results from an exon insertion (exon 4a) between exons 4 and 5 in tau transcripts. Translation of this exon 4a dramatically increases the length of the projection domain relative to the canonical small tau isoforms. ³⁹ , ⁴⁵ , ⁴⁶ , ⁴⁹ , ⁵⁰ , ⁵¹ Early work in rodents demonstrated that big tau expression is not limited to the PNS; it also occurs in selective regions in spinal cord and brain‐stem. ⁴⁷ , ⁵² Big tau expression in the superior cervical ganglion (SCG) of the autonomic nervous system, and in the trigeminal ganglion and dorsal root ganglia (DRG) of the PNS was found to be developmentally regulated, beginning late in the embryonic stage, and increasing postnatally. ⁴⁰ , ⁵³ Investigations using big tau specific antibodies in rodents have documented the expression of exon 4a in the spinal motor neurons, retinal ganglion cells (RGCs), optic nerves, the cerebellum, and CNS neurons that extend processes into the periphery, including cranial nerve motor nuclei. ⁴⁰ , ⁵⁴ The functional rationale for the transition to big tau from the CNS isoform of tau in specific neuronal populations remains undefined, as well as the pathological significance of the additional exon 4a in big tau. ⁵⁵

The big tau sequence in humans could result from the insertion of exon 4a in the tau transcript, ⁴⁵ , ⁴⁶ , ⁵⁰ however, proteomic evidence has not been established. In this study, we investigated the molecular compositions and anatomic distributions of human big tau isoforms across the nervous system. Here, we provide the first MS‐based evidence for the translation and expression of the unique tau insertion domain encoded by exon 4a within the N‐terminal projection region between exons 4 and 5, expanding the known repertoire of tau proteins expressed in the human nervous system. To define the distribution of big tau, we developed an immunoprecipitation‐MS (IP‐MS) strategy to sequence and quantify big tau‐specific peptides alongside canonical tau peptides across human nervous system tissues. Finally, to assess the relevance of big tau with aging and neurodegeneration, we quantified extracellular big tau peptides in CSF from individuals with and without AD. Together, these findings define the molecular diversity and anatomic distribution of big tau in humans and suggest that distinguishing big tau from canonical tau may enable improved biomarker strategies to differentiate CNS and PNS neurological diseases.

Materials and Methods

Human Tissue Samples and CSF

Frozen postmortem brain tissue samples from cases representing different stages of AD neuropathologic change (ADNC) were obtained from the Charles F. and Joanne Knight ADRC Neuropathology Core at Washington University School of Medicine. ⁵⁶ The studies involving postmortem human brain samples were approved by the Washington University Human Research Protection Office (HASD 201105103, ADRC – 201105102, and ACS – 201105305). The brain regions studied included cerebellum (CB), superior frontal gyrus (SFG), superior temporal gyrus (STG), occipital pole (Occ), and inferior parietal lobule (P). This study was approved by the Washington University in St. Louis Institutional Review Board. Brain, spinal cord, and sciatic nerve samples from neuropathologically confirmed sporadic amyotrophic lateral sclerosis (ALS) and disease controls were obtained from the Washington University ALS Postmortem Core. Postmortem nervous tissues across the spinal cord regions and PNS (cauda equina, DRG, sciatic nerve [SN], and brachial plexus [BP]) were obtained from 2 healthy young control donors and stored at −80°C. All participants provided consent for autopsy and research participation. The demographics of the frozen brain (Charles F. and Joanne Knight ADRC) cohort were previously described and shown in Table 1. ⁵⁶ The brain, spinal cord, and PNS sample demographics are also included in Table 2.

TABLE 1.

Demographics of the Human Brain Samples

	Tissue Region	Disease State	Age, yr	Sex, F/M	PMI, h	Total, N = 62
Brain	Frontal cortex (superior + motor)	Disease control a , b , c /normal control	56.5 ± 26.5	2/6	20 ± 6	8
		AD	83 ± 2	1/1	27 ± 20	2
		Aβ+	81 ± 9	2/0	14.15 ± 2.15	2
		sALS	71.2 ± 8.8	6/2	22.7 ± 12.8	8
	STG	Normal control	83.5 ± 3.5	2/0	27 ± 20	2
		AD	83 ± 2	1/1	14.15 ± 2.15	2
		Aβ+	81 ± 9	2/0	23.9 ± 9.4	2
	Parietal	Normal control	83.1 ± 6.4	4/5	47	9
		AD	81.3 ± 6.2	2/13	15.6 ± 7.4	15
		Aβ+	81.3 ± 6.8		17 ± 8.4
	Occipital	Normal control	83.5 ± 3.5	2/0	27 ± 20	2
		AD	83 ± 2	1/1	14.15 ± 2.15	2
		Aβ+	81 ± 9	2/0	23.9 ± 9.4	2
	Cerebellum	Normal control	83.5 ± 3.5	2/0	27 ± 20	2
		AD	83 ± 2	1/1	14.15 ± 2.15	2
		Aβ+	81 ± 9	2/0	23.9 ± 9.4	2

AD = Alzheimer's disease; Aβ+ = amyloid‐β positive; Aβ = amyloid‐β negative; PMI = postmortem interval; sALS = sporadic amyotrophic lateral sclerosis; STG = Superior temporal gyrus.

^a BMD, Becker muscular dystrophy.

^b CSM, Cervical spondylotic myelopathy.

^c PMI, postmortem interval.

TABLE 2.

Demographics of the Human Spinal Cord, Sciatic Nerve, Cauda Equina, Brachial Plexus, and Dorsal Root Ganglion Samples

	Tissue Region	Disease State	Age, yr	Sex, F/M	PMI, h	Total, N = 35
Spinal cord	Lumbar	Disease control a , b /Normal control	48.1 ± 26.9	1/2	26	3
		sALS	71.2 ± 8.8	3/1	22.7 ± 12.8	4
		Normal control	40 ± 2	2/0	26 ± 3	2
	Cervical	Disease control a , b , c /Normal control	56.5 ± 26.5	1/3	20 ± 6	4
		sALS	71.2 ± 8.8	3/1	22.7 ± 12.8	4
		Normal control	40 ± 2	2/0	26 ± 3	2
	Thoracic	Normal control	38	1/0	29	1
	Sacral	Normal control	38	1/0	29	1
Sciatic nerve	Nerve	Disease control c	81	0/1	14	1
		sALS	63.6 ± 3.3	1/4	40.1 ± 8.3	5
		Normal control	40 ± 2	2/0	26 ± 3	2
Cauda equina	Nerve	Normal control		2/0		2
Brachial plexus	Nerve	Normal control		2/0		2
Dorsal root ganglion	Nerve	Normal control		2/0		2

AD = Alzheimer's disease; Aβ = amyloid‐β negative; Aβ+ = amyloid‐β positive; sALS = sporadic amyotrophic lateral sclerosis; sIBM = Disease control‐ sporadic inclusion body myositis.

^a BMD, Becker muscular dystrophy.

^b CSM, Cervical spondylotic myelopathy.

^c PMI, postmortem interval.

CSF samples were collected from 90 participants who underwent human tau Stable Isotope Labeling Kinetics (SILK) protocol as described previously. ⁵⁷ The research with CSF samples was approved by the Human Studies Committee and the General Clinical Research Center (GCRC Tau SILK: 201502091). Clinical and cognitive assessments, including the Clinical Dementia Rating (CDR) and Mini‐Mental State Examination (MMSE) scores at baseline CSF collection for the participants, were used in this study. CSF amyloid‐β 42/40 ratio along with CDR scores were used for clinical group classification as follows: amyloid negative, cognitively unimpaired (CDR = 0); amyloid positive, cognitively unimpaired (CDR = 0, preclinical AD); and mild to moderate AD dementia (CDR = 0.5–2). Amyloid‐β 42/40 ratio was estimated using the IP‐MS assay and amyloid status was defined using amyloid‐β 42/40 ratio cutoff scores, as previously reported. ⁵⁸ The corresponding cutoff ratio (0.11) maximized the accuracy in predicting amyloid‐positivity as determined by Pittsburgh compound B (PiB) positron emission tomography (PET). Amyloid groups were further divided into clinical groups according to their CDR scores, as shown in Table 3.

TABLE 3.

Demographics and Biomarker Values of the Human CSF Study Participants

Variable	YNCs	CDR = 0		CDR >0
		Aβ– a	Aβ+ b	Aβ– c	Aβ+ d
N	25	25	14	4	21
Gender, F/M	13/11 (1 NA)	13/11 (1 NA)	5/9	3/1	10/11
Age, yr	42.45 ± 13.17	72.35 ± 4.68	73.14 ± 5.55	71.24 ± 1.75	76.92 ± 7.03
CDR e	NA	0	0	0.5	0.5–2
MMSE f	NA	29.25 ± 0.97	29.33 ± 0.78	28.75 ± 0.50	23.75 ± 5.29
CSF Amyloid‐β 42/40	0.13 ± 0.03 (1 NA)	0.14 ± 0.02 (1 NA)	0.09 ± 0.01	0.14 ± 0.01	0.07 ± 0.01
pT181/T181 (%)	6.89 ± 1.21	8.08 ± 2.23	8.13 ± 2.13	6.52 ± 2.90	9.61 ± 1.67
pT217/T217 (%)	1.55 ± 0.66	1.52 ± 0.53	3.11 ± 1.24	1.83 ± 0.36	6.37 ± 1.78
MTBR‐tau243 (ng/ml)	0.12 ± 0.05	0.18 ± 0.07	0.22 ± 0.05	0.12 ± 0.04	0.55 ± 0.28
T181 (ng/ml)	4.22 ± 1.74	5.77 ± 2.27	6.49 ± 1.38	3.90 ± 0.77	9.33 ± 2.54
T212 (ng/ml)	3.20 ± 1.43	4.31 ± 1.81	4.96 ± 1.20	2.77 ± 0.53	6.72 ± 2.05
Big tau‐235 (fmol/ml)	0.03 ± 0.03	0.07 ± 0.04	0.06 ± 0.03	0.06 ± 0.03	0.08 ± 0.06
Big tau‐294 (fmol/ml)	0.06 ± 0.06	0.10 ± 0.06	0.08 ± 0.06	0.09 ± 0.10	0.12 ± 0.09
Big tau‐429 (fmol/ml)	0.02 ± 0.02	0.04 ± 0.02	0.04 ± 0.02	0.05 ± 0.03	0.06 ± 0.04
Big tau‐452 (fmol/ml)	0.02 ± 0.02	0.04 ± 0.02	0.04 ± 0.02	0.04 ± 0.02	0.05 ± 0.03
Big tau‐545 (fmol/ml)	0.004 ± 0.005	0.006 ± 0.004	0.008 ± 0.007	0.003 ± 0.002	0.015 ± 0.011
Small tau45‐67 (0 N), (ng/ml)	1.62 ± 0.83	2.05 ± 0.92	2.51 ± 0.66	1.45 ± 0.56	3.41 ± 1.03
N‐term, tau6‐23 (ng/ml)	1.20 ± 0.57	1.65 ± 0.71	2.11 ± 0.40	1.05 ± 0.24	3.44 ± 1.02

Data presented as mean ± standard deviations (SD).

AD = Alzheimer's disease; CDR = Clinical Dementia Rating; CSF = cerebrospinal fluid; MMSE = Mini‐Mental State Examination; YNCs = young normal controls.

^a Age‐matched controls.

^b Preclinical‐AD.

^c Non‐AD cognitively impaired.

^d Mild to moderate‐AD.

^e Clinical Dementia Rating scores.

^f Mini Mental State Examinations; NA Data not available.

Brain Tissue Homogenization

Frozen brain tissue from the Charles F. and Joanne Knight ADRC was sliced using a cryostat at −20°C, weighed, collected in tubes, and stored at −80°C prior to biochemical analyses. Brain homogenization buffer (25 mM tris‐hydrochloride pH 7.4, 150 mM sodium chloride, 10 mM EDTA [ethylenediaminetetraacetic acid], 10 mM EGTA [ethylene glycol‐bis{β‐aminoethyl ether}‐N,N,N′,N′‐tetraacetic acid], phosphatase inhibitor cocktail [Sigma], and protease inhibitor cocktail [Roche]) was added to each tissue sample at a concentration of 0.1 mg/μl of brain tissue at 4°C, and tissue was sonicated (Fisherbrand Model 120 Sonic Dismembrator) for 15 seconds with 1 second on and 1 second off pulses using 35% intensity. Once the samples were homogenized, they were centrifuged at 11,000 g for 20 minutes at 4°C, to pellet cell/tissue debris. Then, 100 μl aliquots of supernatant were carefully transferred into new 0.6 ml Axygen tubes. The remaining supernatant and pellet were stored at −80°C.

Brain, Spinal Cord, and Peripheral Nerve Homogenization

Frozen brain, spinal cord, and sciatic nerve samples from the ALS postmortem core and young control donors were cryo‐sectioned, weighed, and collected in tubes. Cold brain homogenization buffer was added to the tissue samples at 3.25 μl buffer/mg tissue and sonicated (Fisherbrand Model 120 Sonic Dismembrator) for 15 seconds with 1 second on and 1 second off pulses using 35% intensity. After sonification, samples were centrifuged at 11,000 g for 20 minutes at 4°C (Eppendorf Centrifuge 5424R) to pellet debris. Supernatant was carefully removed and aliquoted into new tubes. Sarkosyl (N‐Lauroylsarcosine sodium salt; Sigma) was added to the supernatant to a final dilution of 1% sarkosyl and allowed to incubate for 60 minutes prior to ultracentrifugation at 100,000 g for 1 hour at 4°C (Beckman Optima TLX). The supernatant was aliquoted into new tubes and stored at −80°C prior to use.

Immunoblotting

Soluble lysate samples (20 μg) were mixed with 4X LDS sample buffer (Bio‐Rad) containing a final concentration of 10 mM of 1,4‐dithiothreitol (DTT, Pierce, A39255) and heated at 70°C for 10 minutes. Samples were loaded into each well and electrophoresed at 120 V for 90 minutes on 4 to 12% NuPAGE Bis‐Tris mini protein gel (Invitrogen, NP0322) and then transferred to 0.45 μm PVDF membrane (Immobilon, IPVH00010) and blocked for 1 hour at room temperature (RT) in 5% nonfat milk in phosphate buffer saline with 0.1% Tween20 (PBS‐T). All antibodies were diluted in PBS‐T containing 5% nonfat milk. Membranes were probed with primary monoclonal anti‐tau antibodies Tau‐1 (provided by Dr Nicholas Kanaan's laboratory from Michigan State University; 1:1000 dilution), Tau‐5 (Chemicon, MAB361, 1:1000 dilution), and loading controls monoclonal GAPDH (Invitrogen, MA5‐15738, 1:2000), anti β‐actin (Cell Signaling Technology, 13E5, 1:2000) overnight at 4°C. The membranes were washed in PBS‐T for 3 cycles (3 × 10 minutes) and incubated in goat anti‐mouse IgG‐HRP linked secondary antibody (Invitrogen, A28177, 1:2000), or goat anti‐rabbit IgG‐HRP linked secondary antibody (Cell Signaling Technology, 7,074, 1:2000) for 1 hour at RT and washed with PBS‐T. Immunoblot of cortical and cerebellar brain lysates were developed using West Pico PLUS chemiluminescent substrate (SuperSignal, Cat #34580) and imaged for 120 seconds. In contrast, immunoblots of peripheral nerve lysates were developed using West Atto ultimate sensitivity substrate (SuperSignal, Cat #A38554) and imaged at high exposure (120 seconds) to enhance visualization of low‐abundance HMW and LMW tau species expressed in the periphery using Bio‐Rad Chemidoc Imaging System.

SDS‐PAGE and In‐Gel Protein Digestion

Sarkosyl soluble homogenates (20 μg) from CNS and PNS were reconstituted with 4X LDS sample buffer (BioRad) containing 10 mM of reducing agent dithiothreitol (DTT). Each tissue sample was loaded into an individual well on a 4 to 12% NuPAGE Bis‐Tris mini protein gel. The gel was electrophoresed at 120 V for 90 minutes. After electrophoresis, the gel was stained with Oriole fluorescent gel stain (Bio‐Rad, Cat #1610496) for 30 minutes in the dark. Gel was imaged using Bio‐Rad Chemidoc Imaging System. Gel image was printed out and regions to cut were noted and numbered. Each gel band was cut using a clean scalpel (Med PRIDE) for in‐gel digestion using trypsin protease. The gel bands were de‐stained using 25 mM ammonium bicarbonate (NH₄HCO₃) in 50% Acetonitrile (ACN) solution and completely dried, reduced with 10 mM DTT at 60°C for 30 minutes and alkylated in the dark using 20 mM iodoacetamide (IAA, Pierce, A39271) solution at RT for 60 minutes. Gel bands were dehydrated in the vacuum concentrator and rehydrated using a 40 μl of 10 ng/μl trypsin solution (25 mM TEABC pH 8.0) on ice for 10 minutes. The gel bands were digested for 18 hours at 37°C and the supernatant was desalted on Oasis μElution HLB plates (Waters).

Soluble Tau Quantitation in Human Tissues and CSF

Total protein concentrations of human tissue homogenates were determined using the bicinchoninic acid (BCA) assay method. ⁵⁹ For the soluble tau analysis, tau was immunoprecipitated using Tau‐1, HJ8.5, and HJ8.7 antibody mixture, as previously described with adjustments. ¹⁴ , ⁵⁶ To each soluble brain supernatant (50 μg total protein), 0.01% HSA solution was added along with 0.625 ng of ¹⁵N‐2N4R (0.15 ng/μl, a gift from Dr Guy Lippens, France) and 1.25 ng of ¹⁵N‐0N3R (0.15 ng/μl, Promise Proteomics, Grenoble, France) to a final volume of 500 μL. Tau was immunoprecipitated from 0.5 ml CSF. Soluble tau from each sample (brain or CSF) was immunoprecipitated in detergent (1% NP‐40), chaotropic reagent (5 mM guanidine), and protease inhibitors (Roche), using the antibody cocktail (50% slurry of tau antibody conjugated sepharose (cyanogen bromide activated) beads containing 3 μg antibody/mg beads) by incubating them with rotation for 2 hours at RT. Following incubation, the samples were washed 3 times with 25 mM TEABC (Triethylammonium bicarbonate, Sigma‐Aldrich, St. Louis, MO) buffer. The immobilized proteins were digested on beads for 18 hours using 0.4 μg trypsin (Promega) at 37°C. Then, 50 fmol each of AQUA internal standard peptide for unmodified tau peptides, 5 fmol of phospho‐tau AQUA internal standard, and 25 fmol of tau exon 4a and exon 6 AQUA internal standard peptides were spiked in the digested samples for their quantification. Soluble tau digests were desalted using the Oasis μElution HLB plate (Waters) according to the manufacturer's protocol. The eluent was lyophilized and reconstituted with 25 μl of 2% ACN, 0.1% FA in water prior to MS analysis on Vanquish Neo UHPLC (Thermo Fisher Scientific, USA) coupled to Orbitrap Exploris 480 (Thermo Fisher Scientific, USA). Eighteen common tau peptides were quantified by comparison with the corresponding isotopomer signals from the ¹⁵N internal standard using Skyline software (version 20.2 MacCoss Lab, Department of Genome Sciences, University of Washington). The exon 4a‐L and exon 6 peptides were quantified by comparison with the corresponding isotopomer signals from the AQUA peptides. Relative quantification for common tau peptides, exon 4a‐L, and exon 6 peptides were calculated by taking the “absolute” amount for each peptide and dividing by the “absolute” amount of the reference peptide TPSLPTPPTR (referred to in this manuscript as the “total‐tau peptide”). The protein profile results are plotted as a function of peptide amino acid start position‐end position and each point represents the mean for the given group, with error bars indicating the standard error of mean (SEM).

To validate the endogenous junction peptides for exon 4 to exon 4a‐L (Table 4 and Supplementary Table S1) and exon 4 to exon 4a‐S, custom synthesized AQUA peptides for 0 N‐4a‐L (AEEAGIGDTPSLEDEAAGHVTQEELRVPGRQR), 1 N‐4a‐L (STPTAEAEEAGIGDTPSLEDEAAGHVTQEELRVPGRQR), 2 N‐4a‐L (QAAAQPHTEIPEGTTAEEAGIGDTPSLEDEAAGHVTQEELRVPGRQR), and 0 N‐4a‐S (AEEAGIGDTPSLEDEAAGHVTQEPESGKVVQEGFLR), 1 N‐4a‐S (STPTAEAEEAGIGDTPSLEDEAAGHVTQEPESGKVVQEGFLR) and 2 N‐4a‐S (QAAAQPHTEIPEGTTAEEAGIGDTPSLEDEAAGHVTQEPESGKVVQEGFLR) were purchased from Life Science Technologies (Thermo Fischer Scientific). Then, 50 ng of each of these AQUA peptides (concentrated stock 10 ng/μl prepared in 1% ACN, 0.1% FA) were spiked into 0.1% HSA solution along with 5 ng of ¹⁵N‐2N4R and 5 ng of ¹⁵N‐0N3R to a final volume of 500 μl. Tau immunoprecipitation followed by tryptic digestion and LC‐MS was performed as described above. The LC‐MS traces of the synthetic trypsin digested peptides (Supplementary Table S2) were compared with those of their respective endogenous tau junction peptides derived from human tissue lysates.

TABLE 4.

The List of Exon 4a‐L, Big Tau Variant Isoforms Tryptic Peptides Quantified in This Study by Using Human Nervous Tissues

Sr. No#	Peptide Sequence	Residue	Region	Exon(s)
1	QEFEVMEDHAGTYGLGDR	6–23	N‐terminus	Exon 1
2	DQGGYTMHQDQEGDTDAGLK	24–44	N‐terminus	Exon 1
3	ESPLQTPTEDGSEEPGSETSDAK	45–67	N‐terminus	Exon 2
4	STPTAEDVTAPLVDEGAPGK	68–87	N‐terminus	Exon 2–3
5	AEEAGIGDTPSLEDEAAGHVTQAR	45–68	0 N (small tau)	Exon 4–5
6	STPTAEAEEAGIGDTPSLEDEAAGHVTQAR	68–97	1 N (small tau)	Exon 2–4‐5
7	QAAAQPHTEIPEGTTAEEAGIGDTPSLEDEAAGHVTQAR	88–126	2 N (small tau)	Exon 3–4‐5
8	AEEAGIGDTPSLEDEAAGHVTQEELR	45–70	Big tau 2 N‐4a‐L	Exon 4‐4aL
9	STPTAEAEEAGIGDTPSLEDEAAGHVTQEELR	68–99	Big tau 1 N‐4a‐L	Exon 2–4‐4aL
10	QAAAQPHTEIPEGTTAEEAGIGDTPSLEDEAAGHVTQEELR	88–128	Big tau 0 N‐4a‐L	Exon 3–4‐4aL
11	VVQEGFLR	235–242	Big tau	Exon 4a
12	HQLLGDLHQEGPPLK	294–308	Big tau	Exon 4a
13	VSTEIPASEPDGPSVGR	379–395	Big tau	Exon 4a
14	AAFPGAPGEGPEAR	429–442	Big tau	Exon 4a
15	EADLPEPSEK	452–461	Big tau	Exon 4a
16	YVSSVTSR	545–552	Exon 6	Exon 6
17	IATPR	572–576	Mid‐domain (PRR)	Exon 7
18	TPPSSGEPPK	602–611	Mid‐domain (PRR)	Exon 7
19	SGYSSPGSPGTPGSR	616–630	Mid‐domain (PRR)	Exon 7
20	TPSLPTPPTR	633–642	Mid‐domain (PRR)	Exon 7
21	VAVVR	647–651	Mid‐domain (PRR)	Exon 7
22	LQTAPVPMPDLK	664–675	MTBR	Exon 9
23	IGSTENLK	681–688	MTBR	Exon 9
24	VQIINK	696–701	4R, MTBR	Exon 10
25	LDLSNVQSK	703–711	4R, MTBR	Exon 10
26	HVPGGGSVQIVYKPVDLSK	720–738	4R, MTBR	Exon 10–11
27	VQIVYKPVDLSK	727–738	3R, MTBR	Exon 11
28	IGSLDNITHVPGGGNK	775–790	MTBR	Exon 12
29	TDHGAEIVYK	807–816	C‐terminus	Exon 13

MTBR = Microtubule binding region; PRR = proline rich region.

CSF tau MTBR containing the residue 243 (MTBR‐tau243) peptides were monitored using the previously reported procedures with minor modifications. ¹⁵ Briefly, 0.45 ml of post‐IP (for big tau species described in the previous section) CSF was thawed and 5 μl of ¹³C¹⁵N 2N4R‐tau internal standard (250 pg/μl) spiked into the post‐IP CSF. Then, the tau species containing the residue 243 were immunoprecipitated with HJ32.11 antibody. ¹⁴ The immunoprecipitated tau species were digested by trypsin and desalted by the Oasis μElution HLB plate (Waters) according to the manufacturer's protocol. The samples were reconstituted with 25 μl of 3% ACN, 3% FA in water, and peptides were separated by nanoAcquity ultra‐performance LC system (Waters) which was coupled to Orbitrap Tribrid Eclipse (Thermo Fischer Scientific, USA) MS operating in parallel reaction monitoring (PRM) mode. The resulting MTBR‐tau243 peptide (residue 243–254) and the corresponding isotopomer from the ¹³C¹⁵N internal standard were monitored and quantified similarly as other tau species including big‐tau as described above.

Mass Spectrometry

The peptides were directly loaded onto HSS T3 75 μm × 100 mm, 1.8 μm C18 column (Waters) heated to 65°C using a Vanquish Neo UHPLC (Thermo Fisher Scientific, San Jose, CA) and peptides were separated using a flow rate of 0.4 μl/min with a mixture of buffer A (0.1% FA in water) and buffer B (0.1% FA in acetonitrile). Tau peptides were eluted from the column with a gradient of 4 to 8% of buffer B for 10 minutes, then, 8 to 20% of buffer B for another 8 minutes before ramping up to 95% buffer B in the next 2 minutes and cleaning the column for another 2 minutes. The Thermo Orbitrap Exploris 480 was equipped with a Nanospray Flex electrospray ion source (Thermo Fisher Scientific, San Jose, CA) and operated in positive ion mode. Peptide ions sprayed from a 10 μm SilicaTip emitter (New Objective, Woburn, MA) into the ion source (spray voltage = 2.2 kV), were targeted and isolated in the quadrupole. Isolated ions were fragmented by high‐energy collisional dissociation (HCD), and ion fragments were detected in the Orbitrap (resolution of 15,000 for tau, 30,000 for phospho‐tau, and 120,000 for big tau peptides, mass range = 150–1,500 m/z).

LC‐MS/MS for Bottom‐Up Proteomics

Peptides were loaded directly using HSS T3 75 μm × 100 mm, 1.8 μm C18 column (Waters) heated to 65°C using a Vanquish Neo UHPLC (Thermo Fisher Scientific, San Jose, CA) and peptides were separated using a flow rate of 0.4 μl/min with a mixture of buffer A (0.1% FA in water) and buffer B (0.1% FA in acetonitrile). Peptides were eluted from the column with a stepped gradient of 52 minutes starting from 0.5% buffer B to 15% buffer B, 35 minutes; 30% buffer B, 52.5 minutes. The column was washed with 80% buffer B for 3 minutes before equilibration at 0.5% buffer B for 5 minutes. The proteolytic peptides were sprayed into MS inlet at 2.2 kV in positive ion mode using commercial PicoTip emitters (New Objective). A full mass spectrum scan with a resolution of 120,000 @ m/z 400 was acquired in the mass range m/z 350 to 1,500 (AGC target 2 × 10, ⁶ maximum injection time 54 ms). Peptides were selected for fragmentation using HCD using the monoisotopic precursor selection (MIPS) mode. MS2 scan settings were the following: resolution = 22,500, AGC target = 5 × 10, ⁵ maximum injection time = 22 ms, isolation window = 1.2 m/z, normalized collision energy 30%, filter intensity (intensity threshold 5e⁴), charge states 2–5 selected, undetermined charge states filter on, exclude isotopes on, and dynamic exclusion 30 seconds.

Raw data were analyzed in Fragpipe using human proteome database downloaded in January 2024 from UniProt (only reviewed entries and appended with in‐house fasta containing manually curated tau exon‐4a‐L and exon 4a‐S sequences). Currently, only exon 4a‐S containing big tau sequence is present in UniProt – P10636‐1. We created the in‐house fasta along with P10636‐1 (tau with exon 4a‐S, NCBI Ref seq ID: NP_058519) along with the sequences curated from NCBI protein database (XP_005257423, XP_001364194, and XP_005257419) corresponding to 3 isoforms of tau containing exon 4a‐L. The raw data were searched with this curated protein database using MSFragger database search engine. ⁶⁰ The following settings were applied: trypsin (specificity set as C‐terminal to arginine and lysine) with up to 3 missed cleavages, the mass tolerance was set to ±20 ppm for the precursor and fragment mass tolerance. Fixed modifications included Cys carbamidomethylation (CAM, mass shift +57.0215 Da) with variable modifications including Ser/Thr/Tyr phosphorylation (+79.9663 Da), Gln and Asn deamidation (+0.98 Da) and Met oxidation (+15.9949 Da). False discovery rate (FDR) was set to 1% on peptide and protein levels (PeptideProphet and Philosopher) ⁶¹ with a minimum length of 5 amino acids and was determined by searching a decoy reverse database. For all other search parameters, default settings were used. Similar parameters were used for the trypsin + Asp‐N and tryspin + Glu‐C digested samples with the setting for the corresponding dual enzyme setting. Label‐free quantification was done using the in‐built label‐free quantification (IonQunat) algorithm integrated in MSFragger search. ⁶²

MAPT Targeted IsoSeq

Long read RNA sequencing on MAPT (IsoSeq) was performed as previously described. ⁶³ Briefly, we used a modified single molecule real time bell‐shaped template (SMRTbell) amplicon protocol at the Center for Advanced Genomic Technologies, Mount Sinai. The cDNA synthesis was performed from 100 ng RNA per sample with SuperScript IV First‐Strand Synthesis System (ThermoFisher Scientific, Cat. No. 18091200) with oligo(dT) primers. MAPT‐specific polymerase chain reaction (PCR) amplification used TaKaRa LA Taq DNA Polymerase with GC Buffer (Clontech, Cat. No. RR02AG) and gene‐specific primers MAPT_NF_F1 (5′–ATG GAA GAT CAC GCT GGG AC–3′) and MAPT_NF_R2 (5′–GAG GCA GAC ACC TCG TCA G–3′). Amplicons were purified using AMPure PB beads (PacBio) follow end‐repair, A‐tailing, and ligation of barcoded SMRTbell adapters using the SMRTbell Express Template Prep Kit 3.0 (PacBio). After cleanup, final libraries were sequenced on the PacBio Sequel II platform using circular consensus sequencing (CCS) to generate high‐fidelity (HiFi) reads. Demultiplexing and adapter trimming were performed with lima, and artificial concatemers were removed. Isoform clustering and polishing were done using the Iso‐Seq tool suite, and alignments were performed against GL000258.2 and KI270908.1 references using pbmm2 with the ISOSEQ preset. Redundant isoforms were collapsed and classified using pigeon, referencing GENCODE version 39 annotations. Low‐confidence and single‐exon transcripts were excluded. The resulting isoforms were converted to BED format for visualization and cross‐referenced against an internal isoform atlas to maintain consistent naming. Last, a computational pipeline was used to process the BED files, naming transcripts based on presence of known exons, and plot them to evaluate 4a/4aL expression and isoform composition across brain regions.

Data and Statistical Analysis

Data are represented as mean ± standard deviation (SD) unless mentioned otherwise and were plotted in GraphPad Prism (version 10.5). One‐way analysis of variance (ANOVA) followed by post‐analyses (Tukey's or Dunnett's t tests as appropriate) were performed to compare between multiple brain regions and CSF tau levels. Significance was evaluated at the 0.01 and 0.05 levels. Spearman correlations were used to assess the associations among CSF tau biomarkers, age, and exon 4a‐L and exon 6 peptides. Non‐parametric tests were used for non‐normally distributed data. Spearman correlation was used for continuous variables. Diagnostic performances were evaluated with receiver operating curves (ROC) and area under the curve (AUC) assessments.

Results

Tau Exon 4a‐Long Translation Leads to Insertion of 355 Amino Acids in Human “Big Tau”

Tau transcript could be extended by the insertion of exon 4a (between exons 4 and 5) along with inclusion and exclusion of exon 6 (Fig 1A). Big tau is distinguished from canonical small tau isoforms based on the inclusion and translation of exon 4a and exon 6. Alternative 3′ splicing of tau exon 4a leads to a larger splice variant (exon 4a‐L) in human cancer cells and skeletal muscle tissue (Fig 1B). ⁴⁵ , ⁴⁶ However, 2 different splice variants of exon 4a have been annotated, UniProt ID:P10636‐1 for exon 4a, referred to as exon 4a‐short (exon 4a‐S) and NCBI ID: XP_005257419 for exon 4a‐long (exon 4a‐L; Fig 1A–C). ⁵⁰ These splice variants could theoretically lead to insertion of either 251 (exon 4a‐S) or 355 (exon 4a‐L) amino acids in the N‐terminal projection domain of tau (see Fig 1A,C) in humans, respectively. ⁴⁶ Thus, to determine which big tau protein isoforms are expressed in adult human nervous tissue (Supplementary Fig S1a), we defined their amino acid sequences using MS.

FIGURE 1.

Big tau – inclusion of the exon 4a‐long and exon 6 leads to the longest MAPT isoform sequence. (A) Schematic representation of the full‐length human tau isoforms with and without exon 4a. Insertion of exon 4a‐L and exon 6 result in “big tau” that is 862 amino acids long (top). The insertion of exon 4a‐S and exon 6 will result in big tau isoform with a 758 amino acid sequence (second from top). Big tau isoform can also result from the insertion of exon 4a‐L without exon 6 (third from top). Canonical small tau (2N4R) lacks both the exon 4a and exon 6. Monoclonal anti‐tau antibodies targeting, N‐terminal and proline rich region (HJ8.5, HJ8.7, Tau‐1, and Tau‐5) are depicted on the sequence. (C) Human MAPT genomic sequence highlighting 5′ splice site of the exon 4, BPS, polypyrimidine rich sequence, canonical 3′ splice site, and alternative 3′ splice site within exon 4a‐L (black star), indicating regulatory elements that promote exon 4a‐L insert. (D) The amino acid sequence of the longest human big tau proteins with either exon 4a‐L or exon 4a‐S aligned with the longest common tau isoform found in the adult human brain (2N4R). The N‐terminus of exon 4a‐L is highlighted by the green arrow, whereas the N‐terminus of the exon 4a‐S is depicted by the vertical black line on the amino acid sequence. Tryptic peptide for exon 4 to exon 4a‐L junction peptide (2 N‐4a‐L peptide) is highlighted in the second line. Asterix highlights the exon 4a‐L tryptic peptide (big tau 4a‐L*) corresponding to alternative 3′ splicing. BPS = branch point sequence. [Color figure can be viewed at www.annalsofneurology.org]

Using a combination of antibodies targeting tau N‐terminus (HJ8.5 and HJ8.7) and mid‐domain (Tau‐1) specific epitopes, we immunoprecipitated tau from the tissue lysates originating from CNS and PNS. To determine which big tau isoforms are expressed in adult human nervous tissue (see Supplementary Fig S1a), we defined their amino acid sequences using MS. We performed bottom‐up tandem MS (MS/MS) on pooled homogenates from adult human brain, spinal cord, and sciatic nerve, respectively. Proteins were digested into peptides and analyzed by liquid chromatography–tandem MS (LC–MS/MS). Peptide spectra were searched against a customized protein database containing tau sequences incorporating both exon 4a‐short (4a‐S) and exon 4a‐long (4a‐L) variants. This strategy enabled the identification of exon 4a‐specific peptides and exon‐bridging peptides unique to respective big tau isoforms. Detection of these peptides provided direct protein‐level evidence for the translation of exon 4a‐containing tau isoforms in human nervous tissue and allowed us to distinguish between the predicted 4a‐L and 4a‐S splice products (Fig 2A). We observed the presence of tryptic peptides specific to the translation of exon 4a in the brain, spinal cord, and sciatic nerve, respectively (Supplementary Fig S1B). Multi‐enzyme digestion (see Supplementary Figs S1B and S1C) further validated exon 4a expression at the protein level. The insertion of both the exons 4a‐L and 6 would result in an 862 amino acid long protein sequence of tau, making it the longest isoform of human tau — “big tau” (see Fig 1C). The HCD‐MS/MS spectrum of the tryptic big tau peptide derived from the union of exon 4 and exon 4a‐L (big tau 2 N‐4a‐L; Fig 2B), [M + 4H]⁴⁺ m/z 1064.7520 with the b (b ₂‐b ₂₀) and y ion series (y ₁‐y ₂₆) validates the peptide amino acid sequence “QAAAQPHTEIPEGTTAEEAGIGDTPSLEDEAAGHVTQEELR” (see Fig 2B). This tryptic peptide is different from the sequence “QAAAQPHTEIPEGTTAEEAGIGDTPSLEDEAAGHVTQAR” found in the common small tau isoforms (exon 4‐exon 5 junction) without exon 4a‐L insert (see Fig 2A and Supplementary Fig S2A–C). We found another peptide with “GPAFPKPATTAYLHTEPESGK” (Fig 2C) sequence that is specific to exon 4a‐L. The human exon 4a‐L has been described as the exon 4a‐S plus additional coding sequence at the 5′ end of exon 4a‐S (see Fig 2A). ⁴⁵ , ⁴⁶ Identification of the “GPAFPKPATTAYLHTEPESGK” peptide (see Fig 2C) provides crucial evidence of the alternative 3′ splice junction within the exon 4a‐L. ³⁹

FIGURE 2.

Exon bridging peptides for identification of big tau isoforms in humans (A) Schematic representation of the MAPT exon structure (exons 1–5) illustrating isoforms with and without exon 4a, together with the corresponding amino acid sequences of tryptic bridging peptides. Tryptic bridging peptides from 2 N, 1 N, and 0 N big tau isoforms generated by insertion of exon 4a‐L are distinct from those arising from exon 4a‐S insertion. For reference, the canonical small tau 2 N, 1 N, and 0 N peptides are derived from exon 4 to exon 5 junctions. (B) Representative HCD‐MS/MS spectrum of the unique big tau 2 N‐4a‐L bridging tryptic peptide [M + 4H]⁴⁺ m/z 1064.7520. The inset shows the XIC depicting the distribution of top 4 fragment ions from the MS/MS spectra. (C) Representative HCD‐MS/MS spectrum of big tau 4a‐L* specific tryptic peptide [M + 4H]⁴⁺ m/z 550.5325 corresponding to alternative 3′ splicing (right). The y and b ions are shown in red and blue, respectively. The inset shows the XIC depicting the distribution of top 3 fragment ions from the MS/MS spectra. (D) Representative mirror plots for XIC for 3 endogenous big tau exon 4a‐L bridging peptides (top) following trypsin digestion showing complete match with their respective synthetic standard peptides (bottom) on the LC–MS. The y ₂₀ ²⁺ ion (most abundant fragment ion) was used for comparison between respective endogenous and standard peptides. The peptide sequence for the three exon 4a‐L isoforms (0 N‐, 1 N‐, and 2 N‐4a‐L) that get generated following trypsin digestion are depicted above the XIC plots. (E) Representative mirror plots for XIC for the endogenous big tau exon 4a‐S bridging peptides following trypsin digestion showing complete match with their respective synthetic standard peptides (bottom) for 2 N, 1 N and 0 N isoforms on the LC–MS. The y ₂₂ ²⁺ ion (most abundant fragment ion) was used for comparison between respective endogenous and standard peptides. The peptide sequence for the three exon 4a‐S isoforms (0 N‐, 1 N‐, 2 N‐4a‐S) that get generated following trypsin digestion are depicted above the XIC plots. BP tissue was used for (C) and (D) for generating these mirror plots. (F) Relative distributions (%) of respective big tau isoforms (exon 4a‐L + exon 4a‐S) quantified in CNS (parietal, temporal, occipital, frontal, cerebellum, and spinal cord, each n = 6), and PNS (DRG, n = 2; BP, n = 2 and SN, n = 5) tissues. (G) Representative immunoblot analysis of tau protein isoforms across CNS and PNS tissues in humans, using Tau‐1 antibody. Big tau (HMW tau approximately 115–90 kDa) and small tau (LMW tau approximately 65–40 kDa) protein expression were detected in the PNS (SN, DRG, and BP) and spinal cord tissues. No discernable HMW tau bands were detected across brain regions (parietal, temporal, occipital, frontal, and cerebellum). GAPDH (35 kDa) was used as loading control. BP = brachial plexus; CNS = central nervous system; DRG = dorsal root ganglion; GAPDH = glyceraldehyde 3‐phosphate dehydrogenase; HCD‐MS/MS = high‐energy collisional dissociation‐tandem mass spectrometry; HMW = high molecular weight; LC‐MS = liquid chromatography mass spectrometry; LMW = low molecular weight; MS/MS = tandem mass spectrometry; PNS = peripheral nervous system; SN = sciatic nerve; XIC = extracted ion chromatograms. [Color figure can be viewed at www.annalsofneurology.org]

In canonical small tau, the alternative splicing of exons 2 and 3 with exon 4 predominantly generates 0 N and 1 N isoforms, with less abundant 2 N isoform found in the brain (see Supplementary Fig S2A–C). ²⁷ , ⁶⁴ This led us to investigate whether alternative splicing of exons 2 and 3, together with the expression of exon 4a‐L or exon 4a‐S, could generate additional isoforms akin to small tau isoforms (see Fig 2A). Indeed, along with the big tau 2 N‐4a‐L (see Fig 2B) peptide, we identified specific peptides corresponding to 0 N‐4a‐L, 1 N‐4a‐L isoforms (Supplementary Fig S2D, S2E,S2G–I). Synthetic standard peptides (see Supplementary Table S2) for the corresponding bridging peptides for big tau isoforms (4a‐L and exon 4a‐S) were used for orthogonal validation (Supplementary Fig S3). Tau immunoprecipitation followed by tryptic digestion of the synthetic big tau 2 N‐4a‐L, 1 N‐4a‐L, and 0 N‐4a‐L peptides matched the corresponding endogenous big tau peptides by LC–MS (Fig 2D), confirming the identity of big tau isoforms in human samples. Comparison of the LC–MS analysis of the synthetic big tau 4a‐S peptides confirmed that the corresponding 2 N‐4a‐S, 1 N‐4a‐S, and 0 N‐4a‐S endogenous peptides (Fig 2E and Supplementary Fig S3) can be also detected in human nervous tissues, albeit at low levels. We observed across human CNS (parietal, temporal, frontal, occipital, cerebellum, and spinal cord) and PNS (SN, BP, and DRG), the relative abundance of big tau 4a‐L isoforms (0 N‐4aL + 1 N‐4a‐L + 2 N‐4a‐L) greatly exceeded that of 4a‐S isoforms (0 N‐4a‐S + 1 N‐4a‐S; Fig 2F and Supplementary Fig S3B). We estimated big tau exon 4a‐L isoforms (>99% in CNS brain regions and >97% in CNS, spinal cord, and PNS, SN, BP, >96.5% in PNS, and DRG) > > big tau exon 4a‐S isoforms (<1% in CNS brain regions and <3.5% in CNS, spinal cord, and PNS regions, SN, BP, and DRG; see Fig 2F). Together, these results demonstrate that big tau in the human nervous system is generated predominantly through insertion of exon 4a‐L (>95%) between exons 4 and 5, whereas exon 4a‐S insertion (<5%) provides a minor contribution to big tau isoforms (see Fig 2F and Supplementary Fig S3). ⁵⁰ , ⁵⁵ Our observation aligns with what has been previously reported in human cancer lines and skeletal muscles tissues. ⁴⁶ Along with exon 4a‐L (in the rest of the paper, we have focused on these big tau isoforms) expression across the nervous tissue, we also observed exon 6 peptides across CNS and PNS tissue lysates (see Supplementary Figs S2C and S2F).

Big tau protein, twice as large compared to small tau, has been predicted to have molecular weight of approximately 110 kDa. ⁵² , ⁵⁵ , ⁶⁵ Immunoblotting with tau antibodies targeting the proline rich region (Tau‐1 and Tau‐5; Fig 2G and Supplementary Fig S4) indicated the existence of high molecular weight (HMW; tau, approximately 115–90 kDa) tau bands, consistent with big tau, along with the common low molecular weight (LMW; tau, approximately 40–65 kDa) tau species. These HMW bands were detected in the human PNS tissues (SN, BP, and DRG) and selected CNS region (spinal cord tissues). Notably, no clear discernable HMW bands (approximately 115–90 kDa) were observed in select brain tissues (frontal, temporal, parietal, occipital, and cerebellum) using immunoblotting (see Fig 2G and Supplementary Fig S4). To determine whether the HMW tau species detected by immunoblotting contain exon 4a‐L, we performed SDS–PAGE followed by in‐gel digestion and LC‐MS on nervous tissue lysates from the CNS and PNS (Supplementary Fig S5). Peptides corresponding to big tau exon 4a‐L (2 N‐4a‐L, 1 N‐4a‐L, and 0 N‐4a‐L) were detected in the HMW (75–185 kDa) gel bands from both the CNS and PNS samples. Big tau exon 4a‐L peptides were abundant in the HMW bands in the spinal cord, BP, and DRG (see Supplementary Fig S5). Relatively lower levels of big tau exon 4a‐L isoforms were also observed in the brain HMW gel band in the in‐gel digestion LC‐MS (see Supplementary Fig S5C). We also detected 3R and 4R specific peptides in the HMW bands from the CNS and PNS samples, demonstrating that big tau contains MTBR along with exon 4a‐L (see Supplementary Fig S5).

To complement the protein‐level observations and determine whether big tau transcripts are expressed in the adult human brain, we performed long‐read IsoSeq ⁶³ sequencing on caudate (n = 5) and frontal cortex (middle frontal gyrus, n = 5) from 10 different control brain donors (Supplementary Table S3). Exon 4a‐L transcripts were detected at low levels (Supplementary Fig S6), whereas exon 4a‐S transcripts were not detected. The exon 4a‐L transcripts detected in these 2 brains regions were 1 N‐4a‐L‐3R, 1 N‐4a‐L‐4R, and 0 N‐4a‐L‐3R (see Supplementary Fig S6A). Similar to the proteomic results, big tau transcripts comprised approximately 0.5 to 1% of total tau detected in the 2 brains regions investigated (see Supplementary Fig S6B). Together, our discovery proteomics, in‐gel digestion with LC‐MS, and IsoSeq findings from select brain, spinal cord, and PNS tissues highlight distinct distribution pattern of HMW tau isoform in the human CNS and PNS.

Expression Pattern of Tau Exon 4a‐L and 6 Peptides Across CNS and PNS

Building on our discovery proteomics findings, we developed a targeted immunoprecipitation‐parallel reaction monitoring (IP‐PRM) MS assay to quantify the distribution and stoichiometry of big tau isoforms across the human nervous system. To quantitatively measure big tau peptides alongside other tau peptides, we refined and applied the IP strategy with parallel reaction monitoring (PRM)‐MS technique. ⁵⁶ , ⁵⁷ Using a combination of antibodies targeting tau N‐terminus (HJ8.5 and HJ8.7) and mid‐domain (Tau 1) epitopes (Fig 3A) shared by all tau isoforms (used in the discovery experiments), we immunoprecipitated endogenous tau and quantitatively measured the profiles of tau isoforms using the IP‐PRM assay. ³⁶ We quantified more than 29 peptides across the tau protein sequence, which included peptides that are common to tau isoforms (18 peptides), 3 peptides specific to isoforms lacking exon 4a‐L translation (small tau isoforms with N‐terminal inserts: 0 N, 1 N and 2 N), 3 big tau peptides that are specific to exon4‐exon4a‐L (0 N‐, 1 N‐, and 2 N‐4a‐L big tau isoforms), 5 peptides within the exon 4a‐L itself, and 1 peptide from exon 6 (see Table 4).

FIGURE 3.

Differential expression of big tau across the human nervous system. (A) Schematic of the big tau protein sequence indicating the N‐terminal insert, proline rich domains and microtubule binding repeats. (B) Quantitation of relative tau peptides including the exons 4a‐L and 6 peptides from peripheral tissues, DRG (n = 2, diamond), BP (n = 2, triangle), and SN (n = 6, circle). CNS, spinal cord (n = 15, square), and the brain (n = 39, inverted triangle) tissue lysates. The tau peptides were normalized to total tau (residue 633–642 in reference to 2 N big tau isoform). Big tau (exon 4a‐L, 6), 4R and 3R specific peptides are highlighted. Data are presented as mean ± SD. (C) Bar graphs represent the total tau levels (ng/μg protein) from all tau isoforms (big tau, small tau) were quantified in the brain, spinal cord, SN, BP, and DRG samples used in the cohort in (B). Data were represented as mean ± SD. Statistical significance was performed using Brown‐Forsythe and Welch ANOVA with post hoc multiple comparisons. ****p < 0.0001. (D) Bar graphs of the 3 small tau isoforms (0 N, 1 N, and 2 N) normalized to t‐tau for all the samples used in B. Data are presented as mean ± SD. Statistical analysis was performed using 2‐way ANOVA with multiple comparisons, *p ≤ 0.05. (E) Bar graphs of 3 big tau isoforms (0 N, 1 N, and 2 N big tau) normalized to t‐tau for all the samples used in B. Data are presented as mean ± SD, statistical analysis was performed using 2‐way ANOVA followed by multiple comparisons, ****p < 0.0001. (F) Bar graph depicting the distribution of small tau and big tau N‐terminal isoforms (0 N + 1 N + 2 N) across the nervous system for all the samples used in B. Data are expressed as mean ± SD. Statistical analysis was performed using 2‐way ANOVA with Tukey's multiple comparisons test, ****p < 0.0001. (G) Schematic diagram illustrating the relatively big tau expression pattern across the human nervous system. ANOVA = analysis of variance; BP = brachial plexus; CNS = central nervous system; DRG = dorsal root ganglion; SN = sciatic nerve. [Color figure can be viewed at www.annalsofneurology.org]

Typically, the shorter isoform of tau (small tau) has been referred as being CNS or brain‐specific whereas big tau (exon 4a‐L) is exclusively found in the PNS and periphery. ¹⁶ To test the hypothesis that exons 4a‐L and 6 are differentially expressed across the nervous system, with higher expression in the PNS compared to the CNS, ³⁹ , ⁴⁰ , ⁴⁷ we quantified the common tau isoform peptides and big tau peptides in soluble tau extracted from 6 brain regions including parietal, temporal, occipital, frontal, and cerebellum (CNS), spinal cord (CNS), and SN, BP, and DRG (PNS) samples (Fig 3B and Supplementary Fig S5, Tables 1 and 2). The quantitation of soluble tau peptides from the brain exhibited a profile expected for small tau isoforms, as reported previously. ⁵⁷ We observed the exon 4a‐L peptides in the brains, normalized to the mid‐domain total‐tau peptide chosen as reference (residue 633–642 according to big tau sequence; see Figs 1C and 3B) or residue 212 to 221 according to small tau sequence (Fig 1D), were 200‐fold lower than other common small tau peptides (see Fig 3B, inverted triangle). The levels of exon 4a‐L peptides (residue 235–242, 0.006 ± 0.005; residue 294–308, 0.003 ± 0.003; residue 429–442, 0.003 ± 0.002 compared to total‐tau 633–642, n = 39) are 2‐orders of magnitude lower compared with other common tau peptides (residue 6–23, 1.379 ± 0.247; residue 45–67, 1.298 ± 0.362, residue 602–611, 1.019 ± 0.101 compared to total‐tau) in the human brains (see Fig 3B, inverted triangle and Supplementary Fig S5). These results indicate that the longest isoform of tau, “big tau” with exon 4a‐L in the brain globally combining all the brain regions studied is present at lower proportion (approximately 0.3%) compared to small tau (see Figs 3B and 3F). The exon 6 peptide (residue 545–552, 0.0005 ± 0.0033 compared to total‐tau, n = 39) indicates an even lower contribution to the total tau pool in the brain (less than 0.05% of the total tau; see Fig 3B).

Compared with the brain extracts, we observed 20‐fold more exon 4a‐L peptide in the spinal cord lysates (see Fig 3B, square). Among big tau peptides, the exon 4a‐L peptide (residue 235–242) was the most abundant peptide in the spinal cord (0.100 ± 0.032 compared to total‐tau, n = 15), an order‐of‐magnitude higher than what we detected in the brain lysates (p = 0.026; see Fig 3B). We also observed other exon 4a‐L peptides (residue 294–308, 0.047 ± 0.021 and residue 429–442, 0.060 ± 0.017 compared to total‐tau, n = 15) were 20‐fold higher in the spinal cord tissue (see Fig 3B, square) samples compared with brain lysates. These results indicate that exon 4a‐L is more widely expressed in the spinal cord (approximately 10.7% of total tau) compared to the brain (see Fig 3B, 3F). Compared with brain tissues, we detected 2.5‐fold more exon 6 peptide (residue 545–552, 0.001 ± 0.001 compared to total‐tau peptide, n = 15) in spinal cord lysates (see Fig 3B, square). These results highlight 10‐fold higher exon4a‐L expression compared to exon 6 in the spinal cord.

To further characterize the distribution of the tau exons 4a‐L and 6 in the PNS, we measured soluble tau peptides from adult human SN (n = 6), BP (n = 2), and DRG (n = 2) lysates. We detected exon 4a‐L peptides at sub‐equimolar amounts in the SN (residue 235–242, 0.566 ± 0.260; residue 294–308, 0.268 ± 0.180, and residue 429–442, 0.173 ± 0.082, n = 6; see Fig 3B, circle), when normalized to total‐tau (residue 633–642). We detected exon 4a‐L peptides at similar ratios in BP (residue 235–242, 0.561 ± 0.250; residue 294–308, 0.294 ± 0.126, and residue 429–442, 0.295 ± 0.091 normalized to total‐tau) and DRG (residue 0.441 ± 0.059; residue 294–308, 0.357 ± 0.202, and residue 429–442, 0.577–0.071 normalized to total‐tau). This distribution profile indicates that big tau in the PNS constitutively expresses exon 4a‐L. Our results highlight exon 4a‐L is widely expressed in the PNS (approximately 50%), 2‐orders of magnitude higher compared to the brain. We estimated low abundance of exon 6 peptide (residue 545–552) in the SN (0.002 ± 0.002), BP (0.005 ± 0.003), and DRG (0.006 ± 0.003) lysates (see Fig 3B and Supplementary Fig S1C), reflecting a 10‐fold increase compared to relative abundance in the brain. Overall, these findings suggest there is a distribution pattern for tau exon 4a‐L expression from the CNS to the PNS, increasing from very low levels (0.3% of total tau) in the brain to intermediate levels in the spinal cord (10.7% of total tau) to high levels (approximately 50%) in the PNS.

Additionally, whereas the relative contribution of exon 4a‐L to total tau increases from the CNS to PNS, we found that the concentration of total tau demonstrated an inverse gradient (Fig 3C). We estimated that the brain contains approximately 12 ng/μg protein of soluble tau (small and big tau combined), whereas there is 6‐fold decrease (p < 0.0001) in the total tau levels in the spinal cord (approximately 2 ng/μg protein). The PNS (SN, BP, and DRG) contains roughly 3 orders of magnitude lower levels of soluble tau (0.01–0.03 ng/μg protein) compared to the brain (p < 0.0001). Additionally, we found exon 6 is expressed at a very low level globally compared with other common tau exons in the nervous system. The exon 6 expression is increased by 2.5‐fold in the spinal cord (0.1% of total tau) compared to the brain (less than 0.05% of total tau), however, we detected no significant changes in the SN, DRG, and BP (0.2–0.5% of total tau) compared to the spinal cord (see Fig 3B).

Along with the relative expression pattern of exon 4a‐L across the nervous system, we also quantified the 3R and 4R isoform expression pattern as they are developmentally regulated. ²⁷ , ⁶⁶ We observed a significant decrease in the 4R isoform (residue 720–738) levels in the spinal cord (0.49 ± 0.09 normalized to total‐tau, p < 0.0001) and SN lysates (0.38 ± 0.20, p < 0.0001) compared to the brain (0.66 ± 0.23; see Fig 3B). However, 3R isoform (residue 727–738) remained unaltered in the spinal cord (0.47 ± 0.09) and SN lysates (0.49 ± 0.23) compared to the brain (0.52 ± 0.20; see Fig 3B). These results further highlight that alternative splicing and insertion of tau exon 10 is also differentially regulated across the human nervous system.

N‐Terminal Projection of Small and Big Tau Across the Nervous System

The adult human brain is known to express approximately 40% 0 N, 50% 1 N, and 10% 2 N small tau isoforms (tau without exon 4a‐L insert). ²⁷ , ⁶⁷ , ⁶⁸ However, the distributions of “big tau” isoforms across the nervous system have not been investigated. Guided by the IP‐MS/MS data (see Fig 2 and Supplementary Figs S1–S3), we explored the expression pattern of tau exons 2 and 3 with and without exon 4a‐L across the nervous system (Fig 3D–F). We observed that splicing of MAPT exons 2 and 3 leads to expression of 42.6% 0 N, 43.4% 1 N, and 13.0% 2 N small tau isoforms in the brain (Fig 3D), similar to what has been previously reported. ²⁷ , ⁶⁹ We observed spinal cord has 32% 0 N, 45% 1 N, and 12% 2 N small tau isoforms. The PNS (SN, BP, and DRG) contains 37.7% 0 N, 49.2% 1 N, and 14.3% 2 N small tau isoforms (see Fig 3D).

The big tau isoforms (with exon 4a‐L) were present at low abundance in the brain (0.0004 ± 0.0005 0 N; 0.002 ± 0.003, 1 N and 0.001 ± 0.001, 2 N big tau isoforms compared to total‐tau; see Fig 3E). Compositionally, we detected 0.03% 0 N, 0.16% 1 N, and 0.08% 2 N big tau isoforms out of total tau detected in the brain (see Fig 3E). In contrast, in the adult human spinal cord, we detected 0.005 ± 0.001 0 N, 0.06 ± 0.02 1 N, and 2 N big tau isoforms compared to total tau (see Fig 3E). Compositionally, adult spinal cords contain 4% 0 N and 48% 1 N and 48% 2 N big tau isoforms. Further, we also observed that PNS contains 15.0% 0 N (0.086 ± 0.090 compared to total tau), 40% 1 N (0.244 ± 0.190), and 45% 2 N (0.718 ± 0.150) big tau isoforms (see Fig 3E). These results highlight the differential splicing of exons 2 and 3 for small and big tau isoforms in the adult nervous system. The exclusion of both exons 2 and 3 is more common for small tau, while favoring a larger N‐terminal projection that exclusively contains exons2/3 for “big‐tau”. We demonstrated that, whereas the majority of tau expressed in the CNS, brain (99.7%), and spinal cord (89.3%) is small tau (0 N + 1 N + 2 N small tau isoforms), there is a considerable amount of small tau (approximately 65% in SN, 50% in BP, and 44% in DRG) expression in the PNS (see Fig 3F). Compositionally, big tau isoforms (0 N + 1 N + 2 N big tau) are significantly increased to 56% in DRG (p < 0.0001), 50% in BP (p < 0.0001), 35.1% in the SN (p < 0.0001), and 10.7% in the spinal cord (p < 0.0001) compared to 0.3% in the brain (see Fig 3F). The distribution of these big tau 4a‐L isoforms further validated the relative expression pattern of tau exon 4a‐L from CNS to PNS across the nervous system (Fig 3G).

Regional Distribution of Big Tau in the Human Brain

In rodents, big tau expression appears enriched in the cerebellum relative to cortical regions, although prior studies have reported inconsistent cortical expression. ⁴⁰ , ⁵⁵ To investigate if there is similar regional variance in the human brain, we quantified the distribution of small and big tau peptides from 6 brain regions. We calculated the total contribution of small (Fig 4A) and big tau (see Fig 3B) N‐terminal isoforms in these brain regions. The major contribution of small tau isoforms (0 N + 1 N + 2 N) to the total of tau species remained unaltered across brain cortical regions (temporal, occipital, and frontal) and the cerebellum (see Fig 4A). Parietal small tau isoform levels (1.48 ± 0.18 normalized to total‐tau) were significantly higher than frontal cortex (1.1 ± 0.15; see Fig 4A, p < 0.0001). Regarding the minor big tau isoforms (0 N + 1 N + 2 N), they were significantly increased in the cerebellum (0.018 ± 0.009 normalized to total‐tau, p < 0.001) compared with temporal (0.005 ± 0.004) and parietal (0.002 ± 0.001) brain regions (Fig 4B). The analysis showed that the cerebellum has the highest big tau isoforms (1.2% of total tau) with decreasing levels in the cortical regions in the following order: frontal cortex (0.4%) > occipital cortex (0.3%) > temporal cortex ~ parietal cortex (0.2%; Fig 4C). Big tau isoforms remain relatively low in abundance (less than 1%) in comparison to small tau isoforms across the cortical regions (99%) and the cerebellum (98%).

FIGURE 4.

Regional distribution of small and big tau in the adult human brain. (A) Bar graphs of the relative distribution of the small tau isoforms (0 N + 1 N + 2 N) normalized to t‐tau (residue 633–642) within the human brain: parietal cortex (n = 25), frontal cortex (n = 21), temporal cortex (n = 6), occipital cortex (n = 6), and cerebellum (n = 6). Data are expressed as mean ± SD. Data were analyzed using non‐parametric 1‐way ANOVA followed by Dunn's multiple comparisons test, ****p < 0.0001, ***p = 0.001. (B) Bar graphs of the relative distribution of the big tau isoforms (0 N + 1 N + 2 N) normalized to t‐tau with the human brain regions for all the samples used in A. All values are expressed as mean ± SD. Data were analyzed using non‐parametric 1‐way ANOVA followed by Dunn's multiple comparisons test, ****p < 0.0001, ***p = 0.001. (C) Bar graphs depicting the relative contribution of small tau isoforms and big tau isoforms (0 N + 1 N + 2 N) across the human brain for all the samples used in A. Data are expressed as mean ± SD. Statistical analysis performed using 2‐way ANOVA with Tukey's multiple comparisons test, ****p < 0.0001.(D–F) Bar graphs of the relative exon 4a‐L peptides (residue 235–242 and residue 452–461) and exon 6 peptide (residue 545–552) ratios for all the samples used in A. Data are presented as mean ± SD. Statistical analysis was performed using ordinary 1‐way ANOVA with Tukey's multiple comparisons, ****p < 0.0001, ***p ≤ 0.0006, **p ≤ 0.005, *p ≤ 0.0463. (G) Schematic diagram for the regional distribution in the human brain demonstrates heterogeneity in the expression of big tau protein. ANOVA = analysis of variance. [Color figure can be viewed at www.annalsofneurology.org]

Brain regional distribution pattern for big tau was further validated on other exon 4a‐L peptides that were significantly increased in the cerebellum (residue 235–242, 0.03 ± 0.01 and residue 452–461, 0.002 ± 0.001) compared with other cortical regions (p < 0.0001; Fig 4D–F and Supplementary Fig S7). A significant increase was detected for the exon 6 peptide (residue 545–552) in the cerebellum (0.002 ± 0.001) compared to the parietal (0.0002 ± 0.0001, p < 0.0001) and temporal cortical regions (0.001 ± 0.0005, p = 0.0261; see Fig 4F). These results confirm regional variance of exon 4a‐L and exon 6 translation in the human brain with the cerebellum containing higher levels of big tau compared to the cerebral cortex, which highlighted a 3‐fold increase of big tau isoforms in the cerebellum (1.2%) compared to 0.2 to 0.4% detectable in the cortical brain regions (Fig 4G and see Supplementary Fig S7).

Small and Big Tau Peptides in Human CSF

Soluble tau in the CSF primarily consists of small tau species derived from the brain with a major truncation around the end of the proline rich region (PRR). ¹⁴ , ⁵⁷ Recent investigations of CSF tau species (tau isoform and p‐tau) have mainly focused on the small tau isoforms due to their closer association with pathological changes within the brain in neurodegenerative disorders, whereas the contribution of the big tau isoforms in this biofluid and their disease relevance remain unexplored. Tau was immunoprecipitated (Tau‐1/HJ8.5/HJ8.7) from human CSF and extracellular small tau and big tau specific peptides (exon 4a‐L and exon 6 peptides) were quantified to estimate their corresponding contributions to the total tau concentration in the biofluid (Fig 5A). The CSF concentration of total tau (estimated from peptides comprised of residues 06–23, 45–67, 602–611, 616–630, and 633–642) from young normal control (YNC) participants ranged from 1.65 to 5.8 ng/ml, and that of the truncated tau containing microtubule binding region of tau (MTBR‐tau‐243, estimated from a peptide spanning residues 243–254) was 0.1–0.2 ng/ml (Supplementary Fig S8). ¹⁴ The CSF concentration of big tau from YNCs (estimated from peptides derived from exon 4a‐L, comprised of residues 235–242, 294–308, 379–395, 429–442, and 452–461) ranged from 0.04–0.1 fmol/ml (see Fig 5A [blue highlighted] and Supplementary Fig S8), representing approximately 1% of total soluble tau (total‐tau peptide [30.0 ± 16.9 fmol/ml]) in the CSF, which is comparable to its proportion estimated in the brain (0.3–0.5% big tau compared to total tau; see Fig 3). The exon 6 peptide (spanning residues 545–552) was detected at even lower levels (0.2% compared to total tau; see Fig 5A, red highlighted). In contrast, the small tau specific peptide (small tau 0 N isoform) was detected at relatively higher levels (0.25 ± 0.01 normalized to total‐tau; see Fig 5A, green highlighted) compared to total tau (T212 peptide according to small tau, spanning residues 633–642 in big tau; and as represented in Fig 5A). This analysis confirms CSF tau isoform distribution is essentially identical to that of the adult human brain.

FIGURE 5.

Distribution profile of the small tau and big tau peptides in the human CSF and their respective diagnostic performance as predictive biomarkers in AD continuum. (A) Box plots of log₁₀ normalized CSF big tau profile (n = 70; includes young normal controls, Aβ‐, Aβ+) quantified using HJ8.5/HJ8.7 and Tau 1 immunoprecipitation focusing on N‐terminal tau peptides to mid‐domain of tau. CSF small tau 0 N isoform peptide (residue 45–67), exon 4a‐L, exon 6 peptides, and total‐tau peptide (t‐tau, residue 633–642) is highlighted from left to right. Scatter plots of CSF tau (B) small tau 0 N isoform (residue 45–67), (C) exon 4a‐L (residue 235–242) and (d) exon 6 (residue 545–552) peptide concentrations from a cross‐sectional cohort of YNCs (n = 23), amyloid‐negative CDR = 0 (age‐matched control, n = 26), amyloid‐positive CDR = 0 (preclinical AD, n = 14), amyloid‐negative CDR >0 (non‐AD with cognitive impairment, n = 4) and amyloid‐positive CDR >0 (symptomatic AD, n = 21). Data are presented as mean ± SD. Statistical analysis conducted by using ordinary 1‐way ANOVA with multiple comparisons, ****p < 0.0001, ***p = 0.0001, **p = 0.0038. The ROC curves and AUC values for (E) small tau 0 N (residue 45–67) isoform, (F) exon 4a‐L (residue 235–242) and (G) exon 6 (residue 545–552) peptides demonstrating their respective differential diagnostic accuracies in differentiating symptomatic clinical stage of AD compared to preclinical AD and age‐matched controls. AD = Alzheimer's disease; ANOVA = analysis of variance; AUC = area under the curve; CSF = cerebrospinal fluid; ROC = receiver operating characteristic; YNC = young normal controls. [Color figure can be viewed at www.annalsofneurology.org]

CSF Small and Big Tau Concentrations in AD Cross‐Sectional Cohort

To begin to evaluate if small tau and big tau (exon 4a‐L/exon 6) peptides in CSF are changed in AD, we examined the CSF from a cross‐sectional cohort ⁵⁷ of amyloid‐negative and amyloid‐positive participants representing different clinical stages: YNCs (n = 23), older amyloid‐negative CDR = 0 participants (age matched controls, n = 26), older amyloid‐positive CDR = 0 participants (preclinical AD, n = 14), older amyloid‐negative CDR >0 participants (non‐AD dementia, n = 4), and older amyloid‐positive CDR >0 participants (symptomatic AD, n = 21). The CSF small tau levels (0 N isoform) showed statistically significant differences between symptomatic AD (CDR >0, Aβ+) and age‐matched controls (CDR = 0, Aβ‐, p = 0.0008) as well as preclinical AD (CDR = 0, Aβ+, p = 0.014) and non‐AD dementia (Fig 5B). CSF big tau exon 4a‐L tau peptide levels (eg, that span residues 235–242) were only significantly increased in symptomatic AD (CDR >0, Aβ+) individuals compared with YNCs (p < 0.0001; Fig 5C) and not in comparison to age‐matched controls or across the AD stages. CSF levels of the exon 6 peptide (residues 545–552) were significantly increased in symptomatic AD individuals compared with young and age‐matched controls (p < 0.0001) and those with non‐AD dementia (p = 0.03; Fig 5D), although no significant difference was observed for exon 6 peptide between preclinical AD and symptomatic AD.

Next, we investigated the correlation of CSF small and big tau species with other CSF tau species such as p‐tau and truncated tau. We quantified multiple tau species (non‐phosphorylated tau peptides comprised of residues 06–23, 25–44, 68–87, 45–67 [0 N small tau], 602–611, and 633–642), phospho‐tau (phospho‐tau pT181 and pT217), big tau (peptide comprised of residues 235–242, 2,294–308, 429–437, and 452–461), and MTBR‐tau243 in the CSF from these participants (see Supplementary Figs S8 and S9) and correlated them with small and big tau levels across the whole cohort (Supplementary Fig S10). The small tau (0 N isoform) peptide strongly correlated with other common tau species (eg, ρ = 0.93, p < 0.0001 with total‐tau [T212‐221]; and ρ = 0.97, p < 0.0001 with total‐tau [T181‐190]; see Supplementary Fig S10), correlated weakly with phospho‐tau pT181 (ρ = 0.44, p < 0.0001) and phospho‐tau pT217 (ρ = 0.51, p < 0.0001 with pT217/T217; see Supplementary Fig S10), a biomarker that is highly correlated with amyloidosis in preclinical AD, ⁷⁰ , ⁷¹ and correlated strongly with CSF MTBR‐tau243 (ρ = 0.87, p < 0.0001; see Supplementary Fig S10), a truncated tau species in CSF that exists at baseline in preclinical AD and has been demonstrated to increase and correlate with NFT pathology in AD. ¹⁴ , ⁷² This highlights that small tau 0 N isoform starts to increase when individuals that are positive for amyloid become symptomatic, in contrast to phospho‐tau species that increase when individuals become positive for amyloid, and that have therefore become established biomarkers for preclinical (asymptomatic) AD. We further investigated the association of clinical cognitive measures (MMSE) and CSF biomarkers (Supplementary Fig S11). As previously demonstrated, CSF MTBR‐tau243, a biomarker of AD tau tangle pathology, was negatively correlated with cognitive performance in our cohort (ρ = −0.45, p = 0.0012; see Supplementary Fig 11A), with higher levels associated with poorer cognitive outcomes. ¹⁵ Likewise, N‐terminal tau (represented by a peptide spanning residues 06–23), small tau (0 N isoform), and total‐tau (represented by a peptide spanning residues 181–190) demonstrated significant negative associations with poorer cognitive outcomes (see Supplementary Fig 11B–D,F). Importantly, we found that the percent of occupancy of phosphorylation at pT217 (pT217/T217%) also demonstrated a strong negative correlation with cognitive outcomes (ρ = −0.52, p < 0.0001; see Supplementary Fig 11G), whereas the percent pT181 (pT181/T181%) did not demonstrate a significant association with cognitive outcomes (ρ = −0.11, p = 0.44; see Supplementary Fig 11G). In contrast, the CSF big tau exon 4a‐L peptide (residue 235–242) weakly correlated with total tau (T212; ρ = 0.34, p = 0.02), pT217/T217% (ρ = 0.32), and MTBR‐tau243 (ρ = 0.38; see Supplementary Fig 10). Additionally, we observed weak correlations of CSF big tau exon 4a‐L peptides (big tau‐294, big tau‐429, and big tau‐452) with CSF total tau (T212, ρ = 0.27–0.35), pT217/T217% (ρ = 0.3), and MTBR‐tau243 (ρ = 0.4). These results suggest that CSF levels of soluble exon 4a‐L tau species are not impacted by amyloid status (inferred from pT217/T217%) or tau pathology (CSF MTBR‐tau243 and total tau) in AD. In contrast, the exon 6 peptide weakly correlated with CSF tau species (ρ = 0.50, p < 0.0001 with total‐tau and r = 0.55, p < 0.0001 with MTBR‐tau243; see Supplementary Fig 10). Additionally, CSF big tau exon 4a‐L peptide levels (residues 235–242 and 429–442) demonstrated no correlation with cognitive measures (big tau 235–242; ρ = 0.052, p = 0.72 and big tau 429–442; ρ = −0.013, p = 0.93; see Supplementary Fig 11H, 11I), whereas exon 6 peptide (residue 545–552) demonstrated significant negative correlations with cognitive performance (ρ = −0.46, p < 0.001; see Supplementary Fig S11J). These findings suggest that CSF exon 6 levels change in symptomatic AD and potentially reflect altered tau tangle pathology.

Additionally, we investigated the diagnostic accuracies of the small tau and big tau (exons 4a‐L/6) peptides in the CSF when distinguishing individuals at different stages of AD (Fig 5E–G). CSF small tau levels discriminated between CDR >0, Aβ + and CDR = 0, Aβ + with an AUC = 0.78, 95% confidence interval [CI] = 0.62–0.93 (see Fig 5E, red). When discriminating between CDR >0, Aβ + and CDR = 0, Aβ‐ individuals, CSF small tau had an AUC = 0.85, 95% CI = 0.74–0.96; see Fig 5E, blue). In comparison, CSF big tau peptide (residues 235–242, from exon “4a‐L”) showed essentially no utility for discriminating CDR >0, Aβ + from CDR = 0, Aβ‐ with an AUC = 0.54 (see Fig 5F, blue) or CDR >0, Aβ + and CDR = 0, Aβ + (see Fig 5F, red). In contrast, CSF exon 6 peptide (residues 545–552) discriminated between CDR >0, Aβ + and CDR = 0, Aβ + groups with an AUC = 0.73, 95% CI = 0.57–0.95 (see Fig 5G, red) and between CDR = 0, Aβ‐ and CDR >0, Aβ + groups with an AUC = 0.91, 95% CI = 0.83–0.99 (see Fig 5G, blue).

Finally, we investigated the relationships of small and big tau CSF levels with age for the entire cross‐sectional cohort with and without amyloidosis (Fig 6). We observed that the CSF small tau levels were significantly increased for amyloid positive individuals at age > 70 years compared with amyloid negative individuals at age > 70 years (p = 0.02) and amyloid negative individuals in the age group 40–60 (p = 0.001; see Fig 6A and Supplementary Fig S10). CSF exon 4a‐L big tau peptide increased with age and was significantly increased in amyloid positive individuals in the age group >70 years compared with YNCs in the age group 19 to 40 (p = 0.04; see Fig 6B). CSF exon 6 peptide was significantly increased in amyloid‐positive individuals in the age group >70 years compared with amyloid negative individuals in the age group >70 years (p < 0.001), YNCs in the age group 19 to 40 (p = 0.01) and amyloid negative individuals in the age group 60 to 70 (p = 0.04; see Fig 6C). In amyloid‐negative individuals, we observed big tau peptide (residue 235–242) correlated better with age (r = 0.56, p < 0.0001) compared to MTBR‐tau243 (r = 0.47, p < 0.0001), small tau (r = 0.3), and total‐tau (r = 0.36, p < 0.0001; see Fig 6D).

FIGURE 6.

CSF exon 4a‐L peptides increase with age independently of cerebral amyloidosis. Scatter plots of CSF tau (A) small tau 0 N isoform, (B) exon 4a‐L (residue 235–242) and (C) exon 6 peptide concentrations from a cross‐sectional cohort in different age groups with and without amyloidosis (Aβ‐, n = 41; Aβ+, n = 37). Data are presented as mean ± SD. Statistical analysis was performed using 2‐way ANOVA with multiple comparisons, ***p = 0.0006, **p ≤ 0.0088. (D) Correlation matrix presenting Spearman's correlation for all the tau biomarkers with each other in only Aβ‐ participants. ANOVA = analysis of variance; CSF = cerebrospinal fluid. [Color figure can be viewed at www.annalsofneurology.org]

In summary, CSF small tau isoform (0 N isoform) exhibited changes later in the symptomatic stage of AD, associating better with CSF total‐tau as well as CSF MTBR‐tau243. CSF big tau exon 6 peptide also demonstrated changes associated with AD that will need further exploration in a larger cohort, whereas CSF big tau peptides (exon 4a‐L) did not change with amyloidosis, tau pathology (CSF MTBR‐tau243) or cognitive symptom onset (Supplementary Figs S11–S13).

Discussion

Despite the knowledge of the tau mRNA transcript for the last 3 decades, the protein distribution of the longest isoform of tau – “big tau” has not been actively investigated in the human nervous system. ⁴⁰ , ⁴⁷ , ⁵⁰ , ⁵² Previous qualitative reports have revealed big tau is preferentially expressed in the soma, dendrites, and axons of the CNS neurons that extend processes into the periphery, in addition to PNS neurons and optic nerves. ⁴⁰ Exon 4a‐long has been previously reported in human cancer cells and skeletal muscle tissues (non‐neuronal tissues) as the primary constituent of HMW tau along with the common small tau isoforms. ⁴⁵ , ⁴⁶ This larger splice variant, exon 4a‐L (1,065 base pairs (bp)) was found to be expressed in higher amounts (by RT‐PCR) than the 753 bp exon‐4a‐S. ⁴⁶ The exon 4a‐L contains two 3′ splice sites and when the alternative 3′ splice junction (CAG/A) within this exon is utilized it results in the insertion of an additional 312 bp between exons 4 and exon 4a‐S. ⁴⁵ Using tandem MS/MS sequencing techniques, here, we report that this longer exon 4a (exon 4a‐L) is predominantly expressed across the human nervous system and encodes for 355 amino acids. This longer exon 4a is inserted between tau exons 4 and 5 and has an additional 104 amino acids compared to the exon 4a‐S (251 amino acids). ⁵⁰ , ⁵⁵ Tandem MS/MS using protein database searching led to the discovery of the exon 4 to exon 4a‐L junction/bridging peptides in human nervous system tissues. The alternative 3′ splice site resulting in exon 4a‐L has been proposed as favored splice product due to the presence of consensus (i) 3′ splice junction, (ii) branch point sequence (BPS) upstream in the genomic sequence and (iii) polypyrimidine rich sequence (28 bp) following the BPS. Long read sequencing data from control brain samples (caudate and middle frontal gyrus) further confirmed that the prominent form of big tau in the CNS is the exon 4a‐L isoform. Low transcript levels of exon 4a‐L isoforms in the selected brain regions were consistent with the quantitative proteomic results of low relative abundances of big tau in the brain (<0.5% of total tau). Future long‐read sequencing across additional brain regions (cerebrum, cerebellum, and brainstem) will help define the anatomic heterogeneity of big tau isoform expression (4a‐L vs 4a‐S). Targeted MS using synthetic standard peptides for big tau exon 4a‐L and exon 4a‐S isoforms demonstrated that exon 4a‐L is the predominant form (> 95% of big tau) in human nervous tissues, whereas exon 4a‐S isoforms constitutes a minor fraction (<5% of big tau). Notably, exon 4a‐S was relatively more abundant in the PNS (approximately 3.5% of big tau) than in the CNS (<1% of big tau), although exon 4a‐L remained the dominant isoform across regions. Altogether, our results expand the repertoire of human big tau isoforms beyond what has been previously reported in the literature.

To complement the MS and transcriptomic analyses, we used immunoblotting to resolve tau species by molecular weight, enabling detection of isoform heterogeneity and HMW species not readily distinguished by bulk RNA sequencing or peptide‐based proteomics. We observed a range of HMW bands approximately 115 to 90 kDa in PNS and spinal cord tissue samples on immunoblots using tau‐1 and tau‐5 antibodies, whereas no discernable HMW bands were detected in the brain regions investigated. Tau PTMs (truncation, phosphorylation, acetylation, etc.) could result in multiple HMW bands of big tau. Future investigations using big tau specific antibodies on different tissue lysates will provide further insights on the origin of multiple HMW tau bands beyond the alternative spliced big tau isoforms (exon 4a‐L/4a‐S). SDS‐PAGE followed by in‐gel digestion and LC–MS of these HMW bands from CNS and PNS tissue lysates confirmed the presence of exon 4‐4a‐L junction peptides. The insertion and expression of exon 4a‐L along with exon 6 results in an isoform of tau (big tau) that is almost twice as long (862 amino acids) as the conventional small tau isoform (441 amino acids). The exon 4a between human and rodent nervous system are different. The protein sequences of big tau (exon 4a) in mouse and rat retain only 50% and 51% homology, respectively, with human big tau. ⁵⁰ Another between‐species distinction of big tau sequence could be explained by the presence of an alternative 3′ splice site within human tau genome. In rodents, there is a stop codon within the additional 312 bp sequence and their genomic sequence lacks consensus BPS analogous to a human branchpoint. Thus, whereas our results indicate that human big tau is primarily derived from exon 4a‐L insertion, we expect rodents will primarily express the smaller HMW tau isoform (exon 4a‐S). The expression pattern of exon 4a‐L and exon 4a‐S in rodents needs further investigation.

Ever since its discovery, several descriptive reports have tried to delineate the distribution of big tau in the PNS and optic nerves along with some selected CNS regions. ⁴⁰ , ⁴⁷ , ⁵² In this study, we found relative differential distribution pattern of exon 4a‐L expression increasing from CNS to PNS across the human nervous system in the following order: brain << spinal cord < SN < BP ~ DRG. More importantly, whereas small tau is the major tau isoform in the CNS (approximately 99% in the brain and approximately 90% in the spinal cord), our results indicate there are almost equimolar amounts of small tau and big tau in the PNS (approximately 44–55%). Our results challenge the idea that small tau is CNS specific with big tau being exclusive to the PNS. Instead, we propose small tau being CNS enriched, whereas the PNS is enriched with big tau isoforms. Our results also highlight that although the percentage of big tau contribution increases from CNS to PNS, the total tau concentration from CNS to PNS decreases by approximately 3‐orders of magnitude. We found the brain (approximately 10 ng/μg protein) has almost 1,000‐fold higher total tau levels than the peripheral nerves (approximately 0.03 ng/μg protein). Similar distribution of total tau has been previously reported, where peripheral tissues demonstrated approximately 1.8% (submandibular gland) to 0.16% (liver) total tau compared to the brain (~ 7,800 ng/mg brain). ⁴¹ Another important finding of our study was the relatively low exon 6 peptide abundance in the spinal cord and PNS, indicating exon 6 is weakly expressed in these tissues. This finding could potentially explain why big tau isoforms are different in neural cell lines compared with DRG, spinal cord, and SN, which selectively express the exon 4a‐L and not exon 6. ³⁹ Inclusion of exon 6 in both the 6 and 9 kb tau mRNA and its presence has been reported in both fetal and adult human tau mRNA. ²⁵ , ³⁹ , ⁶⁶ Exon 6 has also been found in mature and immature spinal cord but not in the PNS, decoupled from the peripheral exon 4a‐L. ²⁵ This could lead to a further increase in the repertoire of tau isoforms beyond what we have reported in this study. Our results indicate exon 6 and exon 4a‐L could be mutually exclusive, and splicing could be regulated in a tissue‐specific manner, something that will need further validation using specific antibodies to tau exon 4a‐L and exon 6. Understanding the functional consequences of these differential expression profiles requires further investigation.

Developmental studies of tau expression in human brains have shown 3 major isoforms of tau at early postnatal stages, versus 6 isoforms of tau in adulthood. ² , ²⁷ , ²⁸ , ⁶⁶ , ⁷³ In contrast, big tau isoform stoichiometry and distribution have not been studied until now. Intuitively, cassette splicing of exons 2 and 3 (E2 and E3) should give rise to big tau isoforms with 0, 1 or 2 N‐terminal repeats, as has been observed for small tau. Our characterization of the exon 4a‐L sequence was crucial in identifying the specific exon 4 to exon 4a‐L junction in big tau, which led to the discovery of the 3 distinct N‐terminal isoforms for big tau (0 N‐, 1 N‐, and 2 N‐big tau). Given that 3R and 4R isoforms were also quite abundant in the PNS, we speculate that E10 splicing for big tau would be a common feature, resulting in 6 distinct isoforms of big tau, similar to the 6 isoforms of small tau. Consistent with previous reports, ²⁷ , ⁶⁷ we observed adult human brains have 43% 0 N, 46% 1 N, and 13% 2 N small tau isoforms, with subtle increase in 1 N small tau isoform across spinal cord to PNS. Unlike small tau isoforms, we discovered adult human nervous tissue contains equimolar levels of 1 N and 2 N (40–50%) big tau isoforms, with 0 N big tau isoform being present at less than 5%. These observations, based on multiple CNS (the brain and spinal cord) and PNS (SN) tissue lysates from cognitively normal control individuals, and individuals with AD and ALS, highlighted that the extended N‐terminal projection domain in big tau is a common feature across the adult human nervous system. Alternatively, spliced isoforms of the same protein can have distinct biological functions, with some alternative isoforms being functionally divergent. ⁷⁴ Although the specific physiological roles of big tau remain unclear, the “long” exon 4a insertion is hypothesized to significantly increase the spacing between microtubules, compared to small tau. Similar dramatic structural changes have also been associated with other microtubule associated proteins (MAPs); the LMW isoform (approximately 70 kDa) of MAP2C (juvenile form of MAP2) switches to higher molecular weight MAP2A/B (approximately 250 kDa) during later stages of neuronal development. ⁷⁵ , ⁷⁶ The N‐terminal projection domain of tau, which functions as a variable spacer between neighboring microtubules and regulated by alternative splicing of E2/E3, may play a developmental role in modulating inter‐microtubule spacing. ⁷⁷ , ⁷⁸ Indeed, cross‐sections of axonal microtubules reveal a larger spacing of approximately 35 nm when induced by big tau compared with approximately 20 nm when induced by small tau isoforms in neuronal cultures. ⁷⁸ This larger spacing has been suggested to provide more efficient organelle transport in the axon by lowering resistance of the axoplasm compared to the small tau, that is more conducive to axonal growth. ⁴⁷ The presence of big tau in the long and high caliber neurons in the periphery may be driven by the need for robust and energy efficient axonal transport, whereas small tau isoforms provide more plasticity to the CNS neurons. Developmental studies on big tau expression in rodents indicate active translation in postnatal stages, ⁴⁰ , ⁵⁰ and further exploration would provide clues why we observe a distinct dissociation in relative isoform distribution between small and big tau.

Selective vulnerability is a common phenomenon in many neurodegenerative disorders that feature distinct patterns of neuronal loss and accumulation of protein aggregates within certain brain regions, whereas other regions are resistant to the pathology. ¹⁰ , ⁷⁹ , ⁸⁰ , ⁸¹ Stereotactic spread of tau pathology (NFTs) in AD and other tauopathies ¹¹ , ⁸² , ⁸³ is thought to occur throughout the hippocampus and the cortex, sparing very few brain regions such as cerebellar cortex even in extreme cases. ¹⁰ , ⁸⁴ Tau expression is ubiquitous within the brain, whereas the selective vulnerability of certain brain regions in tauopathy remains an intriguing outstanding question. Although prior studies have examined expression and splicing regulation of tau across different brain regions, ²⁸ , ⁶⁷ currently, the exact reasons for this regional vulnerability are not yet known. Interestingly, exon 3‐encoded inserts have been shown to inhibit tau aggregation, while exon 2 and 10‐encoded inserts increase the aggregation propensity of tau. ⁸⁵ Recent research has highlighted that big tau has significantly diminished aggregation propensity compared to the small tau isoforms. ⁸⁶ They found that exon 4a (exon 4a‐S) has enhanced microtubule‐binding capacity and exhibited less hyperphosphorylation – key mechanisms that ultimately promote tau aggregation. ³ , ⁸⁷ Future investigations are needed to determine whether exon 4a‐L big tau isoforms similarly exhibit enhanced microtubule binding along with reduced aggregation propensity. Indeed, in our bottom‐up proteomic investigation across the nervous system, we did not observe evidence for phosphorylated exon 4a peptides. Moreover, multiple mutations and polymorphisms have been identified on the exon 4a in humans without any known pathogenicity. ⁸⁸ , ⁸⁹ , ⁹⁰ , ⁹¹ Our finding of 3‐fold increase in exon 4a‐L expression in the cerebellum compared with other cortical brain regions aligns with previous reports in human brains and rodents, ⁴⁰ , ⁴⁷ , ⁸⁶ emphasizing the regional heterogeneity of big tau expression in the brain. Most interestingly, the lowest levels of exon 4a‐L in the brain coincide with the vulnerable regions for AD. This raises the intriguing prospect of therapeutically altering the small and big tau ratios in these brain regions to recruit the neuroprotective role of big tau in resisting tangle pathology. ⁵⁵ , ⁹²

CSF is thought to closely reflect brain pathological processes, carrying molecules through passive diffusion from the brain parenchyma. In this study, we finally investigated CSF to understand the tau isoforms distribution in this biofluid and their relation to AD. This is the first study to quantitively measure and compare the small tau and big tau peptides in human CSF. Normalizing the small tau (0 N isoform) peptide in the CSF to the total tau peptide indicated this isoform is quite abundant, whereas big tau (exon 4a‐L/exon 6) accounts for less than 1% of total tau in the CSF. This is consistent with the relative abundance of exon 4a‐L and exon 6 in the brain. CSF total‐tau (t‐tau) measured by monoclonal antibodies binding to the mid‐domain of tau is known to increase during acute neurodegeneration. ⁹³ , ⁹⁴ Assays that measure endogenous brain‐derived fragments of tau species from the N‐terminus and MTBR have been developed to improve the diagnosis of AD and other tauopathies. ¹⁴ , ¹⁹ , ⁷² Recent studies highlight the importance of brain‐derived tau species in the CSF/plasma as neurodegeneration biomarkers useful in identifying and monitoring neurodegenerative progression in AD. ¹⁶ , ¹⁹ , ²⁰ Plasma N‐terminal containing tau fragments (NTA‐tau) and brain‐derived tau (BD‐tau) assays were first validated from cross‐sectional AD cohorts with paired CSF and plasma samples from cognitively normal and symptomatic individuals. Our MS data provide plausible tau species that are also targeted by these assays for detecting tau pathology in AD. In this study, we found significant increase of CSF small tau (0 N isoform) in amyloid‐positive symptomatic individuals (CDR >0). High correlation of the CSF small tau (0 N isoform) with total‐tau and MTBR‐tau243, and not so much with p‐tau measures (pT181/T181 and pT217/T217), suggests a closer link to the clinical phase of AD. ¹⁴ , ¹⁷ , ⁷¹ In contrast, we found CSF big tau exon 4a‐L increased with age but no significant changes of the CSF big tau exon 4a‐L peptides in either preclinical AD or in individuals with mild cognitive impairment due to AD. We also compared our results with CSF MTBR‐tau243, a high performing fluid biomarker specific for AD tangle pathology. CSF big tau exon 4a‐L peptides poorly correlated with total‐tau biomarkers and MTBR‐tau243 (ρ = 0.38). Our results thus identify the small tau (0 N isoform) as an important biomarker target in the blood similar to brain‐derived tau ¹⁶ , ²⁰ that might function as a neurodegeneration biomarker in AD. Translation of this IP‐MS assay, that can simultaneously quantify small and big tau isoforms, into blood derived plasma would supplement current plasma BD‐tau and NTA‐tau assays in disease monitoring and staging. Interestingly, exon 6 peptide had better clinical performance compared with exon 4a‐L peptides, highlighting the complex nature of the CSF tau fragments. Exon 6 peptide demonstrated better correlation with worsening cognitive outcomes (see Supplementary Fig S10) as well as CSF MTBR‐tau243 (ρ = 0.55). Tissue‐specific exon 6 splicing independent of other exons (ie, exon 4a) has been documented. ⁹⁵ The exact mechanism of tau exon 6 splicing, independent of exon 4a‐L, and how that correlates with tau pathology in AD requires further investigation.

The enrichment of big tau isoforms in the PNS compared to the CNS suggests that it may be leveraged as a potential biomarker for PNS injury, such as peripheral neuropathies. Currently, there are no assays that can simultaneously distinguish between the CNS and PNS enriched forms of tau. Recent studies have indicated that plasma p‐tau (p‐T181) is elevated in patients with sporadic ALS with predominant lower motor neuron involvement. ²² A recent multi‐center study assessing serum and plasma p‐tau levels in ALS participants has raised concern about the specificity of the p‐tau assays. Interestingly, they found elevation of plasma p‐tau 181 and p‐tau 217 levels in patients with ALS, overlapping with levels seen in clinically confirmed AD cases. ²³ Furthermore, immunohistochemical analysis revealed significantly elevated sarcoplasmic reactivity for p‐tau 181 and 217 in ALS muscle biopsies compared with disease controls. However, these existing assays cannot differentiate between the phosphorylated tau that could be derived from small tau and big tau isoforms. Such distinction of tau isoform would be crucial in discerning their pathophysiological roles in ALS and other pathologies that impact the lower motor neurons and skeletal muscles. Tau hyperphosphorylation has also been documented in retinal tissues from AD and primary tauopathy participants. ⁹⁶ Another report has highlighted tangle like aggregates of hyperphosphorylated tau in the heart tissue of patients with heart failure (HF) and AD, where the predominant isoform of tau is big tau. ⁶⁵ Although these studies investigated the role of HMW tau, they did not demonstrate whether this big tau protein originated form the insertion from either exon 4a‐L or exon 4a‐S between exons 4 and 5. Our assay thus provides a unique opportunity to probe disease specificity and investigate what role these divergent tau isoforms play in the pathophysiology of neurodegeneration across the nervous system where tau is implicated.

Whereas our study provides valuable new insights into the fundamental biology of big tau and isoform distribution across the human nervous system, several limitations should be acknowledged. First, our MS results are derived from bulk tissue soluble homogenates. Although appropriate for peptide sequencing (especially for characterizing the exon junction peptides) as well as quantitative profiling, this method has limitations regarding the spatial and cellular distribution of the big tau across the brain and spinal cord (grey and white matter). Future spatial localization using immunohistochemistry or other spatial omics technologies would be useful to study the functional role of big tau isoforms. Development of human big tau specific antibodies would be useful to confirm the presence of the big tau protein isoforms across different brain regions (cerebrum, brain stem, and cerebellum) as well as its specific cellular expression pattern. Using PCR to conclusively confirm the cellular origin of big tau/4a‐L and 4a‐S in different brain regions will be highly complementary to anatomic profiling. Second, the lack of functional studies exploring the biological roles of exon 4a‐L/4a‐S limits the comprehensive understanding of big tau and its significance in neurodegenerative processes. Third, we used complimentary proteomic tools to investigate the insertion of exon 4a‐L/4a‐S in the HMW human tau isoform, however, our proteomic digestion limits the investigation of simultaneous exon expression (whether exon 4a‐L/4a‐S and exon 6 are mutually exclusive). Complimentary tools such as long‐read sequencing and size‐exclusion chromatography along with intact/top‐down MS could be used to address this in future research. Last, whereas this study evaluates the diagnostic potential of small and big tau species in the CSF in a cross sectional cohort addressing Aβ and tau pathology, further research correlating these distinct tau proteoforms in diverse neuropathological conditions, clinical parameters, and disease progression (tau pathology) is necessary to enhance their clinical relevance and utility in disease management. A more extensive analysis of CSF and plasma big tau isoforms in diseases that involve peripheral tau changes, such as peripheral nerve injury and lower motor neuron injury, should be investigated to assess its diagnostic potential.

In summary, here, we investigated the tau isoforms with and without exon 4a in the human nervous system. Canonical tau isoforms lacking exon 4a account for 99% of the total tau in the human brain. Insertion of exon 4a between exons 4 and 5 generates a larger tau isoform (big tau), that is twice as large compared to common tau (small tau). Exon 4a undergoes alternative 3′ splicing to generate big tau 4a‐L isoforms along with big tau 4a‐S isoforms. We demonstrate through the detection of exon‐bridging peptides that exon 4a‐L is the predominant splice product in humans. Transcriptomic investigation (Isoseq) from 2 different brain regions identified big tau isoforms with exon 4a‐L. Big tau isoforms containing exon 4a‐L display a striking regional distribution, comprising approximately 1% of total tau in the brain but up to approximately 60% in the PNS. Alternative splicing of exons 2 and 3 further expands big tau diversity, generating multiple isoforms with distribution patterns distinct from the well‐characterized small tau isoforms (0 N, 1 N, and 2 N). Notably, increased expression of big tau isoforms coincides with the brain regions that are resistant to tau aggregation, which raises an intriguing question of whether these tau isoforms are protective against tauopathies and requires further investigation. More importantly, our results demonstrate that, whereas brain derived small tau isoforms in the CSF are increased in AD, CSF big tau peptides remain unaltered. These results provide potential targets for future blood‐based biomarker assay development capable of distinguishing between brain derived tau and peripheral tau, which may be crucial for differentiating CNS and PNS diseases that may coincide or overlap in older populations. Together, our findings expand the known tau isoform repertoire and establish a framework for studying CNS‐ and PNS‐enriched tau species in health and disease.

Author Contributions

S.M., C.M.K., N.R.B., C.V.L., C.P., and R.J.B. contributed to the conception and design of the study; R.K.K., N.R.B., K.H., C.V.L., K.F.R., S.K., R.J.P., E.E.F., C.P., J.O., J.M., T.M., C.S., N.G., A.M.G., C.M.K., R.J.B., and S.M. contributed to acquisition and analysis of data; R.K.K., N.R.B., K.H., C.V.L., C.P., J.O., R.J.P., T.M., N.G., C.S., C.M.K., R.J.B., and S.M. contributed to drafting the text or preparing the figures.

Potential Conflicts of Interest

The authors have nothing to report.

Supporting information

Supplementary Figure S1a. Schematic representation of MAPT isoforms of (A) big tau, characterized by the inclusion of exon 4a‐L and exon 6, (B) exclusion of exon 6 and (C) canonical 6 CNS tau isoforms (referred to as “small tau”) generated by alternative splicing of exons 2, 3 and 10, comprising both 3R and 4R forms.

Supplementary Figure S1b. The protein sequence coverage of tau using bottom‐up tandem mass spectrometry (MS/MS) from the human nervous tissue by using combination of enzymatic digestions. The green highlighted regions are the respective peptide identified in the database search in the MSFragger software.

Supplementary Figure S1c. The log transformed relative intensity of all the tau peptides along with the exon 4a‐L and 6 peptides across the human nervous system derived from bottom‐up LC–MS/MS. Compared to (A) brain, the relative intensity of exon 4a‐L are two‐magnitude higher in the (B) spinal cord and (C) sciatic nerve samples, respectively. The tau peptide sequences are numbered based on the big tau with one N‐terminal insert (1 N big tau) and ordered from N‐terminus to C‐terminus (left to right). PRR, proline rich region; MTBR, microtubule binding region.

Supplementary Figure S2. Tandem mass spectrometry (MS/MS) of small and big tau isoform specific peptides. Illustrative HCD‐MS/MS (using NCE 30) spectra of (A) [M + 4H]⁴⁺ m/z 989.7190, 2 N small tau; (B) [M + 4H]⁴⁺ m/z 753.3523, small tau 1 N; (C) [M + 4H]⁴⁺ m/z 606.7873, small tau 0 N; (D) [M + 3H]³⁺ m/z 1104.1780, big tau 1 N and (E) [M + 4H]⁴⁺ m/z 681.8203, big tau 0 N peptides and (F) [M + 2H]²⁺ m/z 449.7351 peptide from tau exon 6. The y and b ions are shown in red and blue, respectively. (continued in next page). Tandem mass spectrometry of tau exon 4 to exon 4‐S, isoform specific junction peptides (synthetic standard). Illustrative HCD‐MS/MS (using NCE 30) spectra of (G) [M + 3H]³⁺ m/z 942.0967, Big tau 0 N‐4a‐S; (H) [M + 3H]³⁺ m/z 1137.5166, Big tau 1 N‐4a‐S; and (I) [M + 4H]⁴⁺ m/z 1089.7560, Big tau 2 N‐4a‐S. NCE, normalized collision energy.

Supplementary Figure S3. (A) Schematic overview of the immunoprecipitation (IP)‐LC–MS/MS workflow used for the detection of exon 4a‐L and exon 4a‐S, big tau specific junction peptides from human nervous tissue using anti‐tau antibody cocktail (HJ8.5, HJ8.7, and Tau 1) conjugated to sepharose beads and compared with the respective synthetic standard peptides. (B) Quantification of exon 4a‐L and exon 4a‐S, big tau peptides containing different N‐terminal insert isoforms (2 N, 1 N, 0 N) in PNS (DRG, n = 2, BP, n = 2, SN, n = 5) and CNS (parietal, temporal, occipital, frontal, cerebellum, and spinal cord, each n = 6) tissue soluble lysate samples. Peptide abundances are expressed as ratios normalized to total (²¹²TPSLPTPPTR²²¹). Bars represent mean ± SEM. Statistical significance is indicated as ****p < 0.0001, **p = 0.0082.

Supplementary Figure S4. Western blot analysis of tau protein isoforms across central and peripheral nervous tissues in humans. Representative immunoblot of (a) Tau‐1 (190–207, 2N4R) and (b) Tau‐5 (210–239, 2N4R) antibodies showing the expression of big tau (~115–90 kDa) and small tau isoforms (~40–65 kDa) peripheral tissues (sciatic nerve, brachial plexus, dorsal root ganglion) and spinal cord tissues. No discernable HMW bands were detected in the brain regions (temporal, parietal, occipital, frontal cortex, cerebellum) using either Tau‐1 and Tau‐5 antibodies. GAPDH (35 kDa) was used as loading control.

Supplementary Figure S5. SDS‐PAGE gel stained with (A) Oriole (BioRad) fluorescent stain to visualize protein bands from brain, spinal cord, cauda equina (CE), brachial plexus (BP), and dorsal root ganglion (DRG) tissue lysates. A schematic diagram illustrates the processing for in‐gel digestion and LC–MS/MS analysis of the gel bands (HMW and LMW). Gel bands corresponding to big tau (75–185 kDa) and small tau (40–70 kDa) were excised for downstream analysis. LC–MS/MS analysis was performed on peptides excised from brain (frontal), spinal cord (cervical), cauda equina (CE), brachial plexus (BP), and Dorsal root ganglion (DRG) fractions. (B) Shown representative chromatogram displays extracted ion chromatograms (xic) for the major fragments of the targeted.

Supplementary Figure S6. Big tau transcripts expression in the brain. (A) Bar‐plot of read counts for all detectable big tau transcripts (0N3R4aL, 1N3R4aL, and 1N4R4aL) in frontal cortex (MFG) and caudate (CAU), each brain is represented by a dot, only brains with at least one big tau transcript are shown. (B) Heatmap of the 6 canonical isoforms (small tau) and the 3 detected big tau isoforms as percentage expression to total sample count in all the brain tissue samples investigated for transcripts.

Supplementary Figure S7. Bar graph of relative ratio of tau exon 4a‐L peptides to total tau peptide (residue 633–642) in 5 different regions of the adult human brain. Cerebellum (n = 6) contains significantly higher compared with other cortical regions (parietal, temporal, occipital, and frontal cortex) of the brain. One‐way ANOVA statistical analysis with multiple comparisons test using Dunnett statistics, ***p < 0.0001; **p ≤ 0.0058; *p ≤ 0.0432.

Supplementary Figure S8. Scatter plots of concentration (ng/ml) in CSF of (A) N‐terminal tau (residue 06–23), (B) residue 25–44, (C) residue 68–87, (D) residue 45–67 (small tau 0 N), (E) residue 602–611 (total tau T181), (F) residue 633–642 (total‐tau T212), (G) pT181/T181 (%), (H) pT217/T217 (%), and (I) MTBR‐tau243 tau peptides from young normal controls (YNCs), asymptomatic (CDR = 0) Aβ‐, asymptomatic (CDR = 0) Aβ+, symptomatic (CDR = 0) Aβ‐ and symptomatic (CDR = 0) Aβ + individuals. The levels of N‐terminal tau‐6, total tau‐181, total tau‐212 and MTBR‐tau243 are significantly increased in the symptomatic stages of the amyloidosis (CDR = 0, Aβ+) compared to asymptomatic stage (CDR >0, Aβ+) that correlates with the disease progression in AD. One‐way ANOVA with multiple comparisons test using Tukey test, ***p < 0.0001, **p ≤ 0.01, *p ≤ 0.05.

Supplementary Figure S9. Scatter plots for (A) big tau‐235 (fmol/ml), (B) big tau‐294 (fmol/ml), (C) big tau‐429 (fmol/ml) and (D) big tau‐452 (fmol/ml) in the CSF of young normal controls (YNCs), asymptomatic (CDR = 0) Aβ‐, asymptomatic (CDR = 0) Aβ+, symptomatic (CDR = 0) Aβ‐ and symptomatic (CDR = 0) Aβ + individuals. One‐way ANOVA with multiple comparisons test using Tukey test, ***p < 0.0001, **p ≤ 0.01, *p ≤ 0.05.

Supplementary Figure S10: Pair‐wise Spearman's rank correlation (ρ) of the CSF biomarkers reveals while exon 4a‐L peptides correlate highly with each other, exon 6 peptide correlates with MTBR‐tau243. The MTBR‐tau243 correlated highly with the total tau peptides (T181 and T212) and less so with the pT217/T217 and T181/T181. Most interestingly, we observed the exon “4a‐L” and “6” tau peptides correlate with age of the individuals, just like MTBR‐tau243 and total tau in the CSF cohort that includes young normal controls, Aβ‐ and Aβ + individuals.

Supplementary Figure S11. Associations between CSF biomakers and MMSE. Scatter plots depicting the association between (A) MTBR‐tau243 (ng/ml), (B) N‐terminus tau06‐23 (ng/ml), (C) CNS 0 N specific tau (ng/ml), (D) pT181/T181 occupancy (%), (E) pT217/T217 occupancy (%), (F) t‐tau181‐190 (ng/ml) and (G) t‐tau212‐221 (ng/ml), (9H) Bigtau235 (fmol/ml), (I) Bigtau429 (fmol/ml), and MMSE in the CSF cohort, color‐coded by clinical diagnosis (CDR) and amyloid status (A−/A+).

Supplementary Figure S12. Association of CSF biomarkers with age in Aβ + and Aβ‐ individuals. (A) MTBR‐tau243 (ng/ml), (B) Bigtau235 (fmol/ml) in both all Aβ + and Aβ‐ individuals. The CSF biomarkers increase in Aβ‐ individuals with different age groups. (C) N‐terminus tau06‐23 (ng/ml), (D) CNS 0 N specific tau (ng/ml), (E) t‐tau181‐190 (ng/ml), (F) pT181/T181 occupancy (%), (G) t‐tau212‐221 (ng/ml), (H) pT217/T217 occupancy (%), (I) MTBR‐tau243 (ng/ml), (J) Bigtau235 (fmol/ml), (K) Bigtau429 (fmol/ml) and (L) Bigtau545 (fmol/ml). Statistical analysis was performed with Kruskal‐Wallis test with false discovery rate correction for multiple comparison with Benjamini‐Hochberg method.

Supplementary Figure S13. Exon 4a‐L big tau N‐insert isoforms (2 N, 1 N, 0 N) were detected and quantified in cerebrospinal fluid (CSF) alongside small tau N‐insert isoforms. The relative abundance of big tau N‐inserts in CSF was approximately two‐orders of magnitude lower than small tau counterparts.

Supplementary Table S1. The list of peptides identified using multi‐enzyme digest. C, carbamidomethyl.

Supplementary Table S2. The list of synthetic peptides from exon 4‐exon 4a‐L and exon 4‐exon 4a‐S bridging regions and their corresponding tryptic peptide used for validation in this study.

Supplementary Table S3. Clinical demographics of brain tissues used for MAPT Targeted isoseq. CAU, caudate; MFG, middle frontal gyrus; RIN, RNA Integrity Number; F, female; M, male; PMI, postmortem interval.

Data Availability

The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository ⁹⁷ with the dataset identifier PXD059267. Raw iso‐seq data files are deposited in GEO: GSE324361. All the data, tau concentrations, and CSF biomarker data presented in this study are available from the corresponding author upon reasonable request, and such arrangements are subject to standard data‐sharing agreements and approval by the institutional review board.