A microfluidic model of the first sensory synapse for analgesic target discovery

Abstract

The synaptic connections between dorsal root ganglia (DRG) and dorsal horn (DH) neurons are a crucial relay point for the transmission of painful stimuli. To delineate how synaptic plasticity may modulate the excitability of DH neurons, we have devised a microfluidic co-culture model that recapitulates the first sensory synapse using postnatal mouse sensory neurons. We show that DRG-DH co-cultures characterize salient features of the in vivo physiology of sensory neurons. Immunocytcochemical experiments of the cultured DH neurons show a co-localization of Map2 with VGlut2 and of Map2 with Synapsin 1, corroborating the glutamatergic identity of the DH neurons and further suggesting the potential formation of active synapses in this neuronal set. Fluorometric imaging experiments demonstrate the elicitation of calcium responses in DH neurons following the stimulation of DRG cell bodies or axons. Selective NMDA and AMPA receptor blockade appreciably silences DH neuron responses, suggesting that glutamatergic signaling is maintained in vitro. Last, a surrogate model of peripheral nerve injury is introduced in the form of an axotomy, which results in elevated and prolonged calcium responses of DH neurons. Overall, the microfluidic mouse co-cultures provide a method advancement in the study of periphery-to-center pain signaling, where the potential of utilizing the platform for drug target identification is underscored.

Introduction

The pain signalling pathway has established that nociceptors detect and process distinct noxious stimuli.^1,2 The distal axonal projections of nociceptors innervate peripheral tissues such as the skin, visceral organs and muscles. On the opposite end of the bifurcating T-shaped junction, the axonal compartments of nociceptors extend towards the spinal column and innervate distinct laminae of the spinal cord’s dorsal horn.^3,4 Specialized transducer ion channels expressed in nociceptors convert both endogenous and exogenous noxious stimuli of chemical, heat and mechanical nature into electrical signals. ⁵ The evoked potentials follow a propagation trajectory across nociceptors’ long axons and cell bodies and ultimately terminate their peripheral course via synaptic transmission to specific DH laminae. It is challenging to discern how such stimuli contribute to pain sensation in acute and pathological states while aided by immune, glial and other non-neuronal cells. It is crucial to devise optimal strategies for disrupting aberrant painful signalling and provide relief.

The properties of peripheral axon terminals or synapses between the DRG and DH neurons have been less studied than the cell bodies within the DRG. The axonal terminals of sensory neurons are responsible for the initiation and propagation of action potentials (APs) and most often act the original sites of injury. Upon injury, damaged axons upregulate injury-induced gene expression programs. ⁶ These axons may exert local mRNA expression, including the de novo expression of cytoskeletal and immune related molecules as well as second messengers and ion channels.^7,8 Such newly adapted properties of sensory axons may modulate the DH neurons which receive synaptic input from them. Hence, the in vitro recapitulation of the first sensory synapse is important for studying interactions between sensory and DH neurons.

Previously, embryonic rat explants were used to study DRG and DH synaptic properties.^9–11 A micro-island assay was developed to study synapse formation in a precise 1:1 manner between DRG and DH neurons.^12,13 The DRG-DH explant studies are amenable to intracellular recording techniques but are unable to record the excitability of several DH neurons at once. Further limitations include the monitoring of the timing of synapse formation and the potential multi-axonic input that each DH neuron receives in the non-isolated presence of several sensory neurons. No previous techniques have managed to perform large scale, calcium-imaging recordings of several DH neurons with spatially precise pharmacological manipulations in mice.

Technical progress in the field of microfluidics has paved avenues into studying the axonal properties of nociceptors and transmission mechanisms across the first sensory synapse in a spatially precise manner.^14–17 This study regards the establishment of a microfluidic model between DRG and DH postnatal mouse neurons to recapitulate the first sensory synapse in culture. The model introduces the possibility of conducting both pharmacological and genetic manipulations using postnatal mice. A further strength of the model is the isolation of each distinct anatomical compartment to determine respective spatial and functional properties under physiological and disease states. ¹⁶ Combining microfluidic cultures with fluorometric imaging, we can monitor the excitability of multiple DH cell bodies following stimulation of the distal axonal DRG terminals or the DRG cell bodies. As a surrogate model of nerve injury, we finally incorporate the axotomy of peripheral terminals, which is used thereafter to study its effects on the neuronal excitability of DH neurons.

Results

Microfluidic co-cultures of mouse sensory and DH neurons

We developed a microfluidic co-culture system to model synaptic transmission between postnatal mouse DRG and DH neurons. The microfluidic devices were comprised of three compartments (Figure 1(a)): DRG and DH neurons were plated to the middle and side compartments respectively (Figure 1(b)). Sensory neurons showed bidirectional neurite sprouting to both side compartments, henceforth referred to as the peripheral and DH compartments respectively (Figure 1(b)). The sprouted neurites of the DRGs from the middle compartment crossed through 150 μm device corridors towards the peripheral and DH compartments, where the cell bodies of DH neurons were cultured (Figure 1(c)). The DRG cell bodies in the middle compartment sprouted multiple axons towards both sides of the MFC devices (Figure 1(c)). Using this configuration, our aim was to combine electrical current stimulation of DRG cell bodies or their axons in the DRG compartments and measure the ensued activation of the neurons using calcium imaging at the DH neuronal cell bodies as a functional readout (Figure 1(a)). The mitotic inhibitor cytosine arabinoside (AraC, 2 μM) was applied consistently to prevent the proliferation of non-neuronal cells which may contribute to the excitability of either compartment.

Figure 1.

Microfluidic co-culture of DRG and DH neurons. (a) Image of a 3-compartment microfluidic device attached to a 40 mm glass plate. (b) Schematic representation of the co-culture setup. Mouse DRG and spinal cord DH neurons, both harvested from the same mice at P2-P4, were loaded in different compartments of the microfluidic devices and allowed to mature for 8–12 DIV. Peripherally induced electrical stimuli were used to stimulate the axons which in turn triggered activation of neurons in the DH compartment. (c) Illustration with inset fluorescent images of the microfluidic corridors depicting neurites that have crossed from the middle to the peripheral DRG compartment (left), a magnified cluster of DRG cell bodies (middle) and a network between DH neuronal cell bodies (right).

We next sought to determine the percentage of DRG neurons with neurites that traversed bidirectionally to both the peripheral and DH compartments. The lipophilic dyes DiI and DiO were applied at the peripheral and DH compartments, respectively. The dyes were taken up by the sensory neurites and retrogradely stained the corresponding DRG cell bodies in the middle compartment (Figure 2(a)). We found that 43.6% ± 2.8% of stained DRG cell bodies were positive for DiI, hence showing that nearly half of the sprouted axons cross towards the periphery (Figure 2(b)). On the other side, 56.4% ± 4.9% of DRG cell bodies were DiO positive and their correspondent axons crossed towards the DH compartment (Figure 2(c)). The overlapping DRG cell bodies stained by both DiI or DiO, were 27.5% ± 1.2% (Figure 2(d)). Thus we estimated that about 1/3^rd of the DiI/DiO-stained cell bodies crossed towards both the peripheral and DH compartments. These bilaterally crossing neurons may tentatively recapitulate the connectivity between sensory and DH neurons.

Figure 2.

Quantification of DRG axons crossing to the periphery and DH compartments. DiI was applied to the peripheral axonal compartment and DiO was applied to the DH compartment. The dyes retrogradely traversed from the axons to the cell bodies of DRG neurons and the percentage of peripherally crossed, DH crossed, and bilaterally crossed axons was quantified in the same field of view (n = 3 devices from three cultures). (a) Schematic of the staining experiment. (b) DiI staining of DRG cell bodies after application of the dye in peripheral axons. Of the total number of stained DRG cell bodies, 43.6% were DiI positive. (c) DiO staining of DRG cell bodies after application of the dye in the DH axonal compartment. Of the total number of stained DRG cell bodies, 56.4% were DiO positive. (d) DiI/DiO overlap of DRG cell bodies after application of the dyes in both corner sides. From the total number of stained DRG cell bodies, 27.5% were DiI/DiO double stained, indicating axonal crossings into both compartments.

Identification of neuronal properties in the DH compartment

Next, we focused on the DH compartment and characterized the properties of neurons using immunocytochemistry at DIV12. Because of the gelatinous nature of the mouse spinal cord when performing P2-P4 dissections, it can prove demanding to accurately isolate the dorsal horn. We used a microtubule-associated protein 2 (Map2) antibody (Figure 3(a)), a common neuronal marker of cell bodies and dendrites in CNS neurons. ¹⁸ Map2 was co-stained with VGlut2 (Figure 3(b)), a marker of glutamatergic neurons predominantly expressed in laminae I and II of the DH.^19,20 We observed an overlap between Map2 and VGlut2 staining (Figure 3(d)), suggesting that the neurons of that compartment were indeed (i) spinal cord neurons and (ii) glutamatergic, consistent with the identity of DH neurons. Despite the application of the mitotic inhibitor, AraC, we could observe DAPI stained cells that were not MAP2 positive, which may be non-neuronal cells that have survived (Figure 3(c)).

Figure 3.

Synapse formation and glutamatergic expression in the DH neuronal compartment at DIV10. Potentially active synaptic input by Map2 neurons (red) and Synapsin 1, a synaptic vesicle marker expressed in presynaptic membranes. Overlap of Map2 (red) and VGlut2 (green) shows that mouse spinal cord neurons in the DH compartment are glutamatergic. (a) Map2-stained confocal image (red) of the DH compartment showing central neurons. (b) Vglut2-stained confocal image (green) of the DH compartment showing glutamatergic neurons. (c) DAPI-stained confocal image (blue) of the DH compartment. (d) Merged image of DAPI (blue), Map2 (red), Vglut2 (green) highlighting an overlap between Map2 and Vglut2 positive neurons. The overlap confirms that the neurons stained are central and glutamatergic. (e) Map2-staining (red) of the DH compartment shows the spinal cord central neurons in the DH compartment. (f) Synapsin1 expression (yellow) in the DH compartment showing pre-synaptic puncta. (g) DAPI-stained confocal image (blue) of the DH compartment. (h) Merged image of DAPI (blue), Map2 (red), Synapsin1 (yellow) evidence of putative synapses on DH neurons, accompanied by a magnified image of the Map2/Synapsin1 overlap.

We then examined synapse formation within the microfluidic DH compartment, where Map2 central neurons were stained in combination with a synaptic marker, Synapsin1 (Figure 3(f)). The Synapsin1 puncta attaching to the Map2-stained neurons revealed an abundance of potentially active synaptic sites (Figure 3(h)). The observation that Map2 is receiving synaptic input provides initial evidence for the healthy maintenance of DH neurons and underscores their susceptibility to form functional synapses. Taken together, these results provide evidence for the DH identity of the cultured neurons, showing glutamatergic properties. Secondly, they suggest the tentative formation of functional synapses between DH neurons with one another and from the sprouting DRG axons that originate from the cell body compartment and have crossed towards the DH compartment.

Activation of DRG neurons in microfluidic cultures by electrical stimulation

Based on the immunocytochemical evidence for potentially active synapse formation between sprouted DRG neurites and DH cell bodies in the DH compartment, we then methodically evaluated the responsiveness of different compartments to defined electrical current applications. We combined the microfluidic co-cultures with calcium imaging (Figure S1A & B) to quantify changes in Ca²⁺ activity following electrical stimulation. We used a biphasic stimulation board (Figure S1C) connected to two stimulus isolators which in turn supplied two soldered electrodes that provided the devices with current (Figure S1D). Stimulation with eight biphasic pulses was applied consistently at a 4 Hz frequency, of 2 ms duration at 4 μA constant current amplitude (Figure S1E).

We used an electrical stimulation protocol to apply and directly record from the cell bodies of DRG neurons in the middle compartment, using the calcium indicator Fluo-4 (Figure 4(a)). In the chosen ROIs, we measured ΔF/F changes in the amplitude only of DRG cell bodies that responded to stimulation. We found that each electrical pulse led to an increase in the Fluo4 fluorescence as an indication of increase in [Ca²⁺] at the soma (Figure 4(b)). An important concern in applying consecutive electrical stimulations was that either due to electrode fatigue or due to neuronal desensitization, the response to multiple stimulations might decline. In our protocol, we did not find alterations in the average ΔF/F amplitudes after four consecutive stimulations (Figure 4(c)). After establishing that the DRG cell bodies in the middle compartment were responsive to electrical stimulation, we replicated the experiment by stimulating the axonal compartment and recorded from the DRG compartment in the middle (Figure 4(d)). In agreement with the previous result, we found that DRG neuronal cell bodies responded to peripheral axonal stimulation (Figure 4(e)). Consecutive axonal stimulations did not alter the average amplitude DRG cell body responses (Figure 4(f)).

Figure 4.

Electrical stimulation of sensory axons in the periphery activates DRG cell bodies in microfluidic co-cultures. (a) Direct stimulation of sensory axons in the DRG somal (middle) compartment (8 biphasic pulses at 4 Hz frequency, 2 ms duration at 4 μA amplitude) with the same compartment as the readout site of calcium activity, using Fluo-4 b) Representative trace showing the responses at the cell bodies following four consecutive applications of electrical stimulation in the middle compartment. (c) Comparison of the average amplitude of responses following four successive stimulations. Consecutive responses showed no significant changes in amplitude (ns p < .05, One-way ANOVA with Tukey’s multiple comparisons test, n = 53 neurons, n = 3 devices from three cultures). (d) Schematic of the experimental setup. Stimulation of sensory axons in the Periphery compartment (left) with the DRG somal side (middle) as the readout site (e) Representative trace showing the responses at the cell bodies following four consecutive electrical stimulations in the Periphery compartment (left). (f) Comparison of the average amplitude of responses following four successive stimulations. Consecutive responses showed no significant changes in amplitude (ns p < .05, One-way ANOVA with Tukey’s multiple comparisons test, n = 44 neurons, n = 5 devices from three cultures).

These experiments provide evidence for the excitability of DRG cell bodies and their axonal projections in the peripheral compartment. The DRG cell bodies respond directly to electrical stimulation but may also exert excitable output following axonal stimulation. The latter experiment (Figures 4(d)–(f)) advocates that [Ca²⁺] is increased in the axons upon electrical stimulation are propagated towards the cell bodies of DRG neurons. This results in the ensuing Ca²⁺ responses at the cell body.

Activation of dorsal horn neurons in microfluidic cultures by electrical stimulation

After showing that the DRG cell bodies and axons are responsive to electrical stimulation, we focused on the excitability of DH neurons. The DH compartment was previously shown to contain glutamatergic neurons (Figure 3(a)–(c)) that are receiving synaptic input, pinpointing towards the putative formation of functional synapses between DRG and DH neurons (Figure 3(d)–(f)). The DH neurons form a network and while normally at rest, they may sparsely exhibit spontaneous, repetitive firing (Figure 5(a)). We electrically stimulated the DH cell bodies and directly recorded the responses to examine whether the DH neurons are amenable to neuronal excitation (Figure 5(b)). The DH neurons responded to electrical stimulation (8 biphasic pulses at a 4 Hz frequency x 2 ms @4 μA) and observed no significant fluctuations in the amplitude of their excitability (Figure 5(c) and (d)).

Figure 5.

Excitability properties of DH cell bodies following electrical stimulation. (a) Representative trace showing the occasional occurrence of spontaneous activity of a DH cell body in the absence of any stimulation (b) Schematic of the experimental setup. Stimulation of DH neurons (right) (8 biphasic pulses at 4 Hz frequency × 2 ms @4 μA) with the DH compartment itself as the readout site of calcium activity. (c) Representative trace showing changes in calcium influx of DH cell bodies following four consecutive applications of electrical stimulation directly to the DH compartment. (d) Comparison of the average amplitude of responses following four successive stimulations. Responses had no significant changes in amplitude (ns p < .05, One-way ANOVA with Tukey’s multiple comparisons test, n = 10 neurons, n = 3 devices from three cultures).

We showed that DRG cell bodies respond to electrical stimulation generated both through direct stimulation of the middle compartment but also from peripheral stimulation (Figure 4). It compelled us to examine whether peripheral axonal and DRG cell body stimulation will affect DH neuron excitability. Hence, we reprised the protocols of stimulating the DRG cell bodies and the peripheral axons but this time we used Ca⁺² responses of the cell bodies of DH neurons as a functional outcome. Stimulation induced to the DRG cell bodies in the middle compartment resulted in the increased levels of Ca²⁺ in DH neurons (Figure S2A & B), which demonstrates functional incoming synaptic input from the crossed DRG axons. Consecutive applications of electrical stimulation did not alter DH-elicited calcium responses, implying that there was no runup or rundown in synaptic responses (Figure S2C).

To determine the completeness of peripheral nociceptive circuitry recapitulation, we stimulated the peripheral axons and recorded from the DH neuronal cell bodies in the DH compartment (Figure 6(a)). The stimulation of peripheral axons implied that the generated APs traverse through the middle compartment containing the DRG cell bodies and from there to the DH cell bodies compartment axons (Figure 6(b)). Indeed, stimulation of peripheral axons resulted in Ca²⁺ responses in the DH cell compartment (Figure 6(c)), recapitulating the periphery-to-centre circuit.

Figure 6.

Electrical stimulation of peripheral axons leads to [Ca²⁺]_i elevation of DH neurons. (a) Schematic of the experimental setup. Stimulation of the peripheral axons (left) (8 biphasic pulses at 4 Hz frequency × 2 ms @4 μA) depolarizes the cell bodies at the DH compartment (right) (b) Representative trace showing changes in [Ca²⁺] of DH cell bodies following four consecutive applications of electrical stimulation in the axons of the periphery compartment. (c) Comparison of the average amplitude of responses following four successive stimulations. Responses had no significant changes in amplitude (ns p < .05, One-way ANOVA with Tukey’s multiple comparisons test, n = 121 neurons, n = 3 devices from three cultures).

Glutamatergic dependence in synaptic transmission between DRG and DH neurons

The confirmation of functional synapse formation between DRG and DH neurons was crucial towards the establishment of the model. It is well-established that fast neurotransmission at the first sensory synapse is glutamatergic and mediated by NMDA and AMPA receptors. We employed a dual blockade of glutamate receptors NMDA and AMPA to evaluate synaptic transmission to DH neurons in the microfluidic model. The application of CNQX (20 μM) in concomitance with an NMDA receptor blocker, MK-801 (1 μM), largely attenuated the amplitude of evoked signals in DH neurons (Figure S3B). The repetitive stimulation in the presence and absence of CNQX and MK-801 did not alter the decay time of DH neuron calcium responses (Figure S3C). Similarly, electrical stimulation at the peripheral compartment was performed with the Ca²⁺ responses at the DH cells bodies as the readout. The stimulation of peripheral axons in the presence of CNQX and MK-801 at the DH compartment (Figure 7(b)) again produced a stark reduction in the amplitude of DH cell body responses (Figure 7(c)). The decay time represented as a tau value remained unaltered in the presence of glutamatergic blockade (Figure 7(d)).

Figure 7.

Dorsal horn neuron responses to electrical stimulation in peripheral axons before and after glutamate receptor blockers application (a) Schematic of the experimental setup. Stimulation of the peripheral compartment (left) with the DH compartment (right) as the readout for calcium imaging. After two stimulations, the glutamate receptor blockers CNQX (20 μM) and MK-801 (1 μM) were applied and then two more stimulations under the presence of the blockers were made. (b) Representative trace showing changes in calcium influx of DH cell bodies following two applications of electrical stimulation in the absence and two applications of electrical stimulation in the presence of glutamate receptor blockers. (c) Comparison of the average amplitude of responses before and after application of the glutamate blockers (Welch’s t-test, ****p < .0001, n = 10 neurons, three devices from three cultures). (d) Decay time for DH responses to return to baseline represented as tau. The addition of CNQX and MK-801 did not alter the decay time following electrical stimulation of the peripheral axons (ns p < .05, Brown-Forsythe and Welch ANOVA, n = 10 neurons, n = 3 devices from three cultures).

Lastly, glutamate receptor blockade did not affect the excitability of DH neurons when the DH neurons were directly stimulated with electrical current (Figure S4A–D). This result also diminishes any potential doubt about artefactual responses in our previous experiments. Collectively, the dependence on glutamatergic signalling is consolidated between DRG and DH neurons in our model system. This feature signifies the accuracy of recapitulating the in vivo anatomical features of the peripheral nociceptive circuit and subsequent properties of the first sensory synapse.

Axotomy as a surrogate model of peripheral nerve injury in microfluidic co-cultures

Following the functional characterization of the first sensory synapse in the microfluidic platform, we considered which pain models can be approximated using it. We pursued to model peripheral nerve injury, a common aetiology of neuropathic pain. The platform, by design, can discriminate and readily investigate between different anatomical sites of nociceptors. We chose to focus on how peripheral axotomy affects synaptic transmission and subsequently measure the excitability of DH neurons. Our protocol allowed neurite sprouting in the co-cultures for 5 days (Figure 8(a)). An axotomy in the peripheral compartment was subsequently performed (Figure 8(b)) using repeated manual force application with a 1000 μL tip until all axons were cut and removed from the peripheral axonal compartment. The axotomized neurites were allowed to re-grow for another 3-5 days (Figure 8(c)). Following re-growth of the axotomized neurites, calcium imaging was performed by stimulating the peripheral axons and recording changes in the ensued excitability of DH cell bodies (Figure 8(d)).

Figure 8.

Axotomy of distal sensory axons induces increased and prolonged DH responses in comparison to non-axotomized axons (a) Schematic protocol of the axotomy model in microfluidic co-cultures between DRG and DH neurons. From days in vitro (DIV) 0 to 4, the DRG neurons and DH neurons were allowed to grow neurites and sprout (b) On DIV5, axotomy was induced in the axons that sprouted to the peripheral compartment (left). (c) The axons of peripheral neurons were allowed four more days (DIV5-DIV9) to re-sprout following the axotomy. All other neuronal compartments have been left intact. (d) Since the peripheral axons have re-sprouted, calcium imaging of the DH neuronal compartment was used between DIV9-DIV12 (right) to measure changes in excitability following electrical stimulation of the peripheral axons where axotomy had been induced (left). (e) Representative trace showing responses of DH cell bodies following electrical stimulation of sensory axons in non-axotomized cultures (8 biphasic pulses at a 4 Hz frequency × 2 ms @4 μA). (f) Representative trace showing responses of DH cell bodies following electrical stimulation of sensory axons in axotomized cultures (g) DH neurons in axotomized cultures display greater calcium responses following stimulation of peripheral axons. (Welch’s t-test, ***p < .0001, n = 9 devices from five non-axotomized cultures and n = 7 devices from four axotomized cultures) (h) Decay time of DH neuron responses represented as tau following stimulation sensory axons. The decay time of DH neuron responses following axotomy is significantly longer than the non-axotomized responses (Welch’s t-test, *p < .0347, n = 334 neurons, n = 9 devices from five non-axotomized cultures and n = 254 neurons, n = 7 devices from four axotomized cultures) (i) Table of the number of responder DH neurons between axotomized and non-axotomized devices. The non-axotomized devices showed no difference in the number of DH responder neurons compared to the axotomized ones.

Comparisons were then made in matching experiments between intact and axotomized cultures, where DH responses were recorded following stimulation of the axonal compartment (Figure 8(e) and (f), respectively). Using identical stimulation parameters, DH neurons in devices receiving input from axotomized sensory axons evoked larger responses in comparison to those in non-axotomized ones (Figure 8(g)). The result suggested that potential changes in the excitability of sensory neurons due to axotomy results in enhanced synaptic transmission and excitability of DH neurons. The DH responses of axotomized cultures had a larger decay time in comparison to the non-axotomized ones, suggesting longer and amplified DH activation duration (Figure 8(h)). However, we did not observe differences in the number of DH neurons that we recorded from between axotomized and non-axotomized cultures (Figure 8(i)).

In conclusion, we find that axotomy induces Ca²⁺ responses that are larger in amplitude and of longer duration in DH neurons, revealing that ensued changes in the excitability of sensory axons influence transmission mechanisms across the first sensory synapse. The observation that peripheral nerve injury promotes a transition towards an excitable state of DH neurons may provide mechanistic insight into the development of neuropathic pain.

Discussion

The establishment of a microfluidic model of the first sensory synapse in mice is a novel contribution to study the pain pathway. Certain in vivo studies have previously studied the properties of DRG or DH cell bodies in pathological pain models but there has been limited knowledge about their axonal and synaptic mechanisms. There have been scarce studies of mouse DH neuron properties in vitro, as these cells have been considered challenging to survive in culture. Previous efforts of co-culturing DRG and DH neurons were performed using a variation of the Campenot chambers. ²¹ Those experiments were performed in a 2-compartment configuration using embryonic rat neurons, where the stimulation of DRG cell bodies could not be discriminated from sensory axons. ²² Another previous study utilized calcium imaging to record fluorometric responses to noxious stimuli, but the DRG and DH neurons were co-cultured in the same compartment, disabling precision in pharmacological manipulations. ²³

Our microfluidic models of the first sensory synapse have resolved several of these issues. ¹⁶ We demonstrated that utilizing a 3-compartment configuration we can isolate and study all spatially distinct sites. We showed that mouse DRG neurites bilaterally cross towards the two corner peripheral compartments of the microfluidic devices. This configuration allows the separation of sensory axons in one compartment and the recapitulation of the synaptic site with the co-cultured DH neurons on the other. It is henceforth possible to locally apply pharmacological agents, quantify evoked-responses and subsequently measure multiple DRG or DH neurons’ excitability using fluorometric techniques. The use of postnatal mice premises the validation of pharmacological observations with available genetic models. Moreover, given the thorough classification of mouse sensory and DH neuron subtypes using transcriptomics,^24–26 it enables the precise characterization of the cultured neurons, a conceptually important improvement to the previous rat model. ¹⁶

Through the combination of electrical current stimulation with calcium imaging, we methodically established the functionality of the co-cultures. Unlike the cumbersome recording of DH activity using patch-clamp electrophysiology, the utilization of Fluo4 allowed the large-scale recording of several DH cell body responses to DRG stimulation. Stimulation of the axonal or DRG compartment elicited robust responses from the DH cell bodies confirming our observation of functional synapses. Antithetical to potential concerns about electrode fatigue or desensitization confounding the DRG responses, in the DH compartment, the repeated stimulation did not confound our results due to progressive changes in excitability of DH neurons. Repetitive and high-frequency nociceptor stimulation may amplify and prolong responses due to the summation of slow excitatory post-synaptic (EPSPs) which is manifested as ‘windup’.^27,28 It has been shown that wind-up is dependent on glutamatergic transmission from nociceptors to wide-dynamic range neurons (WDRs), which are NMDA-receptor dependent.^29,30 Through our stimulation strategy, we did not observe changes in the amplitude or decay time of DH-evoked responses. Of future interest would be to apply supramaximal frequencies, currents or more stimulus applications to determine whether DH neurons could exhibit windup in vitro. However, our aim was to establish a protocol where the baseline responses are not altered by repetitive stimulations, as it would obstruct the interpretation of experiments involving the pharmacological manipulation of the DH compartment.

Early evidence in explants of embryonic rat DRG and DH neurons showed that glutamate elicits excitatory EPSPs in the synapses formed.^9,10 While these studies did not initially conclude that NMDA antagonism was effective towards EPSPs attenuation, it was later found that the post-synaptic NMDA receptors are essential for ligand-binding of glutamate together with AMPA receptors.^31,32 We were compelled to test the dependence of DH responses to glutamate receptor binding and hence to demonstrate that it leads to excitatory synapse formation in vitro. Application of selective AMPA and NMDA receptor blockers, CNQX ³³ and MK-801 ³⁴ respectively, largely attenuated the peripherally evoked responses. While the DH-evoked responses were attenuated, they were not completely abolished. A plausible explanation would be the concomitant dependence of the DH neurons to more than one transmitters acting together with glutamate, such as the neuropeptides Substance P and CGRP.^35,36 We have excluded the contingency of blocker rundown in our responses, as the application of a selective Nav1.8 blocker did not attenuate synaptic transmission. ³⁷ Despite the attenuation in DH response amplitude due to the glutamatergic blockade, we did not observe significant changes in the decay rate of responses. This implies that the DH neuron calcium efflux mechanism remains unaltered in the wild-type mouse, perhaps in compliance with the recent discovery that the sodium-calcium exchanger NCX3 regulates prolonged temporal responses of DH neurons in mice lacking it. ³⁸ Last, we did not observe attenuations in DH neuron responses after direct stimulation of isolated DH neurons in the presence of glutamate receptor blockers. Although in vivo synaptic responses between DH neurons may be influenced by the network’s excitability, most responding DH neurons in the MFC assay are proximal to the device corridors and presumably receive direct input from the incoming DRG axons. In summary, the evidence corroborates the presence of glutamatergic transmission between DRG and DH neurons in our microfluidic model of the mouse first sensory synapse.

In the microfluidic platform, there is lack of the somatotopic alignment between sensory neurons and DH neurons. At the first relay point of the pain pathway, nociceptors typically innervate the central sites of distinct laminae of the dorsal horn, specifically laminae I and II. However, with the current in vitro techniques, this anatomical feature is not maintained, since both DRG and DH neurons are dissociated and randomly distributed in the devices. Surprisingly, certain elements of the somatotopic alignment are present in these cultures. From a previous study, we have evidence in microfluidic DRG cultures for mainly the survival of small- diameter, unmyelinated sensory neurons. ¹⁴ Concomitantly, the immunocytochemistry experiments at the DH compartment characterized the identity of the cultured cells as spinal cord dorsal horn neurons. The Map2 staining confirmed that these neurons are indeed central and secondarily, through the VGlut2 staining, they were identified as glutamatergic. Thirdly, via the expression of Synapsin1, we showed that these central, glutamatergic neurons may potentially form active synapses within that compartment. The VGlut2-expressing dorsal horn neurons are predominantly localized in laminae I and II of the dorsal horn.^19,39 Previous studies have identified axonal guidance cues of peripheral sensory axons, such as the Slit protein ⁴⁰ and the Runx3 which is responsible for the guiding the dorso-ventral termination of cutaneous DRG axons into the spinal cord. ⁴¹ The in vitro specificity of laminar targeting of DRG axons to distinct DH neuronal bodies is an important avenue for further investigation. It elicits compelling questions about the spatiotemporal expression of such synapse formation cues throughout the postnatal life of sensory neurons both in vivo and in vitro.

In terms of functional plasticity, neuropathic pain is caused by damage or disease and alters the excitability of the somatosensory nervous system. Several in vivo animal models of neuropathic pain have been developed, including that of spared nerve injury (SNI) and spinal nerve ligation (SNL). ⁴² The robust establishment of the microfluidic co-cultures allowed us to further develop an in vitro model to study peripheral nerve injury. We devised an axotomy strategy and utilized the stimulation protocol of DRG axons to measure the ensued excitability of DH neurons. We found that DH neurons receiving input from axotomized DRGs displayed larger responses and with greater duration in comparison to those receiving input from non-axotomized stimulated axons. This suggests that peripherally induced changes have facilitated synaptic transmission and modulated the excitability of DH neurons. In a neuropathic pain model, it has been shown that uptake of glutamate by DH neurons leads to a large elevation of calcium responses. ⁴³ Previous studies have suggested that long-term potentiation (LTP) mechanisms of spinal synapses account for the sensitization of the dorsal horn in the presence of nerve crush or injury.^44,45 Peripheral hyperexcitability of axotomized nociceptors may provide a basis of the induction of LTP in spinal cord dorsal horn neurons. Distal axotomy of pre-synaptic pyramidal neurons has been found to increase glutamate release in CNS neurons. ⁴⁶ Elevated glutamatergic neurotransmission and saturated uptake by the metabotropic glutamate receptor 5 (mGluR5) has been shown in the laminae I and II of the dorsal horn during neuropathic pain.^47,48 We postulate that increased glutamatergic neurotransmission may be responsible for the amplified DH responses in the axotomized cultures. ⁴⁷ We did not examine the synaptic strength or increase in the number of synapses formed between DH neurons and axotomized DRG axons, however there is important such evidence. ⁴⁴ In the future, it would be of interest to quantify the pre- and post-synaptic staining of the DH compartment and compare between the puncta between axotomized and non-axotomized conditions. Overall, our study shows that axotomy in microfluidic co-cultures replicates salient features of the nerve damage-induced changes in the first sensory synapse, by undergoing both pre- and post- synaptic modulation.

Evidence from neuropathic pain injury models have shown a loss of GABA-ergic inhibition in the superficial dorsal horn as well as a loss of glycinergic inhibition which is known to facilitate NMDA receptor activation.^49–52 The spinal disinhibition has been attributed to loss of dorsal horn neurons.^50,53 At the same time, axotomy induces plastic changes in primary sensory neurons, among them a downregulation of the pro-nociceptive neuropeptides Substance P and CGRP ⁵⁴ along with a transcriptional reprogramming of sensory neurons. ⁶ Since there is a downregulation of essential pro-nociceptive transmitters, what could then account for the paradoxical phenomenon of the occurrence of painful responses when a nerve is damaged, instead of exhibiting hypoalgesia? A general framework would regard that the peripherally induced plasticity due to nerve injury may result in glutamate receptor activation and a temporary LTP of DH neuron excitability. ⁵⁵ However, the persistent activation of glutamate receptors and the influx of calcium elicited have been shown to induce apoptosis in DH neurons^53,56 and this observation might provide the basis of a prolonged, central sensitization of the spinal cord. ⁵⁷ In a reductionist demonstration, our experiment suggests that axotomized neurons elicit larger and longer DH responses to electrical stimulation within 96 h following axotomy. The increased neuronal activity as well as the prolonged calcium influx may cause a synaptic modulation that in turn alters DH neuron responses. This may provide a useful link between the temporary, peripheral neuron hyperexcitability after nerve injury and spinal cord central sensitization once peripheral neuron activity subsides.

Last, non-neuronal cells have been shown to modulate the function of neurons in many contexts including pathological pain. ⁵⁸ We chose to examine synaptic transmission in the absence of non-neuronal cells. However, there are peripheral glial cells in the DRG (Schwann and satellite glial cells) or immune cells and keratinocytes which are contributing to peripheral pain plasticity.^59,60 Similarly, resident microglial cells and astrocytes have a long-established role in modulating spinal cord excitability.^58,61 By eliminating these non-neuronal cells from our model, we are also missing their potential contributions following persistent stimulation or their effect after axotomy. However, an advantage of the model is the potential to “add back” the missing non-neuronal cells in future studies.

Overall, we have devised the first in vitro model of the first sensory synapse in mice. Our model reprises robustly several features of the in vivo environment between DRG and DH neurons. We show that the model is amenable to pharmacological manipulations within each distinct compartments while it can also become amenable to genetic manipulations. The platform can be used to mimic peripheral nerve injury as shown by the axotomy model or study inflammatory mediator effects on synaptic transmission. Further, it can be used to study developmental aspects of synapse formation and axonal guidance queues between DRG and DH neurons. Last, the platform can be directly utilized as a screening assay of compounds targeting peripheral nociceptors, where silencing of the DH compartment could be the desired functional output. Whether used as a surrogate model of pain pathology in pre-clinical studies or as a drug screening method, the model will hopefully provide novel insight on the cellular and molecular physiology across the first sensory synapse.

Methods

Animals

C57BL/6 female mice with P2 litter were purchased from Charles-River UK. All animals were maintained in a designated facility according to the UK Home Office Code of Practice for the Housing and Care of Animals Used in Scientific Procedures. Due to resistance to hypoxia, male and female C57BL/6 postnatal mice (P2-P4) were sacrificed only by cervical decapitation in full compliance with UK Home Office Regulations and procedures under Schedule one of Animals (Scientific Procedures) Act 1986. All procedures were approved by the Animal Welfare and Ethical Review Body at King’s College London (PPL U136).

Preparation of microfluidic devices

Microfluidic devices with three compartments consisting of 500 μm microgroove length (Xona Microfluidics, TCND500 design) were used. Glass bottom dishes of 40 mm diameter (WillCo Wells) were cleaned by sonication and sterilized with 70% ethanol. The dishes were previously coated with 0.5 mg/mL poly-L-lysine (Sigma). Microfluidic devices were non-plasma bonded onto the glass. The assembled devices were coated after assembly with 40 µg/ml laminin (Sigma) dissolved in Neurobasal medium (Thermo Fisher).

DRG-DH cultures of postnatal mouse neurons

P2-P4 C57BL/6 mice were sacrificed instantly by decapitation. Preparation of dissociated DRG and DH neuron cultures was performed through initial aseptic excision of the spinal column. The spinal cord is removed carefully while cutting through ventral roots that may still be attached, then the meninges are pulled apart. Approximately 1/3 of the dorsal side was cut off to isolate the DH which is placed in HBSS +2% BSA. Cervical, thoracic, lumbar and sacral DRG-containing cavities were exposed, the DRG were collected and placed in HBSS +2% BSA. The DRGs were digested with enzyme mixture containing 0.125 mg/mL collagenase (Sigma) and 10 mg/mL dispase (Thermo Fisher Scientific) and the DH were digested with 2 mg/mL papain (Sigma) enzyme mixture for 40 min in a 37°C and 5% CO₂ humidified incubator.

Following mechanical trituration of DRG and DH neurons in their respective 30 mm petri dishes, the dissociated neurons were centrifuged respectively on a 10% and 15% bovine serum albumin cushion in HBSS. The cell pellet was washed twice and resuspended in complete medium containing Neurobasal (Thermo Fisher), 1% B27 supplement (Invitrogen) and 1% Glutamax (Gibco), 100 units/ml penicillin and 100 μg/mL streptomycin.

Due to the gelatinous and fragile nature of postnatal mouse DRG and DH neurons, each co-culture contained the extracted and dissociated DRG and DH neurons of 2-3 mice. The DH neurons were loaded into one of the two corner compartments of the prepared microfluidic devices and were left to settle for 5 min. Then the DRG were loaded into the middle compartment of the same devices and were left to settle for 5 min. The co-cultured DRG and DH neurons were left to adhere for 60 min in the incubator before flooding with NGF-containing media. To ensure neurite sprouting, an NGF-gradient between somal and the two corner compartments was applied upon flooding, where 200 μL containing 150 ng/mL NGF (Gibco) on the somal side and 130 μL containing 200 ng/mL NGF on the two axonal sides were added. The devices were then left for 9–12 days in the incubator, with renewal of media every 2-3 days. On DIV3, a mitotic inhibitor, AraC was added to the cultures (2 μM, Fisher Scientific).

Axotomy

In the DRG-DH co-cultures, at DIV5, the peripheral axons of sensory neurons are axotomized using manual application of force to cut them inside the MFC device corridor. A 1000 μL tip was filled with complete medium and was applied and suctioned forcefully and continuously until all axons were cut and removed from the peripheral axonal compartment.

DiI/DiO tracing

The lipophilic trace dyes DiI and DiO were used to mark cell bodies with axons crossing the microgrooves to the axonal side of the microfluidic devices. 24–48 h prior to the recording, the axonal compartments were incubated with the DiI and DiO solutions (1:200 in complete medium) for 1 h at 37°C and 5% CO₂. After washing with Neurobasal-A medium, and replacement with complete medium, the devices were placed in the incubator. The stained cell bodies were visualized immediately prior to live imaging on a Nikon TE200 microscope using a TRITC fluorescence filter set.

Calcium imaging

In the co-cultures, between DIV9-12, all three compartments were incubated for 1 h with 2 μM of the calcium indicator dye Fluo4 (Thermo Fisher) in imaging buffer consisting of HBSS [Ca²⁺ and Mg²⁺- free] supplemented with 10 mM HEPES (Thermo Fisher), 2 mM CaCl₂ (Sigma), 1 mM MgCl₂, 2 mM probenecid (Sigma), pH 7.4 with NaOH. Prior to recording fluorescence activity of DH neurons, the central DRG and peripheral axonal compartment were filled with imaging buffer. The devices were mounted on an inverted fluorescent microscope (Nikon Eclipse TE200) equipped with 20X Plan Fluor 0.5 NA objective and an EasyRatioPro imaging unit (Horiba Scientific). Images were acquired using ORCA 4.2 sCMOS camera (Hamamatsu). Bright field and DiI staining were visualized first, and only neurons with staining in the soma (indicative of axonal crossing) were used for the analysis. The regions of interest (ROIs) were drawn around each corresponding soma and fluorescent images were acquired at 510 nm at 6.3 Hz.

Drug perfusion

A modified perfusion system (Digitimer) has been built for controlled pharmacological administration and avoidance of movement artefacts. The system is comprised of an 8-channel pen with a 360 μm removable tip (Automate Scientific) and regulated by a VC-8 Valve Controller (Warner Instruments) to precisely allow 50 μL of solution every 5 s upon turning on (Figure S1B).

Electrical stimulation

Electrical stimulation was applied to the compartments to quantify the amplitude (ΔF/F ratios, see Image Analysis section) of DH or DRG neuron responses. A biphasic stimulator board was built comprised of three width/delay modules (NL405, NeuroLog/Digitimer), a pulse generator module (NL301, NeuroLog/Digitimer), a pulse buffer module (NL510A, NeuroLog/Digitimer) and an analogue switch (NL506, NeuroLog/Digitimer). The board was connected to two current stimulus isolators (NL800A, NeuroLog/Digitimer) which provided input to handmade soldered electrodes made up of coiled AgCl pellets. The electrodes were set to produce eight biphasic pulses at 4 Hz frequency x 2 ms @4 μA

Image analysis

For calcium imaging experiments, the ΔF/F ratios were determined for designated regions of interest (ROIs) using EasyRatioPro software. The EasyRatioPro sequences saved from the live calcium imaging recordings were converted into frame-by-frame TIFF files. The TIFF files were added to FIJI and ΔF/F traces were acquired using that software. which were saved as comma delimited (.csv) files. The comma delimited files were uploaded on the specialized for electrophysiological analysis software Clampfit 9.0 (Molecular Devices). The amplitudes of the traces were determined using the software.

On each ROI, the responding cell bodies of either DRG or DH neurons were analysed following electrical stimulation. The responses to stimulation were visually verified and confirmed if the maximum increase in F_ratio was at least three times greater than the standard deviations of the baseline calculated from at least 50s prior to drug application. The peak response (ΔF_ratio) was calculated as the maximum increase in fluorescence intensity above the baseline. The detection of Fluo4 fluorescence intensity changes was defined as the (ΔF/F) between baseline responses taken at least 10 s prior to application of stimulus where no fluorescence activity was reported and ΔF maximum peaks within the timespan of application.

In certain calcium imaging experiments, together with changes in amplitude of fluorometric responses represented as ΔF/F values, the decay time (tau) of responder traces until they return to baseline was calculated. The tau values were represented as the following standard exponential curve function:

$$f\left(t\right)={\displaystyle \sum }_{i=1}^n{A}_i{e}^{-t/{\tau }_i}+C$$

Immunocytochemistry

Neurons were washed with D-PBS and fixed in 4% paraformaldehyde (PFA) in phosphate buffer for 10-15 min, then washed and stored at 4°C. After three consecutive D-PBS washes, cells were incubated with blocking solution (10% donkey serum +0.1% Triton-X) for 1 h at 25°C. The primary antibodies for chicken anti-MAP2 (Novus Biologicals 1:4000 dilution), anti-Synapsin 1 (Invitrogen, 1:4000 dilution) or anti-VGlut2 polyclonal (Invitrogen, 1:10000 dilution) were added and incubated for 24 h at 4°C. Upon removal of the primary antibodies, the devices were washed three consecutive times with D-PBS. Respective secondary antibodies donkey anti chicken 649 (Invitrogen, 1:1000 dilution), donkey anti rabbit 488 (Invitrogen, 1:1000 dilution) and donkey anti rabbit 594 (Invitrogen, 1:1000 dilution) were incubated for 1 h at 25°C, which were followed by removal and subsequent D-PBS washes. The acquisition of confocal images was performed in an inverted Zeiss Imager. Z1 microscope with 20x objective and equipped with DAPI, TRITC and FITC filters within 24 h after removal of secondary antibodies. The experiments were performed once and following image acquisition, no further analysis and quantification was conducted.

Statistics

The data are reported as mean ± SEM unless otherwise stated. The number of DH neurons, number of devices and number of cultures are designated in each figure legend. For quantifying changes in brightness intensity with Fluo4, the corresponding values of ROIs were used in the statistical analysis. All figures designed and statistical analysis have been performed using GraphPad Prism 9. Statistical significance was set at p < .05. Student t-test (unpaired, equal variance) was used to compare groups in calcium imaging analysis. One Way ANOVA with Tukey’s multiple comparison test was used for comparisons between three or more independent groups in calcium imaging analysis.

ORCID iD

Georgios Kimourtzis https://orcid.org/0009-0005-4049-1769

Data availability statement

All data discussed in the manuscript and required for the evaluation of the conclusions have been presented in the Results and in the supplementary figures. All data will be available upon request.*