Neurodegenerative disorders are characterized by protein aggregates, the importance of which in inducing neuronal toxicity has been hotly debated. Pathological forms of proteins also appear to spread through the brain in characteristic patterns in most neurodegenerative diseases, with particularly robust evidence for the spread of amyloid beta (Aβ), tau (τ), and α-synuclein (α-syn) [43, 73]. Although each neurodegenerative disorder exhibits accumulation of specific characteristic protein aggregates, there are also many cases that exhibit aggregation of multiple pathological proteins. Studies in model systems including transgenic animals and cell culture support several potential types of interactions between different pathological proteins involved in neurodegeneration including cross-seeding of aggregates and pathological changes in one protein initiating mislocalization and post-translational modifications of another. Here, we will examine the neuropathological co-occurrence of pathologies in human autopsy cases and discuss potential mechanistic links between these diverse proteins in disease pathogenesis. We will focus on the pathological proteins for which there is the clearest evidence of both co-existence in human cases and biological evidence for potential mechanistic interactions. Table 1 summarizes the neurodegenerative diseases we will cover and the protein aggregates most commonly observed in the brains of people with those diagnoses. The co-occurrence of cerebrovascular pathology and protein aggregates is discussed in the other article of this cluster [77]. Here, our aim is to provide an updated view about the co-occurrence of different protein aggregates in neurodegeneration and their possible mutual interactions which should further our understanding of the biology of neurodegenerative disorders.Table 1

Pathological aggregates in neurodegenerative diseases

AD is the most common dementing disorder seen in 50–60% of dementia cases. Neuropathologically, AD is characterized by dramatic atrophy of the gray matter and accumulation of amyloid plaques and neurofibrillary tangles (NFTs) [60]. Amyloid plaques consist of extracellular aggregates of Aβ [93], which is a cleavage product derived from the amyloid precursor protein (APP) by β- and γ-secretase cleavage [53]. The same Aβ peptide also occurs in AD-related vascular deposits known as cerebral amyloid angiopathy (CAA) [7]. Aβ-plaques as well as vascular Aβ deposits first occur in neocortical areas and then expand into further brain regions in a distinct hierarchical sequence [141, 143].

Synucleinopathies encompass Lewy body diseases (LBD) and multiple system atrophy (MSA). The latter are characterized by glial α-syn aggregations in the form of glial cytoplasmic inclusions (Papp-Lantos bodies) while the neuropathological hallmark lesions of LBD are α-syn aggregates in neuronal cell somata and neuronal cell processes termed Lewy bodies (LB) and Lewy neurites (LN), respectively. LBD encompass Parkinson’s disease (PD), Parkinson’s disease dementia (PDD), and dementia with LB (DLB). While PD is primarily a movement disorder clinically characterized by the presence of extrapyramidal symptoms (EPS), clinical core features of PDD and DLB include fluctuating cognition and visual hallucinations [96]. There are no neuropathological criteria that allow to distinguish between PDD and DLB as this distinction can only be made based on clinical grounds; in PDD, EPS precede the onset of dementia by at least 12 months, whereas in DLB dementia is concomitant with or precedes EPS [96].

There are several neuropathological staging systems for LBD, the most frequently used are the Braak LBD stages and the Newcastle/McKeith criteria for LBD. Briefly, Braak LBD stages [14] postulate that nuclei in the medulla oblongata become initially affected (stage 1) and pathology spreads gradually to pons (stage 2), midbrain (stage 3), entorhinal cortex, and hippocampus (stage 4) and finally reaches the neocortex (stages 5 and 6). This topographical progression is not proposed in the Newcastle/McKeith criteria [96], which distinguish between brainstem predominant (PD), limbic (transitional; DLB), and diffuse neocortical (DLB) types of LBD. α-Syn aggregates in the form of LB do not show a continuous increase with increasing age. Instead, the degree and prevalence of Lewy body pathology started in the 41- to 60-year-old age group and then increased up 61–80 years of age. In older individuals, the prevalence of the lesions at the different stages decreased (Fig. 1f). This decrease of Lewy body pathology is unexpected and could be explained by a reduced life expectancy of Lewy body pathology patients compared with that of AD patients. However, confirmation in independent samples and further, more detailed studies are required to clarify these findings.

While the neuropathological hallmark lesions of LBDs are LB/LN other pathologies are highly prevalent in LBDs; Aβ depositions are present in up to 85% of LBD cases with dementia [61, 66, 76] and it has been suggested that striatal Aβ pathology is less frequent in PDD as compared to DLB (e.g., 18 vs. 100% in [76]), while in PD, Aβ pathology is in general less frequent (e.g., 55% in [66]) than in LBD with dementia. Similarly, τ pathology is frequently present in LBD [26, 66, 76] and a recently published study on 213 LBD cases with dementia found that 30% showed high Braak NFT stages V/VI [61]. Such cases which show neocortical LBD (i.e., DLB) and full-blown AD (i.e., Braak NFT stages V/VI) can be neuropathologically diagnosed as having mixed dementia (mixed AD/DLB), since the severity of either AD or LB pathology alone would be sufficient to represent a robust neuropathological correlate for clinical dementia. This raises the question as to why those cases clinically either present as DLB or AD. Using quantitative assessment of Aβ, τ-protein, and α-syn in such mixed AD/DLB we found that those who clinically presented with AD had higher τ loads than those with clinical LBD [156]. Moreover, the hierarchical distribution of τ-protein loads in clinical AD cases was similar to neuropathologically pure AD cases, while clinical DLB cases showed comparatively less hippocampal and temporal τ loads. These findings suggest that in clinical AD cases neuropathologically diagnosed as mixed AD/DLB neurodegeneration may be primarily driven by AD while LB pathology occurs later in the disease, possibly aggravated or even triggered by AD pathology. Conversely, in clinically diagnosed DLB cases, which show mixed AD/DLB at post-mortem examination, classical LB pathology (i.e., α-syn) may be the primary driver of the neurodegenerative process. While these data somehow suggest that in mixed cases the quantitatively predominant pathology determines the characteristic clinical pictures, a large clinico-pathological correlative study using conventional semi-quantitative methodology found in a subset of clinically diagnosed DLB cases only severe neocortical τ pathology and a complete lack of LB pathology [105]. Recently, Irwin and colleagues demonstrated that τ pathology in LBD with dementia independently predicted a shorter time interval between motor symptom and dementia onset as well as shorter overall survival [61], clearly indicating that—if present—τ plays a central role in the disease process of LBD and that the co-occurrence of both AD and DLB pathology is likely to have an aggravating effect on disease progression. Respective data from neuropathological post-mortem studies are corroborated by in vivo imaging studies showing higher amyloid and τ levels in LBD compared to controls [46, 47]. While AD pathology is the most important co-morbidity in LBD cases, they may also show other pathologies such as cerebrovascular disease to varying degrees of severity in up to 75% [67] and TDP-43 pathology in over 30% of cases [95]. More quantitative data on the amount of pathological burden in large clinico-pathological cohorts are needed to clarify the relative influence of Aβ, τ, α-syn, and TDP-43 on the clinical picture. This would pave the way for novel and more subtle clinico-pathological phenotypes, which, together with biomarkers, would allow to diagnose patients based on the actual underlying pathology.

Genetic data provide a tenuous link between synucleinopathies and tauopathies. H1 τ haplotype is associated with increased risk of PD [155], and has recently been associated with increased α-syn aggregation in DLB in a small post-mortem study [25]. There is also cross-talk between α-syn and τ in terms of disease-associated mutations. Mutations in or replications of the SCNA gene encoding α-syn that cause familial PD can also cause cognitive symptoms including dementia [104, 113, 126]. A recent genetic study indicates that the p.A152T variant of MAPT is likely associated with a higher risk of DLB [84]. Another potential genetic link between τ and α-syn is LRRK2. LRRK2 mutations usually give rise to α-syn pathology [52], but sometimes, they also cause tangle pathology. τ and α-syn pathology can also occur hand in hand without amyloid plaque pathology in rare cases, e.g., with hereditary spastic paraplegia due to SPG 7 mutations [148].

In addition to genetic data implying causative links between tauopathies and synucleinopathies, there are also multiple indications that pathological forms of τ and α-syn physically co-aggregate within neuronal cytoplasm [104]. Using double immunofluorescence labeling and mass-spectrometry, τ and α-syn have been observed in several studies to aggregate together in the same tangles, LB, and Lewy Neurites in PD and DLB brain [5, 24, 62]. Importantly, one recent study using novel conformation-specific antibodies demonstrated that toxic oligomeric forms of τ and α-syn accumulate together in PD and DLB [122]. In vitro, α-syn binds directly to τ via the microtubule-binding region of the τ protein and the C-terminus of α-syn [68]. This interaction is affected in a complex and as yet not completely understood manner by both phosphorylation and disease-associated mutations in the proteins [36, 104]. It is possible that τ and α-syn through their binding can cross-seed aggregation. Amyloidogenic proteins share some common characteristics in terms of their aggregation with a typically slow initial aggregation that becomes much faster after the process begins due to the new aggregates templating or seeding further aggregation [81]. This seeding process is not only relevant to the formation of large fibrils but also to the formation of oligomeric forms of proteins, which are likely the toxic species in the brain [81]. The seeding process is not necessarily limited to seeds made from the aggregating protein. Indeed, cross-seeding, in which “seeds”—some type of small aggregate—from one protein can induce aggregation of another protein—has been demonstrated across many different proteins. There is a small amount of evidence suggesting that τ and Aβ may cross-seed the formation of toxic aggregates in vitro [49]. Much more evidence implicates cross-seeding between α-syn and τ. In vitro, co-incubation of α-syn and τ synergistically promotes aggregation of both proteins [41], and in cultured neurons expressing FTDP-17 mutant τ, synthetic aggregates of α-syn seeded τ [50]. However, the cross-seeding of these proteins is not always reproduced, for example, using a sensitive FRET-based detection method, τ and α-syn did not cross-seed [55]. These conflicting results may be due to the ability of specific “strains” or confirmations of aggregates of the proteins, which induce the seeding, which has been observed for τ [22, 37], α-syn [111], and cross-seeding between them [50]. Similarly, distinct strains of Aβ have been observed, which induce different types of aggregates in vivo [133, 159]. Cross-seeding between Aβ and α-syn has been observed in vitro [109], along with hybrid oligomer formation [149]. Further work is warranted on understanding the cross-seeding process, and particularly on whether this causes the formation of toxic oligomeric species of the proteins.

Transgenic mouse models of tauopathies and synucleinopathies also support the idea of synergistic interactions of τ and α-syn in exacerbating neurodegenerative phenotypes. Overexpressing α-syn in transgenic mice causes τ phosphorylation [51, 160]. Combining pathologies in mice overexpressing fAD mutations in APP and PSEN1, FTDP-17 mutant MAPT, and PD associated SCNA, there is a synergistic effect of the pathological proteins resulting in exacerbated pathological and behavioral phenotypes [23]. Knocking out τ, however, does not prevent α-syn-related motor deficits in two models of PD, indicating that these proteins in addition to having synergistic effects can act independently to confer toxicity [103]. In wild-type mice injected with α-syn fibrils, both τ and TDP-43 aggregate in addition to α-syn [94]. There is also a substantial amount of evidence for cross-talk between Aβ and α-syn pathologies. NMR evidence shows molecular interactions between these proteins [91], and in a double transgenic mouse model, Aβ exacerbates α-syn accumulation and neuronal deficits [92]. Aβ induces phosphorylation of α-syn at Ser129 in vitro and pSer129 α-syn in brain tissue homogenates is related to the level of Aβ and Braak NFT stage [135]. Similar to Alzheimer’s disease, there is also evidence that synaptic degeneration is important in synucleinopathies with pathology associated with synaptic toxicity in vitro and in vivo [127, 154]. Less is known about molecular interactions of pathological proteins in synapse degeneration in synucleinopathies. τ reduction did not prevent motor deficits in α-syn overexpressing mice, indicating that τ is not necessary for synaptic deficits in this line [103].

Both FTLD and ALS are heterogeneous diseases that can be sub-divided by their pathological lesions. Over the past few years, a substantial overlap has been observed in the pathologies, genetics, and clinical phenotypes of these complex diseases. FTLD can be subclassified in disorders that accumulate (1) τ, i.e., FTLD-tau, (2) TDP-43, i.e., FTLD-TDP, and (3) other FTLD-forms that accumulate other proteins (e.g., fused in sarcoma (FUS) etc.) [90]. FTLD-TDP and ALS share neuronal TDP-43 aggregates and a certain number of ALS cases also develop FTLD-TDP and vice versa [107].

FTLD-tau comprises several morphologically different tauopathies: Pick’s disease (PiD), corticobasal degeneration (CBD), progressive supranuclear palsy (PSP), argyrophilic grain disease (AGD), and neurofibrillary tangle-predominant dementia (NFTD) [90]. All these diseases have in common that τ is the sole protein that aggregates in these disorders. These diseases usually affect individuals below 70 years of age and concomitant AD-associated Aβ and τ as well as TDP-43 and LBD-associated α-syn may potentially be present but are only rarely seen [147] and hence not assumed to influence the disease process. Moreover, τ pathology in FTLD-tau is, except for NFTD, not restricted to neurons and neuronal processes. It also develops in astrocytes (astrocytic plaques in CBD, tufted astrocytes in PSP, τ-positive astrocytes in AGD) and oligodendrocytes (coiled bodies in AGD, CBD, PSP, PiD) [33]. Except for AGD, FTLD-tau disorders cause dementia and in most cases do not show co-pathologies [147]. AGD is usually present with low-level AD-related pathology and such cases do not show clinical dementia [72, 145]. Demented AGD cases, however, showed significantly more severe AD-related pathology (Aβ-plaques and NFTs) than non-demented AGD cases but less than in demented cases that had pure AD pathology, suggesting additive effects of AGD and AD pathology [72, 145]. τ pathology is also observed in some ALS cases, where it has been reported to correlate with cognitive decline [162].

FTLD-TDP is characterized by intracellular aggregates of phosphorylated TDP-43 in the neuronal cytoplasm, dendrites [107] as well as in oligodendroglial cells [106]. In FTLD-TDP and ALS cases, not only full length TDP-43 aggregates are found, but also truncated fragments, mainly C-terminal fragments of TDP-43, are also found [107]. In a subset of FTLD-TDP and ALS cases, hexanucleotide repeat expansions in the chromosome 9 open reading frame 72 (C9ORF72) have been seen. In these mutation carriers, dipeptide aggregates corresponding to abnormal translation of the pathological hexanucleotide repeats can be seen in the cytoplasm of neurons [101]. However, these aggregates do not correlate with disease severity or the coexisting TDP-43 pathology in the studies reported so far. The amount of TDP-43 pathology correlates more closely with the clinical severity of the disease in the studies conducted to date [89]. TDP-43 pathology in patients with the behavioral variant of FTLD-TDP (FTLD-TDP cases that show characteristic behavioral changes as key feature of the disease = bvFTLD-TDP) and with sporadic ALS show a characteristic pattern that suggests anatomical stages of TDP-43 pathology spreading that differed among bvFTLD-TDP and ALS [16, 18].

TDP-43 pathology is also found in AD cases [4, 153]. Recently, Josephs and colleagues reported that TDP-43 pathology in AD is seen in over 50% of cases where it spreads in a distinct pattern [70, 71], which was also observed in LBD and aged controls by others [95]. Moreover, colocalization of phosphorylated TDP-43 and τ-containing NFTs has been reported but it is not yet clear whether this finding is due to unspecific co-staining of NFTs with the pTDP-43 antibody or whether indeed phosphorylated TDP-43 accumulates in NFTs. A similar questionable colocalization of τ and TDP-43 is speculated in granulovacuolar degeneration (GVD) [75]. GVD has been shown to represent an AD-related lesion [139] exhibiting phosphorylated TDP-43 as well as numerous other proteins in neuronal cytoplasmic vacuoles including τ [34] and phosphorylated Aβ [83].

There is substantial overlap in the genetics of FTD and ALS [115]. Mutations in C9ORF72, TARDBP, FUS, VCP, UBQLN2, SQSTM1, and CHMP2B have all been associated with both ALS and FTD [46, 84], and mutations in hgRNPA2B1 and can cause multisystem proteinopathy and ALS [80]. TARDBP mutations (TDP-43 gene) can lead to FTLD-TDP and/or rare ALS cases. Usually, no other proteins are aggregated in non-age-related amounts in these cases [74]. However, a case report shows a TARDBP mutation in a fAD case indicating a potential link from TDP-43 directly to AD [100]. C9ORF72 mutations go in most instances along with TDP-43 pathology either with a FTLD-TDP or an ALS-related pattern [31, 116]. In single cases, AD-related pathology also occurred as relevant pathological lesion in C9ORF72 mutation carriers, but the interpretation of this finding is controversial [19, 30, 54]. One of the most interesting aspects of C9ORF72 mutations is that in families with a distinct C9ORF72 mutation, the same mutation in some family members causes FTLD, whereas other family members with the same mutation developed ALS [20]. As such, other factors than the mutations appear to act as disease modifiers with impact on the primary vulnerability of distinct neurons. Since the C9orf72-related dipeptide aggregates do not correlate with the clinical symptomatology but with the TDP-43 pathology [107, 121], it is tempting to speculate that TDP-43-specific properties play a role in the primary targeting of the disease. In rare cases, mutations in the valosin-containing protein lead to FTLD-TDP or ALS [158]. Here, nuclear TDP-43-valosin-containing protein aggregates are the predominant lesion although cytoplasmic TDP-43 aggregates can also in different amounts [107, 121]. In this disorder, similar aggregates were found in skeletal muscle and in the myocardium [58, 158].

Pathological proteins typically associated with ALS and FTLD also show multiple mutual interactions including co-aggregation [10]. An interesting phenomenon emerging in the ALS protein aggregation literature is the role of low complexity domains in the RNA-binding proteins implicated in disease pathogenesis contributing to pathological protein aggregation. Low complexity domains mediate liquid–liquid phase separation which in healthy cells contributes to the assembly of membrane-less organelles such as RNA transport granules and stress granules; but in disease, the liquid phase transition initiates fibrillization of aggregates [99]. This may explain the co-aggregation of different low complexity domain-containing proteins such as TDP-43 and FUS [79]. Interestingly, dipeptide repeat proteins produced from unconventional non-ATG dependent translation of C9ORF72 expansions, the most common cause of ALS and FTLD, have recently been found to interact with RNA-binding proteins with low complexity sequence domains in Drosophila [86]. This finding provides a potentially unifying mechanism for the disparate causes of ALS/FTLD converging on disrupted phase transitions that impair the assembly, dynamics, and function of membrane-less organelles that cause widespread cellular abnormalities including impairments in nucleolar function, nucleocytoplasmic transport, RNA splicing, translation, and abnormal protein aggregation.

Here, we have reviewed neuropathological and experimental findings, which together point to pathogenic links between the aggregation of Aβ, τ, TDP-43, and α-syn. These links are most evident in fAD cases with mutations in the APP, PSEN1, or PSEN2 genes because these cases often show aggregates of all four proteins in the brain due to mutation-driven production of high amounts of Aβ. Even a larger number of sporadic AD cases exhibit Aβ, τ, and TDP-43 aggregates [4, 70, 153]. Given the fact that τ-aggregates develop in restricted areas of the brain before Aβ, TDP-43, and α-syn lesions occur, one could speculate that τ pathology initiates AD and the aggregation of other proteins is just a secondary phenomenon. However, most people will accumulate brainstem τ pathology (usually between 40 and 50 years of age) [15] but only 32% will develop symptomatic AD in their lifetimes [3], and genetic evidence strongly implicates changes in Aβ as crucial for initiating the disease process. Thus, τ pathology alone appears not to be sufficient to develop AD as defined according to the recommended criteria [60], Aβ pathology must also be involved. In other words, one can hypothesize that accumulation and aggregation of abnormal τ-protein aggregates is an age-related phenomenon, which becomes overtly pathogenic and allowed to spread throughout the brain in the presence of Aβ aggregates. Post-translationally modified forms of Aβ, such as Aβ_N3pE and _pSer8Aβ, may play an important role in the maturation of these aggregates [117]. Thus, maturation of protein aggregates, especially Aβ aggregates, appears to play a major role in the development of the disease [117]. In animal models for Aβ pathology, Aβ pathology alone is not sufficient to develop the full spectrum of AD pathology [39, 56, 134], although it is possible that this is in part due to the lifespan of rodents and in part due to differences in rodent and human τ. Notwithstanding these caveats, the Aβ-expressing animal model findings do support our hypothesis that τ pathology is an essential prerequisite for the development of AD but only in combination with Aβ. τ-mutations in the absence of Aβ (in man and in transgenic mice) lead mainly to a non-AD tauopathy and only very rarely are associated with AD [27, 59]. Moreover, the AD pathology in APP mutation carriers develops in adulthood after a normal childhood, i.e., when τ pathology becomes more and more prevalent in the published autopsy cohorts meaning that once abnormal τ occurs, the disease starts to develop, implying that τ is a prerequisite for disease.

Interactions between Aβ and TDP-43 or α-syn may follow similar principles as that with τ. The AD-specific progression pattern of AD-related TDP-43 aggregates [70] is different from that in FTLD-TDP and ALS cases [16, 18], which argues in favor of the idea that Aβ also exaggerates AD-related TDP-43 pathology. An argument for the interaction between Aβ and α-syn aggregates may be the frequent coincidence of full-blown AD and LBD pathology in demented patients that were clinically diagnosed as LBD patients (PDD and/or DLB), as well as the detection of LBDs in fAD cases [88, 125]. Moreover, in the absence of Aβ, TARDBP and SCNA mutations do not cause AD but ALS or PD [74, 100, 113].

Other proteins (in addition to Aβ) may also have triggering effects for neurodegeneration. The most attractive candidates are dipeptide aggregates in C9ORF72 hexanucleotide repeat mutation carriers developing ALS or FTLD-TDP with TDP-43 aggregates [101]. Moreover, a case report also suggests that mutant paraplegin (SPG7 mutation) triggers both brainstem τ and α-syn pathology in the absence of Aβ [148]. That one MAPT mutation causes α-syn changes in addition to τ pathology and that a mutation in LRRK2 can induce both, α-syn—and in a few cases τ—pathology [52, 84] further argues for a link between α-syn and τ.

Thus, several proteins can lead to neurodegeneration alone or in concert with one another. Intraneuronal and intracellular protein aggregates appear to be most relevant for the neurodegenerative process because they occur in all of the discussed neurodegenerative disorders and correlate well with cognitive decline and neuron loss [6, 45, 138]. It appears that the neuropathological type of neurodegenerative disorder simply reflects the circumstances of protein aggregation and that the pathogenic process of protein aggregation behind the phenotype has common principles in several neurodegenerative disorders, such as spreading of disease-related pathology [12, 14, 17, 143] and the involvement of post-translationally modified proteins in the disease-specific aggregates [38, 107, 140].

An argument for the hypothesis that the “phenotype of neurodegeneration” is the result of a critical accumulation of soluble and/or insoluble protein aggregates in neurons—regardless of the protein that aggregates—is provided by an artificial mouse model that produces intraneuronal Aβ aggregates after cleavage from an enkephalin-signal-peptide Aβ₄₂ construct, which shows significant neuron loss and an ALS-like phenotype with serious motor dysfunction (APP48 mice) [1]. Under this artificial paradigm, Aβ and/or its soluble and/or insoluble aggregates show different neurodegenerative effects than under the more AD-related paradigm of APP overexpression [118]. However, post-translational modification of Aβ similar to that seen in AD cases also occurred in the intracellular Aβ₄₂ aggregates in this artificial mouse model [118] indicating that post-translational modifications may play an important role for the neurodegeneration process in general. Thus, artificial expression of an aggregation-prone protein in neurons causes neurodegeneration similar to other intraneuronal proteins that develop aggregates such as τ or TDP-43, whereas mainly extracellular deposition of the same protein had only limited effects supporting the idea that the intracellular protein accumulation may be a prerequisite for neurodegeneration, whereas extracellular Aβ aggregates may catalyze spreading of intracellular aggregates while interacting and aggregating with the respective proteins.

The above-mentioned considerations argue against the classical view of the amyloid cascade hypothesis insofar as Aβ does not seem to the one and only player that causes all the rest of AD pathology. Amyloid as a key factor in AD pathogenesis is strongly supported, particularly by genetics of fAD, but it is also clear that alone, amyloid pathology is insufficient to drive neurodegeneration. The weight of the evidence on mechanisms of neuron loss in AD sides with τ being the driving toxic force. The data we have summarized here on co-occurrence of amyloid and τ-pathologies in human brain and their mechanistic interactions in model systems support the intriguing possibility that in sporadic AD, age-related τ-accumulation as observed in PART occurs earlier and independently from Aβ, thus representing a prerequisite for Aβ to kick off the pathogenesis of AD. In the event that both proteins in an aggregated form co-occur in the brain, they appear to become particularly toxic and presumably interact to lead to the devastating neurodegeneration in AD. Lesions typical for other neurodegenerative disorders or vascular lesions may co-exist and either interact with the AD-related protein aggregates or producing brain lesions that are added to the AD-related neurodegeneration. Other tauopathies, TDP-43 proteinopathies, and synucleinopathies may follow a similar cascade of prerequisite lesion plus secondary disease-specific pathogenic event as depicted in Fig. 3.Fig. 3

Cascade hypothesis for neurodegenerative disorders: Pathological protein interactions contribute to neurodegeneration. Schematic representation of the interplay between the different neurodegenerative protein aggregates and their related diseases. As a result of the current knowledge about disease progression and experimental evidence about protein-aggregate interactions in vivo an in vitro, we hypothesize that abnormal τ-protein accumulation is an event that happens during aging in the brain stem of nearly everyone above 40 years of age [15] (1 prerequisite for disease). AD appears to develop when Aβ aggregates occur and initiate more pathological accumulation of τ and its spread through the brain (2 disease-specific pathogenic event and 3 disease). Maturation of plaque-associated and soluble/dispersible Aβ-aggregates thereby appear to be a critical event in the progression of the disease. Abnormal τ-protein aggregates, TDP-43 aggregates, and α-syn aggregates occur in smaller subset of individuals during age as shown in Fig. 1 (it is not clear whether it is a result of aging or the early stages of neurodegeneration). Once Aβ aggregates prevail in a certain amount or biochemical maturation stage, it is tempting to speculate that these Aβ aggregates may also exaggerate/catalyze TDP-43 or possibly also α-syn pathology in pattern similar to that seen in AD or AD-LBD (4 α-syn pathology is typical for PD/DLB and can be interpreted as pure coincidence or as influenced by/influencing AD pathology or its clinical picture). In the absence of Aβ and/or after exposure to other triggers (e.g., disease-specific mutations, excitotoxicity in Guam disease [35]) τ, α-syn, and TDP-43 aggregates develop other neurodegenerative disorders such as FTLD-tau, FTLD-TDP, ALS, or PD/DLB. AD Alzheimer’s disease, FTLD frontotemporal lobar degeneration, ALS amyotrophic lateral sclerosis (synonymous with MND—motor neuron disease), LBD Lewy body disease [including Parkinson’s disease (PD) and DLB]

τ, TDP-43, and α-syn alone lead to specific neurodegenerative disorders that do not necessarily show multiple proteins aggregating in the respective disorders such as in FTLD-tau, FTLD-TDP, ALS, and PD, whereas AD appears to have at least a two-step pathogenesis with alterations in Aβ likely initiating the actual disease process, but with τ accumulation and spread through the brain being the essential step to cause disease symptoms. It is possible that pathological accumulations of τ independent of Aβ are a prerequisite for AD with extracellular Aβ being the second step that exaggerates or catalyzes the pathological process leading to clinical dementia. Similarly, TDP-43 and α-syn pathologies probably develop independently (and can cause clinical disease) but extracellular Aβ may aggravate or catalyze the spread of these pathologies throughout the brain. Accumulating evidence suggests that the fibrillar aggregates of all of these proteins are not likely directly toxic to neurons but reflect—and in some cases exacerbate—the accumulation of toxic soluble forms of the proteins.

Data from human autopsy material and experimental model systems indicate multiple interactions of pathological proteins both directly in terms of co-aggregation and indirectly, for example, involvement in the same molecular pathways to neurodegeneration. Recent data also make it tempting to speculate that pathological proteins interact in the process of spread of the disease through the brain. However, our knowledge on cerebral multimorbidity is still limited as this multimorbidity shows considerable qualitative and quantitative heterogeneity between cases, and hence large-scale studies on human post-mortem brains, which combine both detailed clinical data and quantitative data on the amount of protein-aggregate burden, are needed to further our understanding of protein–protein interactions in the multimorbid old brain. Understanding these interactions will be important to develop effective therapeutic targets to reduce toxicity and possibly to stop these diseases in their tracks (Fig. 3).