Movement and posture depend on sensory feedback that is regulated by specialized GABAergic neurons (GAD2+) that form axo-axonic contacts onto myelinated proprioceptive sensory axons and are thought to be inhibitory. However, we report here that activating GAD2+ neurons directly with optogenetics or indirectly by cutaneous stimulation actually facilitates sensory feedback to motor neurons in rodents and humans. GABAA receptors located at or near nodes of Ranvier of sensory axons cause this facilitation by preventing spike propagation failure at the many axon branch points, which is otherwise common without GABA. In contrast, GABAA receptors are generally lacking from axon terminals and so cannot inhibit transmitter release onto motor neurons, unlike GABAB receptors that cause presynaptic inhibition. GABAergic innervation near nodes and branch points allows individual branches to function autonomously, with GAD2+ neurons regulating which branches conduct, adding a computational layer to the neuronal networks generating movement and likely generalizing to other central nervous system axons.
This is a preview of subscription content, access via your institution
Subscribe to Nature+
Get immediate online access to Nature and 55 other Nature journal
Subscribe to Journal
Get full journal access for 1 year
only $6.58 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Get time limited or full article access on ReadCube.
All prices are NET prices.
All data are available in the manuscript or the supplementary materials. Raw data are available upon reasonable request to the corresponding authors. This study did not generate large datasets or new unique reagents.
Goulding, M. Circuits controlling vertebrate locomotion: moving in a new direction. Nat. Rev. Neurosci. 10, 507–518 (2009).
Capaday, C. & Stein, R. B. Amplitude modulation of the soleus H-reflex in the human during walking and standing. J. Neurosci. 6, 1308–1313 (1986).
Bennett, D. J., De Serres, S. J. & Stein, R. B. Gain of the triceps surae stretch reflex in decerebrate and spinal cats during postural and locomotor activities. J. Physiol. 496, 837–850 (1996).
Andrechek, E. R. et al. ErbB2 is required for muscle spindle and myoblast cell survival. Mol. Cell. Biol. 22, 4714–4722 (2002).
Betley, J. N. et al. Stringent specificity in the construction of a GABAergic presynaptic inhibitory circuit. Cell 139, 161–174 (2009).
Fink, A. J. et al. Presynaptic inhibition of spinal sensory feedback ensures smooth movement. Nature 509, 43–48 (2014).
Hughes, D. I. et al. P boutons in lamina IX of the rodent spinal cord express high levels of glutamic acid decarboxylase-65 and originate from cells in deep medial dorsal horn. Proc. Natl Acad. Sci. USA 102, 9038–9043 (2005).
Eccles, J. C., Eccles, R. M. & Magni, F. Central inhibitory action attributable to presynaptic depolarization produced by muscle afferent volleys. J. Physiol. 159, 147–166 (1961).
Engelman, H. S. & MacDermott, A. B. Presynaptic ionotropic receptors and control of transmitter release. Nat. Rev. Neurosci. 5, 135–145 (2004).
Rudomin, P. & Schmidt, R. F. Presynaptic inhibition in the vertebrate spinal cord revisited. Exp. Brain Res. 129, 1–37 (1999).
Rossignol, S., Beloozerova, I., Gossard, J.P. & Dubuc, R. Presynaptc mechanisms during locomotion. in Presynaptic Inhibition and Neuron Control (eds. Rudmon, P., Romo, R. & Mendell, L. M.) 385–397 (Oxford University Press, 1998).
Ueno, M. et al. Corticospinal circuits from the sensory and motor cortices differentially regulate skilled movements through distinct spinal interneurons. Cell Rep. 23, 1286–1300 (2018).
Jankowska, E., McCrea, D., Rudomin, P. & Sykova, E. Observations on neuronal pathways subserving primary afferent depolarization. J. Neurophysiol. 46, 506–516 (1981).
Zimmerman, A. L. et al. Distinct modes of presynaptic inhibition of cutaneous afferents and their functions in behavior. Neuron 102, 420–434 (2019).
Lucas-Osma, A. M. et al. Extrasynaptic α5GABAA receptors on proprioceptive afferents produce a tonic depolarization that modulates sodium channel function in the rat spinal cord. J. Neurophysiol. 120, 2953–2974 (2018).
Fink, A. J. Exploring a Behavioral Role for Presynaptic Inhibition at Spinal Sensory-Motor Synapses. PhD thesis, Columbia Univ. (2013).
Stuart, G. J. & Redman, S. J. The role of GABAA and GABAB receptors in presynaptic inhibition of Ia EPSPs in cat spinal motoneurones. J. Physiol. 447, 675–692 (1992).
Bardoni, R. et al. Pre- and postsynaptic inhibitory control in the spinal cord dorsal horn. Ann. N. Y. Acad. Sci. 1279, 90–96 (2013).
Szabadics, J. et al. Excitatory effect of GABAergic axo-axonic cells in cortical microcircuits. Science 311, 233–235 (2006).
Howell, R. D. & Pugh, J. R. Biphasic modulation of parallel fibre synaptic transmission by co-activation of presynaptic GABAA and GABAB receptors in mice. J. Physiol. 594, 3651–3666 (2016).
Trigo, F. F., Marty, A. & Stell, B. M. Axonal GABAA receptors. Eur. J. Neurosci. 28, 841–848 (2008).
Barron, D. H. & Matthews, B. H. The interpretation of potential changes in the spinal cord. J. Physiol. 92, 276–321 (1938).
Cattaert, D. & El Manira, A. Shunting versus inactivation: analysis of presynaptic inhibitory mechanisms in primary afferents of the crayfish. J. Neurosci. 19, 6079–6089 (1999).
Alvarez, F. J., Taylor-Blake, B., Fyffe, R. E., De Blas, A. L. & Light, A. R. Distribution of immunoreactivity for the β2 and β3 subunits of the GABAA receptor in the mammalian spinal cord. J. Comp. Neurol. 365, 392–412 (1996).
Walmsley, B., Graham, B. & Nicol, M. J. Serial E-M and simulation study of presynaptic inhibition along a group Ia collateral in the spinal cord. J. Neurophysiol. 74, 616–623 (1995).
Debanne, D., Campanac, E., Bialowas, A., Carlier, E. & Alcaraz, G. Axon physiology. Physiol. Rev. 91, 555–602 (2011).
Goldstein, S. S. & Rall, W. Changes of action potential shape and velocity for changing core conductor geometry. Biophys. J. 14, 731–757 (1974).
Burke, R. E. & Glenn, L. L. Horseradish peroxidase study of the spatial and electrotonic distribution of group Ia synapses on type-identified ankle extensor motoneurons in the cat. J. Comp. Neurol. 372, 465–485 (1996).
Henneman, E., Luscher, H. R. & Mathis, J. Simultaneously active and inactive synapses of single Ia fibres on cat spinal motoneurones. J. Physiol. 352, 147–161 (1984).
Wall, P. D. & McMahon, S. B. Long range afferents in rat spinal cord. III. Failure of impulse transmission in axons and relief of the failure after rhizotomy of dorsal roots. Philos. Trans. R. Soc. Lond. B Biol. Sci. 343, 211–223 (1994).
Swadlow, H. A., Kocsis, J. D. & Waxman, S. G. Modulation of impulse conduction along the axonal tree. Annu. Rev. Biophys. Bioeng. 9, 143–179 (1980).
Wall, P. D. Excitability changes in afferent fibre terminations and their relation to slow potentials. J. Physiol. 142, i3–i21 (1958).
Wu, H. et al. Distinct subtypes of proprioceptive dorsal root ganglion neurons regulate adaptive proprioception in mice. Nat. Commun. 12, 1026 (2021).
Watson, A. H. & Bazzaz, A. A. GABA and glycine-like immunoreactivity at axoaxonic synapses on 1a muscle afferent terminals in the spinal cord of the rat. J. Comp. Neurol. 433, 335–348 (2001).
Lucas-Osma, A. M. et al. 5-HT1D receptors inhibit the monosynaptic stretch reflex by modulating C-fiber activity. J. Neurophysiol. 121, 1591–1608 (2019).
Redman, S. Quantal analysis of synaptic potentials in neurons of the central nervous system. Physiol. Rev. 70, 165–198 (1990).
Zbili, M. & Debanne, D. Past and future of analog–digital modulation of synaptic transmission. Front. Cell. Neurosci. 13, 160 (2019).
Serrano-Regal, M. P. et al. Oligodendrocyte differentiation and myelination is potentiated via GABAB receptor activation. Neuroscience 439, 163–180 (2020).
Zorrilla de San Martin, J., Trigo, F. F. & Kawaguchi, S. Y. Axonal GABAA receptors depolarize presynaptic terminals and facilitate transmitter release in cerebellar Purkinje cells. J. Physiol. 595, 7477–7493 (2017).
Burke, K. J. Jr. & Bender, K. J. Modulation of ion channels in the axon: mechanisms and function. Front. Cell. Neurosci. 13, 221 (2019).
Rudomin, P., Nunez, R., Madrid, J. & Burke, R. E. Primary afferent hyperpolarization and presynaptic facilitation of Ia afferent terminals induced by large cutaneous fibers. J. Neurophysiol. 37, 413–429 (1974).
Aimonetti, J. M., Vedel, J. P., Schmied, A. & Pagni, S. Mechanical cutaneous stimulation alters Ia presynaptic inhibition in human wrist extensor muscles: a single motor unit study. J. Physiol. 522, 137–145 (2000).
Li, Y. et al. Branching points of primary afferent fibers are vital for the modulation of fiber excitability by epidural DC polarization and by GABA in the rat spinal cord. J. Neurophysiol. 124, 49–62 (2020).
Curtis, D. R. & Lacey, G. GABA-B receptor-mediated spinal inhibition. Neuroreport 5, 540–542 (1994).
Pierce, J. P. & Mendell, L. M. Quantitative ultrastructure of Ia boutons in the ventral horn: scaling and positional relationships. J. Neurosci. 13, 4748–4763 (1993).
McCrea, D. A., Shefchyk, S. J. & Carlen, P. L. Large reductions in composite monosynaptic EPSP amplitude following conditioning stimulation are not accounted for by increased postsynaptic conductances in motoneurons. Neurosci. Lett. 109, 117–122 (1990).
Beloozerova, I. & Rossignol, S. Antidromic discharges in dorsal roots of decerebrate cats. I. Studies at rest and during fictive locomotion. Brain Res. 846, 87–105 (1999).
Russ, J. B., Verina, T., Comer, J. D., Comi, A. M. & Kaltschmidt, J. A. Corticospinal tract insult alters GABAergic circuitry in the mammalian spinal cord. Front. Neural Circuits 7, 150 (2013).
Nelson, S. G., Collatos, T. C., Niechaj, A. & Mendell, L. M. Immediate increase in Ia-motoneuron synaptic transmission caudal to spinal cord transection. J. Neurophysiol. 42, 655–664 (1979).
Zehr, E. P. & Stein, R. B. Interaction of the Jendrassik maneuver with segmental presynaptic inhibition. Exp. Brain Res. 124, 474–480 (1999).
Boyden, E. S., Zhang, F., Bamberg, E., Nagel, G. & Deisseroth, K. Millisecond-timescale, genetically targeted optical control of neural activity. Nat. Neurosci. 8, 1263–1268 (2005).
Pinol, R. A., Bateman, R. & Mendelowitz, D. Optogenetic approaches to characterize the long-range synaptic pathways from the hypothalamus to brain stem autonomic nuclei. J. Neurosci. Methods 210, 238–246 (2012).
Zhang, F. et al. The microbial opsin family of optogenetic tools. Cell 147, 1446–1457 (2011).
Chow, B. Y. et al. High-performance genetically targetable optical neural silencing by light-driven proton pumps. Nature 463, 98–102 (2010).
Kralj, J. M., Douglass, A. D., Hochbaum, D. R., Maclaurin, D. & Cohen, A. E. Optical recording of action potentials in mammalian neurons using a microbial rhodopsin. Nat. Methods 9, 90–95 (2011).
Taniguchi, H. et al. A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011).
Madisen, L. et al. A toolbox of Cre-dependent optogenetic transgenic mice for light-induced activation and silencing. Nat. Neurosci. 15, 793–802 (2012).
Madisen, L. et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 13, 133–140 (2010).
Harris, J. A. et al. Anatomical characterization of Cre driver mice for neural circuit mapping and manipulation. Front. Neural Circuits 8, 76 (2014).
Todd, A. J. et al. The expression of vesicular glutamate transporters VGLUT1 and VGLUT2 in neurochemically defined axonal populations in the rat spinal cord with emphasis on the dorsal horn. Eur. J. Neurosci. 17, 13–27 (2003).
Feil, R., Wagner, J., Metzger, D. & Chambon, P. Regulation of Cre recombinase activity by mutated estrogen receptor ligand-binding domains. Biochem. Biophys. Res. Commun. 237, 752–757 (1997).
Harvey, P. J., Li, Y., Li, X. & Bennett, D. J. Persistent sodium currents and repetitive firing in motoneurons of the sacrocaudal spinal cord of adult rats. J. Neurophysiol. 96, 1141–1157 (2006).
Murray, K. C. et al. Recovery of motoneuron and locomotor function after spinal cord injury depends on constitutive activity in 5-HT2C receptors. Nat. Med. 16, 694–700 (2010).
Murray, K. C. et al. Polysynaptic excitatory postsynaptic potentials that trigger spasms after spinal cord injury in rats are inhibited by 5-HT1B and 5-HT1F receptors. J. Neurophysiol. 106, 925–943 (2011).
Street, L. J. et al. Synthesis and biological evaluation of 3-heterocyclyl-7,8,9,10-tetrahydro-(7,10-ethano)-1,2,4-triazolo[3,4-a]phthalazines and analogues as subtype-selective inverse agonists for the GABAAα5 benzodiazepine binding site. J. Med. Chem. 47, 3642–3657 (2004).
Bennett, D. J. et al. Spasticity in rats with sacral spinal cord injury. J. Neurotrauma 16, 69–84 (1999).
Bennett, D. J., Li, Y., Harvey, P. J. & Gorassini, M. Evidence for plateau potentials in tail motoneurons of awake chronic spinal rats with spasticity. J. Neurophysiol. 86, 1972–1982 (2001).
Heckmann, C. J., Gorassini, M. A. & Bennett, D. J. Persistent inward currents in motoneuron dendrites: implications for motor output. Muscle Nerve 31, 135–156 (2005).
Lin, S. et al. Locomotor-related V3 interneurons initiate and coordinate muscles spasms after spinal cord injury. J. Neurophysiol. 121, 1352–1367 (2019).
El-Gaby, M. et al. Archaerhodopsin selectively and reversibly silences synaptic transmission through altered pH. Cell Rep. 16, 2259–2268 (2016).
Li, Y. et al. Pericytes impair capillary blood flow and motor function after chronic spinal cord injury. Nat. Med. 23, 733–741 (2017).
Li, Y., Harvey, P. J., Li, X. & Bennett, D. J. Spastic long-lasting reflexes of the chronic spinal rat studied in vitro. J. Neurophysiol. 91, 2236–2246 (2004).
Vincent, J. A. et al. Muscle proprioceptors in adult rat: mechanosensory signaling and synapse distribution in spinal cord. J. Neurophysiol. 118, 2687–2701 (2017).
Luscher, C., Streit, J., Lipp, P. & Luscher, H. R. Action potential propagation through embryonic dorsal root ganglion cells in culture. II. Decrease of conduction reliability during repetitive stimulation. J. Neurophysiol. 72, 634–643 (1994).
Wall, P. D. Some unanswered questions about the mechanims and function of presynaptic inhibition. in Presynaptic Inhibition and Neuron Control (eds. Rudmon, P., Romo, R. & Mendell, L. M.) 228–241 (Oxford University Press, New York, 1998).
Hubbard, J. I., Llinas, R. & Quastel, D. M. J. Electrophysiological Analysis of Synaptic Transmission (Edward Arnold, 1969).
Munson, J. B. & Sypert, G. W. Properties of single central Ia afferent fibres projecting to motoneurones. J. Physiol. 296, 315–327 (1979).
Dudel, J. The mechanism of presynaptic inhibition at the crayfish neuromuscular junction. Pflug. Arch. Gesamt. Physiol. Menschen Tiere 284, 66–80 (1965).
Katz, B. & Miledi, R. Propagation of electric activity in motor nerve terminals. Proc. R. Soc. Lond. B Biol. Sci. 161, 453–482 (1965).
Armstrong, C. M. & Cota, G. Calcium ion as a cofactor in Na channel gating. Proc. Natl Acad. Sci. USA 88, 6528–6531 (1991).
Stein, R. B. Nerve and Muscle: Membranes, Cells, and Systems (Springer, 1980).
Luscher, H. R., Ruenzel, P., Fetz, E. & Henneman, E. Postsynatpic population potentials recorded from ventral roots perfused with isotonic sucrose: connections of groups Ia and II spindle afferent fibers with large populations of motoneurons. J. Neurophysiol. 42, 1146–1164 (1979).
Leppanen, L. & Stys, P. K. Ion transport and membrane potential in CNS myelinated axons I. Normoxic conditions. J. Neurophysiol. 78, 2086–2094 (1997).
Rushton, W. A. A theory of the effects of fibre size in medullated nerve. J. Physiol. 115, 101–122 (1951).
Arbuthnott, E. R., Boyd, I. A. & Kalu, K. U. Ultrastructural dimensions of myelinated peripheral nerve fibres in the cat and their relation to conduction velocity. J. Physiol. 308, 125–157 (1980).
Blight, A. R. Computer simulation of action potentials and afterpotentials in mammalian myelinated axons: the case for a lower resistance myelin sheath. Neuroscience 15, 13–31 (1985).
Fedirchuk, B. et al. Spontaneous network activity transiently depresses synaptic transmission in the embryonic chick spinal cord. J. Neurosci. 19, 2102–2112 (1999).
Shefner, J. M., Buchthal, F. & Krarup, C. Recurrent potentials in human peripheral sensory nerve: possible evidence of primary afferent depolarization of the spinal cord. Muscle Nerve 15, 1354–1363 (1992).
Boulenguez, P. et al. Down-regulation of the potassium-chloride cotransporter KCC2 contributes to spasticity after spinal cord injury. Nat. Med. 16, 302–307 (2010).
Stalberg, E. et al. Standards for quantification of EMG and neurography. Clin. Neurophysiol. 130, 1688–1729 (2019).
Matthews, P. B. Observations on the automatic compensation of reflex gain on varying the pre-existing level of motor discharge in man. J. Physiol. 374, 73–90 (1986).
Foust, K. D., Poirier, A., Pacak, C. A., Mandel, R. J. & Flotte, T. R. Neonatal intraperitoneal or intravenous injections of recombinant adeno-associated virus type 8 transduce dorsal root ganglia and lower motor neurons. Hum. Gene Ther. 19, 61–70 (2008).
Loeza-Alcocer, E. et al. α5GABAA receptors mediate primary afferent fiber tonic excitability in the turtle spinal cord. J. Neurophysiol. 110, 2175–2184 (2013).
Bravo-Hernandez, M. et al. The α5 subunit containing GABAA receptors contribute to chronic pain. Pain 157, 613–626 (2016).
Wu, K., Castellano, D., Tian, Q. & Lu, W. Distinct regulation of tonic GABAergic inhibition by NMDA receptor subtypes. Cell Rep. 37, 109960 (2021).
Zhou, C. et al. Altered cortical GABAA receptor composition, physiology, and endocytosis in a mouse model of a human genetic absence epilepsy syndrome. J. Biol. Chem. 288, 21458–21472 (2013).
Zhou, W. et al. Loss of function of NCOR1 and NCOR2 impairs memory through a novel GABAergic hypothalamus-CA3 projection. Nat. Neurosci. 22, 205–217 (2019).
Brown, L. E. et al. γ-aminobutyric acid type A (GABAA) receptor subunits play a direct structural role in synaptic contact formation via their N-terminal extracellular domains. J. Biol. Chem. 291, 13926–13942 (2016).
Dzyubenko, E. et al. Inhibitory control in neuronal networks relies on the extracellular matrix integrity. Cell. Mol. Life Sci. 78, 5647–5663 (2021).
Fenrich, K. K., Zhao, E. Y., Wei, Y., Garg, A. & Rose, P. K. Isolating specific cell and tissue compartments from 3D images for quantitative regional distribution analysis using novel computer algorithms. J. Neurosci. Methods 226, 42–56 (2014).
Nicol, M. J. & Walmsley, B. A serial section electron microscope study of an identified Ia afferent collateral in the cat spinal cord. J. Comp. Neurol. 314, 257–277 (1991).
Hines, M. L. & Carnevale, N. T. NEURON: a tool for neuroscientists. Neuroscientist 7, 123–135 (2001).
McIntyre, C. C., Richardson, A. G. & Grill, W. M. Modeling the excitability of mammalian nerve fibers: influence of afterpotentials on the recovery cycle. J. Neurophysiol. 87, 995–1006 (2002).
Cohen, C. C. H. et al. Saltatory conduction along myelinated axons involves a periaxonal nanocircuit. Cell 180, 311–322 (2020).
Stephanova, D. I. & Bostock, H. A distributed-parameter model of the myelinated human motor nerve fibre: temporal and spatial distributions of action potentials and ionic currents. Biol. Cybern. 73, 275–280 (1995).
Olsen, R. W. & Sieghart, W. GABAA receptors: subtypes provide diversity of function and pharmacology. Neuropharmacology 56, 141–148 (2009).
Chua, H. C. & Chebib, M. GABAA receptors and the diversity in their structure and pharmacology. Adv. Pharm. 79, 1–34 (2017).
Chuang, S. H. & Reddy, D. S. Genetic and molecular regulation of extrasynaptic GABA-A receptors in the brain: therapeutic insights for epilepsy. J. Pharmacol. Exp. Ther. 364, 180–197 (2018).
Lagrange, A. H., Hu, N. & Macdonald, R. L. GABA beyond the synapse: defining the subtype-specific pharmacodynamics of non-synaptic GABAA receptors. J. Physiol. 596, 4475–4495 (2018).
Persohn, E., Malherbe, P. & Richards, J. G. In situ hybridization histochemistry reveals a diversity of GABAA receptor subunit mRNAs in neurons of the rat spinal cord and dorsal root ganglia. Neuroscience 42, 497–507 (1991).
Paul, J., Zeilhofer, H. U. & Fritschy, J. M. Selective distribution of GABAA receptor subtypes in mouse spinal dorsal horn neurons and primary afferents. J. Comp. Neurol. 520, 3895–3911 (2012).
Ishizuka, N., Mannen, H., Hongo, T. & Sasaki, S. Trajectory of group Ia afferent fibers stained with horseradish peroxidase in the lumbosacral spinal cord of the cat: three dimensional reconstructions from serial sections. J. Comp. Neurol. 186, 189–211 (1979).
Pan, Z. Z. Transcriptional control of Gad2. Transcription 3, 68–72 (2012).
Esclapez, M., Tillakaratne, N. J., Kaufman, D. L., Tobin, A. J. & Houser, C. R. Comparative localization of two forms of glutamic acid decarboxylase and their mRNAs in rat brain supports the concept of functional differences between the forms. J. Neurosci. 14, 1834–1855 (1994).
Lalonde, R. L. & Bui, T. Do spinal circuits still require gating of sensory information by presynaptic inhibition after spinal cord injury? Curr. Opin. Physiol. 21, 113–118 (2020).
Rudomin, P. Presynaptic selection of afferent inflow in the spinal cord. J. Physiol. Paris 93, 329–347 (1999).
Prochazka, A. & Gorassini, M. Ensemble firing of muscle afferents recorded during normal locomotion in cats. J. Physiol. 507, 293–304 (1998).
Luscher, H. R., Ruenzel, P. & Henneman, E. Composite EPSPs in motoneurons of different sizes before and during PTP: implications for transmission failure and its relief in Ia projections. J. Neurophysiol. 49, 269–289 (1983).
Eccles, J. C., Magni, F. & Willis, W. D. Depolarization of central terminals of group I afferent fibres from muscle. J. Physiol. 160, 62–93 (1962).
Sypert, G. W., Munson, J. B. & Fleshman, J. W. Effect of presynaptic inhibition on axonal potentials, terminal potentials, focal synaptic potentials, and EPSPs in cat spinal cord. J. Neurophysiol. 44, 792–803 (1980).
We thank L. Sanelli, J. Duchcherer, B. Afsharipour and C. K. Thompson for technical assistance and S. Hochman, C. J. Heckman, F. J. Alvarez and T. Bennett for discussions and editing the manuscript. VGLUT1Cre mice cords were kindly donated by F. J. Alvarez. We thank U. Rudolph (McLean Hospital, currently University of Illinois Urbana-Champaign) for providing Gabra5-floxed mice. This research was supported by the Canadian Institutes of Health Research (MOP 14697 and PJT 165823 to D.J.B.) and the US National Institutes of Health (R01NS47567 to D.J.B. and K.F. and R01GM118801 to R.A.P.). These funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
The authors declare no competing interests.
Peer review information
Nature Neuroscience thanks the anonymous reviewers for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
a-b, In the sacrocaudal spinal cord of mice we examined the distribution of GABAA receptor subunits on nodes and terminals of sensory axons, including extrasynaptic α5 subunits, synaptic α1 and α2 subunits, and ubiquitous γ2 subunits (for example forming the common α1βγ2 or α5βγ2 receptors, though less common extrasynaptic α1βδ have been reported)106,107,108,109. We genetically labelled primary sensory axons by their expression of the vesicular glutamate transporter VGLUT1 with a reporter in VGLUT1Cre/+; R26LSL-tdTom mice (tdTom reporter displayed as green, for consistency with Fig. 1). VGLUT1 is mainly only expressed in sensory axons60, especially ventral proprioceptive afferents, as other afferents do not reach the ventral horn15. Axons were reconstructed in 3D as detailed in Fig. 1. 3D reconstructed nodes of Ranvier on myelinated 2nd order ventral branches are shown identified by paranodal Capsr immunolabelling (a, bottom), along with raw confocal image stacks (maximum projection from z-stack) prior to 3D reconstruction (a, top; receptors red). As in rats, nodes were often near branch points (green arrow). Terminal boutons from the ventral horn are likewise shown with raw and 3D rendered images (b). GABA receptors colocalized with the axon are labelled yellow in the 3D reconstructions. Receptor clusters specifically in the plasma membrane are indicated by yellow arrows in (a). Similar to rats, the α5, α1, and γ2 GABAA receptor subunits were found on large axon branches (1st and 2nd order) in the dorsal, intermediate, and ventral cord, near nodes (a), but not on ventral horn terminal boutons (b, 3rd order). As also seen in rats (Fig. 1e), synaptic GABAA receptors were usually in single large membrane bound clusters at nodes, whereas extrasynaptic α5 receptors were often broken up into multiple clusters, with the largest clusters in the membrane (yellow arrows), and smaller cytoplasmic clusters several µm from the edge of the node under the paranodal Caspr (grey arrows; Fig. 1f). Cytoplasmic α5 receptors have been reported previously95. Many receptors were in neighboring neurons (red arrows), and in our previous publication15 these and cytoplasmic axon receptors may have been mistaken for nodal receptors in the axon membrane, though this is corrected here with evaluation of higher resolution confocal images. The presence of these α5, α1, and γ2 GABAA subunits is consistent with their mRNA previously reported in the dorsal root ganglion110. Also, the finding of α1 subunits on these axons is consistent with the recent observation that α1 is only on myelinated sensory axons, rather than unmyelinated C fibres111. c, Synaptic α2 GABAA receptor labelling in mouse and rat axons (rat axon labelled as in Fig. 1) with nodes labelled with both antibodies to sodium channels NaV and Caspr, to confirm the relation of receptors to the node and paranodal region. Receptors are often at the transition between the node and paranodal Caspr, as here. Putative GABAergic synaptic contact labelled with VGAT. d, Ventral terminal boutons in VGLUT1Cre/+; R26LSL-tdTom mouse (with reporter labelling the complete axon, green) immunolabelled with VGLUT1 to verify that these are afferent terminals, which have vesicular VGLUT1+ protein expression. Immunolabelling for γ2 GABAA receptors again showed that terminals lacked this ubiquitous subunit that makes up most GABAA receptors. e, Immunolabelling for GABAB receptors on 3D reconstructed sensory axons, with same format and mice of (a-b). GABAB receptors were generally absent from nodes identified by paranodal Caspr, but present on ventral terminal boutons, as in rats (Fig. 1). Similar results (a- e) were obtained from n = 5 mice. f, Synaptic α1 GABAA receptor labelling at a rat Ia afferent node labelled with both antibodies to sodium channels NaV and Caspr, to confirm the relation of receptors to the node and paranodal region in rat, like in mouse (c, same node as in left of Fig. 1c, but NaV and raw images included). Putative GABAergic synaptic contact labelled with VGAT, where left red arrow shows GABAA receptor contacting VGAT at edge of node. VGAT is near, but does not contact Caspr (in 3D view), but may well contact the paranodal myelin loops, since oligodendrocytes express GABA receptors38. Node is at dorsal 2nd order axon branch point. (a-f) representative of 5 mice. g-h, Box plots of the proportion of ventral terminal boutons with GABA receptors in both mice (g,VGLUT1+) and rats (h), again showing that few boutons contain GABAA receptors, whereas many contain GABAB receptors. * significantly more GABAB than GABAA receptors, ANOVA with Bonferroni correction, P < 0.05, n = 5 animals per condition, with 70–120 terminals examined per animal and receptor. Box plots show the interquartile range (box), median (thin line), mean (thick line), 10 and 90 percentiles (whiskers), and extremes (dots).
Extended Data Fig. 2 Voltage dependence of nodal spike propagation failure following dorsal root stimulation.
a-d, Intracellular recording from proprioceptive Ia afferent branches in the rat dorsal horn with secure spikes at rest, evoked by DR stimulation (1.1xT, 0.1 ms; T: afferent volley spike threshold; sacral S4 DR; a, d). Spike failure was induced by increasing hyperpolarization (failure near rest in d, but not a), with a delay and then abrupt loss of height, reflecting failure of successively further away nodes (a, d). Of the two local nodes adjacent to the electrode, one failed first with hyperpolarization, leaving the attenuated spike from the other node (local FP, about 1 λS away), which eventually failed as well with further hyperpolarization, leaving a much smaller FP from more distal nodes (distal FP). Spike failure in a node was always proceeded by a delay in the nodal spike. Estimated contributions from local nodes (b-c) were computed by subtraction from traces in (d). Spike attenuation was down to about 1/e by the second failure, consistent with the space constant λS being two internodal distances (c), or about 90 µm. Note that due to attenuation of injected current with distance, larger hyperpolarizations were needed to stop spikes with more distal vulnerable nodes (secure spikes), as observed by smaller distal FPs (a verses d), and some axon spikes could not be stopped (Spike only, quantified in e). Data collected in discontinuous current clamp (DCC) mode so electrode rectification during the injected hyperpolarizing current did not affect potential (DCC switching rate 7 kHz, low pass filtered at 3 kHz to remove switching artifact, which also removed stimulus artifact). e, Distribution of spiking in branches with full spikes only, one local nodal spike (local FP), or a distal nodal spike (distant FP) remaining after maximal hyperpolarization, and * significant change when blocking α5 GABAA receptors with L655708 (0.1–0.3 µM) using two-sided χ-squared test, P < 0.05; n = 68 control and n = 47 L655708 treated axon branches from 7 rats. f, Box plots of spike or FP heights and delays in branches and rats indicated in (e), measured just prior to spike failure (or at maximal hyperpolarization for secure spikes) and after failure (for local and distal FPs, as induced in a-d). Delay measured relative to peak of spike at rest. + FP significantly different than spike at rest, n = 7 rats, two-sided paired t-test, P < 0.05. g, Box plots of the hyperpolarization needed to induce failure, in axon branches and rats from (e). * significant change with L655708, two-sided unpaired t-test, P < 0.05, n from (e). Box plots show the interquartile range (box), median (thin line), mean (thick line), extremes, 10 and 90 percentiles (whiskers). h, Incidence of failure at varying potentials (% of total spikes from e). * significantly more failure with L655708, two-sided χ-squared test, n from (e), P < 0.05. Note that secure spikes in rats and mice are not overall different when measure by spike height at rest (Fig. 2g and 3g). However, the spike height shown here in (f) is the height while the cell is hyperpolarized far from rest, and so is much larger than at rest, since spikes generally overshoot to near the reversal potential for sodium, and so spike heights while hyperpolarizing are not comparable to the spikes at rest in mice or rats (measured from holding potential to peak). Thus, during hyperpolarization the nodes produce larger overall spikes, including FPs from local nodes (a-f). Also, during hyperpolarization from current injection, the two adjacent nodes to the electrode fail, unlike during natural failures, and thus when one nodes fails the spike only halves in height, leaving the spike from the second adjacent node (a-b). Thus, these local FPs are large in relation to the FPs from distal nodes with natural failure (Fig. 2).
Simulating spike propagation failure in a proprioceptive axon by applying a brief intracellular current injection to mimic the current arriving from an upstream node (and FP), yielded full spikes evoked at rest, but nearby nodal spikes delayed and then failing as the membrane was held progressively more hyperpolarized with a steady bias current. Also, large steady depolarizations inactivated these spikes, though well outside of the physiological range (> -50 mV). a, Intracellular recording from proprioceptive Ia afferent branch in the rat dorsal horn (sacral S4 axon). A brief current injection pulse (0.5 ms) was applied to simulate the current arriving from distal nodes during normal axon spike conduction, and repeated at 1 s intervals. During these pulses the membrane was held at varying potentials for 1–2 s with steady current injection, with numbers and colours denoting a given holding potential (in DCC mode). At the most hyperpolarized levels spikes failed to be evoked and only the passive response is seen, like a FP (blue, 1). As the potential was depolarized to near the axon’s resting potential (-67 mV) partial spikes occurred (green, 2 and 3), likely from a single adjacent node activating, and then delayed broad spikes occurred, as both adjacent nodes were activated. At more depolarized levels the spikes arose more rapidly and increased in height to full secure spikes (4). b, In the same axon as (a), at holding potentials well above those seen physiologically (near -50 mV, lower plots) spikes started to exhibit sodium channel inactivation and failure, with a decrease in spike height and delay (7–8) and eventually full failure (shown in c). Adjacent nodes started failing at slightly different times with different delays, broadening the spike and eventually separating into two distinct nodal spikes (8*). c, Spike heights plotted as a function of holding potential, including those spikes illustrated in (a and b), with spike number-labels indicated. Left grey line indicates passive leak current response, and shows deviation from passive response near rest. Shaded green region shows all or nothing failure or spikes near the resting potential. Middle grey line shows a region of secure spikes with relatively invariant spikes. Right grey line shows spike inactivation with large depolarizations and outright failure near - 50 mV. Note the split vertical axis. Similar voltage dependence of spike failure occurred for n = 5/5 axons showed similar results, from 4 rats. This demonstrates two modes of spike failure: 1) spikes that fail at rest or at hyperpolarized potentials and 2) spikes that fail with large depolarizations above rest. The latter is likely not physiological, since even the largest PAD that we have observed (5–10 mV; Fig. 4d) does not depolarize axons to -50 mV, since axons rest near -70 mV, and PAD is only large at hyperpolarized levels (Fig. 4d) and decreases steeply as the potential approaches the reversal potential for chloride (-15 mV)15. d, Schematic of recording arrangement and relation to adjacent nodes for data in (e) and (f). e, Expansion of responses 1 (blue) and 3 (pink) from (a), and difference (cyan) to show the first local nodal spike height at threshold, recorded at electrode. Active nodes from schematic in (d) shown with shaded boxes. f, Spikes near sodium inactivation from (b) (7 and 8), with differences indicating local nodal spikes (grey: 7–8, and cyan: 7–7’, both truncated to only show estimated nodal spike). Nodes likely arranged as in (d).
Extended Data Fig. 4 GABAergic innervation of nodes and nearby boutons in viral vector labelled proprioceptive afferents.
a, Sensory afferents labelled in the spinal cord and DR of adult GAD2//ChR2-EYFP mouse (GAD2-EYFP) by a peripheral adeno-associated virus (AAV9-tdTom); confocal image of a transverse spinal cord section. Large myelinated proprioceptive Ia afferents were identified by characteristic extensive ventral horn branching and innervation of motor neurons (as in Fig. 1a). These Ia afferents left the dorsal columns in bundles that were readily traced to the ventral horn in serial sections, with an S-shaped projection path from the dorsal columns to the motor neurons (as in Fig. 3j)112. The relatively sparce afferent labelling helped facilitate this tracing of afferent innervation. The peripheral AAV9-tdTom injection occasionally also labelled one motor neuron in a transverse (lower white arrow), though this only occurred a couple times per spinal cord segment, making them possible to distinguished from afferents elsewhere. No other central neurons were labelled, indicated that the viral vector only affected peripheral sensory and motor axons, as we previously found43. EYFP+ cells (GABAaxo neurons) were only in the dorsal and intermediate laminae (GAD2-EYFP+), with projections into the ventral horn and dorsal columns along the entire path of the Ia afferent bundles, as detailed in Fig. 3. The approximate regions where images in (b-h) were taken are indicated in (a), but in different transverse sections, oriented similarly. b, GABAaxo neurons formed a dense plexus that wound around Ia afferents as they branched out of the dorsal columns into the grey matter (at dashed line). c, Main 1st order myelinated branch of a Ia afferent in intermediate laminae with three nodes at branch points labelled with axonal caspr, approximate myelin regions marked with white lines, and GABAaxo neurons seen wrapping the axon (top images). Node 1 had a GAD2-EYFP + contact that had presynaptic GAD2 labelled with an antibody (GAD2-Ab, cyan; yellow marks contact of afferent, GAD2-EYFP and GAD2-Ab computed in 3D), whereas Nodes 2 and 3 lacked contacts, though the short branch arising off of Node 3 had a nearby unmyelinated terminal branch with a GAD2+ contact. GAD2 (GAD65) is known to be closely associated with GABAergic terminals and vesicular GABA and highly enriched in terminals compared to elsewhere in neurons5,7,113,114, and thus the GAD2-Ab provides a presynaptic contact label, consistent with the intense GAD2-Ab labelling we observed in GAD2-EYFP boutons (also bassoon+), and similar to the GAD65-intense boutons on afferents previously detailed7. Expanded view of Node 1 shown at bottom left, where all caspr labelling is displayed, including in overlapping nearby axons, whereas in other images only the axonal caspr in the 3D volume of the afferent is shown for clarity, together with raw images from other antibodies. Expanded view of Node 1 contact shown on bottom right, where all GAD2-ab labelling is shown, whereas in other images only GAD2-Ab in the GAD2-EYFP neuron volume is shown. Not all GAD2-Ab was in GAD2-EYFP+ neurons, likely due to variability in tamoxifen induced cre or transport of EYFP in GAD2//Ch2R-EYFP mice. d, Nodal contact in the ventral horn, shown with similar format to (c), with again a GABAaxo neuron wrapping around the axon and specifically making a contact at the node (yellow; EYFP + , GAD2-Ab+). White lines: estimated myelin regions. 1st order afferent. e, Two nodes on two 1st order Ia afferent branches in the dorsal horn plexus of (b), again delineated by paranodal caspr, one of which had a GABAaxo neuron contact (GAD2-EYFP, marked yellow again) with presynaptic GAD2-Ab (cyan, shown in 3D volume of GABAaxo neuron) and the other did not. f-g, Nodes from 1st order Ia afferent branches, identified by paranodal Caspr, in control wild-type mice, and so lacking EYFP, but inhibitory innervation examined with VGAT or GAD2 antibodies. In (f) the node had a direct VGAT+ contact and nearby contacts on the small unmyelinated branch arising from the node (contacts computed in 3D labelled yellow). In (g) the node lacked a direct GABAergic contact, but had a nearby GAD2-Ab+ contact on a bouton of a short unmyelinated branch arising from the node. h, Complex node in intermediate laminae with enlarged inter-myelin region and two unmyelinated branches, one with a GAD2-Ab+ contact on the neck of the branch, and the other with contacts on nearby boutons (above top of image). Same format as (d-e). Similar to enlarged nodal boutons of Walmsley25. (a-h) representative of n = 3 mice. i, Quantification of the fraction of total nodes with direct synaptic GABAergic contacts (nodal contacts, GAD2-Ab+, ~25%), nearby contacts at a node or unmyelinated terminal branch/bouton on the same axon (within λS = 90 µm; 98–77%; as in c) and putative extrasynaptic innervation (within 5 µm, ~95%) from GABAaxo neurons, with presynaptic GABA inferred from GAD2-Ab immunolabelling, in n = 3 mice from n = 43–53 nodes, each on 1st and 2nd order branches of Ia afferents. VGAT+ contacts on or near nodes occurred with similar incidence (28% of nodes had synaptic contacts, from n = 50 nodes in n = 3 mice). Similar incidence of contacts occurred in rat afferents labelled with neurobiotin (n = 5 rats). Synaptic contacts occurred at unbranched or branched nodes, as did extrasynaptic innervation, with about half the nodes branched overall. Synaptic contacts that occurred on short unmyelinated afferent terminal branches arising from the node usually occurred at a bouton and are labelled as: Nearby bouton. Synaptic GAD2+ contacts occurred most frequently at or near complex branch points where a parent branch split into large daughter branches (d) or the daughter branches had large boutons (g-h), which theoretically increased the local conductance and probability of spike failure downstream to the branch point (as in Extended Data Fig. 5).
a, Model of a 3D reconstructed proprioceptive afferent, drawn to scale, except myelinated branch lengths all shortened an order of magnitude. Double line segments are myelinated (white) and the rest unmyelinated. Adapted from anatomical studies of Walmsley25. Nodes are indicated with a green dot, and ventrally projecting terminal boutons indicated with a yellow dot. As in our axons of Fig. 1, branch points were always at nodes. GABAA receptors of equal conductance (nS) were placed at each node and associated branch points, and total dorsal columns (dc) depolarization from phasically activating these receptors is shown in inset. The branch lengths (L) and computed space constants (λS) are indicated in gray boxes for each segment of the afferent, the latter computed from subthreshold current injections into each segment. From left to right the gray boxes are for segments spanning from the dorsal columns (dc) to N0, N0 to N1, N1 to N5, N1 to N4, N4 to B2 and N5 to B1. Average space constant was λS = 91 µm, similar to in other axons39. b, Responses to simulated dorsal root (DR) stimulation (0.1 ms pulse, 2 nA, at black arrows) computed at various downstream branch points (nodes) and terminal boutons in the spinal cord, with resting potential indicated by thin dashed line (-82 mV). A sodium spike propagated to the branch point at node N1, but failed to invade into the downstream branches, leaving nodes N2 and N3 with only a passive depolarization from the N1 spike (failure potential, FP). Further downstream nodes and terminals experienced vary little depolarization during this failure (N5, B1 and B2). In this case the GABA conductance was set to zero (gGABA = 0, control), simulating a lack of GABA tone. Note that only the nodes beyond the parent branch at node N1 failed due to the conductance increases in large daughter branches to nodes N2 and N3 that drew more current than node N1 could provide (shunting conductances). Node 3 is particularly interesting as it does not itself branch, though a neighbouring node has a small branch, both contributing to the overall conductance and related failure, and Node 2 has two adjacent branches contributing to its conductance. Other branch points with relatively smaller conductance increases (N0, simpler branching) did not fail to conduct spikes. Generally speaking, if the upstream node of a parent branch cannot provide enough current to activate the nodes of its daughter branches then spikes fail, and this is especially likely with multiple sequential branch points, like in N1–N2. However, failure even occurs at daughter nodes than themselves lack branches (N3), so GABA receptors are useful in aiding spikes at unbranched nodes. c, As seen experimentally, nodal GABAA receptor activation to produce PAD prior to the DR stimulation (~ 10 ms prior; as in Fig. 4f) rescued spikes from failing to propagate. That is, with GABAA receptors placed just at nodes and associated branch points a weak phasic activation of these receptors (conductance gGABA = 0.6–1.5 µS per node shown) rescued conduction down the branch to node N2, with full nodal spikes seen at the distal node N4 and the terminal bouton B2 (DR stimulation at peak of PAD, detailed in (g) with black arrows indicating DR stimulation timing). A larger GABA receptor activation (2.4 nS) additionally rescued spike conduction down the branch to node N3, with full spike conduction to the distal node N5 and the terminal bouton B1. Note that increasing GABA conductance sped up the arrival of distal spikes (for example at N4 and B2), by up to 1 ms, suggesting substantial variation in sensory transmission times induced by GABA, as we see experimentally. Also note that this nodal GABA depolarized the nodes (N1–N3) relative to rest (thin dashed line), thus assisting spike initiation. In contrast, nodal GABA did not depolarize the terminal boutons (B1 and B2), consistent with our recent direct recordings from terminals15. Sensitivity analysis revealed similar results with a wide range of sodium channel and GABA receptor conductances, though increasing sodium conductance sufficiently prevented failure all together (like in Extended Data Fig. 10f). Interestingly, when we put GABA receptors only at node N2 (or at the unmyelinated bouton immediately above N2) then the spike propagating though N2 to the terminal bouton B2 was rescued with the same GABA conductance (0.6 µS) as in the main simulation (c). Likewise with GABA receptors only at node N3 and nowhere else, then the spike was rescued at that node N3 (at 2.4 µS, as in c). d, When instead we removed all nodal GABA receptors and instead place them on terminal boutons (near B1 and B2, yellow, with equivalent total conductance, 2.4 µS condition), then activating them did not rescue the spike propagation failure, since the associated depolarization of nodes is too attenuated at the failure point (N1−N3; no change from resting potential). The GABA receptors did depolarize the terminal boutons (B1 and B2, thick dashed lines) substantially relative to the resting potential (thin dashed lines), but this depolarization was sharply attenuated in more proximal nodes (N1-3). e, Reduction of spike height (shunt) and speeding of spike onset with increasing GABA conductance at a non-failing node (N1; model with nodal and not terminal bouton GABA conductances, c), consistent with actual recordings from axons in Fig. 3d and Extended Data Fig. 8a. f-h, PAD recorded at the dorsal columns (dc) during conditions in (b-d), respectively, as experimentally recorded dorsal PAD. A phasic GABA induced depolarization (PAD) was generated by changing GABA conductances, gGABA, as detailed in Methods, and GABA receptor location varied as in (c-d). DRs were stimulated at the peak of this PAD in (c-d). Note that nodal (g) but not terminal (h) GABAA receptors caused a visible depolarization (PAD) at the dc, due to less electrotonic attenuation over a shorter distance to the dc.
Extended Data Fig. 6 Cutaneous driven trisynaptic circuits mediating PAD and assisting repetitive firing.
a, Cutaneous driven dorsal trisynaptic circuit mediating PAD. A minimally trisynaptic circuit is classically known to depolarize afferents via GABAaxo neurons. This circuit involves sensory afferents activating excitatory intermediary neurons (glutamatergic) that in turn activate GABAaxo neurons that return to innervate sensory axons13,115. Even though GABAaxo neurons are small5 this circuit influences afferents over widespread regions of the spinal cord15. Specifically, the activation of a small group of sensory axons in just one DR or nerve causes this trisynaptic circuit to produce a widespread activation of many axons across the spinal cord, even many segments away and across the midline15. This allowed us to activate PAD from adjacent roots without directly activating an axon in a particular root, as detailed in Fig. 4 and the rest of this figure. One variant of this classic trisynaptic circuit specifically involves cutaneous stimulation activating dorsal intermediary neurons13,14 that activates GABAaxo neurons (likely dI4 neurons115) that in turn innervate cutaneous afferents, which we term the cutaneous dorsal circuit. While this cutaneous dorsal circuit also synaptically innervates some proprioceptive afferents116 (a), its main action on proprioceptive afferents is to produce a pronounced extrasynaptic spillover of GABA that depolarizes these afferents tonically via α5 GABAA receptors (termed: tonic PAD, L655708 sensitive), especially with repetitive cutaneous nerve stimulation (1–200 Hz) that leads to minutes of depolarization15, and we see similar tonic PAD here (detailed next). b, Intracellular recording from a proprioceptive axon branch in rat dorsal horn (sacral S4 axon, DR2). The axon branch spontaneously exhibited spike propagation failure when its was stimulated alone (denoted DR2 stimulation, repeated at 1 Hz, 1.1xT, 0.1 ms; T: afferent volley spike threshold), with only a small failure potential (FP) visible (lower pink traces). Activation of a largely cutaneous DR (caudal Ca1 DR, innervating the tip of the tail, stimulation at intensity for cutaneous afferents, 3xT, 0.1 ms; denoted DR1) evoked a slowly rising tonic PAD when repeated at 1 Hz (blue). When the axon stimulation (DR2 stimulation) was combined with the repeated cutaneous stimulation (DR1, 60 ms prior to each DR2 stimulation) the slowly building PAD prevented spike failure (black spikes), and this outlasted the cutaneous stimulation (after effect). Similar results obtained in n = 20/20 axons tested from 10 rats.
Extended Data Fig. 7 Proprioceptive driven trisynaptic circuit for PAD enabling high frequency spike transmission.
a, Proprioceptive driven ventral trisynaptic circuit mediating PAD. Another variant of the classic trisynaptic circuit involves proprioceptive afferents activating excitatory intermediary neurons (glutamatergic) that then activate GABAaxo neurons that innervate these same afferents, including ventral terminal regions of the afferents13,15,115. This circuit is more ventrally located compared to the cutaneous dorsal PAD circuit13, and thus, we term it the proprioceptive ventral PAD circuit. It likely involves GABAaxo neurons that are from the dILA population115. Importantly, the circuit is recurrent, with proprioceptive afferents causing self-facilitation of themselves (via homonymous PAD). It produces a fast phasic axon depolarization (phasic PAD, fast synaptic; Fig. 4c), as well as a slower tonic depolarization (tonic PAD, Fig. 4c, likely from extrasynaptic GABA spillover), as detailed previously15. Since proprioceptive sensory axons naturally fire at high rates117 where they are vulnerable to spike failure (Fig. 2e, f), we examined the action of self-facilitation by GABA on this failure. b-f, During rapid repetitive stimulation of a DR to evoke spikes in a proprioceptive axon there was an inevitable activation of PAD from low threshold proprioceptive axons (homonymous PAD, b). This PAD helped spikes fire at high physiological rates of up to 200-300 Hz (5–3 ms intervals) before spike inactivation and failure occurred because, in absence of PAD, isolated repetitive activation of the axon with intracellular current pulses (IC) led to failure at much lower firing rates (~100 Hz; longer spike intervals; b, e), even after just two stimuli (doublets; c, e). Additional PAD evoked by simultaneous stimulation of an adjacent DR (2xT; T, spike threshold observed from afferent volley) reduced failure from fast repeated IC stimuli (b, f), repeated DR stimuli (doublet, c, d, f) or hybrid IC-DR stimulation pairs (f). Legend details are as follows: b, Intracellular recording from proprioceptive axon in rat dorsal horn (sacral S4 axon) with spikes securely evoked by fast repeated DR stimulation (top, 1.1xT, 0.1 ms, sacral S4 DR, denoted DR2; resting potential -68 mV), but spikes failing intermittently with repeated intracellular current injection (IC) at the same rate (bottom green), due to sodium channel inactivation. The reason that spike failure does not occur with the fast DR stimulation is that it is accompanied by a build up of tonic PAD (from self-activation) that helps prevent failure, because adding to the IC stimulation a simultaneous conditioning stimulation of other proprioceptive afferents in an adjacent DR (DR1 stim at 1.5xT, 0.1 ms, S3 DR) prevents spike failure (black trace, bottom), via the proprioceptive ventral circuit (a). This DR1 conditioning stimulation does not directly activate spikes in the axon, but it causes a fast depolarization (phasic PAD) that rapidly helps spikes (as early as 6–10 ms later), and a building tonic depolarization (tonic PAD) with repetition that further helps later spikes in the stimulation train (DR1 stimulation alone blue, middle trace). Similar results obtained in n = 7/7 axons from 4 rats. Likely similar tonic PAD and associated increased spike conduction helps explain post-tetanic potentiation of the monosynaptic EPSP, as previously suggested118. c, Repeated DR stimulation at higher rates eventually causes spike failure in proprioceptive axons (sodium spike relatively refractory), as shown in the top panel where a double stimulation (doublet, S4 DR, denoted DR2−DR2, 1.1xT, 0.1 ms, resting at -75 mV) exhibits failure on the second spike (with large FP indicated, magenta). However, additional PAD provided by stimulating an adjacent DR (DR1; 1.5xT, group I intensity, 0.1 ms) about 10 ms earlier helps prevent this spike failure (black trace; blue trace: PAD alone). When the same axon was stimulated slightly slower (with a longer doublet interval, second plot, DR-DR) failure did not occur, which we designate the failure interval threshold, which is quantified in (e). The self-activated PAD caused by the first DR stimulation in this doublet helped prevent failure in second DR stimulation because replacing the first DR stimulation with an intracellular stimulation (IC, 2 nA) to activate the spike leads to failure of the second spike evoked by the DR stimulation at much longer intervals (lower trace, IC-DR). d, Another example of a failed doublet spike (DR2-DR2 stim, 1.1xT, 0.1 ms) that was rescued by PAD evoked by adjacent DR stimulation (DR1 1.5xT, 0.1 ms, resting at -78 mV), as in the top plots of (c), except that in this case the failure is at a more distal node, since the FP is small. e, Failure interval threshold (minimum firing interval prior to failure, or maximal firing rate) with DR doublets (DR-DR), IC doublets (IC-IC) or IC-DR pair stimulation. Note the shorter intervals possible with the PAD evoked by the first stimulation (DR-DR). * significantly longer than minimum DR doublet interval (DR-DR), one-way ANOVA with post hoc Tukey test for multiple comparisons, n = 18 axons each condition from 5 rats, P < 0.05. f, Quantification of the FP heights that were induced by a fast doublet (DR-DR or IC-IC; n = 14 axons each from 5 rats, at failure threshold interval) or IC and DR stimulation (IC-DR; n = 11 axons from 5 rats), and the rescue of spikes by PAD evoked by adjacent DR stimulation for each condition. * significant increase in height with PAD, two-sided paired t-test, P < 0.05. Box plots show the interquartile range (box), median (thin line), mean (thick line), extremes, 10 and 90 percentiles (whiskers).
a, Intracellular recording from a rat sacral S3 proprioceptive Ia afferent branch in the dorsal horn with a secure spike evoked by S3 DR stimulation at rest (DR2, 1.1xT, 0.1 ms, −60 mV rest, rat; T: spike threshold in afferent volley). Sensory-evoked PAD initiated by stimulating an adjacent DR (DR1; S4 DR; 2xT, 0.1 ms pulse, as in Fig. 4) 10 ms prior to the DR2 stimulation only moderately influenced the spike (DR1 stimulation time). It sped up the spike latency and rise time, reduced the fall time and slightly reduced the spike height. Hyperpolarization induced spike failure (lower trace), as in Extended Data Fig. 2a. b, Summary box plots of change in spike peak latency (advance) and height with prior sensory PAD activation as in (a). *significant change, two-sided paired t-test, P < 0.05, n = 26 axons, from 9 rats. c, Intracellular recording from an S3 proprioceptive Ia afferent branch in the dorsal horn with a brief current injection pulse just subthreshold to initiating a spike (near rheobase, bottom trace) at rest, only initiating a passive response with a small failed spike (middle trace). However, prior activation of PAD by stimulating an adjacent DR (DR1 as in A; S4 DR, 2xT, 0.1 ms top trace) allowed the same current pulse to evoke a spike (above rheobase). The passive response to the current injection (double blue arrow; resistance R = V/I) was decreased during the PAD, corresponding to an increase in conductance, that contributed to a shunt (reduction) of the currents generating the spike, though this only caused about a 1% drop in spike height (1 mV; b). DCC recording mode, as in Extended Data Fig. 2a–d. d, Summary box plots of rheobase (current threshold from c) before and during PAD, and change in shunt (conductance = 1/R) with PAD, as in (c). Box plots show the interquartile range (box), median (thin line), mean (thick line), extremes, 10 and 90 percentiles (whiskers). * significant change with PAD, two-sided paired t-test, P < 0.05, n = 37 axons from 11 rats. e, By itself sensory evoked PAD sometimes initiated a spike on its rising phase, when the DR stimulation was large enough, demonstrating a direct excitatory action of GABAA receptors, as previously reported15. These spikes propagate antidromically toward the DR; and are thus termed dorsal root reflexes (DRR). Example shown of intracellular recording from rat sacral S3 proprioceptive afferent branch in the dorsal horn, with PAD produced by a DR1 stimulation (S4 DR stimulation 3.5xT, 0.1 ms) evoking a spike that propagates out the DR (DRR). These DRR occurred in n = 11/120 axons from 15 rats (9% incidence). These PAD evoked spikes occur with a variable latency of 10–30 ms15 and thus make axons refractory for about 30 ms after the DR stimulation119. PAD evokes spikes with a high incidence (79%)15, but these spikes fail to propagate antidomically, yielding the 9% incidence we see. However, these spikes are more likely to travel othodromically (up to 79% incidence) and evoke EPSPs in motor neurons via the monosynaptic pathway8, and thus also produce a post activation depression of the EPSPs for many seconds. We thus kept the PAD low when examining the effects of PAD on EPSPs, to avoid these spikes and their subsequent inhibitory action (in Figs. 5–6).
Extended Data Fig. 9 Estimating the overall spike conduction failure from the dorsal root to the motor neurons.
a, Experimental setup to indirectly measure sensory axon conduction failure following DR stimulation, by examining whether failed axon segments are relatively less refractory to activation after failure, using a double pulse method adapted from Wall30. A tungsten microelectrode (12 MΩ) was placed in the ventral horn (VH) near the sensory axon terminals on motor neurons (S3 or S4 VH), to activate the branches/nodes of the axon projecting to the motor neuron that may have failed (VH stimulation). Spikes from VH or DR stimulation were recorded intracellularly in a proprioceptive axon penetrated in the dorsal columns. b, VH threshold in refractory period. Rapidly repeated VH stimulation (VH doublet; two 0.1 ms pulses) at an interval short enough to produce spike inactivation on the second stimulation (4 ms), with stimulus current adjusted to threshold for inactivation, TVH2. This TVH2 (~15 uA) was always higher than the threshold VH stimulation for evoking a spike with the first stimulation, TVH1 (~10 uA). Recorded in sacral S4 afferent resting at -72 mV, with doublets repeated at 3 s intervals to determine current thresholds. c, VH threshold after DR stimulation. Similar repeated activation of the axon in (b), but with the first activation from a DR stimulation (at 1.5x DR threshold) and the second from VH stimulation at the TVH2 intensity from (b). In this case the VH stimulation readily activated the axon spike, likely because the DR-evoked spike did not propagate to the VH, leaving the silent portion of the axon non refractory. Thus, this VH stimulation evoked spikes with a lower current than TVH2, with this lower threshold denoted TDR,VH (~ 12 µA). This DR–VH stimulation interval was deliberately set too short for the involvement of PAD (which rises in > 4 ms; Fig. 4). d, Computation of spike failure based on changes in VH stimulation thresholds. If the DR-evoked spike entirely fails to propagate to the VH, then the threshold for subsequently activating the VH (TDR,VH) should be the same as the threshold without any prior activation (TVH1 = TDR,VH), whereas if it does not fail then the threshold for activating the VH should be the same as with a VH doublet (TVH2 = TDR,VH). In between these two extreme scenarios, the DR-evoked spike may only partially fail to propagate spikes to the VH; in this case TDR,VH should be between TVH1 and TVH2, with the difference TVH2 - TVH1 representing the range of possible thresholds between full failure and conduction. Overall the % conduction failure can be thus quantified as: (TVH2 - TDR,VH)/(TVH2 - TVH1) * 100%, which is 100% at full failure and 0% with no failure. e, Average spike conduction failure to the VH in proprioceptive axons, and decrease following a DR conditioning stimulation that depolarized the axon (PAD). Box plots of failure estimated as in (b-d). Prior DR conditioning to produce PAD (via adjacent S4 or Ca1 DR stimulation at 3xT) reduced the failure estimated 20 ms later by the paired-pulse conduction testing (repeating DR–VH stimulations of b-d). DR conditioning itself lowered the thresholds for VH activation (n = 5 rats), as previously reported32. We studied two lengths of axons: long axons (intersegmental, n = 11 axons, from 5 rats) with the VH stimulation one segment away from the recording site, and short axons (segmental, n = 12, from 5 rats) with the VH stimulation near the recording site, in the same segment. Box plots show the interquartile range (box), median (thin line), mean (thick line), extremes, 10 and 90 percentiles (whiskers). + significantly less failure with PAD and * significantly less failure with short compared to long axons, two-sided paired and unpaired t-tests, respectively, P < 0.05.
a, Experimental setup to directly record spike conduction failure in proprioceptive axon terminals in the ventral horn (VH) following DR stimulation. Extracellular (EC) recordings from axon terminals in VH, with glass electrode positioned just outside these axons, and for comparison EC recording in the dorsal horn (DH). b, EC field recorded in VH after DR stimulation (S4 DR, 1.1xT; T: afferent volley threshold), with a relatively large initial positive field (magenta arrows, pf) resulting from passively conducted axial current from sodium spikes at distant nodes (closer to the DR; outward current at electrode), some of which fail to propagate spikes to the VH recording site; thus, this field is a measure of conduction failure, as demonstrated in (c-f) below. Following this, a negative field arises (blue arrow, nf), resulting from spikes arising at nodes near the electrode (inward current); thus, this field is a measure of secure conduction. Reducing conduction failure by depolarizing the axon (+PAD) with a prior conditioning stimulation of an adjacent DR (Ca1, 2xT, 30 ms prior), decreased the positive field (pf) and increased the negative field (nf), consistent with increased conduction to the terminals, and in retrospect the same as Sypert et al. (1980)120 saw in cat (their Fig. 4). Large stimulus artifacts prior to these fields are truncated. c, Control recordings from proprioceptive axons in dorsal horn (DH) to confirm the relation of the EC negative field (nf) to spike conduction. Intracellular (IC) recording from axon (sacral S4, resting at -64 mV) and EC recording just outside the same axon, showing the DR evoked spike (IC) arriving at about the time of the negative EC field (nf). There is likely little spike failure in this axon or nearby axons, due to the very small initial positive field (pf). EC fields are larger in DH compared to VH (G, 10x), and thus the artifact is relatively smaller. d, Locally blocking nodes with TTX to confirm the relation of the positive EC field to spike failure. EC recording from proprioceptive axon in the dorsal horn (S4), with an initial positive field (pf) followed by a negative field (nf), indicative of mixed failure and conduction. A local puff of TTX (10 µl of 100 µM) on the DR just adjacent to the recording site to transiently block DR conduction eliminated the negative field (nf) and broadened the positive field (pf), consistent with distal nodes upstream of the TTX block generating the positive field via passive axial current conduction, and closer nodes not spiking. Recordings were in the presence of synaptic blockade (with glutamate receptor blockers, kynurenic acid, CNQX and APV, at doses of 1000, 100 and 50 µM respectively), to prevent TTX spillover having an indirect action by blocking neuronal circuit activity, including GABAaxo neuron activity. This synaptic blockade itself contributed to some spike failure, consistent with a block of GABAaxo neuron activity, as there was a more prominent positive field (pf) compared to without blockade in (c). e, EC field recorded from terminals of proprioceptive axons in the ventral horn near motor neurons (S4), in the presence of an excitatory synaptic block that largely eliminates most neuronal circuit behaviour (with kynurenic acid, CNQX and APV, as in d). In this synaptic block negative fields were generally absent (nf = 0), and only prominent positive fields (pf) occurred (as with TTX block), suggesting that conduction to the VH often completely failed when circuit behavior is blocked, which likely indirectly reduces GABAaxo neuronal circuit activity and its associated facilitation of nodal conduction. f, Rescue of spike conduction to the ventral horn by increasing sodium channel excitability by reducing the divalent cations Mg ++ and Ca ++ in the bath medium80. Same EC field recording as in e, but with divalent cations reduced (Mg++, 0 mM; Ca++, 0.1 mM). The positive field was largely eliminated (pf = 0) and replaced by a negative field (nf), consistent with elimination of conduction failure, and proving that the positive field is not a trivial property of axon terminals78,79,120. g, Conduction index computed from positive (pf) and negative (nf) field amplitudes as: nf / (nf + pf) x 100%, which approaches 100% for full conduction (pf ~0; as in c) and 0% for no conduction (nf = 0; as in e). h-i, Summary of conduction index estimated from EC field potentials, shown with box plots. Without drugs present in the recording chamber, the axon conduction from the DR to the dorsal horn was about 70% (h, n = 17 axon fields, from 10 rats), consistent with Fig. 2, whereas conduction from the DR to the VH was only about 50% (i, n = 11, from 5 rats), suggesting substantial failure at the many branch points in the axon projections from the dorsal horn to the motor neurons. Increasing GABAaxo neuron activity with DR conditioning (PAD, 30–60 ms prior) increased conduction (+GABA, in both the DH and VH, n = 5 and 9 from 5 rats, as in b), whereas decreasing GABA and all circuit activity in a synaptic blockade decreased conduction (-GABA, in both DH and VH, n = 5 and 6 from 5 rats, as in d-e). TTX (n = 5 from 2 rats, h) or removal of divalent cations (Mg++ and Ca++, -Divalent, n = 5 from 2 rats in synaptic block, i) reduced or increased conduction, respectively (as in d and f). Box plots show the interquartile range (box), median (thin line), mean (thick line), extremes, 10 and 90 percentiles (whiskers).* significant difference from control pre-drug or pre-conditioning, two-sided paired t-test, P < 0.05.
About this article
Cite this article
Hari, K., Lucas-Osma, A.M., Metz, K. et al. GABA facilitates spike propagation through branch points of sensory axons in the spinal cord. Nat Neurosci 25, 1288–1299 (2022). https://doi.org/10.1038/s41593-022-01162-x