Abstract
During mammalian embryogenesis, axial elongation of the neural tube is critical for establishing the anterior-posterior body axis, but is difficult to interrogate directly because it occurs post-implantation. Here we report an organoid model of neural tube extension using human induced pluripotent stem cell (hiPSC) aggregates that recapitulates morphologic and gene expression patterns of neural tube development. Axial extending organoids consisted of longitudinally elongated epithelial compartments and contained TBXT(+)SOX2(+) neuromesodermal progenitors, PAX6(+) Nestin(+) neural progenitor populations, and MEOX1(+) paraxial mesoderm populations. Wnt agonism stimulated axial extensions in a dose-dependent manner and elongated organoids displayed regionalized rostral-caudal HOX gene expression, with hindbrain (HOXB1) expression distinct from brachial (HOXC6) and thoracic (HOXB9) expression. Finally, CRISPR interference-mediated silencing of BMP inhibitors induced elongation phenotypes that mimicked murine knockout models. These results indicate the potent morphogenic capacity of hiPSC organoids to undergo significant axial elongation in a manner that mimics early human nervous system development.
One sentence summary Here, the authors introduce an organoid model for neural tube development that demonstrates robust Wnt-dependent axial elongation, epithelial compartmentalization, establishment of neural and mesodermal progenitor populations, and morphogenic responsiveness to changes in BMP signaling.
Introduction
Proper development of multicellular organisms depends on the regulated emergence and coordinated organization of multiple cell types to form tissue structures and organs. An essential example of such morphogenesis is the creation and extension of the neural tube, which eventually forms the spinal cord. Defects in this process result in severe congenital abnormalities (1), highlighting the clinical importance of understanding human spinal cord development. However, many of the molecular mechanisms and cellular behaviors that regulate formation of the central nervous system remain unknown because neural tube morphogenesis is a dynamic process that occurs post-implantation (2, 3).
Formation of the central nervous system in amniotes begins at gastrulation, when neuroectodermal progenitors establish the anterior neural plate. This structure folds to form the anterior neural tube, which will ultimately become the brain and cervical spinal cord. By contrast, the remainder of the spinal cord is generated from a separate pool of neuromesodermal progenitors (NMPs), that reside at the node-streak border and caudal-lateral epiblast immediately post-gastrulation, and in the chordoneural hinge of the tail bud later in development (2, 4-6). NMPs are bipotent progenitors that contribute to the developing neural tube and the surrounding paraxial mesoderm, facilitating extension of the embryo along the anterior-posterior body axis. The subsequent differentiation, patterning, and development of the multiple cell types within the elongating neural tube relies on a series of overlapping morphogen gradients that extend along both the anterior-posterior and the dorsal-ventral axes of the embryo. The anterior-posterior axis of the neural tube is patterned by the Wingless-type MMTV integration site family (Wnts), the fibroblast growth factors (FGFs) produced from the tailbud, and retinoic acid (RA) released from the adjacent somites, which regulate progenitor proliferation and convergent extension (7-9). Dorsal-ventral patterning of the neural tube is dictated by opposing gradients of Sonic hedgehog (Shh) released ventrally from the notochord and floor plate, and bone morphogenic proteins (BMPs) released from the roof plate (10-12).
The development of differentiation protocols for pluripotent stem cells has revealed key factors, particularly morphogen combinations, regulating the emergence of neuronal populations in the developing spinal cord. These advances have enabled robust production of neuronal subtypes in vitro (13, 14). In parallel, organoid models capable of recapitulating structures reminiscent of early embryonic tissue have been developed as tangible platforms for disease modeling, drug testing, and mechanistic study of development (15). More specifically, organoids displaying axonal extensions through neuromuscular junctions (16), mimicking spinal cord elongation through budding extensions of NMPs (17), or 3D gastruloid elongation (18) have extended the ability of in vitro systems to study developmental morphogenesis. However, to date, no organoid model exhibits both extensive unidirectional elongation and emergence of tube-like structures mirroring the endogenous cell behaviors that contribute to human spinal cord morphogenesis. Here we report the development of an organoid model of axial extension derived from human induced pluripotent stem cells (hiPSCs) that recapitulates the uni-axial elongation and morphogenesis of the neural tube in addition to regionalized gene expression profiles of spinal cord development. This model enables the direct examination of features of early human spinal cord development and patterning.
Emergence of Axial Extensions from Neuronal Organoids
In order to model early spinal cord development, we modified a previously described neuronal differentiation protocol (13) for generating hindbrain-level interneurons, with the goal of yielding caudalized neural populations. Organoids were formed by seeding singly-dissociated cells into non-adherent pyramidal inverted wells (10,000 cells/well) and allowed to aggregate for 18-24 hours before being transferred to rotary suspension culture in either 6-well plates or 10cm dishes (Fig. 1A). To caudalize the cells from hindbrain to spinal cord identity (19), we added the small-molecule Wnt agonist CHIR99021 at a concentration of 2μM starting two days prior to aggregation and continuing through day 7 of differentiation (Fig. 1B). After 3-5 days in suspension culture, pronounced singular extensions emerged from the organoid bodies (Fig. 1C).
Elongated organoids appeared to be continuous, with linear striations extending longitudinally from a central, more radially symmetric anterior aggregate body (Fig. 1D). To confirm that elongation did not result from the fusion of multiple organoids during suspension culture, individual organoids from 10cm dishes were transferred into single wells of low-adherent, 96-well plates after 5 days of differentiation, and subsequently imaged for 48 hours, during the time period when extension appeared to be most robust. Approximately 70% of organoids grown in 10cm dishes, and 85% of organoids transferred to 96-well plates demonstrated at least partial elongation (Fig. 1E, S1A, B). Extending organoids doubled the ratio of their longest axis relative to their shortest axis, whereas the length of the shortest axis did not differ between non-extending and extending organoids (Fig. 1E, Fig. S1C, Movie S1-2). Additionally, extending and non-extending organoids displayed several quantifiable morphometric differences (Fig. 1E, Fig. S1C). Time lapse microscopy definitively demonstrated that axial elongation of caudalized neural organoids results from autonomous symmetry-breaking that is comparable to axial elongation observed in model organisms.
In follow-up studies, extensions were found to occur for organoids of various sizes (500, 3,000, or 10,000 cells at seeding) and continued to grow until day 10 of culture. However, the degree of extension varied depending on culture conditions. To determine if culture density affected the efficiency of extension, we varied the number of 3,000-cell organoids cultured in a fixed 2 mL volume, culturing aggregates at 68, 135, or 270 organoids per mL in suspension (Fig. S2). While extension occurred in all conditions, extensions were more frequent and became more pronounced as organoid density decreased; however, at the highest density, elongation morphology was lost by day 15. Density-dependent effects were also observed when we cultured a fixed number of 10,000-cell organoids (270) in low (2 mL) or high (12 mL) media volumes, resulting in densities of 135 or 23 organoids per mL. Robustly elongating organoids formed when cultured at low density, whereas organoids failed to elongate when an equivalent number of organoids were cultured at high density (Fig. S3). In the lowest density condition (23 organoids per mL), extensions persisted through day 15 (Fig. 1C), suggesting that density-mediated signaling parameters, such as paracrine effects or nutrient deprivation, play a role in inducing and maintaining axial extensions.
After analyzing the gross morphology, we examined the structural and gene expression differences between elongating and non-elongating organoids. Histological analysis of extending organoids revealed multiple internal elongated epithelial compartments separated by regions devoid of cells (Fig. 1F, Fig. S4). By comparison, non-extending organoids consisted only of small rosette structures distributed radially throughout the interior. To investigate whether structural differences were associated with distinct spinal progenitor populations, we performed quantitative PCR (qPCR) to identify changes in neural and mesodermal expression patterns. Both extending and non-extending organoids expressed neural progenitor markers such as SOX2, PAX6, HB9, and NeuN by day 17 of differentiation, however the levels of SOX2, HB9, and NeuN were greater in extending aggregates at that timepoint (Fig. 1G). In addition to genes expressed in the early neural plate and neural tube (SOX2 and PAX6), Brachyury (TBXT), associated with primitive streak and mesoderm differentiation, was expressed in all organoids. However, NKX1.2, associated with embryonic axial extension, was expressed for a longer duration and at higher levels in elongating organoids (Fig. 1G). The expression of NKX1.2, SOX2, and TBXT within organoids raised the possibility that NMPs and paraxial mesoderm present within organoids might account for the observed morphogenic phenomena.
Extending Organoids Display Polarization and Region Patterning
We next sought to understand how differences in the identity and spatial distribution of the cell populations in extending and non-extending organoids contributed to differential organoid morphologies. We first performed immunostaining of histological sections to investigate the cellular composition of extending organoids in relation to observed TBXT expression. At day 6, after the start of extension, organoids contained regions of co-localized SOX2 and TBXT expression (Fig. 2A, white arrows), indicative of NMPs, while the surrounding cells expressed Nestin, a marker of neural commitment (Fig. 2A). To more comprehensively visualize the location of TBXT(+) cells in 3D, we imaged extending and non-extending organoids via light-sheet microscopy at day 7 of differentiation. In extending organoids, TBXT(+) cells were localized within the extensions, whereas in non-extending organoids TBXT(+) cells were dispersed around the edges of the organoid body (Fig. 2B, Movies S3-4). In most extending organoids, the TBXT(+) populations formed a streak lining the extension, analogous to the morphology of the notochord, and accumulated at the end of the extension (Fig. S5A). In addition, extending organoids contained a higher percentage of TBXT(+) cells (Fig. S5B). Interestingly, staining in organoids at day 3 (before extension occurred) revealed distinct sorting patterns of either radially symmetrical populations of TBXT(+) cells or polarized clustering of TBXT(+) cells within the organoids (Fig. S6). Both extending and non-extending organoids contained cells expressing the neuronal progenitor marker Nestin and β3-tubulin at day 10 of differentiation, indicating that the majority of cells differentiated to a neural fate (Fig. 2B, Fig. S7, Movies S5-6). However, in the extending organoids, the neural-committed cells organized into distinct compartmentalized layers surrounding lumens within the extensions.
To further interrogate the distinct striations observed by brightfield microscopy and the inner epithelial layers seen in histology, we examined cell polarity by staining for collagen IV (COL IV), a marker of the basement membrane, and zona occluden-1 (ZO-1), a marker of tight junctions at cellular apical domains. An increase in basement membrane deposition and polarized cells was observed between days 7 and 10 of differentiation in extending organoids (Fig. S8). After 10 days, ZO-1 was apparent at the apical domain of cells in the outer epithelial layers of the extensions, as well as in cells facing inward toward the internal lumens (Fig. S8,9). Distinct layers of basement membrane marked by collagen IV were observed basally to ZO1, forming a bilayer between lumens and organoid exterior (Fig. 2B, Fig. S8,9. Movies S7-8). Bilayers of cells were also apparent in non-extending organoids (Fig. S9), but to a lesser extent since the internal lumens were much smaller.
In order to ascertain whether extensions resulted from a polarized population of proliferative cells, we examined EdU incorporation on days 6, 7 and 9 of differentiation (Fig. 2C). A two-hour EdU pulse labeled over 50% of cells in extending organoids, but proliferation was not spatially restricted, indicating that proliferation of cells throughout organoids contributed to their growth. To examine transcriptional differences between the two poles of extending organoids, we manually dissected the “anterior”, or main organoid body from the “posterior”, or extending portion of day 9 extending organoids and performed comparative qPCR (Fig. 2D, Fig. S10).
Expression of hindbrain Hox genes (HOXA2, HOXB4) was enriched in the anterior portion of the organoids. Correspondingly, the posterior regions had increased expression of more caudal HOX genes, particularly HOXB9 and HOXC10, associated with the thoracic and lumbar regions of the spine, respectively (9, 20). The posterior extending regions also expressed higher levels of transcripts associated with caudalization and the tail bud, such as CDX2, FGF8, HES7 and TBXT, indicating that the extending regions of the organoids had a more caudal fate specification than the anterior. The posterior extending portions of the organoids also displayed increased expression of ALDH1A2, indicating an increase in retinoic acid signaling, potentially from TBXT (+) mesoderm.
To determine how well extension morphology and localized HOX gene expression mimicked the regionalization of the spinal cord, we examined HOX gene transcripts marking hindbrain (HOXB1), brachial (HOXC6), and thoracic (HOXB9) regions of the neural tube via RNAscope in day 7 and day 10 organoids (Fig. 2E,F). While non-extending organoids expressed both hindbrain and brachial HOX genes, the expression was distributed radially, and the expression of HOXB9 was delayed relative to extending organoids. In contrast, extending organoids displayed regionalized expression of hindbrain, brachial and thoracic HOX genes along the axis of extension, as seen in the neural tube of developing vertebrates. Specifically, we observed enrichment of HOXB1 (hindbrain) in the central mass of the organoid, and HOXC6 (brachial) and HOXB9 (thoracic) in the extensions (Fig. 2E). Interestingly, HOXC6 and HOXB9 often overlapped at day 7 within extending organoids (Fig. 2E), however by day 10 HOXB9 was more pronounced than HOXC6 in the extension regions, suggesting a transition to a more posterior fate. Although non-extending organoids expressed multiple HOX genes, regionalized expression was infrequently detected and non-extending organoids did not express detectable HOXB9 levels until day 10 (Fig. 2F). These data indicate that while organoid extension is not a requirement for posterior HOX expression, extension enables the stratification of distinct HOX domains. Overall, extending and non-extending organoids maintained similar HOX profiles, but the cell behaviors that contribute to symmetry breaking and subsequent elongation, perhaps through the establishment and maintenance of an NMP pool, allowed for HOX domain segregation similar to that observed in neural tube development.
Wnt Pathway Activation Increases Extension Formation and SOX2(+)TBXT(+) Progenitors
Caudalization of the spinal cord (21) and differentiation of NMPs (2, 22) are critically dependent on Wnt signaling. Therefore, we assessed how increasing canonical Wnt signaling impacted organoid elongation. Escalating doses of CHIR (2μM, 4μM, and 6μM) enhanced the length of extending organoids (Figure 3A,B), and increased the ratio of longest:shortest radial axis in response to Wnt early in the differentiation (Fig. 3C,D, Fig. S11A). Initially, differentiation studies were conducted with a commonly used male hiPSC line (WTC) (23, 24). In contrast, organoids created with a female hiPSC cell line (WTB) failed to elongate in the presence of 2μM CHIR, but did respond to increasing doses of WNT activation (Fig. 3D, Fig. S11B), though not to the same extent as WTC-derived organoids. At later timepoints, aggregates in both cell lines and at all but the highest CHIR dose become more round, suggesting that early polarized extension succumbs to later radial expansion. Overall, the variation in extension length between cell lines implies that inherent differences between hiPSC genomes may influence an organoid’s capacity for axial extension, similar to previously reported differences between various human pluripotent stem cell lines (25, 26).
At day 10 of differentiation, histological sections of extending WTC-derived organoids treated with 4μM- and 6μM-CHIR revealed cells that co-expressed SOX2 and TBXT, with an increasing number of TBXT(+) cells as Wnt signaling agonist was increased (Fig 3E,F). Additionally, TBXT(+) cells were typically surrounded by a bilayer of TBXT(-) cells (Fig. 3E). Organoids exposed to 4μM- and 6μM-CHIR expressed SOX2 and PAX6 in distinct regions of the organoids, whereas organoids exposed to 2μM CHIR displayed SOX2 and PAX6 expression throughout, suggesting a loss of neural progenitor states or a non-neuronal population in higher Wnt signaling environments. All organoids displayed N-cadherin and βIII-Tubulin expression, indicating the emergence of more mature populations of neurons; however, regions without N-cadherin or βIII-Tubulin were observed more frequently in organoids exposed to higher doses of CHIR, suggesting a Wnt-dependent reduction in the overall neural population (Fig. 3E). Overall, increased stimulation of the Wnt pathway increased the NMP progenitor pool, facilitated progenitor maturation, and augmented organoid extension.
Similar Transcriptional Profiles within Extending and Non-extending Organoids
In order to identify the cell diversity within extending organoids and the differences between organoids that did and did not extend, we performed single-cell transcriptomics at day 10 of differentiation. In order directly compare elongating and non-elongating organoids under nearly identical conditions, differentiation was conducted under the original conditions with 2μM CHIR. Shared nearest neighbor computation revealed six clusters containing roughly equivalent percentages of cells from extending and non-extending organoids, with the exception of cluster 2, which contained a majority of cells from non-extending organoids (Fig. 4A,B). Gene ontology (GO) terms for each cluster revealed similar biological processes, related largely to pattern specification and hind limb morphogenesis (Fig. 4C). Neural markers such as SOX2 and Nestin were expressed in all clusters (Fig. 4D,E), however, clusters 0 and 1 had higher expression levels of genes associated with early neural progenitors such as SOX2, PAX6 and NEUROG2 (Fig. 4D, Fig. S12A). Clusters 0 and 1 contained cells with high expression of transcription factors and morphogens associated with the caudal lateral epiblast (CDX2, NKX1.2 and FGF8) (Fig. 4D,E). In contrast, MIXL1, MEOX1, CYP1B1, MEOX2, and PAX3 were almost exclusively expressed in clusters 2-3, suggesting that these clusters contained mesodermal cells (Fig. 4D,E, Fig. S12A). We did not observe expression of markers associated with the endodermal germ lineage such as SOX17, and observed only very low expression of FOXA1 and PAX9 in <1% of cells (Fig. 4E), indicating that both the non-extending and extending organoids consisted entirely of neural and mesodermal cell types, but not endoderm, in contrast to previously reported gastruloid models (18, 36). Further, no single cell cluster expressed genes associated with the developing notochord (Fig S12B), however, single cell transcriptomic data did recapitulate HOX gene expression patterns within organoids (Fig. 4F). Extending and non-extending organoids both expressed genes associated with the hindbrain and the brachial and thoracic regions of the neural tube, but only a minor increase in HOX gene expression was detected in extending relative to non-extending organoids (average 1.3-fold change), which was in agreement with our RNAScope analysis. Overall, these gene expression results indicate similar cell specification within non-extending and extending organoids, suggesting that post-transcriptional regulation or morphogenic differences account for the distinct organoid phenotypes.
Manipulation of Endogenous BMP Signaling Influences Dorsalization and Degree of Extension
BMP signaling is critical to establish both the anterior-posterior Wnt and FGF gradients that drive axial elongation, and the BMP-SHH dorsal-ventral gradient that organizes the developing spinal cord (8, 17, 21, 27). In vertebrate development, Noggin is expressed along the developing neural tube and inhibits dorsally expressed BMP. Noggin deletion in murine models leads to increased BMP signaling at the posterior axis, failure of neural tube closure, and elongation of the developing tail (27). Chordin is a BMP shuttling molecule that concentrates BMP at the dorsal side of the neural tube; in mice, deletion of Chordin results in a shortened body axis and underdeveloped anterior spine (28). Therefore, we assessed the role of endogenous BMP signaling in extending organoids by knocking down (KD) Noggin and Chordin. We introduced RNA guides targeting Noggin and Chordin into a Lamin-B GFP labeled WTC hiPSC line harboring a doxycycline (DOX) inducible CRISPR interference system (CRISPRi) (29, 30), and supplemented the media with DOX at the start of differentiation (Fig. 5A). Noggin and Chordin gene expression significantly decreased 5 days after the addition of DOX (Fig. 5B). Both Chordin and Noggin KD resulted in increased extension compared to wildtype organoids (Fig 5C,D). However, the timing of extension differed, with Chordin KD organoids expanding rapidly within the first 7 days of differentiation then becoming more circular after day 10, while Noggin KD extended continuously, becoming more polarized through day 13 relative to WT organoids. The pronounced extensions in Noggin KD organoids appear analogous to Noggin knockout mice that develop elongated tails. In contrast, Chordin KD organoids initially developed blebs without clear anterior-posterior axis identity, and these disordered, kinked outgrowths were sustained through approximately day 11, similar to the shortened body axis characteristic of Chordin knockout mice.
In addition to contrasting the morphological phenotypes of different organoids, the relative proportions of dorsal-ventral markers of spinal cord progenitor neuron populations were assessed by qPCR at day 17 of differentiation. Expression of LIM Homeobox Protein 2 (LHX2), a marker of dI1 dorsal interneurons, did not significantly change under knockdown conditions. However, expression of LIM Homeobox Protein 5 (LHX5), a marker of V0 ventral interneuron populations, significantly increased under both Noggin and Chordin knockdown conditions (Fig. 5E,F). Expression of Engrailed1 (EN1), a marker of V1 interneuron progenitors, was increased in knockdown populations, while CHX10, a marker of V2a interneurons, and OLIG2, a marker of motor neurons, were unaffected. Attenuation of BMP inhibition would be expected to increase dorsal neural populations, and these data indicate neural progenitors in Chordin and Noggin knockdown lines have a propensity to express more dorsalized markers relative to wildtype. Finally, we assayed markers of neural maturation and its regulation, including Sox2, Pax6, and Hes1. Hes1, which maintains neural progenitors in their immature state, was not significantly upregulated in the Chordin knockdown, nor were the neural progenitor markers Sox2 or Pax6. Altogether these results indicate that BMP signaling impacts the anterior-posterior morphology of the extending organoids in a manner that is similar to neural tube development in vivo.
Discussion
Here, we demonstrate that hiPSC-derived organoids can model axial extension via the specification of progenitors that dictate axial elongation and neural tube formation. Our model allows for robust axial extension as a function of both culture density and Wnt signaling, where either reduced organoid density or increased small molecule Wnt agonism promote elongation phenotypes and increased proportion of TBXT(+) cells. Within extending organoids, we observed a persistent NMP population and TBXT(+) streak along the extension of the organoid axis, reminiscent of similar phenotypic populations present in the caudal lateral epiblast, notochord, and later in the tail bud during axial elongation. Although both extending and non-extending organoids contained cells expressing cervical through thoracic HOX gene profiles, only extending organoids displayed regionalized co-linear patterning of HOX genes, and the timing of HOX gene expression differed between extending and non-extending populations. Single-cell RNA sequencing confirmed similar transcriptomic profiles with slightly differing levels of HOX gene expression between extending and non-extending populations, and also revealed that all organoids contain populations of neuronal progenitors and paraxial mesoderm. Significantly, this shows that despite similar gene profiles, morphological variation may occur due to differential regulation at the protein level. Finally, we demonstrated that the axial extension of organoids is influenced by BMP signaling, such that perturbation of endogenous BMP signaling via KD of inhibitors influences relative expression of neuronal markers along the dorsal-ventral axis. Noggin knockdown resulted in elongation that surpassed that of the wildtype control, mirroring Noggin knockout murine models that exhibit failed neural tube closure and elongated tails. Further, knockdown of Chordin expression yielded organoids that extended quickly in multiple directions, failing to create a uni-directional extension axis, which may be reminiscent of Chordin-deficient mice that have a shortened body axis (28). Overall, the organoid model described herein provides a novel approach to study the gene regulatory networks, multicellular organization, and structural outcomes that contribute to the development of the primitive human central nervous system.
Stem cell-derived organoid model systems that reflect multiple organ systems (16, 31-34) and stages of development (18, 35, 36) can now be used to investigate aspects of human embryogenesis that were previously elusive due to low quantity of tissue, technical difficulties isolating and imaging embryonic tissues, and ethical concerns regarding tissue procurement. Organoid platforms are amenable to varying culture conditions, genetically modified cell lines, or small molecule drug screens, enabling the systematic manipulation of experimental conditions in a higher throughput manner than possible in most model organisms. However, while substantial insights can be garnered from organoids, they do not perfectly recapitulate the spatiotemporal dynamics of gene expression and differentiation that occurs within an embryo (37). Additionally, reproducibility both within differentiations, between cell lines, and across labs remains a large hurdle. Therefore, refining organoid culture protocols and developing new differentiation systems to better reflect developmental conditions is crucial for understanding complex morphological processes that transpire uniquely during embryogenesis, such as axial elongation. Furthermore, refining tractable and manipulatable organoids that mimic embryonic morphogenesis will help to elucidate mechanisms that govern form and function, advancing the field of human developmental biology and enabling the discovery of new therapeutic interventions for congenital defects.
Materials and Methods
Human Induced Pluripotent Stem Cell Line Generation and Culture
All work with human induced pluripotent stem cells (iPSCs) was approved by the University of California, San Francisco Human Gamete, Embryo, and Stem Cell Research (GESCR) Committee. Human iPSC lines were derived from the WTC11 line (Coriell Cat. #GM25256), the WTB line (Conklin Lab)(38), and the Allen Institute WTC11-LaminB cell line (AICS-0013 cl.210). All cell lines were karyotyped by Cell Line Genetics and reported to be karyotypically normal. Additionally, all cell lines tested negative for mycoplasma using a MycoAlert Mycoplasma Detection Kit (Lonza).
Human iPSCs were cultured on growth factor reduced Matrigel (Corning Life Sciences) and fed daily with mTeSR™-1 medium (STEMCELL Technologies)(39). Cells were passaged by dissociation with Accutase (STEM CELL Technologies) and re-seeded in mTeSR™-1 medium supplemented with the small molecule Rho-associated coiled-coil kinase (ROCK) inhibitor Y-276932 (10 μM; Selleckchem)(40) at a seeding density of 12,000 cell per cm2.
The generation of the Chordin and Noggin CRISPRi lines first involved TALEN mediated insertion of the CRIPSRi cassette pAAVS1-NDi-CRISPRi (Gen2) (Addgene) to the AAVS1 locus of the Allen Institute WTC11-LaminB cell line. Following antibiotic selection of clones that received the CRIPSRi cassette, CRISPRi gRNAs were generated targeting Noggin and Chordin (Table S2) using the Broad Institute GPP Web Portal and cloned into the gRNA-CKB (Addgene) following the previously described protocol (30). Guide RNA vectors were nucleofected into the LaminB CRISPRi iPSC line using a P3 Primary Cell 96-well Nucleofector™ Kit (Lonza) and the 4D Nucleofector X Unit (Lonza) following manufacturer’s instructions. Nucleofected cells were allowed to recover in mTeSR™-1 medium supplemented with Y-276932 (10 μM) and then underwent antibiotic selection with blasticidin (ThermoFisher Scientific; 10 μg/ml) following the previously published protocol (29, 30). Knockdown efficiency was evaluated by addition of doxycycline to the daily feeding media over the course of 5 days, collection of mRNA, and subsequent quantification of gene expression by qPCR.
Organoid Differentiation
Organoid differentiations were a modified protocol of previously a published spinal cord interneuron differentiation protocol (13). Human iPSCs were seeded at 125000 cells/cm2 in mTeSR™-1 medium supplemented with the small molecule Rho-associated coiled-coil kinase (ROCK) inhibitor Y-276932 (10 μM; Selleckchem) and small molecule GSK inhibitor CHIR99021 (2μM, 4μM, or 6μM; Selleckchem). Two days later, cells were singularized with Accutase (STEMCELL Technologies), counted using a Countess II FL (Life Technologies), and seeded into 800μm X 800μm PDMS microwell inserts in a 24 well plate (∼270 wells/insert)(23). After ∼18 hours, condensed organoids were transferred to rotary culture in 10cm dishes or 6-well plates in mTeSR™-1 medium supplemented with Y-276932 (10 μM; Selleckchem), CHIR99021 (2μM, 4μM, or 6μM; Selleckchem), ALK5 small molecule inhibitor SB431542 (10μM, Selleckchem), and small molecule BMP inhibitor LDN193189 (0.2μM, Selleckchem). Organoids were fed every other day for up to 17 days. Y-276932 was removed from the media at day 3. At day 5 organoids were transferred to Neural Induction Media (DMEM F:12 (Corning), N2 supplement (Life Technologies), L-Glutamine (VWR), 2μg/ml heparin (Sigma Aldrich), non-essential amino acids (Mediatech INC), penicillin-streptomycin (VWR), supplemented with fresh 0.4μg/ml ascorbic acid (Sigma Aldrich) and 10ng/ml brain derived neurotrophin factor (BDNF, R&D Systems)) supplemented with CHIR99021 (2μM, 4μM, or 6μM; Selleckchem), SB431542 (10μM, Selleckchem), and LDN193189 (0.2μM, Selleckchem). From day 7 onwards, organoids were fed with Neural Induction Media supplemented with retinoic acid (10nM, Sigma Aldrich), purmorphamine (300nM, EMD Millipore) and N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT D5942, 1μM, Sigma-Aldrich).
Organoid Elongation Imaging and Quantification
Day 5 organoids from a single 10cm dish were individually transferred using wide bore pipette tips into the center 60 wells of an uncoated ultra-low attachment 96-well plate (Corning), seeding exactly one organoid per well, with the remaining organoids maintained in rotary culture through day 7. Using an inverted Axio Observer Z1 (Zeiss) microscope with incubation (Zeiss Heating Unit XL S, maintained at 37 °C, 5% CO2), all 60 wells were imaged using an AxioCam MRm (Zeiss) digital CMOS camera at 5x magnification (NA 0.16, 2.6 μm × 2.6 μm per pixel). Each well was imaged in TL Brightfield every 20 minutes for 48 hours giving a total of 145 frames. At the end of imaging (day 7), 31 organoids from the parallel rotary culture were imaged at 5x to generate a comparison image set.
To segment the organoids, all well images were first aligned by fitting a truncated quadratic curve to the average image intensity, then solving for the peak of maximum intensity, which was assumed to be the well center. Next, the average lighting inhomogeneity was calculated as the pixel-wise median of all 60 aligned well images, which was then subtracted from the individual aligned frames. After background correction, individual organoids were isolated by finding objects brighter than 0.83% of maximum intensity, but less than 3.0% of maximum intensity, with object size greater than 2,000 pixels, eccentricity greater than 0.1, and solidity greater than 40%. A bounding box 2 mm × 2 mm around the center of each of these objects was calculated and all frames of the time series cropped to this bounding box to reduce memory usage. To detect the region of maximum motion in the time series, the difference image between each pair of sequential images was calculated, and then the pixelwise standard deviation was calculated over all difference images in a given region. This standard deviation image was then thresholded at between 0.01 and 0.03 (AU) depending on the remaining lighting inhomogeneity in the image, producing a ring-shaped mask around the periphery of each organoid. Finally, using the interior of the mask as the organoid seed and the exterior as the background seed for the first frame, organoids were segmented using anisotropic diffusion (41), evolving the foreground and background seeds using the contour calculated from the previous frame for subsequent segmentations. Segmentation, labeling, and metrology were all performed using the python package sckit-image (42).
Segmentations were manually inspected for accuracy, with 45 of 60 determined as having no or only minor flaws, with the remaining 15 excluded from automated analysis. Using the high-quality segmentations only, each organoid time series was then analyzed to examine geometry change over time. For each contour at each time point, we calculated contour area, contour perimeter, minimum, maximum and mean distance from contour center of mass to the perimeter. As additional non-dimensional measures of shape, we calculated the ratio of maximum to minimum radius and organoid circularity. Organoids were also manually classified as “extending”, “partially extending”, or “non-extending” by examining each video. Organoids assigned to “extending” exhibited at least one, and at most two large protuberances that extended at least 100μm from the main body. Partially extending organoids exhibited at least one, and often many protuberances, all of which failed to extend robustly past the 100 μm demarcation. Non-extending organoids were any organoids that failed to generate any extensions over the observation period.
Dissection of Polarized Organoids
Dissections were performed on a SteREO Discovery.V8 Manual Stereo Microscope (ZEISS) and images were taken with an EP50 Microscope Digital Camera (Olympus). Day 9 organoids were transferred into 12 mL of 37°C PBS in a 10 cm dish. 2 mL of warmed PBS were also added to 2 wells of a 6-well plate. Using two fine point forceps, polarized elongated aggregates were immobilized and pulled or pinched apart. Using a p200 pipet with a cut tip, the ‘head’ and ‘tail’ halves of the organoids were collected and transferred to separate wells of the 6-well plate. After ∼25 organoids were dissected, the collected halves were transferred to Eppendorf tubes and suspended in RLT buffer (RNAeasy Mini Kit; QIAGEN) for RNA extraction. Remaining, whole undissected organoids were also collected and suspended in RLT buffer for comparison. Standard RNA extraction was performed following manufacturer specifications.
Real Time Quantitative Polymerase Chain Reaction
Total RNA was isolated from organoid samples using an RNAeasy Mini Kit (QIAGEN) according to manufacturer’s instructions. Subsequently, cDNA was generated using an iScript cDNA Synthesis kit (BIORAD) and the reaction was run on a SimpliAmp thermal cycler (Life Technologies). Quantitative PCR reaction using Fast SYBR Green Master Mix (ThermoFisher Scientific) and run on a StepOnePlus Real-Time PCR system (Applied Biosciences). Relative gene expression was determined by normalizing to the housekeeping gene 18S rRNA, using the comparative threshold (CT) method. Gene expression was displayed as fold change of each sample versus control. The primer sequences were obtained from the Harvard Primer bank or designed using the NCBI Primer-BLAST website (Table S1).
Due to low RNA abundance in the ‘head’ and ‘tail’ dissected organoids, PreAmp Master Mix (FLUIDIGM) was used to amplify cDNA. Briefly, 20 ng of cDNA were amplified per 5 μL reaction volume for 15 cycles on a SimpliAmp thermal cycler (Life Technologies). Amplified cDNA was diluted 5-fold using nuclease free water. 1 μL of amplified diluted cDNA was used for each 20 μL quantitative PCR reaction using Fast SYBR Green Master Mix (ThermoFisher Scientific) and run on a StepOnePlus Real-Time PCR system (Applied Biosciences), following normal qPCR methods described above.
Histology, Immunocytochemistry and Imaging
Organoids were fixed with 4% paraformaldehyde (VWR) for 40 minutes, washed three times with PBS. Organoids to be used for histology were embedded in HistoGel Specimen Processing Gel (Thermo Fisher) prior to paraffin processing. Parafin embedded samples were sectioned in 5μm sections, and subsequently stained for H&E. For immunofluorescent staining, slides were deparaffinized as for H&E staining. Epitope retrieval was performed by submersing slides in Citrate Buffer pH 6.0 (Vector Laboratories) in a 95 °C water bath for 35 minutes. Samples were permeabilized in 0.2% Triton X-100 (Sigma-Aldrich) for 5min, blocked in 1.5% normal donkey serum (Jackson Immunoresearch) for 1 hour, and probed with primary antibodies against SOX2, PAX6, TBXT, NES, TUBB3, and CDH2 (Table S3) overnight at 4°C and secondary antibodies for 1 hour at RT. Nuclei were stained with a 1:10000 dilution of Hoechst 33342 (Thermo Fisher) included with secondary antibodies. Coverslips were mounted with anti-fade mounting medium (ProlongGold, Life Technologies) and samples were imaged on a Zeiss Axio Observer Z1 inverted microscope equipped with a Hamamatsu ORCA-Flash 4.0 camera.
Whole Mount Lightsheet Imaging
4% paraformaldehyde-fixed paraffin-embedded samples (see “Histology, Immunocytochemistry, and Imaging”) were permeabilized with 1.5% Triton X-100 (Sigma-Aldrich) for 1 hour, blocked in 5% normal donkey serum (Jackson Immunoresearch) for 1 hour, and probed with primary and secondary antibodies (TableS3) overnight. Nuclei were stained with a 1:10000 dilution of Hoechst 33342 (Thermo Fisher) included with secondary antibodies. Samples were then embedded in 1.5% low melt agarose (BioReagent) and drawn up into ∼1mm imaging capillaries and subsequently imaged on the Zeiss Z.1 Light sheet Microscope equipped with two PCO.edge SCMOS cameras at 5X and 20x (NA 1.34, aqueous objective).
Single Cell RNA Sequencing Sample and Library Preparation
Multiple organoid samples were combined and processed together using the MULTI-Seq technology (43). Organoids were singularized using Accutase (STEMCELL Technologies) and washed with cold PBS. Cells were resuspended in PBS with lipid-modified Anchor and Barcode oligonucleotides (gift from Zev Gartner, UCSF) and incubated on ice for 5 minutes. A co-Anchor oligo was then added in order to stabilize membrane retention of the barcodes incubated for an additional 5 minutes on ice. Excess lipid-modified oligos were quenched with 1% BSA in PBS, washed with cold 1% BSA solution, and counted using a Countess II FL (Life Technologies). Single cell GEMs and subsequent libraries were then prepared using the 10X Genomics Single Cell V2 protocol with an additional anchor specific primer during cDNA amplification to enrich barcode sequences (43). Short barcode sequences (approx. 65-100bp determined by Bioanalyzer) were purified from cDNA libraries with two sequential SPRI bead cleanups. Barcode library preparation was performed according to the KAPA HiFi Hotstart (Kapa Biosystems) protocol to functionalize with the P5 sequencing adapter and library-specific RPIX barcode. Purified ∼173bp barcode fragments were isolated with another SPRI bead cleanup and validation by Bioanalyzer.
The sample library was sequenced on an Illumina NovaSeq yielding an average of 41,112 reads per cell and 6,444 cells. The MULTI-Seq barcode library was sequenced on an Illumina NextSeq yielding an average of 9,882 reads per barcode and enabling sample assignment for 4,681 of 6,124 unique UMIs detected (76.4% recovery), using the demultiplexing code provided by the MULTI-Seq protocol (43).
Genome Annotation, RNA-seq Read Mapping, and Estimation of Gene and Isoform Expression
The sample library was aligned to the human GRCh38 reference genome using Cell Ranger v1.2.0 (10x Genomics). Gene expression levels were assessed using the Seurat v3.0.0 analysis pipeline (44). First cells were removed with fewer than 200 detected genes, fewer than 1,000 total detected transcripts, or which had greater than 10% mitochondrial gene expression. Next, expression levels were log normalized, and the top 2,000 variable genes calculated using the VST algorithm. The top 20 principal components were used to group cells into 6 clusters using a resolution of 0.2. Finally, top markers were detected for each cluster by detecting the top differentially expressed genes between one cluster and the remaining data set, where at least 25% of cells in the cluster expressed the gene and the gene was expressed at least 0.25 log2 fold-change different from the remaining population. Clusters and gene expression were visualized on a two-dimensional UMAP projection of the first 20 principal components.
Cluster Analysis
To assign cluster identity, the top markers for each cluster were tested for GO term enrichment using the biological process “enrichGO” function in the R package “clusterProfiler” v3.12 (45). In addition, differentiation maturity in each cluster was assessed by examining expression level of panels of early neuroectoderm markers, proliferation markers, markers of neuron fate commitment, and markers of cell types present in neural tube formation and axial extension (22, 46). Finally, to assess anterior-posterior position of each, panels of HOX genes, as well as marker genes of Wnt and FGF signaling were examined to assign rough position of each cluster along the head-tail axis (7, 9, 47).
RNAScope
In situ hybridization for HOXB1, HOXC6, HOXB9 (probe information in Table S4) was performed on sections of 4% paraformaldehyde-fixed paraffin-embedded samples (see “Histology, Immunocytochemistry, and Imaging”) using the RNAscope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics) and following the protocol outlined in User Manual 323100-USM. Sections were imaged on a Zeiss Axio Observer Z1 inverted microscope equipped with a Hamamatsu ORCA-Flash 4.0 camera.
Statistical Analysis
Each experiment was performed with at least three biological replicates. Multiple comparisons were used to compare multiple groups followed by unpaired T-tests between two groups subject to a post-hoc Bonferroni correction. In gene expression analysis, three replicates were used for each condition, and all gene expression was normalized to control wildtype populations. Significance was specified as P-values< 0.05 unless otherwise specified in figure legends. All error bars represent 95% confidence bounds of the data.
Data Availability
The single cell RNA-seq data generated from this study will be made available on the Gene Expression Omnibus upon acceptance. The data analysis code generated in the live imaging analysis studies will be made available via GitHub repository upon acceptance.
Author contributions
ARGL, DAJ, and TCM were responsible for the conceptualization of the project and designing experiments. Differentiations were conducted by ARGL, DAJ, NHE, EAB, and EAG. Single cell RNA sequencing preparation and analysis were performed by ARGL, DAJ, NHE, and MZK. Image analysis script was written and implemented by DAJ and ARGL. Histological sectioning and staining were conducted by DAJ, NHE, EAB, and EAG. Light sheet microscopy and analysis was conducted by ARGL, DAJ, and EAG. Quantitative PCR was done by ARGL, DAJ, NHE, EAB, and MZK. RNAscope sample preparation and analysis was performed by ARGL. Cell lines were generated by ARGL, DAJ, MZK, and NMC. ARGL and DAJ wrote the original manuscript with input from all co-authors. ARGL prepared the figures with input from all co-authors.
Competing Interests
Authors declare no competing interests.
Supplementary Figures
Supplementary Movie Legends
Movie S1: Elongating organoid dynamics. Movie of an example elongating organoid cultured in a 96-well plate, taken over the course of differentiation from day 5 to day 7 with time steps every 20 minutes.
Movie S2: Non-elongating organoid dynamics. Movie of an example non-elongating organoid cultured in a 96-well plate, taken over the course of differentiation from day 5 to day 7 with time steps every 20 minutes.
Movie S3: TBXT and SOX2 location in non-elongating organoid. Movie of a 3D rendered light-sheet image of an example non-elongating organoid stained for TBXT (yellow) and SOX2 (red).
Movie S4: TBXT and SOX2 location in elongating organoid. Movie of a 3D rendered light-sheet image of an example elongating organoid stained for TBXT (yellow) and SOX2 (red).
Movie S5: COL IV and ZO1 location in non-elongating organoid. Movie of a 3D rendered light-sheet image of an example non-elongating organoid stained for COL IV (yellow), ZO1 (red), and DAPI (blue).
Movie S6: COL IV and ZO1 location in elongating organoid. Movie of a 3D rendered light-sheet image of an example elongating organoid stained for COL IV (yellow), ZO1 (red), and DAPI (blue).
Movie S7: β III tubulin and Nestin location in non-elongating organoid. Movie of a 3D rendered light-sheet image of an example non-elongating organoid stained for β III tubulin (TUBB; yellow) and Nestin (red).
Movie S8: β III tubulin and Nestin location in elongating organoid. Movie of a 3D rendered light-sheet image of an example elongating organoid stained for β III tubulin (TUBB; yellow) and Nestin (red).
Supplementary Tables
Acknowledgements
We would like to thank the Gladstone Light Microscopy and Histology Core, the Gladstone Stem Cell Core, and the Gladstone Graphics Team for their support and experimental expertise.