Causal relationship of CA3 back-projection to the dentate gyrus and its role in CA1 fast ripple generation

Abstract

Background: Pathophysiological evidence from temporal lobe epilepsy models highlights the hippocampus as the most affected structure due to its high degree of neuroplasticity and control of the dynamics of limbic structures, which are necessary to encode information, conferring to it an intrinsic epileptogenicity. A loss in this control results in observable oscillatory perturbations called fast ripples, in epileptic rats those events are found in CA1, CA3, and the dentate gyrus (DG), which are the principal regions of the trisynaptic circuit of the hippocampus. The present work used Granger causality to address which relationships among these three regions of the trisynaptic circuit are needed to cause fast ripples in CA1 in an in vivo model. For these purposes, male Wistar rats (210-300 g) were injected with a single dose of pilocarpine hydrochloride (2.4 mg/2 µl) into the right lateral ventricle and video-monitored 24 h/day to detect spontaneous and recurrent seizures. Once detected, rats were implanted with microelectrodes in these regions (fixed-recording tungsten wire electrodes, 60-μm outer diameter) ipsilateral to the pilocarpine injection. A total of 336 fast ripples were recorded and probabilistically characterized, from those fast ripples we made a subset of all the fast ripple events associated with sharp-waves in CA1 region (n = 40) to analyze them with Granger Causality.

Results: Our results support existing evidence in vitro in which fast ripple events in CA1 are initiated by CA3 multiunit activity and describe a general synchronization in the theta band across the three regions analyzed DG, CA3, and CA1, just before the fast ripple event in CA1 have begun.

Conclusion: This in vivo study highlights the causal participation of the CA3 back-projection to the DG, a connection commonly overlooked in the trisynaptic circuit, as a facilitator of a closed-loop among these regions that prolongs the excitatory activity of CA3. We speculate that the loss of inhibitory drive of DG and the mechanisms of ripple-related memory consolidation in which also the CA3 back-projection to DG has a fundamental role might be underlying processes of the fast ripples generation in CA1.

Keywords: Fast ripples; Granger causality; Hippocampus; In vivo studies; Pilocarpine model; Theta rhythm.

Fig. 1

Representative fast ripple event recorded in CA1 with simultaneous raw intracranial EEG recordings from CA3 and DG. The last row shows a continuous morlet wavelet transform of the signal in CA1, in the 100–600 Hz band

Fig. 2

Bayesian inference of the fast ripple parameters. The surfaces in the plots describe joint posterior probability distribution of (μ) means and (σ) standard deviations in each trisynaptic circuit region of a peak-to-peak amplitude, b power frequency, c duration and d power. The inner surface of each credibility region represents the maximum a posteriori Probability (MAP) estimate value, n = 336 fast ripple events observed in 6 animals

Fig. 3

Statistical comparison of the fast ripple parameters. The box plots describe the distribution of the parameters in our data per trisynaptic circuit region of a peak-to-peak amplitude, b power frequency, c duration and d power. Kruskal–Wallis test with a Wilcoxon test for multiple comparisons was used (*p < .05, **p < .01, ***p < .001, ****p < .0001), n = 336 fast ripple events observed in 6 animals

Fig. 4

Spectral composition of fast ripple events as a function of the posterior distributions of means of power frequency and mean frequency per region

Fig. 5

Multiunit activity at a fast ripple event. a Graphs show the multiunit spike probability distributions in each trisynaptic circuit region, centered at the maximum peak-to-peak amplitude of the fast ripple event. b Graph shows the cumulative probability distributions of (a). CA3 region accumulates a higher spike probability at the beginning of the event, and the CA1 region has a close to normal distribution, as expected

Fig. 6

Time–frequency relationship among trisynaptic circuit regions at a fast ripple event. a Raw intracranial EEG recordings in each trisynaptic circuit region. Arrow is pointing to the maximum peak-to-peak amplitude of the fast ripple event recorded in CA1. b Wavelet coherence of the theta band (3–7 Hz) and gamma band (30–90 Hz) between regions. Note that the fast ripple is centered when theta coherence among all regions is increased. c Statistical comparison of the mean coherence amplitude among the trisynaptic regions 200 ms before and after the maximum fast ripple peak-to-peak amplitude in the theta band (3–7 Hz) and gamma band (30–90 Hz). Calibration bar: 500 μV. The paired Wilcoxon test per connection was used (*p < .05, **p < .01, ***p < .001, ****p < .0001)

Fig. 7

Overall pre- and post-fast ripple causal networks. a Schematic diagram of the resulting pre- and post-fast ripple causal network. Granger causality test was used. Only overall statistically significant connections (Table 1) are drawn (*p < .05, **p < .01, ***p < .001, ****p < .0001). In the precircuit, note the relevance of CA3 back-projection in the generation of fast ripples, and the "conventional" trisynaptic circuit is re-established in the postcircuit. The image was elaborated according the results of data analysis and is our own authorship. b Causal unit density and c causal flow among the trisynaptic circuit regions. Note the changes in b causal density in CA3 and the dentate gyrus (DG) before and after the fast ripple event. CA1 maintains its causal dynamics unchanged, and CA3 is always a causal hub. Also in c, the undetermined causal flow preceding the fast ripple may be interpreted as an increase in entropy in the circuit due to reverberating feedback between regions. The “conventional” causal flow of the circuit after the fast ripple event is re-established with the DG as the causal source and CA3 as the causal sink