Biopsy-derived intestinal epithelial cell cultures for pathway based stratification of patients with inflammatory bowel disease

Abstract

Background

Endoplasmic reticulum stress was shown to be pivotal in the pathogenesis of inflammatory bowel disease. Despite progress in IBD drug development, not more than one third of patients achieve steroid-free remission and mucosal healing with current therapies. Furthermore, patient stratification tools for therapy selection are lacking. We aimed to identify and quantify epithelial ER stress in a patient-specific manner in an attempt towards personalized therapy.

Methods

A biopsy-derived intestinal epithelial cell culture system was developed and characterized. ER stress was induced by thapsigargin and quantified with a BiP ELISA on cell lysates from 35 patients with known genotypes who were grouped based on the number of IBD-associated ER stress and autophagy risk alleles.

Results

The epithelial character of the cells was confirmed by E-cadherin, ZO-1 and MUC2 staining and CK-18, CK-20 and LGR5 gene expression. Patients with 3 risk alleles had higher median epithelial BiP-induction (vs. untreated) levels compared to patients with 1 or 2 risk alleles (p=0.026 and 0.043, respectively). When autophagy risk alleles were included and patients were stratified in genetic risk quartiles, patients in Q2, Q3 and Q4 had significantly higher ER stress (BiP) when compared to Q1 (p=0.034, 0.040 and 0.034, respectively).

Conclusion

We developed and validated an ex vivo intestinal epithelial cell culture system and showed that patients with more ER stress and autophagy risk alleles have augmented epithelial ER stress responses. We thus presented a personalized approach whereby patient-specific defects can be identified which in turn could help in selecting tailored therapies.

2. Introduction

Inflammatory bowel diseases (IBD) comprise a spectrum of intestinal inflammatory conditions with Crohn’s disease (CD) and ulcerative colitis (UC) as the two main entities. They are characterized by chronic inflammation of the gastrointestinal tract and are believed to result from a dysregulated immune response towards the intestinal microbiota in genetically predisposed individuals¹. Physicians and patients still face multiple challenges as no curative treatment yet exists. A significant advance in management of IBD was the introduction of biologic agents targeting tumor necrosis factor (TNF)². Almost two decades following their approval, a second and third class of biologic agents respectively targeting gut-homing α₄β₇ integrin-positive T-lymphocytes and interleukins (IL)-12/23, have been added to the therapeutic options. All approved biologic agents suppress a general adaptive immune response instead of the desired targeting of underlying pathogenic triggers^3–5. Second, steroid-free clinical remission and mucosal healing, two important treatment goals, are observed in no more than 30-35% of patients with large inter-patient variability in treatment response. The search for predictive biomarkers has been unsuccessful so far and as a consequence prediction of therapeutic success is poor⁶.

Genetic association and gene/protein expression studies have uncovered novel underlying pathophysiologic pathways that are currently under (pre)clinical evaluation as a therapeutic target. IL-12/23 and Smad7/TGF-B signaling, endoplasmic reticulum (ER) stress and autophagy are accepted key players in IBD pathogenesis and are targeted by specific small molecules or antibodies that are in different stages of therapeutic development^7–18. It is anticipated that treatment success will vary depending on which pathways drive disease in a given patient. Therefore, it will become increasingly important to identify triggers of disease in order to select the most appropriate therapy ^{3, 6, 19}.

The intestinal epithelium is crucial for intestinal homeostasis and prevention of inflammation as this tightly connected single cell layer limits translocation of luminal microorganisms and other antigens to the lamina propria. Intestinal epithelial cells (IECs) form a physical barrier that is maintained by strong tight junction protein expression and a continuous epithelial cell proliferation in the stem cell compartment at the crypt base^20–22. The role of the epithelial barrier in IBD is underscored by studies that associate barrier defects with disease progression and relapse^23–25. As described above, several epithelial cell integrity pathways such as ER stress and autophagy have been associated with IBD^{21, 22, 26–28}.

ER stress signaling/unfolded protein response and autophagy are two well-characterized homeostatic pathways that closely collaborate and play a key role in the innate and adaptive immune system^29–32. Autophagy serves as an intracellular clearance mechanism for components of endogenous and exogenous origin such as mitochondria, misfolded proteins, signaling complexes and (pathogenic) microorganisms³⁰. The unfolded protein response is triggered by an increased abundance of un- or misfolded proteins in the ER, also called ER stress. Prolonged or uncontrolled ER stress will eventually lead to inflammation and/or cell death²⁹. Both pathways have been genetically associated with IBD; e.g. the ER stress-related XBP1 and ORMDL3 genes and the autophagy-related ATG16L1, IRGM, LRRK2 and ULK1 genes^26–28. Nevertheless, little is known how these genetic variants functionally translate in patients with IBD. We hypothesize that inter-patient differences in the risk allele burden in these pathways will lead to distinctive functional readouts in IBD patient-derived epithelial cells.

Therefore, the aim of this study was to translate an individual’s genetic risk in ER stress and autophagy into quantifiable, functional ER stress-readouts starting from patient-derived epithelial cells. To do so, we developed an ex vivo 2-dimensional epithelial cell culture system starting from human endoscopically-derived biopsies to quantify perturbed pathways in patients with IBD. As a proof of concept, we studied epithelial ER stress levels stratified by the number of ER stress and autophagy risk alleles.

3. Materials and methods

3.1. Patients and ethical statement

Patients with IBD followed at the IBD unit of the University Hospitals Leuven, who were genotyped as part of the international Immunochip project, were selected based on their mutations in ER stress or autophagy genes (Table 1, Supplementary figure 1)^{26, 33}. We selected patients with 0, 1, 2 or 3 ER stress risk alleles and patients who had ≤3, 4, 5 or ≥6 autophagy risk alleles which was based on the risk allele distribution in the immunochipped IBD patient population at the University Hospitals Leuven (Supplementary figure 1). Only SNPs with a call rate > 90% and a minor allele frequency > 0.01 were included. Mucosal biopsies (8/patient) were obtained from the macroscopically non-inflamed colon of 35 patients undergoing endoscopy as part of their IBD management. Patient characteristics are provided in table 2.

Table 1. IBD-associated ER stress and autophagy genes with their specific SNP-IDs, chromosome n°, risk alleles and SNP locations.

Pathway	Gene	Studied SNP	Chromosome	Risk allele	Location
Autophagy	ATG16L1	rs2241880	2	G	non-synonymous coding
	IRGM	rs10065172	5	T	exonic, synonymous coding
		rs4958847		A	intronic
	ULK1	rs12303764	12	T
	LRRK2	rs11175593		T
	MTMR3	rs2412973	22	A	downstream of gene
ER stress	ORMDL3	rs2872507	17	A	upstream of gene
	XBP1	rs35873774	22	C	intronic

(ATG16L1: Autophagy Related 16 Like 1; IRGM: Immunity Related GTPase M; ULK1: Unc-51 Like Autophagy Activating Kinase 1; LRRK2: Leucine Rich Repeat Kinase 2; MTMR3: Myotubularin Related Protein 3; ORMDL3: ORMDL Sphingolipid Biosynthesis Regulator 3; XBP1: X-Box Binding Protein 1.)

Table 2. Patient characteristics.

# Patients	35
Female [%]	20 [57]
Median [IQR] age (yrs)	53 [42-57]
Median [IQR] age at diagnosis (yrs)	27 [21.5-36.5]
Median [IQR] disease duration (yrs)	20 [10.5-28.5]
UC/CD [%]	4/31 [11/89]
Prior abdominal surgery [%]	21 [60]
Smoking [%]
Yes	12 [34]
Former	5 [14]
No	11 [31]
Unknown	7 [20]
Therapy [%]
Antibiotics	2 [6]
Corticosteroids	2 [6]
Thiopurines/methotrexate	4 [11]
Anti-tumor necrosis factor	14 [40]
Vedolizumab	1 [3]

(IQR: interquartile range; yrs: years)

Ethical approval was given by the Ethics Board of the University Hospitals Leuven (B322201213950/S53684) and all patients provided written informed consent.

3.2. Isolation and culturing of IECs

The crypt isolation protocol and cell culture medium were adapted from the colonic organoid culture procedure which was developed by Sato et al.in 2011³⁴. Biopsies were immediately placed in DMEM-F12 (Lonza, Basel, Switserland) containing glutamine (15 mM), hepes (15 mM) and penicillin/streptomycin (100 U/ml, Lonza, Basel, Switserland) at 4 °C for transport (on ice) to the research lab and stored (at 4 °C) for up to two hours until epithelial isolation was performed. First, the biopsies were washed in DMEM-F12 after which they were allowed to settle and the supernatant was discarded. Next, they were thoroughly washed with complete chelating solution (CCS, 0.996 g/l Na₂HPO₄*2H₂O, 1.08 g/l KH₂PO₄, 5.6 g/l NaCl, 0.12 g/l KCl, 15 g/l Saccharose, 10 g/l D-sorbitol, 80 mg/l DTT) by repeated pipetting. Finally, EDTA (10 mM, Thermo Scientific, Waltham, Massachusetts, USA) was added and the biopsy suspension was placed on a rocking platform at 4 °C for 45 minutes after which the EDTA containing solution was removed. The biopsy fragments were passed multiple times through a 10 ml pipette in CCS to mechanically disrupt the IECs from the remaining mucosal tissue leaving them in suspension as the remaining fragments settled down. The supernatant was centrifuged after which the cell pellet was washed one more time in DMEM-F12 before resuspending the IECs in expansion medium (composition see Supplementary table 1) and plating the cells in collagen coated 12-well plates (7 wells/8 biopsies) in a humidified incubator at (37 °C, 5% CO₂). The medium was replaced for the first time after 24 hours and every 48 hours thereafter.

3.3. Immunocytochemistry

To further characterize these intestinal biopsy-derived cell cultures, several immunocytochemic stainings were performed. Isolated crypts were seeded on collagen coated CC2 Lab-Tek chamber well slides (Thermo Scientific, Waltham, Massachusetts, USA). At day 4, cells were washed with PBS and subsequently fixated in paraformaldehyde (PFA, 4%, 20’, 37 °C)cells were washed again in PBS and permeabilized with methanol (10’, RT). After washing, the cells were incubated in glycine (0.1 M, 2x10’, RT) followed by washing and blocking (10 % FBS, 1% BSA, 1h, RT). The primary anti-E-cadherin antibody (ab1416, Abcam, Cambridge, UK) and the anti-platelet derived growth factor receptor-α (PDGFR-α) antibody (sc-338, Santa Cruz Biotechnology, Dallas, Texas, USA) were diluted in 1% BSA and cells were exposed for 1 hour. After washing in 1% BSA, the cells were incubated with the secondary goat anti-mouse antibody (Alexa Fluor 488, Thermo Scientific, Waltham, Massachusetts, USA) and goat anti-rabbit antibody (Alexa Fluor 594, Thermo Scientific, Waltham, Massachusetts, USA) for 1 hour followed by DAPI staining (1 µg/ml). After a final wash step in PBS-T, the cells were mounted with ProLong gold anti fade reagent (Life Technologies, Carlsbad, California, USA). Images were obtained using a BX41 microscope (Olympus, Tokio, Japan) and analyzed with the Scan R software (Olympus, Tokio, Japan). For the immunocytochemic staining of ZO-1, KI67 and MUC2 we used a slightly different protocol provided by the Gastrointestinal Motility and Sensitivity Research Group from KU Leuven. Washed cells (on Lab-Tek chamber slides) were fixated in PFA (4%, 30’, RT) and rehydrated in 100% (3x3’) and 70% (1x3’) ethanol followed by H₂O (2x3’). Antigen retrieval was performed in sodium citrate buffer at 120 °C for 10’ after which the cells were allowed to slowly cool to room temperature and were washed with PBS. Blocking was performed for 10’ with Protein Block Serum-Free (Agilent, Santa Clara, CA, USA) for 10’ after which they were incubated overnight at 4 °C with the primary antibodies: mouse-anti-ZO-1 (1/50, 339100, Thermo Scientific, Waltham, Massachusetts, USA), mouse anti-KI67 (1/150, MONX10283, Sanbio, Uden, The Netherlands), rabbit-anti-MUC2 (1/150, sc-15334, Santa Cruz Biotechnology, Dallas, Texas, USA) diluted in Antibody Diluent (Agilent, Santa Clara, CA, USA). After washing in PBS, the cells were incubated with the secondary goat anti-mouse antibody (Alexa Fluor 488, Thermo Scientific, Waltham, Massachusetts, USA), donkey anti-rabbit antibody (Alexa Fluor 594, Thermo Scientific, Waltham, Massachusetts, USA) and DAPI (all 1/1000) for 30’. After washing, the cells were mounted with citifluor™ (VWR, Radnor, PA, USA). Normal fluorescent images were obtained with the a BX41 microscope (Olympus, Tokio, Japan) whereas a LSM 880 microscope with Airyscan (Zeiss, Oberkochen, Germany) was used to obtain high resolution z-stack images. Analysis was performed with ZEN Blue software (Zeiss, Oberkochen, Germany) and Fiji (ImageJ, NIH, Bethesda, MD, USA).

3.4. Quantitative real-time polymerase chain reaction (qRT-PCR) analysis

Expression levels of epithelial marker genes cytokeratin-18 and -20 (CK-18 and CK-20) were determined at 24, 72, 120 and 168 hours post isolation and compared with expression levels in the fetal human colon (FHC) cell line (positive control) and the IMR-90 lung fibroblast cell line (negative control). Furthermore, the expression of the intestinal epithelial stem cell marker LGR5 was measured at the same time points. Finally baseline expression of the GRP78/BIP gene was measured in 6-day-old untreated IECs from 19/35 patients who were included in this study.

Cells were washed with PBS and mechanically removed in RNAlater using a cell scraper. After centrifugation the pellet was resuspended in RLT-buffer (Qiagen, Hilden, Germany) containing β-mercapto-ethanol and passed repeatedly through a 29G needle or Qiashredder tubes (Qiagen, Hilden, Germany). An equal volume of 70% ethanol was added and mRNA was extracted from this mixture with the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to manufacturer’s protocol. RNA quality was assessed with the NanoDrop 1000 spectrophotometer (Thermo Scientific, Waltham, Massachusetts, USA) and samples were stored at -80 °C until cDNA synthesis with the Qscript cDNA supermix (Quantabio, Beverly, MA, USA) according to manufacturer’s protocol.

The primers used for qRT-PCR analysis are given in supplementary table 2 and 10 µM stock solutions were used to make the reaction mixture (5 µl SybrGreen, 0.2 µM FW & RV primer, 2.5 µl cDNA sample, 2.3 µl RNAse-free H₂O). Samples were analyzed with the Lightcycler 480 (Roche, Basel, Switzerland) and the following amplification program was used: 5’ 95 °C, 45x (10” 95 °C, 15” 60 °C, 15” 72 °C), 5” 95°C, 1’ 60 °C, 4 °C. CK-18, CK-20,LGR5 and BIP/GRP78 mRNA-levels were normalized to the housekeeping gene β-actin and quantified using the comparative (Δ or ΔΔ) C_t method.

3.5. ER stress induction

IEC cultures were treated for 14 hours (from the end of day 5 until the beginning of day 6) with the ER stress inducing compound thapsigargin (0.4 µM, Sigma-Aldrich, Saint Louis, Missouri, USA) in order to enhance potential inter-patient differences.Lysis and total protein measurement

After 14 hours of thapsigargin-treatment, the IECs were placed on ice, the medium was aspirated and the cells were rinsed with ice cold PBS. Next, the IECs were scraped in PBS and spun down, the resulting cell pellet was resuspended in RIPA lysis buffer (Enzo Life Sciences, Farmingdale, New York, USA) and lysis was performed by incubating the suspension 45’ on ice followed by sonication. The cell lysate was spun down at maximal speed to pellet membrane fragments and the supernatant was used to determine the total protein content with the Pierce^® BCA Protein Assay Kit (Thermo Scientific, Waltham, Massachusetts, USA), the remainder was stored at -80 °C until enough samples were acquired to perform an ELISA.

3.6. Binding immunoglobulin protein (BiP)/Glucose-regulated protein 78 (GRP78) ELISA

To quantify the ability of IECs to cope with ER stress, we measured BiP (also known as GRP78 or HSPA5) levels before and after thapsigargin-treatment with a competitive BiP ELISA kit (Enzo Life Sciences, Farmingdale, New York, USA) according to manufacturer’s protocol. Plates were read with the Fluostar Omega microplate reader (BMG Labtech, Ortenberg, Germany) and quantified with a 5-PL logistic regression script in Microsoft Office Excell. ER stress induction rates were expressed as the BiP ratio between treated and untreated IECs ([BiP]_{thapsigargin-treated}/[BiP]_untreated).

3.7. Statistical analysis

As data were not normally distributed, non-parametric tests were used, no multiple testing was performed. The BiP (ELISA) levels and ratios were compared between the different groups using a Mann-Whitney test with Graphpad Prism Software (La Jolla, California, USA). A p-value < 0.05 was considered significant.

4. Results

4.1. IECs and epithelial characterization

After isolation, the intestinal crypts retained their three-dimensional morphology while being suspended in the medium (Figure 1A). During overnight incubation, the crypts sunk and attached to the collagen-coated surface to form an epithelial monolayer. These two-dimensional IEC-islands consisted of cuboidal cells, giving them a pavement-like appearance. Visually each IEC-island was formed around a growth center (Figure 1B, red arrow), suggesting these cells originated from proliferating intestinal stem cells originally located in the bottom of the crypts in vivo. Cells in this center divided and gradually pushed away earlier formed cells resembling the in vivo situation. These observations were confirmed by immunofluorescent stainings for the proliferation marker KI67 (Figure 2). Furthermore, as cells moved away from these areas of proliferation, a significant portion differentiated cells shows presence of MUC2-positive cells with a goblet cell-like morphology (Figure 2, separate fluorescence channel images are provided in supplementary figure 2).

Figure 1.

Brightfield (BF) microscopic image of freshly isolated colonic crypts with intact crypt architecture (A); BF image of an IEC-island with growth center (blue arrow) in a collagen coated well 48 hours post isolation (B); BF image of a 12-day-old IEC culture with typical areas of cell death and detachment (orange arrows) (C).

Figure 2.

Immunocytofluorescent staining for Ki67 (green) and Muc2 (red) in a 4-day-old IEC culture (20x magnification, blue: DAPI staining).

Cells remained viable in culture for approximately 12 days, after which local cell detachment and cell death occurred (Figure 1C, red arrows). The combined isolation and culture success rate was 81% (39 out of 48 isolations). Failure was mostly due to low donor-dependent IEC isolation yields or inefficient cell attachment but was independent of presence of ER stress or autophagy risk alleles.

In order to confirm the epithelial character, 4-day-old IEC cultures were stained for E-cadherin ^{35, 36}. Figure 3A and 3B illustrate how this epithelial transmembrane adherens junction protein was strongly expressed along the cell membranes of the cultured IECs. In order to rule out fibroblast contamination, we also performed a PDGFR-α staining which was negative in 4-days old IEC cultures. A positive control staining for this latter antibody on IMR-90 fibroblast cells is provided in supplementary figure 3.

Figure 3.

Immunocytofluorescent staining for E-cadherin (green) and PDGFR-α (red) in a 4-day-old IEC culture

The epithelial character of these ex vivo cell cultures was further assessed by measuring mRNA levels of CK-18 and CK-20 over time, as indicated in figure 4. In cultured IECs, CK-18 was stably expressed over time (up to 168h), whereas CK-18 mRNA could also be detected (in lower amounts) in FHCs but not in IMR-90 cells (Figure 4A). Figure 4B illustrates how CK-20 is initially expressed at high levels and a gradually decreases over time. Still, also at day 7 (168 hpi), expression levels were strongly present when compared to proliferating cultures of FHCs. Additionally, we also measured the expression of the intestinal epithelial stem cell marker LGR5 in IEC cultures at the same time points (24, 72, 120, 168 hpi) and showed that LGR5 expression increased over time (Figure 5).

Figure 4.

Cytokeratin-18 (A) and Cytokeratin-20 (B) mRNA expression in IEC cultures at 24, 72, 120 and 168 hours post isolation (hpi) and in the FHC and IMR-90 cell lines (ΔΔC_t-method, fold change to expression levels in FHCs and all normalized to β-actin mRNA).

Figure 5.

Lgr5 mRNA expression in IEC cultures at 24, 72, 120 and 168 hours post isolation (hpi). (ΔΔC_t-method, fold change to expression levels at 24hpi and all normalized to β-actin mRNA).

Finally, we assessed the polarity along the apical-basolateral axis, by staining the cells for zonula occludens-1 (ZO-1). We could show that ZO-1-postive signal (green) is distributed apically at a depth of 0-1500 nm, whereas this positive signal disappears completely when moving closer towards the basolateral side (Figure 6 and 7).

Figure 6.

z-stack images at three different depths (300, 1500 and 3000 nm from the apical border) of a 4-day-old IEC culture stained for ZO-1 (green). (blue: DAPI staining)

Figure 7.

Cross sectional view of the reconstructed z-stack images from figure 6.

4.2. Genetic risk in ER stress and autophagy genes and the epithelial ER stress response

The IEC cultures were microscopically monitored daily between the time of isolation and lysis in order to exclude wells that had an aberrant morphology or showed signs of cell death. In order to determine the cells’ intrinsic capability to cope with ER stress, we measured BiP-levels with and without a 14 h treatment with the ER stressor thapsigargin from day 5 until day 6.

Patients were then grouped according to the number of ER stress risk alleles in XBP1 (rs35873774) and ORMDL3 (rs2872507) (Table 1). Median [IQR] thapsigargin-mediated BiP-induction was 2.67 [1.01-6.07], 1.87 [1.50-3.16], 1.70 [1.32-2.41] and 4.48 [3.76-4.64] in IECs from patients carrying 0, 1, 2 or 3 risk alleles respectively. Notice the absence of a group with patients carrying 4 ER stress risk alleles. Because of their low prevalence (0.14 %) in our patient genotype database (Supplementary figure 1A), we were unable to include these patients as this means that we had only 4 patients with 4 ER stress risk alleles in our entire genotyped patient cohort. These specific patients did not undergo an endoscopic investigated at our clinic during the time of inclusion. IECs from patients with three risk alleles had significantly more ER stress induction rates when compared to patients with one or with two risk alleles (Figure 8A, p=0.0262 and 0.0430, respectively).

Figure 8.

Boxplot diagrams showing the thapsigargin (Tg)-induced ER stress (BiP) levels of 6-day-old colonic IEC cultures from IBD patients carrying 0 to 3 ER stress-related risk alleles (n=4, 17, 11 and 3, respectively) in XBP1 and/or ORMDL3 (A); From IBD patients carrying ≤3 (Q1, n=12), 4 (Q2, n=6), 5 (Q3, n=9) or ≥6 (Q4, n=8) autophagy-related risk alleles in ATG16L1, MTMR3, ULK1 and/or LRRK2 (B); from IBD patients carrying ≤4 (Q1, n=8), 5 (Q2, n=10), 6 (Q3, n=7) or ≥7 (Q4, n=10) ER stress and autophagy-related risk alleles in XBP1, ORMDL3, ATG16L1, MTMR3, ULK1 and/or LRRK2 (C). (*: Mann Whitney p-value < 0.05).

We also grouped patients in risk quartiles, based on the number of risk alleles (RA) in autophagy genes ATG16L1 (rs2241880), IRGM (rs10065172 and rs4958847), MTMR3 (rs2412973), LRRK2 (rs11175593) and ULK1 (rs12303764). The distribution of these risk alleles in the sampled population (Supplementary figure 1B) was used to define the number of risk alleles in each quartile: Q1 had ≤3 RA, Q2 had 4 RA, Q3 had 5 RA and Q4 had ≥6 RA. Median [IQR] thapsigargin-mediated BiP-induction was 1.58 [1.13-2.85], 1.78 [1.52-2.64], 3.57 [1.83-4.64] and 2.74 [1.60-3.59] in IECs from patients belonging to Q1 to Q4 respectively (Figure 8B). No significant differences were observed between these groups (Figure 3B) although a trend towards higher ER stress induction rates in Q3 and Q4 compared to Q1 (p = 0.0507 and 0.1535, respectively) was seen.

Finally, given that autophagy and ER stress show a clear interplay^{31, 32, 37, 38}, ER stress and autophagy risk alleles were combined. This combination of risk alleles led to a change in the definition of the genetic risk quartiles (Q1: ≤4 RA, Q2: 5 RA, Q3: 6 RA, Q4: ≥7 RA; Supplementary figure 1C). Median thapsigargin-mediated BiP-induction [IQR] was 1.34 [1.08-1.91], 2.16 [1.68-4.05], 3.60 [1.39-4.48] and 2.41 [1.61-3.27] in IECs from patients belonging to genetic risk groups Q1 to Q4, respectively (Figure 8C). Patients in Q2, Q3 and Q4 had significantly higher ER stress induction rates when compared to Q1 (p = 0.0343, 0.0401 and 0.0343, respectively).

At baseline, there were no significant differences in BiP mRNA or protein expression between the different patient subgroups (Figure 9 and 10, respectively).

Figure 9.

Boxplot diagrams showing the baseline ER stress (BiP) mRNA levels (normalized to β-actin mRNA) of 6-day-old colonic IEC cultures from IBD patients carrying 0 to 3 ER stress-related risk alleles (n=3, 9, 6 and 1, respectively in XBP1 and/or ORMDL3 (A); From IBD patients carrying ≤3 (Q1, n=4), 4 (Q2, n=4), 5 (Q3, n=5) or ≥6 (Q4, n=6) autophagy-related risk alleles in ATG16L1, MTMR3, ULK1 and/or LRRK2 (B); from IBD patients carrying ≤4 (Q1, n=3), 5 (Q2, n=5), 6 (Q3, n=3) or ≥7 (Q4, n=8) ER stress and autophagy-related risk alleles in XBP1, ORMDL3, ATG16L1, MTMR3, ULK1 and/or LRRK2 (C).

Figure 10.

Boxplot diagrams showing the baseline ER stress (BiP) protein levels of 6-day-old colonic IEC cultures from IBD patients carrying 0 to 3 ER stress-related risk alleles (n=4, 17, 11 and 3 respectively) in XBP1 and/or ORMDL3 (A); From IBD patients carrying ≤3 (Q1, n=12), 4 (Q2, n=6), 5 (Q3, n=9) or ≥6 (Q4, n=8) autophagy-related risk alleles in ATG16L1, MTMR3, ULK1 and/or LRRK2 (B); from IBD patients carrying ≤4 (Q1, n=8), 5 (Q2, n=10), 6 (Q3, n=7) or ≥7 (Q4, n=10) ER stress and autophagy-related risk alleles in XBP1, ORMDL3, ATG16L1, MTMR3, ULK1 and/or LRRK2 (C).

5. Discussion

In this study we developed a novel ex vivo two-dimensional IEC culture model allowing characterization and quantification of pathogenic pathways in IBD in a patient-specific manner. We demonstrated that these biopsy-derived epithelial cell cultures remain viable for about 12 days and isolation success was more than 80%. The epithelial character was illustrated by a clear E-cadherin staining along the membranes of IECs, which resembles immunohistochemical E-cadherin stainings on human colonic tissue sections^{39, 40}. We could not detect the fibroblast marker PDGFR-α which indicates that these cultures were free of contamination by mesenchymal cells. The areas where the crypts originally attached, remained a center of IEC proliferation as indicated by the abundance of KI67 positive cells. Daughter cells get pushed outward and either differentiate into intestinal epithelial cell types or retain their proliferative phenotype.

We also analyzed gene expression levels of two epithelial cytokeratins over time. Cytokeratin 18 is a type-1 keratin that is found in all simple epithelial tissues such as the intestinal epithelial lining and the proximal tubule of the kidney^{41, 42}. We could detect stable CK-18 mRNA levels, illustrating our monolayer cultures have an epithelial character that is not lost over time. Cytokeratin 20, on the other hand, also belongs to the type-1 keratin family and is predominantly expressed in differentiated IEC subtypes^41–43. In our IEC cultures, CK-20 mRNA levels were decreasing, suggesting loss of differentiation over time. This is further supported by the inverse correlation between the time-dependent CK-20 and LGR5 mRNA expression: LGR5 expression increases over time indicating a rise in the relative abundance of epithelial stem cells.

Finally, since polarity is an important aspect of a functional epithelial monolayer, we stained the cells for zonula occludens-1 (ZO-1), a tight junction protein which should be located at the apical side of the epithelium. We could indeed show that ZO-1 is distributed apically when compared to the nuclei.

Taken together, these data confirm that the isolated cells form polarized epithelial monolayers that contain both proliferating and differentiated cells. This model therefore shows the potential for measuring specific biologic responses in individual patients stratified on genetic susceptibility, disease location and/or therapies.

As a further proof of concept, we also showed for the first time that the genetic susceptibility in two important pathways associated with IBD, namely ER stress and autophagy, can be functionally translated and quantified in individual patients using biopsy-derived IECs. We chose to focus on these two pathways because of their functional interaction and importance for IEC homeostasis^29–31. We measured intracellular BiP-levels as a quantitative readout for the amount of ER stress. BiP or GRP78/HSPA5 is a molecular chaperone protein that is strongly involved in ER stress signaling. It is upregulated when ER stress increases (eg. after thapsigargin treatment) and controls further activation of all three branches of the unfolded protein response (the ER stress signaling pathway)^{44, 45}.

Two ER stress-related risk loci have been identified so far (rs35873774 and rs2872507) and patients carrying more than two risk alleles in this pathway were rare in our patient population. Therefore, it was impossible to further group patients into genetic risk quartiles. Hence, the highest risk group (carrying 3 RA) contained only three patients. Nevertheless, this patient-group showed a significant increase in thapsigargin-mediated ER stress (BiP) induction when compared to patients carrying two or one risk allele(s). These data illustrate a functional, quantifiable consequence of two confirmed genetic risk variants in the ER stress pathway in patients with IBD.

By clearing un- or misfolded intracellular proteins, autophagy by itself is an essential component of ER stress signaling^{31, 37}. Accumulating evidence underscores the interaction of autophagy and ER stress signaling in the intestinal epithelium^{31, 32, 38, 46}. For example, Adolph et al. showed in mice that epithelial specific genomic deletion of autophagy genes leads to increased ER stress signaling and vice versa. Both mechanisms thus seem to play compensatory roles in maintaining IEC homeostasis and preventing inflammation which is further demonstrated by the spontaneous ileitis that only occurs when both pathways are genetically perturbed⁴⁶. Since it has been clearly demonstrated that dysfunctional autophagy also leads to increased ER stress in IECs^{32, 38, 46, 47}, we tried to confirm these murine findings using our human IEC model but were unable to detect significant differences in ER stress induction rates between patients belonging to different autophagy genetic risk quartiles. However, when ER stress and autophagy risk alleles were combined, a significant association between genetic risk and ER stress induction rates was seen. This indicates that the genetic risk in both pathways should be taken into account when looking at the functional level. Finally, we could show that none of these patient subgroups showed significant baseline differences in the expression of BiP both at the mRNA and protein level. Therefore, these results suggest that it mainly the ability to cope with ER stress-inducing insult (eg. Thapsigargin) is affected rather than the baseline ER stress levels in stress-free conditions.

Our findings do not only show the functionality of this new ex vivo IEC culture system, they also demonstrate that disease-associated molecular pathways can be quantified in an individual patient. This may provide therapeutic opportunities such as the administration of ER stress reducing molecules in patients demonstrating increased ER stress levels in IECs. Despite the fact that ER stress is regarded a key player in the pathogenesis of IBD, it is currently not being considered as possible therapeutic strategy. Yet, the ER stress reducing conjugated bile acid tauroursodeoxycholic acid (TUDCA) may reduce epithelial apoptosis and inflammation and was shown to reduce severity of colitis in multiple IBD mouse models^15–18. Furthermore, oral administration of TUDCA in the context of other diseases has not been associated with any adverse events so far ^48–51. It would therefore be very interesting to study if TUDCA could reduce inflammation in patients with IBD characterized by increased ER stress levels, as demonstrated in our human culture model. Besides TUDCA, other ER stress reducing compounds such as 4-phenylbutyrate (PBA) and glutamine could also be considered as these compounds also have shown some effectiveness in murine IBD models ^13–15.

Likewise, the autophagic inducer rapamycin was effective in IBD case reports but failed to show efficacy in a randomized placebo controlled trial^9–12. We wonder if functional characterization of the patients randomized in this study for defects in autophagy would shed a different light on the results.

Since these cells are grown in two dimensions, the apical side is easily accessible for pharmaceutical compounds or micro-organisms which is a great advantage compared to the 3-dimensional intestinal organoid model originally described by Sato et al³⁴. Intestinal organoids are an excellent model to investigate multiple key aspects of intestinal epithelial physiology and pathologies such as epithelial stem cell proliferation studies. However, our ex vivo monolayer protocol may offer several practical advantages and an easier to use system for exposure studies.

This ex vivo IEC culture system may be used or modified for other applications than the investigation of IBD-associated genetic defects at the site of the intestinal epithelium. Epithelial defects in other diseases like celiac disease, post-infectious irritable bowel syndrome and intestinal cancer could be further elucidated and lead to more personalized therapeutic approaches. Another potential application of the ex vivo cell culture system is personalized drug toxicity screening assays. We are currently further modifying our protocol allowing the IECs to grow on transwell membranes in order to perform permeability assays. This setup could also be used for co-culturing IECs with other relevant intestinal cell types such as macrophages.

In summary, we have developed and characterized a 2-dimensional IEC culture system that allows easy exploration of patient-specific epithelial defects and/or responses. We could detect defects in epithelial ER stress handling in genetically predisposed patients and hereby show that this approach can be used for the detection and quantification of underlying pathogenic mechanisms. Personalized tools such as this will become highly valuable in complex disorders and will allow treatment of a defective pathway instead of a disease phenotype.