Background
Tau immunotherapy was a logical approach following the success of amyloid-β (Aβ) immunotherapies in mouse models but faced resistance as tau was not thought to be accessible to antibodies. However, target engagement was feasible both intra- and extracellularly. Antibodies against tau and other targets have been detected intraneuronally [
1,
2], and studies over the last several decades suggested that all amyloid diseases may be transmissible between cells under proper conditions [
3]. Following our initial report of active tau immunotherapy leading to clearance of tau aggregates in transgenic mice with associated functional improvements, several studies by us and others have confirmed and extended these findings (reviewed in [
4,
5]). Concurrently, spreading of tau pathology between cells in culture and via anatomically connected brain regions in animals has now been shown by several groups (reviewed in [
6,
7]). A few phase I trials have now been initiated on active and passive tau immunotherapies [
5]. The hope is that this approach may be more effective than targeting Aβ in the later stages of the disease as tau pathology correlates better with dementia than Aβ plaques [
8].
Although the efficacy of tau immunotherapy has been confirmed in various models, our knowledge of the mechanisms involved is rather limited. Tau antibodies have now been detected intraneuronally in several studies by a few groups [
1,
9‐
14] and such uptake shown to be necessary for acute tau clearance [
10]. However, some antibodies do not appear to be taken up in appreciable amounts and are likely to primarily work extracellularly [
15‐
17]. Such differences in uptake are well known in other immunotherapy fields and may be related to antibody charge [
18,
19]. Several tau epitopes have been successfully targeted using a similar study design (reviewed in [
5]). However, very limited knowledge exists regarding the ideal affinity of antibodies and which tau species they should bind to be effective in promoting clearance of pathological tau protein. It is conceivable that very high affinity antibodies, at least against certain epitopes, may promote tau assembly or prevent their disassembly.
Most recently, we have developed a novel set of monoclonal antibodies targeting the phospho-serine 396,404 region. Two of these, 4E6 and 6B2, enter neurons and co-localize with tau [
11]. In brain slice cultures, both antibodies reduce soluble phospho-tau after 6 weeks of treatment, and 4E6 has been shown to acutely reduce tau levels in primary neurons via an intracellular mechanism [
10,
11]. The two antibodies display different binding characteristics with 4E6 being phospho-selective and 6B2 having conformational properties influenced by phosphorylation and an apparent higher affinity for tau [
11].
We tested the efficacy of these antibodies acutely in vivo and their ability to prevent toxicity, seeding and transmission of tau pathology in primary neuronal cultures, using paired helical filaments (PHF) isolated from an Alzheimer’s brain. In addition, we examined whether neuronal uptake of antibody was necessary for efficacy, and what role timing of antibody addition had on the observed mechanism of action. Our data indicates that 4E6 acutely improves spatial learning and memory, which is associated with a reduction in soluble phospho-tau protein. Furthermore, 4E6 prevents toxicity, seeding and transmission of tau pathology even though it binds poorly to most forms of tau, whereas 6B2 is ineffective although it binds strongly to most forms of tau. These unexpected results are likely to have major implications for the clinical development of tau immunotherapies, and can be explained by 4E6’s high affinity for solubilized PHF, whereas the ineffective 6B2 binds primarily to aggregated but not to solubilized PHF. Hence, affinity for particular forms of tau predict efficacy. Further, whether the antibody is working outside or inside the neuron depends on the timing of PHF and antibody addition. Antibodies with access to both intra- and extracellular pools of pathological tau protein are likely to be more efficacious than antibodies acting only within one compartment.
Discussion
Our findings indicate that two antibodies against the same epitope region have very different effects on cognition in a mouse model of early stage tauopathy. Both antibodies have phospho-selectivity for the immunogen but differ in many ways. Interestingly, the lower affinity antibody, 4E6, is effective in acutely improving spatial learning and memory and reducing soluble phospho-tau, whereas the higher affinity antibody, 6B2, is ineffective. Importantly, we further show identical efficacy differences in a primary neuronal tauopathy culture model treated with paired helical filaments (PHF) isolated from an Alzheimer brain. This indicates that the ex vivo culture model has similar predictive validity as the mouse model although the measured parameters are not comparable.
Modest tau pathology was detected by immunohistochemistry in brains of htau mice, and the treatment and control groups did not appear to differ. Under such conditions of early stage tau pathology, it is easier to quantitate early tau pathology on western blots than by immunohistochemistry, and on such blots insoluble tau protein was clearly present in the 12–13 month old htau mice. Neither tau antibody induced changes in insoluble tau levels as measured by human specific tau antibody (CP27), although 4E6 markedly improved spatial learning and memory. Analyses of the soluble tau fraction revealed that these cognitive benefits were associated with reduced levels of phospho-tau protein (PHF-1 reactive). It is likely that under such acute treatment conditions, global changes in insoluble tau levels may not be readily achievable, whereas soluble pathological tau protein should be more amenable to clearance. Indeed, the PHF-1 antibody recognizes a phospho-tau epitope within the same region as 4E6, which may explain why this tau fraction is preferentially cleared. However, it does not appear to be oligomer specific clearance, as we did not observe any differences in T22 immunoblots between 4E6 and IgG-treated mice.
We did not observe a functional rescue of associative fear memory following acute treatment with either 4E6 or 6B2. There are many possible explanations for this. First, the training protocol used may have been too ‘strong’ to detect a subtle memory deficit. This may be particularly important because the overall tau pathology that we observed, although present, was mild. In our prior studies we only detected associative memory deficits in aged mice with greater levels of tau pathology [
24].
Despite the different model systems used, the findings obtained from ex vivo and in vivo experiments are consistent and not model dependent, which supports their validity. In both cases, 4E6 shows efficacy in preventing tau pathology and associated toxicity/cognitive impairment, while 6B2 does not.
An insight into the relevant tau species was obtained from ELISA and dot blot studies of antibody binding to soluble, solubilized, or aggregated human tau species. 4E6 recognizes primarily solubilized PHF, in an ELISA and dot blot assay, which may explain lack of more global tau changes in the animals under such acute conditions. Mice at this age with modest tau pathology may be ideal to assess acute effects of therapies, particularly under pairwise cognitive comparison as used herein, which improves the sensitivity of detecting beneficial effects. Such in vivo learning and memory benefits by 4E6 and lack thereof for 6B2 in the htau mice are in agreement with the efficacy results in vivo and in the tauopathy culture model. Interestingly, soluble tau species have recently been linked to LTP and memory in htau mice [
25].
Although the ELISA and dot blot assays provided useful information on the tau binding properties of 4E6 and 6B2, the data obtained from confocal imaging was of greater value in determining mechanism and possible explanations for the differences in efficacy. With co-incubation in the culture assay, extracellular complexes of exogenous PHF and 4E6 formed (Fig.
11l) as 4E6 binds to soluble PHF. This complex formation neutralized PHF and prevented its uptake. However, with 6B2, such complexes did not form, as 6B2 does not bind well to solubilized PHF, and PHF was detected intraneuronally (Fig.
11x). This indicates that 6B2 could not prevent PHF uptake and toxicity. These results support that antibodies can be beneficial while working in the interstitial space between cells. In the living brain, these tau-antibody complexes could then be taken up and cleared by microglia as we have seen previously [
11] and others have studied more extensively [
26,
27].
Alone, addition of 10 or 1 μg/ml PHF dose-dependently induced cell loss, as measured using LDH and NeuN levels, as well as increased total and phosphorylated tau in the remaining neurons. It spread between cell populations, through release and subsequent uptake by other neurons. To test the efficacy of our antibodies, we utilized three different dosing methods differing in the timing of tau and antibody administration. For one of the antibodies, 4E6, two of these methods, addition of the PHF and antibody together, and addition of 4E6 24 h after PHF, prevented PHF toxicity, seeding, and spread. Interestingly, although similarly effective, the mechanism through which the protection occurs differed between the dosing paradigms.
When 4E6 or 6B2 were added 24 h after PHF, these colocalized intracellularly with PHF but only 4E6 prevented PHF toxicity. Based on the confocal data from the co-incubation experiments, as well as dot blot and ELISA data, 4E6 binds better to solubilized PHF than 6B2 which reacts better with aggregated PHF and insoluble tau. Hence, this feature may explain the intracellular efficacy of 4E6. It may prevent PHF polymerization, facilitating access of lysosomal enzymes to clear PHF and/or directly neutralizing soluble PHF and preventing toxicity. However, 6B2-PHF binding may be inert without promoting disassembly. Furthermore, due to poor binding it may be unable to prevent PHF fibril formation and/or toxicity of soluble PHF.
Together, these findings explain the therapeutic efficacy of 4E6. It is capable of both extracellular blockage and intracellular clearance of PHF. Our previous data indicates that 4E6 enters the endosomal/lysosomal system within tauopathy neurons and promotes clearance of native tau, possibly by preventing aggregation [
10,
11,
13]. Other groups have observed internalization of tau antibodies [
9,
12], and lysosomal colocalization [
9]. Further, neuronal colocalization between antibody, target, and endosomal/lysosomal markers has been seen for α-synuclein antibodies in a PD mouse model [
28]. In other experiments, tau antibodies are able to block the uptake of pathological tau or improve experimental outcomes without apparently entering neurons [
15‐
17]. Whether antibodies are taken into neurons is likely influenced by several factors including, charge, target and Fc receptor affinity, and as data presented herein suggests, location of the target and timing of antibody addition.
In contrast, pre-incubation with 4E6 was ineffective at reducing toxicity or seeding of tau pathology. A likely reason is the relative lack of the target epitope under these conditions. Previously, we showed that neuronal 4E6 uptake correlates highly with pathological intracellular tau levels [
10]. When the antibody is added first, efficacy requires retention in neurons until PHF addition 24 h later. However, a shortage of the target means the antibody will remain unbound, and more prone to degradation or recycling out of the cell, as seen via confocal imaging. Although 4E6 was ineffective under these conditions, it does not rule out prophylactic administration of tau antibodies, as circulating antibodies could prevent disease initiation by clearing early-stage tau aggregates. Exogenous antibodies have a half-life of one to three weeks and lower doses could be used in pre-symptomatic individuals at risk.
Notably, the different efficacies between dosing methods were also observed in the spreading assay using the microfluidic chambers. In both model systems, only the co-incubation dosing or PHF → Ab was effective, whereas Ab → PHF was not.
In our previously published findings [
11], both 4E6 and 6B2 showed efficacy in preventing increased phospho-tau levels in a brain slice model, in contrast to 6B2’s ineffectiveness in the primary neurons in the current study. There are likely multiple factors which contribute to these differences. In the slice culture system, treatment with antibodies lasted for up to 6 weeks and no exogenous tau was introduced in that system. In the present experiments, a much shorter time scale is used (7 days as opposed to six weeks) and we are utilizing PHF tau isolated from a human AD brain. The tau found in the PHF fraction also represents a different stage of tangle formation. Furthermore, the PHF isolated from the AD brain may have additional posttranslational modifications not present in the endogenous tau of the slices. Differences in cell health over the course of the experiments are likely also influenced by the culture model. In the primary cultures, neurons lack the trophic support provided by glial cells, which are present in the brain slices.
We have shown that antibody uptake into neurons can be blocked with an antibody against FcII/III receptors or with dansyl cadaverine, which blocks receptor-mediated endocytosis [
10]. Under the co-incubation conditions, blocking antibody uptake had no effect on the outcome. However, when 4E6 was added 24 h after PHF addition, blocking its uptake prevented its beneficial effects. These findings confirm that under co-incubation conditions, the antibody is working extracellularly but when it is added 24 h after PHF, its effects are intracellular.
We have previously shown that both antibodies are taken up into tauopathy neurons in brain slice- and primary cultures, in which they colocalize with tau aggregates in the endosomal-lysosomal system [
10,
11]. Furthermore, 6B2 and its single chain variable fragment derivative can be used to image tau lesions in vivo and end up in the same neuronal compartments after peripheral injection [
14]. Such uptake and colocalization is by itself not an indication of efficacy but we have shown that prevention of 4E6 neuronal uptake blocks acute antibody-mediated tau clearance [
10]. The culture data confirms such intracellular clearance and additionally shows prevention of neurotoxicity by 4E6 in a different culture model, which is more relevant to human disease as AD derived PHF material is used. Furthermore, in our PHF-treated primary culture study, 6B2 was ineffective under various experimental conditions using multiple outcome measures. Overall, 4E6 may be better suited as a therapeutic antibody targeting soluble tau species and 6B2, or ideally its smaller derivatives with better access to the target, useful as an imaging probe for insoluble tau lesions.
Specifically, the ex vivo culture model shows that 4E6, a monoclonal tau antibody targeting the phospho-serine 396/404 region prevented toxicity and reduced tau levels induced by the addition of Alzheimer’s brain-derived PHF material. Importantly, another tau monoclonal, 6B2, which has substantially higher affinity for the tau peptide immunogen and aggregated PHF tau than 4E6, was ineffective under these experimental conditions. Further analyses revealed that 4E6 had higher affinity than 6B2 for the solubilized PHF that was used to promote toxicity in cultured neurons. This likely explains the efficacy of the former antibody and lack thereof for the latter. These findings have major implications for the development of passive tau immunotherapies. Efficacy cannot be predicted by affinity to the immunogen alone or to aggregated tau, but has to be determined in biological models of tau pathology. Combined with imaging data, these results provide information on how affinity and efficacy relate.
Other tau immunotherapy studies have reported efficacy differences between antibodies recognizing epitopes of different sequences of tau and one study between different isotypes of two antibodies of similar affinity against the same epitope (for review see [
5]). 4E6 and 6B2 are of the same isotype, IgG1, and our findings show for the first time that subtle difference in epitope recognition can profoundly affect efficacy. Importantly, we have confirmed and provided mechanistic insight into these in vivo differences in a disease relevant ex vivo neuronal culture model, in which we promote tau pathology with Alzheimer’s brain-derived PHF in primary neurons expressing familial tau mutation. Hence, the contrasting efficacies are seen consistently in different models with or without tau mutation and may have major therapeutic implications for both familial and sporadic tauopathies. The models employed have strong construct and face validity as they are based on sound theoretical rational as normal or familial (mutated) human tau is being expressed, and have the key features associated with tauopathies, namely tau aggregation, toxicity, and associated cognitive impairments in the animals. The human PHF culture model has strong predictive validity for the outcome in the animal model, but it remains to be seen if this holds up in clinical trials.