Coordinated regional association of cathepsins and dipeptidyl peptidases with N-truncated Abeta42, Abeta40, and tau in Alzheimer’s brain

Abstract

Regional distributions of amyloid-β (Aβ) and tau accumulation have provided important insights into the pathomechanisms of Alzheimer’s disease (AD). While such analyses have typically been conducted through histochemical or clinical imaging studies, we previously reported unique regional associations among Aβ species, tau and other proteins by biochemically analyzing multiple postmortem brain regions. Here, using a new cohort and novel ELISAs, we investigated the regional relationships of Aβ species, tau, neuroglial markers, cathepsins and dipeptidyl peptidases (DPP) across AD stages. Despite a relatively small sample size, this study replicated key prior findings, including a strong regional association between Aβ_1-42 and the postsynaptic maker PSD95 particularly in the early stage, distinct regional distributions of Aβ_1-42 and N-terminally truncated Aβ42 (Aβ_t-42), and a significant association between Aβ_t-42 and tau in AD. Moreover, this study observed that Aβ_1-42 was associated with other synaptic proteins, but not neurofilament proteins. Notably, several proteases, particularly cathepsin B, cathepsin D and DPPIV, exhibited strong regional correlations with total Aβ_x-42, Aβ_t-42, Aβ_x-40, and tau accumulations in AD, forming coordinated regional distribution patterns. Such strong regional associations with late-stage Aβ species and tau were not observed for neuroglial markers. At the microscopic level, these proteases displayed abnormal morphologies proximity to Aβ and tau pathologies. Their coordinated regional distributions with Aβ_t-42, Aβ_x-40, and tau may indicate that these proteases cooperatively promote neurodegenerative cascades in AD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-025-02146-1.

Introduction

Alzheimer’s disease (AD) is neuropathologically characterized by the accumulation of amyloid-β (Aβ) peptides and tau proteins, forming amyloid plaques and neurofibrillary tangles (NFTs), respectively. While the etiology of AD, particularly for sporadic cases, remains elusive, the region-specific distributions of Aβ and tau accumulation have provided important insights into the pathomechanisms and clinical symptoms. Traditionally, region-wide analyses were performed by histochemical studies, leading to the development of diagnosis criteria, such as Braak NFT stage and Thal amyloid phase [5, 50]. Moreover, recent advancements in PET imaging ligands enable the in vivo investigation of these regional pathologies [27, 28, 35, 55]. Notably, studies have observed an overlap between brain regions exhibiting high default neuronal activity or aerobic glycolysis and those vulnerable to amyloid-PET signals, suggesting a mechanism that could elucidate the region-specific Aβ accumulation [6, 54]. However, to our knowledge, no studies have yet successfully identified regional associations between tau-PET signals and specific molecules or biological pathways [53], potentially due to the inherent limitation of clinical imaging analyses in detecting a broad range of molecular changes.

In contrast, our biochemical quantification of Aβ, tau, and other molecules across multiple postmortem brain regions has allowed us to analyze their regional relationships throughout AD progression [40, 42, 43, 58]. A key advantage of this approach is its versatility in measuring diverse molecules in parallel. For instance, by quantifying individual Aβ species, we observed their distinct regional accumulation: N-terminally truncated Aβ42 (Aβ_t-42) preferentially accumulated in limbic areas, like the amygdala, while full-length Aβ_1-42 was more prevalent in neocortical regions and the striatum [42]. Notably, Aβ_1-42 is abundant early in the disease, whereas Aβ_t-42 accumulates gradually with disease progression and shows stronger correlations with tau and clinical symptoms [42]. Furthermore, we found a strong regional association between Aβ_1-42 and the postsynaptic marker PSD95, even in non-demented individuals, potentially indicating early involvement of neuronal processes in Aβ accumulation [40, 43]. Aβ_t-42 accumulation also showed significant, albeit modest, regional associations with BACE1 and neprilysin [42]. However, it remains unclear whether other synaptic or neuronal markers also strongly correlate with Aβ_1-42 and if other molecules exhibit stronger regional associations with Aβ_t-42 or tau. To better address the regional distribution of tau accumulation, we recently developed an ELISA that more accurately reflects neuropathological tau burden [41].

Using a new set of brain samples, the current study directly addresses these remaining questions from our prior regional association studies [40, 42, 43, 58]. To this end, we implemented novel ELISAs for neuroglial markers, cathepsins, and dipeptidyl peptidases to identify molecules that regionally correlate with individual Aβ species and pathological tau accumulation.

Materials and methods

Human brain tissues and ethical approvals

Postmortem brain tissues were obtained from the Brain Bank for Aging Research (BBAR) at the Tokyo Metropolitan Institute for Geriatrics and Gerontology under the approval of Institutional Review Board of the National Center for Geriatrics and Gerontology (approval number: 1587) and Institutional Review Board of Tokyo Metropolitan Institute for Geriatrics and Gerontology (approval number: R22-02), both of which were in accordance with the tenets of Declaration of Helsinki. Written informed consent was obtained from the patients’ families before the autopsy. The BBAR routinely performs the standardized neuropathological evaluations as previously described [23, 38]. We analyzed 12 brain regions (Table 1), including five neocortical areas (dorsolateral prefrontal (Brodmann area (BA) = 8/9), orbitofrontal (BA = 12), inferior temporal (BA = 20), inferior parietal (BA = 39/40), and primary visual (BA = 17)), four limbic areas (posterior cingulate (BA = 31), entorhinal (BA = 28), amygdala, and insular), and three subcortical areas (striatum (caudate), thalamus, and cerebellum) of 18 individuals. These individuals were the same as our previous study to develop tau ELISA [41]; 5 normal elderly controls, 5 AD cases, and 8 cases with intermediate pathologies (2 cases with plaque dominant senile change (PSC), 2 cases with AD-type change (ADC), and 4 cases with NFT-predominant change (NFTC)), whose diagnostic criteria were previously described [51]. In brief, the difference between PSC and ADC is the degree of NFT pathology (PSC: Braak NFT stage ≤ II, and ADC: Braak NFT stage ≥ III), while they both have significant senile plaques (Braak senile plaque stage ≥ B). For the current study analyzing macroscopically-dissected tissues, as there were no differences in tau spread pattern between these PSC and ADC subjects, which were only observed in the entorhinal cortex [41], we lumped these four subjects together and called the SP (senile plaque) group. The diagnostic criteria of NFTC are almost equivalent to those of primary age-related tauopathy (PART) [10], except for a clear distinction of argyrophilic grain disease (AGD). In the current study, we call these subjects PART, instead of NFTC, for the sake of simplicity. Regarding vascular pathology, one normal elderly control, one SP subject, and one PART subject had some infarcts in the contralateral hemisphere, although no such lesions were noted in the ipsilateral samples analyzed in this study. The demographics of these four groups are shown in Supplementary Table 1.

Table 1.

Brain regions used for the analysis

Neocortical areas	Limbic areas	Subcortical areas
Dorsolateral prefrontal (BA9, DF)	Posterior cingulate (BA31, PC)	Striatum (caudate, ST)
Orbitofrontal (BA12, OF)	Entorhinal (BA28, EC)	Thalamus (TL)
Inferior temporal (BA20, IT)	Amygdala (AM)	Cerebellum (CB)
Inferior parietal (BA39/40, IP)	Insular (IN)
Primary visual (BA17, VC)

BA = Brodmann area

Sample preparation

We extracted proteins according to a previously described method [41]. In brief, after the removal of meninges and grossly visible major blood vessels, brain specimens were pulverized in a prechilled (on dry ice) BioPulverizer (BioSpec, Bartlesville, OK, USA) and homogenized with a polytron homogenizer (Kinematica, Bohemia, NY, USA) at a ratio of 10 ml/g of wet-weight brain tissue in ice-cold RIPA lysis buffer (Merck Millipore, Billerica, MA, USA) containing 0.1% SDS, a complete protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA), and PhosSTOP phosphatase inhibitor cocktail (Sigma-Aldrich) on ice. After centrifugation at 100,000 g for 1 h at 4 °C, the supernatant was collected as RIPA-soluble fraction, aliquoted, and stored at -80 °C until measurements for neuro-glial markers, cathepsins, and dipeptidyl peptidases. The residual pellet was re-homogenized in TBS plus 5 M guanidine hydrochloride (GuHCl), pH 7.6, and incubated with mild agitation for 12–16 h at room temperature. After centrifugation at 15,000 g for 30 min, the resultant supernatant was collected as the GuHCl fraction, diluted with 9 volumes of TBS, aliquoted, and stored at -80 °C until measurements of Aβ species and tau.

ELISAs

The levels of Aβ_1-42, Aβ_1(2)-42, Aβ_x-42, pyroglutamylated Aβ_pE3-42, Aβ_pE11-42, Aβ_x-40, Tau, PSD95, synaptotagmin-1 (SYT1), gephyrin, neurofilament light chain (NFL), neurofilament heavy chain (NFH), neuron-specific enolase (NSE), glial fibrillary acidic protein (GFAP), CD11b, cathepsin B, cathepsin D, cathepsin L, cathepsin S, cathepsin V, dipeptidyl peptidase II (DPPII), and dipeptidyl peptidase IV (DPPIV) were determined by ELISAs. Some of these ELISAs were previously described [40–44, 58]. The information of coating antibody, detection antibody, detection reagent and standard of in-house ELISAs or commercial ELISAs is summarized in Supplementary Table 2 with detailed ELISA procedure. In a series of our studies, we noticed that anti-Aβ N-terminal Ab5 antibody can detect Aβ_2-x at similar efficiency to Aβ_1-x, we thus call ELISA that uses this antibody together with anti-Aβ_x-42 antibody as Aβ_1(2)-42 ELISA, which was indeed used in our previous studies [40, 42, 43]. In the current study, we thus analyze and show both Aβ_1-42 and Aβ_1(2)-42 levels, in which the former species was detected by ELISA that uses 82E1, Aβ N-terminal-end specific monoclonal antibody [22], together with anti-Aβ_x-42 antibody. On the other hand, we calculated levels of Aβ_t-42 by subtracting Aβ_1(2)-42 levels from total Aβ_x-42 levels to be consistent with our previous study [42]. The levels of the total proteins in each fraction were determined by a Protein Assay BCA kit according to the manufacturer’s instructions (#297-73101, Fujifilm, Osaka, Japan).

Immunohistochemical staining

Formalin-fixed brain tissues were embedded in paraffin wax and cut into sections with a thickness of 6 μm [51]. After deparaffinization, brain sections were submerged in EDTA buffer (1 mM, pH 8.0) and autoclaved for 15 min. Sections were then immersed in 99% formic acid for 30 min for further antigen retrieval. Endogenous peroxidase activity was quenched by incubating the sections with 3% hydrogen peroxide at room temperature. Next, sections were blocked with Protein Block plus Serum Free (#X0909, Agilent, Santa Clara, CA, USA) for 50 min and incubated with rabbit monoclonal anti-cathepsin B antibody (#D1C7Y, Cell Signaling Technology, Danvers, MA, USA), goat polyclonal anti-cathepsin D antibody (#AF1014, R&D systems, Minneapolis, MN, USA), Alexa Fluor 405-conjugated rat monoclonal anti-DPPIV antibody (#FAB1180V, R&D systems), biotin-conjugated mouse monoclonal anti-Aβ antibody (4G8, BioLegend, San Diego, CA, USA), and Alexa Fluor 647-conjugated mouse monoclonal anti-Tau antibody (GT-38, Abcam, Cambridge, CAM, UK), diluted by Antibody Diluent containing Background-Reducing Components (#S3022, Agilent) overnight at 4 °C. Sections were washed three times with PBS at room temperature and then incubated with Alexa Fluor 488-conjugated donkey anti-rabbit IgG H&L (#ab150073, Abcam), Alexa Fluor 546-conjugated streptavidin (#S11225, Thermo Fisher Scientific), and Alexa Fluor 750-conjugated donkey anti-goat IgG H&L (#ab175744, Abcam) diluted (1:250) by Antibody Diluent with Background-Reducing Component (Agilent) for 2 h at room temperature in a light protected chamber. Sections were again washed three times with PBS at room temperature, incubated with 0.3% Sudan Black for 5 min, washed with PBS, and mounted with Vibrance® antifade mounting medium (#H-1700, Vector Laboratories, Newark, CA, USA). Each fluorescent image was taken with appropriate filters under a fluorescent microscope (BZ-X810, Keyence, Osaka, Japan) and analyzed using ImageJ software [39].

Data analysis of neuroanatomical distribution of proteins

To study the regional distribution of each protein level while accounting for the potential changes in Alzheimer’s patients, we introduced a three-step normalization process. First, all values measured by ELISAs were normalized by total protein levels. Second, to assess region-specific differences, this raw protein level, so-called “D” value, was normalized by the average value within an individual to obtain “N” value. These two-step normalization processes are the same as our previous studies, which can evaluate regional differences after adjusting for the influence of differences between individuals [40, 42, 43]. In the current study, we further performed maximum mean-based normalization for data in each group. For each region of each group, the mean of D values and N values among individuals were calculated;

where g is each group, r is region, and i is individual. Then, maximum values across 12 regions were calculated;

These maximum values across 12 regions were then used to compute the normalization scaling factor based on the ratio (R) of maximum values between AD and other groups (g).

Then, we calculated the final normalized value (N’) that was used to plot the regional distribution of each protein;

This approach could estimate the differences in the amount of each protein in each group, compared to AD, keeping the quality of their regional distribution. Lastly, we calculated the mean values of this N’ individual data in each region in each group, which was applied to the Pearson correlation test to address the regional association between each protein across 12 brain regions in each group. As two subjects in the control group, and one subject in PART group lacked amygdala tissue, this region was treated as missing for these subjects. This did not affect analyses using the average values of remaining samples, nor analyses of SP and AD groups. All statistical analyses and data visualization were performed by R statistical software using ggplot2, dplyr, GGally, ggrepel, and tidyr packages (v4.2.2; R Core Team 2022). P-values of less than 0.05 were considered significant.

Results

Region-specific accumulation of individual Aβ species, and tau depending on disease stage

We first evaluated the region-specifically accumulated individual Aβ species and pathological tau in GuHCl fraction across disease stage groups. As expected, control and PART groups did not exhibit Aβ accumulation across all measured species, compared to AD. In contrast, SP group showed significant Aβ accumulation, especially for Aβ_1-42 or Aβ_1(2)-42, with levels approaching those in AD group, while varying by region. Also, regarding regional distribution, Aβ_1-42 and Aβ_1(2)-42 were most abundant in neocortical areas, and less prevalent in subcortical regions, with the notable exception of the striatum, which showed full-length Aβ42 accumulation comparable to those in neocortical areas (Fig. 1A, B). Conversely, total Aβ42 (Aβ_x-42) and N-terminally truncated Aβ_t-42 were generally lower in most areas in SP compared to AD, with Aβ_t-42 being particularly less abundant. In AD, Aβ_x-42 and Aβ_t-42, were relatively more abundant in limbic areas, particularly the amygdala, and less prevalent across all subcortical areas, including the striatum (Fig. 1C, D). The N-terminally truncated species Aβ_pE3-42 and Aβ_pE11-42 showed similar trends to Aβ_x-42 or Aβ_t-42. However, individual variations, especially for Aβ_pE11-42, complicate the assessment of their regional distribution (Fig. 1E, F). Aβ_x-40 exhibited a regional distribution pattern similar to Aβ_t-42, with lower amounts in subcortical areas in AD and no apparent accumulation in SP (Fig. 1G). Pathological tau, as measured by ELISA targeting the middle to C-terminal region [41], was abundant in the neocortical and limbic areas of AD. In contrast, PART and SP groups showed no significant tau accumulation in any region, except for the entorhinal cortex, consistent with the initial Braak stages of these subjects (Fig. 1H).

Fig. 1.

Regional distribution of individual Aβ species and tau and their regional associations. A–H Normalized values of Aβ_1-42 (A), Aβ_1(2)-42 (B), Aβ_x-42 (C), Aβ_t-42 (D), Aβ_pE3-42 (E), Aβ_pE11-42 (F), Aβ_x-40 (G), and tau (H) are shown as individual data points with box and whisker plots by different brain regions in each four groups (control, PART, SP, and AD). The data points of the same subject are connected by line. The blue notation indicates neocortical areas (DF, OF, IT, IP, and VC); the purple notation indicates limbic areas (PC, EC, AM, and IN); and the orange notation indicates subcortical areas (ST, TL, and CB). For normalization, see material methods. I Regional association among individual Aβ species and tau in SP or AD are shown as scatterplot matrix panels. Lower-left panels show the scatter plot of average values of each region with a linear regression line and 95% confidence ellipse, and upper-right panels show its correlation values, by the combination of corresponding diagonal panel variables. The cyan notation indicates SP group, and the red notation indicates AD group. *p < 0.05 and **p < 0.01. The correlation coefficient (r) and p-values were acquired by the Pearson correlation test

While there are several observations about the regional distribution of individual Aβ species and tau in the current study, most of these results are indeed consistent with our previous studies using samples of the Mayo Clinic brain bank [40, 42, 43]. To further confirm our prior observations, we next analyzed regional associations among these Aβ species and tau within SP or AD, as shown in Fig. 1I. The regional associations between Aβ_1-42 or Aβ_1(2)-42 and Aβ_x-42 or Aβ_t-42 were mostly weak and not significant, while the regional associations between Aβ_x-42 and Aβ_t-42 were strong (the 4 × 4 panels in the upper left portion of Fig. 1I). Also, the regional distributions of Aβ_pE11-42, Aβ_x-40 and tau were more associated with those of Aβ_x-42 or Aβ_t-42, than those of Aβ_1-42 or Aβ_1(2)-42 in AD (other remaining panels of Fig. 1I). These results are largely consistent with our previous studies [40, 42, 43], thereby supporting the validity of the current methodology for investigating the regional distribution and associations of Aβ, tau, and potentially other related molecules.

Regional distribution of neuroglial markers and their associations with Aβ and tau

Thus, we next analyzed the regional distribution of synaptic, neuronal, and glial marker proteins in RIPA-soluble fraction to evaluate their association with Aβ species and tau accumulation. Synaptic marker proteins, PSD95, SYT1, and gephyrin, were generally abundant across neocortical and limbic areas, while the striatum also exhibited a comparable amount of these synaptic proteins to those in cortical areas. Also, SYT1 and gephyrin tended to exist more in non-AD groups, suggesting synaptic degeneration in AD (Fig. 2A–C). Regarding NFL and NFH, there exist different trends with these synaptic proteins; all subcortical areas had a comparable amount to neocortical areas, while the dorsolateral prefrontal cortex showed more abundance of these neurofilament proteins (Fig. 2D, E). A neuronal marker, NSE, showed a relatively similar trend, in which subcortical areas had comparable amounts to neocortical areas, and the dorsolateral prefrontal cortex tended to exhibit more abundance of this neuronal protein. Notably, NSE was much abundant in the control group, suggesting neurodegeneration in AD (Fig. 2F). An astrocytic marker, GFAP, showed very different trends with these synaptic or neuronal proteins. It was more abundant in the amygdala in AD group, and much less exists in non-AD groups (Fig. 2G), likely supporting that GFAP is a promising biomarker for AD [3, 36, 44]. In contrast, CD11b, a microglial marker, exhibited a comparable amount between non-AD and AD groups (Fig. 2H). To evaluate whether regional distributions of these proteins were similar or different across each disease stage group, we performed correlation analyses between each group, and observed good associations in each protein among non-AD groups, while their correlations with AD group were relatively weaker, suggesting that the regional distributions of these neuroglial markers change in AD (Supplementary Fig. 1I-P). We then analyzed whether the regional distribution of these neuroglial markers correlated with each other in each group, as shown in Fig. 2I. Synaptic markers well associated with each other especially in non-AD groups (the 3 × 3 panels in the upper left portion of Fig. 2I). Neurofilament proteins and NSE were generally associated with each other, while trends were somewhat different depending on each group (the 3 × 3 panels in the middle portion of Fig. 2I). GFAP and CD11b showed trends of inverse associations with these synaptic or neuronal proteins (the panels in the bottom 2 rows and rightmost 2 columns of Fig. 2I).

Fig. 2.

Regional distribution of neuro-glial markers and their regional associations. A–H Normalized values of PSD95 (A), SYT1 (B), gephyrin (C), NFL (D), NFH (E), NSE (F), GFAP (G), and CD11b (H) are shown as individual data points with box and whisker plots by different brain regions in each four groups (control, PART, SP, and AD). The data points of the same subject are connected by line. The blue notation indicates neocortical areas (DF, OF, IT, IP, and VC); the purple notation indicates limbic areas (PC, EC, AM, and IN); and the orange notation indicates subcortical areas (ST, TL, and CB). For normalization, see material methods. I Regional association among these neuro-glial markers in control, PART, SP, or AD are shown as scatterplot matrix panels. Lower-left panels show the scatter plot of average values of each region with a linear regression line and 95% confidence ellipse, and upper-right panels show its correlation values, by the combination of corresponding diagonal panel variables. The green notation indicates the control group, the yellow notation indicates PART group, the cyan notation indicates SP group, and the red notation indicates AD group. *p < 0.05 and **p < 0.01. The correlation coefficient (r) and p-values were acquired by the Pearson correlation test

We then analyzed the regional associations of these neuroglial markers with individual Aβ species and tau in SP and AD, as shown in Fig. 3. Consistent with our previous studies [40, 43], PSD95, a postsynaptic marker of excitatory neurons [9], showed robust associations (r > 0.80) with Aβ_1-42 and Aβ_1(2)-42 especially in SP (the first two panels of the first column of Fig. 3). SYT1, a presynaptic marker, gephyrin, a postsynaptic marker of inhibitory neurons [13], and NSE, a pan neuronal marker, also showed relatively good associations with Aβ_1-42 and Aβ_1(2)-42, while those of gephyrin and NSE were not so significant compared with PSD95 (the first two panels of the second, third and sixth columns of Fig. 3). Aβ_x-42, which includes Aβ_1-42, still exhibited some significant (but moderate (r < 0.80)) associations with these synaptic markers especially in SP (the third and row of Fig. 3). On the other hand, Aβ_t-42, which do not include Aβ_1-42, was not associated with these synaptic or neuronal markers (the fourth row of Fig. 3). Also, though Aβ_pE3-42 showed some significant (but moderate) associations with synaptic markers, Aβ_pE11-42 did not associate with them (the fifth and sixth rows of Fig. 3), which was consistent with our previous observation that Aβ_pE3-42 relatively show similar regional distribution with those of Aβ_1(2)-42, compared to Aβ_pE11-42 [42]. Also, Aβ_x-40 showed some significant (but moderate) association with SYT1 in AD, but not other synaptic or neuronal markers (the seventh row of Fig. 3). It is of note that tau moderately associated with GFAP (the last row of Fig. 3). Collectively, among neuroglial markers, PSD95 (and SYT1) showed strong regional associations with Aβ_1-42 or Aβ_1(2)-42, which was consistent with our previous studies [40, 43]. In contrast, although there were some significant associations between these neuroglial markers and Aβ_t-42 or tau, the degrees were modest at most.

Fig. 3.

Regional associations between neuro-glial markers and individual Aβ species or tau. Scatter plots of average values of each region with linear regression line and 95% confidence ellipse by the combination of corresponding top and right panel variables. Small rectangles in each panel show correlation values. The cyan notation indicates SP group, and the red notation indicates AD group. *; p < 0.05 and **; p < 0.01. The correlation coefficient (r) and p-values were acquired by the Pearson correlation test

Regional distribution of cathepsins and dipeptidyl peptidases and their associations with Aβ and tau

We then looked for other molecules showing better regional associations with Aβ_t-42 or tau. We thus focused on cathepsins, which have been reported to be closely related to Aβ and tau pathologies and have been extensively studied by many researchers [7, 8, 20, 48]. Using commercially available ELISAs, we measured levels of cathepsin B, cathepsin D, cathepsin L, cathepsin S, and cathepsin V, and determined their regional distributions in each group (Fig. 4A–E). While these cathepsin families have their own distributions, these distributions were different between non-AD groups and AD group. Indeed, when analyzing the correlation of each cathepsin family among these groups, non-AD groups generally associated well with each other, while they exhibited weaker correlations with those of AD group (Supplementary Fig. 1Q-U). In addition to these cathepsins, we addressed the regional distributions of DPPIV, which could participate in truncation of N-terminal Aβ [2, 52], together with another family protein, DPPII (Fig. 4F, G), which is also expressed in brain cells [18]. The regional distributions of these dipeptidyl peptidases were also relatively similar among non-AD groups, compared to those with AD group (Supplementary Fig. 1 V&W). On the other hand, when analyzed intermolecular relationships, the regional distribution of these cathepsins and dipeptidyl peptidases, especially cathepsin B, cathepsin D, cathepsin S, and DPPIV, tended to better correlate with each other within AD group, compared to non-AD groups (Fig. 4H). Indeed, intermolecular regional correlations among all these cathepsins and dipeptidyl peptidases were higher in AD than in other non-AD groups (Fig. 4I), suggesting a machinery leading to such coordinated regional association of these proteases in AD.

Fig. 4.

Regional distribution of cathepsins or dipeptidyl peptidases and their regional associations. A–G Normalized values of cathepsin B (CatB, A), cathepsin D (CatD, B), cathepsin L (CatL, C), cathepsin S (CatS, D), cathepsin V (CatV, E), DPPII (F), and DPPIV (G) are shown as individual data points with box and whisker plots by different brain regions in each four groups (control, PART, SP, and AD). The data points of the same subjects are connected by line. The blue notation indicates neocortical areas (DF, OF, IT, IP, and VC); the purple notation indicates limbic areas (PC, EC, AM, and IN); and the orange notation indicates subcortical areas (ST, TL, and CB). For normalization, see material methods. H Regional associations among these proteases in control, PART, SP, or AD are shown as scatterplot matrix panels. Lower-left panels show the scatter plot of average values of each region with a linear regression line and 95% confidence ellipse, and upper-right panels show its correlation values, by the combination of corresponding diagonal panel variables. The green notation indicates the control group, the yellow notation indicates PART group, the cyan notation indicates SP group, and the red notation indicates AD group. *p < 0.05 and **p < 0.01. The correlation coefficient (r) and p-values were acquired by the Pearson correlation test. I The correlation coefficients of these regional associations are shown as individual data points with box and whisker plots by each group. *p < 0.05 and **p < 0.01; by paired t-test with the Bonferroni correction

Thus, we then examined regional associations between these proteases and AD pathologies (Fig. 5). Indeed, several proteases, particularly cathepsin B, cathepsin D and DPPIV, showed strong regional associations (r > 0.80) with Aβ_x-42 or Aβ_t-42 in AD group (the third and fourth rows of Fig. 5). Also, these proteases, especially cathepsin B, strongly associated with Aβ_x-40 (the seventh row of Fig. 5). Regarding tau, cathepsin D showed strong regional association, and cathepsin B, cathepsin S, and DPPIV were moderately associated in AD. In contrast, these proteases generally did not well associate with neuroglial markers (Supplementary Fig. 2), suggesting that the expression of these proteases were not governed by single type of cells, while analyses of additional cellular makers, such as specific neuronal markers, or disease-associated microglia or astrocytes [11, 19] might be needed. Nonetheless, these results indicate that the coordinated regional distribution of cathepsins and dipeptidyl peptidases in AD reflects the pathological accumulation of Aβ_x-42, Aβ_t-42, Aβ_x-40, and tau.

Fig. 5.

Regional associations between cathepsins or dipeptidyl peptidases and individual Aβ species or tau. Scatter plots of average values of each region with linear regression line and 95% confidence ellipse by the combination of corresponding top and right panel variables. Small rectangles in each panel show correlation values. The cyan notation indicates SP group, and the red notation indicates AD group. *; p < 0.05 and **; p < 0.01. The correlation coefficient (r) and p-values were acquired by the Pearson correlation test

Lastly, we performed quintuple-fluorescent staining using antibodies against cathepsin B, cathepsin D, DPPIV, Aβ, and tau in the dorsolateral frontal cortex of an AD patient (Fig. 6 and Supplementary Fig. 3A). Cathepsin B and cathepsin D mostly exhibited lysosomal vesicle structures and often colocalized with each other (Supplementary Fig. 3B). In contrast, DPPIV, which generally did not colocalize with these cathepsins (Supplementary Fig. 3C&D), displayed entire cellular structures, including cytoplasm, plasma membrane and nucleus, consistent with previous studies [14, 57]. Interestingly, cathepsin B and cathepsin D sometimes showed entire cellular structures without apparent lysosomal vesicles, suggesting the release of these cathepsins into the cytoplasm, due to lysosomal membrane permeabilization [56]. Such abnormal morphologies were often observed around the core senile plaques, which are abundant with N-terminally truncated Aβ as well as Aβ40 [4, 24, 30, 42], or around tau pathologies, sometimes together with strong DPPIV signals (magnified panels of Fig. 6). These proximal and abnormal localization with senile plaques and tau pathologies may underlie their coordinated regional associations observed by the current biochemical analyses in AD.

Fig. 6.

Fluorescent staining of cathepsins, a dipeptidyl peptidase, and AD pathologies. A section of the dorsolateral frontal cortex of an AD case were quintuple-stained using an anti-cathepsin B antibody (green), anti-cathepsin D antibody (magenta), anti-DPPIV antibody (cyan), anti-Aβ antibody (4G8, red), and anti-tau antibody (GT-38, yellow) (upper-left), and its magnifications of a senile plaque with a core (upper-right), a tau-bearing cell (lower-left), and tau surrounded by a senile plaque (lower-right), with individual green (cathepsin B), magenta (cathepsin D), and cyan (DPPIV) channel images. The scale bar represents 50 µm

Discussion

In the current study, we investigated the regional relationships among individual Aβ species, tau, neuroglial markers, cathepsins, and dipeptidyl peptidases across AD stages. The study design and methods were largely similar to our prior studies [40, 42, 43], although the current study analyzed a new set of brain samples from a different brain bank (Mayo Brain Bank in the prior studies vs. BBAR brain bank in the current study), and also introduced novel ELISAs for neuroglial markers, cathepsins and dipeptidyl peptidases. Despite these differences, the current study replicated several key findings of our prior studies [40, 42, 43]. First, a strong regional association between Aβ_1-42 and PSD95, particularly in the early stage of AD [40, 43]. Second, distinct regional distributions of Aβ_1-42 and Aβ_t-42. Aβ_1-42 were most prevalent in neocortical areas and striatum, while Aβ_t-42 were relatively abundant in limbic areas, including the amygdala, but were less abundant in the striatum. Third, a relatively strong association of Aβ_t-42 with tau, suggesting that Aβ_t-42 accumulation is an important clue to understanding the disease progression [42]. Notably, the current study included four to five individuals per group, whereas our previous studies analyzed more than 15 individuals per group [40, 42, 43], indicating that this small sample size may be sufficient to address regional relationships among AD-related proteins and replicate key findings of prior studies. Samples of different brain areas derived from the same individual share identical postmortem interval, freezing durations, and other potential factors affecting protein degradation, which facilitate regional association studies using postmortem brain tissue.

In addition to PSD95, the current study also analyzed other synaptic or neuronal marker proteins, including SYT1, gephyrin, NFL, NFH, and NSE. The strong regional association of PSD95 or SYT1 with full-length Aβ42 in both SP and AD indicates that synapses play important roles in Aβ accumulation from the initial stage as we previously discussed [40, 43], while its exact mechanism, namely their contribution to Aβ production or aggregation, remains unclear from the current results. Of note, these associations were more prominent for PSD95, an excitatory postsynaptic marker [9], but not Gephyrin, an inhibitory postsynaptic marker [13], likely suggesting that excitatory neurons, but not inhibitory neurons, are more involved in Aβ accumulation. Conversely, NFL and NFH were not associated with these Aβ species, while they were associated with each other. Nonetheless, the current results may provide additional clues to understand why default mode neural activity or aerobic glycolysis are regionally associated with Aβ accumulation [6, 54].

This study found that regional distributions of several cathepsins and a dipeptidyl peptidase were strongly associated with those of Aβ_x-42, Aβ_t-42, Aβ_x-40, and tau, particularly in AD. On the other hand, synaptic or neuronal proteins and glial markers did not show such prominent associations. Although many pathological and experimental studies have investigated cathepsins, especially cathepsin B and cathepsin D, in AD pathogenesis [7, 8, 16, 20, 48, 49], their regional distributions have rarely been examined. To our knowledge, this is the first study to characterize their regional distribution and its associations with late-stage amyloid (namely Aβ_t-42 and Aβ_x-40) and tau accumulations. Notably, in vitro studies have shown that cathepsin B can truncate the C-terminal portion of Aβ42, generating Aβ40/38/32 [33]. The strong regional association between Aβ_x-40 and cathepsin B may support the notion that cathepsin B contributes to Aβ40 accumulation by truncating C-terminal Aβ42 in the brains of AD patients. In addition, such significant regional associations apply to DPPIV, a dipeptidyl peptidase, which could potentially be involved in the truncation of the N-terminal Aβ [2, 52], but its localization has rarely been reported in human brains. Thus, in addition to characterizing the regional distributions of these proteases, the current study revealed that their regional distributions correlate with each other in AD brains, forming the coordinated regional distribution patterns. Qualitative histochemical studies or technically-challenging PET imaging studies could not easily address such a point, highlighting the significance of our unique studies biochemically analyzing postmortem brain tissues.

In accordance with their good regional associations with Aβ_t-42, Aβ_x-40 and tau, these proteases existed proximity to amyloid and tau pathologies at the microscopic level (Fig. 6). Interestingly, around these AD pathologies, cathepsin B and cathepsin D often exhibited entire cellular morphologies rather than lysosomal vesicle structures, and DPPIV also showed strong signals, both of which would be indicative of somewhat cellular stress or death [12, 15, 37, 56]. Though their exact roles cannot be determined by the current observational study, these proteases might degrade AD pathologies, including both the N-terminal and C-terminal portion of Aβ [2, 33, 34, 52], while simultaneously enhancing neurodegenerative processes [20, 45, 52]. Thus, their coordinated regional distribution with Aβ_t-42, Aβ_x-40, and tau may indicate that these proteases cooperatively promote neurodegenerative cascades in AD. In this regard, these proteases might be attractive targets to prevent AD progression. Therapeutic potentials of cathepsin B inhibitors have been explored in several experimental models [1, 21, 45]. In addition, recent animal and clinical studies suggest that DPPIV inhibitors, one of the widely used diabetic drugs, may reduce Aβ and tau accumulation, and provide cognitive benefits [25, 46, 52]. However, as it remains to be clarified whether its potential therapeutic effects are mediated through the direct inhibition of DPPIV in the brain, further studies to address the role of DPPIV in AD would be warranted. On the other hand, information obtained from the current study would also be useful in developing PET ligands targeting these proteases [29, 47].

There are several limitations in the current study. First, we used ELISAs to evaluate levels of cathepsins and dipeptidyl peptidases, which may not accurately reflect their proteolytic activities. Measurements of enzymatic activity would be necessary to further confirm their coordinated regional distribution in AD. Second, due to limited availability of ELISA, we did not fully characterize individual N-truncated Aβ species, including Aβ_4-42, which could exert more pathogenic effects [17, 31, 59]. Further, we did not examine buffer- or detergent-soluble fractions, in which Aβ monomers or oligomers may exist, nor did we use formic acid extraction, which could more accurately assess accumulated Aβ (i.e., fibrillar Aβ) than the current guanidine extraction though controversies exist [26, 32]. Also, compared with ELISA, our immunohistochemical staining was restricted to a single dorsolateral frontal cortex section from one AD patient. Further characterization of individual Aβ species across different extraction fractions, together with more extensive immunohistochemical analyses, would help to support and extend our conclusion. Lastly, due to the observational nature of this study, causality between regionally-associated molecules could not be established. Similarly, it remains unclear whether SP or PART represent early stages of AD or constitute distinct entities. Nonetheless, their regional associations provide important implications for understanding the pathogenesis of AD, although it remains to be further validated, ideally through experimental approaches, whether the region- and species-specific patterns of Aβ42 accumulation indeed represent a consecutive cascade. We visually summarize our observations and hypotheses (Fig. 7).

Fig. 7.

Summary of the current observations with a hypothetical model of the progression of AD pathogenesis. Full-length Aβ_1-42 accumulates from the early stage (i.e., SP group), and is most abundant in neocortical areas (NCs) and the striatum (ST), but less prevalent in other subcortical areas (SCs). Synaptic markers, particularly PSD95, strongly associate with this regional distribution, supporting that synapses play important roles in Aβ accumulation. On the other hand, total Aβ_x-42 and N-terminally truncated Aβ_t-42, accumulate substantially in the late stage (i.e., AD group). Regionally, they are more abundant in limbic areas, particularly the amygdala (AM), and less prevalent across all subcortical areas, including the striatum. Tau and Aβ_x-40 mostly show a similar regional pattern (though it is not clear whether Aβ_x-40 represent plaques or cerebral amyloid angiopathy). Several proteases, particularly cathepsin B, cathepsin D and DPPIV, also show a similar regional pattern, forming coordinated regional distributions. These proteases might degrade AD pathologies, including both the N-terminal and C-terminal portions of Aβ42, while simultaneously enhancing neurodegenerative processes (see Discussion). Accordingly, their coordinated regional distribution may indicate that these proteases cooperatively promote neurodegenerative cascades in AD (as indicated by the question mark in the right portion of the illustration). Also, while we hypothesize that region- and species-specific patterns of Aβ42 accumulation represent a consecutive cascade, it requires further validation, ideally through experimental approaches (as indicated by the question mark in the left portion of the illustration)

In summary, we observed that cathepsins and dipeptidyl peptidases exhibit a coordinated regional distribution in AD, which is associated with pathological Aβ and tau accumulation. Because the regional distributions of AD pathologies are neuroanatomically atypical yet consistently observed in AD, the current observations provide clues for understanding how AD pathologies are formed together with these proteases, which may cooperatively promote neurodegenerative cascades.

Abbreviations

Aβ

Amyloid-β

AD

Alzheimer’s disease

BA

Brodmann area

BBAR

Brain Bank for Aging Research

DPP

Dipeptidyl peptidase

ELISA

Enzyme-linked immunosorbent assay

GFAP

Glial fibrillary acidic protein

NFH

Neurofilament heavy chain

NFL

Neurofilament light chain

NFT

Neurofibrillary tangles

NSE

Neuron-specific enolase

PART

Primary age-related tauopathy

PSD95

Postsynaptic density 95

PET

Positron emission tomography

SP

Senile plaque

SYT1

Synaptotagmin-1

Author contributions

M.S. contributed to the concept and study design. N.S. contributed to supervising the study. M.S., G.G., K.K., M.M., S.M., Y.S., G.B. and N.S. contributed to data acquisition and analysis. M.S., G.B., and N.S. contributed to drafting the manuscript and figures. All the authors edited and reviewed the final manuscript.

Funding

This work was supported in part by the Research Funding for Longevity Sciences from the National Center for Geriatrics and Gerontology (19-9, 19-3, 21-12, 24–16 to NS; 20-6 to MS); Grants-in-Aid from Japan Promotion of Science (21H03391, 24K22285 to MS; MEXT24K02361 to NS; JP 22H04923 (CoBiA) to YS & SM); AMED (JP21wm0425019 to YS & SM); Integrated Research Initiative for Living Well with Dementia (IRIDE) of the Tokyo Metropolitan Institute for Geriatrics and Gerontology (to YS); MHLW Research on rare and intractable diseases Program (JPMH23FC1008 to YS); a Takeda Science Foundation Research Encouragement Grant (to MS); a research grant from the Japan Foundation For Aging And Health (to MS); a research grant from the Uehara Memorial Foundation (to MS); a research grant from Japan Brain Foundation (to MS); a research grant from Cyukyo Longevity Medical and Promotion Foundation (to MS).

Data availability

All data generated or analyzed during this study are available from authors upon reasonable request.

Declarations

Ethics approval and consent to participate

Postmortem brain tissues were obtained from the Brain Bank for Aging Research (BBAR) at the Tokyo Metropolitan Institute for Geriatrics and Gerontology under the approval of Institutional Review Board of the National Center for Geriatrics and Gerontology (approval number: 1587) and Institutional Review Board of Tokyo Metropolitan Institute for Geriatrics and Gerontology (approval number: R22-02), both of which were in accordance with the tenets of Declaration of Helsinki. Written informed consent was obtained from the patients’ families before the autopsy.

Consent for publication

Not applicable.

Competing interests

G.B. consults for SciNeuro Pharmaceuticals. All other authors declare that they have no competing interests.