7 September 2007. In the September 6 Neuron, researchers present electroencephalography (EEG) recordings of freely moving hAPP-transgenic mice that reveal evidence of spontaneous non-convulsive seizures. Surprisingly, while these abnormal discharges came on, the mice did not behave as if they were having an epileptic attack. They merely stopped while the electrical storm raged through their brain, and then continued scampering about in their cage. This observation forms the core of an ambitious paper that develops a new hypothesis for a contributing cause to the cognitive deficits in AD. The hypothesis proposes that Aβ-driven overexcitation of entorhinal-hippocampal circuits in turn leads to compensatory inhibitory changes in hippocampal networks, and that both processes together restrict synaptic plasticity and contribute to learning and memory deficits. Lennart Mucke at the Gladstone Institute of Neurological Disease in San Francisco, California, and his coworkers collaborated with epilepsy researcher Jeffrey Noebels of Baylor College of Medicine in Houston, Texas, for the study.
By characterizing epileptiform activity and associated changes in mouse models, the study draws attention to prior observations that people with AD have an elevated risk for seizures. This is known and has been reported as part of the clinical description of AD pedigrees and in epidemiological studies. Seizures occur especially at the early stages of AD and in autosomal-dominant early-onset forms (eFAD). Even so, the human observation has not generated a focused research effort to date. The extent of the problem is currently unknown, in part because caregivers and doctors of people with AD might not notice abnormal brain discharges that stay below the level of the convulsive fits typically associated with epilepsy. Follow-up studies trying to suppress the spontaneous overexcitation with anticonvulsant drugs can test the hypothesis, and also might lead to new therapeutic approaches.
Continuing a research trend, the paper represents a departure from the field's long-standing focus on the historic twin pathologies of AD, i.e., amyloid plaques and neurofibrillary tangles. The word “plaque” appears only tangentially and the word “tangle” not at all, even though Aβ is seen as an upstream cause of all brain changes described and tau as playing a role, as well.
First author Jorge Palop and colleagues developed a hunch that aberrant network activities might be going on when they analyzed biochemical and anatomic changes in the dentate gyrus region of the hippocampus of mutant human APP-transgenic mice that produce high concentrations of Aβ peptide. The Mucke team and others had noticed reductions in the calcium-binding protein calbindin, the immediate-early gene Arc, changes in the phosphorylation of NMDA receptor subunits and other molecular changes that suggest that synaptic plasticity in the networks underlying memory was off. But there was a nagging paradox, in that some of these dentate gyrus alterations seemed to indicate increased neuronal activity and an excitotoxic reason for neurodegeneration, while at the same time the available data for Aβ's effect on synapses suggested that it decreases glutamatergic transmission, possibly by endocytosing AMPA or NMDA receptors. These coexisting clues to both excitotoxicity and synaptic suppression raised the questions of what the net effect of Aβ is. The present paper proposes that both overexcitation and subsequent inhibitory responses are at play.
The authors studied four lines expressing FAD mutant or wild-type human APP. In brief, hAPP FAD (J20) mice had molecular and circuit changes that indicate imbalances between excitatory and inhibitory neuronal activity in the hippocampus. These included increased expression of neuropeptide Y (NPY), axonal sprouting by interneurons, and alterations in the levels of NPY receptors—all in a pattern that suggested overexcitation. Moreover, the authors found increased excitatory mossy fiber innervation of GABAergic basket cells that inhibit these granule neurons. These changes overlap partly though not fully with known changes in epilepsy models. In epilepsy these neurons end up being less inhibited, whereas in the APP models they appear to be more strongly inhibited, and voltage-clamp recordings of the granule cells from J20 APP mice showed that they indeed have more frequent and high-amplitude miniature inhibitory post-synaptic currents (mIPSCs) than do control mice.
A tip-off that these changes might point to some sort of epileptiform activity came from previous research in that field. In an accompanying preview article, epilepsy researchers Soren Leonard and James McNamara of Duke University Medical Center in Durham, North Carolina, cite two prior studies that had found the same calbindin, NPY, and mIPSC changes in the dentate gyrus occurring a day after recurrent seizures in kindling models of epilepsy (Tonder et al., 1994; Nusser et al., 1998). These authors had postulated that the changes were a homeostatic response to the seizures. (Kindling refers to a phenomenon where repeated subconvulsive stimulation in a given area prompts the stimulated fibers to fire spontaneously afterwards; these "after discharges" can wax and sensitize the site of stimulation so that even a weak stimulus there days or weeks later can trigger a full-blown seizure. Kindling involves NMDA receptors and is thought to represent an underlying mechanism of chronic epilepsy.)
Pharmacologic experiments gave Palop and colleagues additional hints that the hippocampal changes in the APP-transgenic mice might be responses to overexcitation. In non-transgenic mice, injection of the excitatory amino acid kainate led to changes resembling those observed in APP mice, that is, increases in NPY, depletion of Arc and calbindin. By contrast, tau reduction prevented kainate-induced alterations in NPY. Recently, the Mucke lab had shown that reducing tau expression by half in the J20 improved cognition in these Aβ overexpressors and made them resistant to induced seizures (Roberson et al., 2007). In this study, coauthor Erik Roberson analyzed these bigenic mice further to show that they suffered neither NPY increase nor calbindin depletion. Together, these data suggested to the investigators that Aβ promotes neuronal excitation, and they conducted two types of experiment to test this notion.
First, Palop and colleagues challenged three different lines of mutant human APP mice with the GABA receptor antagonist pentylenetetrazole (PTZ). This triggered earlier and more severe seizures in all three transgenics than in wild-type controls. A specific difference within the J20 line tracked with the changes in the dentate gyrus: the most susceptible mice, which died from their seizures, had greater increases in NPY and losses of calbindin and Arc than did mice with less severe seizures or controls.
Secondly, the authors collaborated with Noebels' group to monitor EEG activity in six freely behaving adult J20 mice plus control mice with implanted electrodes. The recordings revealed generalized and synchronous discharges in all cortical electrodes in all transgenic but no control mice. Electrodes implanted more deeply into the hippocampus recorded similar discharges. Most striking among the abnormal discharges were non-convulsive seizures that started slow, accelerated, and then ended abruptly into a long cortical depression. During the seizures, the mice stayed still.
Further experiments confirmed that the sorts of synaptic alteration that the authors and other groups have previously linked to Aβ also occur in the dentate gyrus of the J20 mice analyzed for epileptiform activity. These include reduced phosphorylation of the NR2B subunit of NMDA receptors, and reduced levels of the GluR1 and GluR2 subunits of AMPA receptors. Finally, field EPSP recordings from hippocampal slices of these mice revealed synaptic plasticity deficits whose precise nature was specific to the circuit analyzed, suggesting that the dentate gyrus is supremely vulnerable to Aβ-induced network disruptions.
Overall, the authors take the spontaneous seizures to mean that the net effect of excess Aβ on the affected networks is excitatory, and that many of the observed alterations in the mice's brains likely reflect attempts of the networks to tamp down this overexcitation. Both these forces would eventually crimp the ability of entorhinal-hippocampal networks to retain new information. A question the present study leaves open is exactly how excess Aβ would “kindle” the excitability of the affected memory networks in the first place.
The similarities between J20 mice and epilepsy models go only so far, the authors emphasize. In the former, circuit changes tend to dampen granule cell excitability, whereas in the latter, circuit changes rev it up. In the J20 mice, and perhaps in AD, this tug-of-war between Aβ-driven excitation and compensatory inhibition might explain why overexcitation rarely escalates to fully fledged convulsive seizures yet still impairs cognitive function. Overt seizures are not a part of the typical clinical presentation of AD, and specific studies are needed to establish how common spontaneous epileptiform activity is in sporadic and familial AD, the authors state. Conceivably, such activity could explain the daily fluctuations in AD patients that caregivers know so well, where at times the loved one seems sharp but soon after may become confused and disoriented. If this work is independently confirmed, it also raises the prospect that some anticonvulsant drugs, many of which are available and widely used, might suppress the seizures and the depressions between seizures. If a drug does this, would it also prevent the molecular and circuit changes and, most importantly, the cognitive deficits? Both mouse and human studies to test this question are warranted, the authors conclude.—Gabrielle Strobel.
Palop JJ, Chin J, Roberson ED, Wang J, Thwin MT, Bien-Ly N, Yoo J, Ho KO, Yu GQ, Kreitzer A, Finkbeiner S, Noebels JL, Mucke L. Aberrant excitatory neuronal activity and compensatory remodeling of inhibitory hippocampal circuits in mouse models of Alzheimer's disease. Neuron. 2007 Sep 6;55(5):697-711. Abstract
Leonard AS, McNamara JO. Does epileptiform activity contribute to cognitive impairment in Alzheimer's disease? Neuron. 2007 Sep 6;55(5):677-8. Abstract
Q&A with Jorge Palop and Lennart Mucke. Questions by Gabrielle Strobel.
Q: What do the videotaped seizures in the mice look like?
A: It is a type of seizure activity that is not accompanied by the usual twitching and jerking movements seen in many forms of epilepsy. In fact, it took sophisticated brain wave recordings in freely behaving mice by electroencephalography (EEG) and telemetry to detect the seizure activity. These seizures are non-convulsive, so they cannot be easily detected by just watching the mice.
Q: How about in AD patients? Do people just stop what they were doing while a storm of epileptiform activity races through their cortex and hippocampus, then rest?
A: This possibility deserves to be explored. Transient episodes of amnestic wandering and disorientation in AD patients have been associated with epileptiform activity (Rabinowicz et al., 2000).
Q: What exactly is the news of your paper? People have known about seizures in AD patients, and at least some transgenic mouse data exist also.
A: Our paper shows that high levels of Aβ are sufficient to trigger abnormal overexcitation and compensatory inhibitory responses in the very brain networks that are responsible for learning and memory. Notably, most of the seizure activity was non-convulsive, which means it could easily be missed by clinical observations. The increased inhibitory activity we identified may explain why the overexcitation does not more frequently escalate into frank convulsive seizures in hAPP mice and AD patients. Our results also suggest that the suppression of the overexcitation might prevent and possibly even reverse cognitive impairments induced by high levels of Aβ.
Although many people in the field may not realize this, I believe that we were the first to demonstrate in hAPP-expressing hippocampal slices that Aβ inhibits glutamatergic synaptic transmission (Hsia et al., 1999). Subsequently, these findings were beautifully extended by other groups (Kamenetz et al., 2003; Hsieh et al., 2006; Shankar et al., 2007). Extrapolating from these findings, one might have predicted that Aβ decreases overall network activity. Yet, we found the opposite to be true. Several potential explanations were offered in our paper to reconcile these observations. Indeed, we demonstrated experimentally that epileptiform network activity can coexist with deficits in neurotransmission strength or plasticity at specific synapses.
Q: You describe epileptiform activity, presumably as a result of excess Aβ, and homeostatic compensatory reactions that constrain plasticity in learning and memory networks. Turning this around, do some people with epilepsy have learning impairments, and are they at higher risk for dementia?
A: Some people with epilepsy have learning impairments, but the etiology of these impairments is likely as complex as the etiology of epilepsy itself. We are unaware of dementia risk studies in people with epilepsy. Notably, our study suggests that the changes Aβ elicits in the brain overlap with epilepsy only partially, but not completely.
Q: You describe the inhibitory changes as compensation for prior overexcitation by Aβ. How does Aβ cause the original overexcitation? The clues that exist to date—about LTP, NMDAR, and AMPAR subunits—would seem to be inhibitory. In other words, what exactly touches off those spontaneous seizures?
A: We are pursuing this question very actively. As discussed in our study, there are at least three possible ways to reconcile the suppressive synaptic effects of Aβ with the overall network overexcitation we identified:
1. Depressed glutamatergic transmission may be a synaptic compensatory mechanism against overexcitation.
2. Inhibitory interneurons may be more susceptible to the suppressive effects of Aβ on glutamatergic synaptic transmission than excitatory principal neurons, leading to an overall increase in network excitability.
3. Cortical or subcortical regions that control neuronal excitability on a broad scale could be particularly susceptible to Aβ-induced impairments of glutamatergic synaptic transmission, increasing overall network excitability.
Q: Do your data suggest anticonvulsant drugs for treatment of AD?
A: Indeed, particularly during early stages of the disease and in cases with early onset.
Q: Valproate has been tried (ADCS trial; Porsteinsson, 2006), but there are many different anti-seizure drugs. What about others?
A: The valproate studies we are aware of have targeted primarily behavioral alterations in patients with relatively advanced disease. We are unaware of clinical trials of anticonvulsants for MCI and early AD. In addition, our ongoing studies suggest that some anticonvulsants may be better at suppressing Aβ-induced overexcitation than others. We are eager to extend these studies to human subjects.
Q: Are you, or other labs, now testing anticonvulsants in mutant hAPP mice?
A: We are and imagine that other labs will follow suit.
Q: Are seizures more common in FAD because this form of disease is most strongly driven by Aβ overproduction?
A: That is a possibility.
Q: APP- and presenilin-based FAD is rare. What makes your work relevant to sporadic AD?
A: The overlap between familial and sporadic appears wide at the clinical and pathological level. It is likely that there is also substantial overlap at the etiological level. Clearly, Aβ also accumulates to high levels in the brains of people with sporadic AD, and sporadic AD cases also have an increased incidence of seizures, as reviewed in our paper.
Q: In people with AD and MCI, how common is this subtle epileptiform activity?
A: We don't know, but are launching a clinical study to answer this important question.
Q: Where it does not rise to the level of a full-blown convulsive seizure, can it be quantified in people who are cognitively impaired and may be depressed and more passive than normal, as well?
A: Telemetry EEG recordings can be carried out also in humans and can detect seizure activity that is hard to detect clinically.
Q: There is a literature on using EEG to detect AD. Do these groups pick up on the epileptiform signals, or on different changes in EEG recordings?
A: Most EEG studies in AD have focused on frequency analysis, although there have been some reports of epileptiform activity, for example, the report by Rabinowicz et al., 2000 cited above.
Q: Might these groups have datasets already sitting in drawers and computers that would help get at the question of how frequent epileptiform activity is in AD?
A: This is a distinct possibility we would like to explore. Unfortunately, some AD patients with epilepsy have been excluded from EEG studies because it has not been widely appreciated that seizure activity may be part of the pathogenesis of this illness.
Q: More generally, does your study have any implications for the use of EEG in AD?
A: We think so, although we predict that special leads and sophisticated recording techniques may be required to monitor activity in the most relevant networks.
Q: Have you gotten feedback from epilepsy research groups to your data? What kind?
A: Jeff Noebels, coauthor of our paper, is an expert in epilepsy research in humans and mouse models. He was excited when we approached him about this collaboration and even more so when the EEG data started rolling in.
Q: What are the next steps?
A: We plan to use mouse models to identify drugs that can prevent and reverse Aβ-induced network dysfunction. We will also embark on clinical studies to determine if our results and conclusions are relevant to people with AD.