1. Introduction
Dynamic reciprocal interactions between developing organs are a cornerstone of embryogenesis, coordinating the spatial and functional maturation of tissues and organs. During the neurulation stage of embryonic development, dynamic crosstalk between germ layers is essential for proper differentiation, morphogenesis, and organ specification [
1,
2]. For example, signals from the mesoderm induce the ectoderm to form the neural tube, which is the precursor to the brain and spinal cord [
3]. Similarly, during the organogenesis phase, dynamic crosstalk between organs ensures proper cell fate specification, structural patterning, and physiological integration. For instance, the heart and brain exhibit bidirectional interactions [
4]. Neural crest cells guide cardiac outflow tract formation, whereas cardiac-derived signals influence neurovascular development. Disruption of these interactions leads to developmental anomalies, such as congenital heart defects linked to autonomic nervous system dysregulation [
5,
6].
Animal models have long been indispensable in the study of embryonic development, providing valuable insights into the dynamic interactions between multiple organs. These models, ranging from mice to nonhuman primates, have been used to investigate the processes of cell differentiation, morphogenesis, and organ function maturation. For example, transgenic mice have been widely used to study the development of the heart and brain, allowing researchers to manipulate specific genes and observe their effects on organ interactions. However, animal models may not fully recapitulate developmental processes and organ interactions in humans. Additionally, animal models are unsuitable for mechanistic studies of dynamic interorgan communication.
Recent emerging
in vitro research model systems, such as multiorgan-on-a-chip (MOOC) and assembloid models, offer a unique opportunity to study human-biologically relevant multiple organ interactions [
7]. The MOOC connects histotypic or organotypic cultures via microfluidic channels, enabling them to mimic the biochemical and hemodynamic couplings between tissues and organs under dynamic flow conditions [
8,
9]. To date, various MOOC models, such as liver-lung [
10], gut-brain [
11], and gut-liver-brain [
12], have been used to study the effects of organ interactions on functionality, efficacy, and off-target toxicity of anticancer therapeutics [
13,
14]. However, despite their potential, MOOC models are limited in their ability to recapitulate the complex signaling and mechanical cues that drive organogenesis
in vivo. Additionally, the complexity of integrating multiple organoids into a single chip can lead to technical challenges, such as maintaining the viability, phenotype, and functionality of all the organoids over extended periods.
Assembloids, which fuse spatially distinct organoids, can mimic the interactions between different regions of the nervous system or other tissues, allowing for the study of complex biological processes, such as neural migration, axon projection, and circuit formation [
15]. For example, assembloids derived from different brain regions can be combined to model the migration of interneurons and the formation of neural circuits, providing insights into the underlying mechanisms of neuropsychiatric disorders [
16], [
17], [
18]. Assembloids have been used to study the maturation of specific cell types, decipher molecular signals following cell-cell interactions, and investigate the coordination of cell movement during development [
15,
19]. However, assembloids rely on predifferentiated organoids and fail to mimic the coemergence of distinct brain regions from shared progenitor pools during neurodevelopmental processes [
20]. Therefore, there is an unmet need to develop novel, human-biologically relevant research models capable of replicating the dynamic reciprocal interactions between developing organs, particularly the intricate interactions between organs originating from distinct germ layers.
Here, we present a trans-germ-layer codevelopment organoid chip (TGCO-Chip) that permits the formation of a codevelopoid, namely, an organoid complex containing two interconnected distinct organoids codifferentiated from a shared upstream-lineage stem cell aggregate. The rational design of the TGCO-Chip allows precise microenvironmental control over two compartments of the same stem cell aggregate in a well-controllable fashion. Using a TGCO-Chip, we generated heart-brain codevelopoids (HBCs) from embryoid bodies (EBs) and investigated the effects of heart-brain interactions on cell fate specification, morphogenesis, and functional maturation (
Figs. 1(a) and
(b)). Overall, the TGCO-Chip offers a versatile tool for constructing multiorganoid complexes by combining the principles of bioengineering and stem cell biology. Generated codevelopoids hold great promise for elucidating the dynamic crosstalk between two organs that occurs during early human organogenesis, tissue maturation, and disease, which is difficult to dissect
in vivo.2. Materials and methods
2.1. Design and fabrication of a TGCO-Chip
The TGCO-Chip consists of two main components: a polydimethylsiloxane (PDMS; Sigma-Aldrich, Germany) base with a microchamber and a separate polymethyl methacrylate (PMMA) plate (Evonik Industries AG, Germany), as illustrated in
Fig. 1(b). The PDMS base was fabricated through a continuous molding process. First, a microchamber was designed in the shape of two intersecting cylinders using computer-aided design (CAD). The cylinders were 2 mm in diameter and the microchamber had a 1 mm intersection chord and a height of 0.6 mm. Next, a PMMA negative mold was engraved using computer numerical control (CNC) machine tools. A 2% agarose solution, boiled and then poured into the PMMA mold, solidified at room temperature (RT) to create an agarose-positive mold. The PDMS prepolymer was then poured onto the agarose mold and cured at RT for 12 h. After curing, the PDMS layer was carefully removed with tweezers, and a 33 mm punch was used to create holes for a culture chamber that fit a six-well plate. Moreover, a 200 µm thick PMMA medium separation plate was fabricated by laser engraving a 6 mm thick PMMA sheet. The aligned PDMS base and PMMA plate were bonded using PDMS prepolymer, placed in a six-well plate, and cured in an oven at 80 °C for 5 min. Finally, the chips were sterilized by ultraviolet (UV) exposure for 24 h before use.
2.2. Cell culture
The H9 (female) human embryonic stem cell (hESC) line was sourced from the Cell Bank/Stem Cell Bank at the Chinese Academy of Sciences. Cells were cultured in NcTarget™ human pluripotent stem cell (hPSC) Medium (Shownin Biotechnologies, China) and passaged using accutase. After detachment, the cells were reseeded onto vitronectin-coated six-well plates filled with NcTarget™ hPSC Medium supplemented with 10 µmol∙L-1 Y-27632 (Stem Cell Technologies, Canada).
2.3. Generation of HBCs via a TGCO-Chip
Brain compartments of the HBCs were generated using a modified brain organoid culture protocol based on a protocol published by Lancaster et al. [
21]. Cardiac compartments of the HBCs were cultured using the STEMdiff™ Cardiomyocyte Differentiation Kit (Stem Cell Technologies). The hESCs were harvested at approximately 70% confluency. HBCs were generated by seeding 3 × 10
5 cells. Cells were seeded in a volume of 10 μL into a microchamber containing NcTarget™ hPSC Medium supplemented with 4 ng∙mL
−1 basic fibroblast growth factor (bFGF; Peprotech, USA) and 10 μmol∙L
-1 Y-27632. After 5 d of culture, the EBs gradually aggregated toward the center of the culture chamber and blocked the connecting channels.
2.3.1. Generation of the brain compartment of HBCs
The EBs were cultured in suspension in neural induction medium Dulbecco’s modified Eagle’s medium (DMEM)/F12(Gibco, USA), 1:100 N2 supplement (Gibco), 1:100 L-alanyl-L-glutamine (GlutaMAX; Gibco), 1:100 minimum essential medium with non-essential aminos (MEM-NEAA; Gibco), and 1 mg∙mL−1 heparin (Gibco). The medium was exchanged daily for 5 d. On day 11, EBs with neuroepithelial identity were encapsulated in Matrigel and further cultured in neural differentiation medium comprising a ratio of DMEM/F12 and neurobasal (1:1; Gibco) supplemented with 1:200 (v/v) N2 supplement, 1:100 (v/v) B27 supplement without retinoic acid (Gibco), 1:100 (v/v) GlutaMAX, 1:200 (v/v) MEM-NEAA, 1:4000 (v/v) insulin, and 100 μmol∙L-1 β-mercaptoethanol(Merck, Germany) under stationary conditions for 4 d. On day 15, the culture medium in the dishes was replaced with neural maturation medium. The composition of the neural maturation medium was the same, except for the B27 supplement (Gibco) with vitamin acid.
2.3.2. Generation of the heart compartment of HBCs
The EBs were cultured in suspension in medium A (cardiomyocyte differentiation basal medium supplemented with supplement A) to initiate differentiation toward a cardiomyocyte fate. On day 7, the medium was fully replaced with fresh medium B (cardiomyocyte differentiation basal medium supplemented with supplement B). On days 9 and 11, full medium changes were performed using fresh medium C (cardiomyocyte differentiation basal medium supplemented with supplement C). By day 13, the medium was replaced with medium D (cardiomyocyte maintenance medium) to promote further differentiation into mature cardiomyocytes.
2.3.3. Generation of forebrain-heart codevelopoids via a TGCO-Chip
The heart compartment was generated as described above. For the brain compartment, EBs were supplemented with SMADi neural induction medium (Stem Cell Technologies) on day 5 to initiate neural differentiation. On day 13, the medium was replaced with forebrain neuron differentiation medium (Stem Cell Technologies) to promote the differentiation of forebrain neurons. The PMMA medium separation plate was removed on day 28, and the organoids were gently flushed from the chamber using a Pasteur pipette. The organoids were subsequently cultured in forebrain neuron maturation medium (Stem Cell Technologies) until day 33 to ensure forebrain neuronal maturation.
2.4. Immunostaining
2.4.1. Sample collection
Organoids were collected after 21-28 d of culture and fixed in 4% paraformaldehyde (PFA; Sigma-Aldrich) at RT for 30 min on an agitation platform. After removal of the PFA, the samples were stored in phosphate-buffered saline (PBS; Gibco) at 4 °C for further analysis. The samples were incubated in 15% (w/v) sucrose in PBS at 4 °C overnight. Afterward, the samples were embedded in 30% (w/v) sucrose in PBS at 4 °C overnight and frozen in optimal cutting temperature (OCT) compound (Tissue-Tek, USA) at -20 °C. Aggregates with 10 μm sections were cut on a cryostat microtome (Leica CM3050S; Leica Microsystems, Germany), collected on microscope slides (Thermo Fisher Scientific, USA) and stored at -20 °C.
2.4.2. Slice immunofluorescence
The samples were washed with PBS (10 min) to remove OCT, and the tissue boundaries were marked with a hydrophobic pen. Permeabilization/blocking was performed with 5% bovine serum albumin (BSA)/0.3% Triton X-100/PBS (30 min, RT, humidified dark chamber). Primary antibodies in 5% BSA/0.1% Triton X-100/PBS were applied (24 h, 4 °C). The slides were subsequently washed for 10 min in PBS for three times (RT), followed by secondary antibody incubation (1:500 in antibody solution, 2 h, RT). After being subjected to 4′,6-diamidino-2-phenylindole (DAPI; Solarbio, China) staining (2 µg∙mL−1, 5 min) and rinsing with PBS, the slides were stored at 4 °C. Antibody details are in the key resource table.
2.4.3. Whole-mount immunofluorescence
HBCs were fixed with 4% PFA (24 h, 4 °C), permeabilized in 2% PBS containing 0.1% Tween 20 (PBST; 24 h, RT), and blocked with 10% normal goat serum/1% Triton X-100/PBS (24 h, 4 °C). Primary antibodies in antibody dilution buffer (1% goat serum/0.2% Triton X-100/PBS) were applied for 48 h at 4 °C. After washed for 1 h in 0.2% PBST for twice (RT), the samples were incubated in high-salt buffer (3% NaCl/0.2% PBST, 24 h, 4 °C). Secondary antibodies in the same dilution buffer were incubated for 24 h at 4 °C, followed by sequential washes (1 h, twice, RT, 24 h, 4 °C). The samples were subsequently washed with PBS (30 min, three times), stained with DAPI (1:1000, 24 h, 4 °C), and rewashed with PBS. Finally, the samples were mounted in prewarmed (37 °C) RapiClear reagent with orbital shaking (24 h, RT) and imaged using a confocal microscope.
2.5. Microscopy and image analysis
Light microscopy imaging was performed using an Olympus IX-83 microscope, whereas fluorescence images were acquired with a Leica DMI4000 B inverted fluorescence microscope. Image analysis was conducted using ImageJ (v.2.0.0-rc; Fiji, USA).
2.6. Multielectrode array (MEA) analysis of the HBCs
To allow the HBCs to adhere to the MEA plate, 500 μL of 1% Matrigel in PBS was added to the six-well MEA plate and incubated in a 37 °C incubator for 1 h. The selected HBCs were placed in the center of the well with the electrodes, and 50 μL cardiomyocyte maintenance medium was added. The MEA plate was transferred to the MEA device to measure the field potential (FP) and spontaneous spikes. The MEA device was maintained under environmental control (37 °C, 5% CO2). To minimize the error range of the electrical signals, the baseline was recorded by measuring the FP for 3 min after 10 min. After the raw FP and spontaneous data were recorded in AxIS software, they were analyzed using the Cardiac Analysis Tool, Neural Metrici Tool, and AxIS Metric Plotting Tool (Axion BioSystems, Inc., USA).
2.7. Real-time video acquisition and contraction analysis
To compare the heart contractile function of HBCs (n = 6) with that of heart-heart organoids (HHOs; n = 6) on day 15, live image sequences were acquired using a microscope (LS850TM; Etaluma, USA) located in a 37 °C and 5% CO2 incubator. Images were recorded at a frame rate of 100 frames per second for 10 s and converted into Avi files using ImageJ. The contraction amplitude, contraction rate, time to peak, and relaxation time were manually counted and calculated from the video analysis, and representative contraction graphs were generated using open-source MUSCLEMOTION video analysis software, following the provider’s instructions.
2.8. Drug stimulation
Videos were recorded before and after drug administration using a real-time imaging microscope (LS850TM). The HBCs were incubated in cardiomyocyte maintenance medium supplemented with 10 µmol∙L-1 isoproterenol (MCE, China) for 1 h in a CO2 incubator. The beating of the HBCs was then recorded for 1 min. Pre- and post-drug administration were analyzed using MUSCLEMOTION software to measure the distance between boundary surfaces during contraction and relaxation, as well as the contraction rate. Simultaneously, myocardial electrical signals were recorded for 1 min prior to treatment. The culture medium was subsequently replaced with cardiomyocyte maintenance medium containing 10 µmol∙L-1 isoproterenol, and the cells were incubated for 1 h. Then, electrical signals were recorded for 1 min for comparison with baseline measurements. For L-glutamic acid (MCE) treatment, data were recorded before and after the addition of 50 µmol∙L-1 L-glutamic acid.
2.9. Cholera toxin subunit B (CTB) retrograde neural tracing of HBCs
CTB-555 (Brain VTA, China) was stereotaxically positioned 0.50 mm below the surface of the HBC. Employing a microscope, CTB-555 was delivered at a rate of 10 nL∙min−1 via a micropipette connected to the Auto-Nanoliter Injector, with an injection volume of 50 nL per site. After the injection, the micropipette remained stationary for 3 min before being gradually retracted to prevent tracer leakage.
2.10. Transmission electron microscopy sample preparation and analysis
HBCs on day 28 were transferred to new well plates using low-binding pipettes, washed with PBS, and fixed in electron microscopy fixative at 4 °C for 4 h, followed by three washes in 0.1 mol∙L-1 phosphate buffer. After being rinsed with the same buffer, the samples were postfixed in 1% osmium tetroxide on ice for 50 min. Following three additional rinses, the samples were dehydrated in a graded series of acetone on ice and embedded in Agar 100 resin. Ultrathin sections (80 nm) were prepared using an ultramicrotome and placed on copper grids coated with formvar film (WFHM-150; Servicebio, China). The sections were poststained with 2% uranyl acetate and Reynolds lead citrate. Transmission electron microscopy images were acquired using a Hitachi HT7800/HT7700 transmission electron microscope (Japan).
2.11. RNA sequencing (RNA-seq) analysis
To investigate the transcriptional profiles associated with cardiac lineage specification during developmental progression, HBCs (n = 4) and HHOs (n = 4) were harvested on day 15 post differentiation induction. Organoids were mechanically dissociated along the PDMS chip baffle using a sterile syringe under aseptic conditions to isolate the heart and brain regions. For each biological replicate, three organoids per region were pooled to minimize intersample variability, with heart regions from HBCs (n = 4 pooled samples) and HHOs (n = 4 pooled samples) processed independently. The samples were rapidly cryopreserved by immersion in liquid nitrogen to preserve RNA integrity. Total RNA was extracted using TRIzol reagent following the manufacturer’s protocol. The quantity and quality of the RNA samples were subsequently determined using a Qseq-400 system (Bioptic, China). Library preparation for sequencing was performed using an Optimal Dual-Mode mRNA Library Prep Kit (BGI-Shenzhen), and the libraries were sequenced on a G400/T7/T10 platform (BGI-Shenzhen).
Between-group differential gene expression analysis was conducted using DEGSeq under the criteria of a log2(fold change) > 1 and an adjusted p value < 0.05. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed utilizing the database for annotation, visualization, and integrated discovery. Pathways with false discovery rate (FDR) q values < 0.05 were deemed statistically significant. On the basis of the GO and KEGG annotation results, the differentially expressed genes were functionally categorized. Scatter plots, heatmaps, and results of the GO enrichment and KEGG enrichment analyses were generated using R software (v.4.4.1). The transcriptomics data were submitted to the Sequence Read Archive (database identifier PRJNA1202668).
2.12. Fluorescent dextran diffusion assay
The fluorescent dextran diffusion assay was performed by introducing 70 kDa fluorescein isothiocyanate (FITC)-dextran ((131.91 ± 1.91) µg∙mL−1) into the brain compartment on day 5. Dextran levels in both the brain and heart microchamber medium were measured using microplate fluorometry at 0 and 24 h after administration, with dextran-free medium used as a blank control.
2.13. Statistical analysis and reproducibility
Statistical analyses were conducted using GraphPad Prism (v.8.0.2). In accordance with the test requirements, unpaired and paired two-sided Student’s t tests and one-way analysis of variance (ANOVA) were used to determine statistical significance. The corresponding figure captions include all the details regarding biological replicates (N) and statistical tests. All key resources and reagents used in this study are summarized in Table S1 in Appendix A.
3. Results
3.1. Generation of HBCs with a TGCO-Chip
We rationally designed a TGCO-Chip to generate a HBCs containing interconnected heart and brain organoids (
Fig. 1(a)). The TGCO-Chip was constructed on a standard six-well plate and includes a microfabricated construct embedded in a well (Fig. S1(a) in Appendix A). The construct contained a separation plate stuck on a microfabricated PDMS base, which was further stuck on the bottom of a well. The 200 µm thick separate plate was positioned in the center of a well and divided into two culture compartments, enabling the use of two different organoid differentiation media. Additionally, the center of the PDMS base features an intersecting cylinder-shaped microchamber. These cylinders have a diameter of 2 mm and a height of 0.6 mm and are designed with an intersection chord of 1 mm. The microchamber is used to form and culture codevelopoids (
Fig. 1(b), Fig. S1(b) in Appendix A). HBCs undergo different stages including EB formation, lineage determination, co-development, and maturation (Fig. S1(c) in Appendix A).
To generate codevelopoids, a total of 3 × 10
5 hPSCs were seeded into the microchambers, after which 600 μL of culture medium was added to the two sides of each well 1 h after cell seeding. hPSCs gradually aggregated into a peanut-shaped embryonic body and completely blocked the connection channel between the two microchambers on day 5 (Figs. S2(a) and (b) in Appendix A). Fluorescent dextran diffusion assays confirmed complete separation of heart and brain culture compartments by day 5. Following the addition of 70 kDa FITC-dextran to the brain compartment, no significant concentration change was observed in the brain medium over 24 h ((134.28 ± 2.90) vs (131.91 ± 1.91) µg∙mL
−1 at 0 h,
p = 0.1648). Simultaneously, heart compartment concentrations remained at background levels ((0.47 ± 0.23) vs (0.44 ± 0.01) µg∙mL
−1 of the blank control,
p = 0.5248), demonstrating negligible diffusion across compartments (Figs. S2(c)-(f) in Appendix A). Spatially compartmented differentiation of the embryonic body into a codevelopoid containing a cardiac organoid and a brain organoid was subsequently conducted in the two culture compartments of the TGCO-Chip. Specifically, the brain compartment was generated according to Lancaster’s cerebral organoid differentiation protocol [
21]. Three-dimensional (3D) neural differentiation started from the embryonic body and involved three differentiation stages, namely, neuroectoderm induction, neural differentiation, and neural maturation, corresponding to parenchymal cell fate transitions from the ectoderm, neuroectoderm, neuroepithelium, neural stem cells, and immature neurons to mature neurons. Moreover, the cardiac compartment was generated using the STEMdiff™ ventricular cardiomyocyte differentiation protocol with moderate revisions. Cardiac differentiation started from the embryonic body and involved four differentiation stages: mesoderm induction, cardiac progenitor cell induction, cardiomyocyte differentiation, and cardiomyocyte maturation (
Fig. 1(c)). Overall, the TGCO-Chip enabled spatial compartments of neural and cardiac medium to differentiate the two sides of the same embryonic body into brain and cardiac organoids while keeping the two organoids physically interconnected and exchanging biochemical and biomechanical cell signals.
We analyzed the growth and development of the HBCs over a period of 28 d (
Fig. 1(d)). The brightfield image revealed that the area of the brain compartment first gradually decreased, probably because of cell contraction from days 0 to 9, and then increased until day 28. We introduced a wrinkling index (WI =
Li/
Lo, Li and
Lo represent for the inner and outer contour lengths for the HBCs, respectively) as a phenotypic quantification index to evaluate the folding of neural rosettes in brain organoids [
22]. The WI was adapted from the gyrification index (GI) to quantify the phenotypic features of brain organoids [
23]. The brain compartment experienced morphogenesis of neural rosette structures on day 15 and exhibited significantly enhanced morphological complexity on day 21 (WI increasing from (1.10 ± 0.08) to (1.36 ± 0.11),
p = 0.0003;
Figs. 1(e) and
(f)), accompanied by progressive expansion of the neural compartment area ((1.76 ± 0.30) vs (2.84 ± 0.26),
p < 0.0001; Fig. S2(g) in Appendix A). Moreover, the size of the cardiac compartment increased from days 5 to 9 and reached its maximum area, corresponding to the formation of a chamber-like structure (
Fig. 1(g), Fig. S2(h) in Appendix A). Afterward, the area of the cardiac compartment decreased from days 9 to 15 and remained relatively consistent after day 15 (Fig. S2(h)). Coordinated contractions emerged on day 15 (Video S1 in Appendix A), evolving into stable rhythmic beating by day 21 (Video S2 in Appendix A).
Whole-mount immunofluorescence staining of the HBCs on day 28 revealed that the neurokinin 2 transcription factor related locus 5 (Nkx2.5) marker, a transcription factor of cardiomyocytes, was specifically expressed on the codevelopoid side supplemented with cardiac differentiation medium, whereas the neuron-specific class III β-tubulin (TUJ1) marker was predominantly expressed on the codevelopoid side supplemented with neural differentiation medium (
Fig. 1(h)). Moreover, the codevelopoid coexpressed Nkx2.5 and TUJ1 markers in the middle, corresponding to the intersection region of the two microchambers. Quantitative analysis of immunofluorescence intensity further confirmed the region-specific distribution of cell-specific markers (
Fig. 1(i)). These findings indicate that the TGCO-Chip successfully facilitates the codifferentiation of an EB into brain and heart organoid lineages, highlighting its potential for studying organ interactions during development.
3.2. Dual-lineage specification of HBCs
We performed immunofluorescence analysis to characterize neural and cardiac specifications. The brain compartment demonstrated organized hierarchical neurogenesis, characterized by the occupation of proliferative niches by SRY-box transcription factor 2 positive (SOX2
+) neural stem cells, with TUJ1
+ immature neurons and microtubule-associated protein 2 positive (MAP2
+) mature neurons predominantly localized to peripheral regions. Immunofluorescence analysis revealed region-specific expression of the forebrain marker paired box 6 (PAX6) and the hindbrain marker insulin gene enhancer protein 1 (ISL1). Furthermore, T-box brain transcription factor 1 positive (TBR1
+) cortical neurons were specifically detected in the developing cortical architecture of the brain compartment (
Fig. 2(a)).
Immunofluorescence staining revealed that cardiomyocytes exhibited sarcomere microstructures, and electron microscopy revealed that the sarcomeres displayed visible Z lines and alternating band patterns, confirming the maturation of cardiomyocytes (
Figs. 2(b) and
(c)). Cardiac compartments exhibited ventricular-specific differentiation, characterized by the presence of cardiac troponin T positive (cTnT
+)/Nkx2.5
+ cardiomyocytes, which were further segregated into myosin light chain 7 positive (MYL7
+) atrial and myosin light chain 2 positive (MYL2
+) ventricular subpopulations. In addition, vimentin (VIM) cardiac fibroblasts and cluster of differentiation 31 (CD31) endothelial cells in myocardial tissue were abundantly expressed in the cardiac compartment (
Figs. 2(d) and
(e)). The results above indicate that HBCs include various cell types, such as neural stem cells, newborn neurons, mature neurons, cardiomyocytes, atrial cells, ventricular cells, cardiac fibroblasts, and endothelial cell markers. Overall, these results suggest that the HBC possesses spatially compartmented organ-specific cell types and morphological features.
3.3. HBCs display cardiac and neural electrophysiological activities and drug responses
We evaluated the cardiac electrophysiology of HBCs via a MEA system. MEA demonstrated that the FPs of the cardiac compartment resembled human electrocardiogram (ECG)-like activities, reflecting the depolarization, hyperpolarization, and repolarization processes of cardiac physiology (
Figs. 3(a)-(c)). Furthermore, we conducted an in-depth analysis of cardiac electrophysiological signals to determine the signal amplitude, contraction rate in beats per minute (BPM), field potential duration (FPD), and spike slope. The average spike amplitude of the HBCs significantly increased from (0.5 ± 0.1) mV on day 15 to (1.6 ± 0.7) mV on day 21 (
p = 0.0104;
Fig. 3(d)). The contraction rate significantly increased from (46.5 ± 19.1) BPM on day 15 to (88.1 ± 22.6) BPM on day 21 (
p = 0.0085;
Fig. 3(e)). In addition, the peak slope, which represents the electrical conduction velocity, increased from (0.36 ± 0.28) V∙s
−1 on day 15 to (2.36 ± 1.96) V∙s
−1 on day 28 (
p = 0.0160;
Fig. 3(f)). Compared with that on day 15, the FPD on day 28 was not significantly different ((198.48 ± 124.82) vs (183.53 ± 36.61) ms,
p = 0.5897;
Fig. 3(g)). Collectively, these results revealed progressive electrophysiological maturation of HBCs over time with increasing cardiac and neural differentiation processes.
We performed timelapse-based brightfield image analysis of beating movements of HBCs, such as contraction rate and amplitude, which are pivotal parameters of cardiac function [
24]. Isoproterenol, a β-adrenergic receptor agonist, was applied to the HBCs at a final concentration of 10 µmol∙L
-1 in the culture medium for 1 h [
25] (
Fig. 3(h)). The beating frequency of the HBCs significantly increased from (40.00 ± 8.14) to (96.63 ± 14.72) BPM (
p < 0.0001) following the treatment, whereas the contraction amplitude did not significantly change in the HBCs ((245.36 ± 172.01) vs (230.55 ± 130.30)
F/
F0, where
F represents the real-time grayscale value and
F0 represents the baseline grayscale value,
p = 0.8819) (
Figs. 3(i)-(k)). Electrophysiological analyses further revealed a significant increase in the beating frequency and average spike amplitude after adrenaline stimulation, whereas no significant difference was detected in the FPD (Figs. S3(a)-(d) in Appendix A).
To functionally validate neural maturation in the codevelopoid model, we constructed a heart-forebrain codevelopoid model using the TGCO-Chip. Compared with the conventional brain organoid differentiation protocol, our forebrain organoid differentiation protocol enabled faster neural maturation and the formation of neuroelectrophysiological activities (Figs. S3(e) and (f) in Appendix A). MEA electrophysiological analysis on day 33 revealed two distinct firing patterns from those of the codevelopoid in the raster plot (Fig. S3(g) in Appendix A). Specifically, the spike raster plot displayed a rhythmic, synchronized cardiac firing pattern, indicated by a stable beating frequency ((144.12 ± 16.13) BPM, n = 5), consistent action potential amplitude ((2.20 ± 0.76) mV), and uniform depolarization kinetics (mean spike slope: (2.98 ± 1.38) V∙s−1; Figs. S3(h)-(l) in Appendix A). Moreover, the neural firing pattern exhibited high-frequency asynchronous spike signals indicative of neuronal activity, featuring an elevated spike frequency ((577.60 ± 206.42) spikes∙min−1) corresponding to a mean firing rate of (9.63 ± 3.44) Hz, along with periodic burst discharges ((37.60 ± 17.57) BPM). These neural bursts suggest the emergence of network-level synchronization, a hallmark of functional neuronal maturation (Figs. S3(m)-(q) in Appendix A).
Collectively, these findings confirmed the electrophysiological functional maturation of both the cardiac and the neural compartment of the HBCs and the retention of major ion channels and β-adrenergic signaling pathways, which will be beneficial for advanced neurocardiac drug testing in the future.
3.4. Brain compartments enhance the cardiac functionality of HBCs via organ interactions
To explore the impact of dynamic heart-brain interactions on human heart development, we generated HHO complexes using the TGCO-Chip as a control group for HBCs (
Fig. 4(a)). Bright field image analysis revealed that the heart compartment in the HBC group displayed a chamber-like phenotype from days 13 to 15, whereas the heart compartment in the HHO group displayed a more uniform and less structured morphology throughout the same developmental period (Figs. S4(a) and (b) in Appendix A). Transcriptomic analysis further demonstrated the upregulation of 471 genes (
p < 0.05, log
2(fold change) > 1) and the downregulation of 118 genes (
p < 0.05, log
2(fold change) < -1) in the HBC group compared with the control group (Fig. S4(c) in Appendix A). Moreover, 17 645 genes were not significantly different between these two groups (Fig. S4(d) in Appendix A). GO analysis demonstrated that upregulated genes in HBCs were enriched primarily in pathways related to nervous system development, neuromuscular junction formation, and myocardial development (
Fig. 4(b)). Notably, heatmap plots indicated that genes associated with cardiac maturation, such as
CAV3,
CKMT2,
FABP1,
FABP2,
FPGT-TNNI3K,
GJA5, and
HAND2, were significantly upregulated in the HBC group. Moreover, genes linked to neural maturation, including
CHRNA2,
EGR2,
ELAVL3,
FOXG1,
GJB1,
HOPX, and
MAPT, were also upregulated (
Fig. 4(c)), suggesting that bidirectional interactions between developing cardiac and brain organoids promoted their maturation. Time-lapse brightfield image analysis of beating movements of HBCs (Videos S3 and S4 in Appendix A) on day 15 further revealed that compared with the control group, the HBC group exhibited significantly greater contraction amplitude ((113.77 ± 30.52) vs (52.60 ± 26.01)
F/
F0,
p = 0.0039), contraction rate ((44.36 ± 7.26) vs (31.44 ± 3.55) BPM,
p = 0.0029), and time to peak ((0.21 ± 0.04) vs (0.31 ± 0.04) s,
p = 0.0046;
Figs. 4(d)-(g)). Moreover, no significant difference was observed in the relaxation time ((0.33 ± 0.03) vs (0.32 ± 0.10) s,
p = 0.7048;
Fig. 4(h)). Collectively, these findings suggest that brain-heart interactions during early codevelopment enhance myocardial development and promote its maturation.
We employed CTB as a retrograde neural tracing agent to visualize neural projections from the brain compartment to the heart compartment in the HBCs. Principally, CTB binds to the pentasaccharide chain of monosialotetrahexosyl ganglioside (GM1) on the cell surface, after which it is internalized by the axon terminals and transported retrogradely along the axons to the cell body. This process allows CTB to label the neurons that project to the injection site, enabling the visualization of neural projections and connections [
26]. CTB was injected into the heart compartment of the HBCs on day 28, followed by fluorescence imaging on day 33 (
Fig. 4(i)). Fluorescence image analysis revealed CTB
+ neurons aggregately distributed in some specific regions of the brain compartment (
Fig. 4(j)), demonstrating the existence of projections from the brain compartment to the cardiac compartment. Functional neuro-cardiac coupling in heart-forebrain codevelopoids was assessed via MEA recordings during glutamate-mediated neuronal stimulation.
L-Glutamic acid (50 µmol∙L
-1) significantly increased the neuronal firing rate (
p = 0.0172; Figs. S4(e) and (f) in Appendix A). Concomitant cardiac responses exhibited bidirectional modulation: 54.5% of heart-forebrain codevelopoids (6/11) had prolonged beat periods ((1.10 ± 0.18) vs (0.92 ± 0.16) s,
p = 0.0050), whereas 45.5% (5/11) had shorter periods ((0.66 ± 0.11) vs (0.91 ± 0.09) s,
p = 0.0428). In contrast, HHOs showed no significant changes during the cardiac period (
p = 0.3244; Figs. S4(g)-(k) in Appendix A). These results indicate that glutamate-activated neural circuits differentially regulate cardiac activity in heart-forebrain codevelopoids.
4. Discussion
Interorgan communication is a hallmark of embryonic development and adult homeostasis and governs tissue specification, structural patterning, and functional maturation. Despite their centrality, these dynamic interactions remain poorly understood because inadequate models recapitulate lineage-coupled organogenesis. Here, we bridge this gap by integrating developmental biology with bioengineering principles to create a TGCO-Chip. The TGCO-Chip enables the spontaneous differentiation of a single EB into two interconnected organoids under spatially controlled conditions, termed a codevelopoid. Key to this system is its ability to apply compartment-specific differentiation protocols while preserving paracrine gradients and biomechanical coupling between chambers. Physical segregation via controlled cell aggregation prevents unintended signaling crosstalk, ensuring lineage fidelity without disrupting developmental trajectories. By recapitulating the synchrony and bidirectional signaling of embryonic organ pairs, the codevelopoid model offers unprecedented fidelity of interorgan crosstalk. Notably, the TGCO-Chip supports germ layer-mismatched systems (e.g., heart-brain) and lineage-shared tissues (e.g., forebrain-midbrain complexes), providing a versatile platform for studying developmental and regenerative processes.
Current bioengineered models of human interorgan interactions fall into two categories: assembloids and multiorganoid-on-a-chip systems. Assembloids fuse predifferentiated organoids (e.g., cortical-thalamic complexes) to study postmaturation tissue crosstalk, enabling axonal projection analysis and neural circuit modeling in neurological disorders [
16], [
17], [
20], [
27], [
28], [
29]. Conversely, multiorganoid-on-a-chip platforms compartmentalize distinct organoids via microfluidic networks, facilitating metabolite exchange for applications such as drug toxicity profiling [
8,
30]. However, both approaches rely on prespecified organoids, limiting their ability to recapitulate developmental reciprocity during lineage bifurcation. This constraint is exacerbated when germ layer-mismatched organ pairs (e.g., heart-brain) are combined, as conflicting signaling requirements disrupt tissue-specific maturation [
31]. Our TGCO-Chip-derived codevelopoid model overcomes these limitations by enabling parallel differentiation of two organ lineages from a shared EB within compartmentalized microenvironments. Spatial segregation prevents signaling interference, preserving lineage-specific developmental trajectories, whereas controlled paracrine exchange maintains physiological crosstalk. This dual functionality captures early organogenic interactions—such as sonic hedgehog (SHH)-mediated cardiogenesis and Wnt-driven neuroepithelial patterning—that precede tissue specification. Unlike assembloids or multiorgan chips, the TGCO-Chip supports germ layer-mismatched systems (e.g., heart-brain) and lineage-shared tissues (e.g., forebrain-midbrain), providing a universal framework for investigating developmental interdependencies and pathological cascades.
Recent advances in stem cell biology have yielded multiorganoid models recapitulating developmental interactions. Specifically, Yin et al. [
32] demonstrated neuromusculoskeletal tri-tissue assembloids by temporally modulating differentiation cues, revealing skeletal-muscular crosstalk critical for neuromuscular junction formation. Similarly, Silva et al. [
31] engineered cardiac-gut codifferentiation from mesendodermal progenitors, showing that endoderm-derived signals enhance cardiomyocyte subtype specification. Although these models are pioneering in nature, their differentiation methodologies heavily depend on researcher experience and professional knowledge, limiting their extension to broader tissue or organoid interaction models. In contrast, our TGCO-Chip enables the codifferentiation of two interconnected distinct organoids in a well-controlled microenvironment on the basis of their specific differentiation protocols. Thus, our technique can be easily extended to generate diverse types of organoid or tissue complexes that are not limited to heart-brain, forebrain-midbrain, or bone-cartilage interaction models.
Recently, increasing evidence has indicated the crucial role of the heart-brain axis in the development and homeostasis of the heart and brain [
33,
34]. The heart-brain axis facilitates the exchange of signals between these two organs, influencing their structure and function [
33,
35]. During embryogenesis, the heart and brain interact through neural, immune, and endocrine pathways, ensuring coordinated growth and maturation. The heart-brain axis also maintains physiological homeostasis throughout life, regulating processes such as heart rate, blood pressure, and neuroplasticity. Disruptions in this axis have been linked to various diseases, including congenital heart defects and neurological disorders [
36]. Recent research has begun to elucidate the complex mechanisms underlying these interactions, highlighting the potential for targeted therapies to address related pathologies. However, most current studies on the heart-brain axis have been based on conventional animal models, which cannot accurately simulate human heart-brain interactions because of interspecies differences in anatomy, metabolism, and physiology. Although clinical imaging magnetic resonance imaging/positron emission tomography (MRI/PET) provides whole-organ correlation data, its resolution is insufficient to elucidate the molecular mechanisms or single-cell dynamics underlying heart-brain crosstalk.
hPSC-derived microphysiological systems (MPSs) offer powerful platforms for studying the heart-brain axis and dissecting its molecular underpinnings. While early neurocardiac gastruloids have captured initial lineage interactions [
37,
38], existing models fail to establish functionally mature heart-brain systems. In this study, we employed a TGCO-Chip to construct a heart-brain codevelopment model in which interconnected heart and brain organoids codifferentiated from the same EB. Specifically, the characterization of brain compartments during HBC differentiation revealed PAX6
+/ISL1
+ anterior-posterior neural patterning, the formation of neural rosette structures, and the emergence of neuroelectrophysiological activities (neural spikes and bursts), reflecting
in vivo neural tube formation and subsequent neurogenesis and neuronal maturation [
39]. Moreover, heart compartments during HBC differentiation showed rhythmic beating starting on day 15, with the beating frequency increasing over time and reaching 120 BPM and the mean cardiac spike amplitude increasing from (0.54 ± 0.08) to (1.96 ± 0.92) mV by day 28, similar to the heart development level at the 6-7 week stage
in vivo [
40]. Collectively, the HBCs successfully reproduced the gradual development and maturation of the heart and brain separately. Recently, increasing evidence has indicated that the nervous and cardiovascular systems mature together through bidirectional mechanisms. Cardiac tissue influences sympathetic neuron growth, whereas nerve fibers affect cardiomyocyte maturation via neurotransmitters [
41,
42]. Cardiac innervation nerves play a crucial role in regulating cardiomyocyte proliferation and cardiac regeneration [
43]. Sympathetic innervation is crucial for proper cardiomyocyte calcium handling and contraction [
44]. In our HBC model, we observed neural circuits projecting from the brain compartment to the heart compartment using retrograde neural tracing. Gene expression analysis revealed significant upregulation of the expression of cardiac maturation markers (e.g.,
CAV3 and
HAND2) [
45] and neural maturation genes (e.g.,
MAPT and
GJB1) in the cardiac compartments of the HBCs compared with those in the HHO group (
Fig. 4(c)). These results imply that brain-derived neurotrophic factors or electrical activities enhance cardiac development. Notably, compared with the HHO group, the HBC group exhibited significantly greater myocardial contractility (
Fig. 4(e), Videos S3 and S4). These results align with those of prior studies showing that neural tissues improve cardiac sarcomere organization and electromechanical coupling [
44]. Collectively, our HBC model, for the first time, emulates early-stage heart-brain interactions. In this study, we did not generate a brain-brain organoid group as a control to examine the effects of cardiac activity on brain organoid differentiation. We will investigate the influence of cardiac activity on brain organoid differentiation in future work. We expect this HBC model to serve as a high-fidelity human physiologically relevant microphysiological model to study heart-brain development and diseases. Additionally, we believe that TGCO-Chips will provide a transformative technical route to generate bioengineered complex multiple organ or tissue interaction models for basic research and drug discovery.
5. Limitations of the study
With respect to the technical aspect of this study, we did not incorporate the TGCO-Chip with a dynamic culture, which led to the occurrence of necrosis in the brain compartment of the HBCs. In future work, we will update the TGCO-Chip culture system with milifluidic perfusion to improve oxygen delivery and eliminate necrosis in HBCs [
46]. The current trans-germ-layer heart-brain codevelopment model remains in the nascent stage of simulating reciprocal interactions between heart and brain organoids. Future investigations should prioritize neurodevelopmental regulatory mechanisms and incorporate optogenetic manipulation to enhance synaptic maturation and functional coupling efficacy. While the HBC establishes a foundational framework for investigating cardiocerebral crosstalk (e.g., autism spectrum disorders and epileptic cardiomyopathy), its applicability for modeling complex multiorgan pathologies requires systematic validation. This necessitates pathological modeling using patient-derived or genetically engineered hPSC/hESC lines to establish robust genotype-phenotype correlations.
PDF
(4582KB)
