Common belief has held that in Alzheimer disease, buildup of neurofibrillary tangles within neurons switches on caspases that eventually kill the cells. However, an analysis published March 31 in Nature provides evidence for the reverse—namely, that caspase activation may precede tangle formation. Using multiphoton imaging to visualize tangles and activated caspases in real time in the brains of tau-overexpressing mice, researchers led by Brad Hyman at Massachusetts General Hospital, Charlestown, propose a model where caspase activation occurs early, and forthcoming tangles mark, rather than cause, neurodegeneration.

Postmortem studies in tau transgenic mice and AD patients have found that tangles map to brain regions littered with dead (Gómez-Isla et al., 1997; Ramsden et al., 2005) or misshapen (Augustinack et al., 2002) neurons, and that cleaved tau colocalizes with activated caspases and other apoptotic markers (Ramalho et al., 2008; Guo et al., 2004; Rohn et al., 2001). These observations have helped fuel the notion that tangles cause neurodegeneration in AD and other tau-related dementias.

But when Hyman and colleagues began using multiphoton imaging to monitor tau pathogenesis in vivo in Tg4510 mutant human tau transgenic mice, they were surprised to find that tangle-bearing neurons with activated caspases did not die immediately (Spires-Jones et al., 2008 and ARF related news story). In a later analysis, these cells hung around for at least a day after initial imaging (de Calignon et al., 2009), raising the possibility that tau aggregation, even caspase activation, did not spell instant neuronal death in this tauopathy model.

In the current study, lead author Alix de Calignon and colleagues explored these issues further in the Tg4510 model, and in wild-type mice injected intracranially with tau-expressing viruses. Using multiphoton imaging through a cranial window, the researchers detected tangles by staining with thioflavin S, and caspase activation by applying fluorescent indicators that bind activated enzyme. They monitored more than 500 tangle-positive neurons in four seven- to nine-month-old Tg4510 mice (which develop tangles and neuronal loss by seven months) and found they stayed alive through two to five days of observation, reinforcing what they saw previously over shorter timeframes. Consistent with previous studies by the Hyman lab (Spires-Jones et al., 2008), very few neurons (i.e., less than 1 percent) had active caspases, but the vast majority (>87 percent) of those also had tangles. Exceedingly few neurons were caspase-positive without tangles.

Focusing on this rare subclass—of which they found only 22 neurons over 24 imaging sites in three mice—the researchers discovered that 20 (or 91 percent) of the cells formed a new tangle by the time they were imaged the next day (see image below). By contrast, less than 2 percent of caspase-negative neurons formed a new tangle within a day’s time. The authors conclude from these data that tangles can form quickly—in under 24 hours—and that caspases are activated before tangles develop.



Caspases First, Then Tangles
Thioflavin S (green) and caspase indicator (red) were applied on the surface of the brain of Tg4510 mutant human tau transgenic mice. Some tangle-free cells showed caspase activation upon initial imaging (time 0), and most formed a new tangle within a day (time 1). Image credit: Alix de Calignon

Curiously, while virtually all new tangles formed in neurons with caspase activity, the researchers also found that more than 90 percent of tangle-bearing neurons were caspase-negative. This suggested that tangle-containing cells “survive the initial caspase attack and become caspase-negative,” the authors write.

What, then, is responsible for turning on the caspases? Hyman and colleagues figured it might be soluble mutant tau, given that they could not find a single caspase-positive neuron in six wild-type mice or two amyloid precursor protein (APP)-overexpressing mice. To test whether soluble mutant tau was the culprit, they took advantage of the doxycycline-controlled transgene expression of the Tg4510 mice. The researchers monitored for tangles and caspase activation in seven- to eight-month-old Tg4510 and wild-type mice at baseline, and after two or six weeks of doxycycline treatment to block mutant tau expression. Over the six weeks, caspase activation dropped ~20-fold in tangle-bearing neurons, suggesting to the authors that soluble tau molecules are upstream of caspase activation.

Furthermore, they showed that caspase activation colocalized with a truncated form of tau generated by caspase cleavage at aspartate 421 (D421). They did this by imaging the Tg4510 mice in vivo using a pan-caspase indicator, then sacrificing the animals and analyzing their brains for D421-cleaved tau (as measured by an antibody specific for this neo-epitope).

To confirm this in another model, de Calignon and colleagues injected wild-type mice intracranially with viruses carrying wild-type tau, creating a scenario that more closely resembles sporadic AD. In these tau-overexpressing mice, the neurons containing activated caspases also had caspase-cleaved tau. This colocalization of caspase activation and D421-cleaved tau also occurred in older hTau transgenic mice that express wild-type human tau.

The scientists went on to show that this cleaved form of tau could seed aggregation in vivo. Using viruses, they introduced D421-truncated tau into wild-type mice and found that more than a third of neurons expressing this cleaved form developed pathogenic conformations of tau (as measured by Alz50 conformation-specific antibodies). The abnormal Alz50-positive structures also contained endogenous tau. This suggests that truncated tau acts like a loss-of-function mutant, recruiting endogenous tau into aggregates and thereby removing tau from stabilizing microtubules and preventing it performing its normal functions.

In a recent study, Gail Johnson, University of Rochester, New York, and colleagues reported another ill effect of caspase-cleaved tau—compromised mitochondrial function. (Quintanilla et al., 2009). The current paper “provides strong evidence that this form of tau occurs in an AD paradigm, and we can think of it as causing toxic changes to occur,” she said in an interview with ARF. “Everybody's been focusing almost exclusively on tau phosphorylation, but this indicates that caspase cleavage may be a lot earlier and more significant than we realize.” In human disease, Johnson suggested, “You may have cycles of slight caspase activation throughout the life of neurons, until the load of cleaved tau becomes too much for the cell to handle.”

Johnson and others have some concerns. In the author’s proposed model, large amounts of tau accumulate inappropriately in neuronal cell bodies, where they activate caspases that cleave tau to spur tangle formation, at which point caspase activity tapers off and neurons survive in a low-functioning state despite the presence of tangles. This model only describes what could happen when tau is overexpressed and does not address what happens in human disease, noted Jürgen Götz, University of Sydney, Australia, in an e-mail to ARF. Johnson agreed, suggesting that there may be upstream events that cause tau to accumulate in the cell bodies, where they can activate and be cleaved by caspases.

Furthermore, Götz suggested that biochemical evidence is needed to confirm that soluble tau rather than tangles is the trigger for neurodegeneration. Some cells judged to be tangle-free may actually contain fibrillar tau that has not yet reached a level detectable by thioflavin S, he wrote. “Electron microscopy or Western blotting of sarkosyl or formic acid-extracted brain tissue would be helpful.” (See full comment below.)

Others worried that the fluorescent indicator used to stain activated caspases has skewed conclusions because it functions as a pan-caspase inhibitor. The authors said they were aware of this and did a control experiment, which showed that the FLICA indicator wore off quickly relative to the daily imaging they did. “We are confident that even though FLICA covalently binds the active site of caspases, it does not inhibit caspase activity for an extended period in the conditions we used,” de Calignon wrote in an e-mail to ARF (see full comment below). She noted that none of the paper’s reviewers raised concern about the caspase indicator.

All told, the study “really sets up the field and says okay, we have to stop just looking at aggregates,” Johnson said. “It provides strong evidence that neurofibrillary tangles are not toxic. Cells are long-lived after formation of tangles.” Meanwhile, at least several companies are developing drugs that reduce tau filaments. At the 2008 International Conference on Alzheimer’s Disease in Chicago, TauRx Therapeutics, Ltd., a Scottish-Singaporean biotech company, presented Phase 2 data on a drug that interferes with tau aggregation by acting on self-aggregating cleaved tau fragments (see ARF related conference story).—Esther Landhuis.

de Calignon A, Fox LM, Pitstick R, Carlson GA, Bacskai BJ, Spires-Jones TL, Hyman BT. Caspase activation precedes and leads to tangles. Nature. 31 March 2010. Abstract


Make a Comment

To make a comment you must login or register.

Comments on News and Primary Papers

  1. This is a provocative study. It clearly supports the idea that caspase activation lies upstream of tangle evolution. Much of the data in this study validates what our group has been investigating in the postmortem AD brain for the past 10 years.

    As far back as 2001 (Rohn et al., 2001), we put forth a hypothesis that caspase activation and cleavage of tau are early events that may precede tangle formation. We confirmed this idea in a 2002 study (Rohn et al., 2002), whereby we were the first to demonstrate the caspase-cleavage of tau in the human AD brain. In this paper, we actually provided data involving caspase-9 in the human AD brain that are now explained by Brad Hyman's group. In our 2002 study, a quantitative analysis indicated that as the number of neurons containing neurofibrillary tangles (NFTs) increased, the extent of caspase-9 activation decreased, supporting the idea that caspase-9 activation may precede NFT formation. As Hyman and colleagues show, caspase activation is initiated, but then for some reason is no longer evident, in the same tangle-bearing neuron, similar to what we found with activated caspase-9 in the human AD brain.

    We believe the time has come to test whether caspase inhibitors are a viable approach for the treatment of AD. Personally, it has been very difficult to obtain funding for such studies even in mouse models of AD due to concerns by reviewers regarding toxicity. But until studies are actually carried out testing whether caspase inhibitors work, one can only speculate on the relative risk versus reward of using caspase inhibitors as therapeutic agents (Rohn, 2010).


    . Correlation between caspase activation and neurofibrillary tangle formation in Alzheimer's disease. Am J Pathol. 2001 Jan;158(1):189-98. PubMed.

    . Caspase-9 activation and caspase cleavage of tau in the Alzheimer's disease brain. Neurobiol Dis. 2002 Nov;11(2):341-54. PubMed.

    . The role of caspases in Alzheimer's disease; potential novel therapeutic opportunities. Apoptosis. 2010 Nov;15(11):1403-9. PubMed.

  2. Hyman and colleagues use a powerful in vivo multiphoton imaging approach through craniotomy to monitor the sequence of tau pathology in the inducible tau transgenic mouse model Tg4510, and in mice intracranially injected with tau-expressing AAV (adeno-associated viruses). They come up with the conclusion that caspase activation precedes and leads to neurofibrillary tangles. The data would even suggest that tangles are almost something like a salvage pathway.

    The authors also conclude, based on a combination of staining methods, that tangles develop within a day’s time, which is highly reminiscent of their earlier findings, also published in Nature, that amyloid plaques develop within the same time frame (Meyer-Luehmann et al., 2008). The question is, What happens in human disease? Interestingly, in the model presented in the supplement (Supplementary Figure 6), the authors present a “Tangle formation hypothesis” that would operate in tau-overexpressing mice, but not necessarily in humans. Specifically, the authors state that “Expression of human tau (WT or P301L mutant) in a mouse brain, either by transgenesis or viral infection, leads to mislocalization of protein tau in the neuronal cell bodies.”

    Hyman and colleagues, in the first part of the manuscript, count caspase-positive neurons and tangle-bearing neurons, and by determining the relative ratios (90 percent of tangle-bearing neurons are not caspase-positive versus 88 percent of caspase-positive, tangle-negative cell developing a tangle within 24 hours), they conclude that tangle-bearing neurons survive an initial caspase attack and then become caspase-negative. They also identify soluble tau rather than tangles as the culprit in AD. What is missing and what should be addressed in future studies is biochemical evidence backing these findings. Also, while images are shown representing some neurons with soluble tau and others (the end-stage) containing mature tangles (as evidenced by thioflavin S positivity), what happens in between is something like a black box. What is considered as a cell that is tangle-free could, in fact, contain already a lot of fibrillar tau, which however has not reached a level that would be detected by thioflavin S. Here, clearly, electron microscopy or Western blotting of sarkosyl or formic acid-extracted brain tissue would be helpful.

    What is only mentioned in passing is that caspase-positive neurons are not observed in APP transgenic mice. It would be worthwhile determining whether there are synergistic effects in mice with a tau and Aβ pathology.

    While there is an emphasis on truncation of tau, hyperphosphorylation is not considered as an initial step. It is reasonable to assume that the final word is not yet spoken. A problem, certainly, is that it is generally difficult to reduce pathological effects to (tau or Aβ) species that are underrepresented and either escape detection because of this underrepresentation (see, e.g., Aβ*56; Lesné et al., 2006) or because their defining identity is not known.

    A proof whether tangle formation is preceded by caspase cleavage of soluble tau, as the paper would suggest, would be best addressed by crossing tau mutant mice onto a caspase null background.

    View all comments by Jürgen Götz
  3. In response to concerns about the caspase inhibitor, we performed a control experiment, at the beginning of our study, to determine whether FLICA indicator inhibits caspase activation in vivo for an extended period. For this experiment, we used sequentially two different colors of FLICA.

    A craniotomy was performed on an eight-month-old Tg4510 animal, and a solution of red FLICA was applied in the same conditions described in the paper. Caspase-positive cells were imaged. After one hour, a solution of green FLICA was applied. Re-imaging the same sites of the brain showed that caspase-positive cells were labeled with both colors, indicating that, within one hour of the initial application of the FLICA reagent, new caspase activity could be detected.

    We are confident that even though FLICA covalently binds the active site of caspases, it does not inhibit caspase activity for an extended period in the conditions we used it.

    View all comments by Bradley Hyman
  4. There is a good chance that caspases play a physiological role in synaptic plasticity. Mattson and colleagues (1998a; 1998b) showed that activated caspases could be detected in synaptosome preparations and distal dendrites of intact cells. Although their interpretation was focused on pathology, they also concluded that "[o]ur data suggest the possibility that 'synaptic apoptosis' can occur independently of the cell body." This may mean that a caspase-dependent mechanism participates in removal of dendritic spines during normal synaptic remodeling. One of the triggers for such synaptic caspase activation is, in fact, a neurotransmitter. Thus, it should not be too surprising that neurons can persist for several hours or days after caspase activation. Whole-cell apoptosis may be something of an accidental overrun of physiological caspase activity into a quantitative state that crosses a toxic threshold.

    Perhaps it should also be considered that the tau in dendrites is phosphorylated at different sites from that found in axons. This may exist to permit normal caspase activity in dendrites from initiating toxic events stemming from tau modification. Maybe pathology arises from some unfortunate combination of caspase activation with incorrect phosphorylation patterns on tau.

    View all comments by Steve Barger
  5. The findings of Alix de Calignon and colleagues are interesting and support several studies using human AD tissue that show caspase plays a pivotal role in tangle formation. However, given the recent indications that AD has a preclinical phase of between 20-30 years, it remains unknown whether tangles develop within the same relatively short time frame in the human condition as suggested to occur in the brains of Tg4510 mutant human tau transgenic mice. It would have been interesting to determine whether the reported mouse tangle displays filamentous tau at the age examined, which would add additional support to the argument that the evolution of NFT pathology described in the current paper actually produces a true AD-like tangle. Despite these questions, I compliment the authors on an excellent paper!

    View all comments by Elliott Mufson
  6. Evading Apoptosis in Alzheimer Disease
    Comment by Baiyang Sheng, Hyoung-gon Lee, George Perry, Mark A. Smith, Xiongwei Zhu

    Using multiphoton live brain imaging in a mouse model of Alzheimer disease (AD), Hyman and colleagues were able to investigate the in vivo relationship between caspase activation and neurofibrillary tangle (NFT) formation. They suggest a new model whereby the accumulation of free cytosolic tau activates caspases; thereafter, caspase activation cleaves tau to initiate NFT formation, truncated tau recruits normal tau to misfold and form NFT. Neurons that contain NFT were unexpectedly long lived and remain alive, indicating that NFT-forming neurons escape from apoptosis.

    These findings provide compelling support for a number of controversial concepts. First, the term “abortive apoptosis” (1), abbreviated as “abortosis,” was first proposed several years ago as a mechanistic explanation for the disconnect between apoptotic mediators and effectors in AD (2), as well as the mathematical improbability of apoptosis playing a key role in AD (3). A second important concept is that NFTs are not obligate features of neurodegeneration and may in fact be protective (4). Supporting this, tau phosphorylation typically results as a consequence of cellular stress (5) and displays antioxidant and other protective effects (6,7) that enable NFT-containing neurons to survive for a decade (8) or, as shown here, evade apoptotic processes.

    With respect to the relationship between caspase activation and NFT formation, a number of aspects merit further study. For example, how does free cytosolic tau activate caspases? The authors use caspase probes selective for caspase 6, caspase 3/7, and a pancaspase indicator. However, whether initiator caspases (caspases 8 and 9) are also activated is mechanistically important to determine. Moreover, how do neurons survive caspase activation? What is the particular role of NFT in apoptosis? In the latter, it is notable that hyperphosphorylated tau, the major protein subunit of NFTs in AD, has been shown stabilize β-catenin, suggesting a possible mechanism of NFT-bearing neurons to exit from a caspase-induced apoptotic program (9).

    View all comments by Xiongwei Zhu


News Citations

  1. Imaging Tau and Caspases, Aβ’s Synaptic Effects
  2. Chicago: Out of the Blue—A Tau-based Treatment for AD?

Paper Citations

  1. . Neuronal loss correlates with but exceeds neurofibrillary tangles in Alzheimer's disease. Ann Neurol. 1997 Jan;41(1):17-24. PubMed.
  2. . Age-dependent neurofibrillary tangle formation, neuron loss, and memory impairment in a mouse model of human tauopathy (P301L). J Neurosci. 2005 Nov 16;25(46):10637-47. PubMed.
  3. . Specific tau phosphorylation sites correlate with severity of neuronal cytopathology in Alzheimer's disease. Acta Neuropathol. 2002 Jan;103(1):26-35. PubMed.
  4. . Apoptosis in transgenic mice expressing the P301L mutated form of human tau. Mol Med. 2008 May-Jun;14(5-6):309-17. PubMed.
  5. . Active caspase-6 and caspase-6-cleaved tau in neuropil threads, neuritic plaques, and neurofibrillary tangles of Alzheimer's disease. Am J Pathol. 2004 Aug;165(2):523-31. PubMed.
  6. . Correlation between caspase activation and neurofibrillary tangle formation in Alzheimer's disease. Am J Pathol. 2001 Jan;158(1):189-98. PubMed.
  7. . In vivo imaging reveals dissociation between caspase activation and acute neuronal death in tangle-bearing neurons. J Neurosci. 2008 Jan 23;28(4):862-7. PubMed.
  8. . Tangle-bearing neurons survive despite disruption of membrane integrity in a mouse model of tauopathy. J Neuropathol Exp Neurol. 2009 Jul;68(7):757-61. PubMed.
  9. . Caspase-cleaved tau expression induces mitochondrial dysfunction in immortalized cortical neurons: implications for the pathogenesis of Alzheimer disease. J Biol Chem. 2009 Jul 10;284(28):18754-66. PubMed.
  10. . Caspase activation precedes and leads to tangles. Nature. 2010 Apr 22;464(7292):1201-4. PubMed.

Other Citations

  1. Tg4510

External Citations

  1. hTau
  2. TauRx Therapeutics, Ltd.

Further Reading


  1. . Caspase activation precedes and leads to tangles. Nature. 2010 Apr 22;464(7292):1201-4. PubMed.
  2. . In vivo imaging reveals dissociation between caspase activation and acute neuronal death in tangle-bearing neurons. J Neurosci. 2008 Jan 23;28(4):862-7. PubMed.
  3. . Tangle-bearing neurons survive despite disruption of membrane integrity in a mouse model of tauopathy. J Neuropathol Exp Neurol. 2009 Jul;68(7):757-61. PubMed.

Primary Papers

  1. . Caspase activation precedes and leads to tangles. Nature. 2010 Apr 22;464(7292):1201-4. PubMed.