Layer 1 interneurons divide labour by timescale and target, not by oscillation

Question: How do distinct subtypes of layer 1 GABAergic interneurons, identified by molecular markers such as NDNF, VIP, or somatostatin, differentially modulate pyramidal neuron activity during hippocampal sharp-wave ripples and cortical spindles?

Answer

Across the retrieved papers, the only L1 subtype with direct evidence of state-locked action at sharp-wave ripple (SWR) onset is the NDNF/neurogliaform class, and its effect is conditional on spindle coupling rather than on the ripple itself: L1 NDNF cells inhibit the local network at SWR onset "depending on their coupling with spindles," and they do so through slow, primarily GABA_B-mediated inhibition [16]. The SST and VIP classes are not parallel independent channels onto pyramidal cells in this literature. SST cells are the dendritic, delayed, feature-matched inhibitory arm whose axons L1 NDNF cells presynaptically suppress [1], and VIP cells are defined almost entirely by their capacity to disinhibit SST cells, a property demonstrated in slice and modelled in network terms but not measured during SWRs or spindles in any retrieved study [19] [8]. Taken together, the papers support a serial architecture in which NDNF cells time a slow dendritic veto to the spindle-coupled ripple window, SST cells determine which inputs that veto is applied to, and VIP cells set the gain of the SST arm — but the hippocampal half of this claim rests on extrapolation from cortical recordings, since CA1 ripple work has so far resolved perisomatic PV, axo-axonic and SOM populations rather than an L1 NDNF population [13] [14].

NDNF cells inhibit the network at SWR onset only when the ripple coincides with a spindle

L1 NDNF neurons resolve into three populations that assemble into networks tuned to specific sleep stages, and their relation to oscillations differs between subpopulations [16]. NREM- and REM-active L1 NDNF cells inhibit the network upon SWR onset as a function of their coupling to spindles, while L2/3 cells are mostly inactive during SWRs [16]. REM-active NDNF cells and L2/3 neurons show opposite activation patterns during spindles, so the same oscillation can drive the inhibitory and excitatory compartments in antiphase [16]. This positions NDNF cells as coincidence detectors of the two oscillations rather than responders to either alone, consistent with the supralinear, coincidence-detecting dendritic integration described for hippocampal neurogliaform interneurons [20].

The NDNF→pyramidal signal is slow and diffuse, making it a gain control rather than a timing device

L1 NDNF neurons mediate slow inhibition primarily via GABA_B receptors [16]. Systemic GABA_B antagonism reduced the coupling of pyramidal cells but did not change their inhibitory responses during SWRs, which dissociates the synchronising role of GABA_B signalling from the fast inhibitory event itself [16]. NDNF cells inhibit top-down synapses more effectively than SST cells, and thalamic POm input drives L1 neurogliaform cells, a NDNF subtype that elicits unitary GABA_B responses [2]. Volume transmission of GABA from strongly activated L1 NDNF cells onto L2/3 excitatory neurons and PV interneurons has been proposed on the basis of subtype-resolved calcium imaging during cortical stimulation [10]. Consistently, NDNF cells restrain hyperexcitability with a largely GABA_B-dependent effect, and are recruited more slowly than PV cells during epileptiform events [3]. Slow kinetics plus volume diffusion makes this a gain-setting rather than a phase-setting signal.

NDNF cells veto the SST axon, so NDNF and SST inhibition of pyramidal dendrites are not independent

In L1 of visual cortex, NDNF interneurons presynaptically inhibit the axonal outputs of SST interneurons, forming direct contacts onto SST axons and suppressing their output, thereby modulating inhibitory responses in L2/3 pyramidal neurons [1]. This motif is preferentially engaged during locomotion and low-contrast conditions, i.e. it is context-gated [1]. Because NDNF cells are more effective than SST cells at suppressing top-down corticocortical inputs, and this suppression is presynaptic and GABA_B-dependent, the two subtypes converge on the same distal synapses through a serial rather than a parallel arrangement [2]. Any account that treats L1 NDNF and L1 SST as additive inhibitory channels onto pyramidal dendrites is therefore inconsistent with the connectivity reported in these papers.

SST cells provide delayed, input-matched dendritic inhibition whose hippocampal analogue is feedback SOM, not an L1 population

Somatostatin interneuron dynamics are delayed relative to fast-spiking interneurons, and this delayed activation is proposed to drive network-wide suppression and accelerate the decay of visual responses [11]. A PV-fast-feedforward versus SOM-slow-feedback division of labour fits calcium imaging of L2/3 during patterned microstimulation, including the finding that SOM cells reach peak activation later than non-SOM neurons [21]. SST cells receive like-to-like recurrent excitation from pyramidal cells, a precision that explains their feature-matched impact on contextual modulation [15]. They gate top-down input in a behaviourally bidirectional way: increasing SST activity impairs associative learning by altering bottom-up signalling, whereas reducing it facilitates learning by gating orbitofrontal top-down input [7]. In the hippocampal formation, the relevant SST population is not in L1 of CA1 but arrives as long-range axons that densely innervate layer I of lateral entorhinal cortex and drive near-selective inhibition of layer IIa principal neurons [6], and as CA1 feedback SOM inhibition that is potentiated by enrichment and is necessary for sparse coding [12].

VIP cells are defined by disinhibition, but no retrieved study measures them during SWRs or spindles

VIP interneurons comprise at least four populations with distinct laminar distributions, arbors, intrinsic properties and efferent connectivity, and stimulating different subpopulations in vivo produces different effects on the cortical network [19]. Their inhibitory outputs onto hippocampal stratum oriens interneurons are target-cell-specific and express NMDA-dependent long-term depression at OLM/bistratified but not fast-spiking targets after brief activation [9]. Median raphe input recruits a speed-coupled VIP ensemble in CA1 during novelty, and a circuit model links this to dendritic disinhibition and place-cell recruitment [17]. Network models that include E, PV, SOM and VIP populations generate multiple coexisting rhythms, but the VIP effect there is an assumed disinhibitory modulation rather than a measured one during spindles [8]. Where ripples have been resolved by cell type, the identified modulators are perisomatic: axo-axonic cells vary from enhanced to suppressed during SWRs and their silencing increases SWR power and duration, in contrast to the near-uniform excitation of other PV cells [13], and cFos-tagged interneurons preferentially recruited during SWRs come from PV and SST classes [14]. Perturbation effects in these circuits are strongly state-dependent, which further cautions against transferring slice- and model-derived VIP actions to the sleeping brain [4].

Conflicts and gaps

The Brecier et al. result contains an internal tension: L1 NDNF inhibition is described as primarily GABA_B-mediated, yet systemic GABA_B antagonism left the inhibitory response during SWRs unchanged and only reduced pyramidal coupling [16]. Either a non-GABA_B conductance carries the ripple-locked event, or the antagonist acted on the network rather than the NDNF→pyramidal synapse; the paper does not resolve this. Subtype homogeneity is also contested in both directions: NDNF cells split into three sleep-stage-tuned populations with opposite spindle-related activation patterns [16], whereas PV cells are described as near-uniformly excited during SWRs apart from the axo-axonic exception [13]. Whether L1 NDNF inhibition promotes or degrades consolidation is untested: none of the retrieved papers pairs L1 subtype manipulation with a memory readout, and the spindles that gate NDNF action are generated thalamically — local thalamic interneurons induce rebound oscillations in relay neurons and neocortical spindles, and their inhibition lengthens spindles and impairs sensory learning [5] — so cortical L1 subtypes may be downstream gain controllers rather than participants in spindle generation. Finally, VIP's role is entirely inferred: sustained VIP manipulation in corticolimbic regions left anxiety-related and social behaviours unchanged and reduced only mPFC-dependent fear learning [18], which argues against a strong, state-general disinhibitory function.

What would change the answer

Closed-loop optogenetic or chemogenetic manipulation of L1 NDNF, SST and VIP cells locked to SWR onset or spindle trough, with simultaneous imaging of identified pyramidal cells and a paired-associate memory readout, would directly test the coincidence-gating claim that currently rests on correlation [16]. Conditional deletion of GABA_B receptor subunits in NDNF versus SST cells would resolve whether the slow inhibition is monosynaptic or network-mediated. Two-photon imaging of molecularly identified VIP and NDNF cells in the hippocampal stratum lacunosum-moleculare during ripples in freely moving mice would determine whether the cortical architecture transfers to the hippocampus. Voltage or calcium imaging of distal apical tufts during spindles would establish whether NDNF acts on the dendritic compartment that spindles actually engage, and paired recordings of NDNF→SST axon contacts during spindle-band activity would test whether the presynaptic veto is state-gated in sleep as it is in locomotion [1].

References

[1] Alyahyay, M., Dumontier, D., Djama, D. S., Jackson, M., Lugtum, A., Menezes, E. C., Ye, H., Qian, J. C., Sansone, G., Soheib, M., Mirow, A. M. C., Ibrahim, L. A., Pouchelon, G.. "Context-dependent presynaptic inhibition of somatostatin interneuron inputs to Layer 1 of the visual cortex". bioRxiv. 2025-09-25T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.25.678558

[2] Bonekamp, K. E., Ratz, M. L., Gillie, G. R., Xiong, L., Sebek, M. K., Crandall, S. R.. "Presynaptic control of top-down signaling in neocortical layer 1". bioRxiv. 2026-02-01T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.31.703032

[3] Richardson, A., Mercier, M. S., Shimoda, Y., Graham, R. T., Leite, M., Garcia, A., Muller, M., Bond, S. A., Grodzdanova, M., Wu, Q., Lieb, A., Kullmann, D. M., Magloire, V.. "NDNF+ interneurons have a privileged role in regulating cortical excitability". bioRxiv. 2025-11-30T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.27.690945

[4] Moreni, G., Dorcioman, R., Pennartz, C. M. A., Mejias, J. F.. "Cell type specific firing patterns in a V1 cortical column model depend on feedforward and feedback-driven states". bioRxiv. 2025-04-07T00:00:00.000Z. URL: https://doi.org/10.1101/2024.04.02.587673

[5] Simko, J., Ramirez, T. K., Martin, C., Leung, S., Sung, E. A., Makinson, C. D.. "Local thalamic interneurons drive spindle termination and enable sleep-dependent learning". bioRxiv. 2025-09-28T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.26.678870

[6] Nilssen, E. S., Jacobsen, B., Doan, T. P., Girao, P. J. B., Witter, M. P.. "Long-range inhibitory axons from medial entorhinal cortex target lateral entorhinal neurons projecting to the hippocampal formation". bioRxiv. 2026-07-31T00:00:00.000Z. URL: https://doi.org/10.1101/2022.11.29.518323

[7] Studer, F., Zeppillo, T., Barkat, T. R.. "Top-down inputs are controlled by somatostatin-expressing interneurons during associative learning". bioRxiv. 2024-12-31T00:00:00.000Z. URL: https://doi.org/10.1101/2024.12.18.629098

[8] Dey Sarkar, A., Ermentrout, B.. "Multi-stable oscillations in cortical networks with two classes of inhibition". bioRxiv. 2025-10-20T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.03.680414

[9] Jablonska, J., Wiera, G., Mozrzymas, J. W.. "NMDA-dependent coplasticity in VIP interneuron-driven inhibitory circuits". bioRxiv. 2025-01-07T00:00:00.000Z. URL: https://doi.org/10.1101/2025.01.07.631664

[10] Fan, J. L., Lee, K., Tchoe, Y., Ganji, M., Vatsyayan, R., Yoon, H. Y., Garrett, J., Dayeh, S. A., Halgren, E., Ji, N.. "Stimulation with ECoG electrodes modulates cortical activity and sensory processing in the awake mouse brain". bioRxiv. 2025-10-30T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.28.685211

[11] Li, J. Y., Cammarata, C. M., Glickfeld, L. L.. "Synaptic dynamics govern spatial integration in mouse visual cortex". bioRxiv. 2025-05-12T00:00:00.000Z. URL: https://doi.org/10.1101/2025.05.11.653343

[12] Verdiyan, E., Kouvaros, S., Gomes, J. I., Bischofberger, J.. "Enrichment experience improves hippocampal sparse coding via inhibitory circuit plasticity". bioRxiv. 2026-04-02T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.31.715605

[13] Gilbert, E. T., Klaver, L. M. F., Arndt, K. C., Kim, J., Jia, X., McKenzie, S., English, D. F.. "Reciprocal interactions between CA1 pyramidal and axo-axonic cells control sharp wave-ripple events". bioRxiv. 2025-01-16T00:00:00.000Z. URL: https://doi.org/10.1101/2024.07.02.601726

[14] Javed, M. H., Robles-Hernandez, E. M., Patel, R., Flores Camacho, A. L., Nessler, L.-J., Haberl, M. G., da Silva, S. V.. "Inhibitory and excitatory cFos engram neurons are preferentially reactivated by sharp wave ripples". bioRxiv. 2026-07-03T00:00:00.000Z. URL: https://doi.org/10.1101/2024.12.17.628897

[15] Hendricks, W. D., Sadahiro, M., Mossing, D. P., Veit, J., Adesnik, H.. "Feature-tuned synaptic inputs to somatostatin interneurons drive context-dependent processing". bioRxiv. 2025-04-19T00:00:00.000Z. URL: https://doi.org/10.1101/2025.04.16.649203

[16] Brecier, A., Mailhos, G., Jarzebowski, P., Li, Y., Paulsen, O., Hay, A.. "Layer 1 NDNF interneurons form distinct subpopulations with opposite activation patterns during sleep in freely behaving mice". bioRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.1101/2025.01.10.632405

[17] Luo, X., Radhakrishnan, R., Guet-McCreight, A., Topolnik, D., Bergeron, A., Dulam, V., Coulombe, L. D., Besner, M., Fortin-Houde, J., Davies, J., Michaud, F., Amilhon, B., Topolnik, L.. "Median raphe input to dorsal CA1 shapes VIP interneuron recruitment and novelty-guided spatial memory". bioRxiv. 2026-09-18T00:00:00.000Z. URL: https://doi.org/10.64898/2026.09.17.752379

[18] Taylor, M., Nguyen, M. A., Pascual Cuadrado, D., George, Y., Schmeisser, M. J., Hesse, J., Krueger-Burg, D.. "Consequences of sustained modulation of VIP-expressing neurons in mouse corticolimbic regions on anxiety-related avoidance behaviors, social functioning, and associative fear learning". bioRxiv. 2026-10-02T00:00:00.000Z. URL: https://doi.org/10.64898/2026.09.30.755164

[19] Dellal, S., Zurita, H., Kruglikov, I., Valero, M., Abad-Perez, P., Geron, E., Meng, J. H., Pronneke, A., Hanson, J. L., Mir, E., Ongaro, M., Wang, X.-J., Buzsaki, G., Machold, R. P., Rudy, B.. "Inhibitory and disinhibitory VIP IN-mediated circuits in neocortex". bioRxiv. 2025-05-17T00:00:00.000Z. URL: https://doi.org/10.1101/2025.02.26.640383

[20] Griesius, S., Richardson, A., Mercier, M., Kullmann, D. M.. "Coincidence detection in aspiny interneuron dendrites". bioRxiv. 2026-05-14T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.10.675107

[21] Hughes, C. L., Kozai, T. D.. "Temporal modulation of cortical parvalbumin and somatostatin neurons shapes network activation and adaptation during microstimulation". bioRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.17.759837

How this survey was made

This survey answers one question: “How do distinct subtypes of layer 1 GABAergic interneurons, identified by molecular markers such as NDNF, VIP, or somatostatin, differentially modulate pyramidal neuron activity during hippocampal sharp-wave ripples and cortical spindles?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 21 papers whose work this text draws on, chosen from the 136 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 21, 2026 and covered work from the previous two years.