Introduction

Adrenocorticotropic hormone deficiency (ACTHD) is defined by an insufficient production of ACTH by the pituitary, followed by low adrenal cortisol production. Proper diagnosis and management of ACTHD is crucial, as it is a life-threatening condition in the neonatal period, characterized by hypoglycemia, cholestatic jaundice and seizures1. Constitutional ACTHD, which is diagnosed at birth or during the first years of life, can be isolated or associated with deficiencies in other pituitary hormones such as growth hormone (GH) or thyrotropin-stimulating hormone (TSH) and termed combined pituitary hormone deficiency (CPHD). ACTHD is genetically heterogeneous, due to mutations in genes coding mostly for transcription factors responsible for pituitary ontogenesis or their regulation. Responsible genes identified to date include the T-box transcription factor TBX19 (also known as TPIT), NFKB2, LHX3, LHX4, PROP1, HESX1, SOX2, SOX3, OTX2 and FGF8.

Mutations of TBX19 account for approximately two-thirds of neonatal-onset complete isolated ACTHD1. TBX19 is a T-box transcription factor restricted to pituitary pro-opiomelanocortin (POMC)-expressing cells in mice and humans. It is essential for POMC gene transcription and terminal differentiation of POMC-expressing cells in the anterior pituitary (also known as the adenohypophysis)2. POMC is cleaved to produce ACTH in corticotroph cells. TBX19-deficient mice have only a few pituitary POMC-expressing cells, with very low ACTH and undetectable corticosterone levels3. In humans with isolated ACTH deficiency, TBX19 mutations lead to loss-of-function by different mechanisms, such as the K146R (exon 2, c.437 A>G) pathogenic variant located in the T-box region, which results in a loss of DNA-binding ability1.

Thanks to GENHYPOPIT46, an international network aimed at identifying new genetic etiologies of combined pituitary hormone deficiency, we described the first cases of DAVID (Deficient Anterior pituitary with common Variable Immune Deficiency) syndrome, a rare association of hypopituitarism (mainly ACTHD) and immune deficiency (hypogammaglobulinemia) 7. DAVID syndrome is associated with variants in the nuclear factor kappa-B subunit 2 (NFKB2) gene8, 9. In particular, we reported that a pathogenic, heterozygous D865G (exon 23, c.2594 A>G) NFKB2 variant was found in a patient presenting with severe recurrent infections from 2 years of age, who at the age of 5 was diagnosed with ACTHD7. Mutations in the C-terminal region of NFKB2 lead to the disruption of both non-canonical and canonical pathways1012.

Full-length NFKB2 (p100) protein has 2 critical serines, S866 and S870, in the C-terminal domain. Phosphorylation of these sites yields the transcriptionally active form of NFKB2 (p52) by enabling the proteasomal processing of the larger p100 protein. The D865G mutation, located adjacent to the critical S866 phosphorylation site, protects mutant protein from proteasomal degradation, causes defective processing of p100 to p52, and results in reduced translocation of p52 to the nucleus8,11, 12. NFKB2D865G/+ resulted in half of the normal level of processing to p52, whereas NFKB2D865G/D865G exhibited near-absence of p5212. The nuclear factor kappa B (NFKB) signaling pathway is a key regulator of the immune system1214, which likely explains the immune phenotype of patients with DAVID syndrome, including susceptibility to infections and auto-immune disorders15. In contrast, the underlying mechanism causing pituitary disorders remains unknown, with two predominating hypotheses: an indirect autoimmune hypophysitis, preferentially affecting corticotroph function, or a primary developmental defect supported by the expression of NFKB2 in the human fetal pituitary16. However, a developmental role for NFKB2 was not confirmed in the Lym1 mouse model, carrying a heterozygous or homozygous nonsense variant Y868*, which presented apparently normal pituitary development and function although impairing immunity9. Although the mouse continues to be a relevant model for many aspects of pituitary organ development, we sought a higher-throughput method to assess pathogenic effects of new candidate genes or variants on human pituitary differentiation.

Human induced pluripotent stem cell (hiPSC)-derived organoids have emerged as promising models to study many developmental mechanisms and their perturbation in disease17. Pioneering work on human embryonic stem cells and later, hiPSC, established that 3D organoid models can replicate aspects of pituitary development1820, and could thus be of major interest in modeling pituitary disorders18, 21, 22. In the present study, we applied recent progress in genome editing using CRISPR/Cas923 to a refined protocol to derive 3D organoids from hiPSC in order to model ACTHD. We validated the recapitulation of corticotroph cell differentiation in vitro by introducing a TBX19 mutation known to induce isolated human ACTHD and documenting the subsequent deficiencies in corticotroph development. When used to characterize the D865G NFKB2 variant found in patients with DAVID syndrome, hiPSC-derived pituitary organoids displayed dramatically altered corticotroph differentiation in the absence of immune cells, demonstrating for the first time a direct and incontrovertible role for NFKB2 in human pituitary development.

Materials and methods

Culture and maintenance of hiPSC lines

The 10742L hiPSC line, derived from a healthy individual, was used in the study as the WT control (characterized and provided by the Cell Reprogramming and Differentiation Facility [MaSC], Marseille Medical Genetics, Marseille, France). Further information about this line is indicated in Suppl Table S1. Two mutant lines carrying TBX19K146R/K146R and NFKB2D865G/D865G were subsequently generated from 10742L using CRISPR/Cas9 editing as described below. All hiPSC lines were cultured on six-well plates (Corning, #3335, New York, USA) coated with Synthemax II-SC Substrate (working concentration at 0.025 mg/mL, Corning, New York, USA) and maintained undifferentiated in a chemically defined growth medium (StemMACs hiPSC-Brew XF ; MACS Miltenyi Biotec, Paris, France)24. hiPSC lines were maintained in a humidified incubator under conditions of 37°C, 5% CO2, with a daily change of medium, and passaged when cells reached 60-80 % confluency using enzyme-free ReleSR, according to manufacturer recommendations (StemCell Technologies, Canada, #05872).

CRISPR/Cas9 mediated genome editing of hiPSC

CRISPR/Cas9 gene editing was used to introduce the TBX19K146R/K146R and the NFKB2D865G/D865G mutations using a previously described method25. The TBX19K146R/K146R line harbored the c.437A>G, p.Lys146Arg TBX19 (NM_005149) pathogenic missense variant. The NFKB2D865G/D865G line harbored the c.2594A>G, p.Asp865Gly NFKB2 (NM_001322934) pathogenic missense variant.

Preparation of the CRISPR/Cas9 sgRNA plasmid

For each targeted gene, a sgRNA plasmid was prepared as previously described26. Briefly, we designed a sgRNA sequence within 20 nucleotides of the target site, selected with the open-source CRISPOR tool (http://crispor.tefor.net/)27.

The sgRNA sequence used to generate TBX19K146R was (5’>3’): AAGCTGACCAACAAGCTCAA (see also Table S2).

The sgRNA sequence used to generate NFKB2D865G was (5’>3’): GTGAAGGAAGACAGTGCGTA (see also Table S3).

We used the cAB03 (also known as pX459-pEF1alpha) vector as previously described25. In this, the pSpCas9 (BB)-2A-Puro (PX459) V2.0 (Addgene plasmid # 62988, Addgene, Teddington, UK) was modified by an EF1alpha promoter25. The chosen sgRNA was cloned into the cAB03 open plasmid vector as previously described by Arnaud et al25. The final vector expressed the corresponding sgRNA under the control of the U6 promoter, as well as Cas9 under the control of the EF-1alpha promoter, and a puromycin resistance gene.

Single-stranded oligo DNA nucleotides (ssODN) design

We also designed donor sequences to generate TBX19K146R and NFKB2D865G missense mutations. A donor sequence is used for homology-directed repair, allowing the introduction (“knock-in”, KI) of single nucleotide variations with low efficiency. To optimize the efficiency of KI editing, the donor sequences were designed as single-stranded oligo DNA nucleotides (ssODN), carrying the mutation of interest, and also included a silent (or blocking) mutation that mutates the protospacer-adjacent motif (PAM) or disrupts the sgRNA binding region to prevent re-cutting by Cas9 after successful editing25, 2830. The ssODN contained 100 nucleotides with homology arms on each side of the target region. The ssODN sequences were as follows (see also Suppl Tables S2 – Table S3):

For TBX19K146R (5’-3’): TGGATGAAAGCTCCCATCTCCTTCAGCAAAGTGAGGCTGACCAACAAGTTAAATGGAGGCGG GCAGGTACGAATGAGGCGGGCAGGCCTGGCCACCCGCT

For NFKB2D865G (5’-3’): TCCCATTCCTGTCCCCATTTACCCCCAGCAGAGGTGAAGGAAGGCAGTGCCTACGGGAGCCAG TCAGTGGAGCAGGAGGCAGAGAAGCTGGGCCCACCCC.

ssODN were synthesized by Integrated DNA Technologies (Coralville, Iowa, US) at Ultramer™ quality.

Electroporation

hiPSC were transfected with constructed sgRNA/Cas9 vectors by electroporation using the Neon Transfection System 100 µL kit (Invitrogen). A summary of the procedure is described in Suppl. Figure S1.

Clone isolation and screening

On day 3 post-transfection, we used the cleaved amplified polymorphic sequences (CAPS) assay to estimate the efficacy of CRISPR/Cas9 in bulk transfected hiPSC populations and to choose a number of colonies to be picked for further amplification31. A restriction enzyme recognition site was included in each donor template (BseI for TBX19 and MseI for NFKB2). CAPS primers and restriction enzymes were chosen assisted by AmplifX software (version 2.0.0b3; https://inp.univ-amu.fr/en/amplifx-manage-test-and-design-your-primers-for-pcr). CAPS primers are listed in Suppl Table S4. The PCR reaction was performed using the PCR condition reported in Suppl Table S5 – Suppl Table S6. The CAPS products were separated and visualized by electrophoresis on a 2% agarose gel and analyzed for densitometry with ImageJ25, 30.

After 10 days, the CAPS assay was also used to screen and identify possible knock-in (KI) clones. KI clones were then confirmed by Sanger sequencing (Genewiz, Leipzig, Germany). Analyses of Sanger traces were made using Sequencher DNA sequence analysis software (version 5.4.6, Gene Codes Corporation, Ann Arbor, MI USA, http://www.genecodes.com) to align sequences of KI clones with WT sequence to confirm the successful edition. Sanger sequencing primers are described in Suppl Table S4. KI clones were amplified and then cryopreserved.

In vitro pituitary organoids differentiated from hiPSC

Pituitary organoids were produced in parallel using the same protocol for WT and edited hiPSC lines. Two batches of differentiation with two hundred organoids for each hiPSC line were generated, respectively WT versus TBX19K146R/K146R organoids or WT versus NFKB2D865G/D865G organoids.

Differentiation was based on a protocol from Matsumoto et al. with some modifications20, 22. hiPSCs were first dissociated into single cells using Accutase. Ten thousand cells per organoid were plated in 20 µL hanging drops under the lid of a 127.8 x 85,5mm single-well plate (Greiner bio-one) in growth factor-free, chemically defined medium (gfCDM) supplemented with 10% (vol/vol) Knockout Serum Replacement (KSR; Thermo Fisher Scientific, #10828-010) and 20 µM Y-27632 (Sigma-Aldrich, # SCM75). The lid was inverted over the bottom chamber (filled with 10 mL of sterile water to avoid the evaporation of droplets) and incubated at 37°C/5% CO2 for 2 days. Once aggregates formed, organoids were transferred to non-adhesive 100x20mm culture dishes (Corning) containing 10 mL of gfCDM medium supplemented with 10% KSR without Y-27632 and incubated at 37°C and 5% CO2 until day 6. From days 6 to days 17, the medium was supplemented with 10% KSR, 5 nM recombinant human bone morphogenetic protein 4 (BMP4; Peprotech, # AF-120-05ET, USA) and 2 µM Smoothened agonist (SAG; Sigma-Aldrich/Merck, # SML 1314). Half of the medium was renewed every 2 - 3 days.

From day 18, BMP4 was withdrawn and half of the medium was renewed every 2 - 3 days. At this point, organoids were maintained under high-O2 conditions (40%) and 5% CO2 in an MCO-5M incubator (Panasonic). From day 30, gfCDM medium supplemented with 20% knockout serum replacement was used, and all medium was renewed every 2-3 days until at least day 105. Induction into pituitary progenitor cells (defined as LHX3+) and into pituitary hormone-producing cells was evaluated by qRT-PCR (on days 0, 6, 18, 27, 48, 75, 105) and immunofluorescence (on days 48 and 105) (see below).

Validating TBX19K146R targeting and assessing off-target mutations by whole genome sequencing (WGS)

For TBX19K146R, we used whole genome sequencing (WGS) to assess on-target and off-target mutations, because the edited site is far from the cut site (15 bases). DNA extracted from control and TBX19K146R/K146R (5 organoids each) was submitted for WGS (Integragen Genomics platform). PCR-free libraries were prepared with the NEBNext UltraTM II DNA Library Prep Kits (New England BioLabs) according to supplier recommendations. Specific double-strand gDNA quantification and a fragmentation (300 ng of input with high-molecular-weight gDNA) sonication method were used to obtain approximately 400 bp fragments. Finally, paired-end adaptor oligonucleotides (xGen TS-LT Adapter Duplexes from Integrated DNA Technologies) were ligated and re-paired. Tailed fragments were purified for direct sequencing without a PCR step. DNA PCR free libraries were sequenced on paired-end 150 pb runs on the NovaSeq6000 (Illumina) apparatus. Image analysis and base calling were performed using Illumina Real Time Analysis (RTA) Pipeline with default parameters. Sequence reads were mapped on the Human Genome Build (GRCh38) using the Burrows-Wheeler Aligner (BWA)32. Variant calling was performed via the GATK Haplotype Caller GVCF tool (GATK 3.8.1, Broad Institute)33. Mutation enrichment was determined using Fisher’s Exact Test. Variants were annotated with Ensembl’s Variant Effect Predictor, VEP) by Integragen Genomics34. These NGS analyses confirmed the presence of on-target mutation and the absence of off-target mutation in the top four predicted off-target sites.

Validating NFKB2D865G targeting, assessing off-target mutations and differential expression analysis by bulk RNA sequencing (RNA-seq)

For NFKB2D865G, we used bulk RNA sequencing (RNA-seq) for analysis of differential RNA expression, and to assess on-target and off-target mutations, because the target site is close to the cleavage site (6 bases). A total of 500 ng of total RNA was extracted from 10 individual organoids (5 NFKB2D865G/D865G and 5 control organoids). RNA-seq libraries were generated using the KAPA mRNA HyperPrep kit (Roche). The quality and profile of the libraries were visualized on a Bioanalyzer using the High Sensitivity DNA assay (Agilent) and quantified on a Qubit using the dsDNA High Sensibility Kit (Thermo Fisher Scientific). Finally, we performed 2x 76-bp paired-end sequencing on a NextSeq500 (Illumina). After quality control using FastQC (Braham Institute), reads were aligned on the GRCh38 human reference genome using STAR (version 2.7.2b)35. BAM files were ordered and indexed using Samtools36. Read counts on genes were determined using StringTie37. Genes differentially expressed between conditions were identified using the R package DESeq2 (version 1.34.0)38. Genes with an adjusted P-value < 0.05 found by DESeq2 were assigned as differentially expressed. We assessed the presence of desired NFKB2 edits according to the CRISPR sgRNA design by direct visualization of the BAM files and the absence of off-target coding variations by the GATK Haplotype Caller GVCF tool33.

Total RNA isolation and quantitative real-time PCR (qRT-PCR) analysis for assessment of mRNA expression of pituitary organoids development

Organoids were harvested, placed on ice and stored dry at -80°C until RNA extraction. The total RNA was extracted from WT, TBX19K146R/K146R and NFKB2D865G/D865G organoids using the NucleoSpin RNA Plus XS kit (Macherey-Nagel), and measured on a NanoDrop TM 1000 Spectrophotometer (Thermo Fisher Scientific). Organoid RNAs were extracted on days 0, 6 and 18 from 7 or 8 organoids, or on days 27, 48, 75 and 105 from single organoids (3-8 organoids per group for each time point). cDNA was synthesized from 200 ng total RNA using a mix of M-MLV reverse transcriptase (Invitrogen, UK, #28025013), dNTP 10mM (Invitrogen, #10297018), RNAseOut Inhibitor (Invitrogen, #10777019) and random primers (Invitrogen, UK, #48190011). Quantitative PCR was performed using iTaq Universal SYBR Green Supermix (Bio-Rad Laboratories), on a QuantStudioTM 5 Real-time PCR system (Thermo Fisher Scientific) and analysed using QuantStudio™ 5 Design & Analysis Software. The thermal cycling profiles were as follows: initial denaturation at 95°C for 20 seconds, followed by 40 cycles of denaturation at 95°C (3 sec), annealing at 60°C (45 sec) and extension at 72°C (45 sec). All samples were assayed in duplicate. The beta-actin (ACTB), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and beta-tubulin (TUBB) transcripts expressions were used as three endogenous reference controls. Target gene expressions were normalized relative to the mean of the housekeeping genes using the delta Ct method, and then normalized relative to the highest mean values of WT during the 105 days using the comparative 2-ΔΔCt method39. Primer sequences used in the experiments are shown in Suppl Table S7.

Immunofluorescence labeling experiments

At days 48 and 105, ten each of WT and mutant organoids were fixed using 4% buffered paraformaldehyde in standard phosphate-buffered saline (PBS) for an hour at RT before rinsing in PBS. After overnight incubation in 30% sucrose/PBS (w/v) at 4°C, organoids were embedded in Surgipath medium (Leica,# 3801480) and snap frozen. Organoids were cryosectioned at 16 µm. After blocking with 0.01% Triton X100 (Sigma-Aldrich, #329830772), 10% normal donkey serum (Millipore, USA, #S30-100 ML) and 1% fish skin gelatin (Sigma-Aldrich, Canada, #G7765-250 ML) in PBS with 5 ng/mL 4’,6-diamidino-2-phénylindole (DAPI) for an hour at RT, slides were incubated with primary antibody diluted in antibody solution (PBS, Triton as above, 1% normal donkey serum and 0.1% fish skin gelatin) overnight at 4°C, washed and incubated and a 1/500 dilution of secondary antibody in PBS for 2 hours at RT. The primary antibodies and dilutions used in this study are summarized in Suppl Table S8. Secondary antibodies were species-specific conjugates to Alexa Fluor 488, 555 or 647 (Thermo Fisher Scientific). Fluorescence images were obtained using LSM 800 confocal microscopy (Zeiss).

3D imaging of pituitary organoids

We used a whole-mount immunostaining and clearing protocol adapted from Belle et al 40. 4-8 organoids of each genotype (WT, TBX19K146R/K146R or NFKB2D865G/D865G) were fixed by immersion in 4 % buffered paraformaldehyde (PFA) in PBS overnight at 4°C. All steps were carried out on a slowly rotating (70 rpm) agitator. After washing in PBS, organoids were either stored at 4°C or blocked and permeabilized with 5 mL PBSGT for 2 days at RT (PBSGT: 0.2% fish skin gelatin, 0.5% Triton X100 in PBS, NaN3 0.01%). Organoids were incubated with primary antibodies in 3 mL PBSGT (at concentrations used on sections) over 5 days at 37°C to increase penetration. Organoids were then washed 6 times in an excess of PBSGT at RT, before incubation with secondary antibodies in 3 ml PBSGT at RT overnight. The wash step was repeated. Organoids were then embedded in 1% low-melting temperature agarose in 1X TAE after it had cooled to about 45°C.

As adapted from the iDISCO+ protocol41, organoids in agarose blocks were progressively dehydrated to 100% methanol (VWR, #20847.360) with 1h at each graded step. After overnight incubation in two parts dichloromethane (DCM, Sigma-Aldrich, #270997) to one part 100% methanol, organoids were immersed in 100% DCM for 30 minutes before changing to 100% dibenzyl ether (DBE; Sigma-Aldrich, #108014) for transparization over two hours. Samples were maintained in DBE at RT and protected from light in an amber glass vial. Just before imaging, they were transferred to room-temperature ethyl cinnamate (Sigma-Aldrich, #112372). Light-sheet fluorescence microscopy (Miltenyi Biotec UltraMicroscope Blaze™) was used to acquire z-stacks of optical sections of pituitary organoids at 4 µm intervals. After 3D reconstruction with Imaris software (version 9.6, Bitplane), the number of corticotrophs in each organoid was defined by counting individualized cell-sized (7 µm) objects positive for ACTH immunoreactivity with the Spots automatic classifier after setting parameters in a region of interest containing positively labeled WT cells, then applied to all organoids. Using the Surfaces tool of the Imaris software, the volume of organoids was determined separately, based on thresholds corresponding to their respective fluorescence intensities.

Statistical analysis

Statistical analyses were performed and visualized with GraphPad Prism 9.5.0 software (GraphPad Software, Inc). Data are expressed as means ± SEM. Comparisons between the two groups (mutant versus WT) were performed by unpaired two-tailed t-tests (non-parametric Mann-Whitney test). n refers to the number of samples for each experiment outlined in the figure legends. p values of < 0.05 (*), < 0.01 (**), and < 0.001 (***) were considered statistically significant differences.

Results

Corticotroph deficiency can be modeled by directed differentiation of TBX19-mutant hiPSC in 3D culture

To first validate the pituitary organoid model and determine whether the TBX19 mutant can affect corticotroph differentiation, we generated a homozygous TBX19K146R/K146R hiPSC line from the isogenic control line using CRISPR/Cas9 (Figures 1A-1D; Suppl. Figure S1). One KI clone carrying TBX19K146R/K146R was obtained after screening 100 clones by CAPS and confirmed by Sanger sequencing (Suppl Figure S2). This mutant clone was then amplified and differentiated into pituitary organoids, in parallel with the control line.

Design of sgRNA and ssODN to edit TBX19K146R mutation

A, Illustration of the TBX19 gene (HGNC: 11596; NCBI: NG_008244.1; Ensembl: ENSG00000143178; Human GRCh38) and TBX19 protein.

B, Wild-type (WT) sequence containing target site, cut site, target and PAM sequences.

C, ssODN design to edit in a missense K146R mutant of TBX19 using CRISPR/Cas9.

D, Sequence analysis of a TBX19K146R/K146R hiPSC clone 63 obtained by Sanger sequencing after screening by CAPS. This clone was subsequently used in this work to differentiate into pituitary organoids (see below).

E, Summary of our strategy procedure:

Step 1: Production of the knock-in hiPSC lines by CRISPR/Cas9 genome editing.

Step 2: Differentiation into the pituitary organoids from mutant hiPSC lines in parallel with WT line using 3D culture, then comparison of the development of organoids between the two groups.

CAPS: cleaved amplified polymorphic sequences; hiPSC: human induced pluripotent stem cells; KI: knock-in; PAM: protospacer-adjacent motif; sgRNA: single guide RNA; ssODN: single-stranded oligo DNA nucleotides; WT: wildtype; 3D: three dimensions.

The ability of the TBX19K146R/K146R line to differentiate into a hypothalamic-pituitary structure was compared to its control line (200 organoids for each line) using a 3D culture method, in which pituitary-like and hypothalamus-like tissues simultaneously developed (Figures 1E, 2A, Suppl Figure S3A-B). Inside of such organoids, hiPSC differentiate into hypothalamic progenitors, while the outer layer differentiates into an epithelium sharing properties with the oral ectoderm that develops into Rathke’s pouch and then the anterior pituitary (Figure 2A, Suppl Figure S3). The expression of several key markers of pituitary development and differentiation was followed in mutant TBX19K146R/K146R organoids matched to control WT organoids over time, using qRT-PCR (d0, d6, d18, d27, d48, d75, d105) and immunofluorescence (d48 and d105) (Figure 2A). Organoids grew from 0.4 mm on day 6 to 1.9 mm in their largest 2D dimension by day 105 (Figure 2B).

Time course of organoid growth and gene expression in WT and TBX19K146R/K146R organoids

A, Culture protocol and schematic to generate pituitary organoids in three-dimensional (3D) culture from hiPSC. Organoids were collected at days (d) 0, 6, 18, 27, 48, 75, and 105 during differentiation to analyze.

B, Bright-field microscopy views of WT organoid examples at different time points throughout differentiation. Scale bars are indicated in each image.

C-G, Relative quantification (RQ) mRNA expression analysis for key markers of pituitary organoids during differentiation: WT (in black line) and TBX19K146R/K146R organoids (in green line). Data show means ± standard error of the mean (SEM); n=1 on day (d)0, n=3 samples per group (8 organoids per sample) at the time points d6 and d18, n=4-8 organoids per group at each subsequent time point (d27, d48, d75 and d105). Mann-Whitney t-test (unpaired, two-tailed, nonparametric). p < 0.05 (*), p< 0.01 (**).

C, Relative quantification expression of HESX1, one of the earliest pituitary placode markers. The expression of HESX1 was not significantly different in TBX19K146R/K146R organoids vs. WT.

D, Relative quantification expression of PITX1, an anterior pituitary progenitor marker in the ectodermal primordium. PITX1 was significantly downregulated in TBX19K146R/K146R organoids by d48 (** p= 0.0022) and d75 (* p= 0.0159).

E, Relative quantification expression of LHX3, a pituitary progenitor transcription factor. LHX3 was significantly lower in TBX19K146R/K146R organoids as compared to WT on d27 (** p= 0.004), d48 (** p= 0.0022) and d75 (** p=0.0079).

F, Relative quantification expression of TBX19, a corticotroph lineage marker. There was no significant difference in TBX19 expression between the two groups (p= 0.195, d75).

G, Relative quantification expression of POMC, a differentiated corticotroph marker. POMC was significantly downregulated in TBX19K146R/K146R organoids by d48 (* p= 0.026) and thereafter (* p= 0.0159 on d75; * p= 0.0317 on d105).

Successful differentiation was validated in WT organoids by qRT-PCR, at early stages with the expression of a set of pituitary transcription factors, including HESX1, PITX1 and LHX3, and at the latest stage by the expression of corticotroph cell markers TBX19 and POMC (Figures 2A, 2C-2G). HESX1 is the first specific transcription factor of the prospective pituitary and its expression is vital for the early determination of the gland. In WT organoids, HESX1 was expressed on day 6 and rapidly downregulated as of day 18 (Figure 2C). Expression of PITX1 (a Rathke’s pouch marker) and LHX3 (a pituitary progenitor marker) was high as of day 27, then reached stable levels from d48 onwards (Figures 2D-2E). Importantly, immunofluorescence demonstrated effective generation of pituitary progenitors in WT organoids with the appearance of LHX3+ cells in the oral ectoderm-like epithelium by day 48 (Figure 3A, Suppl Figure S3A). This epithelium also expressed PITX1 and E-cadherin (Suppl Figure S3C). Hypothalamus progenitors expressed NKX2.1 (Suppl Figure S3B). By day 105, we observed that WT organoids contained many differentiated corticotroph cells, co-expressing TBX19 protein in the nucleus and ACTH in the cytoplasm as seen in confocal microscopy, a critical feature for the rest of our investigations (n=8, Figure 3B, Suppl Figure S3D). Consistent with these images, qRT-PCR results confirmed the highest levels of TBX19 and POMC transcription at day 105 in WT organoids (Figures 2F-2G). Taken together, these data showed that pituitary organoids in 3D culture mimic important aspects of human pituitary ontogenesis, and were characterized by the ability to differentiate into pituitary progenitors by day 48 and then to corticotroph cells by day 105.

Impairment of corticotroph development in TBX19K146R/K146R organoids as compared with control

A, C, Immunostaining of LHX3 and CDH1 (E-cadherin) expression in epithelial cells, typical of Rathke’s pouch ectoderm in early pituitary primordia, was reduced in TBX19K146R/K146R organoids vs wild-type on day 48 (n=10 organoids for each group). Scale bars : 10 μm.

B, D, Immunostaining showed that ACTH and TBX19 expression was reduced in TBX19K146R/K146R organoids vs wild-type on day 105 (n=10 organoids for each group). Scale bars: 10 μm.

We then tested whether TBX19-mutant organoids could be used to model ACTHD. To this end, pituitary organoids carrying TBX19K146R/K146R were cultured in parallel with WT organoids and analyzed using qRT-PCR and immunofluorescence for the pituitary ontogenesis markers as described above. Our data showed that there was no significant difference in the expression of HESX1 (Figure 2C). However, PITX1 expression was significantly decreased by days 48 (p= 0.0022) and 75 (p= 0.0159) in mutant organoids (Figure 2D). There were significantly lower levels of LHX3 expression (p= 0.004 on day 27, p= 0.0022 on day 48 and p= 0.0079 on day 75 (Figure 2E), although there was no significant change in the expression of the TBX19 transcript itself (p= 0.195 on day 75, Figure 2F). Importantly, POMC expression was significantly decreased in the TBX19K146R/K146R organoids by days 48 (p= 0.026), 75 (p= 0.0159) and 105 (p= 0.0317) (Figure 2G).

Next, we checked several pituitary markers by immunofluorescence (Figures 3A-3D). In line with the qRT-PCR results, TBX19K146R/K146R organoids immunostaining on day 48 showed that LHX3 protein expression was decreased, in contrast to WT organoids (n= 10 organoids for each group, Figures 3A, 3C). By day 105, we observed that there were fewer corticotroph cells expressing ACTH and TBX19 protein in TBX19K146R/K146R organoids (n= 10 organoids for each group, Figures 3B, 3D).

Finally, in order to take into account the possibility of regionalized sampling or damage in the process of histological section, we also confirmed that ACTH production was nearly abolished in TBX19K146R/K146R organoids overall on day 105 as compared to control organoids. To make this assessment, whole organoids were rendered completely transparent using a previously validated clearing protocol and imaged by light-sheet microscopy for quantitative analyses (Figures 4A-4B and Suppl Movies 1-2). Indeed, differentiation into corticotroph cells was not uniform throughout the organoids but was concentrated in one or multiple buds. These data demonstrate that the TBX19K146R/K146R genotype leads to a significant decrease in ACTH+ corticotroph cell numbers by day 105 as compared to control organoids (p= 0.0159, WT n=4, mutant n=5, Figure 4C).

3D reconstruction of whole Wild type (WT) and TBX19K146R/K146R organoids on day 105

A, A representative image of whole-mount immunostaining against ACTH in a cleared WT organoid on d105 using light-sheet microscopy. Scale bar: 200 μm.

B, A representative image of whole-mount immunostaining against ACTH in a cleared TBX19K146R/K146R organoid on d105 as above, showing impaired corticotroph differentiation. Scale bar: 200 μm.

C, The number of corticotroph cells per mm3 was significantly decreased in TBX19K146R/K146R organoids (* p= 0.0159). Means ± SEM for n=4 WT, n=5 TBX19K146R/K146R organoids. Mann-Whitney test (unpaired, two-tailed, nonparametric).

Together, this first set of experiments demonstrated that genome editing of hiPSC bearing a homozygous pathogenic variant of TBX19 prevented their differentiation into ACTH-producing corticotrophs.

NFKB2 signaling is vital for corticotroph development

After validating this organoid model for the study of ACTHD with a known TBX19 missense mutation, we used the same approach to investigate the potential role of NFKB2 during hypothalamic-pituitary development. The NFKB2D865G homozygous mutation was chosen because it was shown to severely affect NFKB2 p52 processing12. To do so, we first established a NFKB2D865G/D865G mutant hiPSC line using CRISPR/Cas9 (Figure 5) from the same isogenic control line. Two homozygous KI clones carrying NFKB2D865G/D865G were obtained after screening by CAPS and confirmation by Sanger sequencing (Suppl Figure S4). RNA-seq analyses confirmed the absence of mutation in the top four predicted off-target sites. We then selected one homozygous missense mutation NFKB2D865G/D865G clone (#7) to amplify and generate 200 mutant organoids in parallel with 200 WT organoids in 3D culture.

Design of sgRNA and ssODN to introduce a NFKB2D865G mutation

A, Illustration of NFKB2 gene (HGNC: 7795; Ensembl: ENSG00000077150; Human GRCh38) and NFKB2 protein.

B, WT sequence containing target site, cut site, target and PAM sequence.

C, ssODN design to introduce the missense mutation D865G into NFKB2 using CRISPR/Cas9.

D, Sequence analysis of a representative NFKB2D865G/D865G hiPSC clone 7 obtained by Sanger sequencing after screening by CAPS. This clone was subsequently used in this work to differentiate into pituitary organoids (see below).

CAPS: cleaved amplified polymorphic sequences; hiPSC: human induced pluripotent stem cells; KI: knock-in; PAM: protospacer-adjacent motif; sgRNA: single guide RNA; ssODN: single-stranded oligo DNA nucleotides; WT: wildtype.

Next, we compared the development and differentiation of the NFKB2D865G/D865G versus WT organoids during culture by qRT-PCR and immunofluorescence as described above. HESX1 showed no significant difference in expression over time (Figure 6A). In contrast, PITX1 and LHX3 were significantly decreased in NFKB2D865G/D865G organoids on days 27 and 48 (Figures 6B-6C). Surprisingly, at the latest stage of differentiation (day 105), we observed a significant increase in TBX19 expression in the mutant group (p= 0.0095, WT n=4, mutant n=6, Figure 6D). In contrast, POMC expression levels were significantly lower in NFKB2D865G/D865G organoids compared to WT (p= 0.019, Figure 6E). Of note, as MRI has revealed pituitary hypoplasia in 43% of patients with DAVID syndrome42, we asked whether mutant organoids might be smaller than WT organoids. However, after calculating the volume of organoids on day 105 using Imaris software, these were not significantly different (p= 0.6, WT n=7, mutant n= 8, Figure 6F). Concordant with qRT-PCR results, immunostaining showed less LHX3 expression in NFKB2D865G/D865G organoids on day 48 compared to WT (Figures 7A-7B). Similar results were obtained from ten individual organoids for each line. NFKB2 was co-expressed in LHX3+ pituitary progenitors on day 48 (Figure 7A). Unexpectedly, NFKB2 was markedly less expressed in NFKB2D865G/D865G organoids than in WT organoids (Figures 7A-7B). On day 105, immunostaining demonstrated fewer ACTH+ cells; in addition, some TBX19+ cells expressed no simultaneous ACTH signal in NFKB2D865G/D865G organoids (Figures 7C-7D).

Time course of organoid growth and gene expression in WT and NFKB2D865G/D865G organoids

A-E, Relative quantification (RQ) mRNA expression analysis for key markers of pituitary organoids during differentiation: WT (in black line) and NFKB2D865G/D865G organoids (in green line). Data show means ± standard error of the mean (SEM); n=1 on day (d)0, n=3 samples per group (8 organoids per sample) at the time points d6 and d18, n=4-6 organoids per group at each following time point (d27, d48, d75 and d105). Mann-Whitney t-test (unpaired, two-tailed, nonparametric). p < 0.05 (*), p< 0.01 (**).

A, Relative quantification expression of HESX1. The expression of HESX1 was not significantly different in NFKB2D865G/D865G organoids vs. WT.

B, PITX1 was significantly downregulated in NFKB2D865G/D865G organoids by d27 (** p= 0.0022) and d48 (* p= 0.0022).

C, LHX3 was significantly lower in NFKB2D865G/D865G organoids as compared to WT on d27 (** p= 0.0087), and d48 (** p= 0.0087).

D, TBX19 was significantly increased in NFKB2D865G/D865G organoids by d48 (* p= 0.026) and d105 (p= 0.0095).

E, POMC was significantly downregulated in NFKB2D865G/D865G organoids by d105 (* p= 0.019).

F, There was no significant difference in volume between WT and NFKB2D865G/D865G organoids (p= 0.6126). Data show means ± SEM; n=7 in the WT group, n=8 in the mutant group. Mann-Whitney test (unpaired, two-tailed).

Impairment of pituitary development in NFKB2D865G/D865G organoids

A, B, Immunostaining of LHX3 and NFKB2 expression in the early pituitary-type epithelium was reduced in NFKB2D865G/D865G organoids vs WT on day 48 (n=10 organoids for each group). Scale bars: 10 μm

C, D, Immunostaining showed that by day 105, there was complete absence of corticotroph cells co-expressing ACTH and TBX19 in NFKB2D865G/D865G organoids versus WT organoids on day 105 (n=10 organoids for each group). Note the population of cells expressing TBX19 without ACTH in the mutant example. Scale bars: 10 μm

Three-dimensional reconstruction of whole organoids highlighted this significantly decreased ACTH expression in NFKB2D865G/D865G organoids on day 105 as compared to WT in both qualitative (Figures 8A-8B and Suppl Movies 3-4) and quantitative (p=0.0007, WT n=6, mutant n=8 Figure 8C) assessments. In the absence of an immune system, this is strong evidence in support of a direct role for NFKB2 signaling in pituitary differentiation.

3D reconstruction of whole NFKB2D865G/D865G or WT organoids on day 105

A, A representative image of whole-mount immunostaining against ACTH in a cleared WT organoid on d105 using light-sheet microscopy. Scale bar: 200 μm.

B, A representative image of whole-mount immunostaining against ACTH in a cleared NFKB2D865G/D865G organoid on d105 as above, showing impaired corticotroph differentiation. Scale bar: 100 μm.

C, The number of corticotroph cells per mm3 was significantly decreased in NFKB2D865G/D865G organoids (* p= 0.0007). Means ± SEM for n=6 WT organoids, n=8 NFKB2D865G/D865G organoids. Mann-Whitney test (unpaired, two-tailed, nonparametric).

To investigate downstream pathways altered in NFKB2D865G/D865G pituitary organoids, we performed bulk RNA-seq of five distinct NFKB2D865G/D865G organoids versus five organoids derived from its isogenic hiPSC control line on day 48. Differential expression (DE) gene analysis identified 2,829 significantly upregulated and 2,586 significantly downregulated genes at adjusted p (pAdj) < 0.05. The global results are depicted as a heatmap to show the similarities across organoids sharing genotypes (Figure 9A) and a volcano plot to highlight the magnitudes of DE between the two groups (Figure 9B).

Whole transcriptome profiles of WT vs NFKB2D865G/D865G organoids on day 48

A, Heat map showing global differential gene expression between WT and NFKB2D865G/D865G organoids (n=5 for each group).

B, Volcano plot of genes showing differential gene expression between WT and NFKB2D865G/D865G organoids. Red dots in the left panel indicate 143 genes involved in pituitary-hypothalamic development. Grey dots indicate all other genes detected in RNA-seq.

C, Volcano plot of genes showing differential gene expression between WT vs NFKB2D865G/D865G organoids. Green dots indicate genes coding for 21 key transcription factors (TF) whose expression is significantly changed (fold change < 0.5 or > 2, padj < 0.05). Red dots indicate fold change in expression between 0.5 and 2. Grey dots indicate all other genes detected in RNA-seq.

Pit-Hpt: pituitary-hypothalamic; KI: knock-in; TF: transcription factors

Global NFKB signaling was not affected in NFKB2D865G/D865G organoids, as most genes assigned to that pathway had fold-change values between -0.5 and 0.5, when they were significant (Suppl Figure S5). However, among a list of 143 genes known to have a functional influence on pituitary-hypothalamic development curated from the published literature and also expressed in our model, 66 of these were found to be DE in NFKB2D865G/D865G organoids (pAdj < 0.05; Figure 9B, and Table 1). 42 genes had transcript level changes of two-fold or more and 21 of these were transcription factors (Figure 9C). Expression of most of these transcription factors was downregulated, with a marked decrease in expression of anterior pituitary progenitor markers such as PITX1 and LHX3, as well as lineage precursor markers PROP1 and POU1F1 (Figure 9C and Table 1). In contrast, the earliest Rathke’s pouch marker, HESX1, showed a 2-fold increased expression in NFKB2D865G/D865G organoids, suggesting an impaired progression of pituitary progenitors towards more differentiated stages, with increased expression of epithelial markers CDH1, KRT8 and CLDN6 and decreased transcription of mesenchymal markers CDH2 and VIM supporting that hypothesis (Figure 9C and Table 1).

RNA-seq expression data for a list of 143 genes known to have a functional influence on pituitary-hypothalamic development. Differentially expressed genes (padj < 0.05) in NFKB2D865G/D865G organoids are highlighted in green.

The transcription factor RAX also showed a moderate but significant decrease in its expression (fold change = -0.6, padj < 0.05) in NFKB2D865G/D865G organoids (Figure 9C and Table 1), suggesting that the phenotype could be due, at least partially, to a hypothalamic defect. Indeed, among the factors known to mediate the interaction between the hypothalamus and oral ectoderm during pituitary development, BMP4 and FGF10 had significant fold-changes of 4.35 and 0.26 in NFKB2D865G/D865G organoids respectively, whereas SHH, WNT5A and FGF8 expression levels were not modified (Table 1). Finally, a 2.4-fold increase in FST expression, a gene recently identified as a marker of differentiating corticotrophs16, and concomitantly decreased levels of mature corticotroph markers AR, NR4A2, PCSK1 and POMC (0.63, 0.4, 0.35 and 0.43-fold changes respectively) (Table 1), corroborate the hypothesis that TBX19-positive precursors fail to achieve terminal differentiation in NFKB2D865G/D865G organoids.

Taken together, our data identify NFKB2-mediated signaling as an important factor acting on pituitary development at multiple levels during its maturation.

Discussion

After describing the rare combination of an anterior pituitary deficit, mostly ACTHD, with common variable immunodeficiency in DAVID syndrome, we and others found that affected patients carry NFKB2 gene mutations affecting specific C-terminal residues of the NFKB2 protein known to play important roles in immunity 79. As the spontaneous mouse knockout mutant of the orthologous gene lacks an endocrine phenotype, yet NFKB2 is transcribed in the prenatal human pituitary gland, we wanted to investigate the mechanisms of ACTHD in a human disease model.

The available in vitro models to study ACTHD and CPHD had strong limitations. For instance, studies based on transfections and luciferase-coupled hormone promoters can only give indirect evidence of the effect of a variant of unknown significance on the activity of a given transcription factor, as they are based on non-physiological amounts of DNA and promoter constructs5. Murine models only partially replicate human CPHD, as already shown for PROP1 mutations in ACTHD (absent in mice but present in 40% of humans)43. However, reprogramming differentiated adult cells into induced pluripotent stem cells and advances in genome editing technologies using CRISPR/Cas9 broadened possibilities for modeling novel aspects of human disease ex vivo44. Organoid culture has allowed the modeling of otherwise inaccessible congenital human disorders affecting the brain25, intestine45, kidney46, liver47, pancreas48, ovary49, and lung50. In the present study, we established a human in vitro model of congenital ACTHD using 3D pituitary organoids differentiated from TBX19K146R/K146R and NFKB2D865G/D865G hiPSC generated by CRISPR/Cas9 genome editing, in order to compare them with their isogenic WT equivalents. After confirming that this model could recapitulate relevant aspects of pituitary development, we then determined whether it could also replicate a disease, namely ACTHD.

As the TBX19K146R homozygous variant is known to cause ACTHD in humans, we designed a human organoid model derived from TBX19K146R/K146R hiPSC as a proof of principle. In line with the Tbx19-/- mouse model3, TBX19K146R/K146R organoids displayed markedly defective corticotroph differentiation, as compared to WT organoids, thereby validating the methodological approach. Interestingly, in organoids carrying a TBX19K146R/K146R missense mutation that affects DNA binding, a few corticotroph cells were still present, suggesting either a role for TBX19 independent of its DNA binding activity or persistence of residual DNA binding activity with this mutation.

Using this same approach with NFKB2D865G/D865G organoids generated by CRISPR/Cas9-edited hiPSC, we demonstrated for the first time a role for NFKB2 in pituitary development. Heterozygous mutations in the C-terminal region of NFKB2, such as D865G, were reported in DAVID syndrome, leading to the disruption of both non-canonical and canonical pathways8, 1113, 51, 52. Although NFKB2D865G has been identified in a heterozygous state in patients with ACTHD, we decided to generate a homozygous model of NFKB2D865G/D865G to determine first whether NFKB2 was involved in human pituitary development at all and could explain the defect in ACTH production in DAVID syndrome (OMIM#, 615577) 7, 8, 53. Indeed, while concomitant common variable immune deficiency (CVID) can be explained by the key roles of NFKB signaling in the immune system, the mechanism of the endocrine deficits caused by NFKB2 mutations was unknown. In mouse models, deletion of Nfkb2 causes abnormal germinal center B-cell formation and differentiation14. However, the spontaneous Lym1 mouse model, which carries a truncating variant Y868* in the Nfkb2 orthologue that prevents its cleavage into a transcriptionally active form, had apparently normal corticotroph function and ACTH expression in the pituitary in both heterozygotes and homozygotes9. Mutant Nfkb2 homozygotes do display reduced fertility and more severe immune abnormalities than their heterozygous counterparts 54. Using pituitary organoids differentiated from a NFKB2D865G/D865G hiPSC line in comparison with its unmutated control and in the absence of an immune response, we have demonstrated that human corticotroph development is directly disrupted by the same missense mutation found in DAVID syndrome.

The precise molecular mechanisms by which NFKB2 impairs corticotroph development remain to be determined. However, our results suggest both early and late effects on corticotroph differentiation. First, NFKB2 may be involved in the early steps of pituitary development. RNA-seq data on day 48 of NFKB2D865G/D865G organoid development showed increased HESX1 and BMP4 transcription, which may reflect blockade in an undifferentiated state evocative of the Rathke’s pouch epithelial primordium: indeed, down-regulation of HESX1 is necessary for proper pituitary development55. This is also supported by RNA-seq results showing downregulation of PITX1, HES1, LHX3 and FGF10 in mutant organoids, as their expression is known to be necessary for progression to a later progenitor state56, 57. Finally, concordant with the hypothesis of a differentiation blockade, upregulation of epithelial markers such as E-cadherin and keratin-8 was observed in mutant NFKB2D865G/D865G organoids on day 48 while in contrast, there were no changes in Wnt and hedgehog signaling pathways. Second, NFKB2 could also be involved in the final steps of corticotroph differentiation. While qRT-PCR showed increased levels of TBX19 mRNA in mutant organoids, this did not lead to the expected increase in POMC mRNA transcripts. Similarly, some NFKB2 mutants had TBX19+ cells without simultaneous ACTH expression. This suggests that NFKB2 is necessary for POMC synthesis in TBX19-specified cells, which NFKB2D865G may prevent. This could be due to abnormal DNA binding of the NFKB heterodimer to the POMC promoter58, 59 and/or regulation of PCSK1, the gene coding for the enzyme necessary to convert POMC to ACTH. Alternatively, NFKB2D865G could also block TBX19+ cells in a pre-differentiated corticotroph state through its derepression of chromatin remodeling factors such as EZH260.

The organoid model described in this paper presents several advantages over other approaches. It is a true human model of ACTHD, emphasizing the limits of mouse models as exemplified by the Lym1 model, which showed that Nfkb2 was dispensable for murine pituitary development. Despite some technical constraints, CRISPR/Cas9-based genome editing of hiPSC is a powerful approach to elucidate gene function in congenital pituitary deficiency patients, as shown for the role of hypothalamic OTX2 regulation of pituitary progenitor cells in congenital pituitary hypoplasia20. While the prevalence of pathogenic variants of known genes in primary hypopituitarism is currently estimated to be below 10%6, exome and whole-genome sequencing regularly identify new genes and variants, for which pathogenicity remains uncertain35. This model or refinements thereof could thus become the gold standard to confirm involvement of a given gene in human anterior pituitary development, or to classify new variants of unknown significance as likely benign or pathogenic. Finally, pituitary organoids derived from human embryonic stem cells have been transplanted subcutaneously into mice with hypopituitarism; grafted cells were able to secrete ACTH and respond to corticotropin-releasing hormone stimulation61. The near future may therefore promise patients with hypopituitarism innovative substitutive treatments with their own reprogrammed cells.

In conclusion, we have designed and validated a model replicating human constitutive ACTHD using a combination of CRISPR/Cas9 and directed differentiation of hiPSC into 3D hypothalamic-pituitary organoids. Not only has this proof-of-concept reproduced the known prerequisite for the TBX19 transcription factor in human corticotroph differentiation, it has also allowed a new classification of a NFKB2 variant of previously unknown clinical significance as pathogenic in pituitary development. From a clinical viewpoint, our organoid model can provide evidence for the pathogenic role of new candidate genes or variants investigated in patients where none other is available, especially in a rare disease affecting a poorly accessible organ like the pituitary gland. From a physiological viewpoint, modeling corticotroph deficiency using hiPSC allows to support the functional role of NFKB2 in human pituitary differentiation.

Acknowledgements

This study was supported by the ADEREM (Association pour le Développement de la Recherche Médicale au CHU de Marseille), the Association Française contre les Myopathies (AFM-Téléthon) MoThARD program and by Pfizer. TTM received funding from France Excellence Scholarship of the French Embassy in Vietnam, from the French Society of Pediatric Endocrinology and Diabetology (SFEDP) and from the Marseille Rare Disease (MarMaRa) Institute of Aix-Marseille University. The authors wish to thank Frederique Magdinier, Natacha Broucqsault and Claire El Yazidi from the Marseille Medical Genetics (MMG) cell reprogramming & differentiation facility (MaSC); Sébastien Courrier, from the MMG microscopy platform; Valerie Delague, Catherine Aubert, Christel Castro and Camille Humbert from MMG’s Genomics and Bioinformatics (GBiM) platform; Carole Siret and Mathieu Fallet from the Centre d’Immunologie de Marseille-Luminy (CIML); Ivo Vanzetta and Alberto Lombardini from the Institut de Neurosciences de la Timone (INT) microscopy platform, and Laurie Arnaud and Emmanuel Nivet from the Institut de Neurophysiopathologie (INP), all in Marseille, for their expert assistance, advice or reagents.

Declaration of interests

The authors declare no competing interests.