Summary
The mechanisms leading to adrenal cortex development and steroid synthesis in humans remain poorly understood due to the paucity of model systems. Herein, we recapitulate human fetal adrenal cortex specification processes through stepwise induction of human induced pluripotent stem cells through posterior intermediate mesoderm-like and adrenocortical progenitor-like states to ultimately generate fetal zone adrenal cortex-like cells (FZLCs), as evidenced by histomorphological, ultrastructural, and transcriptome features and adrenocorticotropic hormone (ACTH)-independent Δ5 steroid biosynthesis. Furthermore, FZLC generation is promoted by SHH and inhibited by NOTCH, ACTIVIN and WNT signaling, and steroid synthesis is amplified by ACTH/PKA signaling and blocked by inhibitors of Δ5 steroid synthesis enzymes. Finally, NR5A1 promotes FZLC survival and steroidogenesis. Together, these findings provide a framework for understanding and reconstituting human adrenocortical development in vitro, paving the way for cell-based therapies of adrenal insufficiency.
Introduction
The fetal adrenal cortex facilitates maturation of various organ systems and is essential for proper functioning of the fetoplacental unit involved in pregnancy maintenance 1,2. In addition, successful development of the fetal adrenals is required for establishment of a functional adrenal cortex postnatally1,3. Accordingly, genetic defects affecting fetal adrenal development (e.g., NR5A1 mutations) or steroid biosynthesis result in primary adrenal insufficiency (PAI), an underdiagnosed but potentially life-threatening disorder necessitating life-long hormone replacement therapy4. As there are currently limited therapeutic or preventative measures for PAI, understanding the genetic and signaling requirements for normal fetal adrenal development and initiation of steroidogenesis is of paramount significance for accurate PAI diagnosis and development of targeted therapeutic interventions.
Our current understanding of human fetal adrenal development has been hampered by marked species-specific differences in development that limit the translational potential of rodent models. For example, adrenal androgens (e.g., DHEA, DHEA-S), the most abundant steroids produced by human fetal adrenals, are only minimally synthesized in rodent fetal adrenals 2,5,6. Because ex vivo culture of human fetal adrenal tissues/cells also suffers from sample-to-sample variability and bioethical constraints7–11, there is a critical need to develop suitable models to understand the genetic and epigenetic mechanisms regulating human adrenocortical development and adrenal steroidogenesis.
In mice, the adrenal cortex originates in the adrenogonadal primordium, a common progenitor for both the adrenocortical and gonadal lineages12. In contrast, organogenesis of the human adrenal cortex starts with fate specification of the NR5A1+GATA4− adrenogenic coelomic epithelium (AdCE) within the WT1+ coelomic epithelium at 3–4 week post fertilization (wpf), which is spatially and temporally distinct from gonadogenesis12. These cells subsequently undergo dorsomedial migration and form a condensed blastematous structure, referred to as the adrenal primordium (AP), which is followed by establishment of two distinct zonal structures, the peripherally located definitive zone (DZ) with putative stem cell/progenitor potential, and the centrally located fetal zone (FZ) with steroidogenic potential by 8 wpf12–14. Our single cell RNA-seq analysis of human fetal adrenals has recently identified the cellular and transcriptional dynamics accompanying specification of adrenocortical lineages12. However, the underlying gene regulatory mechanisms governing human adrenocortical development and concomitant steroidogenesis remain unknown.
To overcome limitations in rodent modes, and establish an alternative to the use of human fetal tissue, we identified a strategy in which human adrenocortical lineages could be induced from human induced pluripotent stem cells (hiPSCs). Using this strategy in multiple hiPSC lines, we observed a robust induction of steroidogenic fetal adrenocortical cells, which we utilized to interrogate human adrenocortical development and steroidogenic function with pharmacologic, genetic and epigenetic approaches.
Results
Induction of early adrenocortical lineage through posterior intermediate mesoderm-like cells derived from human iPSCs
Based on our scRNA-seq analyses of human adrenocortical development, we asked if fetal zone adrenal cortex-like cells (FZLCs) could be established from human iPSCs through stepwise induction. To visualize this process, we generated human iPSCs bearing WT1-p2A-EGFP (WG) (turned on as cells transition to the posterior intermediate mesoderm [PIM]) and NR5A1-p2A-tdTomato (NT) (turned on as cells transition into adrenocortical progenitors) using a CRISPR/Cas9-guided knock-in approach. One male line had both WG and NT (WGNT SV20-211, herein designated as 211) whereas two female lines possessed NT only (NT 312-2121 and NT 1390G3-2125) (Figures S1A–E). As both kidneys and adrenal cortex originate from the PIM15, we exploited a previously established 3D induction platform to derive the metanephric lineage by activating Wnt signaling through high dose CHIR99021 (glycogen synthase kinase 3 inhibitor) for 7 days16. This first step of induction generated transient T+ nascent mesoderm (Figure S2A). Subsequently, at floating culture day (fl)9, retinoic acid-based induction resulted in generation of PIM-like cells bearing high WT1 but low FOXF1 expression; the latter representing a marker of the lateral plate mesoderm (Figures S2A, B)12,15. Consistent with the previous study, further culture of aggregates until fl12 with low CHIR99021 and high FGF9, a potent nephrogenic inducer16,17, resulted in SIX2-, PAX2- and WG-expressing metanephric mesenchyme (Figures S2A, B). As this condition did not activate NR5A1, a key marker of adrenocortical fate (Figures S2A, B)12, we modified conditions to redirect the lineage towards the adrenal fate, using NT fluorescence to monitor our success. Notably, optimization of the initial BMP4 concentration and switching from the original DMEM-based medium supplemented with B27 (DB27 medium) to α-MEM-based medium supplemented with 10% Knockout Serum Replacement (KSR) (MK10 medium) not only increased the induction efficiency of WG, but downregulated metanephric mesenchyme markers (i.e., SIX2, PAX2) (Figures S2C, D). Moreover, while NR5A1 remained undetectable, these conditions facilitated OSR2 expression, one of the earliest markers of adrenocortical fate (Figure S2D)12,15, suggesting that these conditions might be permissive for adrenal fate specification. As CHIR99021, FGF9 and Activin A (homodimers of INHBA) reportedly enhance nephrogenic fate16,18 and FGF9 enhances gonadogenic fate as well19, we next determined if their inhibition promoted adrenal specification. Although antagonism of canonical Wnt and FGF9 by IWR1 and SU-5402, respectively, slightly upregulated NT expression (Figure S2E), addition of SB431542 (NODAL/Activin inhibitor) (fl7-22) together with low dose BMP4 (fl10-15) markedly upregulated NT expression at fl14 and 22 (Figure 1A). Next, as we have previously demonstrated that human fetal adrenal cortex strongly expresses DLK1, a membrane bound protein that negatively regulates NOTCH signaling12,15 and as Sonic Hedgehog (SHH) is critical for adrenocortical development and SHH target genes are highly expressed in human fetal adrenal cortex12,20–22, we assessed the effect of DAPT (NOTCH inhibitor) and SHH on NT expression. Notably, DAPT and SHH in combination further enhanced both NT and steroidogenic gene expression at fl22 (Figures 1A, S2F–J). Notably, this combination also suppressed expression of WT1 and the podocyte marker NPHS2, and lowered the frequency of WGbright+ cells, an apparent off-target nephrogenic population (Figures S2H, J). As this condition (SHH: fl7-22; DAPT: fl10-22) combined with early addition of SB431542 (fl5-22) outperformed that of late addition of SB431542 (fl7-22) (Figure S3A), we employed this new modification for subsequent cultures (SHH: fl7-22; DAPT: fl10-22; SB431542: fl5-22). Based on our success, we reevaluated the role of Wnt signaling at fl10-22 in NT induction and confirmed that IWR1 outperformed CHIR99021 or no-treatment, further suggesting that Wnt activation (fl1-10) followed by inhibition (fl10-22) supports the induction of NT+ cells (Figure S3B). Of note, although the culture condition from fl15 onwards essentially replicated the preceding condition from fl10 (BISSSD medium), overall concentrations of growth factors/inhibitors were reduced and BMP4 was omitted (ISSSD medium). This, however, did not affect NT induction efficiency (Figure S3C) and in this final induction scheme, we confirmed that the initial dose of BMP4 at fl0-1 affects overall induction efficiency at fl22 (Figure S3D).
Figure 1. Induction of early adrenocortical lineage through the posterior intermediate mesoderm (PIM)-like cells from human iPSCs.
(A) Schematic showing key factors for induction of NT+ cells in WGNT SV20-211 (211) hiPSCs (top). Cells were treated with 10 ng/ml ACTIVIN A (fl7-10) or 30 μM SB (in the presence or absence of 50 ng/ml SHH [fl7-22]/ 10 μM DAPT [fl10-22]). Factors depicted above the timeline (fl7-22) and in the scheme in (B) (fl0-7) were used for all culture conditions. CHIR, CHIR99021; RA, retinoic acid; SB, SB-431542; SHH, sonic hedgehog; SU, SU-5402; Y, Y-27632. Fluorescence-activated cell sorting (FACS) analysis at floating culture day 14 (fl14) and fl22 of the indicated culture condition (bottom).
(B) Schematic depicting the finalized induction scheme (top). Bright-field (BF) and fluorescence images of aggregates for WT1-EGFP (WG, green) and NR5A1-tdTomato (NT, red) at indicated stages (bottom). Bar, 500 μm.
(C) FACS analysis of aggregates at indicated stages.
(D) qPCR analysis of expression of key genes in cDNA generated from bulk aggregates derived from 211 hiPSCs at the indicated time point during PIM induction. For each point, the average value with standard deviation is shown. ND, not detected. *, p < 0.05; **, p < 0.01; ***, p < 0.001 vs first detected time.
(E) IF images of the fl10 aggregate for WT1-EGFP (green), WT1 (red) and DAPI (white) with their merge. Bar, 20 μm.
(F) Bright field (BF) and fluorescence (NT, red) images of fl9, fl15 and fl22 aggregates. Bar, 500 μm.
(G) FACS analysis of WGNT expression in (F).
(H) Percentage of NT+ cells, mean fluorescence intensity of NT, and NT+ cell number per aggregate during floating culture as assessed by FACS. Cells were derived from 211 hiPSCs. *, p < 0.05; **, p < 0.01; ***, p < 0.001 vs fl3.
(I) NT+ cell number per aggregate at fl21 in three different batches. The averages are shown (horizontal line).
(J) qPCR analysis of expression of key genes. Cells were derived from 211 hiPSCs.
(K) IF images of fl22 aggregates for CYP11A1, CYP17A1, or SULT2A1 (green), stained with NR5A1 (red) and DAPI (white). Merged images are shown on the right. Bar, 10 μm.
Similar to the original conditions, this new multi-step induction first established WT1+ PIM-like cells (PIMLCs) at fl9-10, in which PIM markers were upregulated while pluripotency associated markers were downregulated (Figures 1B–E). Thereafter, aggregates progressively upregulated NT and adrenocortical markers, NR5A1 and NR0B1 from fl12 onwards while gradually decreasing WT1 from fl15 onwards, consistent with their commitment to early adrenocortical lineages (Figures 1F–J)12. Importantly, neither gonadal markers (GATA4, LHX9) nor kidney markers (PAX2, SIX2) were detectable by fl21, suggesting that cells were directed towards the adrenal cortex but not developmentally related lineages (Figure 1J). We also noted that cells progressively acquired steroidogenic gene expression including STAR, CYP11A1, which was followed by lower level upregulation of CYP17A1 and SULT2A1 after fl18 (Figure 1J). Accordingly, IF identified diffuse expression of CYP11A1 in NR5A1+ cells, weak expression of CYP17A1 in a few scattered cells and undetectable SULT2A1, suggesting that cells had not yet acquired the full steroidogenic activity seen in fetal adrenal cortex in vivo (Figure 1K).
Similar to in vivo human fetal adrenal cortex, we found that DLK1 is highly expressed in NT+ but rarely in NT− cells (Figures S3E, F). Notably, this allowed us to FACS-sort adrenocortical lineages induced from a hiPSC line that did not express NT (Penn067i-312-1, herein designated as 312). Sorted DLK1bright+ cells from 312 hiPSCs subjected to our induction protocol expressed NR5A1, NR0B1, STAR and CYP17A1 at a level equivalent to or slightly higher than sorted NT+ cells from NT 312-2121 hiPSCs, supporting the utility of DLK1 as a cell surface marker of early adrenocortical lineages (Figure S3G).
Induction of fetal-zone adrenocortical-like cells (FZLCs) through air-liquid interface culture
As culturing aggregates beyond fl22 resulted in increased central necrosis, likely due to limited oxygen and/or nutrient diffusion (Figure S4A), we next transferred them onto collagen-type I-coated PET membranes for air-liquid interface (ALI) culture (Figures 2A–C, S4B). Notably, use of ISSD medium (low dose IWR1, SU, SHH, DAPT) tended to outperform basal MK15 medium for promoting survival of NT+ cells during ALI culture (Figures S4C, D). Moreover, ALI culture resulted in progressive upregulation of genes encoding proteins involved in steroidogenesis (i.e., CYP11A1, CYP17A1, SULT2A1, CYP11B1), cofactors (e.g., STAR, POR, FDX1, FDXR, PAPSS2) and other genes related to adrenal steroidogenesis while maintaining NT and NR5A1 expression (Figures 2B, C, S4E–H). Transcription factors such as NR0B1, OSR2 and GATA6 were expressed throughout the ALI culture. Metanephric and gonadal genes were not expressed, suggesting that the culture conditions did not support these cell types (Figures S4G, H). H&E sections of aggregates at ali21-28 identified tightly packed nests of cells with abundant eosinophilic and granular cytoplasm and occasional cytoplasmic vacuoles, and peripherally located round nuclei, highly reminiscent of the FZ (Figures 2E, S4I). IF also identified strong cytoplasmic expression of CYP11A1, CYP17A1, SULT2A1, SOAT1, FDX1, and FDXR, similar to the FZ (Figures 2F–H, S4J)12. ISH further confirmed the expression of NR5A1, NR0B1, STAR, APOA1 (Figure S4K). The characteristic cellular morphology and immunophenotype was maintained until at least ali62 (Figure S4L). During ALI culture, aggregates increased production of IGF-II, a key growth factor of the fetal adrenal cortex (Figure 2I)23. Ultrastructurally, these cells had abundant mitochondria that exhibited a leaf-like contour with tubular cristae and intervening prominent smooth endoplasmic reticulum. There were also frequent round osmiophilic lipid granules of various electron density (Figure 2J). Cells with similar immunophenotypic features were successfully induced from various human iPSC cell lines with or without fluorescence reporters (NT 312-2121, Penn067i-312-1) (Figures S4M, N). As gene expression, histologic, immunophenotypic, and ultrastructural features were all consistent with the FZ of the prenatal human adrenal cortex, we designated these NT+ cells as fetal zone adrenal cortex-like cells (FZLCs).
Figure 2. Induction of fetal-zone adrenocortical-like cells (FZLCs) through air-liquid interface culture.
(A) Schematic illustration of the air-liquid interface (ALI) culture and medium components added.
(B) BF and fluorescence (NT) images of aggregates at indicated time points. Bar, 500 μm.
(C) FACS plots of aggregates for WG and NT expression at indicated time points.
(D) Percentage of NT+ cells, the mean fluorescence intensity of NT during ALI culture of 211 iPSC-derived aggregates as assessed by FACS. *, p < 0.05; **, p < 0.01, vs ali0.
(E) H&E images of the fetal zone (FZ)-like Cells (FZLCs) at ali28 (left) and FZ from a human fetal adrenal gland at 8 wpf (right). Bar, 20 μm.
(F) Low magnification IF images of aggregates at ali28 for CYP11A1, CYP17A1, or SULT2A1 (green), and their merges with NR5A1 (red) and DAPI (white). Bar, 50 μm.
(G) High magnification IF images of aggregates at ali28 (left) and FZ at 8 wpf for CYP11A1, CYP17A1, or SULT2A1 (green), stained with NR5A1 (red) and DAPI (white) and the merged images. Bar, 20 μm.
(H) IF quantification of CYP11A1 (left) and CYP17A1 (right) positive cells among NR5A1+ cells in aggregates at indicated stages.
(I) Insulin like growth factor II (IGF-II) production during induction of FZLCs derived from 211 hiPSCs.
(J) Transmission electron microscope images of the Fetal Zone like cells (FZLCs) in Ali28. Bar, 3 μm.
Robust Δ5 adrenal steroid biosynthesis from FZLCs
Given the marked upregulation of steroidogenic enzymes in FZLCs, we next measured adrenal steroid production, a key function of the adrenal cortex (Figure 3A). Consistent with the expression of proteins involved in Δ5 steroid synthesis (e.g., CYP11A1, STAR, CYP17A1 and SULT2A1) (Figures 2F, G, S4J, L–N), culture supernatant of ali21-28 aggregates contained substantial amounts of Δ5 steroids, dehydroepiandrosterone [DHEA] and its sulfated form [DHEA-S] (Figures 3B, C). In contrast, levels of most Δ4 steroids were low/negligible (e.g., progesterone [prog], 17-OH progesterone [17OHProg], cortisol, aldosterone), except androstenedione (A4), which was produced in moderate levels likely through the action of HSD3B2, which was expressed at low but detectable levels, on DHEA (Figures 3C, S4G). These findings are consistent with the steroid synthesis profile of the human fetal adrenal cortex, which predominantly produces DHEA and DHEA-S1. Finally, consistent with the sequential expression of CYP11A1, CYP17A1 and SULT2A1 (Figures 1J, S4G), we first detected pregnenolone production at fl21 (Figure 3D). Pregnenolone gradually decreased from ali9 onwards, DHEA production peaked at ali9, and DHEA-S was first detected at ali9 and progressively increased until ali21 (Figure 3D).
Figure 3. Robust Δ5 adrenal steroid biosynthesis by FZLCs.
(A) Schematic illustration of adrenal steroid synthesis. See METHODS for abbreviations of steroids.
(B) Validation of ELISA assays by spike and recovery and precision measurements using culture medium of al21 aggregates derived from 211 hiPSCs.
(C) Production of Δ5 and Δ4 steroids in culture media of ali21 aggregates. Fresh ISSD medium was used as a negative control. ELISA measuredΔ5 steroids LC-MS/MS measured Δ4 steroids. Means ± standard deviation (n = 2–6). P-values for the comparison is determined by two-tailed Welch’s t-test (*, p < 0.05; ***, p < 0.001).
(D) Production of Preg, DHEA, and DHEA-S during FZLC induction from 211 hiPSCs. Culture medium was collected at 72 hours and concentrations determined by ELISA. Means ± standard deviation (n = 3). P-values for the comparison were determined by one-way ANOVA for repeated measures followed by Dunnett’s multiple comparison test (*, p < 0.05; **, p < 0.01; ***, p < 0.001 vs fl10).
(E-G) Effects of trophic stimulants (ACTH, forskolin, and 8Br-cAMP) (E), aminoglutethimide (AG) (F), and abiraterone or orteronel (G) on the production of Preg, DHEA, and DHEA-S. Means ± standard deviation (n = 3). Values were normalized with cell numbers. LDL, low density lipoprotein. *, p < 0.05; **, p < 0.01; ***, p < 0.001 vs control (E) and vs LDL + forskolin 1 μM (F, G).
(H) BF and fluorescence images (top) and FACS plots (bottom) of aggregates at ali15 with or without cryopreservation. Aggregates were dissociated at fl21 and cryopreserved before being thawed and reaggregated using 60,000 cells. Bar, 1mm.
(I) IF images of (H) for NR5A1 (red), CYP11A1, and CYP17A1 (green) merged with DAPI (white). Representative images of sections taken from the same aggregate for CYP11A1 (top) or CYP17A1 (bottom) are shown. Bar, 20 μm.
(J) NT+ cell number (per aggregate) after cryopreservation at ali15 (top) and DHEA-S production (bottom). Values were normalized with 103 NT+ cell numbers (bottom). Means ± standard deviation (n = 3). Statistical significance of the differences were evaluated by two-tailed Welch’s t-test.
(K) FACS plot of aggregates at fl22 (left), from which NT+ cells were FACS-sorted and reaggregated for further ALI culture. Bright-field (BF) and fluorescence (NT) images of the ali28 (right). Bar, 500 μm.
(L) IF images of (K, right) for CYP11A1/CYP17A1 (green) and NR5A1 (red) merged with DAPI (white). Representative images of sections taken from the same aggregate for CYP11A1 (left) or CYP17A1 (right) are shown. Bar, 20 μm.
(M) DHEA and DHEA-S production by non-sorted and NT+ sorted cells at ali28 as in (K).
ACTH is a polypeptide hormone secreted by the anterior pituitary gland that plays a central role in the hypothalamic-pituitary-adrenal (HPA) axis by promoting cell differentiation and steroidogenesis in the adrenal cortex through PKA/cAMP signaling pathways2. Accordingly, we found that both ACTH and the synthetic PKA/cAMP agonists, forskolin and 8Br-cAMP enhanced the production of pregnenolone, DHEA and DHEA-S, suggesting that like FZ, FZLCs are responsive to ACTH (Figure 3E).
Given that FZLCs recapitulated Δ5 steroid biosynthesis, we next exploited our novel platform to functionally interrogate the adrenal androgen biosynthesis pathway using pharmacologic inhibitors. CYP11A1 is required to generate pregnenolone from cholesterol, the first step of steroid biosynthesis. Accordingly, aminoglutethimide (AG), an inhibitor of CYP11A1, suppressed the production of pregnenolone, DHEA and DHEA-S (Figure 3F). Likewise, CYP17A1 catalyzes both 17α-hydroxylase activity and 17,20-lyase activity to generate DHEA from pregnenolone. Use of abiraterone, which inhibits both CYP17A1-mediated activities, and orteronel, a a more selective inhibitor of 17,20-lyase activity24, dose-dependently suppressed DHEA and DHEA-S synthesis (Figure 3G). In contrast, pregnenolone levels increased upon treatment with abiraterone, but not with orteronel, suggesting that selective inhibition of 17,20-lyase may promote accumulation of 17α-hydroxypregnenolone (Figure 3G). Altogether, these data suggest that adrenal androgen biosynthesis occurs through the conventional pathway in FZLCs and that our platform can serve as tractable tool for screening inhibitors of this pathway.
Freeze-thawing, fractionation of aggregates before ALI culture
Given the protracted time required to obtain FZLCs (total of ~43 days), we next determined if cell aggregates at fl22 could be dissociated and cryopreserved in liquid nitrogen before starting ALI culture, thus facilitating subsequent studies. After thawing, cells were reaggregated for 1 day in ISSD medium, then transferred to ALI culture for a subsequent 21 days. Notably, these aggregates maintained strong NT and key steroidogenic enzyme expression (i.e., CYP11A1, CYP17A1 and SULT2A1) (Figures 3H–J) and robust production of DHEA-S albeit at slightly lower levels than those that were not frozen (not significant, p = 0.059), supporting use of cryopreserved cells for future studies.
Finally, to further optimize our culture conditions, we next determined if off-target cell generation during ALI culture could be minimized by sorting NT+ cells prior to culture initiation. For this, sorted NT+ cells from fl22 aggregates were allowed to reaggregate in 96 well low-attachment plates and ALI culture for 21 days. These aggregates maintained bright NT fluorescence, expressed steroidogenic enzymes and exhibited robust steroid synthesis, suggesting that NT+ cells can differentiate into Δ5 adrenal androgen-producing FZLCs, even in the absence of NT− cells (Figures 3K–M).
Gene expression dynamics during human adrenal specification in vitro
To capture the global transcriptional dynamics of the FZLC induction process, we performed bulk RNA-seq analyses (Figures 4A–F, S5A, Table S1, 2). To focus on adrenocortical lineages, we FACS-sorted WG+ cells from fl9 aggregates and NT+ cells from fl14 onwards. Unsupervised hierarchical clustering (UHC) segregated the cells into two large clusters, one with hiPSCs, fl9 and fl14 aggregates and the other with fl22 and ali7-28 aggregates (Figure 4A). Consistently, principal component analysis (PCA) showed a progressive transition of transcriptional properties during FZLC induction (Figure 4B). Concordant with qPCR and IF studies, hiPSCs-to-fl9 transition was characterized by upregulation of PIM markers (e.g., WT1, OSR1 and EYA1) and HOX genes (e.g., HOXA1, HOXB2) and accordingly, enriched with GO terms such as “multicellular organism development” (Figure 4C, Table S1). In keeping with NT activation, adrenal specification (e.g., NR5A1, NR0B1, GATA6, RUNX1T1) transcription factors were upregulated in fl14 aggregates12,15. This was followed by progressive upregulation of a number of genes associated with adrenal steroid biosynthesis from fl22 onwards, with enriched GO terms such as “cholesterol metabolic process” (fl14-to-fl22 comparison) or “regulation of lipid metabolic process” (fl22-to-ali21-28 comparison) (Figure 4C, Table S1). Notably, expression of PIM markers was maintained until fl14, then progressively downregulated (Figures 4C, D). Accordingly, fl14 aggregates clustered with fl9 by UHC (Figure 4A), with DEGs enriched in GO terms such as “pattern specification process” (Figure S5B, Table S2) and co-expressed adrenocortical and PIM/coelomic epithelium markers (e.g., WT1, BMP4, KRT18, KRT19), similar to AdCE in human embryos (Figures 4C, D)12. Likewise, fl22 aggregates highly expressed AP markers and to a lesser extent, AdCE markers (Figure 4D). Accordingly, fl9-to-fl22 changes highly resembled transcriptomic dynamics of PIM/early coelomic epithelium (ECE)-to-AP transition in cynomolgus monkey embryos (Figures S5C, D, Table S3)15. Finally, to further validate our model, we compared the transcriptomes of in vitro derived cells with that of in vivo using cDNA generated from whole fetal adrenal cortex at 9 wpf (Ad) and FACS-sorted MME− adrenocortical cells enriched for FZ at 17–19 wpf (Figure S5E)12. Using the same sequencing platform, transcriptomes of FZLCs during ALI culture (ali7-28) juxtaposed to and partially overlapped with those of Ad and FZ in PCA coordinate (Figure 4E), with high correlation coefficients (Figure 4F). FZLCs bore key FZ but not gonadal markers previously defined by single cell RNA-seq of human fetal samples (Figures 4D, G, S5F). Altogether, these findings suggest that aggregates progressively acquire FZLC state through developmental processes similar to that of human adrenocortical lineage in vivo.
Figure 4. Transcriptome dynamics during FZLC induction.
(A, B) Unsupervised hierarchical clustering (UHC) (A) and principal component analysis (PCA) (B) of bulk transcriptomes during FZLC induction. Cells derived from 211, NT 312-2121 and NT 1390G3-2125 hiPSCs at indicated time points were used.
(C) Scatter plots comparing the averaged gene expression values between samples at the indicated time points. Differentially expressed genes (DEGs) (four-fold differences, p <0.05, FDR <0.05) are highlighted in colors. Representative genes and their GO enrichments for DEGs are shown on the right.
(D) Expression of AdCE, AP and FZ marker genes identified as DEGs in previous transcriptome analyses of human fetal samples12 in indicated samples.
(E) PCA of transcriptomes used in (B) projected together with human fetal heart, MME− fetal zone adrenocortical cortex (FZ), and whole adrenal glands (Ad).
(F) Heatmap showing Pearson correlation. Note the high correlation of NT+ cells during ali7-28 with FZ or Ad in vivo.
(G) Expression of markers for fetal Leydig cells (138 genes), FZ (69 genes) or Sertoli cells (225 genes) in indicated samples. These marker genes were identified as DEGs in Figure S5F. Bulk transcriptomes at ali21 and ali28 are shown at the top. Pseudobulk transcriptomes for the FZ and fetal Leydig cells in vivo were retrieved from the previous study and shown at the bottom12.
scRNA-seq identified lineage trajectories of on-target and off-target cell types
To precisely understand the lineage trajectory and transcriptomic dynamics of various cell types derived in vitro, we performed single cell (sc)RNA-seq on five samples, of which three were obtained from floating culture (fl19_1, fl19_2, fl22) and two from ALI culture (ali21, ali28). Samples were FACS-sorted for NT+ cells except for fl19_2 and ali21, which utilized the WG+ fraction or whole live cells, respectively. After QC validation and filtering out low quality cells, 23570 cells remained for downstream analysis (Figure S6A). Transcriptomes of cells in all samples were aggregated and projected onto a UMAP plot after dimension reduction and clustering, which yielded 13 clusters (Figure S6B). All clusters except cluster 13 were successfully annotated based on marker gene expression and differentially expressed genes (Figures 5A–D, S6C–E, Table S4). As expected, we identified NR5A1+CYP11A1+CYP17A1+STAR+ FZLCs (clusters 3 [AP/early fetal zone like cells] and 6 [fetal zone like cells], consisting primarily of ali21, ali28 aggregates) and their putative progenitor cell types including WT1+SFRP2+SHISHA2+ PIM/CE-like cells (cluster 5) and NR5A1+OSR2+DLK1+KRT19+ AdCE-like (clusters 1 and 8, consisting primarily of fl19, 22 aggregates), which aligned along the pseudotime trajectory and actual sample stages (Figures 5A–D). Along this trajectory, developmental genes or genes related to cell cycle were downregulated, suggestive of terminal differentiation into the FZ adrenocortical lineage (Figure 5E). In contrast, genes related to steroidogenesis and angiogenesis were upregulated along the trajectory (Figure 5E)1. Notably, genes related to epithelia-to-mesenchymal transition are transiently upregulated along the trajectory (Figure 5F), similar to in vivo human fetal adrenocortical cells at specification stage12. Interestingly, previously identified definitive zone markers (NOV, HOPX) were transiently upregulated in the trajectory although cells expressing these transcripts did not form independent clusters (Figure S6F). In keeping with this observation, IF analyses identified a few scattered cells that bore DZ markers (NOV, MME) but lower levels of a FZ marker (FDX1), which tended to localize at the periphery of FZLCs. This finding suggests that our platform generated both FZLCs and a small fraction of cells resembling DZ-like cells (Figures S6G–K).
Figure 5. Dynamics of single cell transcriptomes during FZLC induction.
(A-C) Trajectory analysis of single-cell transcriptomes during FZLC induction showing segregation of three distinct lineages. Cells were projected to PHATE embedding. Cells were colored according to cell clusters (1: AdCE-like cells, 2: Transition Cells, 3: Fetal zone-like cells, 4: Capsule-like cells, 5: PIM/CE-like cells, 6: Fetal Zone-like cells, 7: Proliferative CE-like cells, 8: AdCE-like cells, 9: Nephron progenitor-like cells, 10: Podocyte-like cells). Sample origin (B) and pseudotime (C) were also projected to the same embedding.
(D) Expression of key marker genes projected on PHATE embedding.
(E) Heatmap of gene expression along pseudotime trajectory for PIM/CELCs to FZLCs. Genes with similar expression patterns are clustered together, and enriched gene ontology terms are listed alongside the heatmap. The top 2000 variable genes were selected and clustered by UHC. PIM/CELCs, PIM/CE-like cells; CapLCs, capsule-like cells.
(F) Expression dynamics of key transcription factors associated with EMT, aligned along pseudo-time as in (E).
(G) Heatmap of gene expression along pseudotime trajectory for PIM/CELCs to CapLCs as in (E).
Unexpectedly, we also identified minor off-target cell types including CLDN5+KDR+ endothelial-like cells and ASCL1+GATA3+PHOX2B+TH+DBH+ sympathoadrenal-like cells of uncertain origin (Figures S6B, E)12. Notably, trajectory analyses also identified two additional lineages that appear to project from PIM/CE-like cells, thus sharing their origin with adrenocortical lineages. One of them was the nephrogenic lineage (SIX2+PAX2+ nephron progenitor-like cells [cluster 9] and NPHS2+PTPRO+ podocyte-like cells [cluster 10]) (Figures 5A, D, S6B, E), and the other cluster was RSPO3+PDGFRA+ capsule-like cells (CapLCs) (cluster 4), which progressively acquired the characteristic gene expression signature of adrenal capsule cells through a transient proliferative CE-like state (cluster 7) (Figures 5A, G, S6B, E, S7A, B). Accordingly, CapLCs showed marked transcriptional similarities with the adrenal capsule cells (Cap) (Figures 5D, S7B–E)12. UHC also identified that CapLCs bore some similarities to PIM/CE/AdCE clusters, expressing shared markers such as RSPO3 or SFRP2 (Figures 5D, S7B–E). Notably, RSPO3+PDGFRA+ CapLCs localized at the periphery of the clusters of FZLCs, reminiscent of the capsule in fetal adrenals, although CapLCs were not packed as densely as that of the fetal adrenals (Figure S7F). Together, our data suggest that our inductive scheme generates FZLCs through an AdCE-like state and also gives rise to cells with transcriptomic features characteristic of adrenal capsule cells.
Single cell accessible chromatin landscape of the specifying adrenocortical lineages
NT+ cells in fl22 aggregates transcriptionally resembled AP/AdCE of the nascent adrenocortical lineage (Figures 4D, S5C, D). These cells emerged from WG+ PIM-like cells observed in fl9 aggregates (Figures 1E, G, 4B), recapitulating the adrenocortical lineage specification process. To identify cis-regulatory elements, enhancer regions, and co-accessibility changes during adrenocortical lineage specification, we isolated WG+ and NT+ cells from fl09 and fl22 aggregates, respectively, and performed single cell Assay for Transposase-Accessible Chromatin sequencing (scATAC-seq) (Figure S7G). Overall, we obtained 130,598 unique peaks in fl09 WG+ cells, 79,472 unique peaks in fl22 NT+ cells, and 120,090 peaks shared between both samples. The majority of peaks were identified in regions characterized as distal elements or introns, suggestive of enhancer elements (Figure 6A). However, stage-specific peaks showing variability in chromatin accessibility between fl09 and fl22 were enriched in different motifs, with peaks specific for fl09 aggregates showing motif enrichment related to PIM development, such as SIX1, PAX8, LHX1, LEF1, HOXA1 and CDX2 (Figure 6B). In contrast, binding motifs for transcription factors potentially involved in adrenocortical development, such as NR5A1, NR2F2, NR2F1, and PBX2, were enriched in peaks uniquely identified in fl22 aggregates. The overlapping peaks, which might represent housekeeping cis-regulatory elements, were enriched with binding motifs for CTCF, a constitutive transcription factor binding to enhancers and other transcription factors (Figure 6B). To determine if transcription factors might be involved in adrenocortical specification, we integrated our motif enrichment analyses and transcriptome data by first evaluating the dynamics of chromatin accessibility in single cells using latent semantic indexing (LSI) for clustering and PHATE for dimensional reduction (Figures 6C–E). Open chromatin accessibility was heterogeneous in WG+ PIM-like cells at fl09 as reflected by the increased number of cell clusters (clusters 2–8), whereas chromatin accessibility was more homogenous at fl22 (clusters 1, 9, 10) (Figures 6C–E). Cellular trajectory and pseudotime inference obtained using stage-specific open chromatin status were concordant with that of the single cell transcriptomes (Figures 6D, E). We postulated that cell-type specific open chromatin status would correlate with cell type-specific expression of nearby genes and inferred gene activity based on sums of proximal peaks. Notably, inferred gene activities of PIM markers WT1 and SIX1 diminished along the developmental trajectory (Figure 6F), consistent with a respective decline in transcript levels as assessed by RNA-seq (Figure 6B). Concordant with substantial upregulation of transcript levels (Figure 6B), chromatin associated with NR5A1, a marker of early adrenocortical lineage, became increasingly open along the trajectory (Figure 6F). The top 5000 genes with the greatest variance in accessibility exhibited different activity patterns along the pseudotime trajectory (Figure 6G). Consistent with our transcriptomic data, gene activities that were downregulated along the pseudotime trajectory were related to early development (enriched with GO terms such as “multicellular organism development” or “anterior/posterior pattern specification”), whereas genes whose activities were upregulated were related to steroidogenesis or adrenocortical function (enriched with GO terms such as “regulation of lipid metabolic process” or “lipid transport”) (Figure 6G). Moreover, single cell motif enrichment along the pseudotime trajectory identified footprints of transcription factors related to the PIM (e.g., SIX1, LHX1, EMX2) earlier and those related to adrenocortical lineage (e.g., OSR2, NR5A1) later in pseudotime, suggesting that these factors transcriptionally regulate the PIM to adrenocortical cell fate transition through identified cis-elements (Figure 6H). Accordingly, NR5A1 motifs were frequently identified in intronic and promoter regions of NR5A1 itself and genes encoding steroidogenic enzymes (e.g., CYP11A1) providing additional insight into its function (Figure 6I).
Figure 6. Single cell chromatin accessibility landscape of human adrenocortical specification.
(A) Genomic annotation of open chromatin regions identified in fl09 and fl22 were categorized in peaks unique to fl09 or fl22 or overlapped between fl09 and fl22. Stacked barplots show genomic distribution of open chromatin peaks in fl09 and fl22. Different colors indicate different functional regions of the genome.
(B) The top part of this heatmap shows HOMER motif enrichment analyses using different peaks. Each column is a motif, and colors indicate the motif enrichment significance. The bottom part of this heatmap shows the expression of genes coding for transcription factors correlated with motifs in fl09 and fl22. Arrow indicates p-value of 0.01.
(C) Integration of fl09 and fl22 scATAC-seq. Cells are embedded using PHATE and are clustered using Signac. Ten cell clusters were identified; sample information and cell numbers are listed in the table to the right.
(D) Trajectory analysis using PHATE and principal trajectory curve fitted on the embedding. Cells were colored according to pseudotime.
(E) Pseudotime is consistent with the sample information.
(F) Gene activity scores for WT1, SIX1, NR5A1 and motif enrichment scores for WT1, SIX1, NR5A1 motifs.
(G) Pseudotime heatmap ordering of gene accessibility across cell differentiation from fl09 to fl22. Gene accessibility was determined by a gene activity score generated by counting the read counts per cell in the gene body and promoter region (2000bp upstream of TSS).
(H) Pseudotime heatmap ordering of chromVAR TF motif across cell differentiation from fl09 to fl22 (left). The footprint profiles of two representative TFs (right).
(I) Representative genomic browser screenshot of NR5A1 and CYP11A1. In each screenshot, the top two tracks are bulk RNA-seq showing gene expression; the bottom two tracks show the NR5A1 motif loci indicating potential NR5A1 binding sites.
A critical function of NR5A1 in survival and steroidogenesis of FZLCs
In mice, fetal adrenal specification and steroidogenesis is initiated in a dose sensitive manner by the transcription factor, NR5A125–28, with biallelic mutations in murine Nr5a1 leading to adrenal agenesis and PAI as a result of defective proliferation and/or increased apoptosis immediately after the formation of the AP25,26. However, the functional significance of biallelic NR5A1 mutations in human fetal adrenal development is unclear. Taking advantage of our FZLC induction platform, we next determined the role of NR5A1 during early adrenocortical development. As simple disruption of NR5A1 by CRISPR/Cas9 in NR5A1-tdTomato reporter hiPSCs could affect tdTomato expression unpredictably, we established a new NR5A1-knock-in/knock-out hiPSCs in which the DNA binding domain spanning exon 2 and 3 of NR5A1 was replaced with a 2A-tdTomato-polyA cassette in WG reporter hiPSCs, thus enabling simultaneous gene disruption and visualization of the promoter activities of NR5A1 (WGNT-NR5A1−/− SV20-211-19, herein designated as KIKO) (Figures S1F, G). Similar to wild-type isogenic 211 hiPSCs, KIKO iPSCs were induced into NT+ cells at fl22 with roughly equal induction of NT+ cells per aggregate, suggesting that NR5A1 is dispensable for initial lineage specification (Figure 7A). However, NT fluorescence intensity, which was already somewhat lower in the mutant line at fl22, was progressively downregulated after ALI culture, suggesting that NR5A1 may autoregulate its promoter activity (Figures 7A, B), consistent with our scATAC-seq data and a previous report in mice (Figure 6I)29. Remarkably, upon ALI culture, mutant aggregates exhibited thinning and disintegration, and unlike FZLCs in wild-type aggregates, consisted of partly necrotic cells with loss of nuclei, karryorrhexis and scanty cytoplasm (Figures 7B–D). Accordingly, there was markedly diminished recovery of NT+ cells and increased numbers of cleaved caspase 3+ apoptotic cells (Figures 7E, F). Mutant aggregates also showed reduced steroidogenic gene and protein expression (Figures 7G–I) and a consequent dramatic reduction in Δ5 steroid synthesis (Figure 7J). Together, these findings support the critical role of NR5A1 in survival and steroidogenesis of human fetal adrenal cortex.
Figure 7. NR5A1 is essential for FZLC survival and steroidogenesis.
(A) FACS plots of wild-type (211) or NR5A1 mutant (KIKO) aggregates at fl22 (left) and ali21 (right) of induction. MFI, mean fluorescence intensity.
(B) Bright field (BF) and fluorescence (NT) images of fl22, ali7, ali14 and ali21. Bar, 500 μm.
(C) Number of NT+ cells per aggregate at fl22 and ali21 in wild-type or mutant lines. ***, p < 0.001; n.s., not significant.
(D) Hematoxylin and eosin images of wild-type or mutant aggregates at ali21. Mutant aggregates consist of hypocellular stroma with scattered eosinophilic anucleate cells (area 1, arrowhead) and foci of cell clusters with karyorrhexis and scanty cytoplasm (area 2), consistent with necrotic cell death. Bar, 20 μm.
(E) IF images of the wild-type or mutant aggregates at fl22 for active caspase 3 (aCas3) (green), co-stained for NR5A1 (red) and DAPI (white). Merged images for aCas3 and DAPI are shown. Bar,10 μm.
(F) The quantification of active caspase 3 positive population in (E). Statistical significance is determined by Fisher’s exact test (**, p < 0.01).
(G, H) IF images of the aggregates at fl22 (G) or ali28 (H) for CYP11A1 or CYP17A1 (green), co-stained for NR5A1 (red) and DAPI (white). Merged images of NR5A1 and CYP proteins are shown on the right. Bar,10 μm.
(I) qPCR quantification of gene expression for key marker genes in wild-type and mutant aggregates at fl22. Statistical significance is determined by two-tailed Welch’s t-test (*, p < 0.05; **, p < 0.01; ***, p < 0.001).
(J) Production of Preg, DHEA, and DHEA-S by wild-type (211) or mutant (KIKO) aggregates at indicated time points. Statistical significance is determined by two-tailed Welch’s t-test (*, p < 0.05; ***, p < 0.001).
(K) PCA of transcriptomes of mutant NT+ cells at fl22 projected together with those of wild-type NT+ cells used in Figure 4E. Samples at fl14 and fl22 are highlighted in colors.
(L) Heatmap of Pearson correlation showing close correlation of mutant (KIKO) aggregates at fl22 to wild-type (211) aggregates at fl14.
(M) Scatter plot comparison of the averaged values of wild-type (211) and mutant (KIKO) aggregates at fl22 as in Figure 4C.
(N) Scatter plot comparison of the averaged values between wild-type (211) at fl14 and mutant (KIKO) aggregates at fl22.
(O) DEGs upregulated in wild-type fl14 aggregates compared to fl09 aggregates (as defined in Figure 4C, highlighted in red) projected to a scatter plot used in (M).
(P) Schematic representation showing induction of FZLCs and related lineages from hiPSCs.
To gain further insight into the role of NR5A1, we conducted bulk RNA-seq analyses of NR5A1 mutants. Remarkably, PCA and correlation coefficient heatmaps suggested that mutant NT+ cells at fl22 were similar to wild-type NT+ cells at fl14, supporting a critical role for NR5A1 in the fl14-to-fl22 transition (Figures 7K, L). Accordingly, pairwise comparison of mutant and wild-type NT+ cells at fl22 exposed a marked loss of genes related to steroidogenesis in mutant NT+ cells, including CYP17A1, CYP11B1, APOA1, LDLRAD1 and MC2R, which were enriched in GO terms, such as “cholesterol metabolic process” or “adrenal gland development” (Figure 7M, Table S5). Like wild-type NT+ cells at fl14, mutant NT+ cells at fl22 exhibited persistent expression of developmental genes, which were enriched with GO terms such as “multicellular organism development” (Figure 7M). Although the mutant NT+ cells at fl22 were transcriptionally similar to wild-type cells at fl14, a small number of genes were already downregulated in mutants, including NR6A1, NR0B1, APOA1, FDXR, LGR4, which might represent early NR5A1 target genes (Figure 7N, Table S5). In contrast, some genes potentially involved in adrenal development or steroidogenesis (e.g., STRA6, GATA6, DLK1, RUNX1T1, HSD17B8, PDE1C) that were upregulated in wild-type fl14 cells versus those of fl9, were also expressed in mutant cells at fl22, suggesting that initial expression of some early adrenocortical genes is driven by NR5A1-independent mechanisms (Figure 7O). Together, these data highlight the key role of NR5A1 in early adrenocortical survival and steroidogenesis in humans.
Discussion
In this study, we developed an in vitro model system in which adrenocortical lineages derived from hiPSCs recapitulate the normal developmental trajectory. Remarkably, we found that inhibition of NODAL/ACTIVIN or NOTCH signaling are critical for adrenocortical induction (Figure 7P). We also found that SHH is supportive of overall growth/viability of aggregates, but not essential for the induction of the adrenocortical lineage. These findings suggest that the PIM lineage defaults to adrenocortical specification when inductive cues towards other lineages (e.g., nephrogenic lineage) are absent. The elucidation of mechanisms by which NODAL/ACTIVIN or NOTCH dampen adrenocortical specification warrants further investigation.
Like human fetal adrenocortical cells, our hiPSC-derived adrenocortical cells produced substantial quantities of Δ5 steroids (i.e., DHEA, DHEA-S) and low/undetectable quantities of Δ4 steroids (e.g., cortisol, aldosterone)1. This is in stark contrast to earlier studies in which adrenocortical-like cells induced from hiPSCs/ESCs or multipotential stem cells primarily produced Δ4 steroids that are normally produced by adult adrenocortical cells30–32. In contrast to our study, these prior studies over-expressed NR5A1, and it is therefore possible that steroidogenesis was induced directly by NR5A1 without the appropriate cellular or epigenetic context. Moreover, these prior studies utilized agonists of cAMP/PKA signaling for induction of adrenocortical cells. Notably, in our study, adrenocortical lineages were induced in an ACTH/cAMP/PKA independent manner, which is more in accord with human adrenocortical development in vivo where ACTH is critical for growth but not for specification or initial development1,33,34.
Our transcriptome analyses identified progressive developmental trajectories of the adrenocortical lineage with remarkable similarity to those observed in vivo. Surprisingly, single cell transcriptome analyses also identified the presence of minor cell types that could not be detected by bulk analyses, including WT1+ nephrogenic cells and RSPO3+ capsule-like cells that both arose from the same progenitor as the adrenocortical lineage according to trajectory analyses (Figure 5A). In mice, the pre- and postnatal adrenal capsule surrounding the cortex provides a critical niche by providing RSPO3 to subjacent adrenal cortex, thereby maintaining adrenocortical stem cells that are critical for homeostasis of the adrenal cortex and zonation35. Similar cell types were also identified in human fetal adrenals, but their origin was unknown1,12. Our transcriptomic comparison suggests that, like the adrenocortical lineage, CapLCs that closely resemble Cap in vivo originate from the PIM/CE.
We also noted that during the trajectory towards FZLCs, there was a transient upregulation of DZ markers (e.g., NOV, HOPX)12. Accordingly, IF of ali28 identified cells with DZ markers. Our scRNA-seq analysis of human fetal adrenals has recently demonstrated that FZ fate can be established either through direct differentiation from the AP or indirectly through the DZ12. In the current trajectory analyses, DZLCs appear in the midst of the trajectory leading to FZLCs, suggesting that some FZLCs might be generated indirectly through DZLCs (Figure S6F). However, DZLCs were not identified as a distinct cell cluster, presumably due to their relatively small number in our scRNA-seq. Thus, the lineage relationship between FZLCs and DZLCs remains inconclusive, and it is possible that some FZLCs are induced directly. Interestingly, DZLCs preferentially locate at the periphery of the FZLC cluster where it is juxtaposed to mesenchymal cell types including the CapLCs. Thus, it is tempting to speculate that CapLC-provided instructive cues such as RSPO3 might potentiate canonical Wnt/β-catenin signaling to promote the DZ-like fate. The requirement for CapLCs in the formation of DZLCs warrants future investigation.
In this study, homozygous null NR5A1 mutants showed marked loss of cell viability and decreased steroidogenesis in the remaining cells, accompanied by downregulation of genes encoding steroidogenic proteins (e.g., STAR, CYP11A1, CYP17A1) (Figure 7). Interestingly, the transcriptome of mutant NT+ cells at fl22 was remarkably similar to that of wild-type NT+ cells at fl14, suggesting that NR5A1 is the major driver of the fl14-to-fl22 transition, during which steroidogenic genes are upregulated. Notably, some genes upregulated during the fl9-to-fl14 transition that coincides with the upregulation of NR5A1 (NT) were previously annotated as AP markers (e.g., STRA6, RUNX1T1, GATA6, DLK1)12,15. The initial increase in these markers and that of NR5A1 was not affected in the NR5A1 mutant line, suggesting that there are yet unidentified mechanisms involved in their activation. Of note, upon ALI culture, NR5A1 mutant lines gradually reduce NT expression, suggesting that like the developing mouse adrenal cortex, NR5A1 in developing human adrenal cortex autoregulates its expression after its initial upregulation 29. In support, NR5A1 binding motifs were identified in distal and proximal regulatory regions of the NR5A1 loci. Future studies will systematically identify NR5A1-responsive genes and cis-regulatory mechanisms governing their expression through a multimodal sequencing approach to enhance our mechanistic understanding of how NR5A1 orchestrates FZ survival and steroidogenesis in human fetal adrenals. Such findings will not only allow us to provide a precise diagnosis for patients with NR5A1 mutations, but will identify novel gene regulatory networks, defects of which might lead to PAI or adrenocortical carcinoma36,37.
Limitations of the study
This induction platform will serve as a critical foundation for understanding human adrenal development and its dysfunction, and will also serve as a stepping stone for eventual cell replacement therapy for patients with adrenocortical dysfunction. While we demonstrated the robustness of the platform using multiple hiPSC lines, some validation experiments were performed only in 211 hiPSC-derived aggregates including histomorphologic/ultrastructural characterization and steroid measurements. Furthermore, despite remarkable transcriptomic similarities between FZLCs and FZ, comparative epigenetic characterization was not conducted in this study due to our limited access to human embryo materials, but warrants further investigation.
STAR ★ Methods
RESOURCE AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be addressed to, and will be fulfilled by, the lead contact, Kotaro Sasaki (ksasaki@upenn.edu)
Materials availability
hiPSCs, cDNAs generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.
Data and code availability
Accession numbers generated in this study are GSE201794 (10x Chromium scRNA-seq data for hiPSC-derived aggregates), GSE201793 (10X Chromium scATAC-seq data for hiPSC-derived aggregates) and GSE201792 (bulk RNA-seq for hiPSC-derived aggregates and the fetal samples). The codes used for pseudotime analysis are available at GitHub repository: https://github.com/chengkeren/FZLCs).
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Collection of human embryo samples
The fetal adrenal glands (9, 17, 19 wpf) and the heart at 9 wpf utilized for bulk RNA-seq or histologic analyses were obtained from donors undergoing elective abortion at the University of Pennsylvania and Fukuzumi Obstetrics and Gynecology Hospital. All experimental procedures were approved by the Institutional Review boards at the University of Pennsylvania (#832470) and the Hokkaido University (19–066). All donors gave consent for use of aborted fetuses in research. Embryo ages were determined by ultrasonographic measurement of the crown lump length or head circumference. The sex was determined with sex-specific PCR performed on genomic DNA with primers specific to the ZFX/ZFY loci15. The adrenal glands or the heart were submerged in RPMI-1640 medium and isolated from the surrounding connective or adipose tissues under a stereomicroscope.
Feeder free culture of human iPSCs
Parental hiPSC lines, Penn123i-SV20 (male) and Penn067i-312-1 (female), were obtained from the iPSC core facility at the University of Pennsylvania38. 1390G3 (female) was obtained from Dr. Masato Nakagawa at the Center for iPS Cell Research and Application, Kyoto University39. For maintenance of hiPSCs, cells were cultured on 6-well plates (Thermo Fisher Scientific) coated with iMatrix-511 Silk (Nacalai USA) in StemFit Basic04 medium (Ajinomoto) supplemented with 50 ng/mL basic FGF (Peprotech) or StemFit Basic04CT (complete type) medium at 37 °C under 5% CO2. For passaging or use for induction into adrenocortical lineages, hiPSC at day 6–7 after passaging were treated with a 1:1 mixture of TrypLE Select (Life Technologies) and 0.5 mM EDTA/phosphate-buffered saline (PBS) for 12–15 min at 37 °C to dissociate them into single cells. 10 μM ROCK inhibitor (Y-27632; Tocris) was supplemented in a culture medium 24 h after passaging hiPSCs. For single cell cloning of human iPSCs bearing WGNT or NT fluorescence reporters, human iPSCs were cultured in StemFit Basic03 medium supplemented with 50 ng/mL basic FGF.
Authentication of cell lines used
For counting chromosomes, hiPSCs were incubated with 100 ng/ml of KaryoMAX Colcemid solution (Gibco) for 10 h in culture medium. Cells were dissociated using TrypLE Select and incubated with hypotonic solution (75 mM KCl) for 30 min at 37 °C. The cells fixed by Carnoy’s solution were dropped onto glass slides in a moisty chamber to prepare chromosomal spread. The number of chromosomes was counted using DAPI staining and the cell lines confirmed to be harboring 46 chromosomes were further analyzed for G-banding using Cell Line Genetics (Madison, WI). For all genetically modified cells, a master and secondary cell bank were generated in the following fashion. First, cell lines were expanded until there were sufficient cells to cryopreserve at least five vials as a master cell bank. Second, one vial from a master cell bank was thawed and expanded to create at least 20 cryovials as the secondary cell bank. Directed induction experiments were conducted using the secondary cell bank. We verified that all cell lines banked are mycoplasma free using MycoAlert mycoplasma detection assay (Cambrex). All experiments used passage number matched iPSCs (between p25 and p35). Cell lines used for experiments were not passaged more than six times after thawing to minimize the risk of culture-induced genetic changes.
METHOD DETAILS
Generation of hiPSC lines bearing knock-in fluorescent reporter alleles
For a donor vector construction to generate the WT1-p2A-EGFP (WG) alleles, homology arms spanning the 3’ end of WT1 loci (left arm:1157 bp; right arm: 1287 bp) were PCR-amplified from the genomic DNA of male hiPSCs (Penn123i-SV20) and were sub-cloned into the pCR2.1 vector using the TOPO TA cloning kit (Life Technologies). The p2A-EGFP fragment with the PGK-Puro cassette flanked by loxP sites was PCR-amplified from and inserted in-frame at the 3’-end of the WT1 coding sequence using the GeneArt Seamless Cloning & Assembly Kit (Life Technologies). The WT1 stop codon was removed to allow in-frame p2A-EGFP protein expression. Likewise, for donor vector construction to generate the NR5A1-p2A-tdTomato (NT) alleles, homology arms spanning the 3’ end of NR5A1 loci were PCR-amplified from genomic DNA of male hiPSCs (585B1 8-6-8, a gift from Drs. Kotaro Sasaki and Mitinori Saitou, Kyoto University) and subcloned into the pCR2.1 vector40. The p2A-tdTomato fragment with the PGK-neo cassette flanked by the loxP site was also PCR-amplified and inserted in-frame at the 3’-end of the NR5A1 coding region. For the donor vector construction to establish NR5A1−/−; NR5A1-p2A-tdTomato-polyA alleles [NT knock-in, knock-out (NT-KIKO)], the genomic region spanning exons 2 and 3 of NR5A1 loci from Penn123i-SV20 was first cloned in PCR2.1, which constituted a part of homology arms. This vector was linearized by inverse PCR skipping two zinc finger domains (critical domains for DNA-binding activities) constituting most of exons 2 and 3. This left only the first 42 bases of exon 2 and the last 38 bases of exon 3 in the homology arm (left arm: 892 bp; right arm: 1234 bp). Then, the p2A-tdTomato-SV40 polyA fragment with the PGK-neo cassette flanked by the loxP site was PCR-amplified and inserted in-frame to replace the zinc finger domains using a GeneArt Seamless Cloning & Assembly Kit. Finally, to reduce the random integration events, an MC1-DT-A-polyA cassette was PCR-amplified and inserted into the WG, NT or NT-KIKO donor vector, as described previously41.
Pairs of single-guide RNAs (sgRNAs) targeting the sequence at the 3’ end of WT1 or NR5A1, exon 2 or 3 of NR5A1 were designed using the Molecular Biology CRISPR design tool (Benchling). Pairs of sgRNA sequences were as follows: WT1 3’ end (5’-TCTGATGCATGTTGTGATGG-3’ and 5’-ACTCCAGCTGGCGCTTTGAG-3’); NR5A1 3’ end (5’-CTTGCAGCATTTCGATGAGC-3’ and 5’-CAGACTTGAGCCTGGGCCG-3’); NR5A1 exon 2 (5’- AAGGTGTCCGGCTACCACTA-3’ and 5’- ACACCTTGTCCCCGCACACG -3’); NR5A1 exon 3 (5’-CCGCTTCCAGAAATGCCTGA-3’ and 5’-TCTTGTCGATCTTGCAGCTC-3’). The sgRNAs were cloned into the pX335-U6-Chimeric BB-CBh-hSpCas9n (D10A) expression vector to generate the sgRNAs/Cas9n vector (a gift from Dr. Feng Zhang, Addgene, #42335; http://n2t.net/addgene:42335; RRID: Addgene_42335)42.
For generating hiPSCs bearing WGNT double fluorescent reporter alleles (WGNT SV20-211), the donor vectors (5 μg) and sgRNAs/Cas9n vectors (1 μg each) targeting WT1 and NR5A1 loci were simultaneously introduced into one million hiPSCs (Penn123i-SV20, male) by electroporation using a NEPA21 Type II Electroporator (Nepagene). After drug selection with puromycin and neomycin, cells were transfected with a plasmid expressing Cre recombinase (7.5 μg) to remove the PGK-Puro and PGK-Neo cassettes. These cells were plated into 96 well plates (1 cell/well) using the single-cell plating mode of FACSAria Fusion. After 10–12 days of culture using StemFit 03 medium in 96 well plate and 4–6 days in 12 well plates, cells were harvested; half of them were used for genotyping, and the other half were cryopreserved. To generate hiPSCs bearing an NT single fluorescence reporter (NT 312-2121 and NT 1390G3-2125), the donor and sgRNAs/Cas9n vectors targeting NR5A1 loci were introduced into hiPSCs (Penn067i-312-1 or 1390G3, female) followed by neomycin selection. To generate KIKO hiPSCs (WGNT-NR5A1−/− SV20-211-19), hiPSCs (Penn123i-SV20) were transfected with the donor vector and sgRNAs/Cas9n targeting the 3’ end of WT1 loci and subjected to puromycin selection. These cells were subsequently transfected with a donor vector and two pairs of sgRNAs/Cas9n targeting NR5A1 exons 2 and 3, respectively, and subjected to neomycin selection. The cells were then transfected with a plasmid expressing Cre recombinase and subjected to single-cell plating using FACSAria Fusion. Clones bearing targeted alleles, random integration, or Cre recombination were determined by genotyping PCR using the primer pairs listed in Table S6. Sequences of targeted alleles or non-targeted alleles were confirmed to be devoid of indels or other unexpected mutations by Sanger sequencing. Based on this screening, correctly targeted clones without random integration were selected. Thus, WGNT SV20-211 hiPSCs and WGNT SV20-KIKO19 were homozygous for WG and NT, NT 1390G3-2125 hiPSCs were homozygous for NT, and NT 312-2121 were heterozygous for NT.
Induction of FZLCs though floating and ALI culture
To generate three-dimensional aggregates, hiPSCs were plated at 10,000 cells per well in V-bottom 96-well low-cell-binding plates (Greiner Bio-One) in DB27 medium (DMEM/F12 HEPES, 2% B27 supplement, 1% Glutamax, 1% insulin-transferrin-selenium [ITS-G], 1% MEM non-essential amino acids Solution, 90 μM 2-mercaptoethanol, 50 U/ml penicillin-streptomycin [all obtained from Thermo Fisher Scientific]), in the presence of 10 μM Y27632 (R&D Systems) and 2 ng/ml human BMP4 (R&D Systems). The plates were centrifugated at 210 g for 4 min to promote cell aggregation before incubation at 37 °C under 5% CO2 normoxic condition. After 24 h (at fl1), aggregates were transferred to U-bottom 96-well low-cell-binding plates (Greiner Bio-One) in a mesoderm-inducing medium containing 0.3 ng/ml BMP4 (R&D Systems) and 10 μM of CHIR99021 (R&D Systems). On fl3, half of the medium was replaced with fl1 medium containing 0.3 ng/ml BMP4, 10 μM of CHIR99021 and 10 μM Y-27632. On day 5, aggregates were transferred to MK10 medium (Minimum Essential Medium alpha [α-MEM, Thermo Fisher Scientific], 10% KSR (Thermo Fisher Scientific), 55 μM 2-mercaptoethanol, 100 U/ml penicillin/streptomycin (Thermo Fisher Scientific) containing 0.3 ng/ml BMP4, 2 μM SB-431542 (Sigma), 10 μM of CHIR99021 and 10 μM Y-27632. On fl7, aggregates were transferred to an MK10 medium containing 1 ng/ml BMP4, 30 μM SB-431542, 2 μM of CHIR99021, 0.1 μM retinoic acid, 50 ng/mL human sonic hedgehog/SHH (R&D Systems) and 10 μM Y-27632. On fl10, aggregates were transferred to MK10 medium containing 3 ng/ml BMP4, 30 μM SB-431542 (Sigma), 50 ng/mL human SHH, 10 μM of IWR-1 (Sigma), 2 μM SU-5402 (Tocris), and 10 μM DAPT (Tocris). On fl13, half of the medium were replaced with the same medium on fl10. On fl15, aggregates were transferred to MK10 medium containing, 2 μM SB-431542 (Sigma), 50 ng/mL SHH, 1 μM of IWR-1, 1 μM SU-5402, and 10 μM DAPT. On fl22, aggregates were transferred to BioCoat Collagen I Inserts with 0.4 μm PET membranes (Corning, 354444), which were subsequently placed on 24-well plates (Falcon) containing MK10 medium with 1 μM IWR-1, 5 μM DAPT, 50 ng/mL SHH, and 2 μM SU-5402. The inserts were placed in 24-well containing 200 μL of MK10 medium to equilibrate for 1 h prior to use. The inserts in 24-well plates were incubated at 37 °C and 5% to initiate ALI culture. The medium was changed every three days until harvesting aggregates for experiments.
Metanephric mesenchyme (MM) induction from hiPSCs (described in Figures S2A, C) was performed based on a previous study with minor modifications16. Briefly, DB27 medium was used in all processes and the cytokine cocktail was as follows; [fl0; 0.5 ng/ml BMP4 and 10 μM Y-27632], [fl1; 1 ng/mL BMP4 and 10 μM of CHIR99021], [half medium change on fl03 and fl05; 1 ng/mL BMP4; 10 μM CHIR99021; 10 μM Y-27632], [fl7; 10 ng/ml activin A (R&D), 3 ng/ml BMP4, 3 μM CHIR99021, 0.1 μM retinoic acid, and 10 μM Y-27632], [fl9; 1 μM ChIR99021, 5 ng/ml human FGF9, and 10 μM Y-27632].
Fluorescence-activated cell sorting (FACS)
Aggregates were dissociated using 0.1% trypsin/EDTA for 10–15 min at 37 °C with periodic pipetting. After the reaction was quenched by adding an equal volume of fetal bovine serum, cells were resuspended in FACS buffer (0.1% bovine serum albumin [BSA] in PBS). Cell suspensions were strained through 70-μm nylon cell strainers (Thermo Fisher Scientific) to remove clumps before sorting or analysis. To analyze/sort cells using a cell surface marker, the dissociated cells were stained with APC-conjugated anti-human DLK1 (R&D systems). Cells stained with the cell surface marker and those expressing NT were sorted using FACSAria Fusion (BD Biosciences) and were collected in 1.5-mL Eppendorf tubes or 15-mL Falcon tubes containing CELLOTION (Amsbio). All FACS data were collected using FACSDiva Software v 8.0.2 (BD Biosciences) and analyzed using FlowJo software (BD Biosciences). Statistical significance was analyzed using one-way ANOVA for repeated measures followed by Dunnett’s multiple comparison test or two tailed Welch’s t-test.
Quantitative RT-PCR
Cells were pelleted by centrifugation after isolation (500 g for 10 min) and subsequently lysed for RNA isolation using RNeasy Micro Kit (Qiagen). For some assays, cDNA synthesis was performed using Superscript III (Invitrogen) with oligo-dT primers and used for qPCR assays without amplification. In the remaining assays, cDNAs were synthesized and amplified using 1 ng of purified total RNA as described40. ERCC spike-in RNAs developed by the External RNA Controls Consortium (ERCC; life Technologies [4456740]) were added to the samples to estimate the transcript copy number per cell as described previously40,43. Target cDNAs were quantified using PowerUp SYBR Green Master Mix (Applied Biosystems) with StepOnePlus (Thermo Fisher Scientific). The log2 scale gene expression values were calculated using ΔCt normalized to the average Ct values of PPIA and ARBP. Statistical significance was analyzed using one-way ANOVA for repeated measures followed by Dunnett’s multiple comparison test or two-tailed Welch’s t-test. ND, not detected. *, p < 0.05; **, p < 0.01; ***, p < 0.001 vs first detected time points.
Immunofluorescence analyses
The procedure was performed as described previously12. Briefly, aggregates were fixed with 4% paraformaldehyde (Sigma) in PBS for 2 h at room temperature, washed three times with PBS containing 0.2% Tween-20 (PBST) for 15 min, and successively immersed in 10% and 30% sucrose (Fisher Scientific) in PBS overnight at 4 °C. The samples were embedded in OCT (Fisher Scientific), snap-frozen with liquid nitrogen, and sectioned to 10-μm using a cryostat (Leica, CM1800). Sections were placed on Superfrost Microscope glass slides (Thermo Fisher Scientific) and stored at −80 °C until further analysis. The slides were air-dried, washed three times with PBS and incubated with a blocking solution (5% normal donkey serum in PBS containing 0.2% Tween-20) for 1 hr prior to primary antibody incubation. Several antibodies required antigen retrieval before the incubation. Slides were post-fixed with 10% buffered formalin (Fisher Healthcare) for 10 min at room temperature and washed with PBS three times for 10 min. Antigens were retrieved using HistoVT One (Nacalai USA) for 20 min at 70 °C and washed in PBS for 10 min at room temperature. Slides were subsequently incubated with primary antibodies in blocking solution for 2 h at room temperature, followed by washing with PBS six times (total of 2 h). Subsequently, slides were treated with secondary antibodies and 1 μg/ml DAPI (Thermo Fisher Scientific) in a blocking solution for 50 min at room temperature. In some experiments, live EGFP signals were detected without antibody staining (Figure 1E). Slides were washed with PBS six times and mounted in Vectashield mounting medium (Vector Laboratories) for confocal laser scanning microscopy analysis (Leica, SP5-FLIM inverted). Confocal images were processed using Leica LasX (version 3.7.2).
IF analyses on paraffin sections
For IF analyses of fetal adrenal glands and aggregates during ALI culture, samples were fixed in 10% buffered formalin (Fisher Healthcare) with gentle rocking overnight at room temperature. After dehydration, tissues were embedded in paraffin, serially sectioned at 4-μm using a microtome (Thermo Scientific Microm™ HM325), and placed on Superfrost Microscope glass slides. Paraffin sections were then de-paraffinized using xylene. The procedure of IF was as previously described (Cheng et al., 2022). Briefly, slides were treated with HistVT One for 35 min at 90 °C, then 15 min at room temperature for antigen retrieval. The slides were incubated with primary antibodies in blocking buffer overnight at 4 °C. The slides were washed six times with PBS, followed by incubation with secondary antibodies in blocking buffer and 1 μg/ml DAPI (Thermo Fisher Scientific) in blocking solution for 50 min at room temperature. This step was followed by washing with PBS for six times before mounting in Vectashield mounting medium for confocal laser scanning microscopy.
In-situ hybridization (ISH) on paraffin sections
ISH analysis on formalin-fixed paraffin-embedded sections was performed as described 12. Briefly, the samples were hybridized with a ViewRNA ISH Tissue Assay Kit (Thermo Fisher Scientific) with gene-specific probe sets for human NR5A1 (#VA1-3002602-VT), APOA1 (#VA1-10349-VT), STAR (#VA1-3004623-VT), NR0B1 (#VA1-3001628-VT), NOV (#VA1-3000396-VT), RSPO3 (#VA1-13645-VT) and PDGFRA (#VA1-12826-VT). Probe sets against Bacillus S. dapB (#VF1-11712-VT) were used as the negative control. Sections deparaffinized by xylene were treated with pretreatment solution for 12 min at 90–95 °C, followed by a protease solution for 6 min and 30 sec at 40 °C. After fixation in 10% formaldehyde neutral buffer solution for 5 min, sections were hybridized with the ViewRNA type 1 probe set for 2 h at 40 °C. This step was followed by incubation with the preamplifier probe (25 min at 40 °C), the amplifier probe (15 min at 40 °C) and the Label Probe 1-AP (15 min at 40 °C). The sections were subsequently incubated with the AP enhancer for 5 min at room temperature, followed by development using FastRed Substrate Solution for 1 h at room temperature. The slides were counterstained with DAPI (1 μg/mL) in PBS for 1 h, then mounted in Vectashield mounting medium for confocal laser scanning microscopy.
Measurement of insulin-like growth factor II
To measure insulin like growth factor II (IGF-II) production during induction of FZLCs derived from 211 hiPSCs, culture supernatants 72 h after the last medium change were collected at respective time points and IGF-II concentration was determined by enzyme-linked immunosorbent assay (ELISA) (FineTest, EH0166). Means ± standard deviation (n = 3). P-values for the comparison is determined by one-way ANOVA for repeated measures followed by Dunnett’s multiple comparison test (**, p < 0.01 vs fl10).
Measurement of steroid hormones
Medium for aggregates was refreshed three days before collection for steroid hormone measurement. In some experiments, substrates and various agonists or inhibitors were added when the medium was changed 3 days before collecting supernatants. After 72 h of incubation, supernatants were collected and stored at −20 °C until use. Commercially-available clinical-grade ELISA kits (DRG International) were used to detect pregnenolone (EIA-4170), DHEA (EIA3415), and DHEA-S (EIA1562). All procedures were according to the manufacturer’s instructions.
Spike and recovery and precision experiments were conducted using supernatants obtained from 211 hiPSC-derived aggregates at ali21 as validation of the ELISA kit; %recovery was calculated using the following equation: (spiked sample concentration − sample concentration) *100 / spiked standard diluent concentration. Two concentrations of spiked standard diluent were used to calculate %recovery: pregnenolone (3.2 ng/ml and 12.8 ng/ml), DHEA (5 ng/ml and 15 ng/ml), and DHEA-S (2.5 μg/ml and 5 μg/ml). Four technical replicates were measured in a precision experiment. The optical density (OD) value of 450 nm (OD450 nm) was determined using Synergy 2 (BioTek) and processed using Gen5 software (BioTek). Steroid production levels were normalized to 1×103 NT+ cell numbers in some experiments. Treatment effects of trophic stimulants were compared with the non-treated group (control), and inhibitors were compared with the LDL (50 μg/ml) + forskolin 1 μM group. LDL was added as a source of cholesterol that facilitates fetal adrenal steroidogenesis44. Statistical significance was analyzed using one-way ANOVA for repeated measures followed by Dunnett’s multiple comparison test or two tailed Welch’s t-test.
Steroid quantification by liquid chromatography-tandem mass spectometry (LC-MS/MS)
Prog, 17-OHProg, Aldosterone, DOC, A4, 11OHA4, and 11KA4 were measured using LC-MS/MS. Briefly, the culture medium (100 μl) was combined with internal standards, diluted with deionized water, and loaded onto a supported liquid extraction cartridge (Isolute, Biotage, Charlotte, NC). After 5 minutes, steroids were eluted twice with 0.7 mL methyl-tert-butyl ether, dried under vacuum (Savant, Thermo Fisher), and reconstituted in 0.1 mL 40:60 methanol:water. LC-MS/MS was performed as previously described45. The steroids were quantified in dynamic multiple reaction monitoring modes. The lower limit of quantification for each steroid was estimated from the signal-to-noise ratio of pooled samples and extrapolated to a concentration that achieved a signal-to-noise ratio of 3.
Creative Proteomics measured Prog, Cortisol, and T. Briefly, the culture medium (150 μl) was extracted with solid phase extraction and dried under the nitrogen. Samples were then reconstituted with 100 μl methanol (50 ng/mL progesterone-d9 as internal standard for LC-MS/MS analysis). AB SCIEX Qtrap 5500 tandem mass spectrometer connected to a Waters Acquity UPLC was used for the analysis. A Waters Acquity UPLC BEH C18 column (2.1×100 mm, 1.7 μm) was used, and the column temperature was set to 40 °C, and the injection volume was 6 μl. The mobile phase consisted of 0.1% formic acid 2 mM ammonium fluoride solution (phase A), and acetonitrile (phase B). The mass spectrometer was operated in positive mode using an ESI source.
The following abbreviations were used to describe steroids (Figure 3A): Preg, pregnenolone; 17OH Preg, 17α-hydroxypregnenolone; DHEA, dehydroepiandrosterone; DHEA-S, dehydroepiandrosterone-sulfate; Prog, progesterone; 17OH Prog, 17 α-Hydroxyprogesterone; A4, androstenedione; T, testosterone; DHT, dihydrotestosterone; DOC, 11-deoxycorticosterone; 11OHA4, 11-hydroxyandrostenedione; 11KA4, 11-ketoandrostenedione; 11OHT, 11-hydroxytestosterone; 11KT, 11-ketotestosterone; 11OHDHT, 11-hydroxydihydrotestosterone; 11KDHT, 11-ketodihydrotestosterone.
Bulk RNA-seq library preparation
The fetal adrenal glands and the heart, hiPSCs, and floating and ALI cultured aggregates were used for bulk RNA-sequencing. According to the manufacturer’s instructions, the fetal tissues were dissociated using a Multi Tissue Dissociation Kit 1 (Milteny Biotech). After tissues were minced into <1-mm fragments using scissors, samples were collected in 15-ml Falcon tubes and mixed with dissociation solution (enzyme mix mixed with serum-free DMEM). Then, samples were incubated at 37 °C for 30 minutes in a water bath with pipetting every 10 minutes. At the end of incubation, fetal bovine serum was added (15% of total volume), and cell suspensions were strained through 70-μm cell strainers. Cells were washed once in PBS by centrifugation at 300 g for 5 minutes, and subsequently treated with Red Blood Cell Lysis Solution (Miltenyi Biotec) for 2 minutes at room temperature. After adding 3 ml DMEM, cells were centrifuged at 300 g for 5 min and resuspended in FACS buffer before cell number was counted using trypan blue exclusion. In some samples (fetal adrenal glands at 17 and 19 wpf), cells were subsequently stained with APC-anti-CD10 antibody, and CD10 negative cells were sorted using FACSAria Fusion. Cells were pelleted by centrifugation at 300 g for 5 min and pellets were snap-frozen in liquid nitrogen and stored at −80 °C until RNA isolation. Dissociation of hiPSC-derived aggregates was as described previously41. The following samples were prepared: fl0 hiPSC (WGNT SV20-211 [male, n =2]), aggregates at fl9 (whole WG+ fraction, WGNT SV20-211 [n=2]), fl14 (NT+ fraction, WGNT SV20-211 [n=2]), fl22 (NT+ fraction, WGNT SV20-211 [n=2]), fl22 (NT+ fraction, NT 312-2121 [female, n=2]), fl22 (NT+ fraction, WGNT SV20-KIKO19 [male, n=2]), ali21 (NT+ fraction, WGNT SV20-211 [n=1]), ali21 (NT+ fraction, NT 1390G-2125 [n=1]) and ali 28 (NT+ fraction, WGNT SV20-211 [n=1]). RNA was extracted using a RNeasy Micro Kit (Qiagen) and RNA quality was evaluated using High Sensitivity RNA Screen Tape on an Agilent 4200 TapeStation. The cDNA library was prepared using Clontech SMART-Seq® HT Plus Kit (PN R400749) according to the manufacturer’s protocol. AMPure XP beads were used for library cleanup. 75-base pair reads were sequenced on an Illumina NextSeq 500 machine according to the manufacturer’s protocol.
Mapping reads and data analysis for bulk RNA-seq
Raw sequencing data were demultiplexed as Fastq files using bcl2fastq2 (v2.20.0.422). Barcode and adaptors were removed using Trimmomatic (v0.32). Fastq files were mapped to a UCSC human reference genome (GRH38) using STAR (2.7.1a). A raw gene count table was generated using featurecounts. DEGs were analyzed using edgeR (v3.36.0) with log2 fold change above 2, P-value < 0.05, and FDR < 0.05; log2(FPKM+1) was used to make scatterplots.
10x Genomics scRNA-seq library preparation
WGNT SV20-211 hiPSC-derived aggregates were collected for scRNA-seq analyses: aggregates at fl19 (FACS-sorted whole WG+ fraction [n= 1] or NT+ fraction [n=1]), fl22 (NT+ fraction [n=1]), ali21 (whole WG+ fraction [n=1]) and ali28 (NT+ fraction [n =1]). Aggregates were dissociated in 0.1% Trypsin/EDTA for 10–15 min at 37 °C with periodic pipetting. After the reaction was quenched by adding an equal volume of fetal bovine serum, cells were resuspended in FACS buffer (0.1% BSA in PBS). Cell suspensions were strained through a 70-μm nylon cell strainer (Thermo Fisher Scientific) to remove cell clumps before use. All samples were stained with trypan blue and confirmed to be >85% viable. Cells were loaded into a Chromium microfluidic chip B using the Chromium Single Cell 3ʼ Reagent Kit (v3 chemistry) and Chromium Controller (10X Genomics) to generate gel bead emulsions (GEMs). Gelbead emulsion-RT was performed using a C1000 Touch Thermal Cycler equipped with a deep-well head (Bio-Rad). Subsequent cDNA amplification and library construction steps were performed according to the manufacturer’s instruction. Libraries were sequenced using a NextSeq 500/500 high output kit v2 (150 cycles) (FC-404-2002) on an Illumina NextSeq 550 sequencer.
Mapping reads and data analyses for scRNA-seq
Raw data were demultiplexed with the mkfastq command in cellranger (V5.0.1) to generate Fastq files. Trimmed sequence files were mapped to the reference genome for humans (GRCh38) provided by 10x Genomics. Read counts were obtained from outputs from Cell Ranger. Secondary data analyses were performed in R (v.4.1.0) with Seurat (v.4.0). UMI count tables were first loaded into R using the Read10X function, and Seurat objects were built from each sample. Cells with fewer than 200 genes, an aberrantly high gene count above 7000, or a percentage of total mitochondrial genes >20% were filtered out. Of the 30,319 cells for which transcriptomes were available, 23,570 cells passed the quality-control dataset filters and were used in downstream analysis. We detected 3572 median genes/cell at a mean sequencing depth of 44,499 reads/cell (Figure S6A). Samples were combined, and the effects of mitochondrial genes and cell cycle genes were regressed out with SCTransform during normalization in Seurat; then, gene counts were scaled by 10000 and natural log normalized. Mitochondrial genes and cell cycle genes were excluded during cell clustering, dimensional reduction and trajectory analysis. Cells were clustered according to Seurat’s shared nearest neighbor modularity optimization-based clustering algorithm. Clusters were annotated based on previously characterized marker gene expression, and cluster annotation was generated for downstream analyses. Dimensional reduction was performed with the top 3000 highly variable genes and the first 30 principal components with Seurat. Differentially expressed genes (DEGs) across different clusters calculated using Seurat findallmarkers with thresholds of an average log2-fold change above 0.25 and p < 0.01. Developmental trajectories of cells were simulated using PHATE. Trajectory principal lines were fitted with ElPiGraph.R. Pseudotime was calculated using ElPiGraph.R. The cell cycle was analyzed using CellCycleScoring in Seurat. DEGs between two groups in scatterplots were identified using edgeR 3.34.1 by applying a quasi-likelihood approach and the fraction of detected genes per cell as covariates. The DEGs were defined as those with FDR <0.01, p < 0.01 and a log2-fold-change >0.25. For pseudo-bulk analyses, DEGs were defined as those with FDR <0.01, p <0.01, and a log2-fold-change >1. Data were visualized using ggplot2 and pheatmap. Genes in the heatmap were hierarchically clustered according to Euclidean distance, scaled by row and then visualized with pheatmap. Gene ontology enrichment was analyzed using DAVID v6.8.
Library preparation, mapping reads and data analyses for scATAC-seq
FACS-sorted cells of floating day 9 (fl9) and day 22 (fl22) samples [WT1 positive population from fl9 (211), n=1; NR5A1 positive population from fl22 (211), n=1] were used for scATAC-seq library preparation with a Chromium Single Cell ATAC Reagent Kits (v1.1 Chemistry). All cells preparation, nuclear extraction, subsequent cDNA amplification and library construction steps were performed according to the manufacturer’s instruction. Libraries were sequenced using NovaSeq 6000 v1.5 reagents kit on an Illumina NovaSeq 6000 sequencer. We sequenced a total of 11,365 cells (fl09 WG+ cells [8,404 cells] and fl22 NT+ cells [2,961 cells]), with 12,148 average median fragments per cell (Figure S7G).
The scATAC-seq mapping to the human reference genome was provided by 10X genomics using the cellranger-atac-2.0.0 pipeline. The downstream analysis was performed using Signac (v1.6.0). Generally, initial QC metrics were calculated using Signac, and peak_region_fragments > 3000, peak_region_fragments < 100000, pct_reads_in_peaks > 40, blacklist_ratio < 0.025, nucleosome_signal < 4, and TSS.enrichment > 2 were used for filtration of low-quality cells. Clustering and the remaining analyses followed default parameters. Briefly, term frequency-inverse document frequency (TF-IDF) normalization was performed. The top 95% of most frequently observed features (peaks) were used for dimension reduction and clustering. Iterative latent semantic indexing (LSI) was used for dimensional reduction. Cell clusters were identified using FindClusters in Signac with smart local moving algorithm. The gene activity score was calculated based on the number of fragments for each cell that map to gene and promoter regions (gene body plus 2 KB upstream of TSS). The motif activity score was calculated using chromVAR (v1.16.0) and the JASPAR CORE database (v2020). Transcription factor footprinting was performed using Footprint in Signac. Trajectory was constructed using PHATE with the same peak matrix for clustering. Trajectory principal lines were fitted with ElPiGraph.R. Pseudotime was calculated with ElPiGraph.R. Pseudo-bulk analysis was performed using HOMER (v4.6). Peaks were called by MACS2 and annotated by ChIPpeakAnno (v3.28.1). Genome coverage was generated by DeepTools (v3.5.1) and visualized in the IGV genome browser (v 2.12.3).
Quantification and statistical analysis
Statistical analysis was performed in Excel and R (version 4.1, http://www.r-project.org/). The number of biological replicates and the statistical test performed are indicated in the figure legends or elsewhere in the STAR Methods section. A boxplot shows median (center line), upper and lower quartiles (box limits) and 1.5x interquartile range (whiskers). All fluorescence and histologic images and flow cytometric data were representatives of at least two independent experiments with similar results obtained at each experiment, unless stated otherwise in figures.
Supplementary Material
Table S1. DEGs identified from a pairwise comparison between samples at different time points during FZLC induction, related to Figure 4
Table S2. DEGs identified from a multi-group comparison between hiPSCs, aggregates at fl09-14, fl22, and ALI culture, related to Figure 4
Table S3. DEGs identified from a pairwise comparison between the AP and PIM/ECE from cynomolgus monkey embryos, related to Figure 4
Table S4. DEGs identified from a multi-group comparison between cell types in scRNA-seq analysis, related to Figure 5
Table S5. DEGs identified from a pairwise comparison between wild-type and mutant aggregates, related to Figure 7
Table S6. Primers used in this study, related to STAR Methods
ACKNOWLEDGMENTS
We thank L. King for carefully reviewing the manuscript and providing insightful comments. We thank Fukuzumi Obstetrics and Gynecology Hospital and the Women’s Health and Clinical Research Center at the University of Pennsylvania for human sample collection and T. Moriwaki for assisting immunohistochemistry.
This work was supported in part by the Open Philanthropy funds from the Silicon Valley Community Foundation (2019-197906) and the Good Ventures Foundation (10080664) to K.S.
INCLUSION AND DIVERSITY
We support ethical, inclusive, diverse, and equitable conduct of research.
Footnotes
DECLARATION OF INTERESTS
Y.Seita is employed by Kishokai Medical Corporation.
REFERENCES
- 1.Ishimoto H, and Jaffe RB (2011). Development and Function of the Human Fetal Adrenal Cortex: A Key Component in the Feto-Placental Unit. Endocr Rev 32, 317–355. 10.1210/ER.2010-0001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Rainey WE, Rehman KS, and Carr BR (2004). The human fetal adrenal: Making adrenal androgens for placental estrogens. Semin Reprod Med 22, 327–336. 10.1055/S-2004-861549/ID/47. [DOI] [PubMed] [Google Scholar]
- 3.Mesiano S, Coulter CL, and Jaffe RB (1993). Localization of cytochrome P450 cholesterol side-chain cleavage, cytochrome P450 17 alpha-hydroxylase/17, 20-lyase, and 3 beta- hydroxysteroid dehydrogenase isomerase steroidogenic enzymes in human and rhesus monkey fetal adrenal glands: reappraisal of fu. Journal of Clinical Endocrinology and Metabolism 77, 1184–1189. [DOI] [PubMed] [Google Scholar]
- 4.Buonocore F, and Achermann JC (2020). Primary adrenal insufficiency: New genetic causes and their long-term consequences. Clin Endocrinol (Oxf) 92, 11–20. 10.1111/CEN.14109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Rogler LE, and Pintar JE (1993). Expression of the P450 side-chain cleavage and adrenodoxin genes begins during early stages of adrenal cortex development. Molecular Endocrinology 7, 453–461. 10.1210/MEND.7.3.8097866. [DOI] [PubMed] [Google Scholar]
- 6.Keeney DS, Jenkins CM, and Waterman MR (1995). Developmentally regulated expression of adrenal 17 alpha-hydroxylase cytochrome P450 in the mouse embryo. Endocrinology 136, 4872–4879. 10.1210/ENDO.136.11.7588219. [DOI] [PubMed] [Google Scholar]
- 7.Melau C, Riis ML, Nielsen JE, Perlman S, Lundvall L, Thuesen LL, Hare KJ, Hammerum MS, Mitchell RT, Frederiksen H, et al. (2021). The effects of selected inhibitors on human fetal adrenal steroidogenesis differs under basal and ACTH-stimulated conditions. BMC Med 19, 1–17. 10.1186/S12916-021-02080-8/FIGURES/5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mesiano S, and Jaffe RB (1997). Developmental and Functional Biology of the Primate Fetal Adrenal Cortex. Endocr Rev 18, 378–403. 10.1210/EDRV.18.3.0304. [DOI] [PubMed] [Google Scholar]
- 9.Simoniant MH, and Capp MW (1984). Characterization of Steroidogenesis in Cell Cultures of the Human Fetal Adrenal Cortex: Comparison of Definitive Zone and Fetal Zone Cells. J Clin Endocrinol Metab 59, 643–651. 10.1210/JCEM-59-4-643. [DOI] [PubMed] [Google Scholar]
- 10.Johnston ZC, Bellingham M, Filis P, Soffientini U, Hough D, Bhattacharya S, Simard M, Hammond GL, King P, O’Shaughnessy PJ, et al. (2018). The human fetal adrenal produces cortisol but no detectable aldosterone throughout the second trimester. BMC Med 16, 1–16. 10.1186/S12916-018-1009-7/FIGURES/6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Poli G, Sarchielli E, Guasti D, Benvenuti S, Ballerini L, Mazzanti B, Armignacco R, Cantini G, Lulli M, Chortis V, et al. (2019). Human fetal adrenal cells retain age-related stem- and endocrine-differentiation potential in culture. The FASEB Journal 33, 2263–2277. 10.1096/FJ.201801028RR. [DOI] [PubMed] [Google Scholar]
- 12.Cheng K, Seita Y, Moriwaki T, Noshiro K, Sakata Y, Hwang YS, Torigoe T, Saitou M, Tsuchiya H, Iwatani C, et al. (2022). The developmental origin and the specification of the adrenal cortex in humans and cynomolgus monkeys. Sci Adv 8, 8485. 10.1126/SCIADV.ABN8485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hanley NA, Rainey WE, Wilson DI, Ball SG, and Parker KL (2001). Expression Profiles of SF-1, DAX1, and CYP17 in the Human Fetal Adrenal Gland: Potential Interactions in Gene Regulation. Molecular Endocrinology 15, 57–68. 10.1210/MEND.15.1.0585. [DOI] [PubMed] [Google Scholar]
- 14.Sucheston ME, and Cannon MS (1968). Development of zonular patterns in the human adrenal gland. J Morphol 126, 477–491. 10.1002/JMOR.1051260408. [DOI] [PubMed] [Google Scholar]
- 15.Sasaki K, Oguchi A, Cheng K, Murakawa Y, Okamoto I, Ohta H, Yabuta Y, Iwatani C, Tsuchiya H, Yamamoto T, et al. (2021). The embryonic ontogeny of the gonadal somatic cells in mice and monkeys. Cell Rep 35, 109075. 10.1016/J.CELREP.2021.109075. [DOI] [PubMed] [Google Scholar]
- 16.Taguchi A, Kaku Y, Ohmori T, Sharmin S, Ogawa M, Sasaki H, and Nishinakamura R (2014). Redefining the in vivo origin of metanephric nephron progenitors enables generation of complex kidney structures from pluripotent stem cells. Cell Stem Cell 14, 53–67. 10.1016/j.stem.2013.11.010. [DOI] [PubMed] [Google Scholar]
- 17.Barak H, Huh SH, Chen S, Jeanpierre C, Martinovic J, Parisot M, Bole-Feysot C, Nitschké P, Salomon R, Antignac C, et al. (2012). FGF9 and FGF20 maintain the stemness of nephron progenitors in mice and man. Dev Cell 22, 1191–1207. 10.1016/J.DEVCEL.2012.04.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Takasato M, Er PX, Chiu HS, Maier B, Baillie GJ, Ferguson C, Parton RG, Wolvetang EJ, Roost MS, de Sousa Lopes SMC, et al. (2015). Kidney organoids from human iPS cells contain multiple lineages and model human nephrogenesis. Nature 2015 526:7574 526, 564–568. 10.1038/nature15695. [DOI] [PubMed] [Google Scholar]
- 19.Knarston IM, Pachernegg S, Robevska G, Ghobrial I, Er PX, Georges E, Takasato M, Combes AN, Jørgensen A, Little MH, et al. (2020). An In Vitro Differentiation Protocol for Human Embryonic Bipotential Gonad and Testis Cell Development. Stem Cell Reports 15, 1377–1391. 10.1016/J.STEMCR.2020.10.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Laufer E, Kesper D, Vortkamp A, and King P (2012). Sonic hedgehog signaling during adrenal development. Mol Cell Endocrinol 351, 19–27. 10.1016/J.MCE.2011.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gomes DC, Leal LF, Mermejo LM, Scrideli CA, Martinelli CE, Fragoso MCBV, Latronico AC, Tone LG, Tucci S, Yunes JA, et al. (2014). Sonic Hedgehog Signaling Is Active in Human Adrenal Cortex Development and Deregulated in Adrenocortical Tumors. J Clin Endocrinol Metab 99, E1209–E1216. 10.1210/JC.2013-4098. [DOI] [PubMed] [Google Scholar]
- 22.Huang CCJ, Miyagawa S, Matsumaru D, Parker KL, and Yao HHC (2010). Progenitor Cell Expansion and Organ Size of Mouse Adrenal Is Regulated by Sonic Hedgehog. Endocrinology 151, 1119–1128. 10.1210/EN.2009-0814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rainey WE, Carr BR, Wang ZN, and Parker J (2001). Gene profiling of human fetal and adult adrenals. Journal of Endocrinology 171, 209–215. 10.1677/JOE.0.1710209. [DOI] [PubMed] [Google Scholar]
- 24.Schweizer MT, and Antonarakis ES (2012). Abiraterone and other novel androgen-directed strategies for the treatment of prostate cancer: a new era of hormonal therapies is born. Ther Adv Urol 4, 167–178. 10.1177/1756287212452196. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Luo X, Ikeda Y, and Parker KL (1994). A cell-specific nuclear receptor is essential for adrenal and gonadal development and sexual differentiation. Cell 77, 481–490. 10.1016/0092-8674(94)90211-9. [DOI] [PubMed] [Google Scholar]
- 26.Bland ML, Fowkes RC, and Ingraham HA (2004). Differential Requirement for Steroidogenic Factor-1 Gene Dosage in Adrenal Development Versus Endocrine Function. Molecular Endocrinology 18, 941–952. 10.1210/ME.2003-0333. [DOI] [PubMed] [Google Scholar]
- 27.Bland ML, Jamieson CAM, Akana SF, Bornstein SR, Eisenhofer G, Dallman MF, and Ingraham HA (2000). Haploinsufficiency of steroidogenic factor-1 in mice disrupts adrenal development leading to an impaired stress response. Proceedings of the National Academy of Sciences 97, 14488–14493. 10.1073/PNAS.97.26.14488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Sadovsky Y, Crawford PA, Woodson KG, Polish JA, Clements MA, Tourtellotte LM, Simburger K, and Milbrandt J (1995). Mice deficient in the orphan receptor steroidogenic factor 1 lack adrenal glands and gonads but express P450 side-chain-cleavage enzyme in the placenta and have normal embryonic serum levels of corticosteroids. Proceedings of the National Academy of Sciences 92, 10939–10943. 10.1073/PNAS.92.24.10939. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Zubair M, Ishihara S, Oka S, Okumura K, and Morohashi K (2006). Two-Step Regulation of Ad4BP/SF-1 Gene Transcription during Fetal Adrenal Development: Initiation by a Hox-Pbx1-Prep1 Complex and Maintenance via Autoregulation by Ad4BP/SF-1. Mol Cell Biol 26, 4111–4121. 10.1128/MCB.00222-06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sonoyama T, Sone M, Honda K, Taura D, Kojima K, Inuzuka M, Kanamoto N, Tamura N, and Nakao K (2012). Differentiation of Human Embryonic Stem Cells and Human Induced Pluripotent Stem Cells into Steroid-Producing Cells. Endocrinology 153, 4336–4345. 10.1210/EN.2012-1060. [DOI] [PubMed] [Google Scholar]
- 31.Li L, Li Y, Sottas C, Culty M, Fan J, Hu Y, Cheung G, Chemes HE, and Papadopoulos V (2019). Directing differentiation of human induced pluripotent stem cells toward androgen-producing Leydig cells rather than adrenal cells. Proc Natl Acad Sci U S A 116, 23274–23283. 10.1073/PNAS.1908207116/SUPPL_FILE/PNAS.1908207116.SD07.XLSX. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Yazawa T, Inanoka Y, Mizutani T, Kuribayashi M, Umezawa A, and Miyamoto K (2009). Liver Receptor Homolog-1 Regulates the Transcription of Steroidogenic Enzymes and Induces the Differentiation of Mesenchymal Stem Cells into Steroidogenic Cells. Endocrinology 150, 3885–3893. 10.1210/EN.2008-1310. [DOI] [PubMed] [Google Scholar]
- 33.Gray ES, and Abramovich DR (1980). Morphologic features of the anencephalic adrenal gland in early pregnancy. Am J Obstet Gynecol 137, 491–495. 10.1016/0002-9378(80)91134-5. [DOI] [PubMed] [Google Scholar]
- 34.Bocian-sobkowska J, Woźniak W, and Malendowicz LK (1998). Postnatal involution of the human adrenal fetal zone: Stereologic description and apoptosis. Endocr Res 24, 969–973. 10.3109/07435809809032718. [DOI] [PubMed] [Google Scholar]
- 35.Vidal V, Sacco S, Rocha AS, Silva F da, Panzolini C, Dumontet T, Doan TMP, Shan J, Rak-Raszewska A, Bird T, et al. (2016). The adrenal capsule is a signaling center controlling cell renewal and zonation through Rspo3. Genes Dev 30, 1389. 10.1101/GAD.277756.116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Coulter CL (2005). Fetal adrenal development: insight gained from adrenal tumors. Trends in Endocrinology & Metabolism 16, 235–242. 10.1016/J.TEM.2005.05.010. [DOI] [PubMed] [Google Scholar]
- 37.Lalli E (2010). Adrenocortical development and cancer: focus on SF-1. J Mol Endocrinol 44, 301–307. 10.1677/JME-09-0143. [DOI] [PubMed] [Google Scholar]
- 38.Pashos EE, Park YS, Wang X, Raghavan A, Yang W, Abbey D, Peters DT, Arbelaez J, Hernandez M, Kuperwasser N, et al. (2017). Large, Diverse Population Cohorts of hiPSCs and Derived Hepatocyte-like Cells Reveal Functional Genetic Variation at Blood Lipid-Associated Loci. Cell Stem Cell 20, 558–570.e10. 10.1016/J.STEM.2017.03.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Yokobayashi S, Okita K, Nakagawa M, Nakamura T, Yabuta Y, Yamamoto T, and Saitou M (2017). Clonal variation of human induced pluripotent stem cells for induction into the germ cell fate. Biol Reprod 96, 1154–1166. 10.1093/BIOLRE/IOX038. [DOI] [PubMed] [Google Scholar]
- 40.Sasaki K, Yokobayashi S, Nakamura T, Okamoto I, Yabuta Y, Kurimoto K, Ohta H, Moritoki Y, Iwatani C, Tsuchiya H, et al. (2015). Robust In Vitro Induction of Human Germ Cell Fate from Pluripotent Stem Cells. Cell Stem Cell 17, 178–194. 10.1016/j.stem.2015.06.014. [DOI] [PubMed] [Google Scholar]
- 41.Hwang YS, Suzuki S, Seita Y, Ito J, Sakata Y, Aso H, Sato K, Hermann BP, and Sasaki K (2020). Reconstitution of prospermatogonial specification in vitro from human induced pluripotent stem cells. Nature Communications 2020 11:1 11, 1–17. 10.1038/s41467-020-19350-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Cong L, Ran FA, Cox D, Lin S, Barretto R, Habib N, Hsu PD, Wu X, Jiang W, Marraffini LA, et al. (2013). Multiplex genome engineering using CRISPR/Cas systems. Science (1979) 339, 819–823. 10.1126/SCIENCE.1231143/SUPPL_FILE/PAPV2.PDF. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Nakamura T, Yabuta Y, Okamoto I, Aramaki S, Yokobayashi S, Kurimoto K, Sekiguchi K, Nakagawa M, Yamamoto T, and Saitou M (2015). SC3-seq: A method for highly parallel and quantitative measurement of single-cell gene expression. Nucleic Acids Res 43, e60–e60. 10.1093/nar/gkv134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Carr BR, and Simpson ER (1981). Lipoprotein Utilization and Cholesterol Synthesis by the Human Fetal Adrenal Gland. Endocr Rev 2, 306–326. 10.1210/EDRV-2-3-306. [DOI] [PubMed] [Google Scholar]
- 45.Wright C, O’Day P, Alyamani M, Sharifi N, and Auchus RJ (2020). Abiraterone acetate treatment lowers 11-oxygenated androgens. Eur J Endocrinol 182, 413–421. 10.1530/EJE-19-0905. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. DEGs identified from a pairwise comparison between samples at different time points during FZLC induction, related to Figure 4
Table S2. DEGs identified from a multi-group comparison between hiPSCs, aggregates at fl09-14, fl22, and ALI culture, related to Figure 4
Table S3. DEGs identified from a pairwise comparison between the AP and PIM/ECE from cynomolgus monkey embryos, related to Figure 4
Table S4. DEGs identified from a multi-group comparison between cell types in scRNA-seq analysis, related to Figure 5
Table S5. DEGs identified from a pairwise comparison between wild-type and mutant aggregates, related to Figure 7
Table S6. Primers used in this study, related to STAR Methods
Data Availability Statement
Accession numbers generated in this study are GSE201794 (10x Chromium scRNA-seq data for hiPSC-derived aggregates), GSE201793 (10X Chromium scATAC-seq data for hiPSC-derived aggregates) and GSE201792 (bulk RNA-seq for hiPSC-derived aggregates and the fetal samples). The codes used for pseudotime analysis are available at GitHub repository: https://github.com/chengkeren/FZLCs).