Self-assembled Innervated Vasculature-on-a-chip to Study Nociception

Abstract

Nociceptor sensory neurons play a key role in eliciting pain. An active crosstalk between nociceptor neurons and the vascular system at the molecular and cellular level is required to sense and respond to noxious stimuli. Besides nociception, interaction between nociceptor neurons and vasculature also contributes to neurogenesis and angiogenesis. In vitro models of innervated vasculature can greatly help delineate these roles while facilitating disease modeling and drug screening. Herein, we report the development of a microfluidic-assisted tissue model of nociception in the presence of microvasculature. The self-assembled innervated microvasculature was engineered using endothelial cells and primary dorsal root ganglion (DRG) neurons. The sensory neurons and the endothelial cells displayed distinct morphologies in presence of each other. The neurons exhibited an elevated response to capsaicin in the presence of vasculature. Concomitantly, increased transient receptor potential cation channel subfamily V member 1 (TRPV1) receptor expression was observed in the DRG neurons in presence of vascularization. Finally, we demonstrated the applicability of this platform for modeling nociception associated with tissue acidosis. While not demonstrated here, this platform could also serve as a tool to study pain resulting from vascular disorders while also paving the way towards the development of innervated microphysiological models.

1. Introduction

The advent of organoids and microfluidic-assisted organ(s)-on-chip systems has significantly reduced the gap between conventional in vitro platforms and organisms and has emerged as an important model system and a pre-clinical tool[1,2]. Despite the tremendous advances in creating multi-cellular complex tissue models, innervation remains largely neglected[3,4]. While it is common to incorporate neurons into models of the blood-brain-barrier and neurovascular units[5,6], sensory neurons within ex vivo models are largely studied in isolation or in 2D co-culture systems[7–12] and very few multicellular tissue-on-a-chip systems incorporating sensory neurons have been reported[13–16]. Absence of sensory innervation in non-neuronal engineered tissues presents a significant barrier towards gaining a deeper understanding of the role of peripheral nervous system in maintaining organ function and studying processes that regulate nociception and phenomenon such as pain.

Nociception is critical for protecting tissues from dangers and promoting healing when injured. Peripheral tissues are highly innervated by sensory neurons that sense and respond to mechanical, thermal, and chemical noxious stimuli. Nociceptor afferents express ligand- and voltage-gated ion channels, the key molecular transducers of noxious stimuli[17]. The cell bodies of the primary sensory neurons are housed within the dorsal root ganglion (DRG). Vascularization of the DRG and the reciprocal interactions between vasculature and the sensory neuron afferents plays an important role in chemical nociception[18]. Previous studies have indicated the key role played by DRG vascularization in chemically induced peripheral sensory neuropathy, a major side effect observed among cancer and HIV-patients following various drug treatments[19]. Interactions between the vasculature and the sensory neurons are also important in inflammatory pain[20,21]. Inflammatory cytokines are key mediators of nociceptor activity in sensory neurons and pain sensitization[21]. Another clinically important chemical noxious stimulus is associated with tissue acidosis, a condition characterized by increase of proton concentration (i.e., low pH)[22]. Tissue acidosis is associated with several pathological conditions such as cancer, respiratory diseases, and muscular disorders[23–25].

The innervation of the vasculature also plays a crucial role in regulating the blood flow[26]. The main types of neurons innervating the vasculature are sympathetic, parasympathetic, and sensory neurons[27]. The sympathetic neurons innervating the vessels regulate their vasoconstriction while the sensory neurons contribute to vasodilation, which is regulated by the secretion of neuropeptides and neurotransmitters[27]. A number of clinical conditions such as diabetic neuropathy, Raynaud’s disease, and fibromyalgia are associated with dysfunctional neuronal activity of vasculature[26,28–30]. Lack of physiologically relevant in vitro models presents a significant bottleneck in understanding the mechanisms driving these conditions and development of novel drugs including analgesics to relieve pain[31].

In this study, we have developed a microfluidic model of innervated vasculature to study the coupling between sensory neurons and vasculature. Co-culture of human umbilical vein endothelial cells (HUVECs) with primary dorsal root ganglion (DRG) neurons harvested from mice and human donors led to formation of perfusable vascular networks which were innervated by a 3D network of neurites sprouting from the sensory neurons. The neurites wrapped around the vessels, resulting in direct contact with the endothelial cells and thereby promoting the activation of sensory neurons in response to chemical stimuli in the microvasculature. Employing this platform, we also examined the effect of acidosis on sensory neurons by decreasing the pH of the perfusate within the microvasculature.

2. Methods

2.1. Device Fabrication

The device was created by soft lithography using polydimethylsiloxane (PDMS). Briefly, 150 μm SU-8 was spin-coated on silicon wafer (University Wafers) and baked before being exposed to UV light through a photomask to create features with a height of 150 μm. The device design adapted here has been used by a number of studies such as those of vasculature[32,33], blood-brain-barrier[6,34,35], tumor[36–38], and embryo development[39]. The device consists of two parallel media microchannels (500 μm wide) separated by a central hydrogel microchannel (500 μm wide). The microchannels are separated from each other by an array of trapezoidal micro posts with an inter-post distance of 100 μm. Post exposure, the wafer was baked before being immersed in a developer solution to wash away the uncured photoresist. PDMS (9:1; base:crosslinker) (Sylgard 184, Dow Inc.) was poured onto the patterned wafer and placed in a desiccator for 1 hour to remove air bubbles. The wafer with the PDMS was then placed at 60 °C and allowed the PDMS to cure. The cured PDMS was cut around the features and the inlets were punched before bonding it to a glass cover slip using corona treatment. Figure S1 shows an image of the assembled device. For studies involving co-culture of DRG neurons and endothelial cells without direct contact, additional hydrogel and media microchannels were added to the design such that the hydrogel microchannels incorporating sensory neurons and endothelial cells were separated from each other by the media microchannel; this prevents direct contact between the cells but allows cell-cell communication via secreted factors.

2.2. Isolation of murine DRG neurons

DRGs were isolated from mice using a modified version of previously described protocols[40,41]. Briefly, mice were euthanized and sterilized in 70% ethanol to prevent any fur contamination. An incision was made, while the mice were positioned prone, from the neck to the base of the tail along the spine. The spinal column was extracted by separating it longitudinally from the surrounding soft tissue and transversely from the mid thoracic region and lumbosacral joint. A sharp pair of scissors was used to remove the ventral side of the vertebra exposing the spinal cord. The DRGs (situated near the foramen between vertebral levels) were carefully extracted. The collected tissues were digested using a mixture of collagenase type II (Worthington, LS004176) and dispase II (Millipore Sigma, D4693) (1.5 mg/mL each) dissolved in DMEM/F12. DRGs were agitated in the digestion solution on an orbital shaker at 37 °C for 20 minutes. The digestion solution was replaced with trypsin-EDTA (1:10) in DMEM/F12 and incubated for 15 minutes to disrupt the cell cluster. Next, the trypsin solution was replaced with DMEM/F12 (Thermo, 11320033) containing FBS (1:3; FBS:DMEM/F12) to neutralize the trypsin-EDTA. The dissociated DRGs were then triturated and created a cell suspension. The cell suspension was gently transferred on top of a cushion of 15% bovine serum albumin (BSA; Millipore Sigma, A4503) and centrifuged for 6 min at 280g (set for minimal acceleration and no deacceleration) to separate non-neuronal cells and debris. The supernatant was aspirated, and the cells were resuspended in neuronal medium (Neurobasal medium (Thermo, 21103049) + 2% B27 (Thermo, 17504044) + 1% GlutaMax (Thermo, 35050-061) + 1% Penicillin-Streptomycin (Thermo, 15140122)).

2.3. Isolation of human sensory neurons

Sensory neurons were isolated from human DRGs by using a modified version of a previously described method[42]. The dorsal root ganglia of human patients undergoing surgical decompression and fusion for clinically indicated purposes was resected for analysis under Duke IRB Pro00101198. Once the DRG was resected, the specimens were immediately processed. DRGs from 2 different donors were used. Any visible connective tissues were removed and the DRGs were washed with PBS. The DRGs were then minced and incubated in a cocktail of collagenase type II (Worthington, LS004176) and dispase II (Millipore Sigma, D4693) (1.5 mg/mL each) for 40 minutes on an orbital shaker with trituration in between. Most of the collagenase-dispase solution was removed and neuronal medium (Neurobasal medium (Thermo, 21103049) + 2% B27 (Thermo, 17504044) + 1% GlutaMax (Thermo, 35050-061) + 1% Penicillin-Streptomycin (Thermo, 15140122)) was added to the dissociated mixture. The mixture was centrifuged at 300g for 5 minutes after which the supernatant was removed, and the cell pellet was incubated in red blood cell lysis buffer for 5 minutes at 37 °C. The mixture was centrifuged, and the cell pellet was resuspended in neuronal medium. The cells were separated from the debris using a BSA gradient method wherein the cell-debris mixture was carefully pipetted onto a bed of 15% BSA (Millipore Sigma, A4503) solution and centrifuged at 280g for 6 minutes at very low acceleration and no deceleration. The supernatant containing the debris was discarded while the cell pellet was resuspended in neuronal medium.

2.4. Culturing endothelial cells and sensory neurons in the device

The microfluidic devices were treated with 1 mg/ml Poly-D-Lysine (Sigma, P7886) for 4 hours followed by washing with distilled water. The devices were then placed at 60 °C overnight for hydrophobic recovery of PDMS. This prevents the leakage of the hydrogel precursor solution from the hydrogel microchannel into the media microchannels during injection. HUVECs (Lonza, C2519A) (P3 - P5) were maintained in EGM-2MV (Lonza, C3302) medium prior to the experiments. The cells were trypsinized (0.05% Trypsin), centrifuged, and suspended in 4 U/ml thrombin (Millipore Sigma, T6634) solution. Separately, 5 mg/ml fibrinogen (Alfa Aesar, J63276) solution was prepared by dissolving appropriate amount of fibrinogen in 0.9% sodium chloride solution. The DRG neurons were mixed with HUVECs at a 1:25 ratio for mouse and 1:100 for human DRG neurons while keeping the HUVECs concentration at 50 million cells/ml and suspended in 4 U/ml thrombin solution. The cell suspension was mixed with fibrinogen solution in 1:1 ratio and injected into the hydrogel chamber of the device. The solution was allowed to polymerize at room temperature for 15 minutes before flushing the media microchannels with EGM-2MV medium supplemented with 50 ng/ml vascular endothelial growth factor (VEGF) (Peprotech, 100–20). After 24 hours, the medium was switched to EGM-2MV medium supplemented with 50 ng/ml VEGF and 2% B27 as this was found to be the optimum medium that supported both the cell populations (Figure S2). The reservoir of the culture was replenished with fresh medium every 24 hours. Same media composition was used across all experimental conditions.

2.5. Stimulation of sensory neurons within the device and calcium imaging

Intracellular cytosolic Ca²⁺ was assessed using sensory neurons loaded with Fura-2 (Thermo, F1221). Fura-2 stock aliquots (2 mM in DMSO) were mixed with Pluronic F-127 in 1:1 ratio followed by 1:500 in Tyrode’s buffer (140 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 10 mM HEPES, 10 mM glucose). The media microchannels were flushed twice with Tyrode’s buffer before replacing with Fura-2 loading solution and incubated at room temperature for 50 minutes. Tyrode’s buffer was used to wash out the remaining Fura-2 and incubated for another 20 minutes prior to treatment (i.e., exposure to chemical stimuli) and imaging. The devices were mounted on the translation stage of an Olympus IX81 inverted microscope (Olympus America). Fura-2 dual excitation and emission was achieved using 340- and 380 nm excitation filters and a 510 nm emission filter, and Olympus UPlan FLN 20X 1.3 NA water immersion objective was used to view the cells. Digital images at 150 ms exposure were recorded with a Hamamatsu EM CCD camera (Hamamatsu Photonics Japan) at an interval of 1 s.

For capsaicin dose response study, Tyrode’s buffer was introduced into the media microchannel to evaluate the baseline response of the neurons in the absence of any chemical stimulation. This was followed by introduction of 0.1 nM, 1 nM, 10 nM, 50 nM, 100 nM, 500 nM and 2000 nM capsaicin solution at 5-minute intervals into the device. Lastly, 140 mM KCl was added to depolarize the DRG neurons to obtain the total number of neurons by observing the cells with ratiometric change in fluorescent signal. Changes in ratiometric intensity of Fura-2 during capsaicin treatment was performed within the regions of interest (ROIs) identified as neurons through KCl treatment. Selective depolarization of DRG neurons in presence of KCl was validated by CellTracker Red CMTPX Dye-labeled DRG neurons and observing the Fura-2 intensity changes within the labeled neurons following KCl treatment (Figure S4). Cells that exhibited >15% change in ratiometric fluorescent signal intensity at 2000 nM capsaicin compared to baseline were considered “responsive cells”. For acidosis experiment, after recording the baseline response to Tyrode’s buffer of pH 7.4, Tyrode’s buffer solutions with different pHs (5.4–7.4) were introduced into the device at 5-minute intervals followed by KCl treatment and cells’ responses were recorded as described above.

2.6. Immunostaining

Cells were fixed by perfusing 4% PFA in the media microchannels for 15 minutes followed by permeabilization in 0.1% Triton X-100 in PBS for 10 minutes, before incubating with blocking buffer for 1 hour in 3% BSA. Primary antibodies corresponding to β-III Tubulin (Novus Biologicals, NB100-1612) (1:500 dilution), CGRP (Bio-Rad, 1720–9007) (1:300 dilution), and TRPV1 (Novus Biologicals, NBP1-71774) (1:100 dilution) were diluted in blocking buffer and incubated at 4 °C overnight. The devices were washed with PBS thrice and incubated at room temperature for 10 min. The fixed cells were incubated with secondary antibodies donkey anti-goat (Alexa Fluor 555, 1:100), donkey anti-rabbit (Alexa Fluor 647, 1:300), and donkey anti-chicken (Alexa Fluor 488 nm, 1:300), and Hoechst for nuclei (1:2000). Samples were washed with PBS and imaged. For characterization of DRG neurons cultured with or without endothelial cells, Alexa Fluor 488-conjugated anti β-III Tubulin antibody was used at 1:100 dilution. For characterization of the vasculature, VE-cadherin antibody (Bio-techne, AF938) (1:100 dilution) followed by donkey anti-goat (Alexa Fluor 647, 1:200) was used. Phalloidin 555 (Thermo, A34055) (1:40 dilution) was used to visualize actin. Quantification of immunofluorescent staining for TRPV1 and CGRP was performed by analyzing βIII tubulin positive neurons. ROIs were determined based on βIII tubulin positive neurons and TRPV1 and CGRP staining intensity was calculated within the ROIs. Image J was used to characterize neurite numbers by creating a maximum intensity projection image of the Z-stack images to visualize neurites from all the planes in one image. Length of the neurites was calculated by measuring end-to-end distance between the terminal end of the neurites and center of the cell body. Vascular networks were characterized by creating a maximum intensity projection image and the branch numbers and vessel diameters were characterized using skeletonize plugin within Image J.

2.7. Statistics

Comparisons between two groups was performed using t-test when the data was normally distributed and using Mann-Whitney’s test when the data was not normally distributed. For comparisons between multiple groups with one variable, one-way ANOVA was used. For dose-dependent response of capsaicin, two-way repeated measured ANOVA was used to determine significance of interaction with capsaicin concentration being the repeated measure. EC50 values were calculated using a non-linear fit to the dose-dependent capsaicin response data. Between 2–4 independent devices and multiple (≥3) regions-of-interest (ROIs) within each device were used per experimental condition. All experiments were independently performed at least twice with DRGs from different mice and human donors. For all analyses, P<0.05 was considered as statistically significant.

3. Results

3.1. Generation of self-assembled innervated vasculature-on-a-chip

The device used to generate self-assembled innervated vasculature system consisted of a hydrogel microchannel with adjacent media microchannels on either side. The hydrogel microchannel was separated from the flanking media microchannels by an array of trapezoidal posts to confine the hydrogel precursor solution in the microchannel while allowing exchange of nutrients via inter-post regions. One million/ml mouse DRG neurons and 125 million/ml endothelial cells (HUVECs) were mixed in 2.5 mg/ml fibrinogen solution and 2 units/ml thrombin and injected into the hydrogel microchannel (Figure 1A). The device was kept in humidified environment for 15 minutes to allow fibrin gel formation before medium (EGM-2MV medium supplemented with 50 ng/ml VEGF-165) was introduced into the adjacent microchannels. The device was cultured for 5 days post seeding the cells, during which endothelial cells self-assembled into microvascular networks (Figure 1A). During the same time course, neurites were found to be sprouted from the sensory neurons and established a dense innervated 3D network. The microvascular networks were characterized by immunostaining for VE-cadherin and assessed for lumen formation using z-stack imaging (Figure 1B,C, Movie S1). The sensory neurons were characterized by using β-III tubulin staining to visualize the soma and the neurites (Figure 1B,C). Innervation of vessels, defined as direct contact of the neurites with the endothelial cells, was confirmed by visualizing co-localization of β–III tubulin and VE-cadherin staining along the walls of the vascular lumens (Figure 1C). Perfusability of the vascular networks was validated by flowing 70 kDa fluorescein isothiocyanate (FITC)-dextran through the vasculature network, which was found to be confined in the vessels during perfusion (Figure S3).

Figure 1.

Development of innervated vasculature-on-a-chip. (A) Schematic depiction of the cell isolation and process of generating innervated vasculature tissue by co-culturing DRG neurons and endothelial cells (HUVECs). (B) Characterization of the innervated vascular network using immunostaining for endothelial (VE-cadherin) and neuronal markers (β-III tubulin) (C) 3D z-stack images of the tissue showing direct contact between the vasculature and neurites. White arrows show the co-localization of β-III tubulin and VE-cadherin Scale: 100 μm for (B) and 50 μm for (C).

3.2. Reciprocal interaction between the sensory neurons and endothelial cells

To understand the effect of co-culture on each cell type, we cultured mouse DRG neurons and endothelial cells under four different conditions— Monoculture of DRG neurons; (ii) monoculture of HUVECs; (iii) co-culture of DRG neurons and HUVECs without direct cell-cell contact; and (iv) co-culture of DRG neurons and HUVECs with direct cell-cell contact. Neurite formation was assessed to determine the effect of endothelial cells on sensory neurons and vessel structure was quantified to assess the effect of sensory neurons on vasculature. Percentage of DRG neurons with sprouted neurites, as well as the number and the length of neurites was found to increase significantly in the presence of endothelial cells in the non-contact co-culture condition as compared to monoculture of DRG neurons (Figure 2A,B). When the two co-culture conditions were compared, the co-culture condition with direct cell-cell contact demonstrated a significantly greater number and length of neurites (Figure 2A,B). Similar to the effect exerted by the vascular cells on neurons, the coculture also showed a significant difference in the vascular network. Specifically, compared to the monoculture of endothelial cells, the diameter of the vessels was found to be significantly higher; albeit only when endothelial cells were co-cultured with DRG neurons in direct contact. No differences were observed in vascular branch density between the three conditions but the area coverage by the vasculature was significantly higher in the direct contact co-culture condition (Figure 2C,D). Given that the contact between the neurons and vasculature displayed more significant differences in neuronal and vascular phenotypes, we used the direct contact co-culture system for subsequent experiments and compared against monoculture of DRG neurons.

Figure 2.

Reciprocal interaction between the endothelial cells and DRG neurons. (A) Representative images showing β-III tubulin staining of DRG neurons in the monoculture, co-culture without direct cell-cell contact (in adjacent microchannels), and co-cultures with direct cell contact. (B) Culture condition-dependent changes in number of neurons that sprouted neurites, number of neurites, and neurite length. N = 3–4 independent devices where each data point represents average of ≥ 5 ROIs from each independent device. (C) Phalloidin staining (red) of the innervated vascular network with β-III tubulin stained DRG neurons (green) (D) Characterization of average vessel diameter, branch density, and % coverage of the microvascular network under different culture conditions. N = 3–4 independent devices where each point represents an average of ≥ 3 ROIs from each independent device. *P < 0.05, **P < 0.01, ****P < 0.0001, n.s. (not significant) Scale: 100 μm. D: DRG monoculture; H: HUVEC monoculture; D&H: DRG and HUVEC non-contact co-culture; D+H: DRG and HUVEC direct contact co-culture.

3.3. Presence of vasculature sensitizes the sensory neurons

We further characterized the effect of co-culture with endothelial cells on the functions of mouse sensory neurons by using capsaicin as a model molecule of noxious chemical stimuli; capsaicin is also an agonist for TRPV1. We introduced capsaicin into one media microchannel and used the pressure difference between the media microchannels to facilitate its perfusion through the vascular network. We performed Fura-2 calcium imaging on the DRG neurons to measure calcium flux in the presence of capsaicin (Figure 3A). The concentration of capsaicin was gradually increased from 0.1 nM to 2000 nM and the response of neurons was examined. Dose-response curve for peak intensity change of Fura-2 ratio showed that the dose-dependent response of DRG neurons within the innervated vasculature was higher compared to ones cultured in the absence of vasculature (p=0.0615) (Figure 3B). The EC50 (concentration of capsaicin required to achieve 50% of maximum response) value of capsaicin in monoculture (181.8 ± 2.23 nM) was also higher than DRG neurons in co-culture with the microvasculature (34.4 ± 1.72 nM). This effect was more pronounced when we analyzed for the responsive neurons (defined as exhibiting >15% change in ratiometric intensity at 2000 nM capsaicin compared to baseline) where the dose-dependent response of the neurons to capsaicin was significantly higher in the presence of microvasculature (p=0.0024) with significant differences between the two culture conditions at 50 nM, 100 nM and 500 nM capsaicin (Figure 3C).

Figure 3.

Presence of microvasculature sensitizes the neurons. (A) Representative images of Fura-2 ratiometric calcium imaging at varying capsaicin (Cap) concentration and KCl at 380 nm as well as corresponding ratiometric intensity (340 nm/380 nm). (B) Normalized average peak response curve for DRG neurons as a function of capsaicin concentration. N = 43 neurons for DRG+HUVEC co-culture condition and N=29 neurons for DRG monoculture condition (C) Normalized average peak response curve for responsive DRG neurons as a function of capsaicin concentration. N = 29 neurons for DRG+HUVEC co-culture condition and N=21 neurons for DRG monoculture condition. (D) Immunostaining for TRPV-1, CGRP and (E), (F) their quantification. N=49 neurons for DRG monoculture and N=80 for DRG+HUVEC co-culture. *P < 0.05, ***P < 0.001, Scale: 50 μm for (A) and 100 μm for (D).

To evaluate if the observed elevated response in co-culture condition could be a result of increased receptor expression in sensory neurons in the presence of vasculature, we stained the neurons for TRPV-1 (capsaicin receptor) and CGRP (secreted neuropeptide) (Figure 3D). The presence of vasculature was found to increase the expression of TRPV-1 (Figure 3E), while no significant changes in CGRP was observed (Figure 3F), which could be due to the transient nature of CGRP secretion. Taken together, these data suggest that the presence of endothelial cells sensitizes the sensory neurons thereby making them more responsive to chemical stimuli such as capsaicin.

3.4. Modeling nociception associated with acidosis

We next assessed the potential of the innervated vasculature to model nociception associated with acidosis. Blood acidosis usually results from disruption in blood flow such as in ischemia. The resulting buildup of lactic acid causes drop in pH of the blood. Other conditions that can cause blood acidosis are poor lung function, cancer, and muscular dystrophy. The low pH is sensed by the acid sensing receptors and acid sensing ion channels present on sensory neurons resulting in acute/chronic pain[43]. To model acidosis, we decreased the pH of the vascular perfusate. Tyrode’s buffer was perfused through the vascular networks by creating a pressure difference between the two media microchannels and the pH of the buffer was gradually decreased from 7.4 to 5.4 to examine a dose response. This broad range is chosen because while the blood pH goes as low as ~6.5 during acidosis[44], other tissues encountering acidosis can experience lower pH such as ~5.4[45]. Fura-2 imaging was performed during the perfusion to measure the response of the sensory neurons to the pH changes (Figure 4A). As shown in Figure 4B,C, the average ratiometric intensity and peak intensity of Fura-2 were found to increase with decreasing pH in a dose-dependent manner. These results highlight that the DRG neurons retain their function within the innervated vascular tissue and can be used as a platform to study acid nociception.

Figure 4.

Modeling acidosis-induced nociception. (A) Representative ratiometric intensity images of Fura-2 at varying pH. White arrows point to cells whose ratiometric intensity changes compared to a higher pH. (B) Representative plot of average ratiometric response curve for DRG neurons as a function of varying pH. N=16 neurons. (C) Average peak ratiometric intensity of DRG neurons for varying pH. N= 11–16 neurons per pH condition. **P < 0.01, ***P < 0.001, ****P < 0.0001. Scale: 50 μm

3.5. Creating innervated vascular tissue from human sensory neurons

We also validated the ability of this platform to support culture of primary DRG neurons from human clinical specimens. The human primary DRGs, resected from consented patients undergoing spinal decompression and fusion for metastatic epidural compression, were dissociated and the DRG neurons were cultured with endothelial cells within the device. Similar to the murine DRG neurons, the sprouted neurites from human DRG neurons innervated the surrounding microvasculature (Figure 5A,B, Movie S2). Despite the lower number of DRG neurons used (~4 fold reduction in cell number compared to murine DRG neurons), the cells remained viable and functional in the device and innervated the vasculature. While we found no difference in neurite length between the monoculture and co-culture (direct contact between the DRG neurons and endothelial cells) of sensory neurons, the number of neurites per cell was significantly higher in the co-culture (Figure 5C,D,E). Expression of TRPV1 on sensory neurons, as determined by immunostaining, was also significantly higher in the presence of microvasculature (Figure 5F,G). These results are consistent with murine DRG neurons suggesting higher sensitization in the presence of vasculature.

Figure 5.

Generation of innervated vasculature-on-a-chip using human DRG neurons (A) HUVECs and DRG neurons co-cultured within the device and imaged at Day 0 and Day 5, which shows the self-assembly of HUVECs into vasculature and neurites sprouting from the neurons. (B) Immunostaining of the innervated vasculature tissue for endothelial (VE-cadherin) and neuronal (β-III tubulin) markers. (B) β-III tubulin staining of the DRG neurons with and without the vasculature and quantification of (C) average neurite length and (D) number of neurites sprouted per cell. N=13 neurons for DRG monoculture and N=20 neurons for DRG+HUVEC co-culture (E) β-III tubulin and TRPV1 staining of the DRG neurons with and without the vasculature and (F) quantification of TRPV1 expression intensity. N=33 neurons for DRG monoculture and N=31 for DRG+HUVEC co-culture. ****P < 0.0001, n.s. (not significant) Scale: 100 μm

4. Discussion

Despite the vital role played by innervation in controlling and modulating the bodily functions, incorporation of nerves has received limited attention in microphysiological models of tissues and organs[3]. In particular, sensory innervation is yet to be introduced into engineered tissues and organ-on-chip systems, which serve as advanced in vitro models for recapitulating tissue composition, architecture, and functions. Lack of such platforms is a major hurdle in understanding the role played by peripheral nervous system in tissue functions and the mechanisms underlying nociception.

In this study, we described the development of an innervated vascular tissue-on-a-chip. The innervated vasculature was engineered by co-culturing endothelial cells and DRG neurons in a microfluidic device. The endothelial cells self-assembled into perfusable 3D microvascular networks and were innervated by the neurites sprouting from the DRG neurons. The endothelial cells and the DRG neurons were found to influence each other, leading to larger vascular lumens and denser neurite formation in co-culture as compared to monocultures. Among the cocultures, direct contact between the DRG neurons and the endothelial cells showed greater number and length of neurites. These findings are consistent with co-cultures involving endothelial cells and motor neurons[46]. The cellular phenotypic differences observed among the two co-cultures (i.e., with and without the direct contact) could be due to the changes in secretory factors or the differences in diffusion profile and stability of the secreted molecules[47,48].

Our results suggesting a crosstalk between vasculature and sensory neurons are supported by several in vitro and in vivo studies. Prior 2D cell culture studies have reported enhanced axonal outgrowth from DRGs in presence of HUVECs, which was attributed to brain derived neurotrophic factor (BDNF) secreted by HUVECs[49]. BDNF, along with other neurotrophic factors, has been reported to contribute to nociceptive function [50,51]. The influence of sensory neurons on endothelial cells has been attributed to CGRP and substance P secreted by sensory neurons which are shown to upregulate VEGFA, Type 4 collagen, Angiopoietin 1 and Matrix metalloproteinase 2 in endothelial cells[52]. ESCs-derived sensory neurons have been demonstrated to regulate angiogenesis in platforms incorporating ESCs-derived endothelial cells[53]. Addition of CGRP has been shown to promote tube formation of HUVECs on Matrigel similar to VEGF supplementation[54]. These in vitro findings are also supported by multiple in vivo studies. For example, CGRP knockout mice have been shown to display impaired angiogenesis as compared to wild type mice[55,56]. In murine models of developmental, blood vessels have been shown to secrete neurotrophic factors such as nerve growth factor (NGF) [57], neutotrophin-3 (NT-3)[58] and artemin [59] which are crucial for survival and function of axons as they innervate the peripheral tissues. Together, these studies highlight the critical role played by reciprocal interactions between vasculature and sensory neurons in regulating their homeostasis and function.

The neuronal networks in the co-culture displayed enhanced response to capsaicin. While the enhanced mass transport facilitated by the vasculature could have an effect in capsaicin response, immunostaining for TRPV1 shows increased expression in the presence of vasculature, which will have a significant effect on capsaicin response. Another factor that could contribute to increased sensitization is the presence of VEGF[60]. Although the culture medium contains VEGF; however, it was used across all culture conditions and therefore its contributions from the medium could be negligible.

Using our platform, we also modeled nociception associated with acid sensing during acidosis. We gradually decreased the pH of the vascular perfusate and found increasing response of sensory neurons with decreasing pH. Acidosis can develop as a result of vascular diseases such as coronary artery disease where impaired blood flow leads to ischemia, hypoxia, and subsequently acidosis due to accumulation of lactic acid[24]. Respiratory conditions such as asthma and pulmonary fibrosis can also lead to blood acidosis due to inability of the lung to remove the carbon dioxide from blood during gas exchange[61,62]. Several types of acute and chronic pain are associated with tissue acidosis[63]. Several proton sensing receptors and ion channels have been identified in sensory neurons such as TRPV1, proton-sensing G-protein-coupled receptors, acid-sensing ion channels (ASICs)[63]. Acidic environment at the tumor site is also a major cause of pain associated with cancer and metastasis[64]. Platforms such as the one described in this study can serve as important tools towards understanding mechanisms driving nociception associated with particular stimuli by employing targeted blocking of receptors and testing candidate drugs.

Finally, we created a human-specific model of innervated vasculature by using primary sensory neurons harvested from human donors and validated some of the key findings from the murine sensory neurons. While rodent models have served as indispensable tools for understanding the signaling underlying nociception, significant differences have been identified between mice sensory neurons and their human counterparts[65]. While human DRGs remain a largely inaccessible cell source, our results show that innervated vascular network can be generated from human DRG neurons with low cell density.

5. Conclusion

In summary, this proof-of-concept study describes the generation of a 3D innervated vasculature system from both murine and human DRG neurons. We leveraged 3D culture of cells along with microfluidics to generate the perfusable vascular networks innervated with DRG neurons. We have studied the reciprocal interactions between the cells and their proximity on DRG neurons and endothelial cells. The ability of the neurons within the innervated vasculature to respond to noxious stimuli was studied using capsaicin and tissue acidosis as model systems. Such devices can serve as invaluable tools for understanding the molecular mechanism underlying pain as well as platforms for testing therapeutics and analgesics. Besides leveraging the perfusable vascular network to examine the effect of blood borne factors on nociception, microfluidic-assisted microphysiological systems are ideally poised to incorporate dynamic mechanical cues[66–69] as well as vascular specific functions such as vasodilation; all these microenvironmental cues are important in nociception[70,71].