Tissue Engineered Innervated Cardiac Units: A Three-Dimensional Platform Recapitulating the Spatial Organization of Cardiac Sympathetic Innervation

1. Introduction

The rhythm and growth of the heart are profoundly influenced by the autonomic nervous system. Sympathetic and parasympathetic neurons both innervate developing cardiac tissue and orchestrate its maturation and physiology in complementary ways [1,2,3]. Sympathetic innervation promotes cardiomyocyte proliferation, electrical conduction maturity, and increases in heart rate and contractile force during both development and adulthood [1]. Parasympathetic innervation exerts an opposing but equally crucial influence by modulating heart rate and coordinating atrioventricular conduction, thereby refining the developing heart’s rhythm [4]. Cardiac neurons and cardiomyocytes undergo co-maturation: signals from nascent cardiomyocytes regulate the growth and neurotransmitter properties of innervating neurons, while neuronal inputs accelerate cardiomyocyte electrophysiological and metabolic maturation [4]. Disruption of this interplay is associated with abnormal cardiac development and heightened disease susceptibility; genetic ablation of sympathetic guidance cues produces hypo- or hyper-innervation together with arrhythmia and impaired cardiac performance [5,6], underscoring the need to model autonomic influences on the heart in vitro.

Cardiac innervation is highly organized, with sympathetic and parasympathetic neurons occupying distinct anatomical niches [7]. Parasympathetic preganglionic fibers from the vagus nerves project to the heart and synapse onto small intrinsic cardiac ganglia embedded within the surface plexuses [8]. In contrast, sympathetic innervation derives predominantly from postganglionic neurons in the cervical and upper thoracic sympathetic chain—notably the bilateral stellate ganglia—which send long axons to innervate the cardiac chambers [9]. This produces a topographic pattern in which vagal inputs are relatively localized to the pacemaking and atrial regions, whereas sympathetic fibers broadly innervate both atria and ventricles.

This spatial segregation is critical for coordinated autonomic regulation. Because sympathetic somata reside in extracardiac ganglia, a single ganglion distributes a common signal across a wide myocardial territory through bundled long-range axons, allowing a graded, regionally biased chronotropic and inotropic response rather than uniform stimulation [7,9]. Conversely, the intracardiac location of parasympathetic ganglia places cholinergic output immediately adjacent to nodal tissue, supporting rapid, spatially restricted rate control [8]. Segregation also separates the site of neuromodulator synthesis from the site of release, so that transmitter delivery is determined by axon terminal distribution rather than by somatic position, and it preserves the polarity that allows retrograde target-derived trophic signals to reach distant neuronal cell bodies [5,10]. An in vitro model aiming to recapitulate cardiac neurobiology must therefore preserve this spatial pattern—discrete ganglionic sources and long-range projections—to authentically mimic cardiac innervation.

Despite the biological importance of cardiac innervation, most existing in vitro platforms for studying neuro-cardiac interactions are highly simplified and planar. These cocultures lack the structured architecture of the in vivo heart and provide little control over neuronal subtype or spatial arrangement. In many cases, investigators have resorted to heterogeneous or surrogate cell sources, establishing rodent–human hybrid systems that raise concerns due to species differences in maturation and signaling [11,12,13]. Moreover, without guided patterning, the neurons in these setups may not faithfully represent a purely sympathetic or parasympathetic phenotype, as neural cell lines (e.g., SH-SY5Y neuroblastoma cells) or generic induced neurons are often used without robust differentiation into a cardiac-autonomic subtype. This lack of control over cellular phenotype and organization can lead to inconsistent neuromodulatory effects.

Microfluidic systems have significantly improved spatial segregation in neuro-cardiac cocultures by compartmentalizing neuronal and cardiac cells into distinct chambers connected via microchannels that permit directed axonal growth. Human-induced pluripotent stem cell (hiPSC)-derived neurons and cardiomyocytes have been cultured in such devices, most notably in a cardiac innervation chip, in which compartmentalized microfluidics enabled human-on-human neuro-cardiac innervation and functional coupling in an electrophysiologically active system [14]. Early prototypes using polydimethylsiloxane (PDMS) microchips incorporated rat sympathetic neurons and cardiomyocytes separated by microconduits, demonstrating that electrical stimulation of the neuronal compartment could modulate cardiomyocyte beating rates in a frequency-dependent manner [12,15]. Additional platforms integrating hiPSC-derived autonomic neurons with ventricular-like hiPSC cardiomyocytes have used microfluidic designs to dissect neuronal influences on contractility and calcium handling [14,16]. However, these systems largely remain planar or quasi-three-dimensional, confining cellular organization to two dimensions within thin channels and consequently failing to reproduce the volumetric tissue architecture, axon fasciculation, or three-dimensional (3D) polarity observed in vivo.

Recognizing the limitations of two-dimensional (2D) culture, efforts have shifted toward 3D-engineered tissues and organoids that include both neurons and cardiomyocytes [17,18]. While 3D culture offers a more biomimetic microenvironment, current neuro-cardiac 3D models still fall short of recapitulating native heart innervation. A common drawback is the absence of spatial segregation between neuronal and cardiac compartments. In many cardiac organoids or engineered heart tissues, neurons—when present at all—are randomly distributed or intermingled with cardiomyocytes rather than residing in distinct ganglion-like clusters. This blurs the polarity of neuronal projections: instead of extending long axons to distant targets, neurons in unguided 3D cocultures form only short local neurites, so long-range axonal tracts and the clear delineation of axon terminals from cell bodies are rarely observed. Compartmentalized devices address the geometry but not the outcome; even when neurons and cardiomyocytes are physically separated and must connect purely by axon outgrowth, the axons that cross often remain sparse and fail to establish neuromodulation strong enough to alter cardiomyocyte activity, a shortfall that has been particularly persistent in fully human systems where neuronal maturation is slow [19]. A second limitation of current cardiac organoids is incomplete architectural development. Despite some self-organization into chamber-like structures, proper cardiac septation or multi-chambered anatomy is rarely achieved in vitro [20], and human embryonic stem cell-derived heart organoids typically form a single ventricular cavity or primitive heart tube with minimal septation [21,22]. Together, these gaps create a need for a platform that introduces true spatial organization of neurons relative to cardiac tissue, including long-distance aligned axon projections, while supporting the maturation of a more anatomically representative cardiac tissue.

To address these limitations, we developed Tissue Engineered Innervated Cardiac Units (TE-ICUs), which combine tissue engineering and developmental cues to recreate the anatomical and functional coupling of cardiac muscle and nerves. TE-ICUs are fabricated in a modular fashion (Figure 1) and comprise rodent-derived sympathetic ganglia (Figure 1C) seeded at one end of a methacrylated hyaluronic acid (MeHA) hydrogel-based hollow microcolumn (Figure 1A). These ganglia extend long, aligned axonal tracts that innervate rodent embryo-derived cardiac tissue aggregates, termed cardiac units (Figure 1B), located at the opposite end. This configuration establishes neuron-to-cardiac spatial segregation. The sympathetic neuron cell bodies remain clustered within the ganglion source, while their axons are guided to grow longitudinally through the hydrogel column toward the distant cardiac tissue. The resulting architecture mirrors the parallel, bundled sympathetic nerve pathways that innervate the heart in vivo. The TE-ICU design builds on our group’s prior development of micro-tissue engineered neural networks (µTENNs), 3D neural microtissues encased in hydrogel-based hollow microcolumns in which long axonal tracts span the full length of the column, recreating the native architecture of long-range projection pathways such as corticospinal [23], nigrostriatal [24], or peripheral sensorimotor circuits [25]. Here we describe the fabrication of TE-ICUs and evaluate their capacity to recapitulate cardiac autonomic anatomy and physiology. We first characterize cellular morphology, phenotype, contractile behavior and transcriptional responses in planar neuro-cardiac cocultures, in which all gene expression analyses in this study were performed, and then characterize the structure and function of the 3D TE-ICU platform.

2. Results

2.1. SCG-Derived Neurons Exhibit a Robust Sympathetic Phenotype and Innervate Cardiac Aggregates in Planar Coculture

Pieces of SCG isolated from postnatal rats (Figure 1C) were cultured as aggregates on laminin-coated planar surfaces for 10 days in vitro (DIV) (Figure 2A). Immunofluorescence imaging of SCG-only cultures revealed extensive neurite outgrowth from the ganglia. These neurites stained positively for tyrosine hydroxylase (TH) and β-tubulin III, indicating a robust sympathetic neuronal phenotype (Figure 2B). Subsequent coculture of SCG with cardiac tissue aggregates revealed neuronal processes projecting toward and intermingling with cardiac aggregates stained for cardiac troponin T (cTnT), confirming structural contact between sympathetic axons and cardiomyocytes (Figure 2C–F). High-magnification images showed TH-positive fibers penetrating the cardiac tissue aggregates. Beating frequency was assessed by video microscopy at 3 and 5 days after plating of the cardiac aggregates. Cocultured cardiac tissue exhibited a trend toward an increased rate of contraction by 5 DIV, although this difference did not reach statistical significance (Figure 2G).

2.2. Sympathetic Neurons Upregulate Contractile, Conduction and Adrenergic Genes in Planar Neuro-Cardiac Coculture

All transcriptional analyses in this study were performed on planar neuro-cardiac cocultures. RT-qPCR was performed on total RNA isolated from planar cardiac-only cultures and planar neuro-cardiac cocultures to assess transcripts involved in contractility, conduction, adrenergic signaling, cardiac identity, and synaptic machinery (Figure 3). Freshly isolated cardiac tissue from E16 rat embryos was analyzed as the DIV 0 reference sample, and expression was normalized to the housekeeping gene Gapdh.

After 7 days of planar culture (DIV 7), innervated cocultures showed significantly higher transcript levels than the freshly isolated DIV 0 cardiac tissue for the actin-related genes Actc1 and Actn2, the adult myosin isoform Myh6, the β1-adrenergic receptor Adrb1, and the gap junction genes Gja5 (Cx40), Cx43 and Cx45 (Figure 3). The same genes were also significantly elevated in DIV 7 cocultures relative to DIV 7 cardiac-only cultures (Gja5, p < 0.0001; Actc1, Actn2, Adrb1, Cx43 and Cx45, p < 0.01; Myh6, p < 0.05); by contrast, DIV 7 cardiac-only cultures did not differ significantly from DIV 0 tissue for any transcript examined. The presence of sympathetic neurons was therefore associated with a coordinated upregulation of contractile, conduction and adrenergic transcripts over the culture period, whereas cardiac tissue cultured alone remained transcriptionally indistinguishable from freshly isolated tissue.

The cardiac transcription factor Nkx2-5 and cardiac troponin T (Tnnt2) followed the same directional pattern; Nkx2-5 was significantly higher in cocultures than in cardiac-only cultures at DIV 7 (p < 0.05), although neither transcript differed significantly from DIV 0 tissue. No significant differences were detected for the fetal myosin isoform Myh7, the presynaptic marker Snap25, or for TH relative to DIV 0 tissue. Notably, TH remained low in both DIV 0 tissue and DIV 7 cardiac-only cultures, whereas TH levels in the neuro-cardiac cocultures at DIV 7 significantly exceeded those of the cardiac-only group (p < 0.05), reflecting the contribution of SCG-derived sympathetic neurons to the total RNA pool (Figure 3).

2.3. Sympathetic Ganglia Project Functional Noradrenergic Axons in 3D Hydrogel Microcolumns

SCG explants were embedded at one end of MeHA hydrogel-based hollow microcolumns and maintained in culture for at least 30 days. Axons extended longitudinally along the inner walls of the microcolumn in a highly directed manner, with a mean neurite length of ~2 mm by 4 DIV increasing to ~6.0 mm by 17 DIV (Figure 4A–C). Repeated-measures ANOVA yielded a significant effect of time on both neurite length (p < 0.0001) and growth rate (p = 0.0381), with a growth rate of ~0.3–0.55 mm/day sustained throughout the culture period (Figure 4C,D).

Fast-scan cyclic voltammetry (FSCV) was performed at 30 DIV to assess the functional competence of these 3D sympathetic cultures (Figure 4E–G), confirming that the microtissues remain metabolically active well beyond the period over which axon growth was quantified. The ganglion region was electrically stimulated, and the resulting catecholamine release was recorded with a carbon fiber electrode. Peak release reached ~170 nM (Figure 4F). The corresponding cyclic voltammogram displayed an oxidation peak at ~0.67 V and a reduction trough near −0.3 V, consistent with the electrochemical signature of catecholamines such as norepinephrine (Figure 4G). These data establish, as a proof of concept, that 3D sympathetic microtissues respond to stimulation by releasing evoked neurotransmitters.

2.4. TE-ICUs Recapitulate Three-Dimensional Neuro-Cardiac Anatomy

TE-ICUs were fabricated by seeding cardiac aggregates at one end of a MeHA hydrogel microcolumn and SCG explants at the opposite end, and the microtissues were visualized periodically by phase microscopy. Axons were observed projecting from the SCG by 10 DIV; these subsequently bundled into fascicles and reached the cardiac unit (Figure 5). In the ~5 mm construct shown in Figure 5, a physical connection between the two aggregates was established by 14 DIV. Because axon extension proceeds at a relatively constant rate (Section 2.3), the time required to bridge the column scales with its length: shorter columns were connected within several days, whereas the longest constructs (up to 10 mm) required correspondingly more time. This tunability is a feature of the microcolumn format, allowing innervation distance to be set independently of cell source.

Cardiac units within the microcolumns underwent progressive self-organization. Chamber-like cavities emerged around 10 DIV and further resolved into distinct compartments by 14 DIV, reminiscent of rudimentary myocardial partitioning (Figure 5). Axonal bundles remained attached to the cardiac units and maintained structural integrity even while the cardiac units contracted spontaneously within the microcolumn (Video S1).

2.5. Sympathetic Innervation Accelerates Cardiac Beating in TE-ICUs

Confocal microscopy of innervated TE-ICUs revealed TH-positive axons projecting linearly along the column and penetrating the cardiac aggregates, which stained positively for cTnT (Figure 6A,B). Serial z-stacks were merged to visualize the spatial segregation between neuronal and cardiac domains and the established polarity of axonal projection within the column. Control microcolumns containing a cardiac aggregate alone were confirmed by phase microscopy and cTnT immunolabeling (Figure 6C,D); no intracardiac TH signal was detected in these non-innervated constructs.

Contraction rate was quantified as beats per minute at 5 and 8 DIV in microcolumns containing both sympathetic and cardiac aggregates, as well as in cardiac-only microcolumns. Two-way ANOVA revealed significant effects of both culture type (p = 0.0034) and time (p = 0.0089) on contraction rate. Sidak’s multiple comparisons test identified a significant difference between innervated and cardiac-only constructs at 8 DIV (p < 0.01), at which point innervated cardiac units beat at ~24 beats/min compared with ~13 beats/min for cardiac-only controls. At 5 DIV the innervated group showed a higher mean contraction rate (~14 versus ~9 beats/min) that did not reach significance after correction for multiple comparisons (Figure 6E).

2.6. Cardiac Aggregates Do Not Influence Sympathetic Neurite Outgrowth, and Innervated Constructs Show a Trend Toward β-Adrenergic Dependence

To test whether the presence of a cardiac target influences sympathetic axon growth, microcolumns were prepared with SCG explants alone (SCG-only) or with both SCG and cardiac aggregates (SCG-Cardiac). Neurite length and growth rate were quantified 4 and 7 days after addition of the SCG to the microcolumn. Neurite length increased from ~2.0–2.4 mm at 4 DIV to ~3.8–4.1 mm at 7 DIV in both groups, and neither length nor growth rate differed significantly between SCG-only and SCG-Cardiac constructs at either time point (Figure 7A,B).

To probe whether the accelerated beating of innervated constructs depends on β-adrenergic signaling, contraction rate was measured before and after acute application of the β-adrenergic antagonist propranolol (1 µM, 5 min; Figure 7C). Cardiac-only constructs were unaffected by propranolol, whereas innervated SCG-Cardiac constructs showed a reduction in mean contraction rate. Expressed as the ratio of beats after to beats before treatment, cardiac-only constructs gave a ratio of ~1.13 ± 0.28 while SCG-Cardiac constructs gave ~0.80 ± 0.19 (mean ± SEM; Figure 7D). This difference was not statistically significant given the variability across constructs, but the direction of the effect is consistent with a component of the innervation-associated increase in beating rate being mediated by tonic β-adrenergic drive from the sympathetic neurons.

3. Discussion

The TE-ICU described here is a 3D neuro-cardiac platform that reproduces the structural organization of cardiac sympathetic innervation and supports functional coupling between the two tissue compartments. Coculture of cardiac units and sympathetic ganglia within MeHA hydrogel microcolumns produced noradrenergic axonal bundles that spanned the length of the column to innervate the target cardiac unit, chamber-like partitioning of the cardiac unit, and a significantly increased spontaneous beating rate. In parallel planar cocultures, sympathetic innervation was associated with sustained expression of genes encoding contractile and gap-junction proteins.

3.1. Cell Sourcing for Neuro-Cardiac Models

Multiple studies have described neuro-cardiac coculture across planar and 3D platforms. Obtaining appropriate cellular biomass, particularly of autonomic neurons, has been a major limiting factor and a source of variability, with studies relying on neuronal cell lines such as PC12 [12,26,27] or SH-SY5Y [11], or on hybrid cocultures combining neuronal and cardiac cells from different species. Human iPSC lines have emerged as a reliable source of sympathetic and parasympathetic neurons that exhibit structural interactions with and modulation of cardiac activity [16,28,29,30]. However, deriving autonomic neurons from hiPSC lines is non-trivial, time-consuming and expensive, and such protocols often yield low numbers of pure neurons with limited functional maturation.

We therefore used primary sympathetic neurons isolated from the SCG of postnatal rat pups (Figure 1C). Plated on planar surfaces, these explants exhibited robust axonal outgrowth and prominent TH expression, indicating good cellular health and a sympathetic phenotype (Figure 2). Separately, we generated cardiac units from whole heart tissue of E16 rat embryos using the forced aggregation method previously reported by our group [31]. This method produces aggregates structurally similar to spheroids or organoids while affording control over cell number, cell type, and aggregate size. We have previously reported that force-aggregated cardiac tissue exhibits coordinated en masse contraction and stronger local field potentials than dissociated cultures, indicating enhanced functional maturation [31].

3.2. Temporal Matching of Cardiac and Sympathetic Populations

Beyond cell sourcing, the timing of cell seeding matters because innervation occupies a defined window during cardiogenesis. The developmental stages of the heart are highly conserved across species, and functional sympathetic innervation of the rat heart does not occur until E16–E17 [32]; in mice, the first signs of cardiac autonomic innervation appear around E15.5 [33]. We therefore plated cardiac units derived from E16 rat embryos 1 day before introducing SCG-derived sympathetic neurons, so that the cardiac tissue would be present before neuronal arrival, approximating the sequence of innervation in vivo. We emphasize that this is an approximation of developmental sequence rather than a reproduction of it: the interval between cardiac seeding and axonal arrival at the target is set by the geometry of the culture rather than by developmental signaling, and we did not characterize the maturation state of the cardiac unit at the moment sympathetic axons reached it. Cocultures were maintained in nerve growth factor (NGF)-rich medium, consistent with reports of NGF’s requirement in sympathetic cultures. Planar neuro-cardiac cocultures demonstrated innervation of the cardiac unit by sympathetic axonal projections from the SCG, accompanied by a trend toward increased cardiac beat rate (Figure 2).

3.3. Transcriptional Consequences of Sympathetic Coculture in Planar Cultures

Understanding the transcriptional consequences of neuro-cardiac interaction is essential for uncovering how autonomic innervation shapes cardiomyocyte development, electrophysiological competence, and long-term functional integration. Previous studies examining the genetic impact of neuro-cardiac coculture have been limited in scope, often focusing on a narrow set of structural or electrophysiological genes and using rodent–human mixed systems [13]. Our study expands this space by profiling a broader panel of contractile, conduction, and adrenergic transcripts in rodent cardiac tissue cocultured with rodent sympathetic neurons.

In planar coculture, innervated cardiac tissue at DIV 7 showed significantly higher levels of the contractile protein genes Actc1 and Actn2 than both freshly isolated E16 tissue and cardiac-only cultures, while the structural transcript Tnnt2 trended upward without reaching significance, a profile compatible with the enhanced contractility observed in these cocultures (Figure 3). We also assessed Myh6 and Myh7, which encode the α- and β-myosin heavy chain isoforms, respectively [34]. In rodents, Myh7 is highly expressed in fetal and neonatal ventricles and is gradually downregulated postnatally, whereas Myh6 becomes the predominant adult isoform, associated with faster ATPase activity and increased contractile performance [2,34,35,36]. This isoform switch reflects the transition from an energy-efficient fetal phenotype to a more mature, force-generating myocardial state, and re-expression of Myh7 is a documented marker of pathological hypertrophy and normal aging in rodents [37]. In our planar cocultures, Myh6 was elevated relative to both DIV 0 tissue and DIV 7 cardiac-only cultures without a corresponding rise in Myh7, suggesting that the coculture environment supports ventricular-like transcriptional maturation without inducing a stress-associated reversion to the fetal gene program (Figure 3).

The developing heart possesses intrinsic adrenergic activity before sympathetic innervation [38,39], mediated by region-specific and transient expression of enzymes including TH. Intrinsic cardiac TH expression is scattered from ~E11.5, increases markedly in the interventricular septum around E16.5, and declines by E19.5 as functional sympathetic innervation is established [38]. In our hands, TH transcript in E16 cardiac tissue was low at DIV 0 and remained low over 7 days of cardiac-only culture, consistent with the intrinsic adrenergic program already being in decline by this stage (Figure 3). Planar neuro-cardiac cocultures at DIV 7 instead showed significantly higher TH than cardiac-only cultures, which we attribute to TH transcript contributed by the SCG neurons present in the same RNA pool rather than to maintenance of the intrinsic cardiac adrenergic program.

The gap junction genes Gja5 (Cx40), Cx43 and Cx45, which encode proteins central to electrical coupling, were also elevated in innervated planar cocultures relative to both DIV 0 tissue and DIV 7 cardiac-only cultures, suggesting improved intercellular connectivity and conduction potential. These observations are broadly consistent with Kowalski et al. [13], who reported sympathetic neuron-mediated upregulation of sarcomeric and ion channel genes in hiPSC-derived cardiomyocytes. That study observed no change in Cx43, whereas Cx43 was among the transcripts elevated in our planar cocultures. The comparison should be drawn cautiously, as the two systems differ in species, cell source, and neuronal preparation. One possible explanation is that intact ganglia provide a denser and more sustained noradrenergic input than dissociated neurons do, but our data do not test this directly. Adrb1 was significantly elevated in innervated cocultures relative to both DIV 0 tissue and DIV 7 cardiac-only cultures, indicating that sustained sympathetic contact upregulates the receptor transcript itself and not only downstream β-adrenergic responsiveness, in contrast to prior in vitro work in which functional responsiveness was detectable without a corresponding transcriptional change [40,41]. Our propranolol experiments (Figure 7C,D) speak to this point: acute β-blockade reduced the beating rate of innervated constructs by ~20% on average while leaving cardiac-only constructs unaffected, a trend consistent with tonic β-adrenergic drive from the sympathetic neurons, although the effect did not reach significance in our sample. The cardiac transcription factor Nkx2-5 was likewise higher in cocultures than in cardiac-only cultures, consistent with preservation of cardiac identity under innervation, whereas the presynaptic marker Snap25 did not differ across groups, indicating that the transcriptional shift is not simply a function of added neuronal RNA. Overall, these data indicate that structured sympathetic coculture supports transcriptional maturation of cardiac tissue and enhances molecular features underlying excitation–contraction coupling.

3.4. Engineering 3D Sympathetic Networks with Long-Range Noradrenergic Projections

A principal aim of this study was to engineer 3D neuronal microtissue in which a discrete sympathetic population sends long, aligned, noradrenergic axonal projections to innervate cardiac tissue. Our group has developed µTENNs to restore or replace damaged neural circuits. These constructs comprise hydrogel-based hollow microcolumns with phenotypically specific neuronal populations segregated at opposite ends and connected by long axonal tracts that span the microcolumn. We have fabricated and characterized such microtissues using cortical [23,42,43,44], dopaminergic [24,45], and motor and sensory neurons [25], and applied them to the reconstruction of brain pathways and spinal tracts, to neuromodulation as living electrodes, and as peripheral neuromuscular interfaces [23,24,25,46]. Here, we extended this approach to a 3D sympathetic network by placing SCG from postnatal rat pups inside a MeHA microcolumn (Figure 1). SCG within the 3D hydrogel produced rapid, highly directed axonal projections that spanned the microcolumn by 17 DIV (Figure 4). This directed outgrowth contrasts with the radial pattern of planar SCG cultures (Figure 2) and resembles in vivo postganglionic projections toward target organs. These 3D sympathetic networks robustly expressed TH and released transient peaks of norepinephrine (~170 nM) in response to electrical stimulation during FSCV (Figure 4). This concentration is within the range of electrically evoked catecholamine transients measured by FSCV in rodent brain tissue, though the two preparations differ substantially in geometry and electrode–source distance and the comparison is therefore indicative only. Notably, most hiPSC-derived sympathetic cultures either do not exhibit detectable norepinephrine secretion or produce very small amounts, indicating limited functional maturation [40,41].

3.5. Spatial Organization and Cardiac Self-Organization in TE-ICUs

The heart receives sympathetic innervation from axonal projections emanating from the stellate ganglia, a fusion of thoracic and cervical ganglia. During development, these axons grow along the vasculature, passing through the subepicardial layers toward intracardiac targets, including the sinoatrial node, atrioventricular node, and ventricular myocardium. In contrast, parasympathetic input arises from localized intracardiac ganglia on the cardiac wall. Reflecting this spatial relationship, we constructed TE-ICUs by placing cardiac units and SCG-derived sympathetic aggregates at opposite ends of MeHA hydrogel microcolumns 3–10 mm in length. Cardiac aggregates within the microcolumn self-organized into chamber-like cavities reminiscent of the partitioning reported in some human pluripotent stem cell-derived cardiac organoids (Figure 5); formation of internal cavities is considered a key indicator of cardiogenic modeling [21,22]. We have previously observed that dissociated cardiac cultures do not exhibit such chamber-like structures [31]. Within TE-ICUs, sympathetic aggregates extended long aligned projections that fasciculated into thick axonal bundles reaching the cardiac unit (Figure 5). These projections did not break as the cardiac units contracted, indicating durable structural neuro-cardiac coupling (Figure 5, Video S1). At terminal time points, TH+ sympathetic terminals were detected penetrating the cardiac units (Figure 6A,B), whereas no intracardiac TH signal was observed in non-innervated cardiac microtissue (Figure 6C,D), in line with the low intrinsic cardiac TH expression measured in non-innervated planar cultures (Figure 3). TE-ICUs also showed a significant increase in cardiac beating rate at 8 DIV, whereas the corresponding increase in planar coculture remained a non-significant trend (Figure 6E, Figure 2G).

3.6. Cardiac Target Influence on Axon Growth

Rather than being a passive recipient of innervation, the developing heart actively induces and patterns sympathetic growth through vascular guidance structures, bioelectrical feedback and expression of axon guidance molecules [5,6]. Among these, NGF, produced by cardiac myocytes and vascular cells, is the most critical molecule that directs sympathetic innervation [10]. We anticipated that the presence of a beating cardiac unit in TE-ICUs would accelerate axonal outgrowth from the SCG. Instead, cardiac units had no measurable effect on the rate of sympathetic outgrowth along the microcolumn (Figure 7A,B). We attribute this to the high NGF concentration in the medium used for both cardiac-only and neuro-cardiac groups, which may have saturated the trophic requirement and masked any cardiac-derived contribution.

3.7. Limitations

To our knowledge, this is the first report of an in vitro system that models cardiac autonomic anatomy and recapitulates the spatial pattern of cardiac sympathetic innervation, though several limitations should be noted. First, the model lacks parasympathetic input. Although parasympathetic neurons have been derived from human iPSC lines, their isolation from rodent primary tissue is not well established, so we restricted this study to sympathetic effects. Second, we combined E16 embryonic cardiac tissue with postnatal sympathetic neurons, and we acknowledge that this mismatch in maturation state is a limitation. It reflects a practical constraint: postnatal cardiac myocytes are difficult to isolate, require harsher tissue digestion, and survive poorly beyond 2–3 days in culture, while isolating SCG from E16 rat embryos is extremely challenging given their small size and limited anatomical access. Third, TE-ICUs could not be built with fully consistent architecture across constructs. Sympathetic axons occasionally projected along unintended trajectories, and aggregates sometimes shifted position along the column or migrated out of it, contributing to variability in the structural and functional readouts. Fourth, the transcriptional analyses were performed on total RNA from planar cocultures, which limits our ability to resolve cell-specific contributions; the elevated TH signal in cocultures is the clearest instance of this ambiguity. Finally, gene expression was not assessed in the 3D constructs. Future work should compare the transcriptional state of cardiac aggregates in 3D microcolumn culture with planar culture, and of innervated versus non-innervated 3D constructs, to determine how much of the transcriptional response is attributable to dimensionality versus innervation itself. Cell-type-resolved approaches, such as single-nucleus RNA sequencing, would address both the attribution and the cell-specificity limitations.

4. Conclusions

We describe the fabrication and characterization of TE-ICUs as a biofidelic model for studying neuro-cardiac interactions. In planar coculture, SCG-derived neuronal aggregates exhibited a robust sympathetic phenotype, established structural coupling with cardiac units, and were associated with sustained expression of genes encoding cardiac cytoskeletal components, gap junction proteins, and the adult myosin isoform. In 3D microcolumns, sympathetic aggregates produced fasciculated, aligned axonal tracts with demonstrable noradrenergic function. TE-ICUs combined these elements, demonstrating structural and functional neuro-cardiac coupling via intracardiac penetration of sympathetic projections and a significantly accelerated cardiac rate. The platform is modular and can be adapted to human-scale dimensions using hiPSC-derived cells, and extended to incorporate dual autonomic input to more fully recapitulate cardiac autonomic biology.

5. Materials and Methods

5.1. Primary Isolation and Culture of Rat Cardiac Myocytes and Sympathetic Neurons

All procedures were approved by the Institutional Animal Care and Use Committees at the University of Pennsylvania and the Corporal Michael Crescenz Veterans Affairs Medical Center and adhered to the NIH Public Health Service Policy on Humane Care and Use of Laboratory Animals. Primary cardiac myocytes were obtained from E16 Sprague-Dawley rat embryos following previously established protocols (Figure 1) [31]. Hearts were dissected and dissociated using 0.05% trypsin-EDTA for 10–15 min at 37 °C with manual trituration. After centrifugation, dissociated cells were resuspended in Cardiac Medium comprising 78% high-glucose DMEM, 17% Medium-199, 4% horse serum, and 1% penicillin/streptomycin. To form aggregates, 14 µL of a 7 × 10⁶ cells/mL suspension was transferred to pyramidal microwells within a PDMS mold, centrifuged at 1500 rpm for 5 min, supplemented with medium, and incubated overnight prior to seeding [47].

Sympathetic neuron aggregates were isolated from the SCG of P0–P1 Sprague-Dawley rat pups as previously described (Figure 1) [48]. Pups were euthanized by hypothermia and decapitation. The SCG were located at the bifurcation of the carotid arteries lateral to the trachea, extracted, and cleaned of pre- and postganglionic nerves and debris. Pieces of SCG were cultured as sympathetic neuron aggregates without further dissociation. SCG-only cultures were maintained at 37 °C and 5% CO₂ in medium comprising RPMI 1640 with 0.4% penicillin/streptomycin, 1% heat-inactivated horse serum, 10 µM uridine/5-fluorodeoxyuridine and 100 ng/mL NGF. For coculture with cardiac aggregates, the medium consisted of Cardiac Medium supplemented with 100 ng/mL NGF, given the requirement for this growth factor in sympathetic neuron survival and growth. For planar culture, surfaces were coated with 20 µg/mL poly-L-lysine, followed by 20 µg/mL laminin, before seeding cardiac and/or sympathetic aggregates.

5.2. Fabrication and Seeding of MeHA Hydrogel Microcolumns

Cardiac and sympathetic aggregates were cultured under 3D conditions using MeHA hydrogel microcolumns of 3–10 mm in length with outer and inner diameters of 701 and 300 µm, respectively (Figure 1A). MeHA was synthesized by esterification of hyaluronic acid with methacrylic anhydride for ~3.5 h at pH 8.5, purified by dialysis for 5–7 days and recovered by lyophilization [49]. The degree of functionalization of the hyaluronic acid disaccharides was ~44% as determined by ¹H NMR. A 3% w/v MeHA solution in Dulbecco’s phosphate-buffered saline (DPBS) with 0.05% Irgacure 2959 was drawn by capillary action into glass capillary tubes (inner diameter 701 µm) containing an inserted acupuncture needle (outer diameter 300 µm). The solution was photo-crosslinked under 10 mW/cm² ultraviolet light for 5 min. The needle was then withdrawn and the gelled microcolumns were removed into DPBS, sterilized for 30 min under ultraviolet light, and rinsed with fresh DPBS to remove residual free radicals. Columns were cut to the desired length, and an extracellular matrix solution of 1 mg/mL rat tail collagen type I plus 1 mg/mL mouse laminin in Neurobasal (pH 7.2–7.5) was added to the lumen and polymerized for 15 min at 37 °C. Dishes containing microcolumns were then flooded with culture medium and incubated at 37 °C and 5% CO₂ until seeding.

For seeding, cardiac aggregates and SCG were cut with fine forceps under a dissection scope into pieces that fit within the ends of the columns. For cardiac-only cultures, one piece of a cardiac aggregate was placed at one end of the microcolumn. Unidirectional sympathetic microcolumns were fabricated with a sympathetic aggregate at only one end. For TE-ICUs, the cardiac aggregate was seeded first, and a sympathetic aggregate was introduced at the opposite end one day later. Microcolumns were incubated at 37 °C and 5% CO₂ to allow cell attachment to the extracellular matrix and/or MeHA shell and subsequent sympathetic neurite growth.

5.3. RNA Isolation and RT-qPCR Gene Expression Analysis

Gene expression analyses were performed exclusively on planar neuro-cardiac cocultures and planar cardiac-only cultures; no RT-qPCR was performed on 3D microcolumn constructs. Total RNA was isolated after 7 days of culture using TRIzol (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. Freshly isolated E16 cardiac tissue served as the DIV 0 reference sample. Complementary DNA was synthesized from 2 µg of RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Waltham, MA, USA). Real-time PCR reactions of 10 µL were assembled with PowerUp SYBR Green Master Mix (Applied Biosystems), 10 ng cDNA and 300 nM gene-specific primers; primer sequences are listed in Table 1. Reactions were run on a QuantStudio 6 Pro Real-Time PCR System (Applied Biosystems) with one cycle each at 50 °C and 95 °C for 2 min, followed by 40 cycles of 95 °C for 3 s and 60 °C for 30 s, followed by melt curve analysis. Gapdh served as the endogenous reference gene. Relative expression was calculated for each sample as 2^(−ΔCt), where ΔCt = Ct(target) − Ct(Gapdh); values are therefore normalized to the reference gene within each sample and are presented without further normalization to a calibrator group.

5.4. Neurite Growth Characterization

Planar and 3D cultures were imaged in phase-contrast on a Nikon Eclipse Ti-S microscope with a QiClick camera and Nikon Elements software (Nikon, Tokyo, Japan). Imaging was used to quantify the length and growth rate of neurites projected from the sympathetic aggregate as a function of time. Length was measured as the distance between the tip of the longest observed neurite and the edge of the sympathetic aggregate; growth rate was estimated using the backward difference method.

5.5. Immunocytochemistry

At terminal time points, planar and 3D cultures were fixed in 4% paraformaldehyde for 35 min and rinsed in 1× phosphate-buffered saline (PBS). Cultures were permeabilized with 0.3% Triton X-100 in 4% horse serum for 60 min and incubated overnight at 4 °C with primary antibodies in 4% horse serum. Primary antibodies were directed against β-tubulin III (1:500, Sigma-Aldrich, St. Louis, MO, USA, T8578), a neuron-specific microtubule protein; tyrosine hydroxylase (1:500, Abcam, Cambridge, UK, ab113), the rate-limiting enzyme in norepinephrine biosynthesis; and cardiac troponin I (1:250, Abcam, ab47003), a cardiac-specific regulator of actin–myosin interaction. Cultures were then exposed to secondary antibodies (Alexa-488, Alexa-568, Alexa-647; all 1:500) for 2 h at 18–24 °C, followed by 10 min in 1:10,000 Hoechst in PBS. Phalloidin staining was used to visualize actin filaments. Stained cultures were imaged on a Nikon A1RSI laser scanning confocal microscope, with z-stacks presented as maximum intensity projections.

5.6. Fast-Scan Cyclic Voltammetry for Catecholamine Release

To assess evoked catecholamine release from 3D sympathetic aggregates and axon tracts, microtissues were incubated for 2 DIV in medium containing Neurobasal, 2% B-27, 1% fetal bovine serum, 2.0 mM L-glutamine, 100 µM ascorbic acid, 4 ng/mL basic fibroblast growth factor, 100 ng/mL NGF and 0.3% penicillin/streptomycin. Cultures were then incubated in the same medium containing 100 µM L-3,4-dihydroxyphenylalanine (L-DOPA) for at least 30 min. A single microtissue was transferred to a recording chamber and perfused at 37 °C with Neurobasal, 2.0 mM L-glutamine, 100 µM ascorbic acid and 100 µM L-DOPA bubbled with 95% O₂/5% CO₂. A bipolar stimulating electrode was placed to span the sympathetic aggregate region and a carbon fiber electrode (outer fiber length ~300 µm) was positioned in the same area. The carbon fiber potential was scanned linearly from −0.4 V to 1.2 V to −0.4 V versus Ag/AgCl at 400 V/s using a voltammeter/amperometer. After 8 s of recording, a monophasic stimulation train of 20 Hz, 10 pulses of 5 ms width and 8 V amplitude was applied. Cyclic voltammograms were recorded and analyzed with Demon Voltammetry and Analysis Software [50]. The reported current/concentration trace is the average of six recording sessions 10–12 min apart. Current at the peak oxidation potential of catecholamines in consecutive voltammograms was converted to concentration using the slope of the linear regression of current elicited by injection of 3, 6 and 12 µM norepinephrine hydrochloride in 126 mM NaCl, 2.5 mM KCl, 1.2 mM NaH₂PO₄·H₂O, 2.4 mM CaCl₂·2H₂O, 1.2 mM MgCl₂·6H₂O, 25 mM NaHCO₃ and 0.4 mM L-ascorbic acid.

5.7. Analysis of Spontaneous Cardiac Contraction

The effect of sympathetic aggregates on cardiac beating rate was analyzed using video recordings of 1–1.5 min acquired with a Nikon Eclipse Ti-S microscope. Planar cultures were recorded at 3 and 5 DIV, and 3D microcolumn cultures at 5 and 8 DIV, with DIV counted from plating of the cardiac aggregate. Beats per minute were quantified manually in Fiji. Sample sizes were as follows: planar cardiac-SCG coculture and planar cardiac-only cultures, and 3D cardiac-SCG and cardiac-only microcolumns, with n = 8 and n = 7 constructs per group, respectively.

5.8. β-Adrenergic Blockade

To assess the contribution of β-adrenergic signaling to spontaneous contraction, beating rate was recorded in cardiac-only and SCG-cardiac microcolumns immediately before and after acute application of the non-selective β-adrenergic antagonist propranolol. Constructs were incubated with 1 µM propranolol for 5 min before the post-treatment recording. Recordings before and after treatment were acquired from the same construct, and the response was expressed as the ratio of beats per minute after treatment to beats per minute before treatment.

5.9. Statistical Analysis

Statistical analyses were performed in Prism 8.1.1 (GraphPad, San Diego, CA, USA). For neurite growth over the full culture period in 3D microcolumns, repeated-measures one-way ANOVA with Tukey’s multiple comparisons test was used to evaluate the effect of time. For the comparison of SCG-only and SCG-Cardiac microcolumns at 4 and 7 DIV, the effects of group and time were analyzed by two-way ANOVA with Sidak’s multiple comparisons test. For contraction rate, normality was confirmed with the Kolmogorov-Smirnov test, and the effects of DIV and culture type were analyzed by two-way ANOVA, with between-group differences at each time point assessed using Sidak’s multiple comparisons test. Gene expression data were compared across DIV 0 cardiac tissue, DIV 7 cardiac-only cultures, and DIV 7 neuro-cardiac cocultures by one-way ANOVA with Tukey’s multiple comparisons test. Gene expression data are presented as mean ± SD; neurite growth and contraction rate data are presented as mean ± SEM, as indicated in the individual figure captions. Significance thresholds are indicated as * p < 0.05, ** p < 0.01, *** p < 0.001 and **** p < 0.0001; ns denotes not significant.