We performed both proteomic and phosphoproteomic analyses using the label-free quantification of MaxQuant23–25 to analyze the mouse brain hippocampal samples extracted from 4 experimental groups that had been administered AAV-DUSP4 or AAV-GFP into dHc: 5xFAD-DUSP4 (n = 7 females, n = 4 males), 5xFAD-GFP (n = 7 females, n = 5 males), WT-GFP (n = 5 females, n = 7 males), and WT-DUSP4 (n = 5 females, n = 6 males) (Fig. 1A, see Methods and Supplementary Data 1). As a quality control (QC), we verified the genotypes of mice by the western blot analysis using the antibody to transgenic human APP (6E10) and by microscopic observation of the GFP protein activity/fluorescence (see Methods). Based on this analysis, a male mouse originally identified as 5xFAD-GFP was re-classified as WT-GFP, and the downstream analysis was corrected. In addition, we conducted QC on the proteomic and phosphoproteomic data for further downstream processing (see Discussion).
In the present study, we focused our analysis on the two most critical comparisons, i.e., 5xFAD-GFP vs WT-GFP to identify proteins and phosphoproteins that are regulated in the 5xFAD mouse model compared to WT, and in comparisons of 5xFAD-DUSP4 to 5xFAD-GFP, to investigate the impact of DUSP4 overexpression on the proteome/phosphoproteome in 5xFAD. To simplify the presentation, we termed the comparison 5xFAD-GFP vs WT-GFP as 5xFADvsWT, and 5xFAD-DUSP4 vs 5xFAD-GFP as 5xFAD-DUSP4vs5xFAD.
Furthermore, we used the nominal p < 0.05 as a cut-off to include the proteins/phosphoproteins that are regulated by DUSP4 overexpression. Our experimental validation of selected proteins and integration with human proteomics showed that this cut-off is an effective criterion to determine the proteomic/phosphoproteomic signatures that are regulated by DUSP4 (see Discussion). Figure 1B highlights the bioinformatics workflow for data analysis and integration.
Substantial numbers of differentially expressed proteins (DEPs) were regulated in 5xFAD and by DUSP4 overexpression
Together, we quantified 4,459 distinct proteins over the 46 samples. After QC (see Methods), we obtained 3578 unique proteins. We performed DEP analysis to reveal the mouse proteome which is impacted by the 5xFAD transgene and by DUSP4 overexpression. We identified 685 and 564 DEPs comparing 5xFADvsWT for female and male mice, respectively (Fig. 2A; Supplementary Fig. 1A; Supplementary Data 2A, B). We detected more DEPs that were down-regulated than up-regulated in 5xFADvsWT for mice of both sexes (Supplementary Fig. 2). As expected, amyloid precursor protein (APP) expression was substantially elevated in 5xFAD mice of each sex (Fig. 2A; Supplementary Fig. 1A; Supplementary Data 2A, B). In the comparison of 5xFAD-DUSP4vs5xFAD, we found 295 and 335 DEPs for female and male mice, respectively (Fig. 2B; Supplementary Fig. 1B; Supplementary Data 2C, D). In contrast to the comparison of 5xFADvsWT, we detected more up-regulated DEPs than down-regulated ones in 5xFAD-DUSP4vs5xFAD in each sex (Supplementary Fig. 2). As anticipated, DUSP4 protein levels were markedly increased in 5xFAD-DUSP4vs5xFAD for both female (fold-change (FC) = 6.7, p = 0.05) and male (FC = 22.8, p = 4.7E-6) mice, respectively (Supplementary Data 2C-D). Note that the APP protein expression was not altered in 5xFAD-DUSP4vs5xFAD.
We compared the DEP signatures across different comparisons for each sex. We separated up-regulated proteins from down-regulated ones to examine consistency in the directionality of protein expression changes. For each comparison, we observed significant overlap between the male and female DEP signatures in the direction of protein expression changes, and insignificant overlap in the opposite directions (Fig. 2C; Supplementary Fig. 3A). For example, the up-regulated signatures of males and females in 5xFADvsWT significantly overlap (fold enrichment (FE) = 4.2, FDR = 1.1E-68; Fig. 2C) and the down-regulated signatures of males and females in 5xFAD-DUSP4vs5xFAD also significantly overlapped (FE = 1.8, FDR = 0.02; Supplementary Fig. 3A). In contrast, in male mice, the up-regulated signature in 5xFAD-DUSP4vs5xFAD significantly overlap the down-regulated signature in 5xFADvsWT (FE = 7.8, FDR = 1.7E-44; Fig. 2D). Similar results were observed in female mice (Supplementary Fig. 3B and Supplementary Fig. 2). These results show that DUSP4 overexpression reverse the abnormal proteomic changes in the 5xFAD mice in comparison with the wild type mice.
We further looked into the DEPs for cell-type specificity. We observed that the down-regulated signatures were enriched for the markers of neurons, whereas the up-regulated signatures were most enriched for the markers of microglia and astrocytes in 5xFADvsWT in both sex groups (Fig. 2E), consistent with some previous finding of up-regulated immune response and neuronal damage, and down-regulated synaptic transmission 26. However, in 5xFAD-DUSP4vs5xFAD, we found that the down-regulated signatures were enriched for the markers of microglia and astrocytes in both sex groups, whereas the up-regulated signature in only males was enriched for the neuronal markers (Fig. 2E). Thus, overexpression of DUSP4 affected all the major brain cell types, albeit with difference in enrichment significance across sex groups (Fig. 2E).
We also examined biological pathways and functional processes in which these DEPs participated. In male 5xFAD mice, immune and defense response was activated while neuronal and synaptic functions were suppressed (Fig. 2F). Similar results were observed for female 5xFAD mice (Supplementary Fig. 4A). We then examined the effect of DUSP4 overexpression in 5xFAD mice. In male mice, DUSP4 overexpression activated pathways like intracellular signal transduction while suppressed immune and defense responses which were activated in 5xFAD mice (Fig. 2G). However, in females DUSP4 overexpression affected a different set of pathways (Supplementary Fig. 4B). Note that many pathways suppressed by DUSP4 overexpression in female mice (e.g., apoptotic process) are detrimental to cell functions (Supplementary Fig. 4B). These results revealed sex-specific functions of DUSP4.
DUSP4 overexpression caused significant changes in differentially expressed posttranslational modification (DEPTM) sites
We preprocessed the mass spectrometry (MS)-based phosphoproteome profiling using the R package PhosPiR, which removed MaxQuant-marked reverse sequences and potential contaminants, and have summarized the intensities for each phosphosite entry, termed PTM site (see Methods). The expression level (intensity) at each PTM site was obtained from quantile normalization and low-rank approximation imputation27. We removed any PTM site that had no gene name or PTM position information. The expression was further log2-transformed for the downstream analysis.
We obtained 7,124 distinct PTMs across the 46 samples, which spanned 2,222 unique proteins, averaging about 3 PTM sites per protein. We performed differential expression analysis on all the PTMs. We identified 982 and 557 DEPTMs in 5xFADvsWT for female and male mice, respectively (Fig. 3A; Supplementary Fig. 5A; Supplementary Data 3A, B). We detected more DEPTMs that were up-regulated than down-regulated in 5xFADvsWT in both sex groups (Supplementary Fig. 6). In the comparison of 5xFAD-DUSP4vs5xFAD, we found 409 and 425 DEPTMs for female and male mice, respectively (Fig. 3B; Supplementary Fig. 5B; Supplementary data 3C, D). In contrast to the comparison of 5xFADvsWT, we detected more down-regulated than up-regulated DEPTMs in 5xFAD-DUSP4vs5xFAD for mice of either sex (Supplementary Fig. 6). We then compared the DEPTM signatures across different comparisons in each sex in the same way as we conducted the DEP analysis (see above). Overall, a similar trend was observed for the DEPTMs as for the DEPs (Fig. 3C, 3D; Supplementary Fig. 7). In each comparison (5xFADvsWT or 5xFAD-DUSP4vs5xFAD), female and male mice shared a significant portion of DEPTMs with the same directionality, whereas in each sex, 5xFADvsWT and 5xFAD-DUSP4vs5xFAD showed significant overlap between their DEPTMs but with opposite directionality (Fig. 3C, 3D; Supplementary Fig. 7). These results again suggested DUSP4 overexpression might reverse the effects of the 5xFAD transgene on mice at the phosphoproteome level.
We further explored the pathways in which the DEPTMs were involved. Since proteins may possess multiple sites of phosphorylation, we collapsed the DEPTM sites onto their respective protein levels. We define a differentially phosphorylated protein (DPP) as the one that contains at least one DEPTM. We obtained 665 and 418 DPPs in 5xFADvsWT for female and male mice, respectively, and 327 and 340 DPPs in 5xFAD-DUSP4vs5xFAD for female and male mice, respectively. As shown in Fig. 3E, 3F, the most affected pathways are involved in neuronal processes and synaptic function for the DPPs (DEPTMs) across the comparisons in each sex (Supplementary Fig. 8A, B), suggesting that both 5xFAD and DUSP4 might often influence the phosphorylation state of the proteins that are relevant to neuronal and synaptic function.
To more deeply delve into the signals represented in our phosphoproteome profiling, we applied the site-centric pathway analysis28 on our PTMs via the algorithm as described in the R package GSVA29 (see Methods). We examined how the PTMs are enriched for the PTM site-specific phosphorylation signatures28 (PTMsigDB). As shown in Fig. 3G, in female 5xFADvsWT, the PTMs were enriched over more than half of the PTM sets in the mouse PTMsigDB. The top-ranked kinase PTM sets are KINASE-PSP_CAMK2A/Camk2a, KINASE-PSP_ERK1/Mapk3, KINASE-PSP_JNK1/Mapk830 (Fig. 3G), which are critical in AD neuropathogenesis. Strikingly, the PTMs from 5xFAD-DUSP4vs5xFAD in female mice are enriched for the PTM sets in the mouse PTMsigDB yet with an opposite directionality in enrichment score (ES) (Fig. 3G), highlighting that DUSP4 overexpression altered the 5xFAD effects (activated or suppressed) on the PTM sets but in opposite directionality as 5xFAD in WT mice. In contrast, in male mice, the enrichment of PTMs in the mouse PTMsigDB was not very evident in spite of the enrichment in a few of PTM sets (Fig. 3G). These results further suggested that DUSP4 overexpression might counteract the effects of 5xFAD transgene in mice in PTM site-centric pathways.
DUSP4 overexpression resulted in reduction in STAT3 in 5xFAD mice
Hippocampal STAT3, human APP (hAPP), and DUSP4 protein levels were significantly altered in our proteomics data. To validate the changes of these proteins, western blotting was utilized to confirm protein levels. The results showed that hippocampal STAT3 protein levels were increased by about 110% in female 5xFAD mice overexpressing GFP (5xFAD-GFP), while male 5xFAD-GFP increased by about 65%, compared to age- and sex-matched wild type mice overexpressing GFP (WT-GFP) (Fig. 4). STAT3 protein levels were reduced by about 65% in both female and male 5xFAD overexpressing DUSP4 (5xFAD-DUSP4) compared to age- and sex-matched 5xFAD-GFP (Fig. 4). Although STAT3 protein levels were significantly reduced in female 5xFAD-DUSP4 compared to female 5xFAD-GFP, levels were significantly higher than female WT-GFP, while STAT3 protein levels in male 5xFAD-DUSP4 showed no significant differences compared to male WT-GFP (Fig. 4). Western blot analyses confirmed the DUSP4 overexpression in both female and male mice administered with AAV-DUSP4. In addition, western blot analyses detected hAPP protein only in 5xFAD transgenic mice, which confirmed 5xFAD genotypes. These results validate the proteomics analyses.
The DUSP4 DEP and DEPTM signatures are enriched in human AD protein networks
We first compared the mouse DEP signatures in the present study with the human DEPs in AD that were derived from the proteomics profiling in the parahippocampal gyrus (PHG) of the MSBB cohort31,32. We stratified the human subjects over sex and thus obtained the sex-specific DEPs in AD vs normal healthy individual (NL) (Supplementary Data 4, and Methods). The mouse DEPs in 5xFADvsWT significantly overlapped the human DEP signatures, with the same directionality in both sexes though the overlap of male signatures was much less significant (Fig. 5A, B). On the other hand, the DEP signatures in 5xFAD-DUSP4vs5xFAD in the male mice have marginally significant overlap with the human male DEP signatures with the opposite directions (Supplementary Fig. 9A) while the signatures from the female mice don’t significantly overlap the respective human signatures (Supplementary Fig. 9B). These results not only validated the mouse DEPs we identified but also suggested that our findings from the mouse proteomics might be relevant to human AD neuropathology.
We projected the mouse DEP signatures onto the MEGENA co-expression networks from the human proteomics31 to gain further understanding of their functional relevance to human AD. In the MSBB protein co-expression network, more than half (> 15) of the top 30 AD-associated modules were enriched for the mouse DEPs from 5xFADvsWT of both sexes (Fig. 5C). The up-regulated DEPs in both male and female mice are enriched in the astrocyte (M3) and microglia modules (M245) while the down-regulated DEPs overlap significantly the neuronal modules (M5) (Fig. 5C). We also observed the enrichment of the DEPs from 5xFAD-DUSP4vs5xFAD in the network, especially, the down-regulated DEPs in the male mice (Fig. 5C). Similar results were found in the ROSMAP MEGENA network (Fig. 5D). These results further validated the relevance of the mouse DEPs to human AD, and were consistent with the afore-described cell-type enrichment analysis (Fig. 2E).
Furthermore, the mouse DEPTMs are also enriched in a number of top-ranked AD modules in the MSBB (Fig. 5E) and ROSMAP (Fig. 5F) protein co-expression networks. Importantly, the most enriched modules are neuron specific (M5 and M2 in the MSBB cohort, Fig. 5E; M7 and M75 in the ROSMAP cohort, Fig. 5F). These results are consistent with the previous pathway enrichment analysis (Fig. 3E, 3F), and indicate that the DEPTMs are often involved in neuronal and synaptic functions.
DUSP4 protein-centered networks are sex-specific
To formally identify the genes that are co-regulated with DUSP4 in AD, we leveraged a number of human AD cohorts as previously described 33 by examining the genes with significant correlations with DUSP4. We intersected the mouse DEP signatures in 5xFAD-DUSP4vs5xFAD with the human DUSP4-associated genes, and further constructed DUSP4 protein-centric networks for each sex (Fig. 6A, 6B). There are more proteins positively correlated with DUSP4 protein/gene than those negatively correlated with DUSP4 protein/gene (Fig. 6A, 6B). Impressively, the majority of the DUSP4-associated proteins are specifically expressed in either females (Fig. 6A) or males (Fig. 6B). Based on the sex-specific DUSP4-centric networks, we constructed the sex-specific DUSP4 signal maps (Fig. 6C, 6D). As shown in Fig. 6C, in females, DUSP4 is often involved in protein and lipid metabolism, in contrast to its involvement in synapse and myelin functions in males (Fig. 6D). DUSP4 participates in endolysosomal pathways in both male and female mice, but in the opposite directions (Fig. 6C, 6D). In summary, the results demonstrate that DUSP4 plays important roles in AD pathogenesis by regulating biological processes and functions shared by two sexes or distinct in each sex.