Imaging sensory transmission and neuronal plasticity in primary sensory neurons with a positively tuned voltage indicator

Detection of somatosensory inputs requires conversion of external stimuli into electrical signals by activation of primary sensory neurons. The mechanisms by which heterogeneous primary sensory neurons encode different somatosensory inputs remains unclear. In vivo dorsal root ganglia (DRG) imaging using genetically-encoded Ca 2+ indicators (GECIs) is currently the best technique for this purpose mapping neuronal function in DRG circuits by providing an unprecedented spatial and populational resolution. It permits the simultaneous imaging of >1800 neurons/DRG in live mice. However, this approach is not ideal given that Ca 2+ is a second messenger and has inherently slow response kinetics. In contrast, genetically-encoded voltage indicators (GEVIs) have the potential to track voltage changes in multiple neurons in real time but often lack the brightness and dynamic range required for in vivo use. Here, we used soma-targeted ASAP4.4-Kv, a novel positively tuned GEVI, to dissect the temporal dynamics of noxious and non-noxious neuronal signals during mechanical, thermal, or chemical stimulation in DRG neurons of live mice. ASAP4.4-Kv is suciently bright and fast enough to optically characterize individual neuron coding dynamics. Notably, using ASAP4.4-Kv, we uncovered cell-to-cell electrical synchronization between adjacent DRG neurons and robust dynamic transformations in sensory coding following tissue injury. Finally, we found that a combination of GEVI and GECI imaging empowered in vivo optical studies of sensory signal processing and integration mechanisms with optimal spatiotemporal analysis.


Introduction
Dorsal root ganglia (DRG) neurons have pseudounipolar axons that project toward skin where they initially convert external stimuli such as touch, stretch, itch, hot, cold, and/or chemical stimuli into corresponding electrical signals. These electrical signals are integrated and modulated in the cell bodies of DRG located in intervertebral foramen between spinal vertebrae, and then action potentials containing somatosensory information are further propagated to the super cial laminae of dorsal spinal cord.
Electrophysiologic recording has been used as a fundamental tool for measuring neuronal electrical signals for many decades, but this approach is limited by the invasiveness of the procedure, poor anatomical accessibility, the absence of physiological input during commonly used in vitro recordings, and by the di culty of in vivo measurement due to stability issues 1, 2, 3 . Genetically-encoded Ca 2+ indicators (GECIs) allow for monitoring DRG neuronal ring activities, network patterns among neurons and other cell types, and sensory circuits in physiological and pathological conditions with exceptional spatial and populational resolution and limited perturbation 4 . However, Ca 2+ indicators fail to distinguish between action potential-evoked Ca 2+ in ux vs. Ca 2+ transients arising from internal stores and ligandgated Ca 2+ channels. Furthermore, Ca 2+ indicators only report suprathreshold signaling while failing to detect subthreshold membrane potential uctuations due to slow kinetics and limited sensitivity 2, 3, 5 . As an alternative, recording DRG neuronal activity using uorescence generated from genetically-encoded voltage indicators (GEVIs), which can follow not only fast suprathreshold voltage signals but also subthreshold uctuations in membrane potentials, could be an excellent complementary approach to in vivo GECI imaging.

-Kv for in vivo DRG voltage imaging
For in vivo DRG voltage imaging, we intrathecally injected adeno-associated viruses (AAVs) encoding ASAP4.4-Kv into spinal cord to allow for expression in DRG neurons. At 5-7 weeks after injection, in vivo single photon confocal imaging experiments were performed on the right lumbar (L5) DRG, which innervates parts of right hindpaw, leg, and back of the mouse. Fluorescent signals from ASAP4.4-Kv were acquired by confocal microscopy in frame mode to capture the entire population of L5 DRG neurons. We veri ed transduction of ASAP4.4-Kv virus into DRG neurons by imaging ASAP4.4 uorescence in DRG neurons. The basal ASAP4.4-Kv green uorescence intensity was relatively low under in vivo conditions; however, in ammation in hindpaw caused by complete Freund's adjuvant (CFA) injection or chronic constriction injury 22 of sciatic nerves (SNs) yielded a stronger ASAP4.4-Kv uorescent signal ( Fig. 1a and Supplementary Fig. 1a, b). The results showed that ASAP4.4-Kv can be sparsely but highly expressed in DRG neurons in vivo, and can dynamically respond to voltage in the physiological range ( Supplementary  Fig. 2). These are essential properties for carrying out the functional analysis at the cellular level in vivo.
We noticed that many DRG neurons showing brightly uorescent ASAP4.4 were in close proximity to each other, more prominent in CFA and SN-CCI animals ( Supplementary Fig. 1c), implying that "crossexcitation 23 " or "coupled activation 4 " may arise in proximal parts of the in vivo primary sensory neurons. This phenomenon led us to explore the utility of voltage indicators to reveal neuronal crosstalk within the peripheral nervous system of live animals.
Electrical coupling synchronization between adjacent DRG neurons revealed by ASAP4.4 DRG neurons are covered with satellite glial cells (SGCs), grouping with or without SGCs in between neighboring neurons 24 . Under normal circumstances, DRG neurons are loosely connected to each other or to SGCs. In contrast, extensive dye transfer and electrical coupling between adjacent neurons are often observed in many pain conditions, including in ammation in mouse hindpaw and sciatic nerve injury 25,26 , which is attributed to gap junctions present in both DRG neurons and the surrounding SGCs. As a consequence of electrical coupling by gap junction, DRG neurons exhibit coupled activation following peripheral tissue injury examined by Ca 2+ imaging in DRG of live mice 4 .
We thus attempted to simultaneously record electrical activity between two adjacent DRG neurons using ASAP4.4. We randomly selected pairs of adjacent DRG neurons in different regions of the DRG, and performed a single-line scan at about 1.1 kHz across the membrane regions of the two neuronal cells (Fig. 1a-c, images). We analyzed paired data sets from naïve, CFA, or SN-CCI animals by quanti cation of uorescence intensity changes in scanned areas of individual adjacent cells. In naïve mice, very few DRG neurons displayed rhythmic spontaneous subthreshold voltage uctuations, and no temporal cell-tocell coherence or synchronization of voltage signals in neuronal membranes was observed (Fig. 1a).
Under the context of in ammation or nerve injury, subthreshold voltage uctuations were readily detectable in vivo (Fig. 1b, c), with approximately 7-fold increase in average area under the curve (AUC) of ASAP4.4-Kv uorescent signal intensity (Fig. 1d). Strikingly, around 6% of recording neuronal pairs exhibited spontaneous suprathreshold (spiking) activity and strong coincident voltage changes in the range of ten to hundreds of milliseconds, regardless of activity patterns (Fig. 1b, c), whereas gap junction blocker, carbenoxolone (CBX), signi cantly reduced cell-to-cell electrical synchronization (Fig. 1b). Our results indicate that tissue injury increased cell-to-cell connectivity and network communication between DRG neurons leading to enhanced synchronization in DRG neuronal networks, and eventually to better integration and summation of somatosensory signals. To the best of our knowledge, such electrically synchronous neuronal events between cells in the peripheral sensory system in vivo have not been previously described.
To determine whether electrically synchronous events corresponded to global neuronal activity, we included in vivo DRG Ca 2+ imaging of neuronal populations using Pirt-GCaMP3 mice, in which the GECI GCaMP3 was exclusively expressed in primary sensory neurons under the control of the Pirt promoter 27 .
Using Pirt-GCaMP3 Ca 2+ imaging, we could simultaneously monitor neuronal activity of the entire population of DRG neurons 4 . We imaged the entire DRG at ~6.4 to 7.9 s/frame and found that spontaneous activity was rarely seen in naïve animals (1-3 neurons/DRG), but in the presence of in ammation or nerve injury, increased spontaneous neuronal activity was observed (>10 neurons/DRG) ( Supplementary Fig. 3a). This Ca 2+ activity could represent either sporadic Ca 2+ oscillations or steadystate high Ca 2+ (Supplementary Fig. 3b). To this point, however, no synchronized spontaneous activity was observed in GCaMP3 signals. In comparing voltage dynamics seen by ASAP4.4-Kv imaging with Ca 2+ signals seen by GCaMP3 imaging, we found that ASAP4.4-Kv imaging preserved fast temporal signal information, which GCaMP3 imaging failed to convey. ASAP4.4-Kv detected numerously dynamic membrane voltage signal changes associated with in ammation or nerve injury but GCaMP3 did not ( Fig. 1d and Supplementary Fig. 3c). In contrast, GCaMP3 Ca 2+ signals re ected an increasing number of spontaneously activated neurons in the entire DRG after in ammation or nerve injury ( Supplementary  Fig. 1d,e).
Noninvasive optical readout of different afferent subtypes in vivo with ASAP4.4 Primary sensory neurons diverge in function as they express their own unique receptors and ion channels. Classically, DRG neurons are categorized into three subtypes based on somatic action potential shapes and conduction velocity; namely, myelinated Aβ (large diameter and fast conducting) low thresholdmechano receptor (LTMR) and Αδ (medium diameter and medium conducting) afferent nociceptor, along with slowly conducting small diameter unmyelinated C-type nociceptor 28, 29 . Identi cation and functional characterization of different neuronal subclasses in vivo have been a challenge, thus current investigations have mostly relied on the invasive in vitro or ex vivo electrophysiologic recordings 30,31 . As shown in Fig. 1g, we were able to visualize single action potentials by inspecting small spiking area (1-3 µm) of line scan image. We found that the kinetic properties of action potentials were closely related to the sizes of DRG neurons, consistent with conventional electrophysiologic recordings 30,32 . This indicates the feasibility of in vivo voltage imaging for noninvasive optical readout of electrophysiologic features, to the point where afferent subtypes can be inferred and targeted.
ASAP4.4-Kv imaging permits visualization of mechanical stimuli (non-noxious to noxious)-evoked temporal summation of fast voltage signals To understand how DRG neurons encode painful or non-painful mechanical stimuli, we applied stimulation of different strengths to the hindpaw, and visualized evoked ASAP4.4-Kv signals in DRG neurons. At low stimulation strength (light brush, 0.4 g, or 2 g von Frey; Fig. 2), small and transient subthreshold potential changes could be observed in mechanosensitive neurons (Fig. 2a), and only a few neurons exhibited hindpaw stimulation-evoked transient Ca 2+ increases in naïve animals (Fig. 2b). However, peripheral in ammation or nerve injury led to a signi cant increase in membrane electrical signal summation, including both subthreshold and suprathreshold voltage signals (Fig. 2a, d), but not in Ca 2+ responses ( Fig. 2b-d). At an intermediate stimulation strength (100 g press), high-frequency voltage dynamics were observed in neurons of naïve mice (Fig. 3a), while in ammation or nerve injury treatment produced exacerbated voltage uctuations with larger amplitude and longer membrane depolarization ( Fig. 3b-d). On the other hand, GCaMP3 Ca 2+ imaging revealed increased population level activities in injured mice upon exposure to the same press stimulus ( Fig. 3d, e). However, large variations in the magnitude of Ca 2+ transients were found within the same DRG and across different treatment groups.
Consequently, while the data were grouped, neither average amplitudes nor the mean AUCs of Ca 2+ transients differed signi cantly between naïve or injured animals (Fig. 3d, g), despite the fact that increased activated cell numbers (Fig. 3d) and increased amplitudes of Ca 2+ transients (Fig. 3f) were evident in some CFA-injured mice.
At the strongest mechanical stimulus (300 g), long-lasting membrane potential uctuations with sustained membrane depolarization were observed in DRG neurons of naïve mice (Fig, 4a), and voltage uctuations in neuronal membranes were further aggravated by in ammation or nerve injury treatment . In addition, simultaneous in vivo dual color imaging of ASAP4.4-Kv (green) and mCyRFP3 33 , a cyan-excitable red uorescent protein that can be used as a non-perturbing voltage-independent uorescent marker as a control signal for ASAP4.4-Kv voltage imaging, demonstrated that the pattern of evoked electrical activity was distinguishable from rhythmic physiological motions arising from respiration or heartbeat ( Supplementary Fig. 4).

ASAP4.4-Kv imaging reports thermal (heat or cold)-evoked voltage signals with high temporal delity
It has been reported that primary sensory neurons employ different strategies to encode heat vs. cold 34,35 . To discern how heat or cold is represented in vivo, we examined the ASAP4.4-Kv voltage signals from heat or cold-sensing neurons. In naïve mice, the membrane voltage dynamics during noxious heat (50℃) stimulation displayed a slowly depolarizing voltage ramp that returned to baseline within 300 ms ( Fig. 5a). Noxious cold (0℃) stimulation, however, led to two distinctive forms of voltage activity: bursting or non-bursting (Fig. 6a). Bursting neurons displayed burst-frequency ring behaviors, whereas non-bursting neurons generated only single action potentials followed by small membrane uctuations (Fig. 6a). In ammation or nerve injury, in turn, resulted in augmentation of membrane voltage uctuation and electrical amplitude in both heat-and cold-sensing neurons ( After CFA-induced in ammation, numerous DRG neurons were activated upon noxious heat stimulation (50℃) but numbers were similar between naïve and CFA groups (Fig. 5d, e and Supplementary movie 8), while fewer neurons displayed Ca 2+ activity to noxious cold (0℃) compared to naïve animals (Fig. 6d, e).
As with the previous mechanical stimuli, heat-induced increases in Ca 2+ transients were observed in some DRGs of individual CFA-treated mice (Fig. 5f), but not in grouped DRGs (Fig. 5d, g). In contrast, coldsensitive neurons displayed reduced Ca 2+ transients after peripheral in ammation, both individually ( Fig. 6f) and as a group (Fig. 6d, g). These results are consistent with previous reports that cold-mediated Ca 2+ activity was lost in speci c types of cold-sensing neurons following peripheral injury 35 . The discrepancy between voltage and Ca 2+ signals in cold-sensing neurons suggests that, following peripheral in ammation, an individual sensory neuron still retains the ability to encode cold-speci c sensory input; however, summation of the neuronal response to painful cold is suppressed by network activity in DRG.
ASAP4.4-Kv imaging reveals high potassium or capsaicin-evoked strong membrane voltage uctuations Finally, we used the ASAP4.4-Kv voltage sensor to examine how DRG neurons encode noxious chemical nociception. In naïve mice, direct topical application of high potassium (50 mM KCl) or capsaicin (10 µM), a TRPV1 agonist which can initiate activity in nociceptive neurons, onto L5 DRG, resulted in >4-fold increase in voltage uctuations over baseline (Fig. 7a, d). Both CFA and SN-CCI treatments signi cantly increased neuronal responses to KCl or capsaicin, with substantial increases in frequency and magnitude of dynamic membrane voltage uctuations (Fig. 7b-d and Supplementary movie 5, 6). When the same chemical treatments were performed on Pirt-GCaMP3 mice, we observed robust activation of a large population of DRG neurons within the DRG sensory ganglia (Fig. 7f, g). Topical application of capsaicin resulted in DRG neuronal activation primarily in the small and medium diameter neurons within all populations of DRG neurons imaged (Fig. 7g). Small and medium diameter neurons are nociceptors that typically express TRPV1 receptors. On average, the Ca 2+ transients in activated neurons from injured mice were signi cantly higher than those from naïve animals (Fig. 7e). Compared to physical stimulation, direct chemical administration onto DRG neurons produced near-maximal Ca 2+ transients and responses in most DRG neurons in vivo. These ndings lead us to conclude that neuronal hypersensitivity is a common consequence of peripheral injury.

Discussion
GEVIs have been successfully used in analysis of brain regions in vivo in awake behaving mice 12,13,14,15,16,17 , and have encouraged neuroscientists to explore and unlock the full potential of the technological advances. Our current study reports the use of an improved ASAP-family GEVI, ASAP4.4-Kv, to track both spontaneous and evoked voltage activity of mouse primary sensory neurons in vivo. The ASAP4.4-Kv voltage sensor allows direct visualization of distinct temporal features of neuronal dynamics, subcellular voltage dynamics, plasticity induction, and neuronal coding in DRG, the analysis of which have been largely inaccessible and technically challenging in live animals. The ASAP4.4-Kv voltage sensor provides the means for understanding how primary sensory neurons, especially DRG neurons, function or fail to function (changes in dynamics, status, and/or pattern) at any given time under physiological and pathological conditions. Particularly attractive for DRG voltage imaging is the presented ASAP4-Kv with fast kinetics enables optical detection of single action potentials from individual DRG neurons, which allows noninvasive identi cation of somatosensory neuron subtypes in vivo without the aid of conventional invasive electrophysiological recordings.
GEVI imaging as a powerful tool complementary to GECI imaging ASAP3-Kv, a previous ASAP family GEVI with desirable responsivity and SNR for in vivo use, has a negative slope relationship between voltage and uorescence 15,16 . But, new ASAP4.4-Kv produces a depolarization-dependent increase in uorescence intensity 17 , as do most GCaMP GECIs 4,27 . This new feature, combined with somatic targeting and optimal brightness, greatly improves the utility of the GEVI ASAP4.4-Kv for in vivo recording of neuronal electrical signals and activity. In our in vivo DRG voltage sensor imaging studies, we could visualize sparse signals from the membrane surface of neuronal soma expressing uorescent ASAP4.4-Kv with imaging depths of <20 µm below the meninges membrane. We demonstrated superiority of in vivo GEVI recording for resolving long-standing debates over the temporal attributes of neuronal coding by direct comparison with in vivo GECI imaging modalities previously developed by our lab 4 . Compared to GECIs, the main advantage of GEVIs is their ability to examine nonspiking (subthreshold) and/or spiking (action potential) electrical activity, because subthreshold membrane potential uctuation and oscillation do not greatly affect internal Ca 2+ levels or dynamics. ASAP4.4-Kv is able to identify non-spiking subthreshold voltage uctuation events in DRG neurons with duration times in the millisecond range, an order of magnitude faster than the signal integration time of the Ca 2+ indicator GCaMP3 or other advanced GCaMPs.
Unexpectedly, spontaneous neuronal dynamics revealed by ASAP4.4-Kv exhibited cell-to-cell coupled synchronous electrical events following injury that were normally indiscernible in GCaMP3 imaging across all DRG neurons. Highlighting the importance of in vivo voltage imaging, ASAP4.4-Kv imaging fully unmasked altered DRG neuronal electrical activities and signals resulting from peripheral injury at the single-cell level in their native environment. In contrast, variable Ca 2+ activities could be observed in in vivo Prit-GCaMP3 Ca 2+ imaging, but signi cant effects were diminished in comparisons of larger numbers of DRG neurons from multiple animals. Another fascinating aspect of uorescence voltage sensors is their ability to map not only subthreshold depolarizing (excitatory) inputs but also hyperpolarizing (inhibitory) events that occur constantly in almost all neurons 36 . Previous studies have used voltage indicators to monitor membrane hyperpolarization in cultured neurons 37 , brain slices 38 , or in freely moving mice 39 . The hyperpolarizing voltage signals detected by ASAP4.4-Kv in our DRG recordings were evident in response to strong stimuli, and were detected as uorescence intensity dropped to levels below the pre-stimulation baseline. In contrast, GECI imaging can only detect excitatory inputs, and lacks ability to detect signals related to inhibitory inputs. Thus, a plausible use of GEVI imaging would be to examine possible inhibitory signaling involved in controlling peripheral nociceptive or non-nociceptive transmission, or to examine excitatory or/and inhibitory signal summation and integration within sensory ganglia.
Given the heterogeneity of sensory neurons, integrated signals from large-scale DRG neurons need to be collected at high spatiotemporal resolution to establish cell-type or modality-speci c coding strategies. The intrinsic slow kinetics of GECIs permits mapping large numbers of neuronal assemblies in their native environments with conventional confocal microscopic approaches. Intensive studies using GECIs have characterized sensory coding of heat or cold 34,35,40,41 , mechanical 40,42,43 , or chemical stimuli 44 in health and disease conditions. Due to the intrinsic fast kinetics of GEVIs, simultaneous imaging of dozens, hundreds, or thousands of DRG neurons at high spatial (millimeters) and temporal (milliseconds) resolution is an extremely challenging task and limited by current technological advances. To overcome these limitations, we adapted a conventional, low cost, strong laser intensity, upright laser-scanning confocal microscope to be used as a versatile platform, allowing optical reporting of dynamic neuronal activity in DRG neurons at high spatial and temporal resolution by combining GECI-based Ca 2+ signals and GEVI-based voltage signals. With the continuous advances in voltage indicators and optical instruments, simultaneous voltage recording of enormous numbers of neurons in live intact peripheral tissues will enable dissecting functional connectivity in DRG circuits and mapping neuronal coding strategies with better high-throughput and greater accuracy.

Future Outlook
We previously described the phenomenon of coupled neuronal activation within DRG in mouse models of in ammatory or neuropathic pain, which is attributed to an injury-induced increase of gap juntions 4 . Here, our in vivo voltage imaging results provide evidence of gap-junction-mediated electrically synchronous neuronal activity between DRG neurons, further supporting increased neuronal 'cross talk' by gap junctions as an underlying mechanism in the development of hyperalgesia and allodynia. In our ongoing work, we aim to elucidate the pathway and mechanism of neuron-to-neuron transmission.
Our current implementation combining both GEVI and GECI imaging should allow detailed investigations of the relationship between suprathreshold somatic voltage signals and the corresponding Ca 2+ dynamics at single-cell-, population-, or modality-speci c levels. Simultaneous sub-millisecond voltage and Ca 2+ imaging using a voltage-sensitive dye and GECI has been performed on Purkinje neurons in awake animals, and has demonstrated high spatiotemporal variations of suprathreshold voltage signals and Ca 2+ transients between dendritic segments 45 . For in vivo primary sensory neuron studies, dual GEVI and GECI neuronal labeling with different uorescence spectra will help in the analysis of correlation or integration between suprathreshold voltage signals and the resulting Ca 2+ transients at high spatiotemporal resolution.

Limitations Of This Study
As a latest ASAP-family sensor, positively tuned ASAP4.4-Kv with optimal uorescence response and SNR enables identifying and tracking each suprathreshold voltage spike during the designated time in our in vivo DRG recording. In two-photon optical imaging in deep layers of in vivo brain, a previously used ASAPfamily GEVI, ASAP3, reliably detected single spikes and resolved spikes in bursts, with appreciable optical spike amplitudes 15 . However, in our in vivo DRG studies, each optical trace displayed in response to different stimuli is a spatial average of voltage signals from the entire line scan encompassing surface region of 16 µm × 0.5 µm. Spatial averaging can greatly improve the SNR in optical measurements and captures more dynamic signals occurring on the cell membrane, while failing in resolving closely spaced spikes. A few spikes were detectable from a close examination of spiking activity in a small recording region, from their characteristics we were able to identify speci c cell types of primary sensory neurons and their physiological ring patterns in intact living tissues. In conclusion, ASAP4.4-Kv voltage imaging opens new avenues to explore the basic principles of DRG neuron coding, and of the cellular basis for perceptual changes in somatosensation by providing high temporal resolution of individual neurons. The combination of GEVI and GECI imaging allows a more temporally and spatially precise characterization of the neuronal coding and integration strategies in the peripheral somatosensory system.

Animal models
All experiments were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee at University of Texas Health Science Center at San Antonio (UTHSA). C57BL/6J mice (body weight 20-30 g) were obtained and bred in-house. Animals were group housed unless otherwise noted, provided with food and water ad libitum, and kept on a 14/10 light/dark cycle at 23°C. To generate CFA in ammatory injury mice, we made a 1:1 mixture of complete Freund's adjuvant (CFA): saline, and injected 50 μL subcutaneously into the glabrous skin of the hindpaw. In vivo imaging was performed 1-3 days following CFA injection. To generate sciatic nerve (SN) chronic constriction injury 22 , mice were anesthetized by intraperitoneal (i.p.) injection of ketamine/xylazine (0.1/0.015 mg/g body weight). SN was exposed mid-thigh by a small incision and separated from surrounding tissue. Ligatures were loosely tied using 3-0 silk thread around SN. The incision was closed using sutures, and mice were used for in vivo imaging 7-10 days later.
Pirt-GCaMP3 mice were generated and described as done in a previous study 4,25 . Brie y, transgenic animals were generated by targeted homologous recombination to replace the entire coding region of the Pirt gene with the GCaMP3 sequence in frame with the Pirt promoter.

ASAP4.4-Kv virus delivery
AAV8-hSyn-ASAP4.4-Kv and AAV8-hSyn-ASAP4.4-Kv-mCyRFP3.WPRE 17 were generated by the Stanford Viral Core. For the experiments in cell culture, male or female 2 to 3 weeks-old mice were used for intrathecal delivery of the virus to DRG neurons. For in vivo imaging, 2 to 4 months old mice from both sexes were used for intrathecal delivery of virus to peripheral neurons. For intrathecal delivery, mice from both sexes were rst anesthetized with iso urane, shaved and disinfected. The AAVs were diluted in sterile, isotonic saline. A volume of 30 µl containing 2×10 12 virus particles/ml was injected intrathecally (i.t.) by direct lumbar puncture using a 28½-gauge needle and insulin syringe (Becton Dickinson, Franklin Lakes, NJ). A re exive ick of the tail indicated proper needle entry location for intrathecal injection. Following the injection, the animals were returned to recovery cages where they remained for 5-7 weeks until imagining or electrophysiology experiments were performed. DRG exposure surgery DRG exposure surgery was carried out as previously described 4 . Brie y, mice were anesthetized with i.p. injection of ketamine/xylazine (0.1/0.015 mg/g body weight). Mice were kept on a heating pad to maintain body temperature at 37±0.5℃, which was monitored by a rectal probe. Their backs were shaved, and ophthalmic ointment was applied to their eyes to prevent drying. The transverse processes of lumbar L5 were exposed, and the surface aspect of the bone covering the DRG was carefully removed to expose the underlying DRG without damaging the DRG or spinal cord. Bleeding was gently stopped using styptic cotton or gel foam.

In vivo imaging
For in vivo imaging of the whole L5 DRG, mice were placed on a custom-built tilted stage and their spines were secured with custom-built clamps to minimize movement due to breathing and heartbeat. The stage was a xed under a LSM 800 confocal laser-scanning microscope (Carl Zeiss, Inc) equipped with upright 5×, 10×, and 40× objectives. The iso urane-anesthetized animals (1%-2%, vol/vol in 100% O 2 ) were maintained at 37±0.5℃ by a heating pad during the imaging process. Z-stack imaging, which can cover the entire L5 DRG, was typically acquired at eight to ten frames using a 10× C Epiplan-Apochromat objective (0.4-NA, 5.4-mm free working distance, Carl Zeiss) at typically 512×512 pixel resolution with lasers tuned at 488 nm and at 561 nm and emission at 500-550 nm for green and 620-700 nm for red uorescence. DRG neurons were at the focal plane, and imaging was monitored during the activation of were typically discarded due to photobleaching, and uorescent traces with strong motion artifacts were also excluded in the analysis.

Stimulus delivery during imaging experiments
Mechanical or thermal stimuli were applied on the ipsilateral hindpaw in the following order: brush, 0.4 g von Frey, 2 g von Frey, 100 g press, 300 g press, heat (50°C), or cold (0°C). Paw stimuli with 100 g or 300 g press force were delivered using a rodent pincher analgesia meter, and press force was controlled manually by the experimenter. For the ASAP4.4 imaging, the duration of the external stimuli was 4-5 s Fluorescence traces were acquired with cells using whole-cell voltage-clamp mode.
Step voltage was applied to change the membrane potential from a holding voltage of -70 mV to command voltages at -100, -40, +30, or +100 mV in a series of subsequent steps for 0.5-1 s. ASAP4.4-Kv expressing DRG neurons were imaged on an upright Zeiss Examiner.A1 microscope tted with a 40× water-immersion objective (0.75-NA, 2.1-mm free working distance, Carl Zeiss) and with an Axiocam 705 color camera (Carl Zeiss). Images were sampled at 5 Hz.

In vivo imaging data analysis
To analyze confocal line-scan imaging of ASAP4.4, uorescence imaging data were extracted from raw image data, and time-dependent uorescence traces for each neuron were revealed using Mean ROI function in Zen blue software. Because presentation of peripheral stimuli evoked spatially differentiated, large optical signals that were distinguishable from the stimulus-independent component, we averaged the rst 1-2 s before the stimulus onset and designated that as the baseline uorescence (F 0 ). Baseline-normalized amplitudes in the region of interest (ROI) over time were expressed as (F−F 0 )/F 0 ×100% against time. Some experiments were excluded if rundown exceeded 30%.
For GCaMP3 imaging data analysis, individual responding neurons were veri ed by visual examination and con rmed when the uorescent intensity of ROI during stimulus was 15% higher than baseline signals using the Mean ROI function in Zen blue software. Time series recorded uorescence changes were exported to Excel, and were analyzed using GraphPad prism. The average uorescence intensity in the baseline period was taken as F 0 , and was measured as the average pixel intensity during the rst two to ve frames of each imaging experiment. Relative change in uorescence intensity was measured using the formula ΔF/F 0 (%) = (F−F 0 )/F 0 ×100%.

Statistical methods
Group data were expressed as mean ± standard error of the mean (S.E.M.). Student's unpaired t-tests, Mann-Whitney U-tests, one-way ANOVA with a post-hoc Dunnett's t-test, or Kruskal-Wallis test, as appropriate, were employed for comparisons. A two-tailed p value <0.05 was considered statistically signi cant for all analyses. All statistical tests are indicated in gure legends.   In vivo optical recording of mild press (100 g)-induced neuronal activity in intact DRG neurons. a-c

Declarations
Optical voltage recordings of primary sensory neurons in response to a single mechanical force (100 g)