Spatially controlled construction of assembloids using bioprinting

Abstract

The biofabrication of three-dimensional (3D) tissues that recapitulate organ-specific architecture and function would benefit from temporal and spatial control of cell-cell interactions. Bioprinting, while potentially capable of achieving such control, is poorly suited to organoids with conserved cytoarchitectures that are susceptible to plastic deformation. Here, we develop a platform, termed Spatially Patterned Organoid Transfer (SPOT), consisting of an iron-oxide nanoparticle laden hydrogel and magnetized 3D printer to enable the controlled lifting, transport, and deposition of organoids. We identify cellulose nanofibers as both an ideal biomaterial for encasing organoids with magnetic nanoparticles and a shear-thinning, self-healing support hydrogel for maintaining the spatial positioning of organoids to facilitate the generation of assembloids. We leverage SPOT to create precisely arranged assembloids composed of human pluripotent stem cell-derived neural organoids and patient-derived glioma organoids. In doing so, we demonstrate the potential for the SPOT platform to construct assembloids which recapitulate key developmental processes and disease etiologies.

Fig. 1. Magnetic lifting maintains the structural integrity of neural organoids.

a Diameter measurements of MSC and HUVEC spheroids, and hiPSC-derived ventral and dorsal forebrain neural organoids at increasing days of culture. Each data point represents a distinct spheroid or organoid (MSC n = 25, HUVEC n = 27, D25 Ventral n = 27, D50 Ventral n = 25, D100 Ventral n = 25, D25 Dorsal n = 25, D50 Dorsal n = 25, D100 Dorsal n = 12). p values for each diameter comparison are as follows: MSC vs. HUVEC p = 0.9978, all other shown comparisons p < 0.0001. b Mass measurements of spheroids and neural organoids. Each data point represents an average of five neural organoids. p values for each mass comparison are as follows: D25 Ventral vs. D50 Ventral p = 0.1499, D25 Ventral vs. D100 Ventral p = 0.0094, D50 Ventral vs. D100 Ventral p = 0.7933, D25 Dorsal vs. D50 Dorsal p < 0.0001, D25 Dorsal vs. D100 Dorsal p < 0.0001, D50 Dorsal vs. D100 Dorsal p = 0.4743. c Vacuum pressure required to lift neural organoids of increasing diameters within a liquid medium. Each data point represents a distinct organoid (1.0 mm n = 4, 1.5 mm n = 4, 2.0 mm n = 4, 2.5 mm n = 4). p values for minimum lifting pressure comparisons are as follows: 1.0 mm vs. 1.5 mm p = 0.0007, 1.5 mm vs. 2.0 mm p < 0.0001, 2.0 mm vs. 2.5 mm p = 0.0026. d Apparent surface tension of spheroids and neural organoids. Each data point represents a distinct spheroid or organoid (MSC n = 4, HUVEC n = 2, D25 Ventral n = 4, D50 Ventral n = 4, D25 Dorsal n = 4, D50 Dorsal n = 4). p values for apparent surface tension comparisons are as follows: MSC vs. HUVEC p = 0.0235, all other shown comparisons p < 0.0001. e Schematic of vacuum aspiration-assisted lifting of neural organoids. f Representative brightfield (BF) images of a neural organoid prior to vacuum aspiration. g Representative BF images of a neural organoid post vacuum aspiration (6 mmHg). h Representative BF image of a neural organoid that has undergone complete deformation (i.e., is no longer spherical) post vacuum aspiration (6 mmHg). i Quantification of the extent of deformation as a function of the applied vacuum pressure. Each color represents a single neural organoid (n = 3). j Representative quantification of neural organoid deformation during and immediately following vacuum aspiration (6 mmHg, shown in blue). k Long-term neural organoid deformation in response to two vacuum pressures: 6 mmHg and 10 mmHg. Each set of data points connected with a line represents a single biological replicate (6 mmHg n = 4, 10 mmHg n = 4). l Schematic of magnetic lifting of neural organoids. m Representative BF image of a neural organoid post magnetic lifting. Statistical analyses performed as one-way ANOVA with Tukey multiple comparisons test. Unless otherwise noted, all data points represent distinct biological replicates. Data plotted as mean ± SD where *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and ns not significant.

Fig. 2. Cellulose nanofibers mediate the bioprinting of neural organoids with SPOT.

a Schematic of the SPOT platform. b Representative images of a neural organoid being coated in a magnetic nanoparticle (MNP)-laden CNF ink, lifted and transferred into a CNF support scaffold by a magnetic rod attached to an electromagnet-modified 3D printer, and released at a desired position within the support scaffold. c Representative viscosity measurements of inks with 1 wt% MNP and various CNF wt%. d Quantification of the relative degree of MNP dispersion within CNF inks of various wt% over 15 minutes (n = 4 formulations); error bars represent standard deviation. Inset: Representative image of MNP dispersion within a microcentrifuge tube. e Representative image of MNP-laden CNF inks extruded over the top of neural organoids. f Representative storage modulus (filled circles) and loss modulus (open circles) of 0.5 wt% CNF support scaffold exposed to cyclical periods of low (0.1%) and high (300%) strain to evaluate the ability of the material to shear thin and self-heal. Percent G’ recovery following one cycle (mean ± SD): 86.8 ± 6.0 (n = 4 gels); percent G’ recovery following two cycles (mean ± SD): 84.3 ± 25.6 (n = 4 gels). g Storage modulus of CNF support scaffolds of various wt%. Each data point represents a distinct gel (n = 4 for all formulations). p values for storage moduli comparisons are as follows: 0.25 wt% CNF vs. 0.50 wt% CNF p = 0.9001, 0.50 wt% CNF vs. 1.00 wt% CNF p = 0.0097, 1.00 wt% CNF vs. 1.50 wt% CNF p < 0.0001, h Representative viscosity measurements of 0.5 wt% CNF support scaffolds in response to treatment with various concentrations of cellulase. i Quantification of the extent of MNP coverage on the surface of a neural organoid following coating with 1 wt% MNPs in DPBS or a 1 wt% MNP-laden 0.025 wt% CNF ink. Each data point represents a different organoid (n = 3). p values for MNP coverage comparisons are as follows: Pre-Lifting p = 0.048, Post-Release p = 0.0162, Post-Cellulase p = 0.851. j Representative BF image of a neural organoid following SPOT. Statistical analyses were performed as one-way ANOVA with Tukey multiple comparisons test or two-way ANOVA with either Dunnett’s multiple comparisons test or Šídák’s multiple comparisons test. Unless otherwise noted, all data points represent distinct biological replicates. Data plotted as mean ± SD where *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and ns not significant.

Fig. 3. SPOT imparts spatial control over the construction of neural assembloids.

a Schematic of the potential automation of SPOT. Specific media formulations (depicted as red or green) can be portioned into distinct channels within a custom-built chip designed to facilitate the maintenance and differentiation of tissue-specific spheroids and organoids. Dotted lines represent the potential path an electromagnet-modified 3D printer could take to create an assembloid. b Precision, in X and Y dimensions, of automated alginate microgel transfer. Each data point represents a single microgel wherein each microgel yielded data in both X and Y dimensions (n = 10). c Drift (i.e., the distance between where an alginate microgel was intended to be deposited and where the microgel settled) in the Z dimension as a function of microgel diameter. Each data point represents a single microgel (n = 17). Line of best fit: y = −0.4183x + 1.421. Coefficient of determination: r² = 0.45. d Positional stability (i.e., Z dimensional drift relative to the initial displacement) over time of alginate microgels over 72 hours (n = 7); error bars represent standard deviation. e Representative fluorescence image of an eGFP-expressing ventral forebrain neural organoid fused to two mScarlet-expressing dorsal forebrain neural organoids. f Representative fluorescence image of two eGFP-expressing ventral forebrain neural organoids, two mScarlet-expressing dorsal forebrain neural organoids, and two non-fluorescent dorsal forebrain neural organoids from distinct hiPSC lines fused in a ring. g Representative fluorescence image of an eGFP-expressing ventral forebrain neural organoid fused to three mScarlet-expressing dorsal forebrain neural organoids in a multi-layered pyramid in which the ventral organoid is above the dorsal organoids. h Representative immunofluorescence (IF) image of a ventral forebrain neural organoid integrated with a dorsal forebrain neural organoid. i Representative IF image of a ventral forebrain neural organoid integrated with a dorsal forebrain neural organoid with regions of higher magnification to illustrate cell migration. Unless otherwise noted, all data points represent distinct biological replicates. Data plotted as mean ± SD.

Fig. 4. Glioma assembloids predict tumor progression-specific drug response.

a Representative image of a well plate with hiPSC-derived neural organoids fused to patient-derived DIPG organoids. b Representative BF and IF images of a three-part assembloid in which two distinct DIPG organoids, derived from either the pons, which is the tumor origination site (DIPGXIII-P), or the frontal lobe, a brain region into which the tumor metastasized (DIPGXIII-FL), are fused to a dorsal forebrain neural organoid. c Representative IF staining of the apoptosis marker cleaved caspase-3 across an array of permutations of DIPG organoids fused to neural organoids in which a subset of the assembloids were treated with 200 nM panobinostat. d Quantification of the relative degree of apoptosis as determined by cleaved caspase-3 staining within DIPG organoids normalized by DAPI and relative to the untreated control organoids. Each data point represents a different DIPG organoid either alone or within an assembloid (DIPGXIII-P n = 3, DIPGXIII-FL n = 3, Ventral-DIPGXIII-P n = 3, Ventral-DIPGXIII-FL n = 3, Dorsal-DIPGXIII-P n = 3, Dorsal-DIPGXIII-FL n = 3). p values for shown apoptosis comparisons are as follows (additional p values listed in Supplementary Data 1): Pontine:DIPG Only vs. Frontal Lobe:DIPG Only p = 0.9958, Pontine:DIPG Only vs. Pontine:Ventral Fusion p < 0.0001, Pontine:DIPG Only vs. Pontine:Dorsal Fusion p = 0.001, Pontine:Ventral Fusion vs. Frontal Lobe:Ventral Fusion p < 0.0001, Pontine:Dorsal Fusion vs. Frontal Lobe:Dorsal Fusion p = 0.001. e Representative IF staining of H3K27M across an array of permutations of DIPG organoids fused to neural organoids in which a subset of the assembloids were treated with 200 nM panobinostat. f Quantification of the relative number of H3K27M-expressing cells. H3K27M staining within DIPG organoids was normalized by DAPI and shown relative to the untreated control organoids. Each data point represents a different DIPG organoid either alone or within an assembloid (DIPGXIII-P n = 3, DIPGXIII-FL n = 3, Ventral-DIPGXIII-P n = 3, Ventral-DIPGXIII-FL n = 3, Dorsal-DIPGXIII-P n = 3, Dorsal-DIPGXIII-FL n = 3). p values for shown H3K27M comparisons are as follows (additional p values listed in Supplementary Data 1): Pontine:DIPG Only vs. Frontal Lobe:DIPG Only p = 0.983, Pontine:Ventral Fusion vs. Frontal Lobe:Ventral Fusion p = 0.0206, Pontine:Dorsal Fusion vs. Frontal Lobe:Dorsal Fusion p = 0.0055. Statistical analyses performed as two-way ANOVA with Šídák’s multiple comparisons test. Unless otherwise noted, all data points represent distinct biological replicates. Data plotted as mean ± SD where *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and ns not significant.