The intestinal epithelium is a complex and highly dynamic tissue which performs a variety of functions including digestion and absorption of nutrients, barrier formation and maintenance of intestinal homeostasis. Dividing adult stem cells located at the crypt base give rise to all major epithelial cell lineages and enable complete renewal of the epithelial layer over 3–5 days. The identification of reliable markers and key signalling pathways in adult intestinal stem cells has enabled long-term in vitro propagation of intestinal stem cells within self-organising three-dimensional intestinal epithelial organoids (IEOs).1 2 The ability to generate organoids from individual patients provides an unprecedented opportunity to investigate human epithelial cell biology in health and disease. Moreover, organoids from human fetal gut samples can be used to evaluate developmental changes in epithelial cells.3
Epigenetic mechanisms are critical regulators of mammalian development, cellular differentiation and tissue-specific functions. One of the main epigenetic mechanisms is DNA methylation of cytosine in the context of CpG dinucleotides. The role of DNA methylation during intestinal development and homeostasis has been studied in freshly isolated human epithelial cells, biopsies and mouse tissues.4–7 However, it remains unclear whether DNA methylation signatures in the epithelium are a stable cell-intrinsic property or a process dependent on external cues. Despite the recognised importance of DNA methylation in regulating gene expression and the maintenance of cellular identity, investigations of IEO culture systems have so far mainly relied on gene expression analysis.1 2 8 Indeed, some of these studies have demonstrated differences in gene expression in IEOs derived from distinct regions of the gut.9 Yet, an underlying functional mechanism explaining how mucosal IEOs retain regional phenotypic differences has not been identified.
Here, we demonstrate the presence of stable, gut segment-specific DNA methylation signatures in human IEOs, which closely reflect those of matching primary purified epithelium and determine regional transcription and cellular function. In contrast, fetal IEOs showed distinct changes in their DNA methylation profile over time towards the paediatric pattern, suggesting the existence of a cell-intrinsic maturation programme that allows development in vitro without the requirement for other cellular components. Lastly, we demonstrate that IEOs from diseased tissue retains altered DNA methylation signatures, highlighting the potential of organoids to model pathophysiology.
Human intestinal samples
Intestinal biopsies were collected from the terminal ileum (TI) and sigmoid colon (SC) from children under 16 years of age as well as adults undergoing diagnostic endoscopy. All patients included had macroscopically and histologically normal mucosa. Fetal intestine was obtained with ethical approval (REC-96/085) and informed consent from elective terminations at 8–12-week gestational age. Fetal gut was dissected and divided into the proximal (small intestine) and distal (large intestine) sections at the ileocaecal junction. Sample details are listed in online supplementary tables S1 and S2.
Purification of intestinal epithelial cells
Intestinal epithelial cells (IECs) were purified by enzyme digestion and magnetic bead sorting for the epithelial cell adhesion molecule (EpCAM) and purity was assessed as described previously.7 10
Human intestinal epithelial organoid culture
IEOs were cultured according to protocols by Sato et al and Fordham et al2 3 (see online supplementary table S3). Culture of gastric organoids followed a protocol adjusted from Schlaermann et al11 (see online supplementary table S4).
RNA and DNA extraction
DNA and RNA were extracted simultaneously from the same sample as described previously.7
Genome-wide DNA methylation arrays
Bisulfite-converted DNA was measured using the Infinium HumanMethylation450 BeadChip (Illumina, Cambridge, UK). The readout for each CpG is expressed as beta value (from 0=unmethylated to 1=fully methylated) or the log2-beta value (M-value). A subset of IEC samples formed part of previous publications.7 12
RNA sequencing was performed on the Illumina HiSeq 2500 platform. Further information is provided in the supplement.
In vitro experiments
IEOs were in vitro differentiated by culturing in differentiation medium (see online supplementary table S5) for 4 days. For Aza-deoxycytidine (AdC) treatment, 1 µM AdC was added for 48 hours followed by a 4-day AdC-free recovery. For long-term AdC treatment, fetal proximal gut -derived IEOs were treated once with 1 µM AdC for 48 hours at passage 1 and then maintained in regular medium for one or five additional passages.
Bisulfite conversion and pyrosequencing
DNA was bisulfite-converted and locus-specific DNA methylation analysis was performed by pyrosequencing as described previously.7 Primer sequences are provided in online supplementary table S6.
Reverse transcription and quantitative PCR
RNA was reverse-transcribed and used in quantitative PCR (qPCR) as described previously.7 Relative expression was calculated using the ΔΔCt method.13 Primer sequences are provided in online supplementary table S7.
Immunofluorescence and imaging
Fluorescent and brightfield images were obtained using an EVOS FL system (Life Technologies). Organoids were stained using anti-FABP6, anti-MUC5B (Atlas Antibodies) or anti-EpCAM (abcam) antibodies. Further details are provided in the supplement.
Vector construction and genome editing
Plasmid vectors for hCas9 (#41815) and gRNA (#41824) were obtained via Addgene. The targeting vector was generated from human genomic DNA and assembly into pUC118-FLIP-Puro backbone (Addgene #84538).14 Further details are provided in the supplement. Human IEOs were electroporated following the protocol from Fujii et al.15
Statistical analysis for qPCR and pyrosequencing data was performed using GraphPad Prism V.7.00 (GraphPad). Significance levels were determined using multiple t-test with Holm-Sidak correction.
Bioinformatic analysis was performed in R software for statistical analysis V.3.2.3 using Bioconductor V.3.2 packages.
DNA methylation data were analysed using minfi16, sva,17limma18 and DMRcate19. Minfi was used to generate multidimensional scaling (MDS) plots, which allow visualising similarity between large datasets by reducing complexity to a two-dimensional scale. RNA sequencing data were processed using fastq_illumina_filter, cutadapt20, tophat2,21bowtie22, samtools23, htseq-count,24RUVseq.25, DESeq226 and GOseq.27 Further details are provided in the supplement.
Paediatric and adult mucosa-derived intestinal organoids retain gut segment-specific DNA methylation profiles
To investigate DNA methylation and its role in regulating cellular function in human IEOs, we generated organoids from mucosal biopsies from healthy children and adolescents from the distal small bowel (terminal ileum=TI) and distal large bowel (ie, sigmoid colon=SC, figure 1A and online supplementary figure S1A). Additionally, we generated a matching reference sample set of primary, highly purified intestinal epithelium (sorted for the epithelial cell adhesion molecule EpCAM)(figure 1A and refs 7 12). Genome-wide DNA methylation analysis on these samples demonstrated distinct separation of primary epithelium, whole biopsy tissue and non-epithelial, mucosal tissue fraction (figure 1B). Moreover, purified primary IEC displayed highly gut segment-specific DNA methylation profiles as reported previously (figure 1B and refs 7 12). Importantly, IEO cultures clustered closely with primary IEC samples from the same gut segment, indicating that epithelial cells retain a highly gut segment-specific epigenetic profile in culture (figure 1B and online supplementary figure S1B). This was also the case for organoids generated from adult individuals (age 24–60 years, see online supplementary figure S1C and D).
DNA methylation profiling of human intestinal epithelial organoids (IEOs) and primary intestinal epithelial cells (IECs). (A) Summary of experimental setup. Mucosal biopsies were obtained from distal small bowel (terminal ileum=TI) and distal large bowel (sigmoid colon=SC) and used for IEC purification or generation of IEOs. Brightfield images showing established IEOs in culture. (B) Multidimensional scaling plot of genome-wide DNA methylation profiles of purified IEC (labelled as ‘x’), mucosa-derived IEOs (labelled as ‘o’), whole biopsies from SC (labelled as ‘b’) and non-epithelial cell fraction of mucosal biopsies (labelled as ‘n’). (C) Venn diagram indicating the number of significant differentially methylated positions (DMPs) as well as overlap between TI and SC purified primary cells and respective organoids (adjusted p<0.01, n=16 IEC, n=10 IEO for each gut location.) Organoids were passages 1–11. (D) Scatterplots of average DNA methylation beta values of overlapping gut-segment specific DMPs (see figure 1C) in matched IECs and IEOs of the respective gut segment. Pearson correlation r=0.948 (TI) and r=0.955 (SC), p<2.2e-16. (E) Examples of differentially methylated regions displaying gut segment-specific methylation levels in both organoids and primary epithelium. Left panel region upstream of interleukin-6 receptor (IL6R), right panel special AT-rich sequence-binding protein 2-antisense 1 (SATB2-AS1). Chromosomal location, CpGs and DNA methylation (expressed as average beta value (ranging from 0=unmethylated to 1=fully methylated) of sample groups) are displayed. (F) Validation of DNA methylation profiles by pyrosequencing of bisulfite-converted DNA, showing per cent of CpG methylation in genomic regions of IL6R and SATB2-AS1. Data are presented as mean+SD of n=4 per group and representative for two independent experiments. (G) Heatmap of top 1000 DMPs, identified by comparing purified TI versus SC epithelium, in the respective organoids profiled at various passages (P). See also online supplementary figures S1 and S2. SCo, SC organoids; SCp, SC purified; TIo, TI organoids; TIp, TI purified.
To assess to what extent the DNA methylation profiles of organoids reflect gut segment-specific epigenetic signatures of primary epithelium, we performed differential methylation analysis comparing TI and SC for both organoids and primary epithelial samples (supplementary tables S13 and S14) (supplementary files 3 and 4). The majority (87%) of significantly differentially methylated positions (DMPs) between TI and SC organoids were also differentially methylated in the respective primary purified TI and SC IEC (figure 1C). Strikingly, DNA methylation levels of these gut segment-specific DMPs correlated strongly between IEC and organoids in each segment (figure 1D) and the differences (ie, hypo-methylation/hyper-methylation) occurred in the same direction for almost all (99.8%) overlapping DMPs (see online supplementary figure S2A). Combining adjacent DMPs, we identified several thousand gut segment-specific differentially methylated regions (DMRs) between TI and SC with differences highly preserved in the respective organoids, including interleukin-6 receptor (IL6R), special AT-rich sequence-binding protein 2 antisense 1 (SATB2-AS1) and claudin 15 (CLDN15) (figure 1E, see online supplementary figure S2B). We validated the methylation profiles of selected DMRs with pyrosequencing (figure 1F).
Gut segment-specific DNA methylation signatures are stable
Organoids derived from the same gut segment and profiled after a range of culture periods (ie, from 1 to 11 passages) displayed close clustering in the MDS plot (figure 1B), indicating that gut segment-specific DNA methylation profiles are highly stable over prolonged culture periods (up to 3 months). Furthermore, no statistically significant DMPs between low and high passage IEOs were identified in any of the gut segments. Additionally, displaying DNA methylation of the top 1000 gut segment-specific CpGs highlighted the clear clustering of samples according to gut segment with no major differences between higher and lower passage organoids (Figure 1G). These findings were confirmed by locus-specific pyrosequencing of early (passage ≤5) and late (passage ≥10) organoids (see online supplementary figure S2C).
Transcriptional profiles of human IEOs display gut segment-specific signatures of gene expression
In parallel to DNA methylation, we performed transcriptome analysis on organoids and primary IEC by RNA-sequencing (RNA-seq). Unsupervised hierarchical clustering and sample relation analysis of transcriptomes revealed a clear, primary separation of samples according to gut segment (figure 2A). However, within each gut segment organoids were found to separate from primary IEC (figure 2A). The expression levels of several key genes involved in intestinal epithelial cellular function and/or epigenetic regulation are displayed in figure 2B, highlighting both the similarities of samples from the same gut segment and the differences between organoids and primary IEC. Moreover, in contrast to DMPs, only 56% of significantly differentially expressed genes (DEGs) between SC-derived and TI-derived organoids overlapped with DEGs in the respective IEC (figure 2C). However, the majority of overlapping DEGs (98%) showed the same direction when comparing TI and SC. Genes that retain gut segment-specific expression levels in organoid cultures include lysozyme (LYZ), the main glucose transporter solute carrier family 5 member 1 (SLC5A1), CLDN15, cystic fibrosis transmembrane conductance regulator (CFTR) as well as SATB2 (figure 2D). We confirmed expression patterns of some of these key genes identified above by quantitative real-time PCR in an additional sample set (figure 2E; DNA methylation levels are provided in online supplementary figure S3).
Transcriptomic profiling of human intestinal epithelial organoids (IEOs) and primary intestinal epithelial cells (IECs). (A) Hierarchical clustering and sample heatmap of transcriptomes by RNA-sequencing of IECs and organoids from terminal ileum (TI) and sigmoid colon (SC). (B) Heatmap displaying gene expression (ie, rlog-transformed RNA-seq counts) of selected epithelial cell subset markers, genes involved in intestinal epithelial innate immune function and regulation of DNA methylation. (C) Venn diagram displaying number as well as overlap of differentially expressed genes (DEGs) comparing TI versus SC of primary IEC and organoids. Significance cut-off adj. p<0.01 and log2Fold Change>±1.5, n=11 (IEC) and 5 (IEO) for each gut segment. Organoids were passages 1–6. (D) RNA-seq read counts of selected marker genes found to retain gut segment-specific expression patterns in organoid culture. (E) Validation of differentially expressed marker genes by quantitative PCR on a validation sample set, represented as mean+SD, n=3–5 per group, *p<0.05. ALPI, intestinal alkaline phosphatase; CFTR, cystic fibrosis-transmembrane conductance regulator; CHGA, chromogranin A; CLDN15, claudin 15; DEFA, defensin alpha; DEFB1, defensin beta 1; DNMT, DNA methyltransferase; LGR5, leucin-rich repeat containing G protein-coupled receptor 5; LYZ, lysozyme; MUC2, mucin 2; OLFM4, olfactomedin-4; PIGR, polymeric immunoglobulin receptor; SATB2, special AT-rich sequence-binding protein 2; SLC5A1, solute carrier family 5 member 1; TET, ten-eleven translocation; TFF3, trefoil factor 3; TLR, Toll- like receptor.
DNA methylation is stable during in vitro differentiation of IEOs and regulates induction of gut segment-specific gene expression
We hypothesised that the transcriptional differences between organoids and IEC might arise in part from the enrichment for the stem cell niche in maintenance organoid culture. We therefore subjected organoids to in vitro differentiation that mimics migration upwards from the intestinal crypt.2 8 28 Differentiated organoids (dIEOs) displayed subtle microscopic differences, along with reduced expression of stem cell markers and increased epithelial subset markers (figure 3A). Genome-wide DNA methylation analysis of differentiated versus undifferentiated IEOs showed that differentiation did not lead to significant DNA methylation changes. The vast majority of CpGs displayed either minimal (i.e. <0.1 change in methylation beta value) or no change (figure 3B), and differential methylation analysis did not yield any significant DMPs.
DNA methylation-dependent, gut segment-specific induction of gene expression in intestinal epithelial organoids (IEOs) during in vitro differentiation. (A) (I) Brightfield images of terminal Ileum (TI) and sigmoid colon (SC)-derived IEOs in maintenance medium (TIo and SCo), and differentiation medium (dTIo and dSCo), showing change of gross phenotype and (II) change in intestinal stem cell and epithelial cell subset marker expression measured by real-time PCR. (B) Histogram of mean difference in DNA methylation (expressed as beta values) of all ~450K tested CpGs comparing differentiated with matched undifferentiated intestinal organoids, n=8 derived from four patients (C) DNA methylation level of selected CpGs showing stable gut segment-specific differences in undifferentiated and differentiated IEOs, boxplot of n=4 derived from four patients for each group. Array cg-identifier is listed on top. (D) Real-time PCR data showing induction of gene expression during in vitro differentiation, n=5 per group. (E) Immunofluorescent staining of undifferentiated and differentiated IEOs forFABP6 (red, upper panel) and MUC5B (red, lower panel) protein expression. Blue=DAPI, cell nuclei. (F) CpG methylation quantified by pyrosequencing located within ileal and colonic marker genes in IEOs at baseline (Ctrl) and after co-culture with DNA methyltransferase inhibitor (Aza-deoxycytidine (AdC)) over 48 hours followed by a 4-day recovery period. Data shown as mean+SD of n=4 per group. (G) Gene expression of the markers in (F) shown as absolute values normalised to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Mean+SD, n=3–4 per group. KI67, marker of proliferation Ki-67; LRIG1, leucine-rich repeats and immunoglobulin-like domains 1; SI, sucrase isomaltase; SOX9, SRY Box 9; VIL1, villin 1.
Next, we investigated the potential impact of gut segment-specific DNA methylation profiles on regulating mRNA expression in dIEOs. We evaluated genes that had shown limited expression differences between segments in undifferentiated IEOs but contained gut segment-specific DMPs and/or DMRs, including fatty acid binding protein 6 (FABP6), nuclear receptor subfamily 1 group H member 4 (NR1H4), mucins (MUC5B and MUC12) as well as TFF3 (figure 3C). Strikingly, dIEOs displayed gut segment-specific gene expression according to the underlying DNA methylation profile (figure 3D). These patterns were observed for several genes, which displayed both small bowel (ie, TI) and large bowel (ie, SC)-specific mRNA induction profiles. We confirmed that changes in gut segment-specific mRNA expression observed on differentiation reflected changes in protein expression by immunostaining of FABP6 and MUC5B (figure 3E). Our findings suggest that stable DNA methylation profiles contribute to regulating gut segment-specific gene expression and the cellular function of human mucosa-derived IEOs on differentiation.
Disruption of gut segment-specific DNA methylation in IEOs induces aberrant gene expression
To further confirm our hypothesis, we exposed IEOs to the DNA methyltransferase (DNMT) -inhibitor Aza-deoxycytidine (AdC). Treatment with DNMT-inhibitor reduced DNA methylation, primarily at highly methylated loci, and thereby altered gut segment-specific signatures (figure 3F). Importantly, the reduced DNA methylation led to an increase in mRNA expression levels of the respective genes, causing expression of small bowel-specific genes in SC organoids (ie, FABP6, MUC17, CLDN15) and large bowel markers in TI organoids (ie, MUC5B and MUC4) (figure 3G). Hence, DNA methylation is required to maintain gut segment-specific gene expression in IEOs.
Human fetal IEOs display gut segment-specific DNA methylation which undergoes dynamic changes in culture
In addition to paediatric mucosal biopsies, we established IEO cultures from human fetal proximal gut (FPG, ie, small intestine) and fetal distal gut (FDG, ie, large intestine) (figure 4A and online supplementary figure S4A). Epithelial origin and purity of cultures was confirmed by expression of selected epithelial and non-epithelial marker genes (see online supplementary figure S4B and C). Following generation and long-term culture of these fetal organoids, we compared their genome-wide DNA methylation profiles with a sample set of primary purified human fetal IEC as well as paediatric primary IEC. DNA methylation profiles of primary purified fetal IEC separated distinctly from paediatric samples and displayed some differences according to gut segment (figure 4B and ref.7). Similar to paediatric and adult organoids, early passages of fetal organoids clustered close to the respective purified epithelial fraction (figure 4B). Differential DNA methylation analysis indicated that fetal organoids retain a large proportion of gut segment-specific profiles (see online supplementary figure S4D). However, fetal IEOs were found to undergo dynamic DNA methylation changes in culture as higher passage fetal IEOs appeared to cluster closer to paediatric epithelial samples (figure 4B). Given this finding, we aimed to assess whether DNA methylation changes occurring in fetal organoid cultures could indicate a degree of in vitro maturation. We therefore performed differential methylation analysis comparing fetal organoids with primary fetal IEC. We tested for an overlap of identified DMPs with the methylation changes occurring during physiological development from fetal to paediatric epithelium, that is, DMPs between fetal and paediatric primary IEC. Strikingly, the vast majority of identified significant DMPs both for FPG (ie, 83%) and FDG (ie, 80%) indeed overlapped (figure 4C and supplementary tables S15 and S16). Moreover, the direction of DNA methylation changes was the same and correlated strongly in almost all (>99%) overlapping DMPs (figure 4D).
DNA methylation dynamics of human fetal intestinal epithelial organoids (IEOs). (A) Schematic illustration of fetal sample processing and brightfield microscopy images of fetal IEOs at time points 1, 3 and 5 days after seeding and under long-term maintenance conditions. (B) Multidimensional scaling plot of genome-wide DNA methylation profiles of fetal primary intestinal epithelial cells (IECs) and organoids derived from fetal proximal gut (FPG) and fetal distal gut (FDG). Data are displayed in the context of paediatric samples (see figure 1). ‘+’ indicates purified IEC, ‘number’ represents IEO sample and indicates passage. Arrows indicate longitudinal samples of the same IEO line. Fetal organoids: n=10 from eight individuals, paediatric organoids: n=10 from seven individuals (ie, three longitudinal samples). (C) Venn diagrams displaying significant differentially methylated positions (DMPs) (adj. p<0.01) and their overlaps comparing fetal IECs versus fetal IEOs (n=5–10 per group) and fetal IECs versus paediatric IECs (n=5–6 and 16). (D) Correlation plot of fold changes (log2FC) of the overlapping DMPs from (C) (‘maturation-associated DMPs’), indicating that nearly all DMPs show the same direction (ie, higher or lower methylation in fetal IEC with almost 100% concordance for small intestine, 99.7% concordance for large intestine. Correlation: Pearson’s r=0.946 (proximal) and 0.956 (distal), both adj p<0.0001). (E) Heatmap of the top 1000 overlapping DMPs between fetal versus paediatric IECs and fetal IECs versus IEOs (see C) with fetal organoids ordered according to passaging times (P1–P23). As a reference, fetal IECs and paediatric sample values are shown as group average (n=5–16). (F) DNA methylation assessed by pyrosequencing of CpGs located in the promoter region of Toll-like receptor 4 (TLR4), left, and paired-like homeodomain 1 (PITX1), right, during culture of fetal organoids. Sample groups are fetal IEC, early fetal organoids (passage ≤3) and late fetal organoids (passage ≥10) . Mean+SD, n=4–5 per group, *p<0.05 and **p<0.01 versus IEC. See also online supplementary figure S4. FDGo, FDG organoids; FDGp, FDG purified epithelium; FPGo, FPG organoids; FPGp, FPG purified epithelium; SC sigmoid colon; TI, terminal ileum.
To further assess whether DNA methylation differences between higher passage fetal organoids and fetal IEC truly reflect in vitro maturation of fetal organoids over time, we plotted a heatmap of the DNA methylation levels with increasing passage number (figure 4E) for the top 1000 overlapping DMPs that displayed the greatest difference between fetal and paediatric IEC (from figure 4C). While early passage fetal IEOs displayed DNA methylation levels similar to the matching purified fetal IEC, methylation profiles of higher passage organoids showed high similarities with the respective paediatric IEOs and IECs. These maturation patterns occurred in both directions (ie, gain and loss of methylation) and were present in both fetal proximal and distal organoids (figure 4E). Locus-specific pyrosequencing on an additional sample set confirmed DNA methylation of genes we had identified to be differentially methylated in fetal versus paediatric primary IEC, including the pattern recognition receptor Toll-like receptor 4 (TLR4) and the transcription factor paired-like homeodomain 1 (PITX1). These genes displayed dynamic DNA methylation and gene expression changes in fetal organoids during prolonged culture (figure 4F and online supplementary figure S4E).
Dynamic transcriptional changes in human fetal organoids support in vitro maturation
To assess the transcriptional dynamics in human fetal organoids, we performed MDS analysis on RNA-seq profiles, which revealed close clustering with paediatric epithelium from the same gut segment (figure 5A). Similarly to DNA methylation, differential gene expression analysis comparing fetal IEC with fetal organoids demonstrated that the majority of identified DEGs overlap with genes differentially expressed between fetal and paediatric IEC, with almost all expression changes (i.e. >99%) occurring in the same direction (figure 5B, C). In support of our hypothesis that fetal organoids undergo in vitro maturation, a heatmap of the top 100 overlapping DEGs (see also figure 5B) further illustrates distinct transcriptional differences between early and late passage fetal organoids with transcriptional profiles appearing more similar to paediatric epithelium (figure 5D).