Lateral entorhinal cortex inputs modulate hippocampal dendritic excitability by recruiting a local disinhibitory microcircuit

The lateral entorhinal cortex (LEC) provides information about multi-sensory environmental cues to the hippocampus through direct inputs to the distal dendrites of CA1 pyramidal neurons. A growing body of work suggests that LEC neurons perform important functions for episodic memory processing, coding for contextually-salient elements of an environment or the experience within it. However, we know little about the functional circuit interactions between LEC and the hippocampus. In this study, we combine functional circuit mapping and computational modeling to examine how long-range glutamatergic LEC projections modulate compartment-specific excitation-inhibition dynamics in hippocampal area CA1. We demonstrate that glutamatergic LEC inputs can drive local dendritic spikes in CA1 pyramidal neurons, aided by the recruitment of a disinhibitory vasoactive intestinal peptide (VIP)-expressing inhibitory neuron microcircuit. Our circuit mapping further reveals that, in parallel, LEC also recruits cholecystokinin (CCK)-expressing inhibitory neurons, which our model predicts act as a strong suppressor of dendritic spikes. These results provide new insight into a cortically-driven GABAergic microcircuit mechanism that gates non-linear dendritic computations, which may support compartment-specific coding of multi-sensory contextual features within the hippocampus. HIGHLIGHTS Slice electrophysiology experiments investigate how lateral entorhinal cortex influences hippocampal area CA1 LEC drives local spikes in distal dendrites but not in somata of CA1 pyramidal neurons LEC inputs recruit VIP IN and CCK IN populations in CA1, but not SST INs Computational modeling and circuit manipulation experiments identify a VIP IN-mediated disinhibitory microcircuit for gating local dendritic spike generation IN BRIEF Bilash et al. found that a distal cortical input is capable of driving local dendritic spikes in hippocampal pyramidal neurons. This dendritic spike generation is promoted by cortical recruitment of a local VIP interneuron-mediated disinhibitory microcircuit. Their results highlight new circuit mechanisms by which dynamic interaction of excitation, inhibition, and disinhibition support supralinear single-cell computations.


INTRODUCTION
amplitudes of LEC-driven dendritic responses saturated after 6% maximum stimulation strength, 185 with some amplitudes ultimately exceeding 30 mV at higher stimulation strengths. Some  driven dendritic responses exhibited a shape reminiscent of suprathreshold dendritic spikes 187 previously described in the literature (Gasparini et al., 2004;Hausser et al., 2000;Remy et al., 188 2009; Remy and Spruston, 2007). 189 190 To assess how the LEC-driven responses propagate to the soma, we also performed whole-cell 191 recordings from CA1 PN somata (Figure 1I-J, Figure S2E-H), using the same photostimulation 192 paradigms as for dendrites. The somatic post-synaptic responses also exhibited a range of 193 amplitudes and kinetics (Peak amplitude: 1.5 to 13.7 mV (2.9 ± 0.3 mV), Time to peak: 9.8 to 40.5 194 ms, Half-width: 7.7 to 115 ms, n = 51 somata) (Figure 1K-L). The amplitudes of  somatic responses also saturated after 6% maximum stimulation strength ( Figure 1M).  2015). 217 218 To validate that these were indeed LEC-driven dendritic spikes, rather than large-amplitude 219 dendritic PSPs, we analyzed the derivative traces (Gasparini et  the derivative traces and phase plots revealed obvious differences between LEC-driven dendritic 222 PSPs and dendritic spikes ( Figure 2B-C, Figure S3). The derivative traces of LEC-driven 223 dendritic spikes exhibited a recognizable, high amplitude peak that reached a maximum dV/dt 224 value before dipping back below zero. In contrast, the derivative traces of LEC-driven dendritic 225 PSPs exhibited a smooth, flattened curve ( Figure 2B, Figure S3B). Based on these characteristics, 226 we set a dendritic spike classification criterion for our slice electrophysiology dataset: light-evoked 227 dendritic responses with maximum dV/dt > 7.5 mV/ms were categorized as suprathreshold 228 dendritic spikes (dSpikes), while those with maximum dV/dt < 7.5 mV/ms were categorized as 229 subthreshold dendritic post-synaptic potentials (dPSPs). Compared to other measurements, the 230 maximum dV/dt value provided the clearest, quantifiable delineation of subthreshold versus 231 suprathreshold LEC-driven dendritic responses ( Figure 2D, Figure S4). Interestingly, among the 232 recorded dendritic responses, the incidence of LEC-driven dendritic spikes did not appear to be 233 strictly distance dependent within the distal dendritic compartment (Figure 2A, Figure S4B). 234 Eliciting LEC-driven dendritic spikes was possible at lower photostimulation strengths (4-10% 235 maximum stimulation strength), and was not limited to the maximum (100%) stimulation strength 236 protocol ( Figure 2E). LEC-driven PSPs recorded at the dendrite were significantly larger in 237 amplitude and had faster kinetics than those recorded in the soma of CA1 PNs (Peak amplitude p 238 < 0.0001, Time of peak p < 0.0001, Half-width p < 0.0001, Mann-Whitney test, n = 14 dPSPs and 239 52 somatic PSPs, Figure S5). This is expected, given the proximity of the dendritic recordings to 240 the LEC axon terminals and the biophysical properties of the distal dendrites. Taken together, our 241 results indicate that glutamatergic LEC inputs are capable of driving suprathreshold dendritic 242 spikes in CA1 PN distal dendrites. The complete lack of LEC-driven action potentials recorded in 243 over 50 somata (Figure 1) suggests that these local dendritic spikes do not propagate forward to 244 drive suprathreshold somatic activity. 245 246 We next explored the synaptic properties of LEC-CA1 inputs using photostimulation trains. 247 Presynaptic short-term plasticity and postsynaptic summation could provide insight into the 248 facilitatory versus depressing nature of LEC inputs and elucidate whether the repeated activity of 249 LEC inputs enables the propagation of dendritic spikes to drive somatic firing (Figures S6-7). 250 First, we measured the presynaptic short-term plasticity dynamics of the LEC inputs in CA1 PN 251 somata using trains of light pulses delivered across physiologically-relevant frequencies (1-10 Hz, 252 ≤ theta frequency) (Frank et al., 2001;Pilkiw et al., 2017), at low light intensities (2-3% of maximal 253 strength) ( Figure S6). Photostimulation at 4, 8, and 10 Hz was particularly effective at facilitating 254 the synaptic inputs from the ChR2 + LEC axon terminals in area CA1 (Figure S6D-E). 255 Simplistically, this paired pulse facilitation suggested that the glutamatergic LEC axons may have 256 a low release probability, although we cannot rule out the contribution of shifts in excitation-257 inhibition balance. Nevertheless, a low release probability in glutamatergic LEC axons may 258 explain the lack of LEC-driven action potentials in CA1 PN somata in response to a single light 259 pulse. So, we tested whether trains of photostimulation at maximum (100%) strength could evoke 260 summation-driven action potentials in CA1 PN somata. Surprisingly, we did not observe any LEC-261 driven action potentials in the CA1 PN somata, although the maximum photostimulation trains 262 elicited significant summation of the post-synaptic responses ( Figure S7A-F, J). Meanwhile,8 263 Hz photostimulation at maximum strength produced significant summation of LEC inputs in the 264 distal dendrites, and was capable of increasing dendritic spike probability (Figure S7G-J). 265 266 LEC inputs recruit strong feed-forward inhibition onto both compartments of CA1 PNs 267 268 What are the local circuit mechanisms underlying LEC-driven dendritic spikes in CA1 PN distal 269 dendrites? Given that we observe LEC-driven dendritic spikes when inhibition is intact, could 270 GABAergic inhibition be actively gating the incidence of distal dendritic spikes or their spread to 271 the soma? GABAergic inputs synapse onto the entire somato-dendritic axis of CA1 PNs (Glickfeld 272 et al. PSPs exhibited significantly higher amplitudes and significantly wider half-widths in both 293 neuronal compartments (dendrites: 27% increase in peak amplitude (p = 0.025), 95% increase in 294 half-width (p = 0.002); somata: 55% increase in peak amplitude (p = 0.0023), 105% increase in 295 half-width (p < 0.0001), paired t-test, n = 9 dendrites and 11 somata) (Figure 3C, H). Moreover, 296 the time of the peak of the PSPs showed significant delays in CA1 PN somata (56% increase in 297 time to peak, p < 0.0001) ( Figure 3H). There were no significant changes to the time of peak or 298 maximum dV/dt of PSPs recorded at the dendrite (time of peak p = 0.0612, maximum dV/dt p = 299 0.087) ( Figure 3C). While the amount of inhibition recruited by glutamatergic LEC inputs varied 300 across recorded pyramidal neurons, on average, the inferred IPSP was comparable in CA1 PN 301 distal dendrites (-3.33 ± 0.91 mV) and somata (-3.11 ± 0.66 mV). Thus, FFI (IPSPs) recruited 302 simultaneously by LEC excitatory inputs did significantly curb the coincident monosynaptic 303 excitation (EPSPs), but together resulted in a net overall depolarization LEC-driven PSPs ( Figure  304 3D, I). Furthermore, removing inhibition enabled LEC-driven action potentials in CA1 PN somata 305 in some cases (25% of somata, 11% of dendrites, Figure 3E, J), pointing to a complex interaction 306 between the excitation-inhibition balance within the circuit and attenuation of dendritic signals. 307 Together, these results demonstrate that inhibition plays a prominent role in the LEC-to-CA1 308 circuit. Glutamatergic LEC inputs drive strong feed-forward inhibition onto the CA1 PN, which 309 limits the degree and length of depolarization in the dendrites and somata, thereby sculpting 310 excitability and coupling dynamics between both neuronal compartments.  Figure 4F (top), G-I,). Notably, the majority of VIP INs driven to 350 spike by glutamatergic LEC inputs were located in the SLM layer ( Figure S8C).

352
TdTomato-positive CCK INs in CA1 were located around the SR/SLM border and in SP ( Figure  353 4A-B (middle), Figure S8A Figure 4F (middle), G-I). Interestingly, the CCK INs driven to spike by glutamatergic LEC inputs 359 were located exclusively at the SR/SLM border region in area CA1 ( Figure S8C). These belong 360 to the interneuron subpopulation that has been shown to mediate feed-forward inhibition onto CA1 361 PN dendrites, and that is targeted by long-range GABAergic projections from LEC (Basu et al.,362 2016). 363 364 In contrast to the other candidate INs, tdTomato-positive SST INs in CA1 were found almost 365 exclusively in stratum oriens (SO) and SP ( Figure 4A-B (bottom), Figure S8A) and were not 366 monosynaptically recruited or driven to spike by the photostimulation of LEC inputs (26.7% SST 367 INs responded, 0% of them driven to spike, Total n = 15 SST INs, Figure 4E-F (bottom), G-I, 368    trials, 1000 runs per trial, Figure 6D), revealing that LEC overall engages predominantly the 460 disinhibitory function of VIP INs in area CA1 dendrites. The peak amplitude and maximum dV/dt 461 values of dSpikes were similarly affected (peak amplitude: p < 0.0001, maximum dV/dt: p < 462 0.0001, Friedman test, n = 200 trials, Figure 6E, Figure S13). Meanwhile, CCK IN deletion had 463 the opposite effect on dSpike probability (p = 0.001, Figure 6D, Figure 6E, Figure S13) comparisons test, n = 20 trials, 1000 runs per trial, Figure 6F). 60 ms prior to the start of the 470 nm photostimulation to ensure maximum hyperpolarization of 512 Jaws + VIP INs during the LEC-driven light-evoked response in the CA1 PN ( Figure 7D, Figure  513 S14F-H). The 625 nm light pulse ended with a 100 ms-long down ramp, to prevent rebound firing 514 of the VIP INs (Chuong et al., 2014). Given the location of the glutamatergic LEC axons ( Figure  515 1) and LEC-driven VIP INs (Figure 4, S8), we photostimulated with both wavelengths over SLM, 516 as before (Figure 7B-C). 517 518 To determine the effect of VIP IN silencing on subthreshold versus suprathreshold dendritic 519 activity, we categorized the LEC-driven light-evoked responses into dPSPs and dSpikes, based on 520 their maximum dV/dt values (threshold dV/dt = 7.5 mV/ms, same as in Figure 2). Optogenetic 521 silencing of the general VIP IN population in CA1 led to a small, yet significant decrease in dPSP 522 amplitude, compared to the control conditions (p = 0.023, Paired t-test, n = 10 dendrites, Figure  523 7E-G, Figure S15A-D). There were no significant differences in dPSP kinetics ( Figure S15B).

525
Notably, VIP IN silencing had a significant effect on suprathreshold activity ( Figure 7H-J, S15E-526 H) in dendrites that exhibited LEC-driven dSpikes under control conditions. There was a 527 consistent decrease in dendritic spike probability ( Figure 7I), illustrated by the significant 528 reduction in the maximum dV/dt value of LEC-driven dendritic responses after VIP IN silencing 529 (p = 0.008, Wilcoxon matched-pairs signed-rank test, n = 3 dendrites, ~8 sweeps recorded per 530 dendrite, on average, Figure 7J, S15E-H). There was no significant difference in the dSpike 531 amplitude (p = 0.495, Wilcoxon matched-pairs signed-rank test, n = 3 dendrites, Figure S15G).

532
Finally, to test how VIP IN-mediated disinhibition impacts dendritic integration and summation 533 of LEC inputs, we silenced VIP INs while photostimulating LEC inputs with 8 Hz trains. 534 Optogenetic silencing of VIP INs led to a significant decrease in the summation of LEC-driven 535 inputs in CA1 PN distal dendrites, compared to control conditions (p < 0.0001, Two-way ANOVA, 536 n = 3, Figure 7K-L). Thus, local VIP INs in CA1 can dynamically boost the integration of 537 repetitive dendritic inputs in CA1 PNs through disinhibition. 538 539 VIP IN silencing had no effect on LEC-driven somatic PSPs ( Figure S16). Moreover, the level of 540 Jaws expression correlated with the effect of VIP IN silencing seen in the dendrites but not somata 541 ( Figure S17) In summary, our study demonstrated how a long-range cortical input activates a disinhibitory 553 microcircuit in the hippocampus to modulate the generation of local dendritic spikes. 554 Photostimulation of glutamatergic LEC axons in CA1 had asymmetric effects on the dendritic and 555 somatic compartments of CA1 PNs. Even with inhibition intact, photostimulation of LEC inputs 556 could drive local dendritic spikes in CA1 PNs without eliciting somatic action potentials. 557 Additionally, we demonstrated that LEC inputs recruit strong feed-forward inhibition onto the 558 dendrites and somata of CA1 PNs. We, therefore, hypothesized that LEC-driven circuit 559 mechanisms gate the generation of local dendritic spikes by influencing compartment-specific 560 excitation-inhibition balance. Targeted  Stimulation of LEC input alone is sufficient for dendritic spike generation 573 574 A surprising but robust finding of our study is that even in the presence of inhibition, 575 photostimulation of glutamatergic LEC inputs with a single, brief light pulse (at 100% maximum 576 strength) elicited dendritic spikes in up to 30% of dendritic recordings (5/17 in Figure 2, 3/13 in 577 Figure 7). A greatly-reduced photostimulation intensity (5% maximum) could still drive dendritic 578 spikes, which suggests that LEC-driven dendritic spikes do not require maximum or repetitive 579 activation of ChR2 + LEC axons within area CA1, perhaps because a VIP IN-mediated disinhibitory 580 microcircuit, recruited in parallel, is sufficient to overcome the existing dendritic inhibition. 581 Meanwhile, single light pulse photostimulation never elicited somatic action potentials in CA1 582 PNs, indicating that LEC-driven dendritic spikes are local dendritic events. 583 584 In previous in vitro studies, eliciting dendritic spikes in CA1 PNs typically required any 585 combination of the following conditions: . Thus, it was wholly unexpected that we could bypass these 596 usual requirements for dendritic spike generation in our study. We observed, both experimentally 597 and with modeling, that a single cortical input pathway can indeed drive dendritic spike generation 598 while inhibition remained intact in the neural circuit. This was an exciting finding that contrasts 599 the previously-accepted paradigms for dendritic spike generation in vitro, and highlights an 600 important role for single input pathways in driving dendritic nonlinearities in principal neurons by 601 recruiting disinhibitory microcircuits. 602 603 To speculate on their molecular composition, the majority of the LEC-driven dendritic spikes 604 resembled simple sodium spikes ( of a complex, multi-peaked dendritic spike involving additional conductances, possibly calcium 609 (d1, Figure 2). However, its small amplitude suggests a more distal dendritic origin and 610 attenuation of the spike as it traveled toward our recording site. Future research, beyond the scope 611 of this study, is necessary to determine the exact ionic nature of these LEC-driven local dendritic 612 spikes and to verify whether they can be detected by higher-throughput imaging approaches using 613 genetically-encoded calcium or voltage indicators that can be readily applied in vivo.

615
A new GABAergic circuit mechanism for modulating dendritic activity 616 617 Our study demonstrates that input-driven GABAergic microcircuits can also serve this function. In our 626 study, we recorded and modeled LEC-driven activity under more physiological conditions, with 627 inhibition intact. These conditions allowed us to capture the downstream effects of the VIP IN-628 mediated disinhibitory microcircuit, which can sufficiently counteract the dendritic feed-forward 629 inhibition to ultimately lead to dendritic spike generation. This local disinhibitory circuit motif is 630 poised to serve as a powerful gain modulation mechanism to permit dendritic nonlinearities in a 631 fast, dynamic, and pathway-specific manner. Local disinhibition may work in concert with long-632 range disinhibitory circuits such as the GABAergic projections from LEC ( Because we used the general VIP-Cre mouse line to target VIP INs, both our optogenetic mapping 652 (Figure 4; Figure S8 "SR/SLM VIP INs") and silencing experiments (Figure 7)  ours is the first to account for LEC mediated effects and readily reproduces the LEC-driven 695 dendritic spikes and LEC-driven interneuron activity observed in our experiments. Moreover, the 696 model allowed us to implement in silico compartment-specific removal of inhibition ( Figure S11) 697 and deletion experiments for interneuron (sub)populations and their combinations (Figures 6, S13 Our distal dendritic recordings captured LEC-driven dendritic spike activity in the neuronal 716 compartment found closest to the location of glutamatergic LEC synapses. The same 717 photostimulation paradigm that could lead to LEC-driven dendritic spikes failed to generate action 718 potentials in our somatic recordings (Figure 1). From this, we deduced that LEC-driven dendritic 719 spikes are local and do not actively drive somatic spikes. 720 721 What causes the LEC-driven dendritic spikes to remain local? This phenomenon is likely due to 722 attenuation along the dendritic arbor (Golding et al., 2005;Spruston et al., 1994), but the role of 723 compartment-specific inhibitory modulation cannot be ruled out ( Figure 3J) Under what conditions could excitatory LEC inputs drive somatic spike output? High frequency, 754 burst-like activity and/or integration of coincident inputs could elicit somatic spikes by sufficiently 755 depolarizing the somatic compartment. We expanded our computational model, which 756 recapitulated our experimentally-recorded local LEC-driven dendritic spikes (Figure S10), and 757 demonstrated that simulating high frequency, in vivo-like patterns of activity from excitatory LEC 758 inputs, together with somatic depolarization, leads to action potential firing in the CA1 PN model 759 neuron ( Figure S18). Notably, at higher rates of LEC input firing (e.g., 100 Hz, burst-like firing 760 at theta peak), the somatic spike output is differentially modulated by There is sufficient evidence to speculate that LEC may provide a powerful signal to shape context-792 dependent coding in the hippocampus. LEC neurons code for multisensory contextual information   SLM during whole-cell recordings from a CA1 PN dendrite. SR95531 and CGP55845 were 902 perfused onto the brain slice to block inhibition (green X  perfusion of the brain slice with Na + and K + channel blockers, tetrodotoxin (TTX) and 4-1107 aminopyridine (4-AP), respectively, used to isolate monosynaptic connections between 1108 glutamatergic LEC inputs and downstream neurons.

1411
• Key Resources Table  1412 • Resource Availability 1413 • Key Resources Table  1427 1428 Reagent   and adjusted all compartments' lengths ( Table 2). The diameters of all compartments were 1743 updated to follow the d3/2 rule. The model consists of five apical dendritic compartments 1744 simulating the apical trunk (radProx, radMed, radDist_i, ∈ {1,2,3}) and six dendritic 1745 compartments simulating the basal tree, each containing a Ca 2+ pump and buffering mechanism, 1746 Ca 2+ activated slow AHP and medium AHP potassium (K + ) currents, an HVA L-type Ca 2+ current, 1747 an HVA R-type Ca 2+ current, an LVA T-type Ca 2+ current, an h current, a fast Na + and a delayed 1748 rectifier K + current, a slowly inactivating K + M-type current and a fast inactivating K + A-type 1749 current (Poirazi et al., 2003a, b). The proximal and distal apical dendrites consisted of eight 1750 compartments, respectively, each containing a fast Na + , a delayed rectifier K + and a fast-1751 inactivating K + A-type current. The PN model cell was topologically oriented: its soma was located 1752 in the simulated SP layer, its basal dendrites in the SO layer, and its proximal and distal apical 1753 dendrites in the SR and SLM layers, respectively. The type and distribution of ionic mechanisms 1754 in the PN model are described below, while each channel's maximum conductance is given in 1755 Tables 3 and 4. 1756 1757 Table 2. Morphological properties of the pyramidal model cell. Prefix 'radT' in compartments 1758 correspond to the apical trunk, 'rad' to oblique dendrites, 'lm' to apical tuft, and prefix 'ori' to 1759 basal dendrites, respectively. d and L denote the diameter and length of the compartment.  The non-basket cell SR/ SLM CCK IN had 17 compartments, containing a leak conductance, a  1775 sodium current, a fast-delayed rectifier K + current, an A-type K + current, L-and N-type Ca 2+ 1776 currents, a calcium-dependent K + current, and calcium-and voltage-dependent K + current (Table  1777 5, Figure S12). The SR/SLM CCK IN received excitatory connections from LEC to their distal 1778 SLM dendrites and from the CR + VIP IN to their soma. To reproduce the LEC-induced activity on 1779 SR/SLM CCK INs (see Figure 4G-I, S12B-C), we drew the number of excitatory synapses from 1780 an exponential distribution with mean equals 10. At approximately 17% of the trials (n = 10 x 200 1781 trials), the SR/SLM CCK IN receives no LEC connections ( Figure 4H) The OLM IN had four compartments, which included a Na + current, a delayed rectifier K + current, 1792 an A-type K + current, and an h-current (Table 5, Figure S12) The CCK + VIP IN had 17 compartments, containing a leak conductance, a sodium current, a fast-1805 delayed rectifier K + current, an A-type K + current, L-and N-type Ca 2+ currents, a Ca 2+ -dependent 1806 K + current, and a Ca 2+ -and voltage-dependent K + current (Table 5, Figure S12). To replicate the 1807 irregular firing which was experimentally observed, we tuned the calcium influx of this model, 1808 making the decay of Ca 2+ current slower ( Figure S12A). The CR + VIP IN consisted of 17 compartments, including mechanisms for slow K + current, fast 1816 Ca 2+ -activated K + current, and N-type Ca 2+ current (Table 5, Figure S12). Each CR + VIP IN 1817 received excitatory input from LEC. The number of excitatory connections from LEC was drawn 1818 from a Poisson distribution with a mean equal to 5, while in 9% of the trials (n = 10 x 200 trials), 1819 the CR + VIP IN did not receive any connection from LEC ( Figure 5D-F, S12B-C). 1820 1821 Validation 1822 Passive (intrinsic) and active (spiking) properties of each neuronal type were validated against 1823 experimental data. For the validation, we used the frequency vs. injected current (f-I) curve and 1824 the sag-ratio as a function of injected currents (hyperpolarization) (Figure S2). 1825 1826 1827 1828