Abstract
Madar AD, Pfammatter JA, Bordenave J, Plumley EI, Ravi S, Cowie M, Wallace EP, Hermann BP, Maganti RK, Jones MV. J Neurosci. 2021;41(46):9669-9686. doi:10.1523/JNEUROSCI.2439-20.2021 In temporal lobe epilepsy, the ability of the dentate gyrus to limit excitatory cortical input to the hippocampus breaks down, leading to seizures. The dentate gyrus is also thought to help discriminate between similar memories by performing pattern separation, but whether epilepsy leads to a breakdown in this neural computation, and thus to mnemonic discrimination impairments, remains unknown. Here we show that temporal lobe epilepsy is characterized by behavioral deficits in mnemonic discrimination tasks, in both humans (females and males) and mice (C57Bl6 males, systemic low-dose kainate model). Using a recently developed assay in brain slices of the same epileptic mice, we reveal a decreased ability of the dentate gyrus to perform certain forms of pattern separation. This is because of a subset of granule cells with abnormal bursting that can develop independently of early EEG abnormalities. Overall, our results linking physiology, computation, and cognition in the same mice advance our understanding of episodic memory mechanisms and their dysfunction in epilepsy.
Commentary
The dentate gyrus (DG) has long lived in 2 parallel worlds. In one, the DG is considered the brain region that performs pattern separation, the cognitive process by which inputs with some overlapping features are made more dissimilar. In another, the DG acts as gate protecting against excessive excitatory input to the hippocampus. In temporal lobe epilepsy (TLE), the DG gate is proposed to be dysfunctional, resulting in excessive excitation entering the hippocampus and enabling the generation of seizures. Using a multifaceted experimental approach to test pattern separation at the behavioral level in human TLE patients and at the cellular level in a kainate (KA) mouse model of TLE, Madar et al bridge the conceptual gap between the DG as a pattern separator and the DG as an excitatory gate. 1
To address whether TLE patients have memory deficits, as would be expected if the DG is dysfunctional, Madar et al presented a mnemonic similarity task 2 to TLE patients and nonepileptic controls. In this task, subjects are first presented a series of images (e.g., a piano, a pineapple, and a tractor). Next, subjects are presented a second series that contains new images, but also some images identical to those in the first series and some that are similar but not identical to those in first series (e.g., a piano with an opened, not closed, top). In the second series, subjects are asked to classify images as Old, Similar, or New, and their pattern separation score is determined by how well they correctly identify similar images (i.e., those that have only some overlapping features to earlier images). Temporal lobe epilepsy patients scored only about 50% as well as nonepileptic controls, supporting a recent study showing TLE impairment on this task. 3 But what happens in the brain to cause this cognitive symptom?
Using the KA mouse model, the authors searched for clues about the cellular dysfunction that might underlie impaired mnemonic discrimination. Kainate mice were subjected to an object location memory task wherein mice usually spend more time exploring an object that was moved relative to a stationary one. Kainate mice explored both objects equally, indicating they did not remember that one of the objects had moved. This impairment in object location memory is often observed following insult to the hippocampus. Here, the authors show convincingly that the KA mouse model and the human TLE patients have similar deficits on analogous mnemonic discrimination tasks, thereby providing an important validation of the KA mouse model in evaluating disease-relevant memory problems.
The authors then propose that these memory deficits are, in fact, specifically due to DG pathology. To test this hypothesis, hippocampal slices were prepared from the same mice after behavioral testing, and a cellular-level pattern separation assay was performed. This clever assay, developed by the same group, 4 involves stimulating the input to the DG (lateral perforant path) with correlated patterns, recording the response in a DG granule cell (GC), and comparing the similarity across output patterns. If a cell produces output spiketrains that are more correlated than the corresponding inputs, then that cell is biased toward pattern convergence. If outputs become less correlated than their inputs, then it’s biased toward pattern separation. Pattern separation deficits among GCs from KA-treated mice were apparent. The similarity between GC output spiketrains of KA mice was high, and this effect was most prominent when input spiketrains were very similar. This finding is analogous to the earlier finding in TLE patients insofar their impairments are worst when old and new images are similar, and thus harder to differentiate, in the mnemonic similarity task.
Mnemonic discrimination deficits and impairments in neural pattern separation might arise from abnormal GC firing properties. Indeed, a fraction of GCs from KA mice fired more reliably to inputs and fired more bursts of spikes, both of which promoted an overall higher firing rate among GCs. Additionally, clear inverse relationships between firing rate and bursting with pattern separation at the single GC level were apparent. The authors speculate that increased GC excitability is due to decreased inhibitory drive onto GCs because they found no evidence of aberrant synaptic excitation, and because their previous study showed that blocking synaptic inhibition is sufficient to enhance GC excitability. 5 However, without more direct investigation, the link between compromised inhibition and poor pattern separation remains tenuous.
In this study, the authors describe a simple yet elegant experimental design wherein the mice used for behavioral testing were also used for hippocampal slice recordings. By using the same mice in all stages of the study, potential relationships between electrographic seizures, mnemonic discrimination abilities, and cellular-level physiology could be explored. Notably, although the authors track the same animals throughout experimentation, only 2 animals had bona fide seizures: most animals only had interictal spikes. If animals had presented with a broader range of seizure phenotypes, then it would have been illuminating to evaluate mnemonic deficits across that range. Perhaps, if animals presented with more severe seizures, then greater mnemonic deficits would have been observed. Even if seizure severity did not correlate with significant deficits, such an observation would argue that seizure generation and pattern separation may not rely on common mechanisms. Nonetheless, a temptingly simple model wherein cellular-level pathology causes seizures and mnemonic impairments alike must be rejected. There was no obvious relationship between seizure severity and GC physiological pathology. This finding indicates that the relationship between seizures, pathological GC activity, and mnemonic discrimination is complex and requires further study.
One important caveat of the spiketrain correlation assay used in hippocampal slices is that it only considers pattern separation using firing rate or “rate code.” The first model—and many subsequent—to describe how the DG might perform pattern separation emphasized the capacity of the DG to alter the “population code” (i.e., the specific ensemble of coactive cells), rather than the rate codes of individual cells. 6,7 In short, the DG was proposed as an area with so many cells that it could expand the neural space in which a representation is stored and provide a sparser representation with less overlap. 8 In the end, the ways wherein population and rate codes are utilized and combined are complex, and future studies may in fact support the hypothesis that TLE-associated deficits in the DG population code underlie mnemonic impairments. Nonetheless, Madar et al uncover a striking correlation between pattern separation at the cellular and behavioral levels (Figure 9C of their study), thereby providing some assurance that the firing patterns of single GCs can predict behavioral performance.
All considered, this study makes several important contributions to our understanding of TLE and DG function. First, the study further validates the KA mouse model of TLE by demonstrating analogous mnemonic discrimination deficits across KA mice and human TLE patients. Second, the study bridges the conceptual gap between the DG as an activity gate and the DG as a pattern separator. Finally, the authors show that pattern separation by GCs from KA mice is degraded, and that this pattern separation deficit is likely due to pathological physiology. However, the larger question remains: do pattern separation and seizure generation rely on common mechanisms? By evaluating mnemonic impairments across a broader range of seizure severity, future studies may more thoroughly address this question.
