3D culture platform of human iPSCs-derived nociceptors for peripheral nerve modelling and tissue innervation

Functional humanized in vitro nerve models are coveted as an alternative to animal models due to their ease of access, lower cost, clinical relevance and no need for recurrent animal sacrifice. To this end, we developed a sensory nerve model using induced pluripotent stem cells (iPSCs)-derived nociceptors that are electrically active and exhibit a functional response to noxious stimuli. The differentiated neurons were co-cultured with primary Schwann cells on an aligned microfibrous scaffold to produce biomimetic peripheral nerve tissue. Compared to glass coverslips, our scaffold enhances tissue development and stabilization. Using this model, we demonstrate that myelin damage can be induced from hyperglycemia exposure (glucose at 45 mM) and mitigated by epalrestat (1µM) supplementation. Through fibrin embedding of the platform, we were able to create 3D anisotropic myelinated tissue, reaching over 6.5 mm in length. Finally, as a proof-of-concept, we incorporated pancreatic pseudoislets and endometrial organoids into our nerve platform, to build nociceptor innervation models. In summary, we propose here an improved tool for neurobiology research that permits pathology modelling, drug screening and target tissue innervation. We show here the development of a 3D sensory nerve model with biomimetic architecture and applicability for pathology modelling, drug testing and target tissue innervation. First, we developed a method to produce, in large-scale, uniform iPSCs-derived trunk neural crest (NC) spheroids (neurospheres) within an agarose mold. The neurospheres can be harvested from the mold and further differentiated into functional nociceptors that exhibit electrical activity and sensitivity to noxious stimuli (resininferatoxin, RTX). To develop a biomimetic nerve model, we co-cultured the neurospheres with primary rat SCs on an aligned microfibrous scaffold, which enhances neural tissue formation in terms of length, alignment and area, and improves tissue stabilization, compared to traditional glass coverslips. Using our 3D co-culture platform, we modelled diabetes-related myelin damage, through acute hyperglycemia exposure, and show that epalrestat is able to mitigate the damage. To upscale tissue formation, we combined the scaffold platform with a fibrin hydrogel, to create anisotropic myelinated axons, with over 6.5 mm in length, densely packed within a 3D matrix. Finally, using the fibrin/scaffold nerve platform, we demonstrate that pancreatic pseudoislets and endometrial organoids can be incorporated to produce nociceptor-innervated tissue models.


Introduction
The ability to detect external noxious stimuli and internal organ dysfunction signals is essential to maintain physical integrity and homeostasis 1 . This process is mediated by nociceptors, which recognize this input via specialized receptors, such as the transient receptor potential vanilloid 1 (TRPV-1), and conduct the information to the central nervous system (CNS), whilst also locally releasing neuropeptides, e.g. substance P, at the site of stimulus 2,3 . The perception of nociceptive pain is contingent on the continuity of sensory peripheral nerve (PN) fibers between target organs and the CNS 4 . However, the PN is a fragile and exposed tissue, prone to damage by trauma or disease 5 . Most cases of PN damage arise from diabetes type II pathophysiological imbalances 6 , which has prompted significant research into prevention and mitigation of diabetes-related disorders as well as strategies for neural tissue repair 78 .
In other pathological situations, the aberrant presence of nociceptors within organs can lead to abnormal and excessive pain. That is the case of chronic pancreatitis (CP), in which the main symptom is severe abdominal pain, caused in part by peripheral sensitization 9,10,11 . Similarly, women with endometriosis, a benign condition affecting up to 10% of reproductive-aged women, experience nociceptive pain because of infiltrating nociceptors within endometriotic implants and inflammatory sensitization of peripheral nociceptors 12 .
To better understand PNs growth, repair methods and involvement in associated pathologies, it is necessary to develop improved neurobiology research platforms. Furthermore, such platforms are instrumental in discovering compounds for neuropathy prevention, pain alleviation and nerve repair. To this end, biomimetic and functional three-dimensional (3D) in vitro models of PNs and tissue innervation can provide a simple, cost-effective and clinically relevant research tool to substitute the expensive and ethically loaded animal models. In this pursuit, few in vitro models using human sensory neurons and nociceptors in particular, have been reported. Wainger et al. proposed a model using nociceptors from reprogrammed fibroblasts that were electrically active and sensitive to noxious stimuli 13 . Similarly, Jones et al. 14 reported a sensory neuron model, with electrically excitable cells, obtained from human embryonic stem cell differentiation. Both models consist, however, of 2D disorganised and non-myelinated cultures, which do not reflect the biological architecture of mature nerves. To achieve myelination, Clark et al. 15 generated co-cultures of induced pluripotent stem cells (iPSCs) derived sensory neurons with primary rat Schwann cells (SCs) on Matrigel-coated coverslips.
However, the use of Matrigel limits clinical applicability (potential batch-to-batch variability) and the resultant tissue morphology is still flat and random. While 3D cultures exhibiting anisotropic neurites from iPSCs-derived motor neurons have already been reported for the creation of motor nerve 16 or innervated muscle platforms 17,18 , 3D biomimetic models with functional sensory neurons remain to be established.

iPSCs differentiation into nociceptors and neurosphere formation
In order to induce iPSCs differentiation into nociceptors, we adapted and modified the protocol published by Chambers et al 19 . Nociceptor induction was initiated using single cell suspension of undifferentiated iPSCs detached with accutase, followed by seeding of 200 cells/microwell in mTESR1 medium supplemented with 10 µm of Y-27632 and 0.5 % Geltrex (in solution) onto 400 µm agarose microwells. Cell suspension was forced to settle by centrifugation at 1200 rpm for 2 min. Afterwards, cells were incubated for 24 h and were given a complete media change with mTESR1 medium. At this time, the cellular spheroid is formed and cell synchronization is initiated by the addition of mTESR1 medium supplemented with 1% dimethyl sulfoxide (DMSO). The cells were maintained for 72 h in the synchronization medium. Post synchronization cells were given a PBS wash and nociceptor induction was initiated by addition of dual SMAD inhibition media containing Advanced RPMI 1640 supplemented with Glutamax (both Thermo Fisher Scientific), 100 nM LDN-193189 (Tocris) and 10 μM SB431542 (Tocris). The spheres were maintained for 48 h in the dual SMAD inhibition media.
Following this, neural crest commitment was induced via media containing Advanced RPMI 1640 supplemented with Glutamax, 3 µM CHIR99021 (Tocris) and 1 µM retinoic acid (Tocris). The spheres were maintained in the neural crest induction media for 5 days with media change every alternate day.
Finally, the neurospheres, composed of trunk neural crest cells, were collected and seeded on coverslips or scaffolds. In these substrates, cells were cultured in neural maturation medium for at least 5 days to reach the nociceptor phenotype. The neural medium is composed of Neurobasal Medium, 0.5 mM Glutamax, 100 U/ml penicillin and 100 μg/ml streptomycin (all Thermo Fisher Scientific), 100 ng/ml human nerve growth factor (NGF), 50 μg/ml ascorbic acid (all Sigma-Aldrich), 25 ng/ml human neuregulin-1 type III (NRG-1 SMDF) and N21 supplement (both from R&D systems).

Dissociated sensory neuron culture
iPSCs neurospheres were collected on day 9 of differentiation in an Eppendorf containing Advanced RPMI supplemented with Glutamax, and dissociated enzymatically with 1ml of Trypsin EDTA for 10 min at 37°C and 5% CO2. Cells were pelleted at 1200 rpm for 7 min and resuspended in neural maturation medium, followed by mechanical dissociation though gentle repeated pipetting. 5 x 10 3 cells/cm 2 were seeded on Matrigel coated (1:200 in DMEM) 35 mm petri dishes (for the patch clamp experiment) or 5 x 10 4 cells/cm 2 on a 6 well plate (EIA experiment) and extra medium was added. The cultures were maintained in the same medium for up to 45 days, at 37ºC, 5% CO2, with medium refreshments every 3 days.

Primary Schwann Cells harvesting, purification and culture
Primary Schwann cells (SCs) were harvested from the sciatic nerves of neonatal Wistar rat pups, following local and Dutch animal use guidelines. Nerve segments were extracted and digested, followed by cell isolation and purification as described by Kaewkhaw et al. 20 . Briefly, the collected nerves were sliced and digested in a 0.05% (wt/vol) collagenase solution for 60 min at 37 °C, 5% CO2. The cell suspension was filtered through a 40 µm cell strainer, centrifuged for 6 min at 400 g, followed by supernatant removal and cell pellet washing with DMEM containing 10% foetal bovine serum (FBS) and 100 U/ml penicillin and 100 μg/ml streptomycin. Cells were centrifuged again at 400 g for 6 min and the supernatant discarded. Finally, cell pellets were re-suspended with 2 ml of Schwann cell proliferation and purification medium, composed of DMEM D-valine (Cell Culture Technologies), 2 mM L-glutamine, 10% (v/v) FBS, 1% (v/v) N2 supplement (R&D Systems), 20 µg/ml bovine pituitary extract, 5 µM forskolin, 100 U/ml penicillin and 100 μg/ml streptomycin, and 0.25 µg/ml amphotericin B (all Sigma-Aldrich). Cells were then plated on 35 mm petri dish pre-coated with 0.01% (v/v) poly-Llysine (Sigma -Aldrich) and 1 µg/ml Laminin (R&D systems) and incubated at 37 °C, 5% CO2 for 7 days. The use of D-valine in place of L-valine was to inhibit fibroblast growth while permitting SCs survival and proliferation. On day 7, 1 ml of fresh medium was added and the medium changed every 2 days until 70 % confluency. Cells were used between passage 3 and 6 (P3-P6).

Scaffold fabrication and sterilization
The scaffolds were fabricated via a two-step electrospinning (ESP) process with a custom-built apparatus. The first step was the production of a release layer by electrospraying a solution of 50% polyethyleneoxide (PEO, Mn = 3350, Sigma-Aldrich) in Milli-Q onto aluminum foil. For this, the solution flowed through a 0.8 mm inner diameter stainless steel needle (Unimed S.A.) at 2 ml/h, while subjected to 20 kV onto a 60 mm diameter mandrel, placed at 10 cm of distance and rotating at 5000 rpm. Afterwards, a nonwoven polyurethane mesh (6691 LL (40 g/m 2 ), provided by Lantor B.V., The Netherlands) was prepared by punching an array of 12 mm circular holes and placed on the mandrel, over the PEO sprayed-foil. We then produced the scaffolds by ESP of 300PEOT55PBT45 (PolyVation) in 75:25 Chloroform/1,1,3,3-hexafluoroisopropanol solution onto the mesh support on the mandrel. For this process, the solution flowed through a 0.5 mm inner diameter stainless steel needle (Unimed S.A.) at 0.75 ml/h, while applying a voltage of 12 kV and at a distance of 10 cm from a rotating mandrel (at 5000 rpm). During both processes, the humidity was maintained at 35-40% and the temperature at 22-24 °C. Finally, we generated individual scaffolds from the polyurethane mesh by punching 15 mm-outer diameter sections concentric to the 12 mm holes, resulting in a thin ESP membrane supported by a polyurethane mesh ring. To detach the individual scaffolds, these were dipped in deionized water and left in PBS until further use. When required for cell seeding, the scaffolds were transferred to a 24 well plate and immersed in 70% ethanol for 1 h for sterilization, followed by several PBS washes and airdrying. These were then maintained in sterile PBS until needed.

Peripheral nerve platform generation
To fabricate our PN platform we followed the process graphically illustrated in Fig. 1. While the iPSCs differentiate and form neurospheres as described above (Fig. 1 b1), we simultaneously seeded the scaffolds with 100 x 10 3 primary SCs and cultured these for 7 days with SC medium. During this time, cells proliferate and aligned with the scaffold fibers to form highly anisotropic SCs bands (Fig. 1 b2). After 7 days, when SCs bands are fully formed, we manually seeded one neurosphere per scaffold ( Fig. 1 b3). After this, we let the neurospheres adhere to the substrate for at least 6 hours, before adding neural medium.
For comparison, we also cultured neurospheres on scaffolds devoid of SCs, but with laminin coating, matrigel coating or no coating at all. For this, we incubated the scaffolds overnight with laminin solution: 100 μl of 1 μg/ml laminin-1 (R&D systems) and 2 μg/ml poly-D-lysine (Sigma Aldrich) in PBS; or matrigel solution: 100 ul of 1:200 dilution of matrigel stock (Fisher Scientific) in DMEM. The following day, we washed off the coating twice with sterile PBS and followed the same procedure for neurosphere seeding. All culture conditions were maintained for 7 or 21 days at 37 °C, 5% CO2, with medium refreshment every other day.

Coverslip cultures
For sensory neuron marker characterization we coated a 12 mm glass coverslip with laminin (same procedure as scaffold coating) and seeded one neurosphere as described above. The neurons were further cultured for 7 days with neural medium to induce maturation.
As a control for the scaffolds, we cultured neurospheres on bare (no coating), laminin, Matrigelcoated or SCs-seeded 12 mm glass coverslips (same culture area as scaffolds). The laminin or Matrigel coating procedure was identical to the scaffolds. SCs were seeded 7 days prior to neurosphere addition with the same cell density used for scaffold seeding (100 x 10 3 cells in total; ~8.3 x 10 3 cells/cm 2 ). The cultures were maintained in neural medium for 7 or 21 days at 37 °C, 5% CO2, with medium refreshments every other day.

Fibrin embedded peripheral nerve platform
To produce a 3D biomimetic PN platform we combined a SCs-seeded scaffold with a neurosphere in a fibrin hydrogel. To achieve this, we followed a similar process as described above and illustrated in Fig.1. First, 100 x 10 3 SCs were seeded on the scaffolds and induced to proliferate for 7 days. Thereafter, one neurosphere was seeded and the medium change to neural medium. After 1 day, we removed the medium and embedded the constructs into a fibrin hydrogel, composed of 3.5 mg/ml human fibrinogen (Enzyme Research Laboratories), 5 U/ml thrombin (Sigma-Aldrich) and 2.5 mM CaCl2. After gelation (~15 min), neural medium containing 100 μg/ml aprotinin was added. The cultures were maintained for 7 or 21 days at 37ºC, 5% CO2 with medium refreshments every other day.

Pancreatic pseudoislets generation and innervation
To innervate pancreatic pseudoislets, we first generated spheroids composed of a mixture of alpha TC1 clone 6 cells (ATCC, CRL-2934) and INS1E cells (AddexBio), then added these onto a neurosphere seeded scaffold. To do this, we started by expanding alpha TC cells in medium composed of DMEM, 15 mM HEPES, 0.1 mM non-essential aminoacids, 2 g/L glucose (all Thermo Fisher Scientific), 10% FBS, 1.5 g/L sodium bicarbonate (both Sigma-Aldrich), 0.02% BSA (VWR) and

Endometrium spheroids generation and innervation
The human endometrial adenocarcinoma cell line Ishi-M3-HSD-A is derived from Ishikawa cell line after genetic modification to introduce the luciferase fused with the green fluorescent reporters genes 21 . Cell line Ishi-M3-HSD-A used in the present experiment was authenticated by Short Tandem Repeat (AmpFlSTR IdentifilerTM PCR Amplification Kit; Thermo Fisher Scientific) profiling and tested negative for mycoplasma (MycoAlert, Promega).
Cells were maintained in standard monolayer culture as described earlier 21 , using RPMI 1640 medium (Thermo Fisher Scientific) supplemented with sodium-pyruvate, L-glutamine, penicillinstreptomycin and 5% FBS at 37°C with 5% CO2 in humidified air. For organoid formation, cells were detached using with Accutase (Thermo Fisher Scientific), pelleted and resuspended in ice-cold Matrigel (Becton Dickinson). Droplets of 25 µl Matrigel suspension were pipetted into a 6-wells plate to form domes containing 2000 cells each. Matrigel domes were allowed to polymerize for 15 min at 37°C, 5% CO2 and were subsequently covered with pre-warmed endometrium organoid culture medium (RPMI 1640, sodium-pyruvate, L-glutamine, penicillin-streptomycin and 5% fetal bovine serum). Medium was refreshed twice a week, while monitoring the organoids condition.
For the innervation experiments, we carefully picked up the Matrigel domes with a sterile spatula and transferred to a laminin-coated scaffold containing one 7-day grown neurosphere. The spheroids were left to adhere overnight, after which we added 300 μl of fibrin hydrogel composed of 3.5 mg/ml human fibrinogen (Enzyme Research Laboratories), 5 U/ml thombin (Sigma-Aldrich) and 2.5 mM CaCl2. Cultures were maintained for 10 days with medium of neural/endometrium at 1:1 ratio, containing 100 ng/ml NGF, at 37ºC, 5% CO2.

RNA extraction, reverse transcription, and quantitative RT-PCR
Harvested neural spheroids or cells were lysed in Trizol (ThermoFisher Scientific) and total RNA was isolated using an RNeasy Mini Kit (QIAGEN). The RNA was reverse-transcribed with random primers and iScript™ cDNA Synthesis Kit (Invitrogen). Quantitative PCR was carried out using a Real-Time PCR System (Biorad) and iQ SYBR Green Supermix for qPCR (Biorad). Relative mRNA expression levels were analyzed by the ΔΔCT method and normalized to GAPDH gene expression.
Three replicates were used per condition. Detailed primer information is provided in table S1.

Quantification of substance P released from iPSCs-derived nociceptors after resiniferatoxin stimulation
Dissociated neurosphere cultures were prepared as described before and cultured for 10 days.
Before changing the medium for the experiment, the cells were washed thoroughly with PBS. Then, 250 µl of control medium (normal neural medium), neural medium containing 100 nM resininferatoxin (RTX) (Alomone Labs) or neural medium containing 100 mM potassium chloride (KCl) were added to the cultures. These were conditions incubated at 37ºC, 5% CO2 for 5 min with gentle agitation. The supernatant was collected from each condition and immediately stored at -80ºC. To quantify the amount of substance P, we used the human substance P EIA kit (Phoenix PharmaceuticalsT, EK-061-05), and followed the manufacturer protocol. This experiment was performed once and five replicates per condition were used.

iPSCs-derived nociceptors morphology after resiniferatoxin and capsazepine stimulation
Neurospheres were seeded on laminin-coated coverslips as described above and cultured for 7 days. After this period, the culture medium was changed to control medium (normal neural medium), neural medium containing 1% ethanol (EtOH, vehicle control), neural medium containing 10 μM RTX or neural medium containing 10 μM RTX and 100 μM capsazepine (CPZ) (Sigma-Aldrich) (both with 1% EtOH). To evaluate the neurite morphology, the initial 24 h of culture was tracked with brighfield microscopy and after this period the cells were fixed for subsequent immunostaining to βIII-tubulin. At t = 24 h cell viability was also quantified with the PrestoBlue™ Cell Viability Reagent (Thermo Fisher Scientific). Minimum of five replicates were used per condition.

Brightfield microscopy for live cells
To capture micrographs of live cells, we used the Nikon Eclipse TI-E microscope with an Okolabs environmental control. For the tracking of neurite morphology during the RTX/CPZ experiment, we transferred the coverslips containing the cell cultures to a 35mm petri dish (Ibidi) and added control medium or medium supplemented with the drugs. The images were acquired immediately, taking frames every 5 min for 24 h.

Hyperglycemia test
The PN platform (without fibrin embedding) composed of SCs and one neurosphere was fabricated as described above and cultured for 21 days to allow mature myelin formation. At that point the medium was changed to either control medium (normal neural medium), hyperglycemic medium composed of normal medium supplemented with 45 mM glucose (Sigma-Aldrich, G7021) or hyperglycemic medium plus 1 μM epalrestat (Sigma-Aldrich, SML0527). Cultures were kept at 37 °C, 5% CO2 for additional 48 h and then fixed with 4% paraformaldehyde for 20 min at room temperature.
Following this, the samples were prepared for immunostaining and TEM as described below. Five replicates were used per condition.

Immunocytochemistry and fluorescence microscopy
Samples were fixed with 4% paraformaldehyde (PFA) for 20 min at room temperature (RT), rinsed thoroughly with PBS, and left in PBS until further use. Samples were permeabilized for 30 min at RT with 0.1% Triton X-100 in PBS, followed by rinsing with PBS and blocking with blocking buffer composed of 5% goat serum, 0.05% Tween-20, and 1% bovine serum albumin (BSA) in PBS, overnight at 4ºC, under mild agitation. Afterwards, samples were incubated overnight at 4 °C with primary antibody solutions in blocking buffer. The next day, the samples were washed with a wash buffer composed of 0.05% Tween20 and 1% BSA in PBS and incubated for 2 h at RT with secondary antibody solutions in wash buffer. Following this, we rinsed the samples with PBS, stained with DAPI (0.2 µg/ml) for 10 min at RT, and left them in PBS until imaging. The detailed list of used primary and secondary antibodies can be found in table S3 and S4 respectively. Images were acquired using either an inverted epifluorescence microscope (Nikon Eclipse Ti-e) or a confocal laser scanning microscope (Leica TCS SP8).

Transmission electron microscopy
Samples were prepared by fixation with 4% PFA in PBS, followed by washing with 0.

Image Analysis
3D image reconstructions and neurite/myelin volume measurements were processed with Amira (Thermo Fisher Scientific). All other images were prepared and analyzed using Fiji software (https://fiji.sc/). To quantify the different parameters of tissue morphology, we obtained and analyzed images of the whole sample. To measure the orientation degree of fibers, neurites and myelin, we used the OrientationJ plugin 22 and applied the Measure function over circular ROIs that capture the whole tissue to obtain the coherence values (where 0 is full isotropy and 1 is full anisotropy). To measure the neurite length, we used the Simple Neurite Tracer plugin 23 , and measured the distance between the cell bodies and the edge of the respective axons. Cell counts were performed using the standard Analyze Particles function to DAPI + objects. To measure the axonal area, we first converted images of βIII tubulin + cells to binary images and measured the pixel area occupied by the neurites, excluding cell bodies. Then, we divided this value by the total area of the scaffold. For the myelination area, we measured the pixel area of MBP + segments and divided this value by the area of the scaffold. G-ratio measurements were done using the G-ratio plugin (http://gratio.efil.de/) 24 . Myelin decompaction analysis was carried by first measuring the area below the most outer ring and the most inner ring (expected myelin area). Then, the image was converted to binary image and the same areas were determined (actual myelin area). To determine the decompaction area we used the following formula: (1 -(expected myelin area / actual myelin area)) x 100. For all experiments we used at least 5 biological replicates per condition. For the tissue morphological analysis (neurite alignment, axonal area and myelin area) we imaged and analyzed the whole sample. For neurite length quantification we took at least 15 measurements per sample. For myelin morphometric analysis (g-ratio and myelin decompaction) we took at least 5 measurements per sample.

Statistical Analysis
The graphs were built and the data was analyzed using the software GraphPad Prism. Bar graphs are shown as mean ± SD and boxplots represent data point between the minimal and maximal value.

Generation of human nociceptor neurospheres from iPSCs
To generate human nociceptor neurospheres we followed a two-step process detailed in Fig.2a.
First, human iPSCs displaying a pluripotency phenotype (Fig. S1) were seeded on a 400 μm agarose microwell mold (200 cells / microwell; 1580 microwells) and the cells synchronized via DMSO treatment for 3 days. We then followed a differentiation protocol adapted from Chambers et al. 19 , in which human iPSCs were driven towards nociceptors, via a neural crest intermediate, in a quick and efficient manner. At day 2, and for 5 days, retinoic acid was supplemented together with CHIR99021 to promote cell aggregation. Retinoic acid supplementation is essential for the integrity of the spheroids, as its absence leads to poorly formed and loose clusters (Fig. S2b). At day 7, cells reached the neural crest stage denoted by the presence of SOX10 + cells (Fig. S2a). To distinguish between cranial and trunk neural crest cells, we immunostained cells for ETS1, a cranial marker, and Phox2B, a trunk marker 25 .
As shown in Fig. 2b, most cells were Phox2B + rather than ETS1 + , evidencing commitment to the trunk phenotype. To confirm this, we also analysed gene expression at different time points (Fig. 2c). At day 2, the expression of both markers was null. After 3 days, cells displayed a robust expression of ETS1, while Phox2B expression was still null. However, after 4 days of differentiation, we observed an inversion in marker expression, as cells exhibited an upregulation of Phox2B and downregulation of ETS1. Finally, at day 7 cells expressed only Phox2B, while ETS1 expression was null. At day 7 and for 2 days, we inhibited the notch signalling pathway through DAPT and SU5402 supplementation. At this stage, multiple uniform and cohesive spheroids ( Fig. 2d and Fig. S2c) with an average diameter of 308.7 ± 38.6 μm were achieved. For the second part of the process (at day 9 of differentiation), the neurospheres were harvested from the mold, seeded in a substrate of choice, and cultured with NGFcontaining medium for at least 7 days, to finalize the nociceptor differentiation protocol, whilst growing neurites in situ. To validate the acquisition of a nociceptor phenotype, we cultured the neurospheres on laminin-coated coverslips for 7 days and immunostained for the nociceptor-specific markerssubstance P, CGRP and TRPV-1and for the sensory neuron transcription factor -BRN3A. As shown in Fig. 2e, all markers were expressed within grown neurons and located either along the length of the axons (substance P, CGRP and TRPV-1 correlated with βIII-tubulin) or in the nucleus (BRN3A correlated with DAPI). In sum, we were able to generate a large number of even-sized nociceptor neurospheres in simple and quick manner.

Neuron function testing on nociceptor neurospheres
Nociceptor neurospheres were either dissociated to single cells (Fig. 3a-c) or kept intact ( Fig.   3d-g) in order to validate neuron function, i.e., electrical excitability, and nociceptor function in particular, i.e. responsiveness to noxious stimuli. Whole-cell patch clamp measurements from dissociated neurospheres showed a mean membrane capacitance of 16.2 ± 4.2 pF, a mean resting membrane potential (RMP) of -44.1 ± 7.1 mV. All probed neurons (n = 11) were excitable and able to fire APs with a mean AP amplitude of 67.4 ± 26.2 mV and an overshoot peak of 23.3 ± 28.5 mV (n = 8) (Fig. 3b). To probe sensitivity to noxious stimuli, we measured the secretion of substance P after stimulation with RTX, a TRPV-1 agonist and analog of capsaicin 26 . As exhibited in Fig. 3c, RTX exposure led to an immediate release of substance P from the nociceptor population. Specifically, RTXstimulated neurons displayed a 2.12-fold increase of released substance P per DNA (p < 0.05) compared to unstimulated neurons. KCl-depolarized neurons showed an increment of 1.6 times compared to RTXstimulated neurons (p < 0.01) and 3.8 times compared to unstimulated cells (p < 0.0001). Literature reports have shown that capsaicin exposure can lead to nerve fiber retraction and local denervation, after topical application of capsaicin on the skin 27 . Because of this, we explored if RTX exposure would lead to neurite retraction and if that effect could be mitigated by blocking the TRPV-1 channel with the antagonist capsazepine (CPZ) 26 . To this end, the neurite morphology of nociceptor neurospheres was monitored for 24 h after the application of RTX alone or with RTX together with CPZ and compared with unstimulated cells or cells stimulated with vehicle medium (neural medium with 1% EtOH). As visible in Fig. 3d, the neurites of neurospheres stimulated with vehicle ( Fig. 3d left column and movie S1) or RTX plus CPZ (Fig. 3d right column and movie S3) remained unaffected and showed little movement, except for the natural movement of live neurons. Contrarily, when stimulated with RTX only (Fig. 3d middle column and movie S2), we could observe an immediate neurite retraction (first 15 m), followed by a lag period and neurite regrowth. As a consequence, after 24 h of culture in these conditions, neurites of RTX-stimulated neurospheres were significantly shorter than untreated cultures (p < 0.001), vehicle control (p < 0.01) and RTX/CPZ-stimulated neurospheres (p < 0.05) (Fig. 3d and   3f). Notably, neurite length in vehicle control samples and RTX/CPZ-stimulated neurospheres was not significantly different (p > 0.05) than untreated cultures. Finally, we also measured cell metabolic activity after 24 h of culture under these conditions and observed that there was no significant difference (p > 0.05) among them, evidencing that despite neurite retraction, the neurons remain viable after RTX exposure (Fig. 3g). In sum, the differentiated nociceptors were electrically active and could sense and react to noxious stimuli. were used per condition. The boxplots in F) represent data points from the minimum to maximum value.

Fabrication of 2D PN platform: 7 days culture
After characterization of nociceptor functionality, we developed a 3D biomimetic PN platform composed of a neurosphere/SCs co-culture on an aligned microfibrous scaffold (Fig. S3a). The scaffold is composed of aligned microfibers with an average diameter of 1.37 ± 0.20 μm and coherence (alignment degree) of 0.74 ± 0.04. To investigate the impact of the scaffold, we also cultured cells on glass coverslips and evaluated the performance of both substrates after 7 days of culture (Fig. 1b2). The scaffold fibers generated aligned and elongated bands of SCs, in stark contrast to the flattened and isotropic morphology of SCs cultured on glass coverslips (Fig. S3b).
Neurospheres at day 9 of differentiation were placed on SC pre-cultured scaffolds/coverslips (one neurosphere per substrate) to establish the co-culture system. The effect of exogenous SCs on neurite growth and morphology was compared to coverslips/scaffolds that were uncoated, laminincoated or matrigel-coated (Fig. 4). Uncoated coverslips did not even permit neurosphere attachment, while uncoated scaffolds allowed both neuron attachment and neurite outgrowth. This growth was however significantly shorter than all scaffolds with any form of coating (p < 0.0001) (Fig. S4a). When coverslips were pre-coated with laminin, matrigel or SCs, the neurospheres could attach and grow neurites (Fig. 4a, left column). The neurosphere morphology depended on the coverslips coating, as the cluster remained intact when coated with laminin ( Fig. 4a top left image) or seeded with SCs ( Fig. 4a bottom left image), but not when coated with Matrigel ( Fig. 4a middle left image). In the latter, neurons dissociated from the cluster, proliferated and migrated outwards. As a result, we could not measure the axonal area, due to an unclear contrast between cell bodies and neurites. For the other two conditions, neurites were clearly emanating from the neurosphere and covered a total area of 3.6 ± 2.4 % (laminin coating) or 3.4 ± 1.4 % (SC seeding), not significantly different (p > 0.05) compared to uncoated scaffolds (2.4 ± 1.0 %). When comparing neurite length, only coverslips with SCs were significantly higher compared to all other conditions (p < 0.0001) showing a mean length of 644.7 ± 258.2 μm (Fig.   4B). Regarding neurite alignment (Fig. 4c), coherence measurements indicated isotropic growth for all conditions with no significant differences between substrates.
By comparison, neurospheres cultured on scaffolds showed significantly increased (p < 0.0001) neurite length and overall neurite alignment compared to coverslips with equivalent coatings (Fig. 4b   and 4c). Scaffolds coated with Matrigel exhibited the longest neurite length (3071.0 ± 1568.0 μm) and highest coherence value (0.58 ± 0.09). SC-seeded scaffolds performed better than no coating or Laminin-coated scaffolds, showing a mean neurite length of 1612.0 ± 491.0 μm compared to 544.3 ± 189.3 μm for no coating and 966.1 ± 378.2 μm for laminin and an alignment of 0.38 ± 0.06 in contrast with 0.20 ± 0.05 for no coating and 0.38 ± 0.12 for laminin; interestingly, scaffolds with no coating showed a statistically similar neurite length compared to growth on a SC-coated coverslip (p > 0.05), indicating that the lowest performing scaffold is equivalent to the optimal coverslip condition after 7 days of culture. Regarding axonal area, a Matrigel coating led to a significantly larger scaffold coverage (13.3 ± 6.4%; p < 0.001) compared to laminin or SCs (3.4 ± 1.9 % and 3.9 ± 2.4 %, respectively), which was similar to coverslips. Scaffolds with no coating, resulted in a neurite area of 2.4%, the lowest value from all scaffold conditions. SCs that were pre-seeded on either scaffolds or coverslips achieved an intimate association with neurites, although with clear differences regarding overall tissue organization (Fig. S3c). In coverslips ( Fig. S3c top images), neurite projection was radial, isotropic and disorganised. Moreover, SCs tended to migrate towards the neurosphere, grouping around it in high density. In contrast, SCs seeded on scaffolds maintained their aligned morphology (Fig. S3c bottom images) and neurites appeared to follow the directional cues provided by the fiber topography and aligned SCs, leading to an organised and anisotropic pattern. In sum, scaffolds preserved the neuron clustering and promoted longer and more aligned neurite outgrowth compared to coverslips. After 21 days of culture, both neurite length and area increased compared to day 7. In coverslips cultures, SCs gradually led to neurosphere dissociation, similarly to the Matrigel-coated coverslips, resulting in a neural network of multiple interconnected clusters; this hampered further axonal area quantification for both conditions. Only the laminin-coated coverslip preserved neurosphere integrity ( Fig. 5a left column), with a neurite area of 12.4 ± 7.1 % (Fig. 5d) and the longest neurite growth (mean length of 1335.8 ± 772.8 μm) of all coverslip conditions (Fig. 5b). From day 7 to day 21 (Fig. S5), we detected a 3.5-fold increase in growth for the laminin condition, whereas Matrigel condition only imparted a 1.15-fold increase and the SCs condition registered a 0.69-fold decrease. This suggests that the maintenance of the neurosphere integrity is beneficial for enhanced neurite outgrowth, as cluster dissociation leads to the formation of a network with reduced outgrowth. Despite differences in growth, there was a generalized lack of preferred orientation with no differences in neurite alignment among conditions (Fig. 5c).
All scaffold cultures, including uncoated scaffolds, exhibited an increase in length and axonal area compared to the earlier time point (Fig. 5a right column, Fig. S4b and Fig. S5). In terms of neurite length this temporal increment corresponded to 1.61 times for uncoated, 2.2 times for laminin, 1.25 times for Matrigel and 1.9 times for SCs samples (Fig. S5). Compared to coverslips cultures, the neurite length (Fig. 5b) (Fig. 4b and 4c).
In the early stages (up to 7 days), SCs promote a more accelerated neurite outgrowth compared to laminin coatings, but this growth tends to slow down as SCs begin to myelinate axons. Since myelination results in thicker and straighter segments 28 (Fig. 6a and Fig. S6), this hypothetically explains why we detected the highest neurite alignment for SC-seeded scaffolds (0.44 ± 0.16) (Fig 5a). All scaffold conditions at 21 days in vitro (DIV) led to increased (p < 0.05) alignment compared to similarly coated coverslips.
For both coverslips and scaffolds, myelination was observed by day 21 as an abundance of MBP + segments that spatially correlated with βIII-tubulin + structures (Fig 6). In coverslips, the myelin was disorganised, randomly oriented and partially overlapping (Fig. 6a, Fig. S6a and Fig. S6b). These myelinated cultures were mechanical unstable, attributed to the high lipid content that results in poor surface attachment (data not shown). On the other hand, co-cultures in scaffolds resulted in long, straight and anisotropic myelin bundles (Fig. 6a bottom image), with overall great mechanical stability. The presence of other mature myelin proteins, such as myelin protein zero (P0) was also detected (Fig. S6c).
In brief, scaffolds promoted a continuous neural growth and long-term stability of the cluster, as well as longer and more aligned neurites than coverslips. Myelination was also enhanced and more anisotropic in scaffold co-cultures. neurite area (% of total area). The boxplots represent data points between the minimum and maximum values. The graphs bars represent mean value ± SD. In all experiments, a minimum of 5 replicates were used per condition, and the experiments were performed twice. Statistics were performed with two-way ANOVA followed by a Tukey's HSD post-hoc test, where ****p < 0.0001, ***p < 0.001, **p < 0.01, ns denotes p > 0.05. N.A. means not applicable. On the scaffold group, the stars with no associated bar represent comparisons with the no coating condition. . For imaging analysis we took at least 5 images per sample. Statistics were performed with an unpaired t-test, where ****p < 0.0001 and ***p < 0.001.

Hyperglycemia model and drug testing
After establishing the myelinated co-culture scaffold platform, we investigated hyperglycemiainduced damage on myelin and evaluated the mitigating effect of epalrestat, an aldose reductase inhibitor 29 (Fig. 7a). After 21 days of co-culture, the cells were cultured for 48 h in just hyperglycemia (45 mM glucose) or hyperglycemia plus epalrestat (glucose at 45 mM plus epalrestat at 1 μM); normal neural medium was used as control. As shown in Fig. 7b, control cultures (left column) and those supplemented with epalrestat (right column) appeared normal, with no significant damage. Contrarily, hyperglycemia cultures (middle column) showed signs of myelin disruption, particularly visible by MBP immunostaining (in red). TEM micrographs of control cultures (Fig. 7c left image) revealed compact myelin layers with no particular irregularities. In hyperglycemia cultures (Fig. 7c middle image), it was particularly evident that myelin layers separated from each other, exhibiting a decompacted structure.
Finally, myelin morphology in hyperglycemia with added epalrestat condition (Fig. 7c right image) resembled control cultures, with most layers well compacted and with no signs of abnormalities. TEM images were quantified according to g-ratio, a standard parameter, and myelin decompaction percentage, based on an established algorithm 30 . G-ratio measurements showed that both control and glucose plus epalrestat groups produced similar results, with a g-ratio value of 0.66 and 0.63, respectively (Fig. 7d).
The glucose-only group produced a lower value of 0.46 (p < 0.0001), due to myelin layer decompaction.
This was reflected in the myelin decompaction measurements, where the control group had the lowest decompaction (8.4%), the hyperglycemic condition had the highest decompaction (22.2%), and the condition of hyperglycemia with epalrestat registered decompaction value (13.8%) that was equivalent to control and significantly lower than hyperglycemia alone (p < 0.05).

Development of a large PN biomimetic platform
To develop a more representative model that better emulates the 3D hierarchical microarchitecture of a PN, scaffolds containing neural and glial tissue were embedded in a fibrin hydrogel. In the 3D image reconstructions shown in Fig. 8a, scaffolds with no hydrogel generated a anisotropic and planar neurite growth (Fig. 8a left column) with only a small volume increase from day 7 to day 21, from 0.0025 μm 3 to 0.0039 μm 3 , respectively (Fig. 8b). When co-cultures were supplemented with fibrin, neurite growth maintained directional alignment (similar to all conditions and time points, Fig. S7) but was notably multiplanar (Fig. 8a right column), with still a small increase overtime but comparatively larger than bare scaffolds at both 7 DIV (~3 fold; 0.0078 μm 3 ) and 21 DIV (~2 fold; 0.0094 μm 3 ) (Fig. 8b and movie S4). Fibrin embedding also promoted a significant increase (p < 0.001) of myelin volume at day 21 compared to scaffold-only co-cultures (Fig. 8c), with a mean myelin volume of 0.0012 μm 3 versus 0.00046 μm 3 , respectively. This was also reflected in histological crosssections (Fig. 8d), which showed a larger number of myelin rings in fibrin-scaffolds than in bare scaffolds.
To showcase the long-term development and stability of this fibrin-embedded platform, we maintained the co-cultures for 35 days and evaluated the resulting tissue morphology. The engineered neural tissue retained its 3D organisation and exhibited highly aligned myelinated axons that achieved over 6.5 mm in length in all growth planes (Fig. 8e, 8f), which is the largest reported value in literature, to the best of our knowledge. showing the formation of highly aligned and long (over 6.5 mm) neurites throughout the platform. For E) and F) βIII-tubulin is shown in green, MBP in red and DAPI in blue.

Target tissue innervation
To demonstrate the potential of the fibrin PN platform to create innervated tissue models, we show a proof-of-concept using pancreatic pseudoislets (Fig.9 a-d) and endometrial organoids ( Fig.9 eh). Pancreatic pseudoislets were generated with a mean diameter of 41.9 ± 2.98 μm (Fig. S8a and S8b), composed of alpha (alpha TC1) and beta cells (INS1E) at a ratio of 3:7 31 . Neurospheres were cultured on laminin-coated scaffolds for 7 days to initiate neural growth, after which several pancreatic pseudoislets were added and the entire construct embedded in fibrin. The co-cultures were maintained for an additional 10 days, during which time the pseudoislets survived and nociceptors from the neurospheres extended to surround and innervate them (Fig. 9c, 9d). Further reflecting native rodent pancreatic tissue, alpha (glucagon + ; in green) and beta cells (insulin + ; in red) reorganised and formed segregated clusters according to cell type 32 . Immunostaining for substance P (white) revealed that nociceptor axons were able to penetrate the cluster and establish an intimate association with the pancreatic cells, particularly insulin + cells.
To create an endometrium model, a GFP + human endometrial adenocarcinoma cell line (Ishi-M3-HSD-A) 21 was seeded within Matrigel domes to form organoids of regular size, with a mean diameter of 59.9 ± 12.1 μm (Fig. 9e-f; S8c-d). Matrigel domes containing several organoids were picked and positioned on a laminin-coated scaffold close to a 7 DIV neurosphere. After an additional 10 days, we confirmed that the endometrial organoids attached, survived, and retained their spherical shape throughout the co-culture period (Fig. S9). Nociceptors innervated the surrounding endometrial organoids, with neurites able to penetrate through the Matrigel domes and establish an intimate association with endometrial cells (Fig. 9g, 9h, and S9).

Discussion
Nociceptive pain perception is essential for normal organ function and physical integrity maintenance. However, sensory nerves can be affected by a multitude of insults, ranging from trauma to disease-induced neuropathies, which hinder this sensorial ability. At the same time, nociceptive pain is a common symptom of a large range of pathologies, bringing discomfort and reduced life quality to diseased individuals. Current understanding of neuropathies and pain mechanisms in a pathological context is limited by current research tools, which consist of animal models or oversimplified 2D in vitro models. Animal models are expensive, difficult to assess and are not always clinically translatable to humans 33 . In vitro models, on the other hand, can provide an inexpensive, simple and direct translational research platform 34 . To date, some in vitro human nociceptor models, containing neurons that are able to elicit APs and are sensitive to noxious stimuli, have already been reported 13,14 . However, the tissue in these models does not replicate the 3D anisotropic axonal morphology that characterizes the PN (Fig. 1a). Moreover, the reported neurons are unmyelinated, which represents only the slow transmission nociceptors (C-fibers) and not the myelinated fast transmission fibers (Aδ) 1,2 . This lack of proper architectural and cellular representation limits the translational potential of the models and demands for improved biomimetic PN platforms.
In this work, we demonstrate the fabrication of a 3D culture platform containing functional iPSCs-derived nociceptors neurospheres with myelinated or unmyelinated anisotropic neurites. To produce these neurospheres we modified the protocol described by Chambers et al. for accelerated conversion of iPSCs into nociceptors 19 . In our two-step method, we started by seeding iPSCs on an agarose microwell device that induces the formation of spheroids via cell self-aggregation. Because of this, we were able to generate a large number of uniformly sized cell clusters (over 1500) that mimic the dorsal root ganglion (DRG) morphology and can be easily picked and handled manually ( Fig. 1b and   2D). Contrarily, other nociceptor differentiation protocols culture cells on flat surfaces, producing disorganised and non-homogeneous cultures 19,35 . Before commencing differentiation, we adopted a strategy for cell synchronization via DMSO treatment, which has been reported as an effective method to arrest the cell cycle 36 . Following this, the differentiation protocol was initiated and after 7 days, cells exhibited a trunk neural crest phenotype ( Fig. 2b and 2c). Contrary to Chambers et al. protocol, we supplied retinoic acid after day 2, which has been reported as a promoter of trunk neural crest specification 25 and enhanced the spheroids cohesion (Fig. S2b). At day 9 of differentiation, the neurospheres were harvested for the second part of the process, where these were placed on the desired culture substrate and cultured for at least 7 days, to simultaneously promote maturation to a nociceptor phenotype and neurite outgrowth on the substrate. The use of neurotrophic factors in this last stage was reduced to just NGF, which proved enough to promote growth and phenotype acquisition. In just a threeweek period, we were able to obtain several neurospheres (over 1500) exhibiting characteristic nociceptor markers such as substance P, CGRP and TRPV-1 expression (Fig. 2e). The obtained neurons were electrically active, presenting a RMP value slightly higher than other reported iPSCs neurons 37,38 , but all able to elicit APs when stimulated (40 DIV cultures; Fig. 3a; table S2). These differentiated nociceptors also released substance P in response to a noxious stimulus (RTX) (10 DIV cultures; Fig.   3B), denoting the presence of a functional TRPV-1 channel 26,35 . For these experiments, we used dissociated neuron cultures in order to improve the access of the electrical probe to single cells and ensure that substance P release was unhindered by the cell cluster. Clustered nociceptors, representing a more biomimetic form, also revealed functionality, as shown by the reversible RTX-induced neurite retraction from 7 DIV neurospheres (movie S2). RTX (or capsaicin) activation of TRPV-1 promotes calcium influx, which in turn can lead to mitochondrial dysfunction and inhibition of metabolism, resulting in the collapse of nerve endings 39 . This effect is observed on the skin, where after topical capsaicin treatment, epidermal nociceptor fibers are reversibly lost 27 . We could replicate this process using RTX, and additionally, neurite retraction could be significantly reduced by co-addition of a TRPV-1 antagonist (CPZ) (Fig. 3d, 3e, 3f; movie S3). After 24 h, all conditions showed similar metabolic activity (Fig. 3g), evidencing that nociceptors neurites can be targeted, without loss of viability, using a RTX sub-toxic application (10 µM).
After establishing nociceptor functionality, we developed a PN model that replicates the native morphology using an aligned microfibrous scaffold (Fig. 1b), which we previously reported as suitable to induce anisotropic neurite and myelin formation from rat DRGs and PC12 cells 30 . Unlike other strategies that rely on platform wall constraining to induce axonal alignment 16,17 , we simply used a substrate whose fibers are highly efficient in directing parallel axonal growth. Moreover, the fibers coating can be easily customized, which endows the platform with high versatility regarding the chemical and cellular composition. To recreate a PN tissue, we pre-seeded the scaffold with SCs, which proliferated and organised in highly aligned cell bands within 7 days (Fig. S3b). These SC bands replicate the native bands of Büngner, which form in the regenerating PN, to stimulate and guide regrowing axons to their targets 40 . As a cell source, we used the rat sciatic nerves, from which SCs can be extracted in high yield, are easily purified 20 , and have been shown to efficiently myelinate human iPSC-derived sensory neurons 15 . To better evaluate the influence of SCs in promoting neural growth, we compared them with scaffolds coated with laminin (major nerve ECM protein) 41 and matrigel (assortment of ECM proteins) 42 , which are standard coatings in in vitro PN models 5 . Matrigel-coated scaffolds promoted the highest neural growth followed by SCs-containing scaffolds, which additionally were composed of vast and anisotropic myelinated neurites. Compared to glass coverslips, our scaffold promoted higher neurite length, alignment and area after 7 and 21 days of culture. (Fig. 4 and Fig. 5).
Scaffolds also displayed larger amounts of myelin and more aligned segments than coverslips cultures ( Fig. 6) and, unlike coverslip coatings, all scaffold coatings maintained an intact neuron cluster that permitted neurite outgrowth comparable to a DRG explant. Uncoated substrates, i.e. not containing any adhesive units, such as laminin, are traditionally not supportive of neuron attachment, survival and growth 43 . However, uncoated scaffolds supported neural growth (Fig. S4) while uncoated coverslips could not even allow cell attachment. This can be explained by the scaffold favourable chemical and topographical properties. On one hand, PEOT/PBT does not require functionalization to allow nerve growth 44 , contrary to glass 43 . On the other hand, the aligned microfibers promote cell entrapment and induce anisotropic neural growth 45,46,47 , contrary to flat isotropic substrates. In sum, our scaffold provided a superior substrate than glass coverslips, by promoting PN tissue development with higher efficiency, robustness and biomimicry level.
Using the scaffold co-culture system, we produced myelinated tissue to investigate the resulting damage from acute hyperglycemia exposure. Hyperglycemia is a common pathophysiological imbalance resulting from diabetes mellitus type II and a cause of peripheral neuropathy 7,48 . In this situation, excess intracellular glucose is converted to sorbitol by aldose reductase, resulting in sorbitol accumulation, and consequently in increased cellular osmolarity, oxidative stress and mitochondrial dysfunction that lead to cellular damage 8,48,49 . Hyperglycemia damage is characterized by axonal degeneration and myelin abnormalities, such as layer decompaction 50,49 (Fig. 7a). Thus, morphological evaluation can be used to determine the presence of damage, and if so, if it can be prevented by drug supplementation. Among the existing drugs, aldose reductase inhibitors such as epalrestat, have been explored as a method to reduce sorbitol levels, with a positive influence in mitigating hyperglycemic damage 8,29 . To achieve an acute hyperglycemic state, supraphysiological glucose concentrations (45 mM) can be supplied in the culture medium, allowing a rapid modelling of blood glucose spikes 30 . In our experiment, acute hyperglycemic exposure caused axonal and myelin damage, morphologically similar to diabetic type II mice 8 , which was quantitatively manifested in increased layer decompaction and decreased g-ratio (due to layer separation), compared to control cultures. Supplementation of epalrestat to hyperglycemic cultures effectively mitigated the cellular damage, denoting a benefit of sorbitol reduction in damage prevention (Fig. 7). These results highlight the advantages that in vitro platforms provide in comparison to animal models, by permitting a rapid, simple and inexpensive but still accurate modelling of pathologies, as well as testing of therapeutic compounds. Additionally, in diabetic animal models there are several physiological processes altered at once, causing a systemic damage that reduces the level of experimental control and undermines data reliability 50,51 .
To develop a more biomimetic PN model, we embedded the cell-seeded scaffold in a fibrin hydrogel. Fibrin is a natural material present during nerve regeneration and widely used within PN conduits and models 52,53 . Its addition permitted neurites to grow beyond the scaffold, resulting in a neurite and myelin volume enhancement compared to bare scaffolds, while maintaining anisotropy ( Fig.   8 and movie S4). This suggests that the topographical guidance provided by the scaffold, at the bottom of the construct, is sufficient to induce an overall neurite alignment, in a process potentially governed by axonal paracrine signalling 30,54,55 . Using the fibrin/scaffold PN platform, we demonstrated the ability to create myelinated tissue with long-term stability and exhibiting neurite dimensions (over 6.5 mm) that, for in vitro platforms, are the highest reported to date, to the best of our knowledge 16,56 .
Finally, as a proof-of-concept, we build different nociceptor innervation models by incorporating target tissues within the fibrin/scaffold PN platform. As targets, we selected the pancreas, represented by pancreatic pseudoislets, and the endometrium, represented by endometrial organoids.
Both the pancreas, during CP 9,10 , and the endometrium, during endometriosis 57,58 , are innervated by unmyelinated nociceptors, which can become sensitized and trigger nociceptive pain. However, the exact pain pathophysiological mechanisms remain to be elucidated and current research is exclusively conducted in animal models, due to a lack of representative in vitro models. In this work, we established culture conditions that permitted simultaneous survival of all tissues and nociceptor ingrowth towards the target tissue, within 10 days of co-culture (Fig. 9). We did not further explore the mechanisms of nociceptor innervation or sensitization, as we intended to show the versatility of this platform in generating various nociceptor innervation models. With further development, this platform can provide a simple and inexpensive research tool to understand tissue-specific pathophysiology of pain as well as safely and quickly screen analgesic compounds.
In summary, we demonstrate here the formation of a 3D biomimetic nociceptor platform that can be used to assess neural growth in fine-tuned microenvironments, to model diabetes-related pathologies and to produce innervation models. We believe that the platform here proposed could be a new highly relevant tool for neuroscience, and in general for biomedical sciences.

Data availability
The authors declare that all data supporting the findings of this study are available within the paper and in Supplementary Information, and are also available from the corresponding author upon reasonable request.