Amyloid-β and Tau pathologies are significantly increased in DS

AD pathology was assessed using the Braak and Thal staging or equivalent ABC score, for all cases used for proteomics analysis (Table 1, detailed case history in Supp. Table 1). Age was significantly different (p < 0.0001) in the LOAD cohort in comparison to the other experimental groups. However, we included eight controls ≤ 65 years old and the remaining 12 cases ≥ 65 to compensate for the age gap between EOAD and LOAD (Supp. Table 1). In addition, multiple variable linear regression analysis showed that age (p = 0.97) and sex (p = 0.45) did not contribute significantly to the differences observed in the proteomics analysis (Supp. Table 2).

Assessment of the regional distribution of Aβ and Tau pathology (Supp. Figure 1a, b) in all cases showed that Aβ levels in hippocampal and temporal regions were similar in DS and EOAD. However, Aβ quantities in DS were significantly higher (p = 0.013) compared to LOAD (Supp. Figure 1c). PHF-1 immunoreactive Tau pathology was significantly higher in DS compared to EOAD and LOAD (p = 0.0002 and p < 0.0001, respectively) (Supp. Figure 1d). Aβ and Tau pathology were not significantly different between EOAD and LOAD (Supp. Figure 1c–d). These results suggest an exacerbated Aβ and Tau pathology in DS despite the advanced stage of AD for all the cases in the cohorts evaluated.

Protein abundance in amyloid plaques and non-plaque tissue varies across DS, EOAD, and LOADAβ plaque pairwise comparisons

Protein differential expression in Aβ plaques and adjacent AD non-plaque tissue was evaluated using LFQ-MS in the microdissected hippocampus and temporal cortex (Fig. 1). LFQ-MS identified 1995 proteins (Supp. Tables 3–4), detected in at least 50% of the cases in any of the groups. PCA showed minimal segregation by groups (DS, EOAD, LOAD, or control) or by sample type (plaques and non-plaque tissue).

We identified 132 differentially abundant proteins in DS Aβ plaques compared to DS non-plaque tissue (Fig. 2b, d), 192 proteins in EOAD plaques vs. EOAD non-plaques (Fig. 2b, e), and 128 proteins in LOAD plaques vs. LOAD non-plaque tissue (FDR ≤ 5%, FC ≥ 1.5) (Fig. 2b, f). From these sets of proteins, 43 were shared between the three cohorts. We found 45 proteins with differential enrichment in plaques in DS, 97 proteins in EOAD, and 51 proteins in LOAD (Fig. 2b), indicating that enrichment of some proteins in Aβ plaques is variable in each experimental group. We observed a consistent enrichment of AD associated proteins such as the Aβ specific peptide LVFFAEDVGSNK (sequence corresponds to amino acids 17–28 of APP, Fig. 2d–f, j). This peptide does not discriminate between cleaved or full-length sequences. However, previous findings have shown a strong correlation to Aβ pathology [31, 49, 106]. We also identified previously detected amyloid plaque proteins, such as HTRA1, GPC1, VIM, APOE, CLSTN1, and SYT11 within the top ten most significant proteins across groups (Table 2). As expected, APP was within the top ten significantly abundant proteins in DS amyloid plaques (Fig. 2d) and was also significantly enriched in amyloid plaques in EOAD and LOAD (Fig. 2k). The plaque-protein COL25A1 [collagen alpha-1(XXV) chain, also known as CLAC-P] was the most abundant protein in amyloid plaques in all experimental groups, showing more enrichment in plaques than the Aβ peptide (Fig. 2d–f, l). Interestingly, COL25A1 was below mass spectrometry detection threshold in all control tissues (Fig. 2l), suggesting that this protein is highly correlated to Aβ plaque pathology. COL25A1 was increased 129.5-fold in DS, 29.9-fold in EOAD and 71-fold in LOAD (Table 2). In addition, COL25A1 was within the top ten significant proteins only in DS (Table 2). Hyaluronan and proteoglycan link protein 2 (HAPLN2, also known as Bral1) was within the most significant proteins decreased in plaques in the three cohorts studied. In addition, we observed decreased plaque-protein levels of oligodendrocyte proteins. MOG was significantly decreased in all groups, and MAG and MBP were significantly decreased in EOAD and LOAD amyloid plaques, respectively (Supp. Table 3). MAG and MBP levels were also decreased in plaques in DS, although it did not meet our significance criteria. The glucose transport facilitator SLC2A3 (also known as GLUT3) was decreased in amyloid plaques in all groups, yet it was significant only in EOAD and LOAD (Table 2). Overall, we observed similar proteins altered in Aβ plaques in all groups evaluated. However, most of the proteins show different abundance levels in plaques of DS, EOAD, and LOAD, accounting for the differences observed among groups.

Fig. 2

Principal component analysis (PCA) and differential protein expression in Aβ plaques and non-plaque tissue. a PCA shows the distribution of the n = 20 cases per each experimental group, with minimal segregation. b Venn diagram of differentially abundant Aβ plaque proteins shows 43 common proteins for all the AD subtypes evaluated, 45 for DS, 97 for EOAD, and 51 for LOAD. c Venn diagram of differentially abundant non-plaque proteins depicts 138 proteins in DS, 76 proteins in EOAD, 148 proteins in LOAD, and 65 common proteins for all AD subtypes. d–f Volcano plots indicate differentially expressed proteins (enriched in red, decreased in blue) in Aβ plaques compared to non-plaque tissue in DS (132 proteins, d), EOAD (192 proteins, e) and LOAD (128 proteins, f). g–i Volcano plots depict differentially expressed proteins in DS non-plaque tissue compared to controls (263 proteins, g), EOAD non-plaques (269 proteins, h), and LOAD non-plaques (301 proteins, i). j–l Normalized protein expression obtained from the label-free quantitative mass spectrometry proteomics of Aβ peptide (j), APP protein (k), and COL25A1 (l). Significance was determined using a student’s two-tailed t test (FDR < 5%, fold-change > 1.5). P values are indicated based on the pairwise comparisons. *** p < 0.001, **** p < 0.0001. Error bars indicate standard error of the mean (SEM). Significant pairwise comparisons are indicated for those analyses that were performed, and controls are shown as reference

Table 2 Top 20 significant proteins in Down syndrome, and early-onset and late-onset AD for ‘plaque vs. non-plaque’ pairwise comparisonsAD non-plaque tissue pairwise comparisons

We identified 263 differentially expressed proteins in DS non-plaque tissue compared to control non-plaque tissue (Fig. 2c, g), 269 proteins in EOAD non-plaque tissue vs. control non-plaque tissue (Fig. 2c, h), and 301 significantly altered proteins in LOAD non-plaque tissue vs. control non-plaque tissue (Fig. 2c, i). We identified 65 altered non-plaque proteins compared to control tissue that were common between all cohorts evaluated (Fig. 2c). We also observed 138 proteins with differential enrichment levels in DS non-plaque tissue, 76 proteins in EOAD, and 148 proteins in LOAD (Fig. 2c). Notably, we identified among the top ten enriched proteins in DS non-plaque tissue CLU, VIM, HSPB6, and SYNM (Supp. Table 5), which we also found enriched in amyloid plaques in all disease groups. CLU was consistently enriched in non-plaque tissue in the three groups evaluated when compared to control tissue (Supp. Table 5). VIM and HSPB6 were also among the most enriched proteins in EOAD non-plaque tissue (Supp. Table 5). Conversely, we identified the actin-binding protein destrin (DSTN) as the only protein among the top ten significantly decreased proteins in non-plaque tissue from DS, EOAD, and LOAD cohorts compared to controls (Supp. Table 5). We also observed that parvalbumin (PVALB) was the most decreased protein in DS non-plaque tissue compared with controls (Fig. 2g), whereas the levels of PVALB in EOAD and LOAD were not significantly different from controls (Supp. Table 4). Our proteomics findings in non-plaque tissue showed that there were more differences in protein levels in non-plaque tissue between groups, in comparison to the more consistent protein levels in plaques, highlighting the largely similar plaque proteome between AD subtypes despite differences in baseline, non-plaque-protein expression.

Amyloid plaque proteomes of DS, EOAD, and LOAD are highly correlated

We performed correlation analyses to compare the proteomes of Aβ plaques and non-plaque tissues in DS, EOAD, and LOAD. Proteins included in the correlations were significant and FC > 1.5 at least in one of the groups evaluated. For amyloid plaques, there was a positive correlation between DS and EOAD (R2 = 0.77, p < 0.0001). We observed 65.5% (164/250) of the proteins changing in the same direction (i.e., fold-change for a protein is positive or negative in both groups), where 29.6% (74/250) of the proteins were significantly altered in DS and EOAD plaques (Fig. 3a). We only observed 4.8% (12/250) of the proteins changing in different directions (i.e., fold-change for a protein is positive in one group and negative in the other) (Fig. 3a). DS and LOAD plaque proteomes also correlated positively (R2 = 0.73, p < 0.0001), with 66.2% (135/204) of the proteins with same fold-change direction and 27.5% (56/204) of the proteins significantly altered in both groups (Fig. 3b). Similar to DS and EOAD, only 6.3% (13/204) of the proteins were changing in opposite direction (Fig. 3b). There was also a positive correlation between EOAD and LOAD differentially abundant plaque proteins (R2 = 0.67, p < 0.0001), similar to what we observed between DS vs. the AD subtypes evaluated. We identified 66.4% (234/256) of the proteins changing in the same direction, and 25% (64/256) of the proteins were significant in both groups (Fig. 3c). The proteins changing in opposite direction accounted for 8.6% (22/256) of the total (Fig. 3c). Our analysis shows high similarity among the proteins altered in Aβ plaques vs. non-plaques of DS, EOAD, and LOAD, with the majority of the proteins changing in the same direction.

Fig. 3

Correlation analyses of differentially abundant proteins in Aβ plaques and non-plaque tissue. a–c Correlation analyses for significant proteins in Aβ plaques vs non-plaque tissue and d–f DS, EOAD and LOAD non-plaque vs control non-plaque tissue. Yellow dots represent proteins changing in the same direction (highly abundant or less abundant proteins in both groups evaluated) and that are significant for both groups compared. Magenta dots represent proteins changing in the same direction, but are significant only in one of the groups evaluated. Green dots represent proteins changing in opposite direction (i.e., abundant in one group and less abundant in the other group evaluated). Numbers are colored to match the dots. Proteins were selected for the correlation analysis if they were significant at least in one of the groups compared and its fold-change > 1.5. We observed a positive correlation between DS vs. EOAD a (p < 0.0001, R2 = 0.77, b DS vs. LOAD (p < 0.0001, R2 = 0.73) and c EOAD vs. LOAD (p < 0.0001, R2 = 0.67). There is also a positive correlation when comparing non-plaque proteins in d DS vs. EOAD (p < 0.0001, R2 = 0.59) and e DS vs. LOAD (p < 0.0001, R2 = 0.33). h. Correlation between EOAD and LOAD non-plaque proteins (p < 0.0001, R2 = 0.79)

Correlation analyses of DS, EOAD, and LOAD non-plaque differentially abundant proteins showed positive correlations between DS and EOAD (R2 = 0.59, p < 0.0001) and a weaker correlation between DS and LOAD (R2 = 0.33, p < 0.0001) (Fig. 3d–e). We observed 65.9% (275/417) of the proteins changing in the same direction in DS and EOAD Aβ plaques, where 27.6% (115/417) of the proteins were significantly altered in both groups. We observed 6.5% (27/417) of proteins changing in the opposite direction (Fig. 3d). Similarly, 67.1% (328/489) of the proteins in DS and LOAD were changing in the same direction (Fig. 3e). We observed that 15.3% (75/489) of the proteins were significant in both groups, whereas 17.6% (86/489) of proteins had opposite fold changes (Fig. 3e). Moreover, we observed a higher positive correlation between EOAD vs. LOAD non-plaque proteomes (R2 = 0.79, p < 0.0001), with 63.9% (273/427) of the proteins were changing in the same direction, with 33.5% (143/427) being also significant in both groups (Fig. 3f). Only 2.6% (11/427) of the proteins were changing in opposite directions (Fig. 3f). Overall, we observed a similar ‘amyloid plaques protein signature’ across the experimental groups. Nonetheless, correlations of the non-plaque tissue proteomes suggest a higher similarity between EOAD and LOAD differentially enriched proteins in comparison to DS.

Protein-coding genes present in Hsa21 are not associated with protein enrichment in Aβ plaques

We performed chromosomal mapping of significantly altered proteins identified through proteomic analysis across all human chromosomes using the UCSC Human Genome Browser to evaluate the distribution of these proteins across DS, EOAD, and LOAD. Supplemental Figure 2 illustrates the percentage of significantly altered proteins for each group. The overall percentage of proteins from each chromosome was below 20%, and no single chromosome exhibited a markedly overrepresented protein expression pattern. This suggests that proteins from all chromosomes, not just Hsa21, contribute to the molecular differences observed in both DS and AD.

Of the 1995 proteins identified in this study, 22 were from Hsa21 (Fig. 4). We compared these proteins with those reported in a previous DS plaque proteomics study [31], identifying a total of 26 Hsa21 proteins between the two studies. A significant portion, 69.2% (18/26), of these proteins were shared between the current and previous studies (Fig. 4). Among the proteins identified, APP was significantly altered in Aβ plaques in all cohorts (Fig. 4). GART was significantly abundant in LOAD and DS non-plaque tissue (Fig. 4a, c), and PCP4 was differentially expressed in LOAD and EOAD non-plaque tissue (Fig. 4a, b). CXADR was differentially expressed in EOAD amyloid plaques (Fig. 4b). APP was also significantly enriched in DS non-plaque tissue (FDR < 0.05, Fig. 4a). NCAM2, CBR1, CBR3, PDXK, CSTB, and COL6A1 were significantly enriched in DS non-plaque tissue (Fig. 4a). Taken together, these results along with the chromosomal mapping of all significantly altered proteins suggest that Hsa21 triplication does not necessarily lead to the enrichment of those gene products in Aβ plaques or in the surrounding non-plaque tissue.

Fig. 4

Mapping protein-coding genes to chromosome 21 (Hsa21). a Dashed box contains Venn diagram of proteins from genes in Hsa21 identified in the current study vs. Drummond et al. 2022, [31]. a–c The figure depicts fold-change (Log2 FC) of the 22 Hsa21 genes whose corresponding protein products were found in Aβ plaques (circles) or neighboring non-plaque tissue (squares) in LOAD (a) EOAD (b) and DS (c). Paired two-tailed t tests (plaques vs. non-plaques) or unpaired two-tailed t tests (non-plaques vs. control) with permutation correction at a 5% FDR are indicated. Aβ peptide is shown as reference

Aβ plaque-protein signature is related to APP processing, immunity, and lysosomesAβ plaques functional analyses

We identified functional associations for the significantly abundant proteins in Aβ plaques and AD non-plaque tissue by performing ‘GO enrichment analysis’ (FDR < 0.05, Supp. Tables 6–13). Top enriched biological process (BP) GO terms in DS included lytic vacuole organization, lysosome organization, and lysosomal transport (for the three terms, p = 1.29 × 10−5, Fig. 5a, Supp. Table 6). We also identified terms cell activation (p = 0.00024), regulation of immune system process (p = 0.00027), and leukocyte activation (p = 0.00016), which were also observed in EOAD (Fig. 5a). For cellular component (CC), we identified as the top terms vacuole, lysosome, lytic vacuole (p = 9.56 × 10−14), and endosome (p = 9.71 × 10−14, Fig. 5a, Supp. Table 10), similarly as BP GO terms. In contrast, EOAD most enriched BP terms were regulation of immune system process, B-cell-mediated immunity, immunoglobulin-mediated immune response, and lymphocyte-mediated immunity (p = 4.33 × 10−5, Fig. 5a, Supp. Table 6). Top CC GO terms in EOAD were secretory granule (p = 1.13 × 10−6), vacuolar lumen, and collagen-containing extracellular matrix (both p = 8.75 × 10−7, Fig. 5a, Supp. Table 10). LOAD also showed BP GO terms related to lysosomes as observed in DS, yet with a lower significance. For instance, we identified lysosomal transport and organization and lytic vacuole organization (p = 0.0288 Fig. 5a, Supp. Table 6). CC GO terms included lysosome and lytic vacuoles (p = 2.47 × 10−7), collagen-containing extracellular matrix (p = 9.41 × 10−6), and endosome (p = 0.00063) (Fig. 5a, Supp. Table 10), highlighting functional similarities of plaque-associated proteins between DS and LOAD.

Fig. 5

Gene ontology annotation and protein–protein interaction networks of significantly abundant proteins in Aβ plaques. a GO terms heatmap depicts top ten enriched BP and CC GO terms for significantly abundant Aβ plaque proteins in DS, EOAD, and LOAD. Color indicates the adjusted p value < 0.05 (− Log10 [adj. p value]). b–d Protein networks (PPI Enrichment p = 1 × 10−16) show functional and physical amyloid plaques protein associations in DS (b), EOAD (c) and LOAD (d). Node color indicates fold-change (log2 [FC]) and node size depicts adjusted p value (-log10 [p value]) from the student’s two-tailed t test. Disconnected nodes are not shown in the network. Colored dotted lines highlight groups of proteins based on functions/pathways observed in the GO terms; blue: APP protein metabolic process, red: immune response and inflammation, green: lysosomal-related functions, and purple: intermediate filament proteins, glial cells. GO terms annotation was performed using R package clusterProfiler v 4.8.2. PPI networks were created in Cytoscape v 3.10.0 using STRING database v 11.5

We also evaluated the physical and functional protein interactions of significantly abundant proteins in Aβ plaques, using Cytoscape and the STRING database (Fig. 5b–d). The networks for amyloid plaque proteins for all the cohorts evaluated showed a significant degree of protein–protein interactions (PPI enrichment p = 1 × 10−16). We observed a consistent group of proteins in all forms of AD evaluated, which were grouped based on functional enrichment (Fig. 5b–d). For instance, we identified proteins related to APP and Aβ metabolism (APP, APOE, CLU, CLSTN1, NCSTN, APLP2, and SPON1), immune response and inflammation (HLA-DRB1, HLA-DRB5, C1QC, C4A, and C3 consistent in DS and EOAD; CD44, ICAM1, and MSN in EOAD and LOAD), and lysosomal-related functions (PPT1, TPP1, LAMP1, PSAP, and CTSD). APOE was highly abundant in Aβ plaques in DS and LOAD (Fig. 5b, d) compared to EOAD, being the most significant in DS (Fig. 5b) in comparison to EOAD and LOAD. We also identified a group of glial-related proteins in EOAD network, namely VIM, DES, and GFAP (Fig. 5c). Overall, our findings suggest a similar plaque-protein signature in the three groups, which were functionally associated mainly to APP and Aβ processing, immunity-related responses, and lysosomal functions.

In addition, an analysis of the ten most abundant proteins (ranked by FC) differentially enriched in Aβ plaques in DS, EOAD, or LOAD further showed the relationship of Aβ plaque-associated proteins with lysosomal and immune-related functions (Supp. Table 14). According to the GO annotation, we found that the significantly enriched amyloid plaque proteins in DS predominantly relate to endo/lysosomal functions, including CLCN6, ATG9A, and VAMP7 (Fig. 6, Supp. Table 14). Oligodendrocyte protein MOG was significantly decreased in plaques for all cohorts, but fold-change suggests an increased reduction in DS (Supp. Table 3, Fig. 6a) in comparison to the other groups. We identified protein ITM2C, which is involved in Aβ peptide production [29] (Fig. 6b). We also observed proteins with functions linked to presynaptic signaling and axon guidance, namely, RUNDC3A and NTN1 [60, 104] (Fig. 6). The calcium-binding protein and marker of inhibitory neurons PVALB was significantly enriched in DS plaques but was unaltered in EOAD and LOAD (Fig. 6f). In contrast, we observed that Aβ plaque proteins significantly abundant in EOAD are mostly related to immune response, immunoglobulin-mediated immune response (S100A7, HPX, and IL36G), as well as vacuole lumen and secretory vesicles related (GGH, TTR). The protein EPPK1 is linked to cytoskeletal organization functions such epithelial cell proliferation and intermediate filament organization (Supp. Table 14). In LOAD, we observed a series of proteins involved in bounding membrane of organelle, collagen-containing extracellular matrix, and vesicle membrane (CYB5B, VWF and PTPRN2). Although we did not observe particular association with GO terms, other amyloid plaque LOAD proteins, including TIMM8A, ACSS3, and SFXN5 (linked to mitochondrial functions) [89, 133, 138], THUMPD1 and RPS7 (related to RNA-binding activity and ribosomes) [14, 128] and NRXN2 (protein–protein interactions at the synapses) [76] were identified (Supp. Table 14). These observations support our findings in the GO functional enrichment and protein interaction networks, providing evidence that some of the most abundant proteins in DS plaques are primarily linked to lysosomal pathways.

Fig. 6

Enriched Aβ plaque proteins of interest in DS compared with EOAD and LOAD. (a–f) Normalized protein expression obtained from the label-free quantitative mass spectrometry proteomics of abundant Aβ plaque proteins of interest in DS. Proteins are shown by order of decreasing significance. Proteins of interest were defined as significant (FDR < 5%, fold-change > 1.5) only in DS and also have known or predicted roles in AD and DS. Pairwise comparisons p values are indicated. * p < 0.05, **** p < 0.0001. Error bars indicate standard error of the mean (SEM). Significant pairwise comparisons are indicated for those analyses that were performed, controls are shown as reference. Additional symbols on top of the control bar indicate that the given protein is not significantly abundant in non-plaque AD tissue compared to controls in # DS, † EOAD, and ‡ LOAD, respectively

Non-plaque tissue functional analyses

GO terms for abundant non-plaque proteins showed chromatin remodeling as the top BP term for all experimental groups (DS p = 0.00128, EOAD p = 5.79 × 10−9, LOAD p = 1.69 × 10−10, Supp. Figure 3a, Supp. Table 8). Importantly, top BP GO terms in DS were associated with integrin-mediated signaling, extracellular structure, and extracellular matrix organization (p = 0.00684, Supp. Figure 3a, Supp. Table 8). In contrast, EOAD and LOAD top BP GO terms included protein–DNA complex assembly (p = 4.74 × 10−6 and p = 1.14 × 10−8, respectively), regulation of gene expression (EOAD p = 5.08 × 10−5, LOAD p = 1.68 × 10−8), and nucleosome assembly (EOAD p = 4.74 × 10−6, LOAD p = 3.25 × 10−8) (Supp. Figure 3a, Supp. Table 8). Top CC GO terms for DS were collagen-containing extracellular matrix, which was also observed in EOAD and LOAD, external encapsulating structure, and extracellular matrix (p = 3.52 × 10−8, Supp. Figure 3a, Supp. Table 12). Top CC GO term for EOAD was nucleosome (p = 4.44 × 10−6), which was also identified in DS and LOAD. Other EOAD top CC GO terms were DNA packaging complex (p = 8.01 × 10−6) and protein–DNA complex (p = 2.23 × 10−5) (Supp. Figure 3a, Supp. Table 12). In a similar fashion, LOAD top CC GO terms were DNA packaging complex, protein–DNA complex (both p = 3.78 × 10−14), and nucleosome (p = 1.71 × 10−12) (Supp. Figure 3a, Supp. Table 12).

We also created protein interaction networks of non-plaque tissue DS, EOAD, and LOAD proteomes, which showed a highly significant degree of protein–protein interactions (PPI enrichment p = 1 × 10−16, Supp. Figure 3b–d). We observed groups of RNA-binding proteins, such as SRSF4, eukaryotic initiation factors (eIF4), and the heterogeneous nuclear ribonucleoproteins (hnRNP) protein family, primarily in EOAD and LOAD networks (Supp. Figure 3c, d). We also observed a set of intermediate filament and glial proteins, such as GFAP, AQP4, DES, VIM, ALDH1L1, and GART (Supp. Figure 3b–d). Additionally, there were groups of histone proteins related to the nucleosome, such as H2A, H2B, and H1 protein families (Supp. Figure 3b–d). Particularly, the DS protein interaction network exhibited a set of collagens, laminins, cell adhesion proteins, proteoglycans, and heparin sulfate proteins (Supp. Figure 3b) as well as proteasome and chaperone proteins also involved in regulation of gene expression, including SQSTM1, PSMB4, PSMD4, and HSPB6 (Supp. Figure 3b). Our findings highlight a pivotal role of extracellular matrix (ECM) and structural components in DS besides the proteins associated to Aβ plaque pathology.

Comparative analysis with previous human AD proteomics and identification of novel plaque proteins

We compared the differentially abundant proteins found in Aβ plaques and AD non-plaque tissue with previous human AD proteomics studies compiled in the NeuroPro database [4]. We observed that 77.7% of altered proteins identified in amyloid plaques in our study were also identified in previous AD plaque proteomics studies (Fig. 7a). From the 301 significantly altered plaque proteins that we identified in the present study, 13.6% have not been found in previous plaque proteomics studies, but only reported as significantly altered in bulk brain tissue proteomics studies (Fig. 7a). Similarly, 85.2% of the proteins we identified in the non-plaque tissue have been described in previous plaque and bulk tissue proteomics studies, whereas 10.9% have been identified in bulk human brain tissue but not in plaque proteomics studies (Fig. 7a). Interestingly, we identified in our study 34 proteins that have not been described previously in any human AD proteomics study, either in plaques or in bulk tissue (Fig. 7a, Supp. Table 15–16).

Fig. 7

Comparison of protein changes with previous advanced AD proteomics studies. a Altered proteins identified in the current study were compared with proteins found altered in previous AD proteomics compiled in NeuroPro [4] (v1.12; https://neuropro.biomedical.hosting/). Pie charts show that 77.7% (234/301) of altered plaque proteins in the present study have been identified in previous AD plaque proteomics studies (gray). 13.6% (41/301) of the proteins have been seen only in bulk tissue proteomics studies (white), and 8.6% (26/301) of the altered proteins observed in the current study have not been described in previous AD proteomics (purple). In a similar fashion, 85.2% (478/561) proteins altered in AD non-plaque tissue have been observed in AD plaque proteomics, 10.9% (61/561) only in bulk tissue proteomics, and 3.9% (22/561) have not been described in previous AD proteomics studies. b Venn diagrams illustrate the altered proteins identified in Aβ plaques and AD non-plaque tissue for each AD subtype evaluated, in comparison to the 5104 altered proteins in advanced AD registered in NeuroPro database. c Heatmaps depicting the fold-change (Log2 [FC]) of the plaque and AD non-plaque altered proteins identified in the present study that have not been described in previous AD proteomics. Numbers in the cells represent the significance (FDR < 0.05) values observed in the pairwise comparisons, n.s represent no significant differences regardless of the fold-change

In DS specifically, we identified seven amyloid plaque proteins and eight non-plaque tissue proteins significantly altered in our study, which have not been found in past AD brain tissue proteomics studies (Fig. 7b, Supp. Table 17). Similarly, we identified in EOAD 21 significantly altered proteins in plaque and eight in non-plaque tissue, which have not been described previously (Fig. 7b, Supp. Table 17). In the case of LOAD, we observed four significantly altered proteins in amyloid plaques and 15 in non-plaque tissue that have not been identified in previous AD plaques or bulk brain tissue proteomics studies (Fig. 7b, Supp. Table 17). From this group of proteins, LAMTOR4 (late endosomal/lysosomal adaptor and MAPK and MTOR activator 4) was significantly enriched in Aβ plaques in all the cohorts analyzed (Fig. 7c). The proteins HLA-DRB5, ALOX12B, and SERPINB4 were significantly enriched in DS and EOAD amyloid plaques (Fig. 7c). In contrast, LAMA2 was significantly decreased in DS and EOAD amyloid plaques (Fig. 7c). On the other hand, we observed the histone protein H2BC11, the basal cell adhesion protein BCAM, and the DNA-binding protein FUBP3 significantly enriched in non-plaque tissue in DS, EOAD, and LOAD (Fig. 7c). The protein centrosomal protein of 290 kDa (CEP290) showed a marked decrease in DS Aβ plaques compared to DS non-plaque tissue; however, it was detected in few cases of the 20 evaluated in that cohort (Supp. Table. 3), reason why it did not reach FDR < 0.05 (Fig. 7c). The protein FAM171A2 was significantly enriched only in EOAD and LOAD, contrary to the protein DCAKD that was significantly decreased in EOAD and LOAD non-plaque tissue (Fig. 7c). Overall, our proteomics findings are consistent with previous proteomics studies. Notably, our comparative analysis allowed us to identify novel proteins in AD human proteomics.

Validation of the Aβ plaques protein signature in DS and novel plaque proteins in human DS proteomics

The NeuroPro database is a powerful tool to investigate proteomic changes in AD human brains. However, by the time of writing this article, the database does not include DS proteomics data. Therefore, we compared our DS amyloid plaques proteomics findings with our previous study (Drummond et al., 2022 [31]) where unbiased localized proteomics was used to interrogate the DS amyloid plaques proteome. In the study led by Drummond and colleagues, any Aβ plaque detected by IHC was sampled regardless of plaque morphology. We observed 2522 proteins between both DS plaque proteomics datasets, comprised of 1981 proteins in the present study and 2258 proteins in our previous work (excluding isoforms). We observed 68.1% (1717/2522) of proteins overlapping between both studies, with a total of 228 significantly altered plaque proteins in either dataset. Among these, 21.9% (50/228) were common to both studies (Fig. 8a). Particularly, 36% (82/228) of the significantly altered proteins in the present study were not significant in Drummond et al., and conversely, 42.1% (96/228) of the proteins identified in the previous study were not detected in the current dataset (Fig. 8a, Supp. Table 18). This variance may reflect differences in statistical thresholds and increased sample size, providing higher power in this study to identify more plaque-enriched proteins in DS with greater confidence. For instance, 35 proteins that were significantly enriched proteins detected in the Drummond study but not significant in ours were nonetheless observed in our dataset, with many showing increased abundance trends that nearly reached significance. In addition, from the proteins that were different between both studies (Fig. 8a), only 12 had a different direction of change, suggesting that most of the differences observed between the datasets are due to the differential stringency applied and the number of samples. Despite these differences, we observed a significant positive correlation between the Aβ plaque proteomes of the DS cohorts (p < 0.0001, R2 = 0.60, Fig. 8b). In fact, the 50 common proteins between both studies were changing in the same direction (48 enriched and 2 decreased in plaques, Fig. 8b). Within these set of amyloid plaque proteins, we identified Aβ peptide, APP, COL25A1, and a set of previously described plaque proteins, such as APOE, SMOC1, CLU, C3, and CLCN6 among others (extended data in Supp. Table 18), thus validating a plaque-protein signature also observed in DS Aβ pathology. Interestingly, from the seven novel DS plaque proteins regarding the NeuroPro database (Supp. Table 17), only ACP2 was also observed in the previous DS plaque proteomics study (Supp. Table 18). Our study is consistent with previous similar proteomics studies on AD brains, and further expanded the proteins present at these pathological lesions.

Fig. 8