Alzheimer’s Disease is defined by the deposition of senile plaques, neurofibrillary tangles and progressive neuronal loss. A longstanding controversy surrounds what precedes the neurodegeneration—what is the trigger, or series of triggers, that leads to neuronal dysfunction and, subsequently, neural system failure? Pinpointing the discrete stages at which neurons degenerate will elucidate targets for therapeutic intervention. A growing body of evidence has suggested that calcium dyshomeostasis might be an important, and possibly critical, step towards initiating cell death mechanisms. Most of this evidence, however, comes from reduced systems in which cultured cell lines, or acute brain slices are used to study the intracellular regulation of calcium. These reports demonstrated that exogenous amyloid-β application or expression of mutant PS1 can lead to altered calcium levels. Without measuring such changes in an intact neural system, and alongside AD-relevant pathology, it is difficult to draw conclusions as to whether calcium overload is indeed relevant in AD and, just as importantly, which mechanisms might be responsible.
In vivo multiphoton microscopy provides an ideal approach to measure changes to brain structure and function in aged, adult mouse models. To accurately measure intracellular changes in calcium in neuronal processes, along with neuritic morphology requires a fluorescent probe that can fill not only soma, but also dendrites, axons and spines. We found that yellow-cameleon 3.6 when expressed via an adeno-associated vector (AAV2) allowed such a measurement. As a FRET-based probe, it is capable of reporting quantitative [Ca2+]i when calibrated appropriately. Just as importantly, it exhibits relatively fast kinetics and robust response properties to physiologically relevant stimuli making it useful for a wide variety of neuroscientific applications. We exploited gene transfer and intravital multiphoton microscopy to image YC3.6 in layer 5 pyramidal neurons in transgenic mice that have mutations associated with AD—both mutations in APP and PS1.
We found that amyloid-β plaque deposition was required to induce calcium overload. This was confirmed using four different mouse models—two of which develop plaques, the APPswe/PS1- ΔE9 mouse line which has an accelerated deposition of plaques and the Tg-2576 (APPswe) line. No calcium overload was seen in either of the PS1-mutant mice (PS1-ΔE9 or PS1-M146V) that we investigated. Importantly, neither of these FAD-linked mutations alone leads to plaque deposition. Furthermore, in 3.5 mo-old APP/PS1 transgenic mice, an age before plaque deposition, there was no appreciable calcium overload. These data provide evidence that mutant PS1, though important in accelerating plaque deposition when coupled with mutant APP, does not play a role in the calcium overload in neurites that we measured. It does not preclude, however, a more subtle role for mutant PS1 in calcium handling. In fact, based on recent evidence from acute brain slices in multiple transgenic mouse lines, it is likely that PS1 mutations might affect intracellular calcium signaling by modulating release from the ER (Stutzmann et al., 2006
). These more subtle changes to intracellular calcium might work alongside the calcium overload we describe.
The calcium overload we observed in vivo
in transgenic mouse models was coupled to the deposition of senile plaques, and was most pronounced in the immediate vicinity of senile plaques. This suggests that senile plaques are the proximal event in the cascade, but does not preclude a diffusible toxic species, e.g. Aβ oligomers, enriched near plaques. We propose a multi-stage degenerative process in which amyloid-β aggregates induce calcium influx as an initial acute trigger. The resulting moderate calcium overload initiates a second pathological stage that activates the phosphatase, CaN. Increased CaN activity then triggers the final stage of neurite structural degeneration and even more severe calcium overload. Recent evidence suggests that CaN activation is directly linked to dendritic structural degeneration and neuritic beading (Zeng et al., 2007
). By inhibiting CaN with FK-506, we showed that we can prevent the third stage—structural degradation of calcium overloaded morphologically-intact dendrites and axons—thereby halting the progressive degeneration at an intermediate phase. In behavioral studies, CaN inhibition with FK-506 partially restored learning and memory in Tg-2576 mice (Dineley et al., 2007
). Our data now allow a mechanistic interpretation of these results. We suggest that the restoration of behavioral deficits may reflect the prevention of neuritic beading and severe calcium overload that otherwise would irreversibly interrupt neural circuit function important for learning and memory.
We have also shown another major functional consequence of calcium overload: disruption of the spino-dendritic signaling interface. Normally, spines compartmentalize calcium from the local parent dendrite thereby allowing independent signaling of multiple spines along the same dendritic cable. Dendrites overloaded with calcium no longer have spines that can compartmentalize their calcium. Rather, the dendrite and spine are fully coupled, no longer allowing independent signaling events. This deficit may underlie the neuronal network dysfunction in AD.
Taken together, the activation of CaN and the alterations of spino-dendritic coupling suggest another important consequence—one with specific implications for cellular mechanisms of learning and memory. The calcium overload in layer 1 spiny dendrites in APP mice was approximately 500 nM; this range is similar in magnitude to levels that have been associated with long-term depression (LTD) of synaptic strength (Sabatini et al., 2002
), suggesting a link between the cellular mechanisms of LTD and the pathophysiology of AD. Multiple lines of evidence have shown that activation of CaN leads to LTD (Malenka and Bear, 2004
; Mulkey et al., 1994
)—taken together, we posit that the calcium overload and subsequent CaN activation is sufficient to induce LTD. This hypothesis is further supported by evidence that a sustained increase in [Ca2+
can initiate LTD (Tanaka et al., 2007
) and that local application of Aβ can induce LTD-like mechanisms in vitro
(Snyder et al., 2005
The calcium overload we observed in the transgenic animals and associated LTD mechanisms could contribute to the reduced spine density present throughout the cortex and which is particularly striking within 25 microns of a senile plaque (Spires et al., 2005
). Our in vivo
data suggest that calcium overload is also most acute near plaques, reinforcing the idea that senile plaques are focal sources of toxicity. Neurites with highly elevated [Ca2+
are more likely to lack spines; in neurites from WT or APP/PS1 mice with normal [Ca2+
, the ratio of spiny to aspiny processes is ~4:1. This ratio changes almost four-fold to nearly 1:1 among neurites with calcium overload—consistent with the hypothesis that calcium overload is associated with spine loss via a CaN-mediated pathway (Hsieh et al., 2006
). Taken together with recent studies, this suggests that [Ca2+
-elevations might yield changes in the strength of synaptic connections which might, in turn, lead to the chronic loss of spines. Thus amyloid-β accumulation impacts spines in at least two ways: the structural loss of spiny processes near plaques as well as marked functional impairment of calcium compartmentalization at the spino-dendritic interface. The magnitude of these changes in the cortex is remarkable, with nearly 20% of all neurites exhibiting calcium overload. These data provide novel in vivo
support for synaptic plasticity mediated changes to network properties that are crucial for learning and memory.
We have determined at least three specific downstream functional consequences of calcium overload—spino-dendritic calcium de-compartmentalization, structural neuritic alterations and CaN activation—and highlighted an important intermediary stage that can be targeted for therapeutic intervention—inhibiting CaN activity leads to amelioration of the structural and functional deficits and may have behavioral implications. A critical factor that remains to be determined is the specific pathway that amyloid-β activates to induce calcium influx. The effect may be mediated by the formation of calcium-conducting pores comprised of the Aβ peptide (Arispe et al., 1993a
; Arispe et al., 1993b
; Pollard et al., 1993
; Simakova and Arispe, 2006
), modulation of voltage-gated calcium channels (Brorson et al., 1995
; Ekinci et al., 1999
; MacManus et al., 2000
), or an Aβ-mediated oxidative stress that results in altered calcium regulation (Behl et al., 1994
; Yan et al., 1996
; Yuan and Yankner, 2000
). Nonetheless, we show that mutant APP processing and subsequent plaque deposition play a critical role in inducing calcium overload in these AD models and activating calcineurin-dependent neurodegenerative processes.