VDR is expressed in human colon crypt stem cells

To analyze the expression of VDR protein in human colon mucosa, we performed immunohistochemistry assays in healthy tissue from CRC patients. In the same sections, to identify the colon stem cells, we carried out RNAscope in situ hybridization for detection of the stem marker LGR5 [9] as all available anti-LGR5 antibodies showed a lack of specificity in test experiments (data not shown). We observed a strong predominantly nuclear VDR signal in both top differentiated enterocytes/cells and a population of cells located at the crypt bottom, while those in the mid-crypt displayed a lower signal (Figs 1A and S1). Quantification of data at cell level in 100 crypts from four CRC patients (25 crypts/patient) revealed that VDR is expressed in 64% of crypt bottom cells and that 87% of LGR5⁺ cells were positive for VDR expression (Fig. 1B). We additionally performed double immunostaining of tissue microarrays of healthy colon from 25 CRC patients with antibodies against VDR and PTK7, a stemness-related cell marker [12]. In line with LGR5 data, VDR (nuclear) and PTK7 (extranuclear) proteins were largely co-expressed in crypt bottom cells (Fig. S1). These findings provide evidence, for the first time, that VDR is expressed in human colon stem cells in vivo.

Human normal and tumor colon organoids express VDR and respond to calcitriol

To investigate the effect of vitamin D on human colon stem cells, we successfully generated a living biobank of long-term normal and tumor organoid cultures from 39 CRC patients who had undergone surgery for curative purposes (Table S1). Normal organoids had relatively homogeneous cystic morphology formed by one or two cell layers, whereas tumor organoids were frequently multilayered and showed two alternative morphologies, cystic or compact (Fig. 1C) and resembled the primary tumor morphology (Fig. 1D). The analysis by RT-qPCR of VDR RNA expression revealed slightly higher levels in normal than in tumor organoids (Fig. 2A). RNA expression of CYP24A1, the most responsive calcitriol-target gene that encodes the enzyme responsible for its degradation, was higher in tumor than in normal organoids, as occurs in tumor vs. healthy colon tissue [13]. Calcitriol treatment increased CYP24A1 RNA to comparable levels in the two types of organoids (Fig. 2B). In addition, calcitriol increased both VDR and CYP24A1 proteins to a variable extent in matched normal and tumor organoids from a subset of patients (Fig. 2C) and CA2, a protein typically expressed in normal colon organoids [14] (Fig. 2C). We also performed RNAscope in situ hybridization for LGR5 detection in organoids and, as in colon tissue, VDR protein and LGR5 RNA were mostly co-expressed in normal and tumor organoids. VDR signal was increased by calcitriol in both types of organoids whereas that of LGR5 seemed to increase apparently only in normal organoids (Fig. 2D). Altogether, these results showed that both types of organoids, derived respectively from human normal and cancer stem cells, express VDR and respond to calcitriol.

Calcitriol differentially modulates cell phenotype and proliferation in normal and tumor colon organoids

Based on the induction of differentiation by calcitriol on carcinoma cell lines [15], we next explored the effect of this hormonal agent on the cell phenotype in both types of organoids. Ultrastructural analysis of untreated normal organoids from three patients consistently showed an undifferentiated cell phenotype with decondensed euchromatin and large nucleoli typical of high transcriptional activity, absence of intercellular adhesion structures and microvilli, many free ribosomes, and few Golgi complexes and intermediate filament bundles. Unexpectedly, calcitriol treatment did not change the undifferentiated cell phenotype in normal organoids (Fig. 3A, upper panels). On the contrary, cells of three tumor organoid cultures that showed a similar undifferentiated phenotype responded to calcitriol with changes compatible with differentiation: heterochromatin patches, prominent microvilli and adhesion structures (desmosomes) (Fig. 3A, lower panels), bundles of intermediate filaments, abundant, well-developed, rough endoplasmic reticulum cisternae, autophagic vacuoles, and Golgi complexes (Fig. 3B).

We then explored whether the antiproliferative effect of calcitriol reported in carcinoma cell lines [2, 3] was reproduced in colon organoids. Calcitriol consistently diminished cell proliferation in all seven patient-derived normal organoids under study and, to a variable extent, in tumor organoids (Fig. 4A,B). No differences were found in the number of organoids, which suggests that calcitriol does not affect the clonogenicity of either normal or cancer stem cells (Fig. 4C). Taken together, the VDR RNA level in normal and tumor organoids show a statistically significant correlation with the antiproliferative response to calcitriol (Fig. 4D,E), which may also depend on other calcitriol-related features and/or the tumor-specific mutational status. Together, these results indicate that calcitriol contributes to maintain the undifferentiated phenotype and restricts the proliferation of normal colon stem cells and their progeny, while it induces differentiation and variably inhibits the proliferation of cancer stem cells.

Calcitriol differentially regulates the expression of stemness, differentiation, proliferation and tumorigenesis genes in colon cancer patient-derived normal and tumor organoids

Next we studied the long-term effect of calcitriol on the transcriptomic profiles of six matched normal and tumor organoid cultures by RNA-seq analysis upon 96 h of treatment. Given the differences in function and mutational profiles between colon segments (right-side, left-side) [16, 17], to attenuate interindividual variability [18], we selected patients with left-side colon (descending plus sigma) tumors (Table S1). Mutational analysis of tumor organoids confirmed that they harbored patterns of genetic alterations that are typical of sporadic CRC, with APC (5/6) and KRAS mutations (4/6) (Fig. 5A). Accordingly, nuclear staining of MLH1, PMS2, MSH2, and MSH6 proteins in all cases (data not shown) and lack of POLE mutation ruled out a hypermutated MMR phenotype.

Transcriptomic profiles of untreated normal and tumor organoids showed a high number of differentially expressed genes, which mostly coincide with those previously reported (Fig. 5B; GSE64392) [14]. Paired analysis of data from normal and tumor organoids of each patient followed by joint analysis of all patients’ data revealed that calcitriol significantly changed the expression of 2,107 genes (943 upregulated, 1,164 downregulated) in normal organoids and 1,182 genes (643 upregulated, 539 downregulated) in tumor organoids, which were partially coincident (532, 19.3%) (Fig. 5C,D and Table S2; GSE100785). A strong correlation was found between genes induced by calcitriol in normal and tumor organoids, whereas the correlation was much weaker for repressed genes (Fig. 5E). Common calcitriol-target genes in both types of organoids included known VDR-regulated genes such as CYP24A1 and TRPV6 (Fig. 5F and Table S2). Notably, some clusters of genes showed different quantitative or even inverse regulation by calcitriol in normal and tumor organoids (Fig. 5G, left). The expression level of these genes in untreated normal and tumor organoids is shown (Fig. 5G, right). Calcitriol targets also included genes involved in a variety of metabolic, detoxification, nuclear receptor-related, and other pathways (Table S3). Supporting post-translational regulation by vitamin D, as described in several systems [19, 20], VDR was not identified as a significant regulated gene in the RNA-seq analysis.

Among the top up/downregulated genes by calcitriol in normal and tumor organoids, there were some involved in cell stemness (MSI1, SMOC2), differentiation (CEACAM7, HMGCS2, and MUC2), proliferation (GRK5, RARRES1), and tumorigenesis (ALDH3A1, S100P) (Table S4). So, next we used an independent cohort of 18 normal and 9 tumor organoids to a) confirm the effect of calcitriol on the expression of several genes that were found to be significantly regulated in the RNA-seq assay and, b) study the effect of calcitriol on genes involved in those processes or pathways relevant for colon biology. As expected, differentiation genes were expressed at higher levels in untreated normal than in tumor organoids (Fig. S2). Interestingly, the opposite was found for stemness-related genes, which showed lower expression in untreated normal than in tumor organoids (Fig. S2). In line with the RNA-seq data, RT-qPCR analyses confirmed the upregulation by calcitriol of multiple stemness-related genes including LGR5, MSI1, PTK7, SMOC2, LRIG1, and MEX3A, a marker of slow-dividing subpopulation of LGR5⁺ stem cells, in normal organoids (Fig. 6A) [9, 12, 21, 22]. In tumor organoids, the effects of calcitriol on stemness-related genes were weaker as only MSI1 and LRIG1 were upregulated (Fig. 6A).

Consistent with the ultrastructural analysis, calcitriol induced desmosomal (DSC2), microvilli (PLS1), and adhesion (CDH1) genes in tumor organoids (Table S2). We also confirmed by RT-qPCR that several genes expressed in differentiated enterocytes (KRT20) or in goblet/secretory cells (KLF4, SPDEF, and MUC2) were downregulated by calcitriol in normal but not in tumor organoids (Fig. 6A). Additionally, Gene Set Enrichment Analysis (GSEA) showed that the differentiation signature EPHB2^low vs. EPHB2^high of human colon cells [12] was significantly enriched in calcitriol-treated tumor, but not normal, organoids (Fig. 6B).

We also observed that a series of genes involved in cell proliferation (LRIG1, GRK5, RARRES1, TFF2, and TNS4) and tumorigenesis (ALDH3A1, BCAS1, S100P, PPP1R14D, and SERPINE1) were regulated by calcitriol, preferentially in normal organoids, in a way compatible with the previously described antiproliferative action of calcitriol and a potential antitumoral effect (Fig. 6A). Concordant with its effects on cell proliferation, GSEA revealed an inverse association of the calcitriol profile (RNA-seq) with E2F, mTOR, and c-MYC proliferative signatures (Fig. 6C). GSEA also showed that untreated tumor organoids were enriched in the ‘Core transcriptional embryonic stem cell (ESC)-like’ pro-tumorigenic signature that, as described by Wong and cols [23] is frequently activated in aggressive human epithelial cancers. Remarkably, calcitriol was able to abolish this association in both types of organoids (Fig. 6D). In contrast to data on carcinoma cell lines cultured in 2D conditions [15], calcitriol did not repress classical genes of the Wnt pathway except DKK4 (Fig. 6A), a gene that increases invasiveness, angiogenesis, and chemoresistance [24, 25] (Table S5).

To confirm the specificity of the RNA-seq results, we silenced VDR expression in normal organoids through lentiviral transduction of a shRNA^mir against VDR, which inhibited the regulation of several calcitriol-target genes (Fig. S3). Overall, these results show that calcitriol has different/opposite regulatory effects on stemness and differentiation genes in human normal and tumor organoids.

In addition, we generated colon organoids from wild-type (wt) and, as control, Vdr-deficient (Vdr^−/−) mice and analyzed the transcriptomic profile induced by calcitriol treatment using RNA microarrays. As expected, no genes were regulated by calcitriol in organoids from Vdr^−/− mice, while 143 target genes were identified in wt organoids. Remarkably, only 32 genes overlapped between those regulated by calcitriol in human (RNA-seq analysis) and mouse (RNA microarrays, GSE105117) colon organoids (Fig. S4A), while 111 genes were exclusively regulated in mouse organoids (Table S6). This result agrees with the reported lack of conservation of VDR response elements between primates and rodents [26]. We also show that the calcitriol expression profile in wt mouse organoids was related with the Mex3a^high/Lgr5^high mouse signature of slow-dividing intestinal stem cells [22] (Fig. S4B).

Identification of direct calcitriol-target genes in human colon organoids

To further elucidate the action of calcitriol in colon homeostasis, we searched for direct transcriptional target genes in normal organoids. To this end, three independent chromatin immunoprecipitation-sequencing (ChIP-seq) assays were performed using six additional normal patient-derived organoids (2 per ChIP-seq) treated with calcitriol for 2 h. A variable number of DNA binding sites for VDR were found after calcitriol treatment (Fig. 7A). As in other systems [27], a large number of VDR-binding sites were located in intronic or intergenic regions (Fig. S5A). Identified peaks were assigned to the nearest gene at a maximum distance of 10 kb. As many as 559 genes were coincident between two ChIP-seq assays and 182 genes were coincident in all three, including CYP24A1 and TRPV6 (Table S7; GEO107283). Moreover, GSEA revealed that the calcitriol-direct target gene signature, defined as the 182 genes commonly identified in all ChIP-seq assays, was strongly associated with calcitriol-upregulated genes in the RNA-seq analysis (Fig. 7B). This result agrees with the current model of VDR/nuclear receptor action: direct binding to upregulated genes and predominant indirect action by signaling interference, protein–protein interaction or via microRNAs on downregulated genes [28]. Remarkably, 68 genes of the 182-gene signature overlapped with those regulated (62 up- and 6 downregulated) by calcitriol in the RNA-seq assay (Table S7). The stemness-related MSI1 and SMOC2 genes, several genes with antiproliferative and/or tumor suppressive function (KCNIP3, PPP1R1C, PRKG2, PRR5L, GRK5, and RARRES1, the latter with a VDR-binding site 37 kb away from 5′UTR), and CYP24A1 and TRPV6 (Fig. 7C) were among the calcitriol-direct transcriptional targets (Table S7). ChIP-seq assays also showed enrichment in the consensus direct repeat (DR)3 VDR motif (Fig. 7D). The comparison of our RNA-seq data analysis with public transcription factor (TF) ChIP-seq datasets confirmed VDR as the most significant TF involved in gene upregulation (Fig. S5B).

Human samples

Fresh human tissues were provided by IdiPAZ and Fundación Jiménez Díaz Biobanks, integrated into the Spanish Biobank Network ( www.redbiobancos.es), from individuals diagnosed with colorectal cancer and subjected to surgery between 2013 and 2018. Normal tissue samples were obtained from the area distal to the tumor and the histology of the biopsies was evaluated by the pathology services of La Paz University Hospital and Fundación Jiménez Díaz. All human subjects gave informed consent. The study complied with ethical regulations and was approved by the Ethics Committee of La Paz University Hospital (HULP-PI-1425) and the Fundación Jiménez Díaz (PIC-15/2014).

Establishment of 3D colon and tumor organoid cultures

To establish human normal colon organoids, colon crypts were isolated from human biopsies as previously described [10, 11]. Briefly, colon mucosa biopsies were incubated with a mixture of antibiotics [Primocin (Invivogen, San Diego, CA, USA), Gentamycin and Fungizone (Thermo Fisher Scientific, MA, USA)] for 30 min in rotation at room temperature (RT). Next, tissue was cut into small pieces and incubated twice with 10 mm dithiothreitol (DTT) for 5 min at RT. Samples were then transferred to 8 mm EDTA solution for 5 min at RT and 60 min in slow rotation at 4 °C. Samples were washed in PBS until complete EDTA removal and transferred to a 50 mL conical tube in fresh PBS. Colon crypts were separated from the mucosa after shaking and supernatant was centrifuged at 115 g for 5 min at RT. Crypts were washed twice in washing buffer [Advanced DMEM/F12, 10 mm HEPES, and 10 mm Glutamax (Thermo Fisher Scientific)] and finally pelleted crypts were embedded in Matrigel (Corning, MA, USA). Drops were seeded on pre-warmed, 48-well culture dishes. After Matrigel solidification at 37 °C complete ‘normal’ culture medium was added [50% Advanced DMEM/F12, 50% Wnt3a-conditioned medium, 10 mm HEPES, 10 mm Glutamax, 10 mm Nicotinamide (Sigma-Aldrich, St. Louis, MD, USA), 1x N2 (Thermo Fisher Scientific), 1x B27 (Thermo Fisher Scientific), 1 mm N-acetyl-l-cysteine (Sigma-Aldrich), 1 : 500 Primocin, 0.1 μg·mL⁻¹ Noggin ( Peprotech, New Jersey, NJ, USA), 1 μg·mL⁻¹ Gastrin (Tocris, Bristol, UK), 1 μg·mL⁻¹ RSPO1 (Sinobiological, Beijing, China), 50 ng·mL⁻¹ EGF (Peprotech), 0.02 μm PGE2 (Sigma-Aldrich), 1 μm LY-2157299 (Axon-Medchem, Groningen, The Netherlands), and 10 μm SB-202190 (Sigma-Aldrich)].

To establish mouse normal colon organoids, colon crypts were basically isolated as described for human crypts with minor modifications. Mouse colon was washed in PBS, incubated for 30 min with antibiotics (Primocin) and cut into small pieces. Next, tissue was incubated for 5 min with 8 mm EDTA at RT and 30 min at 4 °C in slow rotation. Then, it was washed in PBS to remove EDTA. Crypts were isolated after shaking and collected in a 50 mL conical tube after passing the solution through a 70-μm mesh filter. The crypts solution was centrifuged at 260 g for 5 min at 4 °C. The pellet was washed in washing buffer and pelleted crypts were embedded in Matrigel on a prewarmed 48-well culture dish. After Matrigel solidification, ‘mouse’ culture medium was added (‘normal’ culture medium minus Nicotinamide, Gastrin, PGE2, and SB-202190).

Human tumor organoid cultures were generated as follows. Human tumor biopsies were washed in PBS several times and incubated in a mixture of antibiotics (Primocin, Gentamycin, and Fungizone) for 30 min in rotation at RT. To obtain single cells, tissue was cut into small pieces and digested enzymatically in a suspension of 1 mg·mL⁻¹ collagenase type IV (Sigma-Aldrich) (in PBS) for 30 min at 37 °C with continuous shaking in a waterbath. We forced cell disaggregation by passing the suspension through a 18G syringe. Next, we added 5% FBS and single cells were collected in a conical 50 mL tube after passing the solution through a 70-μm mesh filter and centrifugation at 250 g for 5 min at 4 °C. To lyse erythrocytes, the pellet obtained after centrifugation was incubated in 157 mm NH₄Cl for 5 min, washed in PBS and centrifuged again. Cells were then washed in washing buffer and finally pelleted cells were embedded in Matrigel and seeded on prewarmed 48-well culture dishes. After Matrigel solidification, ‘tumor’ culture medium was added (‘normal’ culture medium minus Wnt3a-conditioned medium, Nicotinamide and RSPO1).

Growth and expansion of organoid cultures

Normal and tumor culture medium was changed every other day. For passaging, we followed the protocol described [10, 11], with modifications. Briefly, Matrigel-embedded organoids were collected using a scraper in a 15 mL conical tube and incubated with Cell Recovery Solution (Corning) for 30 min on ice with continuous shaking to remove Matrigel. After centrifugation at 260 g 5 min at 4 °C, organoids were washed in washing buffer and centrifuged again. Next, organoids were incubated with disaggregation buffer [washing buffer solution containing 1 mg·mL⁻¹ dispase (Thermo Fisher Scientific)] for 10 min at RT in orbital rotation. Immediately afterwards, 2 mm EDTA was added and the mixture was incubated for an additional 5 min. Following this, the solution was homogenized by passing it through a 21G syringe, collected in a 15 mL conical tube, centrifuged at 250 g for 5 min at 4 °C and washed twice in washing buffer. Pelleted cells were embedded in Matrigel and seeded on culture dishes.

RNAscope in situ hybridization

In situ hybridization for LGR5 on formalin-fixed paraffin-embedded samples (colon mucosa and normal/tumor organoids) was performed using the RNAscope^® 2.5 HD Assay-RED (Advanced Cell Diagnostics; External Service, Dr. Ilse Rooman's laboratory at Vrije Universiteit Brussel) according to manufacturer's instructions and with standard sample pretreatment (15 min Target Retrieval and 30 min Protease Plus). Subsequently, sections were stained for VDR using a rabbit monoclonal antibody (D2K6W #12550, Cell Signaling, Danvers, MA, USA) diluted 1:50 in Antibody Diluent (Agilent, Santa Clara, CA, USA) and applied for 20 min at RT. For detection, a donkey anti-rabbit Alexa Fluor 488-labeled F(ab’)₂ secondary antibody (Jackson ImmunoResearch Europe, Cambridgeshire, UK) was used, diluted 1:500 in PBS and applied for 30 min at RT. Nuclei were counterstained with Hoechst 33342. Images were taken with Zeiss LSM710 confocal microscope (Zeiss, Oberkochen, Germany). VDR and LGR5 colors were changed using imagej software for a better visualization (VDR in red and LGR5 in green). Images of at least 25 crypts/patient from four patients were taken in a confocal microscope. Quantification of VDR and LGR5 staining was performed manually by two people, considering the percentage of positive cells from a total of 46 cells of the crypt bottom per crypt.

Western blot

Normal (passages 1–5) and tumor (passages 3–8) organoids were pelleted after removing Matrigel with Cell Recovery Solution and whole-cell extracts were prepared using RIPA buffer (0.05 m Tris/HCl pH 7.5, 0.1% SDS, 0.15 m NaCl, 1% Triton X-100 (Sigma-Aldrich) and 1% sodium deoxycholate) containing protease and phosphatase inhibitors (10 μg·mL⁻¹ leupeptin, 10 μg·mL⁻¹ aprotinin, 1 mm PMSF, 1 mm orthovanadate, and 1 mm NaF, all from Sigma-Aldrich). Cell extracts were separated by SDS/PAGE, transferred to PVDF membranes and incubated with the following primary antibodies: rabbit monoclonal-VDR (D2K6W) (Cell Signaling, #12550), rabbit polyclonal-CA2 (CUSABIO, #PA004370HA01HU, TX, USA), rabbit polyclonal-CYP24A1 (Santa Cruz Biotechnology, #sc-66851, Dallas, TX, USA) and goat polyclonal-β-actin (Santa Cruz Biotechnologies, #sc-1616). Different exposure times of the films were used to ensure that bands were not saturated. ImageJ was used for the semi-quantification of VDR protein level.

Electron microscopy

For electron microscopy analysis, samples from three human normal (patients #11, #12, and #37; passages 2–11) and tumor organoid cultures (patients #4, #29, and #38; passages 10–21) treated with 100 nm calcitriol or vehicle for 96 h (n = 3 per group) were fixed with 3% glutaraldehyde (Merck-Millipore, Burlington, MA, USA) in 0.12 m phosphate buffer, pH 7.4. Samples were then rinsed in 0.12 M phosphate buffer, postfixed in 2% osmium tetroxide (Sigma-Aldrich), dehydrated in acetone and embedded in araldite (Durcupan, Sigma-Aldrich, St. Louis, MO, USA). Ultrathin sections stained with uranyl acetate and lead citrate were examined with a JEOL 201 electron microscope.

Proliferation and clonogenicity assays

Normal (passages 2–7) and tumor (passages 8–27) organoids were disaggregated with 1 mg·mL⁻¹ dispase and around 6000–7000 single cells were seeded on 20 μL Matrigel drops in 48-well culture dishes. Organoids were incubated with normal or tumor complete medium plus Y-27632 (Tocris) in the presence of 100 nm calcitriol or vehicle for 10 days. Medium and calcitriol/vehicle were replaced every other day. At day 10, normal and tumor organoids were used to study RNA expression and cell viability by estimating the amount of cellular ATP using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI, USA) following the manufacturer's instructions. To analyze the clonogenicity, pictures of drops were taken at days 0 and 10 of the cell viability assay and the number of organoids in each drop was counted using mipar software [40]. Phase-contrast images of cultured organoids were captured with a DFC550 digital camera (Leica, Wetzlar, Germany) mounted on an inverted TS100 microscope (Nikon, Tokio, Japan).

Mutational status

Mutations present in six tumor organoid cultures from CRC patients (#1–6; passages 1–3) were analysed. Briefly, Matrigel was removed using ice-cold Cell Recovery following manufacturer's instructions and DNA was extracted from pelleted organoids by incubation overnight (O/N) at 56 °C with lysis buffer [50 mm Tris/HCl pH 8.0, 100 mm EDTA pH 8, 100 mm NaCl, 1% SDS, and 20 mg·mL⁻¹ proteinase K (Merck-Millipore, MA, USA)]. Saturated NaCl buffer was added for 5 min and DNA was precipitated with isopropanol and washed twice with 70% ethanol. DNA was genotyped by sequencing the amplified product of a multiplexed-PCR reaction (Amplicon sequencing) using a proof-reading polymerase. Indexed libraries were pooled and loaded onto a MiSeq instrument (Illumina). Initial alignment was performed with BWA after primer sequence clipping and variant calling was done with the GATK Unified Genotyper and VarScan2 followed by ANNOVAR annotation. Mutations were called at a minimum 3% allele frequency. SNPs were filtered out with dbSNP and 1000 genome datasets. All detected variants were manually revised and confirmed.

Real-time quantitative PCR

Matrigel-embedded normal and tumor organoids (passages 1–6) were washed twice in PBS, lysed with TRIZOL (Thermo Fisher Scientific) and total RNA was purified using the NucleoSpin miRNA extraction kit (Machery-Nagel, Düren, Germany). For cDNA retrotranscription iScript cDNA Synthesis kit, (Bio-Rad, Hercules, CA, USA) was used. RT-qPCR analyses were performed with the Taqman^® Universal PCR Master Mix (Applied Biosystems, Waltham, CA, USA) using the following FAM-labeled TaqMan probes (Applied Biosystems): ALDH3A1 (Hs00964880_m1), AXIN2 (Hs00610344_m1), BCAS1 (Hs00180227_m1), CA2 (Hs01070108_m1), CCND1 (Hs00765553_m1), CEACAM7 (Hs00185152_m1), CYP24A1 (Hs00167999_m1), DKK1 (Hs00183740_m1), DKK4 (Hs00205290_m1), GRK5 (Hs00992173), HMGCS2 (Hs00985427_m1), IFITM1 (Hs01652522_g1), IFITM2 (Hs00829485_sH), JSRP1 (Hs00376079_m1), KLF4 (Hs00358836_m1), KRT20 (Hs00300643_m1), LGR5 (Hs00173664_m1), LRIG1 (Hs00394267_m1), MSI1 (Hs01045894_m1), MUC2 (Hs03005103_g1), MYC (Hs00153408_m1), PPP1R14D (Hs00214613_m1), RARRES1 (Hs00161204_m1), S100P (Hs00195584_m1), SERPINE1 (Hs01126606_m1), SLC2A1 (Hs00892681_m1), SMOC2 (Hs00405777_m1), SPDEF (Hs01026050_m1), TFF2 (Hs00193719_m1), and TNS4 (Hs00262662_m1). RNA expression values were normalized vs. the VIC-labeled TaqMan probe for housekeeping gene large ribosomal protein (RPLPO) (H99999902_m1) using the comparative C_T method. Additionally, we used Power SYBR^® Green PCR Master Mix (Applied Biosystems) and the following primers to VDR (forward 5′ AACGCTGTGTGGACATCGGC-3′; reverse 5′-GTCATGGCTTTCGTTGGACT-3′), MEX3A (forward 5′-CAAGCTCTGCGCTCTCTACA-3′; reverse 5′-ATGAACACTGGTTCCTCGCC-3′) and PTK7 (forward 5′-TCTGGGAGACCTCAAGCAGT-3′; reverse 5′-ATGCACAAAGCGGTTGTTGG-3′) analysis, and their values were normalized vs. the housekeeping gene Succinate Dehydrogenase Complex subunit A (SDHA) (forward 5′-TGGGAACAAGAGGGCATCTG-3′; reverse 5′-CCACCACTGCATCAAATTCATG-3′) using the comparative C_T method. All RT-qPCR were performed in a CFX384 Touch Real-Time PCR Detection System (Bio-Rad). SW480-ADH cells were used as relative value to compare VDR and CYP24A1 expression in Figs 2A,B and 4D. To study the mRNA levels of other genes, organoids treated with vehicle were used as control.

RNA-sequencing

Matched normal and tumor organoids from six patients (#1–6; passages 1–3) were seeded in 48-well culture dishes and 48 h later were treated with 100 nm calcitriol (Sigma-Aldrich) or vehicle for 96 h. Medium and calcitriol/vehicle was replaced every other day. Total RNA was extracted and RNA integrity was confirmed as optimal for all samples on an Agilent 2100 Bioanalyzer (RIN range 9.1–9.9). Six hundred nanograms of RNA samples was used for library construction. PolyA⁺ fraction was purified and randomly fragmented, converted to double stranded cDNA and processed through subsequent enzymatic treatments of end-repair, dA-tailing and ligation to adapters as in Illumina's TruSeq Stranded mRNA Sample Preparation kit (Illumina, CA, USA). Adapter-ligated library was completed by PCR with Illumina PE primers (eight cycles). The resulting purified cDNA library was applied to an Illumina flow cell for cluster generation and sequenced on an Ilumina HiSeq2000 by following manufacturer's protocols. Eight samples were combined per sequencing lane and a minimum of 25 million 50 base single-reads were generated for each sample.

Sequencing reads were aligned to the transcriptome with TopHat2 [41]. Novel transcript discovery was not attempted. TopHat was provided with known gene annotations and other transcript data obtained from Gencode (Version 26 – Ensembl 75) basic gene set for the GRCh37/hg19 human genome assembly [42]. Gene expression level was calculated from TopHat alignments as the number of reads per gene computed using HTSeq [43] using default settings and gene features as defined in the GRCh37.75 release of the human genome (gtf file). Differential gene expression analysis was performed with the Bioconductor [44] DEseq2 package for the r statistical Software [45]. Single enrichment analysis was performed with differentially expressed genes between calcitriol vs. vehicle treatment. The analysis was carried out with the functional annotation tool included in the PathVisio bioinformatic resources [46, 47].

Chromatin immunoprecipitation-sequencing

We performed three independent ChIP-seq experiments with human normal organoids (two culture organoids/experiment: ChIP #1, patients #34&35 at passages 12 and 6; ChIP #2, patients #18&19 at passages 3 and 4; and ChIP #3, patients #38&39 at passages 2 and 4) following the ChIP-seq Millipore protocol with modifications. Briefly, normal organoids from two patients cultured on 6-well dishes were treated with 100 nm calcitriol or vehicle for 2 h. DNA-protein crosslinking was done by incubation of Matrigel-embedded organoids with 1% formaldehyde (methanol free) for 8 min shaking at 4 °C. Crosslinking was quenched with 1.25 mm glycine in PBS for 5 min. We transferred the organoid solution to a 50 mL conical tube for centrifugation at 700 g for 5 min at 4 °C. Supernatant was removed and the pellet was washed twice in cold PBS, transferred to a 1.5 mL tube and lysed in 1 mL ChIP-lysis buffer [1% SDS, 10 mm EDTA, 50 mm Tris/HCl pH 8.1 and 1x protease cocktail inhibitors (Complete, EDTA-free, Sigma-Aldrich)] for 1 h. We used a 21G syringe to homogenize the solution and force the lysis. Samples were sonicated for 20 min at maximum intensity at 4 °C in a S220 Focused-ultrasonicator (Covaris). We took 40 μL of calcitriol and vehicle-samples as input control, and diluted 10-fold the remaining sonicated cell supernatant in ChIP-dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mm EDTA, 167 mm NaCl, Tris/HCl pH 8.1 and 1x protease cocktail inhibitor). After 45 min precleaning incubation using magnetic beads (Magna ChIP^™ Protein A+G Magnetic Beads, Merck-Millipore), the supernatant was incubated with 10 μg rabbit monoclonal-VDR (D2K6W) antibody (Cell Signaling #12550, 2 mg·mL⁻¹, made to order) O/N in orbital rotation at 4 °C. Immunocomplexes were recovered using magnetic beads for 2 h in orbital rotation at 4 °C. Immunoprecipitated material was then washed in the following buffers in the indicated order: ChIP-dilution buffer, ChIP-low salt buffer (0.1% SDS, 1% Triton X-100, 2 mm EDTA, 150 mm NaCl, 20 mm Tris/HCl pH 8.0 and 1x protease cocktail inhibitor), ChIP-high salt buffer (0.1% SDS, 1% Triton X-100, 2 mm EDTA, 500 mm NaCl, 20 mm Tris/HCl pH 8 and 1x protease cocktail inhibitor), ChIP-LiCl buffer (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mm EDTA, 10 mm Tris/HCl pH 8.0 and 1x protease cocktail inhibitor), ChIP-TE buffer (two washes) and finally eluted in freshly prepared ChIP-elution buffer (1% SDS, 0.1 M NaHCO₃). The eluted solution and input controls were incubated with 0.2 M NaCl at 65 °C ON in order to de-crosslink the DNA complexes, and subsequently we added 9.5 mm EDTA, 36.5 mm Tris/HCl pH 6.5 and 70 μg·mL⁻¹ proteinase K for 1 h at 45 °C. DNA extraction was performed using the QIAquick^® PCR purification kit (Qiagen, Venlo, The Netherlands).

Fragmented DNA were processed through subsequent enzymatic treatments of end-repair, dA-tailing, and ligation to adapters with NEBNext Ultra II DNA Library Prep Kit for Illumina (New England BioLabs, Ipswich, MA, USA). Adapter-ligated libraries were completed by limited-cycle PCR and extracted with a (single) double-sided SPRI size selection. Median fragment size is 340 bp from which 120 bp correspond to adaptor sequences. Libraries were applied to an Illumina flow cell for cluster generation and sequenced on an Illumina instrument by following manufacturer's protocols. Sequencing reads were aligned to the GRCh37/hg19 genome assembly with Bowtie2 [48]. Enriched peak calling was performed with MACS2 [49] callpeak function, with –q-value 0.15, –nomodel and –extsize 200 options using as control concatenated same-patient alignment files from input treated (calcitriol) and untreated (vehicle). MACS2 bdgdiff function was used to obtain statistically enriched DNA regions using the pileup data obtained from the peak calling. Peaks annotation was performed using custom parameters in order to assign the peaks to the nearest gene in a range of 10 kb upstream/downstream from the gene body. Manual inspection of data and RNA-seq identified genes of interest was performed using integrative genome visualization (igv) Tool v.2.3.79 [50, 51].

Functional enrichment analysis

We used Gene Set Enrichment Analysis (GSEA) [52] to assess the degree of association between our RNA-seq profile (genes regulated by calcitriol in human colon normal/tumor organoids), or microarray expression (genes regulated by calcitriol in mouse colon normal organoids), with other signatures defined with genes enriched (differentiation gene set: EPHB2^low vs. EPHB2^high human normal colon epithelial cells [12]; proliferation gene sets: E2F, MYC and mTORC1 hallmark gene sets from MSigDB; pro-tumorigenic gene set: Core ESC-like from MSigDB; slow-dividing mouse intestinal stem cells gen set: Mex3a^high/Lgr5^high mouse colon cells [22] and VDR ChIP-seq in human colon normal organoids, Table S7). The over-representation of transcription factor binding sites in genes regulated by calcitriol was computed with the Bioconductor TFEA.ChIP package.

Statistical analysis

Statistical analysis was performed using GraphPad Prism (GraphPad, CA, USA). P-values < 0.05 were considered significant and * indicates P < 0.05, **P < 0.01, and ***P < 0.001. The exact P-values are indicated for nonsignificant results. Box plots shown in Figs 2A,B, 4A,C and 6A represent median ± max/min. Graphs showing the quantification of VDR and LGR5 positive cells in human crypts (Fig. 1B) represent mean ± standard error of the mean (SEM). Graphs showing the RNA levels in VDR interference (Fig. S3), and relative VDR RNA levels of normal and tumor organoids (Fig. 4D) represent mean ± standard deviation (SD).