Multimodal determinants of phase-locked dynamics across deep-superficial hippocampal sublayers during theta oscillations

Abstract

Theta oscillations play a major role in temporarily defining the hippocampal rate code by translating behavioral sequences into neuronal representations. However, mechanisms constraining phase timing and cell-type-specific phase preference are unknown. Here, we employ computational models tuned with evolutionary algorithms to evaluate phase preference of individual CA1 pyramidal cells recorded in mice and rats not engaged in any particular memory task. We applied unbiased and hypothesis-free approaches to identify effects of intrinsic and synaptic factors, as well as cell morphology, in determining phase preference. We found that perisomatic inhibition delivered by complementary populations of basket cells interacts with input pathways to shape phase-locked specificity of deep and superficial pyramidal cells. Somatodendritic integration of fluctuating glutamatergic inputs defined cycle-by-cycle by unsupervised methods demonstrated that firing selection is tuneable across sublayers. Our data identify different mechanisms of phase-locking selectivity that are instrumental for flexible dynamical representations of theta sequences.

Fig. 1. Bimodal distribution of theta phase-locked firing across CA1 sublayers.

a Representative example of a CA1 pyramidal cell recorded juxtacellularly from a head-fixed running mouse. See average action potential waveform and autocorrelation at right, and morphological identification as deep cell at bottom. Scale bar 50 µm. b Theta phase-locked firing histogram of the cell shown in a. Raw local field potential (LFP) traces and juxtacellular signals are shown below: SO, stratum oriens; SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum moleculare. c Theta phase firing histogram from single cells recorded in awake head-fixed mice, ranked according to their preferred phase (n = 12 cells). Individual (gray) and mean (black) theta filtered LFP signals at top are z-scored (Z). The population histogram at bottom represents the distribution of mean preferred phases from individual cells. d Same for single pyramidal cells recorded juxtacellularly in freely moving rats (n = 28 cells). e Individual and mean ± SD values of vector length per cells reported before in head-fixed and freely moving conditions. No differences between groups. f Awake/REM sleep (left) and RUN/no-RUN (running vs. other motor behavior) dependency of theta phase preference of single pyramidal cells. See legend in (g). g Multivariate analysis of morphologically identified cells demonstrate effects per location (Harrison–Kanji test for deep-superficial location: χ²(2) = 10.7, p = 0.0136) and preparation (χ²(2) = 8.5, p = 0.0046); no interaction (χ²(1) = 2.1, p = 0.148). Phase preferred data from each morphologically identified cell is depicted separately for deep and superficial cells in both preparations. h Optogenetic tagging of single units using micro-LED optoelectrodes in two transgenic lines allowed identifying superficial (Calb1-cre + AAV5-DIO-ChR2-YFP) and deep CA1 pyramidal cells (Thy1-ChR2). Raster at right shows response of one representative unit to 79 trials of nanowatt blue light stimulation in a Thy1-ChR2 mouse. Bottom, ChR2 signal (white) and Calbindin immunostaining (magenta) are shown in false colors (one confocal plane from each line). Scale bar 25 µm. i Phase-locked firing histograms from Calbindin+ superficial (n = 7 units from 3 mice) and Thy1+ deep opto-tagged units (n = 14 units from 2 mice; statistically different Watson–Williams test, F(1,20) = 17.7, p = 0.00048. Note that bimodal distribution of preferred theta phases can be explained by different cell-type contribution.

Fig. 2. Genetically constrained models of CA1 pyramidal cell activity.

a Activity of morphologically identified CA1 pyramidal cells was modeled using the multi-compartment Hodgkin–Huxley formalism. b Genetic algorithms restricted the intrinsic parameter space (GA factors) by fitting experimental somatic and dendritic responses to current pulses (green). Different combination of GA factors (individuals) giving experimentally valid input-output responses (Error = 0 in target; n = 45 individuals) are shown in Supplementary Fig. 3C. c Same as in (b) for fitting synaptic responses to Schaffer collaterals stimulation (n = 60 individuals). See valid GA factors and examples in Supplementary Fig. 3D. Note synaptic responses similar to those recorded in deep and superficial CA1 cells in vivo. d Theta-modulated activity simulated in the synthetic cell shown in (a) in response to a biologically realistic collection of glutamatergic (CA3, CA2, ECIII, and ECII) and GABAergic (Axo, Bis, CCK, Ivy, NGF, OLM, PV, and SCA) theta-modulated inputs (Supplementary Table 3). e Theta phase-locking dynamics of same set of intrinsic features (individual #5) in four different morphologies. Upper traces show membrane potential at the soma (black), proximal (gray), and distal dendrites (light gray). Data from 1000 theta cycles. f Effect of individual variability of intrinsic properties. Data were ranked according to results from a permutation test that maximized correlations between individuals and phase preference. Mean data from each intrinsic individual in the four morphologies is represented as box-whisker plots. The number of dots per individual varies according to inclusion criteria (individuals with realistic firing rate values in at least three out of four morphologies, n = 3 for individuals 9–11, n = 4 for the rest). The matrix at right shows mean GA factors across individuals. Note Pearson’s correlation coefficient R and significance at top: gC (p = 0.0003), gKDR (p = 0.0002), gL (p = 0.0006), gNa (p = 0.002), gA (p = 0.002), gHCN (p = 0.003), gCa (p = 0.02), gM (p = 0.05), gAHP (p = 0.02), and Ra (p = 0.03). g Same as in (f) for morphologies. The number of dots per morphology varies according to inclusion criteria: n127 (n = 10 synthetic cells), n409 (n = 11), sup1 (n = 11), and n128 (n = 12). Note significant effect of perisomatic inhibition (p = 0.041) and the total dendritic branches (p = 0.013). ***p < 0.001; **p < 0.01; *p < 0.05.

Fig. 3. Interneuron-specific control of theta phase preference.

a Effect of different distribution of perisomatic GABAergic inputs from PV and CCK basket cells. Note reliable phase-locking behavior determined by inputs typical of deep (30% CCK, 100% PV; blue) and superficial (100% CCK, 30% PV; magenta) pyramidal cells, consistent with experimental data. Data from 1000 theta cycles (individual #12 in four morphologies). Bottom histograms show effect of fully removing inhibition. Traces at the right-hand side show mean membrane potential dynamics at different somatodendritic compartments of deep- (blue) and superficial-like (magenta) synthetic cells shown at left. b, Effect of different PV/CCK inputs on 12 individuals from the four different morphologies (n = 48 synthetic cells), quantified by the TP index. Individual and mean +/− SD values are shown. Data tested with paired t-test for deep- versus superficial-like groups within each morphology. Full dataset tested with one-way ANOVA, F(2,108) = 41.9 and post-hoc t-tests. ***p < 0.0001 (from left to right: p = 6 × 10⁻⁵, p = 10⁻⁴, p = 8 × 10⁻⁷, p = 5 × 10⁻⁵, p = 4 × 10⁻⁵, p = 3 × 10⁻⁶, p = 4 × 10⁻⁷). c Chemogenetic approaches allowed to test predictions on the effect of silencing PV+ interneurons in PV-Cre mice injected with hM4D(Gi). Immunohistological validation of interneuronal cell-types affected by hM4D(Gi): Bistratified cells (Bis, n = 2), oriens-lacunosum moleculare cells (OLM, n = 4), PV basket cells or axo-axonic cells (PV, n = 23), Ivy cells (n = 12), All PV+ (n = 42), and Off target (n = 4). Data from 3 sections from a representative mouse. Scale bar 50 µm. d Model prediction of the population level effect of blocking GABAergic inputs from PV basket cells, bistratified cells, OLM interneuron, and a minority of Ivy cells (n = 32 synthetic cells; 1000 cycles). e Examples of chemogenetic effects and histogram quantification of the full dataset (n = 5 juxtacellular; n = 23 single units from multisite recordings). Note similar distribution as compared with simulations (circular Watson–Williams multi-sample test, F(1,53) = 0.02, p = 0.88). All cells/units significantly theta modulated as tested by Rayleigh (p < 0.05). f, Change of theta phase preference in a subset of n = 12 putative pyramidal cells recorded before and after CNO in PV-hDM4(Gi) mice. Significant differences, circular two paired sample test, t(11) = 5.7846, p = 0.0048.

Fig. 4. Effect of glutamatergic input pathways.

a Effect of CA3 and ECIII input pathways on phase preference of synthetic cells across the PV/CCK axis. b Mean phase-locked firing histograms and membrane potential dynamics at the proximal and apical dendritic trunk from all synthetic cells examined under low and high levels of activity of the CA3 (yellow) and ECIII inputs (orange). Data from 48 synthetic cells in each condition (1000 theta cycles each). c Relationship between the firing phase and the phase of the maximal dendritic depolarization measured at the proximal apical trunk. Each dot represents data from one spike from 48 synthetic cells in the high CA3 and high ECIII conditions. Arrowheads indicate uncorrelated phase-locking dynamics between firing and dendritic potentials. Bimodality was stronger during higher CA3 inputs. Data from 1000 theta cycles in each condition. d Results of a multinomial logistic regression model suggest phase-specific contribution of different individual factors. Black, upregulation. Gray, downregulation. Data from 731 heterogeneous synthetic cells and >300,000 theta cycles. See Supplementary Fig. 7.

Fig. 5. Dynamic phase shifts of deep and superficial CA1 pyramidal cells.

a Unsupervised classification of theta cycles in different classes allowed evaluating changes of firing dynamics from single-cells recorded juxtacelullarly and using multisite probes. Cycles classified as tSC1 and 2 are associated to larger CA3 inputs (yellow) while tSC3 and 4 reflect larger entorhinal layer III inputs (orange). b Results from the surrogate tests (1000 shuffles) on significant theta modulation of single-cell data during theta cycles subclasses. Only cells/units meeting significance (p < 0.05 Rayleigh test) for both tSC1-2 and tSC3-4 firing histograms were considered for further analysis (n = 7 juxtacellular cells, n = 96 units from multisite recording). c Rank order tests were used to identify sequences of units repeating consistently across classified theta cycles as tested against chance order (500 shuffles). d Example of the effect of high ECIII versus high CA3 inputs in deep and superficial units firing together over several theta cycles. Upper plots show mean rank order ± SD for each unit for all high CA3 (from up to bottom: 19, 13, 17, 17, 18, and 10 cycles) versus high ECIII cycles (from up to bottom: 13, 15, 13, 12, 14, and 13 cycles). Bottom plots show same data plotted as theta phase firing histograms. The order was established by high CA3 cycles. Note significant phase shift of putative superficial units #21 and 34 at bottom and the deep unit #40 at top. e Phase shift differences (high CA3 phase – high ECIII phase) caused by tSC3-4 cycles (>ECIII) vs. tSC1-2 cycles (>CA3) in deep and superficial cells. Data from juxtacellular (4 deep and 3 superficial cells; open dots) and multisite recordings (70 deep and 26 superficial units; filled dots). Horizontal lines indicate mean ± SD from each sublayer (deep: black; superficial: gray). Circular ANOVA, F(1,101) = 14.3, p = 0.0003 for all cells together; F(1,5) = 4.8, p = 0.0795 for juxtacellular cells alone; ***p < 0.001. f Mean population polar plots showing phase shift effects of high ECIII inputs in deep and superficial cells separately. Same data set as in (e). The arrows reflect the main trends reported in (e).