Direct voltage imaging of peripheral axons in vivo reveals ongoing nociceptor firing in healthy mice

Using genetically encoded voltage indicators and holographically targeted widefield imaging, this paper records membrane voltage directly from the terminal branches of pain-sensing neurons in the cornea of living mice — and finds that a substantial fraction of these terminals fire action-potential-like spikes continuously, with no stimulus, in healthy tissue. The firing is generated peripherally (it survives silencing the parent ganglion with lidocaine), is synchronized across sister branches, and depends on NaV1.8 sodium channels and HCN/Ih conductances. If it holds up, it changes the baseline picture of what a "resting" nociceptor is doing, and it opens the axon — the part of the neuron that actually carries the output — to direct optical electrophysiology in vivo.

One caveat before we start: only the abstract of this preprint was available to me, so everything below is built on that plus necessary background. Where I reason beyond the abstract, I'll flag it.

The blind spot: we've never really watched an axon work in a living animal

Think about what we can actually measure in an intact nervous system. Patch-clamp electrodes give you exquisite voltage recordings, but only from structures big enough to seal a pipette onto — cell bodies, occasionally thick dendrites. Extracellular electrodes pick up spikes near somata. Calcium imaging is the workhorse of systems neuroscience, but calcium is a slow, lossy proxy for voltage: it smears a 1 ms spike into a ~100 ms transient, misses subthreshold dynamics entirely, and in fine axonal compartments the relationship between calcium influx and membrane potential is murky.

Meanwhile, the axon — especially its distal branches and terminals — is where much of the interesting electrical business happens: spike initiation, conduction failures at branch points, integration of local inputs. For most neurons we've simply assumed the soma is a reasonable readout. For nociceptors, that assumption is not merely convenient; it's structurally wrong.

A nociceptor is a bizarre neuron. Its cell body sits in a sensory ganglion (here, the trigeminal ganglion, tucked at the base of the skull) and does essentially no sensing. A single axon leaves the soma and splits: one branch runs to the spinal cord or brainstem, the other runs out to the periphery — in this case, the cornea — where it ends in a sprawling arbor of bare nerve terminals. Those terminals are the actual sensor: they carry TRPV1 channels (capsaicin/heat), TRPM8 (menthol/cold), and the voltage-gated machinery that converts a noxious stimulus into a spike train. The action potential is born in the terminal, centimeters away from the soma, in branches well under a micron thick.

So every time you record from a nociceptor's soma — which is what nearly all pain electrophysiology does, because it's what's accessible — you're listening to a telephone line and inferring what happened at the microphone. Conduction failures, terminal-local dynamics, sub-threshold events: all invisible. The field has known this is a problem for decades. Nobody could do anything about it, because you cannot patch a 0.5 µm free nerve ending in a living animal.

The key move: make the terminal report its own voltage, and only illuminate the terminal

Genetically encoded voltage indicators (GEVIs) are fluorescent proteins whose brightness tracks membrane potential, fast enough to resolve individual action potentials. Over the last decade they've matured to the point where you can image spikes from cortical somata in behaving mice. But GEVIs come with brutal constraints that get worse for axons:

• The signal lives only in the membrane, so total photon counts scale with membrane area. A soma is a big bag of membrane; a terminal branch is a thread.
• You need kilohertz-class frame rates to catch 1 ms spikes, so each frame collects very few photons.
• Widefield illumination excites every labeled structure in the volume, drowning your thin axon in out-of-focus background and burning the tissue.

The authors' solution combines three things. First, express the GEVI in trigeminal neurons, so the corneal terminals and their parent axons are labeled. Second, exploit the cornea itself: it is transparent, avascular, thin, and optically accessible from outside the animal — arguably the single best window onto peripheral nerve terminals that exists in a mammal. No cranial window, no implanted lens; the tissue is a built-in coverslip. Third — and this is the enabling trick — use holographic targeted illumination: a spatial light modulator sculpts the excitation light so that photons land only on the specific terminal branches and axon segments you want to record, not the surrounding tissue. Background fluorescence collapses, photobleaching and phototoxicity drop, and the shot-noise-limited SNR of a hair-thin structure becomes workable at high speed. High-speed widefield detection then reads out voltage from many points along the arbor simultaneously — something no electrode arrangement could ever do.

What they saw

Two classes of result. The first is a validation-and-then-some: individual terminals depolarized and fired trains of spikes in response to capsaicin (the TRPV1 agonist, "hot chili" channel) and menthol (TRPM8, "cool mint" channel). That's the expected physiology, now observed directly at the terminal with millisecond voltage resolution rather than inferred from calcium or somatic recording. Seeing capsaicin- and menthol-evoked spike trains born in the terminal itself is already a first.

The second result is the surprise. A substantial fraction of terminals fired ongoing, action-potential-like spikes with no stimulus applied at all, in healthy corneas. Spontaneous nociceptor activity is normally associated with injury, inflammation, or neuropathic pain — it's a hallmark of pathology. Finding it as a baseline feature of intact tissue is the kind of observation that makes you re-examine what "resting" means for a pain fiber.

The abstract lays out a clean chain of controls establishing that this is real, sodium-driven, peripherally generated physiology:

• It's spiking, not indicator noise or a slow artifact: the ongoing activity is sodium-channel-dependent. Block the sodium channels and the events go away, which is the defining signature of an action potential, not a fluorescence fluctuation.
• It's coherent across the arbor: firing is synchronized across sister terminal branches and shares its dynamics with the parent axon. A GEVI artifact or illumination flicker would not respect the neuron's branching topology; a propagating spike would. This spatial-coordination argument is, to me, the strongest single piece of evidence that they're watching genuine electrical events travel through a real axonal tree.
• It's born in the periphery: delivering lidocaine (a sodium-channel blocker / local anesthetic) to the trigeminal ganglion did not stop the terminal firing. If the spikes originated at or central to the soma and merely propagated outward, silencing the ganglion would kill them. They persist — so the pacemaker is distal to the ganglion, in the peripheral arbor itself.
• It has a specific molecular basis: the firing was abolished by inhibiting either NaV1.8 or HCN/Ih. NaV1.8 is a nociceptor-enriched, TTX-resistant sodium channel long implicated in repetitive and ectopic firing. HCN channels carry I_h, the hyperpolarization-activated "funny" current that drives rhythmic pacemaking in the heart and in thalamic neurons. Their joint necessity paints a picture of the terminal as a local oscillator: I_h providing the slow depolarizing drive back toward threshold, NaV1.8 firing the spikes.

That combination — a peripheral pacemaker built from HCN plus NaV1.8, read out optically at the terminal — is a concrete, mechanistically satisfying story, not just a phenomenon.

Why this matters

Two reasons, on two timescales.

The immediate scientific payload: ongoing nociceptor firing in healthy tissue reframes the baseline. If terminals are tonically active, then pain and sensitization may be modulations of an ongoing rhythm rather than the switching-on of a silent line. That reframing touches how we think about spontaneous pain, about the "irritable nociceptor" phenotype, and potentially about NaV1.8 and HCN as analgesic targets — both of which are already of intense pharmaceutical interest (NaV1.8 blockers have reached the clinic for acute pain). A local peripheral oscillator dependent on these two conductances is a testable, druggable hypothesis.

The methodological payload is arguably bigger and broader than pain. The generic problem the paper solves — recording membrane voltage from thin axonal compartments in intact tissue — has blocked whole classes of questions across neuroscience. Where does a spike initiate in a branching arbor? Do branch points fail selectively, gating which downstream targets get the signal? How do terminals integrate local input? Targeted-illumination GEVI imaging of identified axons in vivo is a general instrument for those questions, and the cornea is a particularly clean proving ground before the harder cases (central axons, peripheral fibers in opaque tissue).

What to scrutinize

The reviewer's instinct here is exactly right, and it centers on one worry: is the "ongoing firing" real physiology or an artifact of the measurement itself? GEVI overexpression can alter membrane properties and excitability; sustained illumination can heat and depolarize tissue and even photo-activate channels; the cornea is being surgically/optically manipulated. Any of these could manufacture spontaneous-looking activity.

The paper's controls address this better than most. The sodium-dependence, the NaV1.8/HCN pharmacology, and especially the synchrony across sister branches and shared dynamics with the parent axon are hard to fake with an optical or expression artifact — a topology-respecting propagating event is a strong positive signature. The lidocaine-at-the-ganglion experiment localizes the generator without appealing to the imaging at all. Still, from the abstract alone I can't judge several things that matter: how many terminals were imaged and what fraction "substantial" really means; whether ongoing firing rate correlates with illumination intensity or GEVI expression level (the cleanest artifact control); whether the corneal preparation is mechanically or thermally perturbed in a way that itself drives firing; and whether GEVI-expressing animals show any behavioral or physiological signs of altered corneal sensitivity. Comparison against a non-perturbing readout — even careful calcium imaging or single-unit recordings in the same preparation — would strengthen the claim that healthy corneas fire spontaneously in the un-imaged state.

The section to read first, when the full text is available, is the one describing the ongoing-firing controls — the illumination-intensity and expression-level dependence, the counts and fractions, and the lidocaine and NaV1.8/HCN pharmacology. That's where the surprising claim lives or dies. The methods on holographic targeted illumination and the achieved SNR at kHz rates are the second priority, because that's the transferable instrument that could reshape how the field studies axons.

Bottom line: a genuine measurement advance (direct in-vivo voltage imaging of identified peripheral axons) that immediately produced a counterintuitive, mechanistically specific finding (a peripheral HCN/NaV1.8 pacemaker driving spontaneous nociceptor firing in healthy tissue). Both halves are worth your attention; both deserve the skeptical read the authors seem to have anticipated.