Formation of Human Colonic Crypt Array by Application of Chemical Gradients Across a Shaped Epithelial Monolayer

Abstract

Background & Aims

The successful culture of intestinal organoids has greatly enhanced our understanding of intestinal stem cell physiology and enabled the generation of novel intestinal disease models. Although of tremendous value, intestinal organoid culture systems have not yet fully recapitulated the anatomy or physiology of the in vivo intestinal epithelium. The aim of this work was to re-create an intestinal epithelium with a high density of polarized crypts that respond in a physiologic manner to addition of growth factors, metabolites, or cytokines to the basal or luminal tissue surface as occurs in vivo.

Methods

A self-renewing monolayer of human intestinal epithelium was cultured on a collagen scaffold microfabricated with an array of crypt-like invaginations. Placement of chemical factors in either the fluid reservoir below or above the cell-covered scaffolding created a gradient of that chemical across the growing epithelial tissue possessing the in vitro crypt structures. Crypt polarization (size of the stem/proliferative and differentiated cell zones) was assessed in response to gradients of growth factors, cytokines, and bacterial metabolites.

Results

Chemical gradients applied to the shaped human epithelium re-created the stem/proliferative and differentiated cell zones of the in vivo intestine. Short-chain fatty acids applied as a gradient from the luminal side confirmed long-standing hypotheses that butyrate diminished stem/progenitor cell proliferation and promoted differentiation into absorptive colonocytes. A gradient of interferon-γ and tumor necrosis factor-α significantly suppressed the stem/progenitor cell proliferation, altering crypt formation.

Conclusions

The in vitro human colon crypt array accurately mimicked the architecture, luminal accessibility, tissue polarity, cell migration, and cellular responses of in vivo intestinal crypts.

Graphical abstract

See editorial on page 161.

Summary.

Human colonic epithelia were cultured on a microfabricated scaffold under a growth factor gradient to generate crypt structures. Responses to cytokines or bacterial metabolite gradients were assessed from altered stem and differentiated cell numbers and locations.

Organ-on-chip technology is a rapidly advancing field that is expected to usher in completely new approaches to drug testing and biological study.¹ Organ-on-chip devices strive to combine microengineered environments with living cells to produce physiologic systems. Devices are being developed to recapitulate the structure and function of a variety of organs, including liver, heart, and lung.² The expectation is that these devices, particularly when incorporating human cells, will create a revolution in the study of human biology and drug development.3, 4 Among these systems, the large intestine especially represents an important organ for in vitro study of intestinal physiology and the evaluation of the effects of pharmaceutical agents and microbial metabolites. For example, pharma and food companies are intensely interested in screening the gut microbiome and the effects of prebiotics and probiotics by virtue of their roles in metabolism and their influences on the human body.5, 6 Despite this importance, major challenges exist in creating an in vitro intestinal epithelium because the intestinal lining is a highly polarized tissue and primary gut epithelium rapidly dies in standard culture. Current gold standard methods for intestinal assays use tumor cell lines, such as Caco-2, and animals.7, 8, 9, 10 This is problematic because tumor cells lack many of the features of normal intestinal tissue and animal models are expensive and increasingly fraught by ethical concerns.

Recently developed intestinal stem-cell culture methods are expected to dramatically improve this situation.11, 12 It now is possible to create multicellular structures known as organoids or mini-guts from primary animal and human stem cells.13, 14, 15 These structures possess self-renewing stem cells and their differentiated progeny to reproduce intestinal epithelium in a culture dish. The potential for the organoid technology is enormous; nevertheless, the enclosed cystic, spherical architecture of organoids presents severe limitations.16, 17, 18, 19, 20 Because of the bulk properties of the matrix and lack of spatial control of growth factors (eg, biochemical gradients), intestinal organoids form embedded within a paddy of Matrigel (Corning, Tewksbury, MA) with an enclosed, inaccessible lumen and random buds lacking distinct stem/transit-amplifying and differentiated cell compartments.21, 22 These characteristics preclude the use of the organoid system in numerous applications and a true living construct suitable for assay of dietary metabolites, cytokines, microorganisms, and drug interactions has not been available.23, 24 What is needed is the ability to recreate the epithelium of the organoid system in an open-faced geometry while maintaining its cellular composition, polarity, and physiology. Our group has been striving to achieve such an intestine-on-chip system for both small and large intestines.25, 26, 27, 28, 29, 30, 31, 32

To address the limitation of the enclosed lumen of organoids, we surveyed a variety of biomaterials, and identified a collagen hydrogel scaffold that enabled a self-renewing planar monolayer culture of colonic epithelial cells with properties similar to those of organoids, but whose luminal surface was readily accessible.²⁵ Nevertheless, these monolayers lacked the 3-dimensional architecture and tissue polarity of in vivo colon crypts. To overcome this shortcoming, in the current study we incorporated microfabrication of the hydrogel scaffold and spatial control of growth factors in a simple-to-use open format. Primary cells from a colonoscopic biopsy were first expanded in the aforementioned monolayer system. The collagen hydrogel scaffold was microfabricated in a microwell architecture possessing the shape of human colonic crypts. A gradient of growth factors along the crypts’ z-axis was used to induce the polarization of the crypts such that stem/progenitor cells were confined to the basal region, while nonproliferative cells were situated along the upper and luminal aspects of the crypts. The in vitro human colonic crypts were compared with native in vivo colon crypts in terms of architecture, luminal patency, tissue polarity, and cell migration. To show the utility of this organ-on-a-chip system, the platform was used to study the impact of metabolites and cytokines on cellular proliferation and location within the tissue.

Materials and Methods

Cell Culture Media

The media compositions are listed in Tables 1 and 2. The culture media (expansion medium [EM], stem medium [SM], and differentiation medium [DM]) for human colonic crypts and epithelial cells were prepared from a mixture of advanced Dulbecco's modified Eagle medium/F12 medium (12634010; ThermoFisher, Waltham, MA) and Wnt-3A, R-spondin 3, noggin (WRN) conditioned medium (see later) at a volumetric ratio of 1:1, and supplemented with 1× GlutaMAX (35050061; ThermoFisher), 1× B27 supplement (12587010; ThermoFisher), 10 mmol/L HEPES (15630-080; ThermoFisher), 1.25 mmol/L N-acetyl cysteine (194603; MP Bio, Santa Ana, CA), 10 mmol/L nicotinamide (N0636; Sigma-Aldrich, St. Louis, MO), 50 ng/mL epidermal growth factor (315-09; Peprotech, Rocky Hill, NJ), 10 nmol/L gastrin (AS-64149; Anaspec, Fremont, CA), 10 nmol/L prostaglandin E2 (14010; Cayman Chemicals, Ann Arbor, MI), 3 μmol/L SB202190 (S1077; Selleckchem, Houston, TX), 100 U/mL penicillin-streptomycin (15140122; ThermoFisher), and 50 μg/mL primocin (ant-pm-1; InvivoGen, San Diego, CA).³³ A total of 10 μmol/L Y27632 (A3008-200; ApexBio, Houston, TX) was used in the first 48 hours after cell plating to prevent dissociation-induced cell apoptosis. WRN conditioned medium was prepared from L-WRN cells (CRL-3276; ATCC, Manassas, VA) following a published protocol.¹⁴ This cell line produces Wnt-3A, R-spondin 3, and noggin. EM was used to expand the epithelial cell numbers as monolayers or organoids. A83-01 (SML0788; Sigma), a transforming growth factor-β inhibitor, was not included in the EM. SM and DM both contained A83-01 (500 ng/mL) because the cells adopted a more columnar morphology under these conditions.

Table 1.

Supplier, Catalog Number, Stock Solution Concentration, and Storage Condition of Reagents

Reagent	Suppliers	Catalog number	Stock solution	Storage
WRN–conditioned medium	Made in-house			-20°C
Advanced DMEM/F12	ThermoFisher	12634-010		4°C
GlutaMax	ThermoFisher	35050061	100×	4°C
HEPES	ThermoFisher	15630-080	1 mol/L	4°C
Gentamycin	ThermoFisher	15750060	50 mg/mL	4°C
Primocin	InvivoGen	ant-pm-1	50 mg/mL	-20°C
B27	ThermoFisher	12587010	50×	-20°C
N-acetyl cysteine	MP Bio	194603	1 mol/L in PBS	-20°C
Murine EGF	Peprotech	315-09	250 μg/mL in 0.1% BSA	-20°C
Nicotinamide	Sigma	N0636-100G	1 mol/L in PBS	-20°C
Gastrin	Anaspec	AS-64149	1 mg/mL in 0.1% BSA	-20°C
Prostaglandin E2	Cayman Chemicals	14010	1 mmol/L in DMSO	-20°C
A83-01	Sigma	SML0788	5 mmol/L in DMSO	-20°C
SB202190	Selleckchem	S1077	30 mmol/L in DMSO	-20°C
Y27632	ApexBio	A3008-200	10 mmol/L in PBS	-20°C
Sodium butyrate	Acros Organics	263190050	500 mmol/L in PBS	-20°C
DAPT	Xcessbio	M60023-5	10 mmol/L in DMSO	-20°C
TNF-α	Peprotech	300-01A	10 μg/mL in 0.1% BSA	-20°C
IFN-γ	Peprotech	300-02	10 μg/mL in 0.1% BSA	-20°C

DMEM, Dulbecco's modified Eagle medium; DMSO, dimethyl sulfoxide; EGF, epidermal growth factor.

Table 2.

Formulation of Culture Media for Human Colonic Epithelial Cells

Reagent	EM	SM	DM
WRN–conditioned medium	50 vol%	50 vol%
Advanced DMEM/F12	50 vol%	50 vol%	100 vol%
GlutaMax	1×	1×	1×
HEPES	10 mmol/L	10 mmol/L	10 mmol/L
Primocin	50 μg/mL	50 μg/mL	50 μg/mL
N-acetyl cysteine	1.25 mmol/L	1.25 mmol/L	1.25 mmol/L
Murine EGF	50 ng/mL	50 ng/mL	50 ng/mL
Nicotinamide	10 mmol/L
B27	1×
Gastrin	10 nmol/L
PGE2	10 nmol/L
A83-01		500 nmol/L	500 nmol/L
SB202190	3 μmol/L
Y27632	10 μmol/La

DMEM, Dulbecco's modified Eagle medium; EGF, epidermal growth factor.

^aUsed in the first 48 hours after cell plating to prevent dissociation-induced cell apoptosis.

Isolation of Crypts From Human Colonic Biopsy Specimens

Biopsy specimens of human colonic epithelium were obtained during routine screening colonoscopies performed at University of North Carolina's Hospital Meadowmont Endoscopy Center with consent of the patient (under the approved University of North Carolina Institutional Review Board #14-2013). Crypts were isolated from the biopsy specimens by incubation with EDTA (2 mmol/L) and dithiothreitol (0.5 mmol/L) in an isolation buffer for 75 minutes at 20°C followed by vigorous shaking in a 15-mL conical tube. The isolation buffer was composed of 5.6 mmol/L Na₂HPO₄, 8.0 mmol/L KH₂PO₄, 96.2 mmol/L NaCl, 1.6 mmol/L KCl, 43.4 mmol/L sucrose, and 54.9 mmol/L D-sorbitol, pH = 7.4.¹³ Hundreds of crypts were released from each biopsy, and the crypts were placed into either monolayer or organoid culture within 30 minutes of isolation.

Organoid Culture With Matrigel Embedding

Released crypts were embedded in Matrigel for organoid culture as described previously.13, 14 Briefly, 500 crypts were suspended in 100 μL of cold Matrigel (356235; Corning, Tewksbury, MA) and then 20 μL of this suspension was added to each well of a 24-well plate. After Matrigel gelation at 37°C for 15 minutes, 500 μL of EM was added to each well. The medium was changed every 48 hours. Y27632 was added to the medium only for the first 48 hours. Every 5 to 7 days the organoids were dissociated with Accutase (07920; Stemcell Technologies, Cambridge, MA) and cultured on a new 24-well plate after dilution of the cells by one third.

Monolayer Culture on a Neutralized Collagen Hydrogel

Human colonic epithelial cells were expanded as a monolayer on a neutralized collagen hydrogel as described previously.²⁵ To prepare 36 mL of a neutralized collagen solution (1 mg/mL), 7.4 mL of collagen (356236; Corning; rat tail, type I, 4.89 mg/mL in 20 mmol/L acetic acid) was mixed with 170 μL of sodium hydroxide (stock concentration of 1 N, working concentration of 23 mmol/L), 720 μL of HEPES (15630080; ThermoFisher; stock concentration of 1 mol/L, working concentration of 20 mmol/L), 2.16 mL of sodium bicarbonate (25-035-CI; Corning; stock concentration of 7.5% wt/vol, working concentration of 0.45% wt/vol), 3.6 mL of 10× phosphate-buffered saline (PBS) (46-013-CM; Corning; working concentration of 1×), and 21.95 mL of deionized water on ice. The mixture was added to 6-well plates (T1006; Denville, Holliston, MA) at 1 mL per well. The plates were incubated at 37°C for 1 hour to generate a clear hydrogel. Four milliliters of PBS was added to each well to hydrate the hydrogel. The plates were placed in a sealed plastic bag and stored at room temperature for up to 3 months. The hydrogel was incubated in PBS buffer for at least 15 days before cell culture. This incubation step was critical for successful culture of the self-renewing epithelial cells and may result in a change in stiffness and other hydrogel properties as the gel matured over the 15-day time span.

Before plating crypts or epithelial cells, the collagen hydrogels were rinsed 3 times with 4 mL of PBS for 5 minutes to remove accumulated salt as a result of water evaporation during storage. Isolated crypts were placed on the top of the collagen hydrogel at a density of 500 crypts/well of a 6-well plate and overlaid with EM (4 mL). The medium was changed every 48 hours. When the cell coverage was greater than 80% (typically 5–7 days), the monolayers were passaged by a gentle 2-step dissociation method.²⁵ The first step was to detach the monolayer from the collagen hydrogel by breaking the hydrogel into pieces by repeated pipetting followed by incubating the hydrogel with 500 U/mL collagenase at 37°C for 10 minutes. The second step was to further split the cell fragments into smaller pieces by incubating the mixture with 150 μL of EDTA (0.5 mmol/L) and Y27632 (10 μmol/L) in PBS at 37°C for 5 minutes. The monolayers again were dispersed by pipetting up and down 30 times using a 200-μL pipet tip. The cell fragments were resuspended in medium and subcultured on a new collagen hydrogel at a passage ratio of 1:3. To ensure that the cells possessed normal chromosomes, cells were karyotyped (KaryoLogic, Inc, Research Triangle Park, NC) at passage (P)7 and P11. Twenty cells from each sample were analyzed. All experiments used the cells between P5 and P10.

Lineage Differentiation

Lineage differentiation was performed using a protocol adapted from that published previously.³⁴ Cells were cultured as monolayers in EM for 4 days followed by culture in DM for 4 days to initiate cell differentiation. Both EM and DM were changed every 48 hours. DM did not contain growth factors (Wnt-3A, R-spondin, or noggin). Sodium butyrate (B; 5 mmol/L, 263190050; Acros Organics, Waltham, MA), γ-secretase inhibitor DAPT (N-[(3,5-Difluorophenyl)acetyl]-L-alanyl-2-phenyl]glycine-1,1-dimethylethyl ester; 10 μmol/L, M60023-5; Xcessbio, San Diego, CA) were added as indicated to induce differentiation into either absorptive colonocytes or mucus-producing goblet cells.

Microfabrication and Surface Modification of Polydimethylsiloxane Stamps

Polydimethylsiloxane (PDMS) stamps were used to micromold collagen scaffolds with an array of microwells. The PDMS stamps possessed an array of microposts with slanted walls and rounded top surfaces (height, 430 μm; base diameter, 125 μm). A custom-formulated 1002F epoxy photoresist was used to generate high-aspect ratio microposts. Formulation of 1002F photoresist was described in a previous publication.³⁵ A film of 1002F photoresist of 150-μm thickness was created on a glass slide by spin-coating 1002F photoresist (formulation 100) on a glass slide (75 mm × 50 mm × 1 mm; Corning). After baking at 95°C for 60 minutes, a second layer of 1002F of 150-μm thickness was spin-coated on the top of the first layer. After baking at 95°C for 60 minutes, a third layer of 1002F of 130-μm thickness was spin-coated on the top of the second layer. The film (total thickness, 430 μm) was baked at 95°C for 60 minutes to evaporate the solvent. To create the sloped post sidewall, the film was exposed to UV light at a dose of 1500 mJ/cm² from the backside (ie, through the glass slide). A postexposure baking was next performed in a 95°C oven for 10 minutes followed by baking on a 120°C hotplate for 10 minutes. The resist then was developed with propylene glycol monomethyl ether acetate for 60 minutes, and baked on a 120°C hotplate for 60 minutes to harden the film. PDMS stamp 1 was created from the earlier 1002F master mold by spreading PDMS prepolymer on the mold followed by degassing under a vacuum. PDMS on the mold was baked in a 95°C oven for 30 minutes followed by demolding. The demolded PDMS possessed an array of microwells. The PDMS mold was treated with plasma for 2 minutes, and coated with octyltrichlorosilane using a vapor-phase process for 16 hours. A second PDMS stamp (stamp 2) was replicated in the same manner from stamp 1 to create a negative replica with an array of posts with slanted sides.

The PDMS stamps were coated with poly(ethylene glycol) to eliminate collagen adhesion during micromolding. The PDMS stamps were surface-grafted with poly(ethylene glycol) as described previously using UV-mediated graft polymerization.³⁶ Briefly, PDMS stamps were placed in a glass tube with a screw-cap (GL 25, polytetrafluoroethylene [PTFE] protected seal; Schott, Elmsford, NY). The tube was filled with a mixture of 10 wt% poly(ethylene glycol) methyl ether acrylate monomer (454990; Sigma-Aldrich, St. Louis, MO), 0.5 mmol/L sodium periodate (311448; Sigma-Aldrich), 0.5 wt% benzyl alcohol (108006; Sigma-Aldrich) in water. The tube then was exposed to UV radiation in a Loctite 7411-S UV Flood System (Henkel, Düsseldorf, Germany) for 4 hours. The stamps then were rinsed with deionized water and incubated in deionized water overnight to remove noncovalently attached polymer and monomer. The grafted stamps were stored in 75% ethanol until use.

Micromolding Collagen Scaffolds on a Modified Transwell Insert

A cross-linked collagen scaffold possessing microwell structures was micromolded on the top of a porous membrane in a modified 12-well Transwell insert (3401; Corning, Tewksbury, MA) using a PDMS stamp. To construct the modified insert, the polycarbonate porous membrane was removed from the Transwell insert using sandpaper. Next, a hydrophilic PTFE porous membrane (BGCM00010; Millipore, Burlington, MA) was attached to the insert using a biocompatible transfer adhesive (1504XL; 3M, Moncure, NC). To reduce the effective area of the porous membrane (ie, from 12-mm diameter to 3-mm diameter), the backside of the porous membrane was blocked by attaching a nonpermeable cyclic olefin copolymer plastic film (4-mL thick, TOPAS 6013; TOPAS Advanced Polymers, Florence, KY).

Collagen in the absence of acetic acid was used for micromolding because the acetic acid (low pH) interfered with the cross-linking reaction. Acetic acid was removed by freeze drying the collagen solution (20 mmol/L acetic acid) on a lyophilizer for 72 hours, followed by redissolving the dried collagen solid in 2-(N-morpholino)ethanesulfonic acid buffer (0.1 mol/L, pH 5) to generate a 5 mg/mL collagen solution. Collagen solution (400 μL), 600 mmol/L 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC; 50 μL), and 150 mmol/L N-hydroxysuccinimide (NHS; 50 μL) were mixed in a conical tube on ice by pipetting up and down 40 times. The trapped air bubbles in the mixture were removed by centrifugation at 3000×g for 1 minute. The modified Transwell inserts were placed in a 12-well plate. Collagen mixture (50 μL) was added to the center of the Transwell insert, and a PDMS stamp was placed on the top of the mixture. (Note that the mixture starts to gel within 5 minutes upon mixing the collagen with EDC and NHS. Thus, the earlier-described steps must be completed within 5 minutes.) To remove the trapped air bubbles among the microposts of the PDMS stamp, the 12-well plate was placed inside a pressure pot (448PP; Dental Planet, Wichita Falls, TX), and pressurized to 40 psi using nitrogen for 2 hours. After removal of the plate from the pot, the insert was incubated in deionized water, and the PDMS stamp was demolded from the solidified collagen scaffold. The collagen scaffold was incubated in 2 L of deionized water for 24 hours to remove residual EDC/NHS. The scaffolds were sterilized with 75% ethanol for 5 minutes, rinsed with PBS for 5 minutes 3 times, and stored in PBS at 4°C until use.

Characterization of Chemical Gradients Applied Across the Scaffold

An impermeable plastic film was attached to the porous membrane on the side opposite to the scaffold to reduce the available area for transport across the membrane from 113 mm² (standard 12-well Transwell size) to 7 mm². The film decreased the area available for diffusion so that upper and lower reservoirs could be approximated as an infinite source and sink. To characterize the chemical gradient profile over time, 100 μg/mL of fluorescein-dextran (molecular weight, 40 kilodaltons) was added either to the top compartment (0.5 mL) or to the bottom compartment (1.5 mL). PBS then was added to the other compartment (bottom compartment, 1.5 mL; top compartment, 0.5 mL). The fluorescein-dextran concentration was measured at 24, 48, and 72 hours in the compartment originally filled with PBS.

Generating In Vitro Crypts

In vitro crypts were formed by culturing cells on the micromolded scaffolds followed by application of a chemical gradient across the scaffold. Before plating cells on the micromolded collagen scaffold, the scaffold was coated with 10 μg/mL human collagen (5007-20ML; Advanced Biomatrix, San Diego, CA) in PBS at 37°C overnight. The scaffold was rinsed with 2 mL PBS. Monolayers of primary cells were expanded on the neutralized collagen hydrogel in a 6-well plate as described earlier in the “Monolayer Culture on a Neutralized Collagen Hydrogel” section. When the monolayers reached >80% confluency, the monolayer was detached from the collagen by incubation with collagenase and then fragmented by EDTA and mechanical agitation as described earlier. The cells then were plated on the top surface of the micromolded collagen in the modified Transwell insert. Cells from 1 well of the 6-well plate were dispersed into 4 separate modified 12-well Transwell inserts. The cells were cultured in 3 mL EM per well (1 mL in the top compartment, and 2 mL in the bottom compartment), and the medium was exchanged every 48 hours. When the cells covered the entire surface of the scaffolds, typically approximately 7 days, a gradient of growth factors was applied across the shaped collagen scaffold and cells. DM (0.5 mL) was placed into the top compartment, and SM (1.5 mL) was added into the bottom compartment. The DM and SM were changed daily to maintain a stable gradient across the collagen scaffold or long axis of the in vitro crypt. Typically, polarization of the in vitro tissue was observed in 4 days under these chemical gradient conditions. When maintained under this chemical gradient, the polarized tissue was maintained for up to 32 days (the longest time tested). To assay the impact of other chemical gradients on crypt polarity, short-chain fatty acids (SCFAs) or cytokines (100 ng/mL tumor necrosis factor-α (TNF-α) [300-01A; Peprotech] and 10 ng/mL interferon-γ (IFN-γ) [300-02; Peprotech]) were added to either the top or bottom compartment as indicated.

Simultaneous Assay of 5-Ethynyl-2-Deoxyuridine Incorporation, Alkaline Phosphatase Activity, Mucin 2 Presence, and DNA Content

Living cells were pulsed with 5-ethynyl-2-deoxyuridine (EdU) and assayed for alkaline phosphatase (ALP). The cells then were fixed and sequentially stained for EdU incorporation (S-phase cells), mucin 2, and DNA. Cells were first pulsed with 10 μmol/L of EdU for 24 hours at 37°C, rinsed with PBS, and incubated with a red ALP substrate (SK-5100; Vector Laboratories, Burlingame, CA) in Tris buffer (0.15 mol/L, pH 8.4) for 30 minutes at 37°C. The cells were rinsed with PBS, fixed in 4% paraformaldehyde for 15 minutes, and permeabilized with 0.5% Triton X-100 in PBS for 20 minutes. Incorporated EdU was labeled with Click-iT EdU Alexa Fluor 647 (C10340; ThermoFisher). The cells were then rinsed with 0.75% glycine in PBS for 5 minutes 3 times, followed by blocking with 10% donkey serum (017-000-121; Jackson Immunoresearch, West Grove, PA) for 1 hour. The cells were incubated in rabbit mucin 2 (Muc2) antibody (1:200, sc-15334; Santa Cruz Biotechnology, Dallas, TX) at 4°C overnight, and stained with donkey anti-rabbit IgG-conjugated Alexa Fluor 488 (1:500, 711-545-152; Jackson Immunoresearch) for 45 minutes. Finally, the DNA was stained with Hoechst 33342 (2 μg/mL, B2261; Sigma-Aldrich) for 15 minutes.

Quantification of Mucin Secretion and ALP Activity

The protein level of Muc2 secreted by the cells into the supernatant was determined quantitatively by a direct enzyme-linked immunosorbent array (ELISA) adapted from a previous publication.³⁷ Cells were incubated for 48 hours and their supernatant collected and diluted with coating buffer (0.1 mol/L sodium carbonate, pH 9.5) at a ratio between 1:10 and 1:100, and 100 μL was added to each well of an ELISA plate (15042; ThermoFisher). The plate was incubated in a 40°C oven overnight to let the samples dry. The wells then were washed 3 times with 200 μL PBS, blocked with 200 μL of 1% bovine serum albumin (BSA) (A9647; Sigma) in PBS for 1 hour at room temperature, and incubated with 100 μL of horseradish peroxidase (HRP)-conjugated MUC2 primary antibody (sc-515032; Santa Cruz Biotechnology; diluted in 1% BSA to 15 ng/mL) for 16 hours at 4°C. The wells then were washed 5 times with 200 μL of 0.05% Tween-20 in PBS. SuperSignal ELISA Femto horseradish peroxidase substrate (37075; ThermoFisher; 10× dilution in deionized water, 100 μL) was added to each well. After 10 minutes, the luminescence was measured using a plate reader (SpectraMax M5; Molecular Devices, Sunnyvale, CA). Media from wells without cells were used as a control. The results are expressed as a percentage of the control medium and were normalized to the total protein content of the cells from which the supernatant was collected.

To measure the ALP enzymatic activity, the cells were detached from the collagen hydrogel by incubation in collagenase for 15 minutes at 37°C as described in a previous section, and washed twice with 5 mL PBS. The cell pellets were lysed with 200 μL lysis buffer (1% Triton X-100, 150 mmol/L KCl, 50 mmol/L HEPES, 5 mmol/L EDTA, 5% glycerol), vortexed, and incubated on ice for 20 minutes. After centrifugation at 10,000×g (10 min, 4°C), the supernatants were collected to quantify ALP activity and the total protein content. The ALP activity of the cell lysates was measured using the Sensolyte luminescent ALP assay (AS-72122; Anaspec, Fremont, CA). ALP from calf intestine was used as a positive control and standard. The protein concentration of the cell lysates was measured by a Pierce is now a ThermoFisher trademark, not a company. There is no need to provide address for Pierce BCA protein assay (23225; ThermoFisher). The ALP activity is expressed as picograms equivalent to the calf intestine standard and normalized to the cell protein amount. Three wells were used for the measurements of both mucin secretion and ALP activity (n = 3).

Immunofluorescence Staining

Immunofluorescence was used to show the presence of different cell types and subcellular structures of a tissue (eg, biopsies, monolayers, organoids, in vitro crypts). To obtain tissue slices, the specimen was first fixed in 4% paraformaldehyde for 15 minutes, rinsed with PBS, and incubated in 30% sucrose at 4°C overnight. The tissue then was embedded in a cryo-embedding optimal cutting temperature (OCT) medium (Tissue-Tek O.C.T. Compound, VWR, Radnor, PA), and the frozen tissue block was sectioned into 10-μm–thick films by a cryostat. For zonula occludens-1 (ZO-1), E-cadherin, occludin, and β-catenin, the cells were fixed with ice-cold methanol and placed at -20°C for 15 minutes. To preserve mucins for Muc2 staining on in vitro crypts, the tissue was fixed with Carnoy’s solution (90% methanol and 10% acetic acid) for 30 minutes. For immunostaining, the primary antibodies used were antibodies to olfactomedin-4 (Olfm4) (1:500, 14369; Cell Signaling Techology, Danvers, MA), cytokeratin 20 (KRT20) (1:500, 60183-1-lg; ProteinTech), Muc2 (1:200, sc-15334; Santa Cruz Biotechnology), ezrin (1:200, PA5-17518; ThermoFisher), integrin β4 (1:200, sc-9090; Santa Cruz Biotechnology), ZO-1 (1:100, 21773-1-AP; Proteintech, Rosemont, IL), β-catenin (1:200, sc-9090; Santa Cruz Biotechnology), occludin (1:100, 13409-1-AP; Proteintech), and E-cadherin (1:100, 20874-1-AP; Proteintech). Antigen retrieval was performed when specimens were stained for Oflm4 and KRT20. For other stains, antigen retrieval was not required to visualize immunofluorescence staining. For antigen retrieval, the specimen sections were rinsed with PBS 3 times to remove the cryo-embedding OCT medium. The slides were placed in a retrieval solution (RV1000MMRTU, Reveal Decloaker; Biocare Medical, Pacheco, CA) in a plastic Coplin jar in a 95°C water bath for 15 minutes. The jar then was placed at 20°C for 15 minutes. The sections with and without antigen retrieval then were rinsed with PBS and blocked with 10% donkey serum for 1 hour at 20°C. The sections were incubated with primary antibodies overnight at 4°C. The sections then were stained with secondary antibody, donkey anti-rabbit, or mouse IgG conjugated with Alexa Fluor 488 or 594 (1:500, 711-585-152 or 715-585-150; Jackson Immunoresearch). The nuclei were stained with Hoechst 33342 (2 μg/mL) for 15 minutes.

Electron Microscopy

For scanning electron microscopy (SEM), samples were dried with a critical point dryer (PVT-3, Tousimis Semidri, Rockville, MD), coated with 10-nm metal by a sputter coater (Cressington 108, Cresington Scientific Instruments, Watford, UK), and inspected by a SEM (FEI Quanta 200 ESEM; FEI Company, Hillsboro, OR).

Fluorescence Microscopy

Specimens were imaged using a Nikon Eclipse TE300 inverted epifluorescence microscope (Nikon, Melville, NY) equipped with 4′,6-diamidino-2-phenylindole/fluorescein isothiocyanate/Texas Red/Cyanine 5 (CY5) filter sets. Images were acquired at randomly selected locations within a specimen using a 10× (N.A. numerical aperture (N.A.) = 0.3) or 4× (N.A. = 0.13) objective. S-phase cells stained with EdU were imaged using a CY5 filter (excitation filter, 604–644 nm; emission, 672–712 nm), ALP stain was imaged with a Texas Red filter (excitation filter, 542–582 nm; emission, 604–644 nm), Muc2 immunofluorescence was visualized with a fluorescein filter (excitation filter, 450–490 nm; emission, 520 nm long pass), and DNA stained by Hoechst 33342 was identified using a 4′,6-diamidino-2-phenylindole filter (excitation filter, 352–402 nm; emission, 417-477 nm).

Confocal microscopy was performed using a confocal laser scanning microscope (Fluoview FV3000; Olympus, Waltham, MA) with laser-based excitation and emission wavelengths selected using a holographic transmission diffraction grating.

Image Analysis

Images were empirically thresholded using ImageJ software (https://imagej.nih.gov/ij/). ImageJ also was used to quantitatively measure the monolayer surface area possessing suprathreshold fluorescence for the EdU, ALP, or Muc2 stains. The identified surface area then was divided by the total cell area calculated from the image area positive for Hoechst fluorescence (DNA stain) and expressed as a percentage cell area positive for the EdU, ALP, or Muc2 stain. All data used 5 randomly selected locations of the same monolayer (n = 5), and the average with a single SD is shown unless otherwise specified.

Statistics

Statistical analysis was performed using a 1-way analysis of variance followed by a multiple comparison test with the Tukey honest significant difference procedure conducted at the 5% significance level. All analyses of variance returned P values less than .004. Multiple comparison testing was performed on all pairwise comparisons between experimental groups. Analyses of variance and subsequent multiple comparisons were performed using MATLAB (MATLAB 2014b; The MathWorks, Inc, Natick, MA). An unpaired 2-tailed Student t test was used to compare the outcomes from 2 experimental conditions in the cytokine perturbation experiment.

All authors had access to the study data and reviewed and approved the final manuscript.

Imaging and Assay of Whole Crypts Detached From the In Vitro Tissue

Before removal from the tissue, the crypts were stained as described earlier for EdU incorporation, ALP activity, Muc2 presence, or DNA content. To assess the polarity of the in vitro crypts, a side view of the intact crypts was obtained by scraping the crypts loose from the tissue and scaffold using a tungsten dissecting needle (RS-6063; Roboz Surgical Instrument Co, Gaithersburg, MD). Detached crypts were overlaid onto a glass slide onto which they settled, lying in a horizontal position. The crypts then were imaged using standard fluorescence microscopy. The position of proliferating cells (EdU⁺) along the basal–luminal crypt axis was obtained by evenly dividing the crypt into 8 regions. The fluorescence intensity ratio of the EdU stain relative to that of Hoechst fluorescence (DNA) from each region was obtained by ImageJ. The EdU/Hoechst ratio then was plotted against position along the basal–luminal crypt axis. Twenty crypts were quantified under identical fluorescence measurement conditions.

To assay the impact of SCFAs or cytokines on the polarity of in vitro crypts, the crypts were scraped from the array and imaged as described earlier. The number of S-phase cells per crypt was obtained by manually counting the number of EdU⁺ cells per crypt using ImageJ. The relative proliferation length (expressed as a percentage) is defined as the length of the crypt possessing EdU+ cells divided by the total length of the crypt. The normalized ALP activity of the crypts was obtained by measuring the total fluorescence intensity of the ALP stain divided by the fluorescence intensity of the Hoechst DNA stain. More than 10 crypts were quantified for each data point under identical imaging conditions.

Unless otherwise specified, the data shown for a single experiment used crypts or cells obtained from a single biopsy (male; age, 52 y). Experiments used material from biopsy specimens from 3 different human beings and the results were consistent over time without outliers. Data from representative experiments are presented.

Results

Organoid Culture of Human Colonic Epithelial Cells

The luminal surface of the human colonic epithelium possesses a high density of crypts or invaginations interspersed among luminal columnar cells (Figure 1A). Between these cell-covered invaginations lies a lamina propria that shapes and supports the crypt and its cells. Proliferative stem/progenitor cells (Olfm4⁺) are confined to the basal side of the crypts, while the nondividing differentiated cells are located at the luminal surface.38, 39 This exquisite tissue polarity with compartmentalization of proliferative and differentiated cells is thought to be maintained in vivo by a gradient of well-balanced growth signals (Wnt-3A, Notch, bone morphogenic protein [BMP], and other factors) spanning the long axis of crypt from the luminal to the basal region.⁴⁰ This microenvironment also ensures the unidirectional migration of cells from the proliferative to the differentiated cell zone so that the luminal epithelial cells are replaced every 5–7 days.⁴¹ The U shape of the crypt permits the stem cells to be in a contiguous monolayer with their differentiated progeny while surrounded by a chemical microenvironment distinct from that of their differentiated offspring. The growth-factor–rich microenvironment supports the stem cells as they proliferate and protects these cells from the harsh milieu found within the colonic lumen.⁴² Isolation of crypts from the colon and placement within a Matrigel patty with appropriate supporting growth factors enables the stem cells to proliferate but with transition into a single-layered, hollow organoid, with loss of the crypt anatomic features (Figure 1B).³⁹ When maintained under a growth-factor–rich medium (EM), the organoids predominantly comprise proliferative cells (EdU⁺, Olfm4⁺), with a paucity of differentiated cells (KRT20⁺) (Figure 1C). The loss of tissue polarity likely is owing to the absence of a polarity-defining growth factor gradient within the surrounding hydrogel as well as the inability of the Matrigel to guide the growing organoids into the proper sized and shaped crypt structures.22, 25 Although the Matrigel-based organoid system has been an enabling technology for experimentation into intestinal stem cell behavior and the creation of many intestinal disease models, the system does not fully recapitulate the architecture and physiology of in vivo crypts. The organoid morphology and surrounding hydrogel remain incompatible with many desired investigations, especially those focusing on crypt dynamics or studying exposure to organisms and chemicals selectively impacting one face (luminal or basal) of the intestinal epithelium.

Figure 1.

Organoid culture of human colonic epithelial cells. (A) Immunofluorescence staining of human colonic tissues. A fluorescence image of tissue immunostained for KRT20⁺ (red) and Olfm4⁺ (green) is shown. DNA was stained with Hoechst 33342 (blue). (B) Human colonic organoids embedded in Matrigel using a crypt as the starting material. The brightfield images on days 0, 2, and 4 of culture at passage numbers 0 and 1 are shown. (C) EdU pulse labeling and immunofluorescence staining (Olfm4 and KRT20) of organoids after 4 days in culture. DNA was stained with Hoechst 33342. Scale bars: 100 μm.

Monolayer System Allows Self-Renewal and Differentiation of Human Colonic Epithelial Cells in Response to External Biochemical Cues

Wang et al²⁵ recently developed a collagen hydrogel scaffold that supported a self-renewing monolayer of human colonic epithelial cells with both stem/proliferative cells as well as the various differentiated cells (goblet cells, enteroendocrine cells, and colonocytes). Under the reported growth-factor–rich culture conditions, both monolayers on collagen and organoids within Matrigel predominantly comprised proliferative cells, with a minority of differentiated cells.²⁵ Prior reports have shown that the stem/proliferative cells of organoids can be chemically differentiated into either absorptive cells or goblet cell lineages by modulating Wnt and Notch signals.³⁴ To determine whether the monolayer cells also might be force differentiated, the monolayers were cultured in EM for 4 days and then to media promoting differentiation for 4 days: SM (with growth factors but without inhibitors/hormones), DM (without growth factors, inhibitors/hormones), DM plus 5 mmol/L butyrate (DM-B), or DM plus 10 μmol/L DAPT (DM-D) (Figure 2A and B). Cells remained proliferative in SM owing to the presence of growth factors, but grew at a slower rate relative to cells in EM (Figure 2B and E). Cell proliferation largely was eliminated in DM as evidenced by a decrease in S-phase cells as a result of the absence of growth factors. However expression of differentiation markers (ALP⁺ for colonocytes, Muc2⁺ for goblet cells) increased relative to that of cells in SM (Figure 2B and E). Addition of butyrate to DM increased ALP activity but not Muc2 expression, suggesting that the monolayers predominantly comprised absorptive colonocytes (Figure 2B and E). Butyrate is produced in vivo by bacterial fermentation of fiber and is thought to act as a Notch activator.43, 44 In contrast, DAPT, a Notch inhibitor, increased Muc2 expression, suggesting that Notch inhibition increased Goblet cell production under these conditions.43, 44 These results were consistent with the apical morphology of the monolayers when inspected by SEM (Figure 2C): proliferative cells (under SM) showed a low density of stunted microvilli and no visible secretory granules, colonocytes (under DM-B) showed a high density of long microvilli but again without visible secretory granules, and goblet cells (under DM-D) showed secretory granules lining their apical surfaces but without microvilli. In addition, tight junction proteins (β-catenin, ZO-1, occludin, and E-cadherin) were expressed in the monolayers (Figure 2D), suggesting that this can be optimized to interrogate barrier function.

Figure 2.

Forced differentiation of human colonic epithelial cells grown as a monolayer. (A) Schematic showing the forced differentiation of the cells. Cells were cultured in expansion medium (EM) for 4 days and then either stem medium (SM) or differentiation medium (DM) for 4 days. Additional molecules added to the DM were 5 mmol/L butyrate (DM-B) or 10 μmol/L DAPT (DM-D). (B) Fluorescence microscopy of monolayers grown under the different differentiation media. Colors as follows: EdU, green; ALP, red; Muc2, yellow; DNA, blue. (C) Apical surface topography of human colonic monolayers inspected by SEM. Left: low-power magnification; right panels, high power magnification. (D) Immunofluorescence staining of epithelial monolayers (6 days in expansion medium) using antibodies directed against β-catenin (green), ZO-1 (red), occludin (red), and E-cadherin (green). Nuclei were stained with Hoechst 33342 (blue). (E) The percentage of the monolayer surface area showing fluorescence from the EdU, Muc2, and ALP stains under the various culture conditions. (F) ALP activity of cell lysates under the various culture conditions. (G) Mucin secretion by the monolayer under the various culture conditions. IF, immunofluorescence; Mag, magnification. *P < .05 and **P < .005.

The presence of ALP activity and Muc2 was spatially heterogeneous in the monolayers, which led to a large variability in their detection when measured by fluorescence microscopy (Figure 2B and E). For this reason, ELISA and enzymatic assays able to measure the ALP activity and Muc2 presence from the entire monolayer also were performed. ALP enzymatic activity in the presence of DM-B was significantly greater than that for monolayers incubated with SM, DM, or DM-D, again suggesting that the DM-B monolayer was primarily colonocytes and that Notch activation drove colonocyte formation under these conditions (Figure 2F). Muc2 secretion was increased significantly by incubation in DM-D relative to that of DM, and both of these media yielded substantially more Muc2 than incubation in either SM and DM-B (Figure 2G). Notch inhibition under these conditions enhanced the formation of Goblet cells and mucin presence in the monolayer cultures (Figures 2B and 4M). These data taken together suggest that cells in the monolayer behave in a similar fashion to those in the organoids in response to external biochemical cues.³⁴ Thus, the collagen hydrogel scaffold is a suitable matrix for culture of proliferative monolayers or differentiated human colonic epithelial cells. The ability to redirect monolayer cell fate into either of the 2 differentiated lineages provides additional evidence for the presence of malleable stem cells within the monolayers.

Figure 4.

Generation of in vitro human colon crypts. (A) Schematic showing the process of converting in vivo crypts to in vitro crypts. Crypts were isolated from a biopsy, and the cells were expanded as monolayers on a planar scaffold (for 5–25 days, depending on the size of the biopsy specimen and the number of cells banked for other experiments). The cell fragments were cultured on the shaped scaffold for 7 days to create the crypt-like geometry. The cells on the shaped scaffold then were cultured under a biochemical gradient for at least 4 days to polarize the crypts. Differentiated and stem/proliferative cells are shown in red and green, respectively. (B) Time-lapse brightfield images showing the growth of cells across the surface of the microwell array. Microwells appeared darker in these brightfield images because their walls were lined with cells. (C) Side view of an in vitro–formed crypt. (D) Side view of an array of crypts. (E) Low-power image (top view) of a 3-mm array under a growth factor gradient. EdU incorporation (green), ALP (red), DNA (blue). (F) Fluorescence image (side view) of 5 interconnected in vitro crypts after an EdU pulse (green), ALP stain (red), and Hoechst-DNA labeling (blue). (G) A cross-section from a confocal image of a polarized crypt showing the hollow lumen. EdU incorporation (green), DNA (blue). (H) Normalized EdU incorporation along the crypt axis. Zero marks the crypt base while 1 denotes 430 μm from the base (crypt top). (I) Directional migration of EdU⁺ cells over time. Images were at 0 and 96 hours after EdU pulse (24-h EdU pulse duration). EdU is green and DNA is blue. (J) Box plots of the relative proliferation length, which was defined as the length of EdU⁺ tissue divided by the crypt length (430 μm) at 0 and 96 hours after EdU pulse. Ten crypt units were quantified. P < .0001 based on t test. (K) Cross-section of in vitro crypt immunostained for KRT20 (red) and Olfm4 (green) and DNA labeled with Hoechst 33342 (blue). (L) Overlaid brightfield and fluorescence images of cross-sections through the luminal surface between the in vitro crypts. (M–O) Mucus layer on the tissue. (M) A mucus layer was observed above a planar monolayer after fixing the tissue with Carnoy’s solution (after 4 days of culture in EM, and 4 days in DM + 10 μm DAPT). (N) Mucus was detected within the in vitro crypts (4 days in EM, 6 days under a polarization gradient). Top: Brightfield microscopy image. Bottom: Immunofluorescence image. Muc2⁺ mucus protruded from the inside of the crypts up into the media. (O) Crypts released from the scaffold were imaged by fluorescence microscopy. (N and O) Red depicts Muc2 immunofluorescence and blue is Hoechst 33342 fluorescence. Scale bars: 1 mm (E), 100 μm (B–D, F, G, I, and K), and 10 μm (L).

Microfabrication of a Collagen Scaffold to Support Chemical Gradients

To shape the monolayers into the architecture of in vivo human colon crypts, we micromolded collagen hydrogels on a porous membrane using a PDMS stamp (Figure 3A–E). A hydrophilic, 0.4-μm PTFE membrane was selected because of its optical transparency, low fluorescence, and high permeability. The stamps possessed an array of high-aspect ratio posts with a slanted sidewall and rounded end (Figure 3A, C, and D). During the molding process, the PDMS stamp with posts was pressed into the collagen to create microwells with a size and shape similar to that of the human colon crypts (Figure 3E and H–J). To strengthen the collagen scaffold and prevent cells from deforming the microwells, the collagen was mixed with EDC/NHS to initiate collagen cross-linking during the molding process.⁴⁵ The micromolded collagen scaffold possessed 245 microwells in the central 7 mm² area of a modified 12-well Transwell insert (Figure 3F–J). The Transwell platform was used because of the ease in which a gradient could be initiated across the long axis of the microwells (z-axis of the scaffold) (ie, simply placing different media in the basal and luminal reservoirs). The platform also can be adapted to other chamber types (such as Seahorse Metabolic Analyzer, Agilent, Santa Clara, CA) by modifying the insert design.

Figure 3.

Scaffolding and gradient cassette for the culture of in vitro crypts. (A) Fabrication process for the PDMS stamps used to micromold collagen scaffolds mimicking the topography of epithelium of human colon. (B) Transmitted light microscopy of a human colonic crypt from a biopsy sample. (C) Geometry of the PDMS stamp that was used to micromold collagen. Units are shown in micrometers. (D) SEM image of the PDMS stamp. (E) Schematic of the micromolding process used to generate a shaped, cross-linked collagen scaffold on a modified Transwell insert. (F) Image of the 12 scaffolds within a 12-well plate. To show the trans-scaffold transport occurred only at the central area, the insert was briefly exposed to 0.1% toluidine blue from the basal side. The toluidine blue only stained the scaffold within the central 3-mm diameter area in contact with the basal compartment. (G) Image of a single scaffold within an insert. (H) Top view of an array of microwells created in the scaffold. (I) Close-up view of a region of the scaffold in panel H. (J) Side view of a cross-section through a fluorescein-labeled collagen scaffold. (K) Concentration of fluorescein dextran (molecular weight, 40 kilodaltons) measured at the luminal (solid squares) and basal (empty circles) compartments at 24, 48, and 72 hours. Top plot: Diffusion from basal to luminal. At 0 hours, 0.5 mL PBS was added to the luminal compartment and 1.5 mL fluorescein dextran (100 μg/mL) was added to the basal compartment. Bottom plot: Diffusion from luminal to basal. At 0 h, 0.5 mL fluorescein dextran (100 μg/mL) was added to the luminal compartment and 0.5 mL PBS was added to the basal compartment.

To determine the period of time over which the basal and luminal reservoirs could act as an infinite source and sink to support a stable gradient of growth factors across the collagen scaffold, the diffusion of a model compound, fluorescein-dextran, across the scaffold/membrane was measured (Figure 3K). Fluorescein-dextran (40 kilodaltons) was chosen because it is of similar molecular weight to Wnt-3A (39.7 kilodaltons), R-spondin (40.0 kilodaltons), and noggin (46 kilodaltons). At 0 hour, the fluorescein-dextran concentration was (source) = 100 μg/mL and (sink) = 0 μg/mL. At 24 hours, when the basal reservoir was the source, the concentration was (source) = 96.6 ± 3.0 μg/mL and (sink) = 2.9 ± 0.7 μg/mL (luminal reservoir). When the direction of diffusion was reversed (ie, the basal reservoir was now the sink), the (sink) = 0.9 ± 0.2 μg/mL and the (source) = 90.2 ± 0.9 μg/mL (luminal reservoir). These data suggested that a stable, steep chemical gradient was established across the collagen scaffold when the fluids in the basal and luminal reservoirs were replenished every 24 hours.

Construction of Polarized Colonic Crypts

Colonic crypts were constructed by growing cells on the collagen scaffold until the cells covered the surface including that of the sides and bottoms of the microwells (Figure 4A). When cultured in the presence of EM, which promotes rapid cell proliferation, the cells typically covered the scaffolds within 7 days (Figure 4B). In contrast to native collagen scaffolds, the cross-linked collagen scaffolds possessed sufficient strength to resist cell-induced deformation of the microwell structures. The crypts within the monolayer could be mechanically dislodged from the scaffolding and placed on a glass slide for microscopy as single crypts or groups of connected crypts (Figure 4C–F). Each in vitro crypt possessed a shape and dimensions nearly identical to that of the in vivo structures and showed an open luminal end mimicking in vivo crypts.

Polarization of the in vitro crypts was accomplished by adding SM to the basal scaffold side and DM to the luminal scaffold side. SM and DM are identical except that DM does not contain growth factors Wnt-3A, R-spondin, and noggin. Thus, a gradient of growth factors was established along the basal–luminal axis of the crypt structures with the growth factors high at the basal crypt region and low at the luminal epithelial surface (Figure 4A). After 4 days of culture under the SM/DM gradient, a polarized in vitro crypt array was created (Figure 4E–G). Proliferative cells were localized to the base of the microwell or crypt (ie, to the region with high growth factor concentrations), while differentiated, absorptive colonocytes were located on the luminal surface (Figure 4E–G). Application of a growth factor gradient was sufficient to re-create an asymmetric crypt showing a proliferative and nonproliferative compartment (Figure 4G and H). Upward cell migration at a rate similar to that occurring in vivo was shown by EdU pulse labeling and tracking the labeled cells over 96 hours as they migrated up the crypt axis (30 μm/day) (Figure 4I and J).⁴⁶ Similar to in vivo crypts (Figure 1A), cells of in vitro crypts expressing a differentiation marker (KRT20⁺) were located at the luminal surface, while cells expressing a marker of colonic stem cells (Olfm4⁺) were restricted to the basal crypt region, showing a stem cell niche (Figure 4K).³⁸ Goblet cells and mucus also were observed when the cells were cultured for 4 days in EM and 6 days under gradient (Figure 4M–O). When cells lining the luminal tissue surface (not within the crypt) were immunostained, the cells themselves were asymmetric, displaying properties of an absorptive colonocyte, for example, expressing the microvilli marker ezrin on the cell side adjacent to the media and integrin-β4 on the cell side contacting the collagen scaffold (Figure 4K and L). The cells also adopted the columnar shape (height, 35 μm; width, 9 μm) typical of differentiated absorptive colonocytes in vivo. When these human in vitro crypts were maintained under a growth factor gradient for 32 days, the epithelium remained as a monolayer, with polarized crypts creating a long-term human epithelial replica.

Response of In Vitro Crypts to Gradients of SCFAs

SCFAs are produced in large quantities by bacterial fermentation of fiber in the colon with the SCFAs such as acetate, propionate, and butyrate present at millimolar concentrations in the lumen.⁴⁷ Colonic epithelial cells use these SCFAs as their primary energy source, decreasing the SCFA concentrations near the stem cell compartment to low micromolar concentrations and creating steep SCFA gradients along the length of the crypt.48, 49 The gradient of butyrate (but not acetate or propionate) is thought to play a regulatory role modulating stem cell proliferation by acting as a histone deacetylase inhibitor at low concentration.⁴⁸ To assess whether the in vitro tissue was responsive to SCFAs, a gradient of acetate, propionate, or butyrate in addition to that of the growth factors was generated across the in vitro crypts by placing the SCFA into the luminal compartment above the tissue (Figure 5A).47, 49, 50 The crypts were cultured under the SCFA gradient for 4 days (Figure 5B and C). All 3 SCFAs increased alkaline phosphatase activity indicative of absorptive colonocytes at the luminal surface, but butyrate was most potent by several fold (Figure 5C and F). Acetate and propionate gradients did not impact cell proliferation along the crypt axis consistent with their reported behavior in organoid systems (Figure 5B–E).⁴⁸ In contrast, the butyrate gradient eliminated cell proliferation in the luminal crypt region as shown by the absence of EdU⁺ cells from the upper crypt half (Figure 5B–E). Cell proliferation also was diminished but not eliminated near the crypt base so that butyrate acted to restrict the proliferating cells to the very base of the crypt. Dose-response experiments showed that each butyrate concentration (0.5–5 mmol/L) diminished stem/progenitor cell turnover (Figure 5G and H), whereas only the higher concentrations (2–5 mmol/L) enhanced absorptive colonocyte formation. Previous SCFA studies have been reliant on colon cancer cell lines that do not accurately model the colonic epithelium.⁵¹ Using our in vitro crypt platform, we show that butyrate suppresses stem cell activity, which supports the hypothesized effect of butyrate in human beings,⁴⁸ and is capable of inducing the differentiation of colonocytes into the absorptive lineage. This latter finding has broad-ranging implications because it is evidence linking microbial metabolites with cell-fate decisions in human beings.

Figure 5.

Modulation of in vitro human crypts by SCFAs. (A) Biochemical gradients applied to the tissue. (B) Side view of representative crypts from the arrays (scale bar: 100 μm) under different SCFA gradients: no SCFA gradient (control), acetate (0–15 mmol/L), propionate (0–5 mmol/L), and butyrate (0–5 mmol/L). Red, ALP; green, EdU-based stain; blue, DNA. (C) Top view of human crypt array. Upper panel scale bar: 1 mm; lower panel scale bar: 100 μm. (D) Number of EdU⁺ cells per crypt under different gradients. (E) Relative proliferation length, defined as the length of EdU⁺ crypt over the total length of the crypt for the different gradients. (F) Normalized ALP activity for the various SCFA gradients. (G and H) EdU incorporation and ALP activity under different butyrate gradients. Basal butyrate (0 mmol/L) with the luminal butyrate concentration listed on the x-axis. N, noggin; R, R-spondin; W, Wnt-3A. ∗P < .05 and ∗∗P < .005.

Perturbation of In Vitro Crypts by Cytokines

Inflammatory cytokines such as IFN-γ and TNF-α are produced in significant quantities by the inflamed intestine.⁵² For example, IFN-γ and TNF-α were found to be the most highly induced cytokines in dextran sulfate–treated colon.⁵² However, the exact effect of these cytokines has been quite challenging to identify because of the plethora of changes ongoing in animals with inflamed intestine. As a result, many studies have relied on tissue-cultured cell lines, such as the T84 colon carcinoma line, as surrogate reporter systems. In this cell system, TNF-α and IFN-γ synergistically suppress proliferation.⁵² Further complicating the understanding of cytokine impact is that these molecules likely exist at different concentrations within the crypt, which may modulate their impact on the various cell types.52, 53, 54 To determine whether the in vitro colon crypts are responsive to inflammatory cytokines, the tissue was cultured under a gradient of growth factors and cytokines (100 ng/mL TNF-α and 10 ng/mL IFN-γ) applied from the basal side for 4 days (Figure 5A). The cytokines were included in the basal compartment because they are produced by various cells of the mucosal immune system in response to environmental triggers.⁵³ Cell proliferation as measured by EdU incorporation under a gradient of cytokines was dramatically suppressed in the crypts when compared with the vehicle control (0.1% BSA) (Figure 5B–E). The cytokines eliminated tissue polarity because most crypts exposed to the cytokines possessed no EdU⁺ cells at any location within the crypt. In the small number of crypts that did possess EdU⁺ cells, the cells were located randomly (Figure 5B and E). The number of S-phase (EdU⁺) cells per crypt decreased from 30 ± 13 for the control tissue to 1 ± 1 in that exposed to cytokine (Figure 5D). In contrast, the cytokines only slightly decreased the ALP activity relative to that of control (Figure 5F). To determine whether the decrease in proliferation under the cytokine gradient was a result of cell death, the crypt arrays were stained with Hoechst 33342 and propidium iodide. The cytokine-exposed tissue possessed significantly more dead cells (propidium iodide positive) than that without cytokine (Figure 6A–H); however, the absolute numbers of dead cells per crypt was low in both instances. Thus, the loss in cell proliferation was not owing to the death of the cells within tissue but rather owing to diminished numbers of viable cells in the S phase. These data show that the basally released inflammatory cytokines diminish the self-renewal power of the crypts.

Figure 6.

The effect of the cytokines, IFN-γ and TNF-α, on in vitro human crypts. (A) Biochemical gradients applied to the tissue. (B) Side view of representative crypts from the arrays under gradients of vehicle (0.1% BSA, control), or cytokines of 10 ng/mL IFN-γ and 100 ng/mL TNF-α. Scale bar: 100 μm. Red, ALP; green, EdU-based stain; blue, DNA. (C) Confocal 3-dimensional reconstruction of human crypt array viewed from basal side. Upper panel scale bar: 1 mm; lower panel scale bar: 100 μm. (D) Effect of cytokine gradient on number of EdU⁺ cells per crypt (N = 20 crypts). (E) Box plot showing the relative position of EdU⁺ cells along the basal–luminal axis of crypts. Zero represents the basal end of the crypt and 1 marks the luminal end of the crypt. (F) Normalized ALP stain for the array with the focal plane at the luminal side with a depth of field of 60 μm (N = 4 arrays). (G) A cross-sectional image through an in vitro crypt obtained by confocal fluorescence imaging. Left: Crypt in the absence of a cytokine gradient. Right: Crypt in the presence of an IFN-γ/TNF-α gradient. The cells were stained with Hoechst 33342 (blue) and propidium iodide (PI) (red). Scale bar: 100 μm. (H) Box plot depicting the number of dead cells per cross-section of crypt (N ≥ 24). Unpaired 2-tailed Student t test: *P < .05; **P < .005. N, noggin; R, R-spondin; W, Wnt-3A.

Conclusions

Stem cells obtained from a human colon biopsy specimen were used to create self-renewing mimics of colonic crypts. Chemical gradients applied across the tissue construct led to compartmentalization of the stem/progenitor cells and promoted appropriate lineage differentiation and cell movement replicating human crypt biology. Application of gradients of microbiota-derived fermentation products drove alterations in the size of the crypts’ proliferative and differentiated compartments as predicted to occur in vivo. Construction of primary epithelium from human stem cells yields a physiologically relevant mimic of the colonic epithelium, bringing to fruition an in vitro human tissue platform for the study of drugs and microbiome-derived compounds.