de Calignon, Alix MSc; Spires-Jones, Tara L. DPhil; Pitstick, Rose BA; Carlson, George A. PhD; Hyman, Bradley T. MD, PhD
Neurofibrillary tangles (NFTs) composed of abnormally hyperphosphorylated tau protein assembled into paired helical filaments are, along with senile plaques and neuronal loss, a major pathological hallmark of Alzheimer disease (1). Intraneuronal NFTs are also pathological hallmarks of tauopathies, a large group of disorders of cognition and movement that includes progressive supranuclear palsy, corticobasal degeneration, Pick disease, frontotemporal dementia, and parkinsonism linked to chromosome 17 (2). Although it is widely known that accumulation of insoluble tau aggregates correlates with disease progression, current understanding of the mechanisms involved is insufficient to conclude that there is a causal relationship between NFTs and neuronal death.
The rTg4510 mice reversibly express P301L mutant human tau and develop neurofibrillary pathology, neuronal loss, and memory impairment (3-5). In this model, NFTs appear as early as 2.5 months of age; neuronal loss is dramatic and approaches approximately 50% loss of cortical neurons by 8.5 months of age (5). Both tangle formation and neuron loss are region specific and seem to be dissociated processes. For example, loss of neurons occurs before neurofibrillary lesions do in the dentate gyrus; conversely, neurofibrillary pathology appears without major cell loss in other areas such as the basal ganglia (5). In the cortex and hippocampal CA fields, however, the extent of tangle accumulation and neuronal loss is correlated, suggesting that tangles either precede and lead to neuronal death or are independent but concurrent pathophysiological events.
We previously showed that it is possible to visualize NFT in the brains of live rTg4510 mice using 2-photon in vivo imaging through a cranial window. We tested the idea that apoptosis-like mechanisms were responsible for neuronal death in rTg4510 by developing a method to image activation of caspase during in vivo imaging (6). Caspase activation could be detected in vivo in a small percentage of neurons; there was nearly universal colocalization of caspase-positive neurons and NFTs, but these neurons did not proceed to an immediate apoptotic death during the subsequent hours of observation. We hypothesized that the caspase-positive neurons are under cellular stress, but that they do not die immediately. Therefore, we postulated that using a later marker for cell death would be a better method for revealing all dying neurons.
In the present study, we sought to develop another marker that might reflect additional (i.e. nonapoptotic) cell death mechanisms and that would not be dependent upon enzymatic activation of a probe. We used propidium iodide (PI), a widely used fluorescent marker of cell death in tissue culture assays (7). Propidium iodide is excluded from cells with intact membranes but can access dead or dying cells. When PI is intravenously injected into mice subjected to middle cerebral artery occlusion, PI-positive neurons can be observed postmortem in association with ischemic damage (8). Thus, PI labels compromised neurons in mouse brains. Moreover, its fluorescence increases 20- to 30-fold upon binding to DNA so that background signal is low and positive cells are identified as having bright red nuclei (7). We also used the Hoechst fluorescent DNA-binding dye that freely diffuses through membranes and labels nuclei of living cells blue. We found that during in vivo imaging of 7- to 9-month-old rTg4510 mice, PI labels a small number of neurons. These neurons nearly universally also stain with the active caspase reagent and contain NFTs. In accordance with our observations using the caspase reagent alone, neurons that are caspase and PI positive remained intact 24 hours after initial imaging, thereby supporting the concept that NFT-containing neurons are under substantial stress but do not immediately die.
MATERIALS AND METHODS
Surgery and Imaging
All animal experiments were carried out following National Institutes of Health and Massachusetts General Hospital Committee on Research Animal Care guidelines. We used 7- to 9-month-old rTg4510 mice that reversibly express P301L mutant human tau (n = 6) (3). The mice are bred by Rose Pitstick and George A. Carlson at McLaughlin Research Institute, Great Falls, MT. Age-matched littermates that express only the activator transgene were used as negative controls (n = 4). A craniotomy was performed as previously described (9). Briefly, animals were anesthetized with isoflurane (0.5%-2%; Baxter, Deerfield, IL) in balanced oxygen, and a 6-mm circle was drilled on the dorsal surface of the skull (centered on the midline with the anterior tip of the circle immediately anterior to bregma), and the 6-mm circle of encompassed skull was removed. Before a glass coverslip was sealed over the craniotomy, the dura was resected from the cortical surface, and PBS containing thioflavin S (0.025% thioflavin S; Sigma, St Louis, MO), a fluorescent indicator of caspase activation (FLICA indicators, 5× concentration, polycaspase indicator; Molecular Probes, Eugene, OR) and PI (10 μg/mL; Molecular Probes) was applied for 20 minutes. For longitudinal imaging, Hoechst 33342 (Invitrogen, Carlsbad, CA) solution in 1 mmol/L PBS was applied with the other fluorescent markers (n = 2) and imaged the next day.
In vivo imaging was performed using a Bio-Rad 1024ES multiphoton microscope (Bio-Rad Laboratories, Hercules, CA) with 800-nm excitation (Ti:Sapphire laser; Mai Tai, Spectra-Physics, Mountain View, CA) for thioflavin S, as previously described (10). We empirically found that 800 nm excites PI, caspase indicator, and thioflavin S optimally on our Mai Tai laser system. We used 750-nm excitation on the following day to detect Hoechst 33342 fluorescence. For quantification of NFTs and cell death markers, images were taken in a systematic random fashion through the window area to ensure unbiased sampling (11). For each animal, z-stacks from 8 different areas of 307.92 × 307.92 × 52 μm (representing 5% of the window area) were collected. Observations of the same neurons on consecutive days were possible using XY coordinates recorded for each high-resolution (5×) image of cells of interest; lower-resolution (1×) images of the surface of the brain were used as reference to identify the same site based on the blood vessels; 2× images of the same focal plane provided an overview of the area with neighboring cells.
After completion of the experiments, the animals were killed with an overdose of Avertin (2,2,2-tribromoethanol in t-amyl alcohol [Sigma], 400 mg/kg). The brains were harvested, fixed in 4% paraformaldehyde containing 15% glycerol cryoprotectant for 2 to 3 days, and then either sectioned or stored at 4°C in a 20% sucrose solution.
The 1× and 2× images were taken as a location reference with a normal rate of scan speed. The rate of scan speed was low for high-resolution (5× and 10×) images to assess detailed cell morphology. Laser intensity was 30%, which was presumably low enough not to cause cell damage. The photomultiplier tube voltage was unchanged for all imaging sessions at 750 V for the blue channel, 850 V for the green channel, and 1,050 V for the red channel. Imaging times were between 1 and 20 seconds, depending on the depth of the scan.
Images were taken as z-stacks. All images were processed with ImageJ and presented as z-projections. Intensity, contrast, and color balances were used to pseudocolor the images and to enhance the quality of some images.
The brains were frozen, and serial coronal sections 40 μm thick were cut on a freezing microtome. An unbiased stereological counting method was used to determine the percentage of thioflavin S-positive neurons containing a Hoechst-positive nucleus. The optical fractionator method was used as previously described (12). The image analysis system CAST (Olympus), mounted on an upright BX51 Olympus microscope with an integrated motorized stage (Prior Scientific, Rockland, MA), was used for outlining regions, sampling, and counting. Tangle-bearing neurons with or without a visible nucleus were counted in a 21.8 × 21.8 × 36-μm counting frame with a guard volume of 2 mm at the edges of the section to avoid sampling tangles that had lost their nuclei during sectioning. Dissectors were placed using a meander sampling paradigm (a grid placed at random angles on the region of interest) to sample 2% of the tangles in the area under the cranial window.
To assess whether the cells with PI-positive tangles observed in vivo died leaving "ghost tangles" (those without a healthy-appearing Hoechst-positive nucleus postmortem), we designated cells with PI-positive tangles and ghost tangles as "dying" neurons. The χ2 statistic from contingency table analysis (using the Stat View program; SAS Institute, Cary, NC) was used to determine whether the percentage of putative dying tangle-bearing neurons (PI positive) in vivo was the same percentage as those that are actually dying (as assessed by the absence of a nucleus associated with the tangle postmortem). A χ2 value of p < 0.05 was considered significantly different, implying that PI-positive cells in vivo are not dying.
Small Numbers of Neurons With NFTs Contain Cell Death Markers
To quantify the PI uptake into neurons, we used in vivo multiphoton imaging through a cranial window and sampled randomly throughout the cortical area accessible for imaging.
A small proportion of cells with NFTs (6%) had PI-positive nuclei, indicating disruption of membrane integrity (Table, Fig. 1). To ensure that there was no interference between the compounds, we performed a control experiment by applying PI first, imaging PI-labeled cells, and then applying thioflavin S or Hoechst and reimaging the same brain area. Results with these sequential applications were consistent with our earlier findings (data not shown) (6). We also confirmed by stereological counts on postmortem tissue that the numbers of cells with PI uptake in the window area were comparable to the numbers found by multiphoton imaging (data not shown); this suggested that the relative rarity of PI-positive neurons detected was not caused by insensitivity of the in vivo imaging methodology.
Simultaneous imaging of NFTs and caspase activity showed that a similarly small number of cells with tangles had some caspase activation. No labeled cells were detected by either method in nontransgenic littermate animals. Most of the NFT-containing neurons were not positive for cell death markers, indicating that the presence of NFTs does not necessarily lead the neuron to a cell death pathway.
The Cell Death Markers Colocalize and Appear Specifically Within Tangle-Bearing Neurons
In most of the labeled cells, caspase indicator or PI was found in tangle-bearing neurons (Fig. 2A), indicating that NFTs are not only associated with caspase activation but also with loss of membrane integrity, a later hallmark of cell death. Using a green polycaspase indicator in combination with PI (Fig. 2B), we determined whether the 2 cell death markers were colocalized. They were almost always seen together in cells (Fig. 2C), suggesting that activated caspase is associated with membrane destabilization. The caspase indicator was mostly within the cytoplasm, whereas the PI label was exclusively within the nucleus, thus allowing visualization of both markers in green NFT-bearing cells (Fig. 2D). Taken together, these data are consistent with our previous study showing caspase activation, and they further show that caspase activation is associated with loss of membrane integrity in a minority of NFT-bearing neurons.
Death Marker-Positive Tangles Persist For 24 Hours
Despite the presence of caspase activation, NFT-bearing neurons were previously found to be unchanged for hours of imaging (6). Here, we observed a similar absence of degeneration for hours when the cells were PI positive.
To investigate whether cell death markers are associated with subsequent degeneration of NFT-bearing neurons, we reimaged the same neurons 24 hours after initial imaging (Fig. 3). All tangle-bearing neurons, including those labeled with cell death markers on the first day of imaging, were detected at both time points. Prolonged excitation of fluorescent probes can lead to relatively fast quenching, especially in vivo, where antiquenching agents cannot be applied. Thioflavin S fluorescence remained detectable despite light photobleaching, but PI fluorescence was rarely seen on the second day; caspase indicator fluorescence faded after several hours.
Because both cell death markers faded, we could not be certain whether the tangles that persisted on the second day were in living neurons that continued to have a nucleus or whether the cells had died, leaving behind ghost tangles. Ghost tangles in Alzheimer disease are observed in entorhinal cortex and hippocampus but rarely in neocortex. To address this, we used Hoechst dye as a marker of living cells. A solution of Hoechst 33342 was applied on the surface of the brain at the first imaging session. Because it is a highly lipophilic dye, Hoechst 33342 penetrated the brain tissue and crossed intact cell membranes. The next day, Hoechst fluorescence had reached Layers II to III of the cortex, where thioflavin S-labeled NFTs were observed. It was detectable at a 750-nm excitation. No evidence of ghost tangle formation was observed; that is, cells with NFTs associated with death markers on the first day of imaging contained Hoechst-positive nuclei on the second day (Fig. 3H). Postmortem analysis of the brains showed that only 1 of 233 NFT-containing neurons did not contain a clearly stained Hoechst-positive nucleus (Table), confirming that ghost tangles are, at most, a rare occurrence in this model.
To investigate the relationships between the presence of NFTs and neuronal death, we imaged markers of these events in the brains in live mice in a model of tauopathy. Our observations suggest that only a very small proportion of neurons show signs of degeneration at any individual point in time, as might be expected for a neurodegenerative process that progresses for many months.
We hypothesized that caspase activation would be an early sign of cell distress caused by apoptosis-like events and that PI staining of the nucleus (which is a widely used marker of dead and dying cells reflecting loss of membrane integrity) would be a later marker of neuronal death not dependent on a particular pathophysiological process. Our earlier studies suggested that a minority of neurons was caspase positive and that these were associated with NFTs because they did not lead to immediate death during a short period, that these were associated with NFT, and relatively stable over a short time scale since they did not lead to immediate death (6). Our calculations suggested that the number of caspase-positive cells was insufficient (by an order of magnitude) to account for the number of neurons dying each day. Therefore, we sought to develop another marker of cell death that might provide insight into whether the caspase-positive NFT-containing neurons were really dying and possibly reveal neurons undergoing non-caspase-mediated death processes.
Our results show that PI can be used as an in vivo marker of loss of membrane integrity because there was a robust staining of neurons by multiphoton microscopy. These PI-positive neurons were almost certainly distressed because they nearly always contained NFTs and displayed both early and late markers of apoptosis; caspase activation and cell membrane disruption occurred together. Previously in the same model, we found that initiator and executioner caspases seemed to be activated in tangle-bearing neurons (6). Here, we found significant colocalization of caspase activation with cell permeabilization to PI-associated caspase activation with downstream cell damage. This finding associates caspase activation with downstream cell damage.
Our data show several surprising features. Although PI is a robust marker of cell distress in vivo, we observed only a very small number of neurons that were PI positive at any imaging session. Based on cross-sectional stereological analysis collected from the entire neocortex, we calculate that there would be an approximate loss of 680 neurons per day in the volume we imaged, whereas we observed only 20% of that number to be PI positive. In this estimate, we assumed that all cortical regions undergo similar neuronal loss because the entire cortical mantle considerably shrinks in these mice. If PI staining was transient, for example, if neurons that were PI positive died and were cleared rapidly by apoptosis-related mechanisms, this dissociation could be explained using a kinetic argument. However, our longitudinal imaging data show that PI-positive neurons persist for at least 24 hours, as judged by the continued presence of the NFT, our ability to detect Hoechst-positive nuclei 24 hours after observing PI staining, and the absence of ghost tangles. This surprising result suggests 2 major conclusions: first, that neurons containing tangles, caspase activation, and even PI staining nonetheless can survive for longer than we would have expected based on in vitro use of these cell death markers. Wilde et al (13) investigated the use of PI as a marker for immediate degeneration in vivo in rat brain and found that PI-labeled CA3 pyramidal cells had no obvious histologically detectable cell damage for hours after stress was induced. After ischemic injury in mouse brain, intracerebroventricular administration of PI showed that most PI-labeled cells were not positive for terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling in areas with mild injuries; PI incorporation has been observed in cells that do undergo apoptosis in areas displaying more severe ischemic changes (8). Nonlethal caspase activation has been observed, however, in studies of synaptic plasticity and neuroprotection (14). A more recent study showed the limits of PI as a cell viability indicator for bacteria and that, in addition to nonviable cells, it also stains cells with cell-cycle activity, a stage where cell membrane integrity is temporarily affected (15).
Our results reinforce the idea that tangle-containing neurons are much more likely than non-tangle-bearing neurons to show signs of immediate distress, including caspase activation and PI staining. The large majority of tangle-bearing neurons are not caspase or PI positive, however, suggesting that tangles are associated with but do not necessarily cause activation of cell death pathways. The mechanisms that allow the damaged tangle-bearing neurons to survive remain to be understood but suggest the possibility that endogenous transitory neuroprotective mechanisms are activated in tangle-bearing neurons.
One of our goals for developing PI as a marker of cell death was to better understand the relationship between tangle formation and cell death in this model. Our calculations, based on rate of neurodegeneration from cross-sectional studies of the rTg4520 mice, suggest that there is neuronal loss occurring in the cortex that is not accounted for by rare neurons that contain a tangle and contains caspase and PI neurodegeneration markers. We are now trying to develop markers of earlier lesions, perhaps before tangle formation, to further explore this paradox.
The authors thank Karen H. Ashe for development of the rTg4510 mouse line.
1. Buee L, Bussiere T, Buee-Scherrer V, et al. Tau protein isoforms, phosphorylation and role in neurodegenerative disorders. Brain Res Brain Res Rev 2000;33:95-130
2. Trojanowski JQ, Lee VM. "Fatal attractions" of proteins. A comprehensive hypothetical mechanism underlying Alzheimer's disease and other neurodegenerative disorders. Ann N Y Acad Sci 2000;924:62-67
3. SantaCruz K, Lewis J, Spires T, et al. Tau suppression in a neurodegenerative mouse model improves memory function. Science 2005;309:476-81
4. Ramsden M, Kotilinek L, Forster C, et al. Age-dependent neurofibrillary tangle formation, neuron loss, and memory impairment in a mouse model of human tauopathy (P301L). J Neurosci 2005;25:10637-47
5. Spires TL, Orne JD, SantaCruz K, et al. Region-specific dissociation of neuronal loss and neurofibrillary pathology in a mouse model of tauopathy. Am J Pathol 2006;168:1598-607
6. Spires-Jones TL, de Calignon A, Matsui T, et al. In vivo imaging reveals dissociation between caspase activation and acute neuronal death in tangle-bearing neurons. J Neurosci 2008;28:862-67
7. Arndt-Jovin DJ, Jovin TM. Fluorescence labeling and microscopy of DNA. Methods Cell Biol 1989;30:417-48
8. Unal Cevik I, Dalkara T. Intravenously administered propidium iodide labels necrotic cells in the intact mouse brain after injury. Cell Death Differ 2003;10:928-29
9. Skoch J, Hickey GA, Kajdasz ST, et al. In vivo imaging of amyloid-beta deposits in mouse brain with multiphoton microscopy. In: Sigurdsson EM, ed. Amyloid Proteins: Methods and Protocols. Totowa, NJ: Humana Press, 2004;349-64
10. Spires TL, Meyer-Luehmann M, Stern EA, et al. Dendritic spine abnormalities in amyloid precursor protein transgenic mice demonstrated by gene transfer and intravital multiphoton microscopy. J Neurosci 2005;25:7278-87
11. Hyman BT, Gomez-Isla T, Irizarry MC. Stereology: A practical primer for neuropathology. J Neuropathol Exp Neurol 1998;57:305-10
12. Irizarry MC, McNamara M, Fedorchak K, et al. APPSw transgenic mice develop age-related A beta deposits and neuropil abnormalities, but no neuronal loss in CA1. J Neuropathol Exp Neurol 1997;57:965-73
13. Wilde GJ, Sundstrom LE, Iannotti F. Propidium iodide in vivo: An early marker of neuronal damage in rat hippocampus. Neurosci Lett 1994;180:223-26
14. McLaughlin B. The kinder side of killer proteases: Caspase activation contributes to neuroprotection and CNS remodeling. Apoptosis 2004;9:111-21
15. Shi L, Gunther S, Hübschmann T, et al. Limits of propidium iodide as a cell viability indicator for environmental bacteria. Cytometry A 2007;71:592-98