Insoluble aggregates of the microtubule-associated protein tau characterize a number of neurodegenerative diseases collectively termed tauopathies. These aggregates comprise abnormally hyperphosphorylated and misfolded tau proteins. Research in this field has traditionally focused on understanding how hyperphosphorylated and aggregated tau mediates dysfunction and toxicity in tauopathies. Recent findings from both Drosophila and rodent models of tauopathy suggest that large insoluble aggregates such as tau filaments and tangles may not be the key toxic species in these diseases. Thus some investigators have shifted their focus to study pre-filament tau species such as tau oligomers and hyperphosphorylated tau monomers. Interestingly, tau oligomers can exist in a variety of states including hyperphosphorylated and unphosphorylated forms, which can be both soluble and insoluble. It remains to be determined which of these oligomeric states of tau are causally involved in neurodegeneration and which signal the beginning of the formation of inert/protective filaments. It will be important to better understand this so that tau-based therapeutic interventions can target the most toxic tau species.
Intracellular accumulation and aggregation of the microtubule-associated protein tau is a pathological feature of a group of neurodegenerative diseases collectively called tauopathies. These include AD (Alzheimer's disease), frontotemporal dementia [including FTDP-17 (frontotemporal dementia with parkinsonism linked to chromosome 17)], progressive supranuclear palsy, Pick's disease and corticobasal degeneration (reviewed in ). In addition to being aggregated, tau in these diseases is also abnormally hyperphosphorylated and misfolded. The pathological significance of all of these aberrations in tau has been the subject of much debate over the last few years. Although the mechanisms are unclear, the general consensus now is that these tau protein abnormalities mediate neuronal dysfunction and toxicity and actively contribute to neurodegeneration, even if they coexist with other pathological proteins. In such ‘secondary’ tauopathies, such as AD in which aggregated hyperphosphorylated tau coexists with Aβ (amyloid β-peptide) oligomers and plaques, the aberrant Aβ proteins are believed to require aberrant tau species to mediate their toxic effects ([2,3] and reviewed in ).
Tau aggregates display different morphologies in different tauopathies (reviewed in ). The type of aggregate formed is determined by the tau isoforms involved and the presence of mutations in the tau gene (e.g. FTDP-17). Tangles containing insoluble paired helical filaments and straight filaments characterize many tauopathies including AD and some FTDP-17 cases, whereas twisted ribbon-like filaments are evident in other FTDP-17 cases, as well as in some sporadic tauopathies . Argyrophilic grains which contain aggregated tau not deposited into tangles, but instead accumulated within spherical structures, are evident in argyrophilic grain disease . Despite having different morphologies, a common feature of all of these tau aggregates is that they usually consist of hyperphosphorylated tau proteins. It is generally believed that hyperphosphorylation of tau precedes its aggregation and it is well known that a pool of soluble hyperphosphorylated tau species also exists alongside the tau aggregates in all tauopathies .
Research in this area is focused on identifying the pathological role of each of these abnormal tau species. It has been shown that soluble hyperphosphorylated tau proteins cause dysfunction and toxicity [8–10], but whether this changes once larger insoluble aggregates form is not clear. Some reports show that formation of insoluble aggregates is protective [11,12], whereas others demonstrate that these structures are directly toxic . The present review seeks to only briefly discuss this literature and to then focus on studies describing the less well-studied tau oligomers.
What are tau oligomers?
Tau oligomer is the term used to describe any complex comprising two or more tau molecules in a multimeric structure. They can form from both hyperphosphorylated and non-phosphorylated tau proteins and they have been shown to be both soluble (when they are made up of small numbers of tau molecules) and insoluble (when they consist of larger numbers of tau molecules). They have been identified in AD [14,15] and FTDP-17 brains  and are considered to be intermediates between soluble tau monomers and insoluble tau filaments. Given that oligomeric precursors of the pathological hallmarks in many neurodegenerative diseases are now strong candidates to be the mediators of toxicity, it is tempting to speculate that tau oligomers may be crucial players in tauopathies. However, because they can exist as soluble hyperphosphorylated species as well as insoluble hyperphosphorylated and non-phosphorylated species, their potential role in tauopathies may be more complicated.
In the present paper, we describe the various tau species that have been demonstrated and briefly summarize the evidence implicating them in disease pathology.
Soluble hyperphosphorylated tau
Phosphorylation of tau is a normal physiological event, but it is believed to become dysregulated in the brain in early stages of disease. Indeed, hyperphosphorylation and misfolding are among the earliest disease-associated changes observed in AD [17,18]. This prompted the hypothesis that there are two consequences of tau hyperphosphorylation in tauopathies: tau loss of function [reduced binding to MTs (microtubules)] (reviewed in ) and an increased likelihood of tau aggregation, either because phosphorylation stimulates filament formation or because the increased free cytosolic tau would be more likely to interact with itself and/or other factors, thus creating favourable conditions for filament formation.
One can envisage how phosphorylation may indirectly promote tau aggregation by increasing the availability of free cytosolic tau, but it is less clear how it actively promotes it. Some groups have shown that hyperphosphorylated tau self-assembles into filaments in vitro , whereas others have shown that filament formation is not phosphorylation-dependent .
Although the relationship between tau phosphorylation and its subsequent aggregation is not clear, the toxicity associated with tau phosphorylation is not refuted (reviewed in ). Phosphorylation at some sites (e.g. Ser202/Thr205 and Ser396) is associated with disruption of cytoskeletal integrity, defects in axonal transport and behavioural phenotypes in one Drosophila model of tauopathy [22,23]. Phosphorylation at other sites (including Ser262 and Ser356) is associated with learning and memory deficits in another Drosophila model . Similarly, tau toxicity appears to be critically dependent upon phosphorylation at Ser262 in yet another Drosophila model of tauopathy . In other Drosophila and rodent models of tauopathy, tau phosphorylation is evident at early time points when behavioural phenotypes emerge in the absence of filament formation [9,16,26]. These (and many other similar reports not described in the present paper) collectively show that soluble hyperphosphorylated tau proteins mediate dysfunction and toxicity. Whether these soluble hyperphosphorylated tau species are entirely monomeric tau proteins or also include small soluble tau oligomers is not clear and has not always been assessed in such studies.
There are a variety of pre-filament tau aggregates that have been identified by various groups, which are termed tau oligomers. They vary in the number of tau molecules that they contain, their relative size, their solubility and the toxicity with which they are associated. When generated in vitro, they form from non-phosphorylated tau proteins, but in vivo, they are believed to form from highly phosphorylated and misfolded tau (although this may not always be the case; C.M. Cowan, S. Hands, D.W. Allan and A. Mudher, unpublished work).
Small soluble tau oligomers
Soluble tau oligomers of differing sizes have been reported by various groups. Preparing small oligomers from recombinant tau in vitro, dimers have been reported with apparent sizes of 180 kDa  and 130 kDa , and trimers with an apparent size of 120 kDa . Sahara et al.  generated tau oligomers which first formed soluble dimeric structures that developed into oligomers containing six to eight tau molecules. Since all of these tau oligomers are formed from recombinant tau in vitro, they are unlikely to be phosphorylated. It is unclear whether these variously reported dimers and trimers are indeed different tau species or whether they represent subtle variations of the same structure.
The effect of these soluble tau oligomers on a number of physiological processes has been investigated by these researchers. Makrides et al.  have shown that their tau dimers decorate MTs and speculate that such MT-dependent oligomerization of tau may be a physiological process required for optimal MT stabilization by tau. However, such tau oligomers decorating MTs have not been reported in living tissues, so their physiological relevance has yet to be demonstrated. Patterson et al.  show that their dimers suppress fast axonal transport in a squid axoplasm model. The oligomers of Lasagna-Reeves et al. [29,32] have been shown to be toxic in vitro  and in vivo . When exposed to SH-SY5Y cells in vitro, these oligomers cause significantly more death than tau monomers or filaments . Intra-hippocampal injections of these oligomers causes profound synaptic loss and neuronal death culminating in memory defects which are not seen following injections of tau monomers or fibrils . Overall, these findings imply that small oligomeric tau species are toxic. In addition, more than one group has demonstrated that in vitro generated tau dimers aggregate to form larger tau oligomers [27,30]. Whether aggregation into larger oligomers alters the toxicity of the small tau oligomers has not been determined.
To confirm the physiological relevance of these small soluble tau oligomers, studies were undertaken to assess whether they existed in vivo. The Binder and Kayed laboratories raised antibodies specific to the tau oligomers that they made and used them to probe AD brain [27,33]. Both antibodies, TOC-1 from the Binder laboratory and T22 from the Kayed laboratory, react with pre-tangle structures in early to mid-Braak stages of AD [27,33]. However, whereas TOC-1 co-localizes with highly phosphorylated tau epitopes, T22 co-localizes with some, but not all, such epitopes. This implies that the dimers and trimers found in AD may be phosphorylated at some, but not all, tau phospho-epitopes associated with disease. Other studies have also confirmed the existence of small soluble oligomers in vivo. Ali et al.  detected soluble oligomers of 150–250 kDa in brain homogenates from transgenic Drosophila expressing R406W mutant human tau. Berger et al.  and Sahara et al.  independently identified small soluble tau oligomers of approximately 140 and 170 kDa (believed to be dimers and trimers) in brain homogenates of P301L transgenic mice. The oligomers detected by Berger et al.  appeared at very early stages of disease when memory deficits were evident in the absence of tangle formation or neuronal loss. The majority of the 140 kDa oligomers appeared to be soluble and did not contain hyperphosphorylated tau proteins. Conversely, most of the 170 kDa oligomers were detected in sarkosyl-insoluble fractions and were made up of tau proteins hyperphosphorylated at a number of epitopes (Ser202/Thr205, Ser396 and Ser422). A significant negative correlation was seen between the levels of both tau oligomers (as well as a monomeric hyperphosphorylated 64 kDa tau species) and memory function in these mice, implying that it is these tau species rather than larger tau filaments and tangles that cause this phenotype. Like the oligomers described by the Binder and Kayed laboratories [27,33], these oligomers were also present in the brains of AD and FTDP-17 patients. It is not clear whether all of these oligomers are one and the same tau multimer or whether they represent tau oligomers at different stages of maturation during the disease process.
The mechanism leading to the formation of these small soluble tau oligomers has also been investigated. It has been shown that incubation of tau with other oligomeric proteins (such as α-synuclein or Aβ) in vitro leads to the formation of 120 kDa tau oligomers, which are believed to be trimers . This shows that any oligomeric protein can seed the formation of tau oligomers. The authors speculate that tau oligomerization is stimulated by other oligomers in secondary tauopathies, e.g. by Aβ oligomers in AD and by α-synuclein in Parkinson's disease and FTDP-17. This is supported by the recent report by Henkins et al.  that both tau oligomers and Aβ oligomers co-localize in synaptosomal fractions from AD patients . It has also been postulated that oligomer formation occurs as a result of impaired clearance of misfolded tau by the ubiquitin–proteasome or chaperone systems. Two groups have shown that members of the chaperone family, HSP70 (heat-shock protein 70) and NMNAT (nicotinamide mononucleotide adenylyl transferase), selectively bind to tau oligomers both in vitro  and in vivo . Patterson et al.  demonstrated that co-incubation of HSP70 with small soluble tau oligomers suppressed the inhibition of fast axonal transport by these oligomers. Ali et al.  demonstrated that co-expression of wild-type or R406W mutant human tau in Drosophila with NMNAT, a protein believed to promote HSP70 activity, led to the clearance of the 150–250 kDa tau oligomers. These studies collectively show that promoting chaperone activity clears tau oligomers and are consistent with the suggestion that deficient chaperone activity may promote oligomer formation.
Large insoluble tau oligomers
Large insoluble tau oligomers were first described by the Takashima group and given the name GTOs (granular tau oligomers) . They consist of an average of 40 molecules of tau, with an average size of 1800 kDa and diameter of 20 nm . They have been identified in early Braak stages in human brain and can also form from recombinant non-phosphorylated tau in vitro [14,15]. GTO levels decline in late Braak stages when tangle formation is greatest, implying that they are pre-tangle structures . Indeed, in an in vitro assay of tau aggregation, GTO formation occurs first, followed by filament formation , supporting further the idea that GTOs are precursors to mature tau filaments. Since there is a significant inverse correlation between GTO levels and a number of chaperone proteins in the late Braak stages, it is postulated that GTOs form because of saturation of the chaperone system and thus inefficient removal of misfolded tau .
Whether the GTOs are in themselves toxic, protective or inert is not entirely clear. Since they are believed to be precursors of tau filaments and tangles, they could contribute to toxicity, although the toxicity of tau filaments and tangles is debatable (discussed below and reviewed in ). Similarly, they are generally believed to consist of phosphorylated species of tau because phosphorylated tau levels are high in the AD brain at the time points when GTOs are abundant. However, we have observed recently the formation of GTOs in a Drosophila model of tauopathy which are neither phosphorylated nor toxic (C.M. Cowan, S. Hands, D.W. Allan and A. Mudher, unpublished work). Clearly, the precise composition and pathological significance of GTOs has yet to be fully understood; it is possible that different GTOs form in different circumstances and the phosphorylated status of their constituent tau proteins is what determines their toxicity.
Tau filaments and tangles
Tau aggregated into filaments and tangles is believed to play an intimate part in neurodegeneration because tangle pathology has been shown by many to correlate with disease severity . This was supported further by the demonstration that mutations in the tau gene which lead to tangle pathology are associated with FTDP-17 . Moreover, extensive cell death has been reported in animal models expressing aggregate-prone tau proteins  and a rescue of toxicity is evident when aggregation is suppressed . Although these studies imply that neurodegeneration in tauopathies is mediated by toxic tau filaments, there is a growing body of evidence that refutes this hypothesis. In many transgenic rodent models of tauopathy, behavioural deficits precede filament formation [16,26]. Furthermore, in some studies, rescue of neuronal loss or behavioural phenotype by suppression of tau transgene expression does not alter tangle formation [8,38,39]. Similarly, in a number of Drosophila models of tauopathy, neuronal dysfunction, neurodegeneration and behavioural phenotypes have been described in the absence of overt tau filaments and tangles [9,10,22,23,40,41]. These studies suggest that it is a pre-tangle tau species (possibly a hyperphosphorylated soluble tau monomer or oligomer) that causes dysfunction and toxicity in tauopathies (reviewed in ). Some would even speculate that filament and tangle formation is a physiological means of sequestering these toxic pre-tangle tau species and that tangle-bearing neurons are therefore healthier and longer-lived .
This idea, although in its infancy for tau pathology, has been proposed for other misfolded proteins that constitute hallmarks of various neurodegenerative diseases: the amyloid fibrils and plaques in AD, the Lewy bodies in Parkinson's disease and the inclusion bodies in Huntington's disease. In all of these diseases, there are examples from animal models and even human immunization studies (in the case of amyloid plaques ) of a dissociation between pathological hallmark formation and neuronal dysfunction/death (see, e.g., [43–45]).
Oligomeric species of tau exist and can be identified in the AD brain. It is generally believed that they form early in the disease process, possibly as a result of inadequate clearance of misfolded tau by the chaperone/ubiquitin–proteasome systems. As the disease progresses, the tau oligomers go on to form filaments and tangles. Their pathological significance is currently not clear. However, since questions are being raised about the role of tangle pathology (as discussed above), it is conceivable that pre-tangle highly phosphorylated tau species (whether in monomeric or oligomeric states) are the key players in neurodegeneration in tauopathies. However, because tau can form a variety of oligomers, which can comprise both phosphorylated and non-phosphorylated tau proteins, and can be soluble or insoluble (depending on how many tau molecules constitute them), it will be important to understand how they relate to each other and what determines their toxic potential. It is tempting to speculate that those tau oligomers made up primarily of highly phosphorylated tau are toxic, whereas those made up of non-phoshorylated tau are inert, irrespective of their solubility. This has to be formally tested, so that therapeutic interventions can target the correct aberrant tau species in all tauopathies.
The Biology and Pathology of Tau and its Role in Tauopathies II: A Biochemical Society Focused Meeting held at Robinson College, Cambridge, U.K., 8–9 January 2012. Organized and Edited by Amritpal Mudher (Southampton, U.K.) and Makis Skoulakis (BSRC Alexander Fleming, Greece).
Funding is provided by the Alzheimer's Society UK and BUPA Foundation UK.