A soluble truncated tau species related to cognitive dysfunction is elevated in the brain of cognitively impaired human individuals.

Neurofibrillary tangles are a pathological hallmark of Alzheimer's disease, and their levels correlate with the severity of cognitive dysfunction in humans. However, experimental evidence suggests that soluble tau species cause cognitive deficits and memory impairment. Our recent study suggests that caspase-2 (Casp2)-catalyzed tau cleavage at aspartate 314 mediates synaptic dysfunction and memory impairment in mouse and cellular models of neurodegenerative disorders. Δtau314, the C-terminally-truncated cleavage products, are soluble and present in human brain. In addition, levels of Δtau314 proteins are elevated in the brain of the cognitively impaired individuals compared to the cognitively normal individuals, indicating a possible role for Δtau314 proteins in cognitive deterioration. Here we show that (1) Δtau314 proteins are present in the inferior temporal gyrus of human brains; (2) Δtau314 proteins are generated from all six tau splicing isoforms, (3) levels of both Casp2 and Δtau314 proteins are elevated in cognitively impaired individuals compared to cognitively normal individuals, and (4) levels of Δtau314 proteins show a modest predictive value for dementia. These findings advance our understanding of the characteristics of Δtau314 proteins and their relevance to cognitive dysfunction and shed light on the contribution of Casp2-mediated Δtau314 production to cognitive deterioration.

occur before or without the formation of NFTs, and that switching off tau expression ameliorates memory impairment even though NFTs remain [12][13][14] . In addition, NFT-bearing neurons in the visual cortex function normally in electrophysiological studies 15 . Taken together, these findings suggest that soluble tau species, but not NFTs, could be responsible for cognitive dysfunction.
A recent study identified Δtau314 16 in the rTg4510 mouse line 12,17 that overexpresses transgenic human tau 0N4R isoform with the proline-to-leucine mutation at amino acid residue 301 (P301L) associated with frontotemporal dementia and parkinsonism linked to chromosome 17 18,19 . Δtau314 are C-terminally-truncated, soluble tau species that are produced from caspase-2 (Casp2)-catalyzed cleavage at aspartate 314 (D314, tau 2N4R isoform numbering system hereafter unless specified). Lowering endogenous levels of Casp2 reduces Δtau314 production and ameliorates memory deficits in rTg4510 mice; whereas rendering tau non-cleavable by an aspartate-to-glutamate mutation at D314 (D314E) blocks tau mislocalization to dendritic spines and attenuates synaptic function impairment in cultured primary neurons expressing the tau P301L mutant 16 . Further, transduction of mice with adeno-associated virus (AAV) carrying this additional tau D314E mutation prevents cognitive deficits caused by tau P301L-carrying AAV-mediated transduction 16 . Together, these findings support Casp2-mediated tau cleavage at D314 as a pathogenic event that leads to cognitive dysfunction. In addition, the relevance of Δtau314 to AD was established in a cohort from the Memory and Aging Project at the Rush University, Chicago, IL 20 , in that levels of Δtau314 proteins are elevated in the disease-vulnerable inferior temporal gyrus (ITG) of the cognitively impaired (CI) individuals compared to the cognitively normal (CN) individuals 16 .
To confirm the existence of Δtau314 proteins in human brains and their relevance to cognitive dysfunction, we analyzed post-mortem ITG specimens in a cohort from the Mayo Clinic Study of Aging at the Mayo Clinic, Rochester, MN. This cohort consisted of ninety persons (twenty-four with amnestic dementia who had relatively pure AD pathology at post-mortem examination (AD dementia), thirty-three individuals who died while still being diagnosed with mild cognitive impairment (MCI) and who had at autopsy primarily AD pathology, and thirty-three individuals who were CN within 24 months before death (twenty-eight of them remaining CN within 18 months before death)). This cohort was analyzed using conventional immunoprecipitation (IP)/Western blotting (WB), combined with IP/mass spectrometry (MS) and an ultra-sensitive, enzyme-linked immunosorbent assay (ELISA). We found Δtau314 proteins derived from all six tau splicing isoforms in ITG and elevated levels of both Casp2 and Δtau314 proteins in the CI individuals (AD dementia and MCI) compared to the CN individuals. This study further characterizes Δtau314 proteins and their relationship to AD and cognition, and it confirms the finding that levels of Δtau314 proteins are elevated in CI individuals in a second cohort. These results contribute to our understanding of AD pathogenesis and development of biomarkers for synaptic deficits.

Results
Antibodies recognize Δtau314 with specificity. We first assessed the binding specificity of the H1485 and 4F3 antibodies developed to target the truncated tau proteins ending C-terminally at D314 (Table 1). For this, we synthesized the full-length human tau 0N4R splicing isoform (fl-tau) and Δtau421 and Δtau314, two truncated tau proteins ending C-terminally at aspartate 421 and D314, respectively. Both antibodies detected Δtau314, but not fl-tau or Δtau421, though the polyclonal H1485 also recognized several endogenous proteins present in the synthetic system (Fig. 1). The results thus indicate that both H1485 and 4F3 recognize Δtau314 with specificity.
Δtau314 proteins are present in post-mortem human brains. We asked whether Δtau314 proteins are present in the ITG of human brains. We performed IP/WB analysis of post-mortem specimens to detect the tau-13-immunoprecipitated proteins that are reactive to H1485 or the biotin-conjugated 4F3 antibody (Table 1). We identified protein bands that electrophoretically migrated at ~45-, ~40-, ~35-, and ~30kilo-Dalton (kDa) (Fig. 2a), potentially representing Δtau314 proteins that are produced from six tau splicing isoforms. To confirm this, we performed liquid chromatography (LC)-tandem MS (MS/MS) analysis of the trypsin-digested peptides extracted from gel pieces containing the H1485-reactive proteins. We identified the peptide HVPGGGSVQIVYKPVD, the C-terminal fragment of Δtau314 proteins specifically produced from the tau isoforms harboring four microtubule-binding motifs (4R-tau), from the gel pieces containing the ~45-, ~40-, and ~35-kDa proteins ( Fig. 2b-d, Supplementary Figs. S1 and S2); in parallel, we identified the peptide VQIVYKPVD, the C-terminal fragment of Δtau314 proteins produced from the tau isoforms harboring three microtubule-binding motifs (3R-tau), from the gel pieces containing the ~40-, ~35-, and ~30-kDa proteins ( Fig. 2e-g, Supplementary Figs. S1 and S2).
To further verify that Δtau314 proteins are products of all six tau isoforms, we immunoprecipitated the intracellularly-enriched (IC) fraction of ITG homogenates using H1485 and probed blots using monoclonal antibodies specifically targeting the 4R-and 3R-tau, respectively ( Table 1). The binding specificities of antibodies to the 4R-and 3R-tau isoforms were verified (Fig. 3a). In randomly selected samples, we showed that the ~40-, ~35-, and ~30-kDa proteins were detected by the monoclonal tau-5 antibody directed against an epitope present in all six tau isoforms (Fig. 3b). The ~40-kDa proteins are likely comprised of the Δtau314 proteins that are produced from the 1N4R and 2N3R tau isoforms; the ~35-kDa, the 0N4R and 1N3R tau isoforms; and the ~30-kDa, the 0N3R tau isoform. The anti-3R-tau antibody detected all the three proteins (Fig. 3c), and the anti-4R tau antibody detected the ~40-and the ~35-kDa (Fig. 3d). We did not detect the ~45-kDa protein that presumably matches the Δtau314 protein produced from the 2N4R tau isoform using either tau-5 or anti-4R tau as the detection antibody. This failure is likely due to 1) the scarce level of this ~45-kDa Δtau314 protein present in the studied brain tissue and 2) detection interference from non-specific protein bands. Taken together, results of IP/WB substantiate the finding that Δtau314 proteins in human ITG are produced from both the 4R-and the 3R-tau isoforms.
Levels of Δtau314 proteins are elevated in the brain of cognitively impaired individuals. We asked how levels of Δtau314 proteins differ in the AD-affected ITG of a cohort containing CI (including individuals with MCI and AD dementia patients who had no other pathologies other than plaques and tangles) and CN individuals (Supplementary Table S1). First, we compared the demographic and neuropathological features of the AD dementia, MCI, and CN individuals. We found that neither the sex distributions nor the post-mortem intervals (PMIs) differed among the three groups; however, the ages at death of the AD dementia patients were higher than the CN individuals by 6% (Table 2, Supplementary Fig. S3). As expected, these three different diagnostic groups are distinct in Braak stage neuropathology 21,22 (Table 2, Supplementary Fig. S3) and neuritic plaque density 23 (Table 2, Supplementary Fig. S3).

Antibody
Host/Isotype Epitope a Source Usage  Table 1. Antibodies used in this study. a Amino acid residues of tau protein are counted using the 2N4R splicing isoform numbering system. Ms = mouse, Rb = rabbit, conc. = concentration, n.d. = not determined.

Figure 1.
The antibodies H1485 and 4F3 recognize Δtau314 proteins with specificity. Three µL of the rabbit reticulocyte lysates containing synthetic full-length 0N4R tau (fl-tau), Δtau421, or Δtau314 proteins were subjected to Western blotting. The blots were probed by the monoclonal antibody tau-13 directed against an N-terminal epitope (a.a. 20-35) of human tau (left panel) and the polyclonal antibody H1485 (middle panel) and the monoclonal antibody 4F3 (right panel), both of which targets tau epitopes ending C-terminally at D314. While tau-13 detects fl-tau (hash mark), Δtau421 (arrowhead), and Δtau314 (asterisk), both H1485 and 4F3 recognize only Δtau314 (asterisks). Notably, proteins migrating similarly as Δtau314 in samples containing Δtau421 and fl-tau in the tau-13-probed blot may represent the tau proteins of the similar size to, but different from, Δtau314. Bands other than the ~35-kDa are likely non-specific, as all these bands appear in a sample in which no tau synthesis DNA template was included. The three full-length blots were prepared from three different 10-20% Tris-Tricine precast gels, and the same exposure time was applied to the development of the three blots. There was no grouping within each blot.
www.nature.com/scientificreports www.nature.com/scientificreports/ We then performed IP/WB to quantitatively analyze levels of Δtau314 proteins in the CI and the CN individuals. We showed that levels of Δtau314 proteins, detected by H1485 (Fig. 4a,d, Table 3) and the biotin-conjugated 4F3 (Fig. 4b,e, Table 3), in the CI individuals were 1.6-and 1.4-fold higher than in the CN individuals, respectively. Levels of Δtau314 proteins in the AD dementia patients and the individuals with MCI did not differ ( Supplementary Fig. S4, Supplementary Table S2). Not surprisingly, levels of the H1485-detected Δtau314 proteins were highly correlated with levels of the 4F3-detected (Fig. 4f), indicating that these two antibodies share antigen-recognizing specificity. Next, to compare the proportion of soluble total tau (T-tau) comprised of Δtau314 proteins between groups, we normalized levels of Δtau314 proteins to levels of T-tau proteins that are immunoreactive to tau-13. Interestingly, levels of T-tau proteins were 1.2-fold higher in the CN individuals than in the CI individuals ( Fig. 4c,g, Supplementary Fig. S5, Table 3), and levels of T-tau proteins in the AD dementia patients and the individuals with MCI did not differ ( Supplementary Fig. S4, Supplementary Table S2). The lower levels of T-tau proteins in the AD dementia patients and the individuals with MCI are unrelated to neuronal loss since levels of βIII-tubulin, a molecular proxy for neuron number, did not differ between the CN and the CI individuals (Table 3).
When normalized to levels of T-tau proteins, levels of H1485-and 4F3-reactive Δtau314 proteins in the CI individuals were 1.8-and 1.6-fold higher than in the CN individuals, respectively (Fig. 4h,i, Table 3). Levels of Δtau314 proteins in the AD dementia patients and the individuals with MCI were comparable ( Supplementary  Fig. S4, Supplementary Table S2).
Next, we performed protein quantification using Quanterix single molecule array (simoa), an ultra-sensitive ELISA. We showed that levels of the 4F3-captured, BT2/HT7-detected Δtau314 proteins (Table 1)   Using an anti-3R tau antibody (middle panel), three of the six tau-13-reactive bands (single asterisks) were detected, representing the three 3R-tau isoforms. In parallel, an anti-4R tau antibody (right panel) detected the other three tau-13-reactive bands (double asterisks), representing the three 4R-tau isoforms. The three fulllength blots were prepared from three different 10% Tris-HCl precast gels, and the same exposure time was applied to the development of the three blots. (b-d) Immunoprecipitation/Western blotting (IP/WB) analyses show that the H1485-immunoprecipitated Δtau314 proteins from representative samples (three cognitively normal (CN) individuals and six cognitively impaired (CI) individuals including three individuals with mild cognitive impairment (MCI) and three Alzheimer's disease dementia patients (AD dementia), randomly selected within each group, and 300 µg of proteins per sample) were detected by the biotin-conjugated tau-5 antibody that recognizes all six human tau isoforms (b), the anti-4R antibody (c), and the anti-3R tau antibody (d) (arrows), respectively. Hash marks, non-specific bands. rbtIgG/br. ex., rabbit IgG-immunoprecipitated brain extracts. Ab only, H1485 antibody IP without brain extracts. These full-length blots were prepared from different 10% Tris-HCl precast gels, and different exposure times were applied to the development of the three blots to achieve the best signal-to-noise ratio for each blot. There was no grouping within each blot. higher in the CI individuals than in the CN individuals (Fig. 4j, Table 3), and levels of these Δtau314 proteins were comparable between the AD dementia patients and the individuals with MCI ( Supplementary Fig. S4, Supplementary Table S2), which are consistent with the IP/WB findings. Notably, the Δtau314 protein species detected by the two methods may not be fully identical as N-terminally truncated Δtau314 proteins, if present, can also be included in the quantification using ELISA. We also quantified T-tau proteins using an antibody (Q C ) targeting tau mid-region as the capture reagent and an antibody (Q D ) targeting N-terminal tau as the detection reagent (Table 1). We showed that levels of T-tau proteins did not differ between the CN and the CI individuals (Table 3) or between the AD dementia patients and the individuals with MCI (Supplementary Table S2), which is discrepant to the results of IP/WB. This discrepancy may arise from differences in the capture and the detection antibody combination between IP/WB and ELISA (Table 1), which could result in measuring two different sets of tau proteins by the two methods. Not unexpectedly, levels of Δtau314 proteins, when normalized to levels of T-tau proteins, did not differ between the CN and the CI individuals (Table 3) or between the AD dementia patients and the individuals with MCI (Supplementary Table S2).
Further, we asked how levels of Δtau314 proteins measured by IP/WB (normalized to levels of T-tau proteins) correlate with levels of Δtau314 proteins measured by ELISA. We found that levels of Δtau314 proteins measured by the two methods correlated well (Fig. 5), despite the possibility that different Δtau314 proteins may be measured by the two experimental approaches.
We then asked whether the demographic and neuropathological features of our participants were associated with levels of Δtau314 proteins (normalized to levels of T-tau proteins). Interestingly, we identified that the normalized levels of Δtau314 proteins correlated with not only amyloid pathology densities ( Supplementary Fig. S6), but also, to a lesser degree, Braak stages ( Supplementary Fig. S7). This suggests a potential link between Δtau314 proteins and AD pathology. In addition, the ages at death of participants did not correlate with the normalized Δtau314 protein levels ( Supplementary Fig. S8). The normalized levels of Δtau314 proteins were also comparable between the female and the male individuals ( Supplementary Fig. S8). PMIs correlated notably with the normalized Δtau314 protein levels ( Supplementary Fig. S8), indicating that PMI is likely a confounder that may affect levels of Δtau314 proteins in the studied participants. Sample size, N (% within group)   21 was assigned based on a procedure previously described 22 . Specifically, 0 = AD-type neurofibrillary degeneration not present, 1 = Braak stage I, 2 = Braak stage II, 3 = Braak stage III, 4 = Braak stage IV, 5 = Braak stage V, 6 = Braak stage VI. d The consortium to establish a registry for Alzheimer's disease (CERAD) score for density of neocortical neuritic plaque 23 was assigned based on a procedure previously described 22 . Specifically, 0 = no neuritic plaques, 1 = sparse neuritic plaques, 2 = moderate neuritic plaques, 3 = frequent neuritic plaques. www.nature.com/scientificreports www.nature.com/scientificreports/ We then reassessed levels of tau protein species following adjustment for age at death, sex, and PMI. When adjusted for the three demographic factors, levels of Δtau314 proteins measured by IP/WB, whether normalized to T-tau levels or not, remained higher in the CI individuals than in the CN individuals (Table 3). Thus, these data confirm that levels of Δtau314 proteins are elevated in the ITG of the CI individuals. Notably, the reassessed levels of Δtau314 proteins measured by ELISA, when normalized to T-tau levels, were no longer significantly different between the CI individuals and the CN individuals (Table 3). This is likely due to differences in T-tau species measured by ELISA versus IP/WB as discussed above.

Levels of Δtau314 predicts cognitive impairment of individuals.
To assess the strength of the associations between cognitive impairment and Δtau314 proteins, we generated receiver operating characteristic (ROC) curves in which participants were categorized as CN or CI. The area under a curve (AUC) values were 0.67 and 0.69 for levels of H1485- (Fig. 6a) and 4F3- (Fig. 6b) immunoreactive Δtau314 (normalized to levels of T-tau), respectively, and the AUC value was 0.68 for levels of Δtau314 determined by ELISA (Fig. 6c). The AUC values for Braak staging (Fig. 6d) and neuritic plaque density (Fig. 6e) were 0.84 and 0.75, respectively. Statistical analyses of ROC curves showed that the AUC values of levels of Δtau314 proteins were lower than that of Braak staging but comparable to that of neuritic plaque density ( Table 4).
Levels of Casp2 are higher in the brain of cognitively impaired individuals. We sought to understand the extent to which Casp2 is involved in the production of Δtau314 proteins in the CI individuals. At present, there is no reliable method for measuring Casp2 activity in human brain tissue. Therefore, we measured levels of Casp2 detected by a monoclonal anti-Casp2 antibody directed against a C-terminal epitope (Table 1), which were previously shown to be elevated in Lewy body dementia 24 , Huntington's disease 25 , and AD 26 .
Levels of Casp2 proteins, normalized to levels of total proteins, were 1.3-fold higher in the CI individuals than in the CN individuals (Fig. 7, Supplementary Fig. S9, Table 3) and trended higher following adjustments for age at death, sex, and PMI of brain tissue harvest (Table 3). This finding supports the active involvement of Casp2 in Δtau314 production in persons affected with AD.

Mann-Whitney
Two-tailed, unpaired t-test a Multiple linear regression a,b Casp2:total proteins p = 0.020 p = 0.040 p = 0.089 Table 3. A statistical comparison of protein levels of cognitively impaired and cognitively normal individuals. a The two-tailed, unpaired t-tests and multiple linear regressions were performed on the log-transformed outcomes. b Multiple linear regression was used to analyze protein levels that were adjusted for age at death, sex, and PMI of brain tissue harvest of individuals.

Discussion
In this study, we detected and measured the levels of Δtau314 proteins, a C-terminally-truncated, soluble tau species whose production is responsible for mouse memory impairment and synaptic dysfunction 16 in the ITG, one of the earliest brain regions affected by AD pathology 27 . In addition, we showed that Δtau314 proteins are derived from all six tau splicing isoforms, that levels of both Casp2 and Δtau314 proteins are higher in cognitively impaired than in normal individuals, and that levels of Δtau314 proteins are moderately predictive of cognitive impairment.
Our results support the findings of a previous study 16 , in which similar experimental procedures and quantitative biochemical analyses were utilized, and levels of Δtau314 proteins were measured in the same brain structure (i.e., ITG). Specifically, in the Zhao et al. study, levels of Δtau314 proteins in the CI individuals are 1.4-fold higher than in the CN individuals ( Supplementary Fig. S10), and levels of T-tau proteins are 1.5-fold lower in the CI individuals than in the CN individuals ( Supplementary Fig. S10), resulting in a 2.0-fold higher Δtau314:T-tau ratio in the CI individuals than in the CN individuals ( Supplementary Fig. S10). However, unlike the Zhao et al. study, we did not normalize levels of Δtau314 and tau proteins to levels of βIII-tubulin because levels of Δtau314 proteins correlated inversely with levels of βIII-tubulin (Supplementary Fig. S11).  www.nature.com/scientificreports www.nature.com/scientificreports/ When measured by IP/WB, the levels of soluble T-tau proteins in the CI individuals were lower by approximately 20% than in the CN individuals, and yet quantification of βIII-tubulin levels indicates no obvious difference in the degree of neuronal loss between the two groups. This may be related to a higher degree of neurofibrillary tau pathology in the CI individuals than in the CN individuals ( Supplementary Fig. S3b). It is reported that higher severity of neurofibrillary tau pathology associate with lower soluble tau levels in cortices of patients with AD dementia, but such an association is not obvious in age-matched controls 28 . While neuronal loss and age-related decrease in tau production may lower soluble tau levels in elderly CN individuals, a further loss of soluble tau proteins may be linked to the presence of neurofibrillary tau pathology in the studied cortical region of CI individuals. This is likely accomplished, at least partially, through reallocation of tau proteins from the soluble to the insoluble pool 28 .
The levels of soluble T-tau proteins, to which levels of Δtau314 proteins are normalized, show discrepancy between IP/WB and simoa. In our opinion, this is largely due to difference in experimental settings. Specifically, levels of full-length tau proteins were measured in direct WB; however, the use of an antibody targeting a mid-region tau epitope as the capture reagent in simoa likely results in measuring additional C-terminally truncated tau species besides the full-length proteins. To minimize discrepant results between IP/WB and simoa, the two assays should be set up to allow measuring virtually, if not exactly, the same proteins and their variants, and it is reasonable to use the same set of capture and detection reagents in IP/WB and simoa for quantification of proteins of interest, though the accessibility of antibodies to their targeted epitopes may differ under the denaturing WB versus the non-denaturing simoa detection conditions (see below for more discussion).
Although full-length tau isoforms are generally considered highly soluble proteins free of structure, recently solved atomic resolution structures of tau microtubule-binding domain in complex with cytoskeleton proteins 29,30 suggest that tau adopts particular tertiary structures upon interacting with its binding partners in biological systems. Under non-denaturing experimental conditions like simoa and immunoprecipitation, it is conceivable www.nature.com/scientificreports www.nature.com/scientificreports/ that certain tau epitopes targeted by antibodies were inaccessible to the capture/detection reagents, resulting in inaccurate quantification and discrepant outcomes between different types of assays. Despite this, the consistent findings of Δtau314 proteins between IP/WB and simoa using 4F3 under different schemes (the capture reagent of simoa under non-denaturing conditions versus the detection reagent of IP/WB under denaturing conditions) suggest the issue of epitope mask does not apply to this Δtau314-specific monoclonal antibody. However, it may affect the measurement of soluble T-tau. To minimize the inaccuracy and inconsistency of immunoassay, protein quantities may need to be more comprehensively determined using available reagents that target different regions of proteins of interest, and assays may need to be carried out under light denaturing conditions to rid or attenuate the influence of epitope mask.
It is well established that tau proteins are digested by multiple proteases to produce various types of cleavage products (for reviews, see, for example 5 ). A growing body of evidence supports the presence of truncated tau species produced in brains of individuals with dementia and their relevance to cognitive dysfunction [31][32][33][34][35][36][37] . Further, recent studies addressed the contribution of protease-mediated tau cleavage to synaptic dysfunction and animal behavioral abnormalities. There are other pathogenic pathways triggered by the proteolysis of tau besides Casp2-mediated tau cleavage at D314 16 . For example, Zhang et al. showed that asparagine endopeptidase (AEP)-mediated tau cleavage at asparagines 255 and 368 is responsible for synaptic dysfunction and cognitive deficits of tau transgenic mice 38 modeling frontotemporal dementia with parkinsonism [39][40][41] , and that the tau cleavage product ending C-terminally at asparagine 368 is present in AD brains 42 . Indeed, our IP/MS analysis of tau proteins extracted from brain tissue identified both AEP-mediated cleavage products ( Supplementary  Fig. S12). Interestingly, besides peptides ending at the C-termini of Δtau314 species and AEP-cleaved tau products, using IP/MS we did not detect any other tau fragments that match the C-termini of truncated products (e.g., Δtau421 43 , tauΔCsp-6 44 , and C-terminally truncated tau at glutamate 391 45 ) comprised of paired helical filaments of tau pathology in human brains. This failure is possibly due to the scarcity of these aggregation-prone tau proteins in the prepared brain homogenates.
Of particular note, we need to bear in mind the following points when interpreting the results. First, that the levels of both Casp-2 and Δtau314 proteins in the CI individuals are higher than in the CN individuals and that levels of Δtau314 proteins moderately predict cognitive impairment are correlative findings. Therefore, these results do not necessarily mean that Casp-2 or Δtau314 proteins play a causative role in deteriorating cognition. Rather, they suggest that Casp-2 and Δtau314 proteins are potentially involved in molecular pathways that lead to cognitive impairment. Interestingly, elevated levels of Casp-2 and Δtau314 proteins have been reported in multiple neurodegenerative disorders other than AD dementia [24][25][26] , thus indicating that Casp-2 and Δtau314 proteins may play a role in distinct disease processes. What remains unclear is whether the elevation of Casp-2 and Δtau314 proteins initiates or is secondary to neuropathologies and/or disease symptoms. Second, although beyond the scope of this study, the disease relevance of the finding that Casp-2-mediated Δtau314 production triggers cognitive impairment needs to be further validated. The identification of the presence of Δtau314 proteins and the role of Casp-2-mediated Δtau314 production in affecting cognition was carried out initially using mouse models that overexpress human tau transgene, a circumstance distinct from pathophysiological conditions of neurodegenerative disorders. Although we have subsequently confirmed the presence of Δtau314 proteins in the brain of individuals with varied tauopathies 16,24,25 , future studies are required to validate the cognition-affecting role of Casp-2-mediated Δtau314 production in systems that better model the diseases (e.g., human tau knock-in mice). Third, the possible physiological function of Casp-2-mediated Δtau314 production warrants an investigation. A growing body of evidence indicates that post-translationally modified tau (e.g., phosphorylated tau) is essential to neuronal development (e.g., neurite polarity and axon outgrowth) and normal cellular activities (e.g., axonal transport and synaptic plasticity) (for reviews, see for example 46,47 ). While Casp-2-mediated Δtau314 production impairs cognition in tau overexpression models, and Δtau314 proteins are associated with CI individuals with long-standing pathology, that Casp-2-mediated Δtau314 production exerts physiological effects cannot be ruled out.
We were aware of a low sample size (n = 55 for the CI group and n = 30 for the CN group) used in the Zhao et al. study 16 ; unfortunately, we have maximized the number of available individuals that met the criteria for specimen selection in that cohort. We, therefore, took an alternative strategy by conducting a similar investigation on a different cohort, hence this study. We admitted that conclusions of this study were drawn from a small population as well; however, the consistent findings from both cohorts increase the likelihood that our findings hold true in a larger population. In addition, we are currently extending the study by investigating the connection between levels of Δtau314 proteins in ITG and various functional and structural biomarkers derived from longitudinal studies of CN elderly individuals. Results of this ongoing study, with a sample size of nearly one hundred, would help us better understand the correlation of brain levels of Δtau314 proteins to the degree of progression in pathology and the rate of deterioration in cognitive function.
Given the minute portion of the brain in which Δtau314 measurement was made, the moderate associations between Δtau314 and clinical status is important. One limit of this study is that levels of Δtau314 proteins were only measured in inferior temporal gyrus. As the distribution of tau pathology varies dramatically among brain regions 48 , quantitative analysis of Δtau314 proteins in other (AD dementia-affected) regions, as one of our future study subjects, may help us better understand the connection between this tau fragment and cognition. In addition, simoa for quantification of Δtau314 proteins in lumbar puncture cerebrospinal fluid (CSF) and blood is currently under development. Successful measurement of Δtau314 protein levels in biological fluids of living persons allows their quantities to be tracked longitudinally and their predictive value for cognitive impairment to be better assessed. Further, the possibility of using Δtau314 proteins, alone or together with other established biomarkers, for dementia disease diagnosis can be investigated.

Scientific RepoRtS |
(2020) 10:3869 | https://doi.org/10.1038/s41598-020-60777-x www.nature.com/scientificreports www.nature.com/scientificreports/ conclusions Overall, our results confirm-in a second cohort-the presence of Δtau314 proteins in human brains, and the finding that levels of Δtau314 species are elevated in elderly CI individuals. In parallel, we showed elevated levels of Casp2 in CI individuals. We also validated 4F3, a novel Δtau314-specific monoclonal antibody, and we developed a highly sensitive ELISA for measuring Δtau314 levels using this antibody. We showed that Δtau314 proteins were generated from all six tau splicing isoforms. Finally, within the limits of our MS detection method, there appear to be only two classes of N-terminally intact, C-terminally truncated soluble tau species in the brain-those cleaved by Casp2 and AEP. These findings add to our knowledge the composition of Δtau314 proteins and reveal the relationships between Δtau314 proteins and cognition and neuropathology of individuals with AD pathology. Moving forward, the following subjects warrant further investigation: the topographic distribution of Δtau314 proteins in various anatomical brain regions of individuals at different stages of AD, the relationships of Δtau314 proteins to AD risk factors (e.g., aging and Apolipoprotein E genotype), the detection and quantitative analysis of Δtau314 proteins in biological fluids (e.g., lumbar puncture CSF and blood), and the correlation between catalytically active Casp2 and Δtau314. These studies would advance our understanding of the contribution of Casp2-mediated tau cleavage at D314 to disease pathogenesis and the potential use of Δtau314 proteins as molecular diagnostic markers of synaptic dysfunction.

Methods
In vitro synthesis of human tau proteins. Human tau proteins was synthesized in vitro as previously
For the detection and quantitative analysis of Casp2, aqueous brain protein extracts were prepared based on a protocol previously published 50 . Specifically, human tissue specimens of ~150 mg (wet weight) were cut, transferred to 5 volumes (1 g per 5 mL) of ice-cold extraction buffer A (25 mM Tris-HCl (pH 7.4), 140 mM NaCl, and 3 mM KCl with the protease and phosphatase inhibitors mentioned above), and homogenized using a Dounce homogenizer. The resulting material was centrifuged for 90 min (16,100 g; 4 °C). The supernatant was subsequently depleted of endogenous IgG with Protein G Sepharose 4 Fast Flow beads (GE Healthcare) and stored at (2020) 10:3869 | https://doi.org/10.1038/s41598-020-60777-x www.nature.com/scientificreports www.nature.com/scientificreports/ −20 °C until further use. As a control, brains of C57BL/6J and Casp2 KO mice were homogenized using ice-cold extraction buffer B (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% (v/v) NP-40, and 0.5% (w/v) deoxycholate with the protease and phosphatase inhibitors mentioned above). The resulting material was centrifuged for 90 min (16,100 g; 4 °C). The supernatant was subsequently stored at −20 °C until further use.
To understand which tau isoforms produce Δtau314 proteins, 300 μg of brain proteins from IC extracts of representative CN individuals, individuals with MCI and AD dementia patients were immunoprecipitated with 4 µg of the custom-produced rabbit polyclonal antibody H1485 (New England Peptide, Gardner, MA) (or rabbit IgG as negative control) using the protocol described above.
To understand which tau isoforms produce Δtau314 proteins, blots were probed with the tau-5 and the anti-3R and 4R tau antibodies as described above.
To measure levels of T-tau proteins, 10 μg of brain proteins from IC extracts were electrophoretically separated by SDS-PAGE using Criterion 10% Tris-HCl Precast gels (Bio-Rad) and then transferred onto nitrocellulose membranes. Blots were probed with tau-13 (MilliporeSigma; 1:30,000, final concentration not determined). Following 5 × 5-min washes in wash buffer, membranes were incubated at room temperature with the HRP-conjugated anti-mouse IgG (Thermo Fisher Scientific; 1:200,000, final concentration = 5 ng/mL) for 1 hr. The membranes were then washed again (5 × 5-min in wash buffer) prior to being developed.
To measure levels of βIII-tubulin, a protein almost exclusively expressed in neurons, 10 μg of brain proteins from IC extracts were electrophoretically separated by SDS-PAGE using Criterion 10% Tris-HCl Precast gels (Bio-Rad) and then transferred onto nitrocellulose membranes. Blots were probed with the monoclonal anti-βIII-tubulin antibody (MilliporeSigma; 1:200,000, final concentration = 5 ng/mL). Following 5 × 5-min washes in wash buffer, membranes were incubated at room temperature with the HRP-conjugated anti-mouse IgG (Thermo Fisher Scientific; 1:200,000, final concentration = 5 ng/mL) for 1 hr. The membranes were then washed again (5 × 5-min in wash buffer) prior to being developed.
In addition, total protein levels were used as a loading control for normalization of levels of Casp2. Proteins transferred to nitrocellulose membranes were revealed following the manufacturer's instruction before blots were probed with anti-Casp2.
Western blots were developed using the West Pico electrochemiluminescence (ECL) detection system (Thermo Fisher Scientific), except for the blots probed with biotin-conjugated 4F3, which were developed using the West Femto detection system (Thermo Fisher Scientific). The blots probed with anti-Casp2 were developed using the Clarity Max Western ECL detection system (Bio-Rad). The Kodak Scientific Imaging film X-OMAT Blue XB (PerkinElmer, Bostin, MA) was used to reveal the blots except for total proteins and Casp2, which were revealed using Image Lab 6.0 (Bio-Rad). For each blot, a series of exposures, ranging from 1 sec to 5 min, was used to ensure that bands of interest fell within the linear range of detection. Signal intensities were quantified densitometrically using Optiquant (Packard Cyclone, Perkin-Elmer Life Sciences Inc., Boston, MA) except for levels of total proteins and Casp2, which were quantified using Image Lab 6.0.
Experimenters performing (IP/)WB were blind to the demographic and clinical features of participants. The mean levels of Δtau314 proteins and βIII-tubulin were determined from two independent experiments, and the mean levels of T-tau proteins were determined from three independent experiments. The mean protein levels were used for statistical analysis.
Antibody immobilization on protein G magnetic beads. The tau-13 antibody was immobilized as previously described 52 . Briefly, the antibody was covalently linked to Protein G magnetic beads (Thermo Fisher Scientific; 200 µg of antibody per 1 mL of bead slurry) using dimethyl pimelimidate (Thermo Fisher Scientific) as the crosslinker. Prior to use, beads were pre-treated with IP buffers A and B to wash off any antibody not covalently linked to beads. To monitor any shedding of antibody from the beads, the beads were treated in elution buffer (100 mM glycine-HCl (pH 2.8) and 1% (w/v) n-Octyl β-D-thioglucopyranoside (OTG)). The eluate was collected and analyzed by WB using HRP-conjugated goat-anti-mouse IgG (Jackson ImmunoResearch Laboratories Inc., West Grove, PA).
Preparation of samples for mass spectrometry. IC protein extracts (0.9 mg) of a randomly selected specimen were incubated overnight at 4 °C with tau-13-bound Protein G-coupled magnetic beads. Beads were then washed twice with wash buffer 1 (0.1% (w/v) OTG in IPDB (pH 7.4)) and subsequently twice with wash buffer 2 (1% (w/v) OTG in IPDB (pH 7.4)). Proteins were eluted using 30 µL of elution buffer at room temperature by agitation for 5 min. Elution was carried out three times, and the eluates were mixed and allocated into three parts, which accounted for 10%, 45%, and 45% of the total volume, respectively. These allocated eluates were size-fractionated on a Criterion 10% Tris-HCl Precast gel (Bio-Rad) and then cut into three pieces. The piece of gel containing 10% of the eluates was subjected to WB using H1485 as the detection antibody, during which time a piece of gel containing 45% of the eluates lanes was stored in a moist chamber at 4 °C, whereas the other piece was subjected to silver stain using a SilverXpress silver staining kit (Thermo Fisher Scientific) following the manufacturer's instruction. After WB, the untreated gel pieces were overlaid on the film record using the molecular weight standards for alignment, and the pieces of unstained gel overlaying the bands of interest (migrating at ~45, ~40, ~35, and ~30 kDa) were excised. In parallel, the bands of interest (migrating at ~45, ~40, ~35, and ~30 kDa) shown in the gel piece with silver stain were excised as well.
Mass spectrometry. In-gel trypsin digestion. In-gel trypsin digestion was performed according to a previously published procedure 53 . Briefly, gel pieces were cut into a size of approximately 2 mm 2 × 1 mm (thickness of the gel), washed with addition of wash solution (50 mM ammonium bicarbonate (pH 7.8) and 50% (v/v) acetonitrile) at room temperature for 15 min, and then aspirated. The washing and aspiration steps were repeated once. The gel pieces were then incubated for dehydration with 100% (v/v) acetonitrile at room temperature for 1 min. The acetonitrile was aspirated, and buffer 1 (50 mM ammonium bicarbonate (pH 7.8) and 10 mM dithiothreitol) was added to the gel pieces and incubated at 56 °C for 1 hr. Buffer 1 was then aspirated, and 55 mM iodoacetamide was added and incubated with gel pieces at room temperature in the dark for 30 min to alkylate cysteine residues. The solution was then aspirated, and the gel pieces were washed twice at room temperature with wash solution. After washing, 100% (v/v) acetonitrile was incubated with the gel pieces at room temperature for 1 min and then aspirated. Trypsin solution was made (5 ng/µL sequencing grade trypsin (Promega) in buffer 2 (50 mM ammonium bicarbonate (pH 7.8) and 5 mM calcium chloride)) and added to each sample to cover the gel pieces. The samples were incubated on ice for 10 min, after which any non-absorbed trypsin solution was aspirated. Buffer 2 was added to cover the gel pieces. Samples were placed in a warm air incubator at 37 °C for 16 hr. The solution from each sample was transferred to a new 1.5 mL microcentrifuge tube, and 50% (v/v) acetonitrile/0.3% (v/v) formic acid was added to each sample for further extraction of peptides from gel pieces. The same sample extracts were pooled, and the process was repeated with 75% (v/v) acetonitrile/0.3% (v/v) formic acid. The pooled digests were frozen at −80 °C and dried in a speed vac. The resolubilized peptide mixtures were desalted using the Stage Tip technique with 3 M Empore styrenedivinylbenzene extraction disks according to a previously published protocol 54 . The final peptide mixtures were dried in vacuo.
Mass spectral database search. Peaks Studio 56 8.5 (Bioinformatics Solutions, Inc, Waterloo, Ontario, Canada) was used for interpretation of MS/MS (mass spectra) and protein inference. Search parameters were: human (taxonomy ID 9606) protein sequence database from UniProt (http://www.uniprot.org/) downloaded on December 13, 2016 concatenated with the common lab contaminant database from http://www.thegpm.org/ crap/; precursor mass error tolerance: 50.0 parts per million (ppm); fragment mass error tolerance: 0.1 Dalton; precursor mass search type: monoisotopic; no enzyme specificity; variable modifications: methionine oxidation and di-oxidation, cysteine carbamidomethylation, pyroglutamic acid, and protein N-terminal acetylation; maximum variable modifications per peptide: 2; false discovery rate calculation: On; spectra merge: Off; no charge state correction; and spectral filter quality: >0.65.
Mass spectral data interpretation. Support for the detection of peptides from each supporting MS/MS spectrum was based on: 1) a minimum of five consecutive b-or y-type peptide fragment ions, 2) 1% peptide and protein false discovery rate threshold, and 3) precursor mass accuracy <7 ppm.
Mass spectral data presentation. Extracted ion chromatogram (XIC) precursor peptide profiles for MS identification of Δtau314 proteins were plotted in MZmine 2.29 56 , and XIC fragment ion profiles for MS identification of Δtau314 proteins were generated from Skyline 4.1 viewer 57 .
Quantitative measurements of Δtau314 proteins were performed using 4F3-coated magnetic beads for protein capture and a combination of biotin-conjugated BT2/HT7 antibodies for detection. For the quantification assay, serially diluted recombinant human Δtau314 (tau 0N4R derivative) protein samples were used to generate the calibration curve. The lower limit of detection (LLOD), calculated as 2.5 standard deviations above the mean of background signal, was approximately 0.1 pg/mL. The lower limit of quantification (LLOQ) was determined to be approximately 1 pg/mL. The quantification of Δtau314 protein levels of all the studied individuals were carried out within the linear range of the calibration curve.
Soluble T-tau proteins were quantitatively measured using the Simoa Tau 2.0 reagent kit (Quanterix), which has been widely used in multiple published studies (see, for example [58][59][60] ). The LLOD and LLOQ are reported as 0.019 and 0.061 pg/mL, respectively. Serially diluted recombinant full-length human tau (0N4R splicing isoform) protein samples were used to generate the calibration curve. The quantification of soluble T-tau protein levels of all the studied individuals was carried out within the linear range of the calibration curve.
Experimenters performing ELISA were blind to the demographic and clinical features of participants. The experiments of quantitative determination of the levels of both Δtau314 and T-tau proteins were performed in duplicate. The mean protein levels were used for statistical analysis.
Statistical analysis. Statistical analyses were performed using GraphPad Prism Version 7.04 (GraphPad Software, La Jolla, CA) and R Version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria). The demographic and neuropathological characteristics were compared among the AD dementia, MCI, and CN individuals using Kruskal-Wallis tests for continuous variables and chi-square tests for binary variables. When overall differences were detected between groups, post hoc analyses comparing each pair of groups were performed. The associations between demographic and neuropathological characteristics and levels of Δtau314 proteins (normalized to levels of T-tau proteins) were estimated using Spearman's rank-order correlations for continuous variables and Mann-Whitney tests for binary variables. Finally, levels of Casp2, Δtau314, T-tau, βIII-tubulin, and total proteins were compared using Mann-Whitney tests between the CN and the CI individuals and between the AD dementia and the MCI, respectively. In addition, multiple linear regressions of the log-transformed protein levels were fit with adjustment for sex, age at death, and PMI of brain tissue harvest. The unadjusted version of this analysis (i.e., (2020) 10:3869 | https://doi.org/10.1038/s41598-020-60777-x www.nature.com/scientificreports www.nature.com/scientificreports/ a two-tailed, unpaired t-test of the log-transformed protein levels) was also shown. P < 0.05 was considered statistically significant. No formal corrections were made for the testing of multiple outcomes. The area under a curve (AUC) values were compared using DeLong's tests for correlated receiver operating characteristic (ROC) curves.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information Files.