Olfactory Bulb Proteomics Reveals Widespread Proteostatic Disturbances in Mixed Dementia and Guides for Potential Serum Biomarkers to Discriminate Alzheimer Disease and Mixed Dementia Phenotypes

The most common form of mixed dementia (MixD) is constituted by abnormal protein deposits associated with Alzheimer’s disease (AD) that coexist with vascular disease. Although olfactory dysfunction is considered a clinical sign of AD-related dementias, little is known about the impact of this sensorial impairment in MixD at the molecular level. To address this gap in knowledge, we assessed olfactory bulb (OB) proteome-wide expression in MixD subjects (n = 6) respect to neurologically intact controls (n = 7). Around 9% of the quantified proteins were differentially expressed, pinpointing aberrant proteostasis involved in synaptic transmission, nucleoside monophosphate and carbohydrate metabolism, and neuron projection regeneration. In addition, network-driven proteomics revealed a modulation in cell-survival related pathways such as ERK, AKT, and the PDK1-PKC axis. Part of the differential OB protein set was not specific of MixD, also being deregulated across different tauopathies, synucleinopathies, and tardopathies. However, the comparative functional analysis of OB proteome data between MixD and pure AD pathologies deciphered commonalities and differences between both related phenotypes. Finally, olfactory proteomics allowed to propose serum Prolow-density lipoprotein receptor-related protein 1 (LRP1) as a candidate marker to differentiate AD from MixD phenotypes.


Introduction
Alzheimer's disease (AD) and vascular dementia (VaD) are the most common causes of dementia in the elderly [1]. In medical practice, the term mixed dementia (MixD) is mostly referred to cases where there are clinicopathological evidences of both AD and vascular disease [2]. Around 25% of demented patients have pure AD pathology, while more than 50% present different vascular lesions (such as micro/macroinfarcts, microhemorrhages, lacunar strokes, among others), either alone or associated with AD [3]. Furthermore, atherosclerosis is evidenced in cerebral arteries in AD patients [4]. Vascular risk factors (hypertension, obesity, and diabetes mellitus) are associated with an elevated dementia and amyloid overproduction risks [5,6]. A frequent comorbidity of cerebrovascular and AD pathologies is confirmed in aged subjects [7][8][9]. At the mechanistic level, a plethora of tissue and molecular events have been proposed to interplay between the neurodegenerative process and the cerebrovascular damage (blood-brain barrier leakage, inflammation, oxidative stress) [10,11], however, the complete knowledge of this potential cause-and-effect relation is still lacking [10].
Although AD can be frequently diagnosed with a considerable accuracy, the distinction between MixD, isolated AD and VaD remains controversial, being a difficult diagnostic challenge [11].
Particularly relevant to neurologists is the fact that olfactory dysfunction can be considered a clinical, or in some cases a preclinical, sign of different dementias like AD and VaD [12,13]. Although a score below normal performance in olfactory test has been observed in VaD patients [14], further sensorial studies with larger and longitudinal cohorts are necessary to evaluate potential differences in the olfactory performance between AD and VaD [14,15]. It is important to note that olfactory dysfunction has been associated with increased mortality from neurodegenerative and cardiovascular diseases [16]. Several studies point out that cardiovascular and cerebrovascular disease, subclinical atherosclerosis, stroke, and diabetes are considered predictors of accelerated odor identification decline [17][18][19]. Ischemic or structural damages in brain areas involved in olfaction have been proposed as potential drivers of this olfactory decline [20].
Given the global prevalence of MixD-associated cognitive impairment and the lack of therapeutic strategies, there is a clear unmet need for vascular therapies targeting mechanisms that precipitate the neurodegenerative process. It is well known that the molecular homeostasis of olfactory structures is deeply altered in the context of AD pathology [21][22][23][24]. However, the impact of MixD on olfactory areas remains to be clarified. In this study, we have applied an olfactory proteotyping strategy [25] to partially reveal the missing relationships in the pathobiochemical knowledge when AD and vascular damage coexist, deciphering common and differential olfactory protein mediators between pure AD and MixD. Moreover, we have used olfactory neuroproteomic data as a strategy to define potential fluid biomarkers for the diagnosis and discrimination of patients affected by AD and MixD.

Human Samples
According to the Spanish Law 14/2007 of Biomedical Research, inform written consent forms of the Brain Bank of IDIBAPS (Barcelona, Spain) were obtained for research purposes from relatives of patients included in this study. Post-mortem fresh-frozen olfactory bulbs of 6 Mixed dementia (MixD) patients, and 7 controls were obtained from the Brain Bank of IDIBAPS (Barcelona, Spain) following the guidelines of Spanish legislation. The control group was composed of elderly subjects with no histological findings of any neurological disease. The study was conducted in accordance with the Declaration of Helsinki and all assessments, post-mortem evaluations, and procedures were previously approved by the Local Clinical Ethics Committee (PI_2019/108). All human brains considered in the proteomics and follow-up phases had a post-mortem interval (PMI) lower than 19 h (Table  1). On the other hand, in order to check potential disease biomarkers, serum samples from MixD (n = 19) and AD (n = 31) patients together with serum samples from healthy subjects (n = 32) were collected (Supplementary Table S1). In all cases, neuropathological assessment was performed according to standardized neuropathological guidelines [26][27][28][29].

Olfactory Proteomics
Whole OB specimens (70-80 mg) derived from controls and MixD cases were homogenized in lysis buffer containing 7 M urea, 2 M thiourea, 50 mM DTT. After ultracentrifugation, protein extracts were precipitated, pellets were dissolved in 6 M Urea and Tris 100 mM pH 7.8 and Bradford assay kit (Bio-Rad) was used for protein quantitation. Whole proteomes were concentrated in the stacking/resolving SDS-PAGE gel interface. After staining, protein digestion (10 ug) was carried out with trypsin (Promega; 1:20, w/w) at 37 • C for 16 h as previously described [30]. Prior to LC-MS/MS, peptides were purified and concentrated using C18 Zip Tip Solid Phase Extraction (Millipore, Burlington, MA, USA). Label free LC-MS/MS analyses were performed on an EASY-nLC 1200 liquid chromatography system interfaced with a Q Exactive HF-X mass spectrometer (Thermo Scientific, Waltham, MA, USA). Chromatographic/elution conditions and mass-spectrometry parameters were as previously described [31]. Data were acquired using Xcalibur software (Thermo Scientific, Waltham, MA, USA).

Data Analysis
Mass spectrometry raw data were processed using the MaxQuant software (v.1.6.3.3) (Max Planck Institute, Munich, Germany) [32] following the next parameters: (1) Homo Sapiens UniProtKB database (February 2019) containing contaminants and the reversed version of all sequences, (2) main peptide search (4.5 ppm) and first search tolerance (20 ppm), (3) trypsin digestion with a maximum of two missed cleavages, (4) variable modifications (methionine oxidation and N-terminal acetylation), (5) fixed modification (carbamidomethylation), (6) peptide length (7 amino acids), (7) fragment mass deviation (40 ppm) and (8) false discovery rate (FDR) for peptide spectrum match (PSM), peptide and protein identification (1%). The analysis of the Maxquant output file and subsequent visualization was done by Perseus software [33]. Potential contaminants and proteins identified as reverse were removed. The data were transformed into log2 values and normalization was performed using a width adjustment strategy. Protein identification and quantitation criteria was performed as previously described [31]. The protein identification was considered valid with at least two unique or razor peptides whereas protein quantification was calculated using at least two unique peptides. For differential analysis, a 1.3-fold change cut-off was used (two-way Student T-test; p < 0.05). Hence, proteins with ratios below the low range of 0.77 were considered downregulated whereas those with higher range than 1.33 were considered up-regulated. MS data and search results files were deposited in the Proteome Xchange Consortium via the JPOST partner repository (https://repository.jpostdb.org, accessed on 13 April 2021) [34] with the identifier PXD025368 for ProteomeXchange and JPST001128 for jPOST. Interactome and pathway analysis were performed using BioGrid [35], Ingenuity (Qiagen), or Metascape [36] tools.

Results and Discussion
Unbiased omics approaches have been proposed as essential tools to increase our understanding of the AD pathogenesis subtype variety as well as the common presence of vascular effects present in mixed pathologies [37]. Specifically, proteomics has already been aimed to provide more insights into VaD at cortical level [38,39]. Although it is widely known that patients suffering from AD and VaD experience olfactory dysfunction [12,13], no studies have examined the impact of this sensorial impairment at molecular level.

Olfactory Bulb Proteome-Wide Characterization in Human MixD
Since olfactory system is considered a potential gateway for the access of environmental insults and a prion-like propagation site in different forms of dementia [40,41], we have used OB label-free quantitative proteomics to deeply characterize the olfactory proteostatic imbalance in MixD (Table 1).

MixD Induces Olfactory Disruption in Functional Tau/APP Interactomes and Specific Survival Pathways
Bearing in mind that the characterization of unexpected connections between seemingly unrelated proteins and neuropathological substrates is a straightforward approach for the identification of novel MixD related-targets, we explored whether Tau (MAPT) and APP were functionally interconnected with DEPs in MixD OBs. Proteome-scale interactome maps merging the OB DEPs were performed using the IPA software ( Figure 2 and Supplementary File S1). Interestingly, 20 differential functional interactors for Tau were identified, suggesting the involvement in related biological functions. Specifically, olfactory Tau is central to an interconnected molecular network between plasma membrane (CLTA, SNAP25, DNAJC5, STX1B, AGRN, CLTB) and nucleocytoplasmic region (PRKCG, PPIA, NDUFS4, TUBB6, BASP1, ENO1, ATP5F1D, GSTP1, S100B, TOP2B, VSNL1, GFAP, PRDX6, SOD1). However, the deregulated olfactory APP interactome impacts across extracellular space (VWF, NES, OGN, NPTX1), plasma membrane (THY1, RAC1), and cytoplasm (SERPINB6, DYNC1H, RTN3, AK1, CRYL1, PFN2) ( Figure 2).  In a wider scale, a whole proteome comparison revealed that signaling mediators like the nuclear factor kappa B (NFκB) and PI3K complexes and cell-stress related such as PKA appeared as principal hubs in functional interactome maps (Figure 3, Supplementary File S1). As shown in Figure 3A, the deregulation of several mitochondrial-related proteins (ATP5ME, ATP5F1D, ATP5MG, NDUFS4, COX5B, Cytochrome c oxidase, CYB5A, SOD1, GSTP1) suggested an impairment in mitochondrial function in the OB of MixD subjects. In accordance, mitochondrial dysfunction constitutes an early and well-known feature of neurodegenerative processes [46] and our group has previously described alterations in the mitochondrial sensor PHB complex across several-related neurological disorders, including MixD [23]. On the other hand, although the NFκB constitutes a master regulator of many essential signaling cascades when activated travelling from the cytoplasm to the nucleus [47], the recently described presence of mitochondrial NFκB suggest its influence on important mitochondrial processes [48]. Therefore, subsequent experiments were performed in order to study the activated status on NFκB. As shown in Figure 4A,D, although no significant changes were found when analyzing all the study samples at the same time, a significant increase in the activated levels of NFκB was observed in subjects diagnosed with the highest Braak stages (Braak VI). Of note, NFκB role on cell survival can be either neuroprotective or induce neurotoxicity by proinflammatory mechanisms. Depending on the pathological state, its overexpression can result in damage to the vessel walls and impaired vascular cell function [49]. On the other hand, in order to enhance the analytical outcome of our proteomic experiment, the activated status of the PI3K complex and the cAMP-dependent protein kinase A (PKA) was also monitored ( Figure 3B,C, respectively). As shown in Figure 4B,D, although a slight upregulation of p110a (PI3K catalytic subunit) protein levels was evidenced, more predominantly in Braak V stages, an AKT inactivation was observed, suggesting a potential role of phosphatases such as PP2A, PTEN, or others in this context [50]. Likewise, the PI3K/Akt signaling pathway mediates cell survival and differentiation, and participates in learning and memory processes [51]. Interestingly, significant changes in AKT levels were not detected in the OB of AD subjects [52], suggesting that the vascular damage may be responsible for this deregulation at the level of the OB. On the contrary, significant increased levels of the catalytical subunit of PKA (PKAc) were found ( Figure 4C), where the tendency was mainly observed in the Braak III stages ( Figure 4D), thus, indicating alterations in cAMP levels. In this sense, similar alterations in AD subjects have been observed, suggesting common OB molecular mechanisms between these two pathologies.

Comparison of OB Deregulated Proteome in Pure AD and Mixed Dementia: An Specificity Analysis
To further study in detail the OB metabolic modulation in MixD, the differential proteome map was functionally characterized. As shown in Figure 1F, synaptic transmission, nucleoside monophosphate metabolism, carbohydrate metabolism, response to metal ion, aromatic catabolism, intracellular transport, neuron projection regeneration and VEGF signaling were part of the most significantly overrepresented bioprocesses in MixD subjects (Supplementary Table S2 and Supplementary Figure S2). Bioinformatic analysis also revealed that a subset of OB proteins was linked to AD and/or vascular processes such as blood vessel development, atherosclerosis, and coagulation ( Table 2).

Comparison of OB Deregulated Proteome in Pure AD and Mixed Dementia: An Specificity Analysis
To further study in detail the OB metabolic modulation in MixD, the differential proteome map was functionally characterized. As shown in Figure 1F, synaptic transmission, nucleoside monophosphate metabolism, carbohydrate metabolism, response to metal ion, aromatic catabolism, intracellular transport, neuron projection regeneration and VEGF signaling were part of the most significantly overrepresented bioprocesses in MixD subjects (Supplementary Table S2 and Supplementary Figure S2). Bioinformatic analysis also revealed that a subset of OB proteins was linked to AD and/or vascular processes such as blood vessel development, atherosclerosis, and coagulation (Table 2). In order to partially decipher unspecific and specific proteostatic alterations due to the presence of concomitant AD, we interlocked our MixD differential dataset with OB differential proteome data previously obtained from pure AD cases [23]. Due to the MixD samples used in our proteomics workflow derived from subjects with concomitant AD (Braak stages III-VI), only differential proteins detected across these Braak stages in our previous work were considered. As shown in Figure 5A, 32 protein mediators were deregulated not only in MixD but also in pure AD. However, only the protein expression profile corresponding to 6 proteins (BASP1, CALM, SOD1, ERP44, TPM4, ALAD) was similar across AD and MixD OBs ( Figure 5B). Interestingly, functional clustering unveiled common and distinct imbalanced biological processes between MixD and AD ( Figure 5C). Specifically, DEPs that map in functional categories such as integrin-mediated signaling pathway, CDC5L complex, regulation of cell projection organization, leukocyte migration, synaptic protein-protein interactions and neuron projection regeneration were exclusively and significantly enriched in MixD ( Figure 5 and Supplementary Tables S3  and S4). These data indicate that the comparative analysis between omics outputs may be considered a useful tool to potentially distinguish pure AD and MixD pathologies through Specifically, DEPs that map in functional categories such as integrin-mediated signaling pathway, CDC5L complex, regulation of cell projection organization, leukocyte migra-tion, synaptic protein-protein interactions and neuron projection regeneration were exclusively and significantly enriched in MixD ( Figure 5 and Supplementary Tables S3 and S4). These data indicate that the comparative analysis between omics outputs may be considered a useful tool to potentially distinguish pure AD and MixD pathologies through the elucidation of specific biological process as well as the identification of potential discriminatory biomarkers. On the other hand, other biological pathways were commonly deregulated between the two correlated pathologies ( Figure 5C). Among them, cell-stress related pathways were potentially deregulated according to the differential proteomic signature in both AD and MixD. Therefore, we decided to monitor a kinase panel gathering essential biological pathways due to (i) previous findings showing alterations in cell-survival and stress related pathways across different neurodegenerative disorders at the level of the OB [24,44,45], and (ii) the absence of similar reports focused on MixD contexts. First, regarding the MAPK pathway, a significant decrease in the activated levels of ERK 1 2 was observed in the MixD cases diagnosed with Braak III staging ( Figure 6A and Supplementary Figure S3), interestingly opposite to the hyperactivation previously observed in the OB of AD subjects [23]. Concerning the PDK1-PKC axis, a significant increase in the activated levels of PDK1 was observed in the Braak V stages, just as an increment in the total levels of the PKC family in the Braak III and VI staging ( Figure 6B and Supplementary Figure S3). Deregulations in the PKC signaling cascades are known to be early features in the brain of patients with AD [53] and previous reports in olfactory AD samples have also reported stage-specific deregulations in this axis [23]. However, to our knowledge, this is the first report linking alterations in this route in MixD backgrounds. On the other hand, since the p38 MAPK signaling has been extensively linked to neurodegeneration and inflammatory processes [54,55], we further evaluated the activated status of p38 MAPK in MixD OBs. As shown in Figure 6C and Supplementary Figure  S3, both activated and total levels of P38 were upregulated but only in Braak VI stages, suggesting a neuroinflammatory environment at the level of the OB. In line with this findings, recent studies have shown that the activation of this pathway may lead neuronal apoptosis and functional deficits in vascular dementia [56]. In accordance, stage-specific alterations were also found for both activated and total levels of SEK1, again demonstrating altered cell-stress responses at olfactory level during MixD disorders.

Protein Serum Profile Across AD and MixD: A Pilot Study Targeted to the Analysis of Neurexin-3 (NRXN3), Tenascin-R (TNR) and Prolow-Density Lipoprotein Receptor-Related Protein 1 (LRP1)
More than half of the patients meeting clinicopathological AD diagnostic criteria also have vascular lesions [3]. Based on the neuropathological co-existence and the controversy regarding the differential diagnosis between AD and MixD, there is an urgent need for a better clinical differentiation of these pathologies. Bearing in mind that fluid proteomics is considered a valuable molecular repository for diagnosing/targeting the neurodegenerative process and olfactory neurodegeneration is among the earliest features, the application of olfactory proteomics is an ideal bridge to detect olfactory proteins that might be tested in fluids as potential biomarkers [57]. Aiming to discover potential biomarkers to differentiate neurological syndromes, we have focused our attention on three proteins (NRXN3, TNR, LRP1) because they are tentatively present in biofluids and are involved in exclusive altered biofunctions enriched in MixD (Supplementary Table S3). Specifically, NRXN3 is related to cognition, learning/memory ( Table 2) and protein-protein interactions at synapses (R-HSA-6794362; Figure 5C); TNR is also related to cognition, learning/memory ( Table 2) and regulation of cell projection organization (GO: 0031345; Figure 5C); and LRP1 is related to AD, blood vessel development (Table 2) and neuron projection regeneration (GO:00311102; Figure 5C). For that, serum samples belonging to AD and MixD phenotypes together with non-neurological controls were included in the study (n = 32/control; n = 31/AD; n = 19/MixD;~50/50 female/male) (Supplementary  Table S1). To our knowledge, no experimental evidence have linked NRXN3, TNR, and LRP1 with MixD. Alterations in presynaptic adhesion NRXN3 protein levels have been linked with a major AD risk [58]. Interestingly, a significant decrease in NRXN3 serum levels was observed between both AD and MixD samples ( Figure 7A) and neurological intact controls. However, although our data suggest a more prominent decrease in NXRN3 serum levels in AD, no significant changes were observed between both pathologies. On the other hand, while serum TNR protein levels were unaltered between neurological contexts and healthy controls ( Figure 7B), a significant increase in LRP1 serum protein levels was observed in MixD subjects maintaining normal levels in AD ( Figure 7C). In this sense, although an effort to find specific sex differences was performed analyzing our data (Supplementary Figure S4), no significant changes were observed for any of the biomarkers. In particular, LRP1 is an ApoE receptor that plays a role in clearance of Abeta and regulates glucose uptake and insulin signaling in the brain, being a key regulator of Tau uptake and spread [59][60][61]. In this sense, being aware of the limited cohort analyzed, our data indicate that LRP1 may be a potential biomarker able to distinguish between both syndromes. Raw quantifications are shown in Supplementary Table S5. These results should be further evaluated in larger cohorts and in combination with other biochemical markers in order to improve the current diagnostic assays.
vival and stress related pathways across different neurodegenerative disorders at the level of the OB [24,44,45], and (ii) the absence of similar reports focused on MixD contexts. First, regarding the MAPK pathway, a significant decrease in the activated levels of ERK ½ was observed in the MixD cases diagnosed with Braak III staging ( Figure 6A and Supplementary Figure S3), interestingly opposite to the hyperactivation previously observed in the OB of AD subjects [23]. Concerning the PDK1-PKC axis, a significant increase in the activated levels of PDK1 was observed in the Braak V stages, just as an increment in the total levels of the PKC family in the Braak III and VI staging ( Figure 6B and Supplementary Figure S3). Deregulations in the PKC signaling cascades are known to be early features in the brain of patients with AD [53] and previous reports in olfactory AD samples have also reported stage-specific deregulations in this axis [23]. However, to our knowledge, this is the first report linking alterations in this route in MixD backgrounds. On the other hand, since the p38 MAPK signaling has been extensively linked to neurodegeneration and inflammatory processes [54,55], we further evaluated the activated status of p38 MAPK in MixD OBs. As shown in Figure 6C and Supplementary Figure S3, both activated and total levels of P38 were upregulated but only in Braak VI stages, suggesting a neuroinflammatory environment at the level of the OB. In line with this findings, recent studies have shown that the activation of this pathway may lead neuronal apoptosis and functional deficits in vascular dementia [56]. In accordance, stage-specific alterations were also found for both activated and total levels of SEK1, again demonstrating altered cell-stress responses at olfactory level during MixD disorders.  Although this study has uncovered many intricacies in the OB homeostasis in the context of MixD, there are potential limitations that warrant discussion. Due to the technological approach used, we failed to accurately monitor many protein species with low expression levels. Both AD pathology and cerebrovascular disease independently are strongly related with cognitive decline and/or dementia. Frequently, they appear together, showing an additive or even a synergistic effect and the weight of each component may be different among patients. In those cases, it remains challenging to distinguish between AD and MixD. This difficulty in diagnosis limits the number of MixD patients available in our cohort. Furthermore, even in absence of cognitive impairment, it is difficult to find pure age-matched controls without any sign of amyloid pathology or cerebrovascular disease given the high incidence of these lesions in the elderly and the fact that both findings increase with age. This factor is the reason why our control group has decreased age compared with MixD group. Therefore, we could not exclude the possibility that part of the differences found between both groups could be influenced by other age-related factors apart from the MixD ocurrence. Regarding the vascular component of MixD, different cerebrovascular lesions and locations induce different phenotypes of dementia. Given a big enough cohort of subjects, OB proteomics may prove to be useful to discriminate different types of dementia according to different cerebrovascular lesions. Although this study has uncovered many intricacies in the OB homeostasis in the context of MixD, there are potential limitations that warrant discussion. Due to the technological approach used, we failed to accurately monitor many protein species with low expression levels. Both AD pathology and cerebrovascular disease independently are strongly related with cognitive decline and/or dementia. Frequently, they appear together, showing an additive or even a synergistic effect and the weight of each component may be different among patients. In those cases, it remains challenging to distinguish between AD and MixD. This difficulty in diagnosis limits the number of MixD patients available in our cohort. Furthermore, even in absence of cognitive impairment, it is difficult to find pure age-matched controls without any sign of amyloid pathology or cerebrovascular disease given the high incidence of these lesions in the elderly and the fact that both findings increase with age. This factor is the reason why our control group has decreased age compared with MixD group. Therefore, we could not exclude the possibility that part of the differences found between both groups could be influenced by other age-related factors apart from the MixD ocurrence. Regarding the vascular component of MixD, different cerebrovascular lesions and locations induce different phenotypes of dementia. Given a big enough cohort of subjects, OB proteomics may prove to be useful to discriminate different types of dementia according to different cerebrovascular lesions.

Conclusions
Overall, the present study provides new clues regarding the molecular mechanisms concerning the olfactory dysfunction that occurs during MixD. Part of the differential OB protein set was not specific of MixD, being also deregulated across different tauopathies, synucleinopathies and tardopathies. However, functional analysis has unveiled OB commonalities and differences between pure AD and MixD. Based on olfactory proteomic data, LRP1 may be considered a potential serum biomarker to differentiate AD and MixD phenotypes.  with the mass spectrometric data deposit in ProteomeXChange/PRIDE. Mass spectrometry data acquisition was performed in the Proteomics Core Facility-SGIKER at the University of the Basque Country (UPV/EHU, ERDF, EU). The Proteomics Platforms of Navarrabiomed and University of the Basque Country are member of Proteored (PRB3-ISCIII) and are supported by grant PT17/0019/009, of the PE I+D+I 2013-2016 funded by ISCIII and FEDER. The Clinical Neuroproteomics Unit of Navarrabiomed is member of the Global Consortium for Chemosensory Research (GCCR) and the Spanish Olfactory Network (ROE) (supported by grant RED2018-102662-T funded by Spanish Ministry of Science and Innovation).

Conflicts of Interest:
The authors declare no conflict of interest.